Shift work and heart disease - Lægeforeningen

Shift work and heart diseaseEpidemiological and risk factor aspectsPh.D. thesisHenrik BøggildThis PDF-version is not necessarily in the same style as the printet version, andpagenumbering is different. The wording is however the same.Due to transfer of copyright the five papers in the printet version are not includedin this version.Centre for Working Time ResearchDepartment of Occupational MedicineAalborg Regional HospitalAalborgFaculty of Health SciencesUniversity of Aarhus2000

Forelæsningen finder sted onsdag den 23. august 2000 kl. 14.00i Auditoriet, Aalborg Sygehus Syd

Shift work and heart diseaseEpidemiological and risk factor aspectsPh.D. thesisHenrik BøggildCentre for Working Time ResearchDepartment of Occupational MedicineAalborg Regional HospitalAalborgFaculty of Health SciencesUniversity of Aarhus2000

Praise be doubt! I advise you to greetCheerfully and with respect the manWho tests your word like a bad penny.I'd like you to be wise and not to giveYour word with too much assurance.Bertolt Brecht“In praise of doubt”Brecht Poems 1913-1956Methuen Publishers

Shift work and heart diseaseAcknowledgementsAcknowledgementsThe original idea for some of the studies, thatthis thesis is based upon originates more than10 years ago, before I even knew that shiftworkers presumably had a higher risk ofheart disease. Hans Jeppe Jeppesen, psychologistat the Department of OccupationalMedicine in Aalborg, upon reading a studyby Knutsson and coworkers, drafted an ideafor an intervention study. Shortly after Istarted working at the Department in 1991,he introduced me to the field of shift workand - well - we have been working togethersince. Thank you, Jeppe, for many hours ofdiscussions and cooperation. I guess we havelearned a lot from each other and maybebrought medicine and psychology a littlecloser together.My other coworkers, Anders Knutsson fromthe Department of Public Health and ClinicalMedicine, Occupational Medicine, UmeåUniversity, Hermann Burr and Finn Tüchsenfrom the National Institute of OccupationalHealth, Copenhagen, and Poul Suadicani,Hans Ole Hein and Finn Gyntelberg from theCopenhagen Male Study have all contributedsignificantly to the work that constitute thebase of this thesis. Especially thanks to Andersfor inviting me to the northern most partof Sweden for discussions, and for giving methe opportunity to further explore some ofthe questions raised later in this thesis in theSHEEP-VIP database.Jens Peter Bonde, Department of OccupationalMedicine in Århus and Bo Netterstrøm,Department of Occupational Medicine inHillerød volunteered as supervisors when Idesperately needed them, they have bothcontributed with ideas, discussions of theresults, and have been very supporting in theprocess of bringing this thesis together.Ben Jansen and Alex Klein from ATOSConsultancy in Holland generously providedthe Rota Risk Profile Analysis program usedin publication IV and V.The work has been carried out, while I wasworking as registrar and research fellow atthe Department of Occupational Medicine,Aalborg Regional Hospital from 1995-99.Part of the work was funded by grants fromthe Danish Working Environment Foundation(1995-2) and the Danish Heart Foundation.Sven Viskum has been acting as mylocal supervisor, and Jens Peter Johansen isacknowledged for finding ways of financingthe months of my work that was not coveredby the grants.I would also like to thank my other colleguesat the Department, who patiently have heardabout shift work and heart disease atnumerous occasions over the years.All the nurses and nurses aides, that volunteeredto participate in the NARFE II study(publication IV and V), and especially theworking group members at the wards, arewarmly thanked for answering all our questions,going to the laboratory for bloodtests and for allowing us the opportunity todiscuss working hours with them.Aino, my wife, you have been asking manyquestions, that often made me think it allover once more. Things would have lookedvery differently without you.Frederikshavn, December 1999.3

Shift work and heart diseaseContentsAcknowledgements ............... 3Contents........................ 4Introduction..................... 8Background ..................... 9Shift work..................... 9Shift work and health ......... 10Heart diseases................. 10Heart disease epidemiology....... 11Risk factors for heart disease...... 11Traditional risk factors ........ 11Other risk factors ............ 11Social class................. 11Work environment ........... 12Personality factors ........... 12Biochemical risk factors (biomarkers) 12Markers of atherosclerosis ..... 13Markers of hemostasis ........ 14Stress and heart disease.......... 14Definition of stress ........... 14Biomarkers of stress.......... 15Shift work and CVD ............ 15Epidemiological evidence ...... 15Mechanisms ................ 16Individual modifying factors .... 20Gender difference ............ 20Shift scheduling and prevention.... 20Ergonomics and shift schedules . 21Rota risk profile analysis (RRPA) 21Intervention in shift scheduling . . 22Shift intervention and biomarkers 22AcknowledgementsSummary of results .............. 30Publication II ................. 30Publication III................. 30Publication IV................. 30Publication V ................. 31Discussion ..................... 33Methodological problems ........ 33Selection bias and confounding .... 33Choice of reference groups..... 35Social class................. 35Work environment differences . . 37Other risk factors ............ 37Confounding or mediator ........ 37Exposure assessment............ 38Detailed shift work exposure . . . 39Measuring ergonomic criterias . . 43Measurement of biomarkers .... 47Information from questionnaires . 49Intervention and biomarkers .... 50Evidence of a relationsship ....... 51Shift work and stress ......... 52Is shift work a a risk factor for IHD? 53Is it a causal risk factor? ....... 53Is it a causal risk factor for all? . . 54Conclusion and perspectives....... 60Shift work as a risk factor for IHD . 60Future research ................ 61Preventive perspectives.......... 61Abstract ....................... 63Resumé........................ 64References ..................... 65General aims of the studies ........ 25Material and methods ............ 26Publication I .................. 26Publication II ................. 26Publication III................. 26Publication IV and V............ 274

Shift work and heart diseaseContentsThe thesis is based upon the following publications, referred to in the text by their romannumerals:I: Henrik Bøggild and Anders Knutsson. Shift work, risk factors and cardiovascular disease.Scand J Work Environ Health; 1999: 25: 85-99II:III:IV:Henrik Bøggild, Poul Suadicani, Hans Ole Hein and Finn Gyntelberg. Shift work, socialclass, and ischaemic heart disease in middle aged and elderly men: a 22 year follow up in theCopenhagen Male Study. Occup Environ Med 1999; 56: 640-5Henrik Bøggild, Hermann Burr, Finn Tüchsen and Hans Jeppe Jeppesen. Work environmentamong Danish shift and day workers (manuscript)Henrik Bøggild and Hans Jeppe Jeppesen. Intervention in shift schedule and change inbiomarkers of heart disease and stress at hospital wards (manuscipt)V: Henrik Bøggild and Hans Jeppe Jeppesen. Shift schedule characteristics and biomarkers ofstress and heart disease (manuscript)Abbreviations:AP Angina pectoris RRPA Rota Risk Profile AnalysisApo-A Apolipoprotein A RE RegularityApo-B Apolipoprotein B PE PeriodicityBMI Body Mass Index LS Load/shiftCI Confidence interval LW Load/weekCMS Copenhagen Male Study ON Opportunity for rest at nightCVD Cardiovascular disease PE PredictabilityHbA 1c Glycated hemoglobin OH Opportunity for household activitiesHDL High density lipoproteins OE Opportunity for evening activitiesIHD Ischaemic heart disease OW Opportunity for weekend recreationLDL Low density lipoproteinsMI Myocardial infarctionOR Odds ratioNARFE Not an abbreviation; A giant from Nordic mythology, the father of day and grandfather of nightNWECS National work environment cohort studyt-PAI tissue-Plasminogen activator inhibitorRR Relative rate ratioTC total cholesterolTG triglyceridesWHRWaist Hip Ratio5

Shift work and heart diseaseIntroductionIntroductionShift work was not a consequence of theindustrial revolution - The munks of themonastery, guards at the castle, and seafarershave been at work at odd hours since biblicaltimes. And the hazards (albeit not heart disease)of night work was recognized by thefather of occupational medicine, BernardiniRamazzini (1633-1714), who wrote in “DeMorbis Artificium” (1713) that bakers “workat night, and when other men have finishedthe day’s task and are asleep recruiting theirenergies, they must work all night and sleepall day, like bats”. Also learned men workingby the burning oil lamps have “the spirits(diverted) to other organs (causing) theirstomachs to abound in the acid of undigestedfood...” (quoted from Harrington 1 (p. 3)).Whether gastrointestinal problems are suggestedcaused by “spirits” or metabolicdesyncronisation the shift worker have at alltimes recognised that working hours in thenight influenced their health.The knowledge of an apparent risk of heartdisease is of newer date. The first studiesdealing with shift work and heart disease wasconducted in Norway by plant physicians,Thiis-Evensen at Norsk Hydro, Porsgrunn 2and Aanonsen at electrochemical plants inOdda 3 . Heart disease was included as part ofstudying morbidity in broader sense. Bothstudies have been regarded as essentiallynegative but when reanalysing the study byThiis-Evensen, a higher risk is detected (I).The first cohort study 4 was also interpretedas not showing any differences. It did, however,show that workers having left shift workhad a higher risk of heart disease. With thefairly inconclusive literature Harringtonconcluded in his classical review of shiftwork and health that “no excess cardiovascular(...) morbidity or mortality has been demonstrated”(p. 19) 1A second wave of studies on shift work andheart disease was published from the beginningof the 1980'ies and with three Swedishstudies 5-7 in the process of publication thefirst review to suggest an association betweenshift work and heart disease was publishedin 1984 8 . Especially the historical cohortstudy of 504 papermill workers by Knutssonet al 7 has been influential, as it was wellconducted and showed an increasing riskwith increasing number of years on shiftwork.Since then, the number of epidemiologicalstudies has slowly increased and recentreviews both of shift work and health ingeneral 9-12 , on work environment and heartdisease 13-15 , and on shift work and heartdisease in particular 16,17 all conclude that shiftwork and heart disease is associated.At the same time, studies on the possiblecauses of this relationship have been published,suggesting that the risk could be conferredthrough traditionel factors like smoking,blood pressure and diet (I).Another type of study related shift schedulingand the introduction of ergonomic criteriasfor scheduling to health and well-being,and a single study had shown that rotationwith the clock led to lower triglycerides andglucose than counter clockwise rotation 18 .This prompted the idea that the risk of heartdisease among shift workers could be reducedthrough the design of shift schedules.When we started the research related to thisthesis in 1993, I was rather convinced by thereviews. The work has broadend through theperiod and this thesis examines whether therelationship between shift work and heartdisease is causal, a prerequisit for rationalprevention.6

Shift work and heart diseaseBackgroundBackgroundShift workThere is no universally accepted definition ofshift work. It encompasses situations whereworking hours are at least partly situatedoutside normal day work. Day work, however,also lacks a definition, often just beingstated by two clock hours for instance between6 am and 6 pm but differing fromcountry to country.Shift work covers a temporal organisation ofwork that includes part or all evening andnight hours. In some definitions, amongothers the one used by ILO 19 , shift work isfurther delimited to situations where crewsor teams alternate to ensure a continuousoperation. In this definition, shift workwould not cover for instance waiters, whowould instead be labeled as having permanentdisplaced evening work.In this thesis shift work is used in the broadersense and defined as "work outside normalday working hours". Others would not regardthis as shift work, but it was chosen asthe effect of working hours is probably notrelated to whether the shift workers is alternatingwith others or not.Shift work may be organised in very differentways, several hundreds schedules beingdescribed in the literature 20 . In the perspectiveof the shift worker the organisation of theshifts in a shift schedule is viewed in terms ofhaving either one or more types of shifts. Inthe first instance where working hours forinstance are always 3 pm to 11 pm or from11 pm to 7 am, shift work is labeled permanentevening and permanent night work, inthe second instance working hours alternatebetween at least two different shifts and thisis labeled rotating shift work. Rotating schedulescan involve two or more (often three)types of shifts, in most instances coveringalso the day shift. Contrary to this perspective,from the organisations point of view, aschedule is considered as a way of organisingwork coverage, regardless of the type of shiftwork. A continous operation may this waybe covered by fixed shift staff or by rotatingshift workers without any real difference forthe organisation or employer. The focus ofthis thesis is the shift work of individuals.In industry shift work is traditionally done inteams, groups of workers having the sameshift schedule. One type of shift in this area isthe traditional 5 day shift, 2 days off, 4 eveningshifts, 3 days off, 4 night shifts, 3 daysoff. The schedule can be written DDDDD--/AAAA---/NNNN---/[1]. This way threeteams could cover a continuous operationfrom Monday morning to Friday at noon. Ifa longer operation is desired, for instance byincluding the weekend, more than threeteams would be necessary. In the aboveexample the schedule is three weeks inlength, in the fourth week starting over withthe day shifts. The teams would work oneweek delayed from one another. A schedulenot including work in weekends is labeleddiscontinous shift work.In other parts of society shift work is bytradition done in more irregular types ofschedules, examples being the transportationand the hospital care sectors. Here shift workis often not done in teams but instead everyemployee have their own schedule.Relating to the problems of defining shiftwork the number of shift workers inDenmark is not known. Recent publicationrates between 9% 21 and 18% 22 as shift work-[1] Throughout the thesis I use the notation “D” forday shift, “A” for afternoon shift (evening), “N” fornight shift, “-“ for day off and “/” to mark the beginningof a new week (sunday/monday)7

Shift work and heart diseaseBackgrounders. This is somewhat lower than figures C problems with sleep and sleepiness 33-35 ,from other Western countries 23-25 , where 25- eventually leading to effects on performanceand higher risk of accidents (at30% are reported shift working. The need ofshift work is relating to technical, economical work or while commuting) 36 .and service obligations, of which the firstC Social disruption with problems relatingmay explain the lower Danish figures. Bothto the interaction of work and socialof the latter reasons (for instance work organisationslike “production on demand”,32 .(family and spare time) lifewhere stock is being held at a low level C Risk of adverse pregnancy outcomes asrequiring flexibility and the increasing use of preterm birth, small for age and prolonged“around the clock” opening hours of shops) waiting time for conception 37-39 .explains why shift work seems to increase.C Disease. This covers psychological distressand maybe psychiatric disease asThere are no Danish figures, but in Swedenthe proportion of the work force with nightdepressionshift work has doubled in 15 years 23 .40,41 , gastrointestinal disordersand disease 1,42-44 , and heart disease. The12-hour shifts, dividing the day in two, is thesis will only cover the last of thesebecoming more and more popular in the items.United States and is also making its way intoDenmark. Besides the problems of shift workit also brings along problems relating to long Heart diseasesworking hours 26-31 but this is not coveredThe term "cardiovascular disease" (CVD)directly by this thesis.comprises ischaemic heart disease, cerebrovasculardiseases and diseases in larger arteriasand veins, especially in the form ofShift work and healthclaudicatio intermittens.By displacing the working hours, shift workIschaemic heart disease (IHD) or coronarypotentially affects the spare time and familyheart disease (CHD) comprises myocardiallife of all shift workers. The shift worker mayinfarction (MI), angina pectoris (AP), suddenexperience positive social effects of shiftdeath, cardiac failure, and cardiac arrhythmias45 . MI is a regional infarction in the heartwork, for instance when two spouses cantake turns looking after smaller children aton the basis of an occlusion of the coronaryhome 32 . Most shift workers regards the socialarteries. AP is attacks of sudden severe chestimpact as negative. Shift work that includepain caused by inadequate perfusion of a partnight work always leads to circadian disruption(see p. 15) and thus to physiologicalof the myocardium relative to the metabolicdemands. Coronary Artery Disease denoteschanges.coronary arteries affected by a pathologicalIn short, evening working hours is mostly process, and is often thought of as preexistingIHD by many years. While the reviewrelated to social and domestic problems, andnight working hours to physiological problems.Theories for the mechanisms of this is thesis will primarily relate to IHD.(publication I) included CVD broadly, thisdealt with in a later section (p. 14). TheThe pathological processes leading to IHDresulting negative outcomes falls in fourcategories 1,11 can be divided in two. The forming of anatheroma in the coronary vessels, and the8

Shift work and heart diseasesecondary forming of a clot leading to ischemia.The atheroma forms in the tunica intimaof the wall as a pocket with cholesterol,cholesterol-esters and other lipids being thefat-containing nucleus. Migration and proliferationof smooth intimal musclecells, collagen,eventually hydroxyapatit and dead cellsforms the plaque. The atheromatous plaqueproduced in this process can lead to a narrowingof the artery lumen, which diminishesthe coronary artery flow, eventually leadingto AP.IHD is secondly attributable to the formationof a clot often on the basis of a ruptured orulcerated atheromatous plaque, and leadingto a sudden trombosis of the lumen.Heart disease epidemiologyIHD is uncommen below the age of 40 and ismore prevalent among men than womenbefore menopause. After menopause, womenhave the same incidence as men. The genderdifference has been related to the protectingeffect of endogene estrogenes.In Denmark approximately 16,000 deathyearly are attributable to IHD, notably MIand sudden death, and 100-150,000 suffersfrom IHD (mostly MI and AP) 46 .In the last decades incidence and mortality ofIHD has been declining 46-48 . Some of thedecline has been related to a better prognosisand consequently a lower letality, but it alsoseems as if the incidence of new IHD isfalling 46,49 .Risk factors for heart diseaseBackgroundSince the late 1940'ies explanations for thedevelopment of IHD and detection of peopleat elevated risk of IHD have been pursued 50 .It was shown that the risk of IHD could bepredicted on the basis of conditions in clinicallyhealthy people, and these conditionswas labelled risk indicators or risk factors.Although some advocate the use of the termrisk factor for situations that predicts risk 51,52 ,others have suggested that the term shouldbe reserved for factors that are causallyconnected with the evolution of disease,while the term risk indicators could be usedfor factors that were just associated withIHD 53 . I use it in the latter meaning. Morethan 300 risk indicators have been described54,55 . Most of these are only proxy measuresof other underlying risk factors, and thenumber of risk factors are probably lower.On the other hand, risk factors discovereduntil now have only been able to explain halfof the changes in incidence, both in individualsand in populations 56 .Traditional risk factorsHypertension, diabetes, smoking and hypercholesterolaemiaare labeled modifiabletraditional risk factors 51 . Other conditionshave moved from indicator to factor status asmore knowledge has been gathered, and lackof exercise, overweight and a J- or U-formedcurve for alcohol consumption are nowadded as lifestyle related risk factors 51,57 .Other risk factorsAmong the enormous number of risk factors,only a few - relating to the papers and thisthesis - will be discussed.Social classSocial class can be measured in many ways,but will most often encompass either measurementof some kind of social prestige(normally education) or control with production58 , or both. In Denmark the scale ofSvalastoga 59 has been used in a revised ver-9

Shift work and heart diseasesion combining information on education andthe number of subordinates 60 . In this scale,social class I encompass self-employed withmore than 20 employees, people with longereducations and civil servants with more than51 subordinates, class II covers self-employedwith 6-20 employees or a middle-longeducation and civil servants with 11-50subordinates, social class III encompass selfemployedwith less than five employees andcivil servants with 1-10 subordinates. Socialclass IV encompass civil servants withoutsubordinates and skilled workers, and socialclass V covers unskilled workers.A large number of studies showed socialclass to be strongly inverse related to IHD 61-63. The possible causal effect of social classon IHD is not known. The social class termis in some instances a proxy for other riskfactors being more prevalent among lowersocial classes. It has been shown, however 64 ,that social class acts as a independent riskfactor, not explained by differences in otherrisk factors. One possible risk factor thatsocial class could be a proxy of is life style,as smoking and a diet rich in saturated fat hasbeen shown to be more prevalent in lowersocial classes. It has also been suggested thatpeople in lower social classes don’t quitsmoking and that prevention campaigns donot have the same effect among those. Besideslifestyle factors, the effect of social classhas been related to material conditions,selection processes, cultural differences andeffects in utero or infancy 65-67 . The differencesin work environment has also been suggested68 to be partly responsible, but thework environment in general is only explaininga minor part of the risk 69 .So, even when control of smoking and dietaryhabits between shift and day workers ismade, this would not necessarily outbalancethe confounding effect of social class.Work environmentBackgroundA number of work environment factors havebeen shown to relate to the risk of CVD.Sedentary work has been known since theearly 1950, when Morris compared busdrivers and conductors, and later on monotonouswork, passive smoking, noise, heat andcold have been included 13 . Occupationalstress is discussed on p. 12. Also differentchemicals are cardiotoxic 70 most known iscarbondisulfid shown in a series of wellconducted epidemiological studies 71 . Olsenand Kristensen 72 calculated the attributablerisk (etiological fraction) of work environmentrisk factors on the risk of CVD to bearound 20% that is that 1/5 of CVD caseswould not exist if all work environment riskfactors were entirely removed.Personality factorsIHD has been associated with the so-calledtype-A personality, persons being competetiveand with what has been labeled coronaryprone-behaviour,but it has later been shownthat it is the hostility and agression componentsthat are related to IHD 51,73 and toatherogenic lipids 74 .Biochemical risk factors (biomarkers)Morbidity and mortality of IHD is the endpoint of concern. In order to obtain informationon risk at an earlier stage, considerableinterests have been on the possible use ofbiochemical factors as cholesterol and lipoproteinsas proxy measures of later risk.Apart from serving as markers of apparentdisease when they are high - as hypercholesterolaemia- the levels of biochemical substancesincluded in the disease process is ofimportance. There is a dynamic equilibriumof proteins and lipoproteins between the10

Shift work and heart diseaseblood and the arterial wall, especially for thesmall high density lipoprotein (HDL)-cholesteroland mediumsized low density lipoprotein(LDL)-cholesterol. It is thought that theconcentration of for instance LDL-cholesterolin “normal” levels is connected to theprogression of disease so that the transportof lipids into the vessel wall will be related tothe concentration of LDL.As the concentration of these biochemicalfactors represent early, subclinical adversehealth effects they can be regarded asbiomarkers 75 . I use this notion to separatewhat is regarded as internal biological effectsfrom external behavioral risk factors. Asnoted above, the traditional nomenclaturewould label the biomarkers as risk factors.Markers of atherosclerosisThe blood lipids cholesterol and triglyceridesare in different concentrations integrated inlipoproteins of different densities.High total-cholesterol and LDL-cholesterol(especially its oxidized form 54 ) and lowconcentrations of HDL-cholesterol have beenshown to be causally related to risk ofIHD 51,76,77 .It has been debated, whether high triglyceridesare a causal risk factor, and the level isprobably not linearly related, but may bedependent on HDL-cholesterol levels 51,78 .In subjects without other risk factors, atherosclerosisis normally not present whencholesterol < 3.5 mmol/l, LDL < 2.5 mmol/l,triglycerides < 1.5 mmol/l, and HDL > 1mmol/l.It has been suggested that total cholesterol/HDL-cholesterolratio is especiallyusefull as a predictor. The ratio should belower than 5 52 .BackgroundThe risk factors and biomarkers are interdependent,and this may explain why for instanceJapan has a much lower incidence of IHDdespite a high proportion of smokers, due toa low level of cholesterol 79 .These biomarkers are influenced by genderand age and of life style factors. Femaleshave lower total cholesterol, LDL-cholesteroland triglycerides, and higher HDL thanmen. The relation to the other factors areshown belowTC HDL LDL TGhigher age (until> >55)Saturated fat> >Carbohydrates?Obesity? >Exercise> ?Smoking> ? > >Alcohol> >Table 1: Relation between risk factorsand biomarkers of IHD (adapted from 80 ).TC=total cholesterol, TG=triglyceridesA 1% reduction in total cholesterol has beenassociated with a risk reduction for IHD of2-3% 76 , a 10% reduction in cholesterol to20% reduction in IHD 52 , and a reduction of0.6 mmol/l at any point on the scale to a 50%reduction at age 40, a 40% reduction at age50 etc 81 .As a high blood glucose has been related tohigher risk of IHD, glycated hemoglobin(HbA 1c , see also p. 13) that mirrors theblood-glucose concentration in the preceedingmonths also act as a marker of atherosclerosisrisk 82 .Markers of hemostasis11

Shift work and heart diseaseWhen an atheroma is formed the second stepis the formation of a clot. This is regulated bythe hemostatic balance.The end point of the coagulation system isthe generation of an enzyme (trombin) thatconverts fibrinogen into the gel-like fibrin,and thus gives hemostasis. The fibrinolyticprocess, on the other hand, activates plasminthat turn fibrin into soluble degration products83 dissolving the clot.Markers of this delicate web are numerousfor among others fibrinogen, tissue-plasminogenactivator inhibitor (t-PAI) and factorVII there is knowledge to designate a highconcentration as causally related toIHD 51,54,84-87 . Also other markers of the coagulationand fibrinolytic system (trombocytes,platelet aggregation 52 , factor VIIc 88,89 , t-PA 51 and complement 90 ) are suspected to becausally related to IHD.Fibrinogen is - like the lipids - affected bydiet, tobacco, exercise and obesity 85 .Triglycerides are supposed to be a strongerand cholesterol a weaker predictor of risk inwomen than in men 77 , and the hemostasis andinsulin regulation is also acting differently inwomen 51 .Stress and heart diseaseDefinition of stressBackgroundThe word “stress” has a variety of definitionsbut can be anticipated as a “particular relationshipbetween the person and the environmentthat is appraised by the person as taxingor exceeding his or her ressources and endageringhis or her well-being” (Lazarus andFolkman 91 , p. 19). This interaction is characterisedby being unpleasant and tense, andcontains both emotional, cognitive, behavioraland physiological components. In thisviewpoint stress is an individual and relationalphenomenon implying that people canreact very differently to the same stimuli.Conditions in the work environment can besources of stress (in other definitions labeledas stressors) 92 : physical surroundings (light,heat, noise), new technology, quantitativeand qualitative demands, lack of goal, unclearroles, conflicts, insecurity, lack of participation,unemployement etc.In relation to IHD occupational stress havemostly been examined in two distinct models.In the Karasek-Theorell model 93 quantitivedemands and descision latitude is making upa two dimensional plane, in which both selfreported and “objective” scores can place theindividual. The quadrant with low controland high demands is labeled “strain” and isespecially stress inducing. The model hasbeen expanded with the inclusion of socialsupport as a third dimension and with iso(-lation)-strain as the most stressfull situationin this three dimensional model 93 .More than 20 studies, mainly finding a linkbetween strain and IHD have been published94-96. Recently, several studies have pointedat control as the important factor 97 .Another model, the efford-reward-imbalancemodel (ERI), has been proposed by Siegristet al. 98,99 stating that the demands of workare outbalanced by rewards, among othersmoney, promotion aspects and esteem. Ifpeople have certain personality traits (previouslylabeled “immersion” 100 , now “overcommitment”),a percieved inbalance betweenthe two are particularly stressful.Stress as defined by this model has also beenshown to be a risk factor for IHD 100,101 .Also a few of the other occupational factorshave been connected to IHD, especially anoisy environment 13 .12

Shift work and heart diseaseBiomarkers of stressIn situations taxed by the individual as stressful,an acute release of epinephrine, norepinephrineand a more slowly release ofcortisol is made 102 . This gives as an effect arelease of glucose and free fatty acids, usedas “fuel” in the “fight or flight” reaction.The effect of acute stress can be measured byepinephrine, norepinephrine and cortisol butneither of them are well suited for field experimentsbecause diurnal measurements areneeded 103 and the half-life in blood is minutesto hours 104 .If the situation anticipated as stressfull continuesa change in the release of hormonesare seen. In a chronic strained person anabolichormones such as testosteron, dehydroepiandrosteronand estradiol are depressed.Other anabolic functions and the immunesystem are also lowered.Stress has been shown to alter lipid levels innon-laboratory situations, but eventually onlythe more severe forms 74 for chronic stressmostly studied in relation to demands andcontrol and to job insecurity. Only few studieshave found a relation 74 although anotherreview found that stress may bring about arise of 25% in cholesterol 105 . One explanationmight be that the perception of threat in thesituations hardly ever is evaluated.Triglycerides has been proposed as a indicatorfor stress as epinephrine mobilizes freefatty acids 106 .Clotting factors have been shown to berelated to stress among tax accountantsincreasing blood coagulation 107 , but theliterature in general does not agree whetherstress is related to clotting factors 104Because stress is followed by higher bloodglucose, glycated hemoglobin as an integratedmarker of blood glucose for the previous1-2 month (corresponding to the half-life ofBackgrounderythrocytes) will be higher as a marker ofthe metabolic regulation 108 109 . Also otherglycated proteins as glycated fructosamincould be used 110 104 .Prolactin has in several studies by Theorell etal. 111,112 been shown to increase in stressfullencounters, especially when the person ispowerless and passive (compared to being incommand and active) or responding to thestressfull encounter with passive coping.The stress response may, however, be morecomplicated than outlined. The body seeks tomaintain equilibrum and changes may beleveled out leaving surprisingly steady concentrationsdespite high stress levels 104 . Fromthis point of view biomarkers not part of themetabolic equilibrium itself may be moresteady markers (eg. HbA 1c ).Shift work and CVDEpidemiological evidenceThe epidemiological literature on shift workand CVD has been reviewed in publication I.It examines 17 studies (including publicationII) and it suggests that shift work is associatedwith an increased risk of CVD. The bestavaliable estimate is a relative risk of 1.4from the study by Knutsson in 1986 7 andsupported by other studies of different methology.Some studies were not included inthe review due to a demarcation to Englishor Scandinavian publications with comparisonsbetween day and shift workers. A Germanreview 17 included 8 studies in the Germanlanguage. One study was excluded forcomparing three and two shift workers 113 ,and one study because it was not clear whetherthe control groups was in fact dayworking 114 . [2][2] Two reviews on shiftwork and heart disease havebeen published since publication I:13

Shift work and heart diseaseAmong the 17 studies included in the reviewsome did not find an association betweenshift work and CVD. This prompted a furtherand deeper study of the literature 115 . In ameta-analytic approach of the 17 studies weidentified two groups of longitudinal studieswith individual exposure classification, oneconsisting of four “positive“ studies, and oneof five “negative” studies (table 2). As thebasic idea of a meta analysis is that the includedstudies should be measuring the samerelationship between well defined exposureand outcome, this is probably not fullfilled -as is often the case in meta analysis of observationalstudies. The chi-square test of heterogeneityis not very powerfull, but theresult suggests that at least two differentgroups of studies are present.BackgroundWe further tried to identify study characteristicsthat influenced the meta-regressionanalysis. This only suggested that studieswith an aggregated exposure classificationgave higher risk estimates than studies usingindividual exposure classification. This wasnot anticipated.MechanismsSeveral models have been proposed to explainthe relation between shift work andhealth 16,122,123 , mostly suggesting a web ofcauses. None of the models have been dedicatedto explain the risk of IHD.The other part of the review (publication I)discussed possible explanations for a causalrelationsship between shift work and heartdisease. Expanding the model proposed byGroup of studies OR meta(95% CI) P 2 (p-value)“positive” 116;7;117;118 1.31 (1.17-1.45) 0.26 (0.97)“negative” 119;120;121;4;(II) 0.96 (0.85-1.08) 2.06 (0.72)Table 2. Meta analysis of longitudinalstudies of shift work and IHD with individualexposure classification. Aggregatedodds ratio with 95% confidenceintervals (CI) and a chi-square test ofhomogenicity with n-1 degrees of freedom.(From 115 )Steenland K. Shift work, long hours, and cardiovasculardisease: a review. In: Schnall P, Belkic K,Landsbergis P, et al, eds. The workplace and cardiovasculardisease (Occupational Medicine: State ofthe Art Reviews). Philadelphia: Hanley & Belfus,Inc., 2000:7-17Dong W. Shift work and cardiovascular disease.Israel Journal of Occupational Health 2000;4:33-38Neither adds substantially to the knowledge (appendedjuly 2000)Figure 1. A possible causal web betweenshift work and IHD. Width of arrowsare not related to importanceKnutsson 16 we suggested four possible interactivemechanisms as part of a causal weband identified studies that had dealt withthese (see figure 1)[3].[3] An updated version will appear: Knutsson A,Bøggild H. Shiftwork and cardiovascular disease:14

Shift work and heart diseaseA (see figure 1): Most of the functions in thebody has a rhythm during the day, so that forinstance concentration of hormones andlipids, temperature, hand grip strength, airwaynarrowing etc. each day follow theirown pattern (figure 2). These biological andpsychological circadian[4] rhythms aredriven by an autoregulatory genetic clock 124in the suprachiasmatic nuclei of thehypothalamus 125 and exists even in darkcaves where no zeitgebers[5] as thedark/light pattern are present. The robustclock which in most individuals runs atapproximately 25 hour is synchronized to the24-hour external environment by entrainmentof dark/light pattern, watches, socialactivities etc. Melatonin released from thepineal gland may act as the internalmessenger 126 .When working at night and sleeping duringday these circadian rhythms move but evenafter weeks of night work no completeadjustment of the rhythms are made 127 . Thesingle rhythm moves at its own pace andFigure 2. Circadian rhythm fitted inmeasurements. A hypotheticaldesynchronisation with a phase-delayof 3 hours is shown in grey.A=amplitudereview of disease mechanisms. Reviews on EnvironmentalHealth 2000; (in press) (appended july 2000)[4] From Latin: circa dies, around 24 hours[5] From German: time signalBackgroundseveral rhythms may come in disharmonywith each other or the surrounding. This islabeled internal desynchronization 35,128 .In a rhytmic function the acrophasedescribes the timing of a rhythm in relation toa reference time point, representing the crestof the cosine curve best fitting the data (seefigure 2). The mesor is the midline estimatingstatistic of the rhythm, being the valuemidway between the highest and lowestvalue of the curve best fitting the data. Theamplitude is the distance between the mesorand the highest point on the fitted curve 129 .Also factors relating to the cardiovascularsystem follows circadian rhythms 130,131 andmay be desynchronized during night work.Some factors (like blood pressure) areprimarily externally influenced for instance ofactivity 132 but the rhythm is still controlled bythe internal clock, this external influencecalled masking. In some 133-136 , but not all 137-139of the studies on night shift work andblood pressure, a longer plateau of highsystolic pressure is described in the nightrhythm.It has been shown that also lipids have acircadian rhytmicity, and in day orientationthe circadian variation (given as percentageof the total variation) was 5.6% for HDL/totalcholesterol ratio, 30.5 and 31.6 for HDLand total cholesterol, 33.5% for LDL and38.5% for triglycerides 140 (see figure 5).These circadian rhythms follow a primarilyendogenic circadian rhythm 141 , withtriglycerides having the acrophase early inthe morning, the others later in the day.Fibrinogen peaks in the morning with a 10%amplitude 142 . t-PAI has its acrophase at 3 amwith a amplitude of 50% 131 . Protrombin(factor II-VII-X) seems to be lowest at night(although the results are conflicting) 142 .There is also a yearly rhythmicity, cholesteroland LDL-cholesterol being up to 3-5%15

Shift work and heart diseasehigher in winter (cold period), while HDLand triglycerides are only modestly changed80 .Several studies in the last years have examinedthe influence of meal timing on the postprandiallevels of lipids and glucose. In onestudy the postprandial response were lowerfor cholesterol-rich lipoproteins at 1 amcompared to 1 pm, while the response oftriacylglycerol-rich lipoproteins were higherat night 143 . In another study the insulin responsewas found to change during nighteating 144 , but not replicated in a new study bythe same group 145 using a different diet.These studies suggest that the timing of foodintake may influence the metabolic function.[6]The amount of carbohydrates ingested in thenight hours was associated to cholesterollevels 146 , and it was suggested that this wasdue to an internal desynchronisation betweeneating and metabolism.The literature on shift work and desynchronizationin relation to IHD is very fragmentedwith only very few studies available (I). Onthe other hand there is indications to believethat this pathway may be important.B) Shift work leads to changes in lifestyle,for instance in relation to diet, smoking,alcohol and exercise, all being risk factors ofIHD. The literature shows that changes in forinstance smoking habits must be very slow 147 ,with shift workers followed for ½ year notBackgroundbeginning to smoke[7]. Furthermore, smokingmay be viewed as a way of coping withthe stress of shift work. Lifestyle changesmay also be more directly influenced by theshift work, as for instance cantinas may beclosed at night 148 leaving little possibilities forbetter nutrition.Most of the literature on shift work and IHDhave concentrated on lifestyle (I), and weconcluded that shift workers smoked moreand ate more irregular than day workers,while exercise and alcohol consumptionswere not affected by shift work.C) The interaction between work and familylife may lead to stress. Shift workers have tosleep in the day time, may be unavailable forthe family in weekends, and may haveproblems in attending regular social arrangements,such as evening school. Sleep is oftencompromised for both physiological andsocial reasons and may be compromisedwhen shift workers choose between familytime and sleep.Several studies 149-153 have suggested twomajor contributors of stressfull experiencesin shift work:C Physiological disruption of circadianrhythms, leading to sleep problems, fatigueand exhaustion.C Social conflicts, as working hours inevening and night collide with familyinteractions.[6] In a recent study (Spiegel K, Leproult R, VanCauter E. Impact of sleep debt on metabolic andendocrine function. Lancet 1999;354:1435-9) it wasshown that sleep deprivation (4 hours/day for 6 days)gave a 40% decrease in glucose tolerence. (appendedjuly 2000)[7] However, in a recent study (van Amelsvoort L.Cardiovascular risk profile in shift workers: cardiaccontrol, biological and lifestyle factors (thesis).Wageningen: Grafisch Service Centrum Van GilsB.V., 2000:1-131) 81 smoking shift workers startingon a new job smoked 2.4 cigarettes more after oneyear of follow up. It was calculated, that it wouldincrease the risk of IHD with 10% over 20 years.(appended july 2000)16

Shift work and heart diseaseThe identified areas have not been associatedwith emperical data, but is mostly based onknowledge of the effects of shift work ingeneral and on theoretical considerations.The strain of shift work is this way tied toother factors than normally examined inwork related stress. The shift workers maymoreover be subjected to the same stressorsas other workers, for instance in relation tomonotony, high demands/low control ornoise, and it has been theoretically arguedthat the shift worker may be more sensitive154 .Neither social strain nor sleep problems havebeen tied to IHD in shift work (I)Recently a couple of studies have examinedstress in shift workers in relation to IHD.The articles have been using the demandcontrolmodel of Karasek and Theorell,finding no difference between shift and dayworkers155 or the Effort-Reward-Imbalancefinding that imbalance had an influence onthe association between shift work and hypertensionbut not with lipids 156 .In a couple of studies of shift work and IHDdemand-control measures have been used tocontrol for stress 117,118 . In the Helsinki HeartStudy shift workers had lower control butthe same demands as day workers, also aftercontrolling for social class (own reanalysis).D) Biomarkers of stress and IHD could bethe link to disease, the literature finds shiftworkers in general to have higher total cholesterol,triglycerides, and a slightly higherblood pressure (I). Markers of hemostasisand stress have not been dealt with, exceptfor one study using HbA 1c . These shift workerswere, however, also exposed to otherstressors (noise, overtime) 157 .BackgroundAs stated, the focus of the literature has beenon lifestyle factors, and almost no studieshave examined the other two suggestedmajor pathways, circadian rhytmicity disruptionand stress as a consequence of lack ofsleep and social disruption (I).Since the review, another three studies onshift work and blood pressure have beenpublished 135,138,158 , Morikawa et al finding ahigher risk of hypertension among youngshift workers in a cohort design.Two new possible explanations for an associtationbetween shift work and IHD have beenput forward since the review:In an analysis of cross sectional data from theSwedish WOLF study 159 shift workers havebeen shown to have higher prevalence of the“metabolic syndrome”, a collection of symptomsand signs related to the metabolism,which is also a risk factor for IHD. Theanalysis comprised 734 shift and 855 dayworkers. Shiftworkers had a higher prevalenceof triglycerid above 1.7 mmol/l, lowerHDL-cholesterol, higher waist-hip-ratio(WHR) but no difference in BMI. Shiftworkers had lower prevalence of hypertension.The shift workers had more often severalrisk factors present than day workers.The higher WHR has also been reported in across sectional sample of japanese shiftworkers 86 .In two studies from the same Japanese group,cross sectional 160 and longitudinal data 161show shift workers to have a prolonged -QTc-interval on their ECG, which has beenshown to raise the risk of CVD. In the longitudinalstudy 161 158 blue collar shift and 75white collar day workers were followed for10 years. HDL, triglycerides and bloodpressure did not differ. At baseline, shiftworkers had already prolonged QTc, butamong those with normal QTc (

Shift work and heart diseaseged interval after 10 year.[8]Individual modifying factorsShift workers may react very differently tothe shift work exposure. A variety of personalityfactors have been proposed to influencewhether a shift worker will be able tocope with working hours. The factors suggestedto lower adaptation to shift work includeneurotism, introversion, lack of hardinessand sleep patterns (rigid sleep, morningness(lark)) 162,163 . Besides this age, experiencewith shift work, an individual circadian variation,the experience of having chosen shiftwork, and satisfaction with work have beenrelated to succesfull adaption to shift work.Outer factors, such as possibilities foremployment, risk of unemployment, motivation,work load, challenges in work, housing(ability for sleep), transportation etc may alsoinfluence the ability to cope with shiftwork 154,164 . The effect of these differenceshas not been related to IHD.The individual dimension in the stress reactionmakes people react differently to thesame stimuli. If, for instance, noise is expectedand with a function for the individual thebody reacts differently than if the noise isunexpected and without a function 165,166 . Thismay parallel shift work where problems withwork-family life interactions may be accepteddue to economical incentives, and shift workersthat feel they voluntered to shift work[8] A third possible explanation was recently proposedby van Amelsvoort (van Amelsvoort L. Cardiovascularrisk profile in shift workers: cardiac control,biological and lifestyle factors (thesis). Wageningen:Grafisch Service Centrum Van Gils B.V.,2000:1-131). Shift workers had more ventricularextrasystoles in 24-hour Holter monitoring after oneyear of follow up than day workers, mostly so inworkers having > 5 night shifts a month. (appendedjuly 2000)Backgroundmay act differently than those feeling shiftwork to be forced upon them. It has, however,not been examined.Gender differenceIn relation to stress, it has been shown thatgender differences exist in the cognition ofstress reflecting both biological (metabolic)and social differences being present betweenmen and women 102 . Males have a rewindingof the stress hormones after going homefrom work, while females have a persistenthigh concentration because of the responsibilyto look after the home. Also female shiftworkers experience more responsibility forchild care and often have a more fragmentedsleep 167,168 . In the epidemiological literaturethe risk of IHD for female shift workers is inthe same magnitude as for men (I).Shift scheduling and preventionWithin the framing of a shift schedule, that isthe number of workhours covered, the numberof persons needed at different hours etcetera the schedule can be made up quitedifferently. The schedule may more or lesstake circadian disruption and social/familylife - work conflicts into account.Danish legislation offers no direct help indrawing up schedules. The law 169 only statesthat workers (regardless of being shift or dayworkers) should have at least 11 hours off inevery 24 hour span, and a full 24 hour off inconnection with the 11 hour daily rest periodonce every 7. day. Both periods may bepostponed until respectively 8 hours of dailyrest and a rest day every 11. day. This isnormally done after consultation with unionrepresentatives. The EU-directive on workinghours 170 did not change these constrains.Unions have agreements with regulations18

Shift work and heart diseaseconcerning for instance the mean number ofhours worked a week, but normally no constraintson the shift schedule are given. Sothere is very wide opportunities to scheduleshift systems.It has been discussed what type of shiftsystem that would lead to the least strain.There is no universal agreement 171-173 . Theprevaling European view has been to keepcircadian rhythms as close to day orientationas possible, while Americans have suggestedvery long spells (months) of night shifts inorder to allow the circadian rhythms tosynchronize to night working. A problem inthe American tradition is that even permanentnight shift workers maintain a day orientedlife at days off, allowing the circadianrhythms to a rapid synchronization with dayorientation. In a recent publication fromNIOSH 174 [9] the European view has thereforebeen adaptedSeveral studies have shown day sleep to belongest and best in slowly rotating and permanentsystems 171,175 and permanent shiftworkers have more often the feeling of havingchoosed to work at night 176 . There isonly limited knowledge of the effect of permanentshift work and IHD. In an aggregatedstudy permanent night shift was found tohave the same risk of hospitalization asrotating shift 177 .Ergonomics and shift schedulesSeveral groups have evolved ergonomiccriterias for shift scheduling with the aim ofminimizing the adverse effect of shift- andnight work 174,178-181 . It it important to noticethat the criterias are predominantly based onphysiological and circadian knowledge althoughsocial factors are also covered. The[9] National Institute of Occupational Safety andHealthBackgroundcriterias may be in conflict with each other.Different European proposals do not differmuch, and one of the most widespread list isfrom the BEST-group 181 , suggesting 14criterias. They are:C Minimise permanent nightsC Minimise sequence of nights (to 2-4 nightshifts in succession)C Avoid fast double backs (with a shorttime off between two shifts (e.g. 8 hours))C Plan rotas with some weekends offC Avoid overlong work sequences (compressedworking weeks; 12 hour shifts)C Fix shift length to task loadsC Consider shorter night shiftsC Rotate forwards (e.g. morning shift, thenafternoon, then night)C Delay morning start (to after 7 am)C Allow flexibility in shift change timesC Keep rotas regular (for planning a normallife)C Allow some individual flexibilityC Limit short-term rota changesC Give good notice of rotasRota risk profile analysis (RRPA)There have been a number of papers onalgoritms for evaluation of shift schedules onergonomic principles 182,183 , but no commonaccepted way of doing it. Some of the algoritmshave been translated to computerprograms. Besides the inventors own use noformal evaluation has been made of theirappraisability and no comparison betweenthe algoritms or programs. Other computerprograms have been written that can helpdesigning ergonomic better schedules 184 .19

Shift work and heart diseaseThe Rota-risk-profile-analysis (RRPA) byJansen et al. 185 computes nine scores fordifferent schedule characteristics. The basicfeatures of a schedule are keyed in the program.These include types of shifts (day,evening, night etc with start and stop hours)and the actual schedule. The calculation of aschedule results in two types of scores, onebasic score ranging from 0-10, with 0 for badand 10 for excellent. The basic score is calculatedas a mean of all weeks in the schedule,and a constancy score provide a figurethat is high if the weeks in the cycle are alikeand low if the weekly scores are very different.The constancy score range from 0-100. The combined score is then calculatedas basic score - (100-constancy score)/100The scores and their definition have beengathered in table 3, p. 22.Eventually, the program calculates meanworking hours/week, operating time, workinghours between 22-06 in rota, the numberof times with less than 11 hours betweenshifts, the number of times with less than 36hours rest between blocks of shift, and thenumber of sundays off in the rota.The scores depend heavily on the assumptionsin calculations and in some instancesother values could have been selected (thenumber of night shifts in a row is for instancereflected in the PE (cumulating circadiandisturbance) and could be valued higher).The benefit of the RRPA, however, is thatthe schedules are summarized in a variety ofscores in a meaningfull and “objective” manner.Intervention in shift schedulingBackgroundHarrington 1 stated that “surprisingly few[scientists] have tried to evaluate new rotasbefore, during, and after, introduction” (p.3).20 years later this is still so, at least if interventionsare seen as planned and appliedactivities, designed to produce designatedoutcomes 186 . Most intervention studies in thearea still seems to be coincidental in design,being made when shift schedules are changedat management decisions, and not from atheoretical idea, followed by a change inschedule. Furthermore, several interventionstudies lack a control group 187-189 , whichgives the possibility for a Hawthorne effectthat the interest itself is capable of producinga positive outcome, regardless of its content.Most of the controlled studies are eitherexamining a change from counter clockwiseto clockwise rotation 18,187,190 , a reduction inthe consecutive number of nightshifts 189,191,192 or a change towards permanentnight shift work 188,193 .A couple of studies are examining severalchanges simultanously, including both achange in the direction of rotation and in thenumber of consecutive night shifts (from 6-7to 3-4) 194,195 , a change towards permanentcombined with forward rotation 196 , an interventionaimed at forward rotation,predictability and participation 197 , or fasterrotation, extra day off after last night shiftand fixed evenings off 198 .In general, satisfaction in all of theseinterventions were high, although theemployee in some instances - despite higherwell-being in the new schedule - chosed theold schedule, most often due to longer spellsof days off 18,197 . In most instances the changeled to better sleep quality 187,188,191,194,196,197 ,while symptom scores in general did notchange.Shift intervention and biomarkersTwo studies have evaluated certaincharacteristics of ergonomically scheduledshift systems in relation to biomarkers of20

Shift work and heart diseaseheart disease.Orth-Gomér 18 followed 46 male police officers.In a cross-over design, half of the menwere first followed in clockwise rotation forfour weeks, then counter-clockwise foranother four weeks. The other group workedvice-versa. Sleep quality was better at clockwise,but with no difference in symptomscores, smoking pattern or self reportedphysical and psychological strain. There waslower irritability at clockwise rotation. Therewas a change in triglycerides so that valueschanged toward lower values in the clockwiseperiod and worsened in counter-clockwisecompared to the middle period. For glucoselower values appeared in both clockwise andcounter-clockwise rotation, but mostly so inthe clockwise period. Also systolic bloodpressure was lower after clockwise thancounter-clockwise rotation.Kecklund et al. 199 followed an interventionwith a change from 7 to 3 and 4 night shiftsin a row at a power plant. 72 shift workerswere followed during two years, and 48 ofthese and 15 of 60 day working controls hadblood pressure, triglycerides, cortisol andtotal-, HDL, and LDL-cholesterol measured.Sleep was improved, the social problemswere lower, and there was a general satisfactionwith the new scedule. In the period HDLrose and LDL/HDL-ratio fell in the interventiongroup. Also both systolic and diastolicblood pressure fell.Two other studies with pseudochanges in theshift schedule and measuring of biomarkersshould be mentioned:12 new shift workers and 13 new day workerswere examined before starting and afterBackgroundsix month of work 147 . All had normal cholesteroland triglycerides. Height, weight,blood pressure, cholesterol, triglycerides,apoB and apoA1 were measured. At baselinethe shift workers were younger with a higherBMI and higher systolic blood pressure butlower cholesterol and apo-B than the dayworkers. During the six month follow up theapoB/apoA-1 ratio rose more in the shiftworkers, than in the day workers. No otherdifferences were found.In a uncontrolled study 200 , 447 workers at acar factory had their schedule changed fromtwo shifts 8-17.30 and 21-6 to a new twoshift schedule 6.30-15.15 and 16.15-1.00.Using data from the company medical recordswith measurements before and afterthe change it was shown that cholesterol fellfrom 201.8 to 195.2 mg/dl, HDL rose from53.3 to 62.2 and triglycerides fell from 109.2to 98.9. Systolic blood pressure fell from122.1 mmHg to 116.9 and diastolic from77.2 to 74.5 mmHg.These four studies all suggests that the outlineof the schedule will influence biomarkersof IHD.[10][10] van Amelsvoort (van Amelsvoort L. Cardiovascularrisk profile in shift workers: cardiac control,biological and lifestyle factors (thesis). Wageningen:Grafisch Service Centrum Van Gils B.V., 2000:1-131) showed in a one year follow up of 150 day and227 shift workers that HDL rose and total- and LDLcholesteroldecreased in both groups. LDL/HDL-ratiodecreased mostly in shift workers . Also decreases inBMI and WHR were seen in shift workers. (appendedjuly 2000)21

Shift work and heart diseaseBackgroundTable 3. Rota risk profile analysis (RRPA) scores and their definitions (adapted fromJansen and Kroon) 185scoreRegularity (RE)Periodicity (PE)Load/shift (LS)Load/week (LW)Opportunity for rest atnight (ON)Predictability (PE)Opportunity forhousehold activities(OH)Opportunity for eveningactivities (OE)Opportunity for weekendrecreation (OW)descriptionChanges in work time, frequency and nature1) number of changes in start and stop hours in continous shifts2) the timing of the change3) changes in start hours are considered twice as difficult as stop hoursThe changes are linearly standardizedExtent of circadian disturbance by the schedule1) weighting the worked hours2) accumulation over consecutive shifts (no adaptation)3) correction for direction of rotation (10%)The corrected changes are linearly standardizedMean number of hours pr. shift, standardized with a parabolic figureassuming a cummulative effect of long work hoursMean number of hours pr. week, standardized with a parabolic figureassuming a cummulative effect of long work hoursMean number of hours worked between 23-07 transformed liniarlyA measure of the extent that the schedule can be “mentally” foreseen1) length of rota in weeks2) multiplying the number of diffent shift types and clusters found over thecycle3) the number of stand-by shifts (maximum 10% of basic figure)liniarly transformed.The constancy score depict the variation in the weeks in the number ofshift clusters and the number of diffent shiftsAmount of daytime free on weekdaysAmount of eveningtime (19-23) on weekdaysNumber of saturdays or sundays with at least 21 hour consecutive hoursoff. An entire free weekend is multiplied with 222

Shift work and heart diseaseGeneral aims of the studiesGeneral aims of the studiesThe overall aim of the thesis and the studiespresented is to examine whether shift work isa risk factor for IHD. I will use two differentapproaches; The first (publication II and III)is to address two possible problems in theepidemiological literature, relating to theselection of comparison groups. One is theapparent need for controlling for social class,the other is the possible differences in workenvironment conditions between shift- andday workers.If shift work is related to heart disease, thenpossible causal links between exposure anddisease will be needed to explain the association.While these possible links in a causalweb is discussed in the review (I) the secondapproach would be to change the shift schedulingin accordance with ergonomic criteriasand examing if biomarkers for heart diseasechange. If they do, it will add to the credibilityof the claim of a causal relation. Thesecond group of publications (IV and V)attacks the problem from this point of view.A third approach uses all five publications ina discussion of the definition of shift work asan exposure. In the literature shift work ismost often used as a bivariate exposure;However, problems relating to this approachand a suggestion of several underlying modalitieswill be presented. If shift work is causallyconnected to IHD, then it should bepossible to identify characteristics of theschedule that are more harmfull than others.The aims of the studies are thus to investigate:- whether shift work is a causal risk factorfor IHD, or whether confounding from socialclass and work environment could be responsiblefor the association.- whether changes in shift work scheduling inaccordance with ergonomic criterias can leadto lower biomarkers of IHD- how different shift work exposure definitionsmight influence the biomarkers of IHD- whether stress can be responsible for therelationship.As the modern society is based on workersbeing on shift, and as shift work is thereforenot avoidable, the demonstration of lowerbiomarkers for IHD after an intervention andthe possible identification of schedule characteristicsrelated to biomarker levels wouldleave potential possibilities for preventionthrough shift work scheduling.This would be the ultimate goal of the workbeing done; to show ways of lowering apossible shift work induced risk of heartdisease among shift workers.23

Shift work and heart diseaseMaterial and methodsMaterial and methodsPublication IPublication I reviews the shift work literatureboth on the risk of CVD and on possiblecausal mechanisms. The literature was foundthrough Medline (from 1966) and OSH-ROM, and we handsearched the proceedingsfrom meetings in the ICOH Commissions onNight and Shift Work and on Work andCardiovascular Disease. Eventually listedreferences from the literature and personalcontacts were used. The search revealed 17epidemiological studies, comparing shiftworkers with day workers on CVD endpoints(mortality and morbidity). The secondpart of the review used a model to explorethe possible mechanisms in relation to knownor suspected risk factors, risk indicators andbiomarkers.The review has already been cited in thebackground chapter.Publication IIPublication II is based on a reanalysis of theCopenhagen Male Study (CMS). CMS is aprospective longitudinal cohort study,established in 1970 to address physical activityas a risk factor for ischemic heart disease.The participants were examined again in1985-86. The original cohort comprised5249 men aged 40-59 in 14 larger companiescovering railway, public road construction,military, post, telephone, customs, nationalbank and medical industries. At the secondbaseline all survivors were invited, and 3387participated. In this reanalysis only informationon working time was used from the1985-6 baseline. Information of exposure atboth baselines came from a questionnaire,and comprised of shift work, night work andirregular working hours, and of daytimework.Social class and risk factors (alcohol, smoking,leisure time physical activity, sleep,history of diseases) were also collected fromquestionnaire, and in a physical examination,fitness value, height, weight and blood pressurewere measured.Information on hospitalisation of IHD andmortality of all causes up until 1993 wasfound in the National Health Service registerand from the Danish Institute of ClinicalEpidemiology.Baseline differences of the shift and dayworking cohorts were compared with logisticregression models, and the relative risks ofIHD and all cause mortality over 22 yearswere calculated by way of Cox proportionalhazards regression models.Publication IIIPublication III is based on a reanalysis of theDanish National Work Environment CohortStudy. In 1990 a sample of 9653 was drawnrandomly from the Central Population register,of which 8664 were interviewed. 5940worked as employees within the last twomonth prior to the interveiw. The telephoneinterview comprised information on a numberof working condition. Working hourswere divided in four categories, night workand three-shift work, evening work and twoshift,other irregular working hours includingmorning work, and day work. The reanalysisexamined the prevalences of workenvironment factors, known or suspected tobe risk indicators or risk factors for IHD inshift and day workers. The workenvironment factors examined were length ofworking hours, ergonomic exposures, repetitiveworking tasks, dust, heat and noiseesposure, walking or standing postures,quantitative and cognitive psychological24

Shift work and heart diseasedemands, decision authority, social support,conflicts at work and job insecurity. Informationon social class was also included toexamine whether differences were due tosocial class.Odds ratio for prevalence of the work environmentfactors in each of the three groupscompared to day work was calculated inlogistic regression models, controlling forsocial class. In order to examine the impactof age, the material was stratified on age inthree groups and the models were repeated.Publication IV and VPublication IV and V are based on the NAR-FE II-project. It was a longitudinal controlledquasi experimental intervention studyamong nurses and nurses aides at 10 hospitalwards at Aalborg Regional Hospital. Theintervention consisted of a change in principlesfor shift scheduling introducing fourprinciples consisting of more regularity andpredictability, a maximum of 3-4 consecutivenight shifts in a row followed by an extra dayoff after last night shift, maximizing thenumber of weekends off, and a change fromthree to two types of shifts. Six wards wereoriginally included in the discussion of changingschedules. One ward was divided in twojust before the intervention and of the sevenwards, four choosed to start scheduling theirrotating shift work according to ergonomicprinciples. Two wards who could not findways of changing schedule agreed to continuein the project as control wards of theintervention. Among the four interventionwards two introduced all four principles, andthe other two only one-three principles.Furthermore, the change only encompassed(some) of the staff having rotating shifts, andnot the fixed evening or night shift staff. Fouroutpatient clinics with day work only, werealso included as controls. The staff wereMaterial and methodsindividually approached and gave informedconsent to participate in the evaluation.254 staff members were invited, of which172 (67.7%) agreed to participate at baseline.At six months follow up, 17 had left thewards and of the remaining 155 nurses andnurses aides, 101 participated at the seconddata collection. At the intervention wards athird data collection was done after 12months but participation rate was only 36%(36 ward personnel of 100 at four wards).The data collection included:At six of the wards (intervention and interventioncontrol ward), a questionnaire wasused in the pre-intervention process in September1996, with questions on attitudestowards shift scheduling, social factors, andsymptoms. These informations were used toevaluate differences between participants andnon-participants in the intervention period.Just before the intervention was put intoeffect in April 1997, a questionnaire, a requisitionfor blood tests and a diary was distributedat the intervention and interventioncontrol wards. For a smaller group the materialwas supplied in May. At the day controlwards the material was distributed oneyear later, in April and May 1998.The questionnaire was based on the StandardShiftwork Index 201 , and among others includedinformation on age, having spouse andchildren, personality factors, sleep type andquality, heredity, moonlightning, attitudes,impact and satisfaction with schedule. Thesymptom scores included selfrated health,physical health questionnaire, scales onstress, vitality and mental health. Lifestylefactors included exercising, smoking, alcohol,and coffein consumption. Social impactand conflicts were included, as also questionsrelated to commitment with collegues, worksatisfaction, demands, control, and socialsupport at work (for further information on25

Shift work and heart diseasequestions, see publication IV, appendix).Blood sampling was done between 8 and 10am at the local Department of Biochemistryby trained bioanalytics, at least three daysafter the last night shift, and after eight hoursof fasting. Analyses included glycated hemoglobin,immunoglobulin A and prolactin asmarkers of stress, blood glucose to controlfor diabetes, and triglycerides, total cholesteroland HDL-cholesterol. LDL-cholesteroland total/HDL-ratio were calculated. Fibrinogen,tissue-plasminogen-activator-inhibitorand factor II-VII-X were analysed as markersof trombosis.The diary consisted of a double page for eachof up to five days, consisting of one day, oneevening, two night shifts and a day off. Thediary included information on dietary distribution,self reported stress, physical activity,energy, sleep length and sleep quality, meansof commuting, and a selfassessment of sleepiness.After 6-8 months and after 12 months at theintervention wards, a smaller questionnaire,requsition for the same blood samples andthe same diary were distributed.Information in the diary is not used in thesepapers, as the participation rate was low.From two eight week periods, one just priorto the intervention, and one a year later, shiftschedules were collected from interventionand intervention control wards, and enteredinto the RRPA computer program 185 .In publication IV data on the interventioneffect from the first to the second datacollectionperiod was analysed. Two approacheswere employed. In order to analyse accordingto a “intention to treat”-principle (here:“intention to change”), participants weredivided into three groups according to wardand questionnaire data. Participants stating inthe second questionnaire that their scheduleMaterial and methodswas changed due to the project were dividedinto two groups, according to ward. The twowards that had introduced all four principlescomprised one group, and another was formedfrom the other two intervention wards,agreeing to 1-3 of the four principles. Participantsthat indicated they did not change shiftschedule constituted the reference groupregardless of ward, as did the day workingoutpatient clinics.The second approach was to analyse changesin the RRPA-scores regardless of whetherthe participants were intented to have theirschedule changed due to the intervention.In publication V, the baseline data was usedto examine relation between different shiftschedule characteristics and biomarkers in across sectional design. Different shift definitions,including or excluding night work,schedule characteristics as regularity, predictability,periodicity, number of night shifthours worked, part time employment, andnumber of night, and evening shifts in a rowwere examined.The RRPA scores for regularity, predictabilityand periodicity were divided into tertilesand compared with the lowest values asreference group.All scales in the questionnaire were subjectedto control of the psychometric properties,with calculation of Cronbachs alpha, itemscalecorrelation and factor analysis.Nominal scales were compared with the P 2 -test and Fishers exact test, ordinal scales andnon-parametric interval scales were comparedwith Mann-Whitney and Jonckheere-Terpstra tests, and paired data with theWilcoxon test. For interval scales followinga normal distribution linear regression analysiswas used, eventually after logaritmictransformation.26

Shift work and heart diseaseIn both publications, adjustment was donefor age and for lifestyle factors known to berelated to the biomarkers (smoking, alcoholand exercise). In publication V some of theregression models also included a step thatadjusted for the number of night shift hoursworked.Material and methodsAs the outcome in both publication IV and Vwere correlated biomarkers it was also plannedto use MANOVA analysis, but as wewere not able to harmonise variance andobtain a normal distribution, this is not presented.27

Shift work and heart diseaseSummary of resultsSummary of resultsPublication IIShift work was not associated to all causemortality or to incidence of IHD. For thewhole 22-year follow up, the age adjustedrelative risk was 1.0 (0.8-1.2) for IHD and1.1 (0.9-1.2) for all cause mortality. Controllingfor social class did not change theseestimates, and further controlling for sleeplength, tobacco, age, weight, height, andfitness were not changing the estimates.Men being shift workers in both 1971 and1985 had the same risk as ex-shift workers ina eight year follow up from the 1985-6 baseline.Both a screwed distribution of socialclass between shift and day workers and asocial class gradiant in the risk of IHD wereseen. Social class would, by this, act as apotential confounder. Due to the large groupof shift workers in social class III that had anintermediate risk, this heavily influenced theresults. When analysing only participants inthe outermost social classes, I, II and V, theunadjusted RR was 1.3, which fell to unitywhen adjusted for social class. Although adistribution of participants like this would beunlikely to be encountered, it suggests thatthe confounding effect of social class isrelevant and should be controlled.Publication IIIIn NWECS we examined the correlations ofshift work with other working environmentfactors that might act as risk factors of cardiovasculardisease. Compared to day, shiftwork was associated with shorter workingweek length, especially in females, exposureto noise (males), heat (males), passive smoking(mostly males), walking or standing postures(females), ergonomic exposures (females),monotony (males), high cognitive(but not quantitative) demands, low decisionauthority and skill discretion, and conflicts atwork. High quantitative demands were moreseldom seen among male three shift workersthan day workers.The only work environment factors not beingmore prevalent among shift workers in atleast one group were dust exposure andquantitative demands. The different types ofshift work were associated to different workenvironment factors.Especially evening and night shift work werecorrelated with lower social classes in men,and controlling for social class changed oddsratio estimates especially in males for cognitivedemands and conflicts at work.Publication IVPublication IV showed that an intervention inshift scheduling aiming at more regular andpredictive schedules with a reduction in nightshifts in a row and with a reduction of shifttypes lead to lower biomarkers for IHD, andto a lesser extend to lowering robust stressbiomarkers.The change in shift schedule was validatedwith the use of RRPA. It showed that shiftschedules in the intervention group at wardsintroducing all four principles changed nonsignificantlytowards higher regularity andhaving only two types of shift (paired analysis).In the control group schedules changedtowards less regular and predictable schedules,having also fewer Sundays off. Theirschedules were also changed towards fewerhours worked a week. When comparingintervention and control groups, the interventionled to more regular and predictablescores, and to a higher degree of having twoshift types. There were no statistically significantdifferences with regard to the number of28

Shift work and heart diseasenight shift in a row, or to circadian disruption.The changes in the group introducing 1-3 principles were smaller, and differenceswere not significant, neither in paired analysesnor when comparing with the controlgroup, except for a tendency of a higher LWscore and having more night work hours.The perception of flexibility and predictabilityin the questionnaire changed according tothe change in the intervention group.The intervention group introducing four principlesdid not change sleep quality, there wasa slightly better self reported health, but ingeneral it did not lead to fewer symptoms ofneither stress nor disease. The interventiongroup introducing 1-3 principles had lowerstress symptom scores and higher scores onmental health after the intervention.In the group introducing all four principles,both paired and age adjusted statistical significantcomparisons with the controlgroupshowed lower values for glycated hemoglobin,LDL-cholesterol and total/HDL-cholesterolratio, and higher values of HDL-cholesterol.In the second intervention group onlyfibrinogen was found statistically significantlower, but the changes were in the predicteddirection for the rest of the lipids. The controlgroup had raised total/HDL-cholesterolratio and lower fibrinogen (paired analysis)during the intervention period.Due to the problems of explaining the changein schedules in the control group, we alsoanalysed differences between schedules oneyear apart, whether being a result of theintervention or not. Higher regularity wereassociated with a statistical significant fall intriglycerides and total/HDL-cholesterol.LDL-cholesterol was lowered with higherpredictability. The reduction in number ofconsecutive night shifts in a row reducedtotal/HDL-cholesterol ratio.There was a highly significant trend betweenSummary of resultsthe number of ergonomic changes in a scheduleand change in HDL, LDL, andtotal/HDL-ratio. Glycated hemoglobin alsotended to be lower, all in parallel with the“intention to change” analyses conducted.Publication VPublication V shows that different shift workdefinitions (day versus non-day; night versusday, and non-night versus day) are related tobiomarkers in different ways. The non-nightversus day definition was associated to glycatedhemoglobin and nonsignificantly withLDL-cholesterol, while the definition includingnight work was both associated withtriglycerides and glycated hemoglobin. Especiallythe number of hours worked between10 am and 6 pm in night shift workers wassignificantly related to lower HDL-cholesterol,higher triglycerides and highertotal/HDL-cholesterol ratio, and to glycatedhemoglobin. The broad definition of shiftwork was associated to higher levels offactor II-VII-X, mostly so with the non-nightshift work.Part time employment was associated withhigher total cholesterol and LDL levelsamong shift workers. Regularity of the schedulewere (non-significantly) associated toglycated hemoglobin, to higher prolactin, andto lower total- and LDL-cholesterol levels.For predictability, the associations were withtotal cholesterol and HDL, with triglycerides,and with a non-significant trend for fibrinogenand t-PAI, all, expect for regularity andprolactin, in the predicted direction. Forperiodicity, there were significant differencesin triglycerides, HDL- and total/HDL-cholesterolratio, and a non-significant trend forfibrinogen. A high number of consecutivenight shifts in a row were associated to lowerHDL- and higher total/HDL cholesterolratio,and to trends for higher prolactin and29

Shift work and heart diseaselower factor II-VII-X. This was no longersignificant, when the number of night shifthours worked was entered into the models.The number of consecutive evening shiftswere associated with a trend for higherSummary of resultsimmunoglobulin A and a significant lowerfibrinogen, somewhat lower when the numberof evening shifts were entered into themodel. The differences were not changedmuch when lifestyle factors were included.30

Shift work and heart diseaseDiscussionDiscussionIn the five studies presented together withthis thesis it was shown, that:C The epidemiological literature suggests arelation between shift work and CVD (I),and that several pathways may be presentto explain this relationship (I). However,these findings are not present in somemethodological well conducted studies.One methodological problem encountered inthe literature is that confounding might bepresent, and it was shown that:C Social class differences are present betweenshift- and day workers (II and III), andC Work environment differs between shiftand day workers in a random sample ofthe population, with the least favourablefor the development of IHD belonging tothe shift workers (III).From these observations, the appropriatechoice of comparison group will be discussed.In another approach to explain the relationshipbetween shift work and IHD, it wasshown that:C Intervention in shift scheduling with theintroduction of more ergonomically correctschedules with regularity and predictabilityas the main changes lowers biomarkersfor IHD (IV), andC There seems to be different associationsbetween shift work in- and excludingnight work, the night work schedulesmostly associated with the lipids, and witha linear relation between the number ofnight hours worked and biomarkers (V).These results will be used to discuss bothmethodological problems relating to exposureassessment and information, and todiscuss whether shift work can be regardedas a causal risk factor for IHD.Methodological problemsThe heterogeneity of the epidemiologicalstudies has not been explained. We haveshown that at least two different groups ofstudies exists 115 , and that none of the studycharacteristics (publication year, area ofstudy, exposure characteristics, outcome(MI, IHD, CVD), type of study, follow upperiod, confounding control) explain theheterogeneity. This is somewhat surprising,1/3 of the literature is cross-sectional, andthe selection processes discussed next willmake it probable that the shift workers willbe a survivor population. However, this wasnot seen in the meta-analysis. Another surprisingfinding was that the studies using aggregatedexposure assessment had higher riskestimates than studies relying on individualassessment. Normally, aggregation would bebiasing the estimate toward unity.However, this analysis took only into accountthe information given in studies, andtwo major problems mostly unaccounted forin the literature emerge.Selection bias and confoundingThe traditional methodological problemdescribed in relation to shift work research1,8,11,202,203 is secondary selection bias. Alarge proportion of shift workers leave shiftwork within few years, Harrington 1 suggestedthat 10% enjoy shift work, 70% acceptit, and 20% leave shift work. In an olderstudy in one company 2 , 25% left shift workfor day work within the first 10 years, 42%of these due to medical reasons. These figuresfor secondary selection are probablylow, as people leaving for other factories or31

Shift work and heart diseasetrades would not have been counted. It alsopoints at the assumption that the possibilitiesfor finding other work will influence the rate.Some leave due to management changes, forinstance when a company changes workschedule from three to two shift 200 , anothergroup leaves due to changes in the socialvalue given to the shift work, and a part dueto health problems, for instance when experiencinggastrointestinal disorders. The latterexplanation could both relate to the shiftworker finding another job with day workinghours, or to a referral within the company toa day time job, eventually by way of a companyphysician. The prevalence of workersselected out of shift work for the latter reasonwould be suspected to be higher whenfocus was on the relationship, and whenfactory doctors could dismiss people fromshift- to day work. This means that the healthbased secondary selection will be larger whenthe physician know and trust that the selectionis appropriate and possible. For IHD thisknowledge would be at hand from the latterhalf of the 1980'th, after the study in Lancet 7and the review by Åkerstedt et al. 8 firstdescribing the risk. The fact that the selectionprocess is relevant has been shown in literaturefinding a higher risk of CVD for ex-shiftworkers, a standardised mortality rate (SMR)of 160 (122-210 (analysed by us (I)) inTaylor and Pococks study 4 and an incidenceof 6/100 in shift workers after changing today work compared to 2/100 in day and shiftworkers (no testing possible) 119 . In one studythe difference between health based secondaryselection and secondary selection due tomanagerial changes has been described.Thiis-Evensen found that ex-shift workers,that changed to day work due to healthproblems had higher absentism for CVD thanex-shift workers leaving shift work due tochange of job 204 .Explaining the health based secondary selectionin relation to IHD indicates that someDiscussionkind of warning signals make the shift workerapply for a day job, or that some commonfeature is leading to both IHD andproblems in adapting to shift work. One suchpossible common denominator could bestress (see page 16). The group continuingwith shift work has been labeled a survivorpopulation.An anonymous referee for publication IIsuggested what could be labeled tertiaryselection, meaning that people having IHDleaves shift work. The differentiation betweensecondary and tertiary selection in shiftwork is relevant in relation to the risk of exshiftworkers. If the shift worker has IHDwhen he or she leaves shift work, he or sheshould be regarded as belonging to the shiftwork exposed group, while the secondaryselection process would be in operationbefore clinical signs of IHD. The distinctionis not easy as for instance hypertension canbe the first sign of IHD, and misclassificationwould be easy. The tertiary selection processwill be a potential problem in case-referentand cross sectional studies only.Less focus has been on the primary selectioninto shift work 202,203 . People choose to applyfor a shift work job with some kind of anticipationof whether it will suit them or not 205 .This judgement can be based upon formerexperience or solely on self judgement of forinstance social possibilities or sleep habits. Ina small study 203 applicants for shift and daywork were compared before commercingtheir new job. Neither had formal shift workexperience, and besides being younger, theshift workers were more eveningness personalitytypes with less rigid sleep patterns.This suggests that self-judgement is relevant.There were no differences in relation tosocial factors, symptoms or biomarkers(blood pressure, cholesterol, triglycerides) atbaseline. This study was thus not supportingthe idea that shift workers are a priory healt-32

Shift work and heart diseasehier than day workers, or that they are morestress resistant 202 .The literature has related the feeling of havingchosen shift work to better adjustmentto shift work 176 , but it has not been shownwhether the feeling of self selection influencebiomarker levels or the risk of IHD.In terms of primary selection the applicationfor a shift working job can also be seen asthe only chance of getting a job. In someinstances day work is used as a reward forsome years of shift work (when only internalapplications are considered), forcing the newapplicants to a shift work job in order toobtain a day time job at a later time. In thesesituations the volontary application could bea choice between shift work and no work atall. The situation at the labour market mightthus influence both the primary and secondaryselection processes.Both in publication II, III and IV/V selectionmechanisms may have been present. In CMSthe population was middle-aged at entry (40-59 years of age), and they might thus havehad 20-40 years of work experience beforeinclusion. The shift workers in danger ofIHD could by this have left shift work. Thisis, however, not different from the majorityof the longitudinal epidemiologic studies withindividual exposure classification 4,116,118 usingcohorts with the same age structure, and ofwhich two studies found a relation. From anideal point of view, however, a cohort followingthe shift workers from their first entryshould be preferred. In publication III, thepopulation was a random cross sectionalsample, and among the elderly with possibilitiesof selection. The same is true for theNARFE project, where most of the dayworking controls had previously had shiftwork.In other words, it is not a random sample ofpeople that ends up doing shift work. ThisDiscussionhas potential impact on the first of the encounteredmethodological problems:Choice of reference groupsThe choice of the comparison group is crucialwhen the impact of shift work is assessed.Every comparison is done with reference toa nonexposed group. In principle the comparisongroup should be what the shift workerswould have been if they were not shiftworkers. And that is not necessarily being aday worker. The shift workers might differfrom day workers socially, economically, andin personality factors. If the comparisongroup differs from the exposed groups inother relevant ways than the working timearrangement, eventually as the result ofselection bias, potential confounding may bepresent. And as information on all possibleconfounders can’t be obtained the differenceswill eventually not be controlled, leavingresidual confounding.In almost all of the literature encounteredshift workers has been compared to dayworkers without questionning.Social classOne of the differences between shift and dayworkers are the social class they belong to.In publication III covering a random sampleof employees we found that the distributionof shift work was highly screwed acrosssocial classes (table 4). In social class I morethan 90% had day work, while this was onlythe case for 3/4 of social class V. Furthermore,the type of shift work differed, withnight and three shift work being most prevalentin the lower social classes, while socialclass I was nearly only related to irregularworking hours. In a random sample of thepopulation shift work would thus be con-33

Shift work and heart diseaseDiscussionSocial classTotal (n)I II III IV V UndereducationDay time 91 80 81 84 74 87 4791Irregular working hours 8.4 13 10 4.9 11 7.5 600Evening or two shift 0.6 3.9 5.0 3.7 8.5 3.1 283Night or three shift 0 3.0 3.4 4.2 5.8 2.0 212Total (n) 716 1048 2049 569 1249 255 5886Table 4. Working time arrangement by social class. Percent. Due to roundingcolumns do not sum to 100%founded by social class as day work is moreprevalent in higher social classes, and thetype of shift work is screwed with shift workincluding night work being more prevalent inthe lower social classes. Among females thedifferences were smaller (III)The association between social class, shiftwork and IHD can be further analysed inpublication II. The Copenhagen Male studyis not a random sample of the population, butrather consisted of men in larger companies.The effect is that shift work is most prevalentin social class III with nearly 50% shift working.Figure 3 depicts the 22 year incidenceof IHD. It shows that the incidence nearlydoubles from social class I and II to socialclass V. Within each social class the incidencebetween shift- and day workers are nearlythe same, a little lower for shift workers insocial class I - III and a little higher in socialFigure 3. Incidence of IHD over 22 yearsin day and shift workers by social class34class V. It shows that social class will act asa strong confounder if not controlled.Even if a single factory is studied day workersin the production will tend to be skilledworkers, while shift workers will more oftenbe unskilled. This would - in the CopenhagenMale Study - be a comparison between socialclass IV and V, the two right most groupsand would lead to the social class differencesin incidence still being larger than the differencesbetween shift and day workers. Inthe Knutsson et al study of Swedish papermillworkers 7 this was in fact the comparisonmade. Shift workers at the papermill wereunskilled production workers, and they werecompared to day working skilled maintenancepersonnel. It means at the same time thatcontrolling would be difficult if interactionwas taking place, when there are only fewskilled shift workers in this type of study.Of the remaining epidemiological studies (I)social class was controlled together withother potential confounders in the Kawachi 116study of female nurses (for spouse’ education),in the McNamee 120 study (jobstatus),the Knutsson case-referent study 117 (education),and in the Tenkanen study 118 controlfor jobstatus was done by stratification in away that enables a closer look at the effect.The relative risk for blue and white collarworkers joined (age adjusted) was 1.52 (95%CI 1.11-2.07) and for blue collar alone, (age

Shift work and heart diseaseadjusted), RR was 1.35 (0.94-1.93).The figures do not give the possibility toexamine the effect of shift work in whitecollar workers, and in general, the literatureis focussing on the lower social classes.The missing control for social class differenceswill bias risk estimates upwards.Work environment differencesAnother potential confounder is differencesin work environment between day and shiftworkers.In publication III all IHD risk factors analysedexcept quantitative demands, dust andlong working hours were more prevalent inthe shift than in the day working population.Furthermore, differences in the distributionof factors between irregular working hours,two shift/permanent evening and threeshift/permanent night shift were present.Even when controlling also for differences insocial class between day and shift workers,differences were still present between thegroups in some instances even with higherodds ratios than when unadjusted. Bothobservations are probably a result of samplingin different work situations (shift workbeing more prevalent in certain industrial andservice sectors (especially the hospital)), andit would be expected that if a shift and dayworking group were doing the same tasks atthe same work place, differences would bediminished.Comparing day and shift workers withoutproperly controlling for work environmentfactors would this way cover differences inother working environment factors, and shiftwork might be seen as a proxy measure forthese other relevant risk factors. The resultsagain suggests a bias upwards if not controlled.DiscussionAs shift work types differed in relation toprevalences of work environment factors,and shift work type was also screwed amongsocial classes, it would seem to be relevant tocontrol for both factors.The literature has in general not controlledfor work environment differences. It haspreviously been shown that shift workers aremore likely exposed to noise 206 and passivesmoking 207 . Only a few studies have controlledfor differences in relation to IHD, andmostly so in relation to job strain 118,208 , whileÅkerstedt et al. 209 controlled for (aggregated)information on several work environmentfactors.Other risk factorsIt has repeatedly been shown that day andshift workers differs in relation to personalityfactors, shift workers scoring higher onneurotism and hardiness scales, although thefindings are not consistent 163 . In the NARFEII project, constituting the base for publicationIV and V, internal comparisons betweenshift workers with fixed and rotating shiftsrevealed differences in relation to Locus ofControl and coping strategies 210 . The literaturedoes not contain studies that have examinedthe distribution between day and shiftworker of personality factors relating toIHD, for instance hostility 73 , and the personalityfactors shown to be related to shift workhave not been shown to be risk factors forIHD (e.g. sleeping patterns, neurotism etc).Confounding or mediatorOn the other hand, not all differences shouldbe controlled. One problem in the examinedliterature is the apperent confusion overmediating variables and confounders. Inmany of the reviewed articles (I) controllinghave been made of every difference found35

Shift work and heart diseasebetween the groups. The question has beenposed: Is shift workers “born or made” 11 (p.5), that is, were the differences present beforestarting on the job or are the differencesa result of shift work experience. In the firstinstance, controlling of the differences shouldbe made, in the second it may be more appropriateto see it as an effect of shift work.A confounder is defined as a factor thatmeets all of the following three conditions: 1)It is a risk factor for the outcome of interest(i.e. heart disease), 2) it is associated withthe exposure under study (shift work status)in the source population (the population atrisk from which the cases are derived), and3) it is not affected by the exposure or disease211 (p. 123-5). A mediating variable, on theother hand, is an intermediate step in thecausal path between exposure and diseaseFigure 4. The difference between a confounderand a mediating variable(figure 4).As an example, some studies have controlledfor differences in blood pressure betweenshift and day workers 116,118,120 . Blood pressureis a risk factor for IHD and is associatedwith shift work status in some studies, but itmight be affected by the exposure in this waybeing a potential link between shift work andIHD with blood pressure higher because ofshift work 133 , thus not fullfilling conditionnumber three. When controlling by way ofrestriction to normotensives in the studydesign or by adjusting in multivariate regressionanalyses the effect will normally be aartificially lower risk than is really present,Discussionexcept when confounders are nondifferentiallymisclassified. In that case bias can be ineither direction.In the study by Tenkanen et al. 118 , differencesin lipids were also controlled but as highercholesterol can likewise be regarded as aneffect of being a shift worker (cholesterolrelated to the amount of night work (V)),lipids should also be regarded as a mediatingvariable.Smoking and BMI have been controlled in allof the studies where it was measured5,7,116,118,120,207 and also in (II) but it is openfor debate, whether differences in smokingand BMI acts as confounders or whetherthey are a consequence of shift work. Forinstance smoking may be regarded as acoping strategy of stress related to shiftwork, and higher BMI could be due to achanged metabolism at night. Likewise,whether exercise should be controlled canalso be discussed.In these circumstances, it would seem prudentin general to give both estimates uncontrolledand controlled for these factors,which was done in publication IV and V.Of the discussed confounders related to IHD,the following can under no circumstances beregarded as potential mediators: age, gender,social class, and work environment differences,and they should always be controlled.Exposure assessmentThe other main problem in the shift workliterature is relating to the definition of shiftwork, as already pointed out by Åkerstedtand Knutsson 13 years ago 205 . In their contemporaryreview 8 shift work was defined as“work beyond the conventional daytime thirdof the 24-hour cycle” (p. 409), and theypointed at the permanency of work hours,completeness of 24-hour coverage and regu-36

Shift work and heart diseaselarity in alternations as possible explanationfor consideration. Not much work has beendone since then.Most of the epidemiological studies havebeen on rotating shift work. Some evenseems to have included fixed evening andnight shift in the control group 116,212,213 . Froma theoretical point of view there is, however,no indication that permanent night shiftshould not be related to IHD, and the studyby Tüchsen 177 also finds higher hospitalisationrates among bakers that at an aggregatedbasis have permanent night work. Inanother study 121 permanent third shift wasnot associated with (acute) risk of IHD. Thedemonstration in publication V of a relationbetween night shift work, the number ofnight hours in night shift workers and biomarkerseven suggests that permanent shiftworkers might be worse off than rotating.On the other hand, the two-shift and permanentevening shift has neither been investigatedin any details. Even so, it is encompassedin the commonly used ad hoc definitions inthe literature. In the study by Tüchsen aggregatedevening work was a risk factor. Thegroup covered taxidrivers, workers in hotelsand restaurants, cooks and waiters, all ofwhom may be night working as well. Inpublication V the different associationsbetween shift work with and without nightwork and biomarkers suggests that nightshift work could be related to a higher risk ofIHD. It still remains to be shown that permanentevening or two-shifts without nightwork is a risk factor for IHD, and secondarywhether the risk is of the same magnitude aswith night work. From a theoretical point ofview it would be surprising if it were; asstated previously, evening work is mostlyrelated to social problems while night workis related to physiological desynchroniztionof circadian rhythms.Detailed shift work exposureDiscussionIn general the characterization of an exposureshould encompass its nature, the amountand the time relationship 214 . Translated toshift work a thorough shift work exposureshould encompass at least the followingmodalities:C The type of shift work. Whether eveningor night work is included, whether theschedule is rotating or permanent, andwhether the schedule is covering thewhole week (continous) or only weekdays(discontinous). Even among shift workersthe spare time in weekends is valuedhigher than any other time of the week,and the inclusion of weekend work wouldnormally lead to negative social effects. Itcould then be anticipated that shift workincluding weekends would be more strainfullthan non-continous work. As with thequestions raised above on evening workand permanent night work this statementis not based on emperical ground.C The schedule. As presented previously, aclockwise rotation 18 , and the number ofconsecutive night shifts in a row are leadingto lower biomarkers 199 . One unrecognisedproblem is that if the ergonomiccriterias are of importance in diminishingthe strain of the shift schedule, then aschedule following the ergonomic criteriasshould lead to a lower risk of IHD than aschedule not made according to ergonomiccriterias, even when the amount ofe.g. night work is the same. This make itimportant to incorporate some kind offormal description and comparisons ofshift work schedules (see p. 19). Evenmore so, a schedule that changes in onedimension will often have changes inother modalities as well, and often probablyin a negative direction. As anexample, the change in schedule fromcounter-clockwise to clockwise 18 leads to37

Shift work and heart diseasea better circadian rhythm, but at the sametime diminishes the amount of socialdesirable weekend hours. One possiblesolution would obviously be to give theactual schedule with the start and stophours of the shifts, and preferable with anindication of weekends. This has in factbeen done by several authors (see table 5,p. 54), although they have not discussedit.C The amont of shift work per month oryear. If shift work is a risk factor then the“concentration” would probably be relevant.The figure should probably be countingthe number of different shift typesworked (number of evening shifts, nightshifts, weekends). Also this modality hasonly modest emperic data, but publicationV showing a relation between the numberof night shift hours and HDL-cholesterol,total/HDL-cholesterol ratio, and triglyceridespoints out the relevance of at leastthe amount of night work in the schedule.[11]This would also be helped by theinclusion of the worked schedule, as thereader would be able to make the figuresout for him or herself.C The number of years in shift work. Thereis some support to a dose-response in thesense of years on shift work. The study onSwedish papermill workers 7 showed analmost linear relation between years inshift work and relative risk up until thegroup with more than 21 years of exposure.Also the studies by Tenkanen 118 andKawachi 116 report dose-response betweenyear in shift work and an increasing riskof IHD, while the essentially negative[11] Also the study by van Amelsvoort et al., showing(see note 8) that having more than five nightshifts a month was related to more ventricularextrasystoles points at the importance of night work(appended july 2000).Discussionstudy by McNamee 120 did not detect anydose-response.One would then have to summarize thisinformation. One possible exposure variableof interest would be an “integrated lifetimeexposure”, that is the sum of the products ofthe number of for instance night shift hoursper month and the number of year worked onthis particular schedule. The approach hasindirectly been used in the Knutsson 1986study, where people at the factory wereallocated by the number of years as shiftworkers. It works when the shift schedulehas been the same in all of the years, this waymaking the first three bullets above the samefor the whole period. In fact two differentschedules were used 215 , one up until 1975and then a slightly different schedule up til1983 when the study was concluded (seetable 5, p. 54). The “older” shift workerswho also would have the longest exposurewould have had higher chance of working inboth, and as the old one was worse in anergonomically context, they would in theorybe more exposed (bullet two above).In studies, where different populations areincluded as in the other two studies mentioned116,118 (and as also would be the case inpopulation based studies) the first threebullets is not the same for all of the shiftworkers and this makes it problematic tocompare “years of shift work” between themembers of the cohorts. The “integratedlifetime exposure” would have to be able totake into account that shift schedules differas stated in bullet two and three above.In population based studies with questionnairedata the exposed groups classified asshift workers might be covering eveningworkers, night shift workers with only acouple of nights a month etc, up to permanentnight shift work in continous operations.In general this misclassification of exposurewould tend to diminish any real association38

Shift work and heart diseasebetween shift work and outcome, and itsuggest that population based studies wouldshow lower estimates. This is apparently notthe case 115 .It seems that exposure should be delimited toabove a certain level, in order to have enoughcontrasts between shift and day work.Another problem is that many studies haveused only information on exposure at onepoint in time (prevalent) and the respondentmight this way have very little exposure.Another problem with the “prevalent” shiftexposure is that it would be possible that athreshold effect was present, meaning that acertain amount of shift work (e.g. hours ofnight work per month or years of shift work)should be present, before the risk of IHDwas raised. It would not be reasonable tobelieve that few months of shift work in alifetime would raise the risk. Other studieshave demanded at least 10 years of shiftwork 4 in order to be classified as shift worker.This would on the other hand open forthe possibility that only a well-adapted andmaybe thus less affected survivor populationwould be selected as exposed giving a lowerrisk of IHD.It would be relevant to include more informationon past shift work experience in thepopulation studies, and better questionnairesshould be developed.One worry is whether the respondent understandwhat is meant by shift work. In a fewstudies some kind of validation of the informationof shiftwork 16,120 has been undertaken,but in most circumstances the exposurehas been taken at face value. In anunpublished observation we studied theanswers in the SHEEP-VIP[12] databasewhere two questions were used to constructinformation on exposure 117 . One question[12] Stockholm Heart Epidemiology ProgrammeandVästernorrland Infarction ProjectDiscussionwas whether the respondent had shift work(“skiftarbete”), and if so whether it was twoshift,three-shift with continous running,scheduled work or other types. Anotherquestion asked whether the majority ofworking hours had been between 6 am-6 pm,between 6 pm-10 pm or between 10 pm and6 am (or combinations). Inconsistencies werepresent, people stating that they did not haveshift work at the same time also indicated atype of shift work, or indicated that theirworking hours was mostly including the 6pm - 6 am period. In numbers (excludingthose with a type of schedule not assessable),115 of 4106 (2.8%) stating that they did nothave shift work had inconsistent answers. Asthe same type of misclassification among dayworkers would not be possible, differentialmisclassification would be the result. Thiscould be the effect of (at least Danish orSwedish) respondents that percieve the word“skiftarbejde” as belonging only to industry.Four studies used aggregated exposureassessment 5,6,177,209 . Instead of applying theexposure on every participant in the study,information on the prevalence of shift workin different occupations from smaller samplesare utilised to give every participant in aspecific occupational group the same exposureinformation. So if for instance employmentas a bus driver is known for 75% of busdrivers to include rooster work then everybus driver in the population is viewed asexposed to shift work, including the 25%,that did not have shift work. Due to themisclassification this type of study wouldnormally lead to lower risk estimates. This isnot so in general, the studies seem to givehigher risk estimates than studies usingindividual exposure assessment (I) 115 . This iseven more surprising as the actual cut offpoint in two of the studies 177,209 was as low as20% exposure, with a potential misclassificationof up to 80% of an occupation.39

Shift work and heart diseaseBut even in the population based studies withan individual exposure classification theactual schedules are not reported. So forinstance in the American Nurses study 116where shift work was defined as havingworked for more than 1 year with more thanthree nights a month (and excluding permanentnights), this covers schedules with awide variety of shifts. Having publication Vin mind, the actual exposure might be verydiluted.The problem of misclassification of exposureis also related to the past exposure of theshift workers. In the McNamee study 120 alarge part of the shift workers had been dayworking at some time before inclusion asshift worker. The median exposure of shiftwork was 10.7 years among the shift workinggroup that had median 15.5 years ofemployement at the factory. So it would berelevant to count the total actual exposure.Another problem in exposure assessmentrelates to the comparison group. Traditionallythey have been day workers, but somestudies included permanent night (see above)and one study included two shift workers 3 inthe “non exposed” group. Furthermore, inmany studies the day workers could in facthave been having shift work before the daywork (ex-shift worker), especially if onlyprevalent information is used. This is inparticular the case for population basedstudies, and would dilute an effect as controlsare misclassified as non exposed[13].The severity of the problem is relating to ahitherto unaccounted characteristics of therelationship between shift work and IHD; Ifshift work is a risk factor, would exposed[13] Furthermore, ex-shift workers have been shownto have a higher risk than both day and shift workers.This could lead to substantial bias toward lowerestimates, if the secondary selection was large(appended july 2000)Discussionpeople then have the same risk after the endof exposure or would the risk change? (as itis the case of cigarette smoking where thehigher risk for IHD disappears within twothreeyears after cessation 76 ), so that it diminishesfor instance over a couple of yearsafter the cessation of shift exposure. In thecase of a diminishing risk with time theproblem of misclassification of exposure innon exposed would be largest when shiftworkers had just left, and be a less problemif ex-shift work was several years back. In anunpublished examination from the SHEEP-VIP case-control study we actually foundthat the risk of IHD related to shift workdiminished within 3-5 years after shift workwas left.The information on exposure assessment inthe studies from publication I is collected intabel 5 at the end of the chapter. Severalstudies do not even give a definition of whatis meant by shift work. Of the rest, approximately½ give some details of the shift schedulein some instances with the possibilitiesto calculate the amount of for instance nightshifts. Within these, most studies are threeshift, rotating, and most of these with fewnight shifts in a row (see also p. 43).In the Copenhagen Male Study (II) informationon shift work were collected throughself reported questionnaires at two occasions,in 1971-72 and in 1985-86. The informationincluded one question, asking whetherthe participant had irregular hours, fixedevening or night work, or whether theyworked day time only. The information wasconfirmed during a later interview. Thismeans that the exposure was “prevalent” andit did not include neither type of schedule,amount of shift work, schedulecharacteristics nor life time exposure. Throughthe second baseline using the samequestionnaire 14 years later some indicationof the stability of exposure was obtained,40

Shift work and heart diseaseDiscussionsuggesting that those stating to have shiftwork at both occations would be exposed.For social class III with approximately 50%shift workers we compared the self reportedoccupations from day and shift workers, andfound that shift work was inherent as a naturalpart of the reported occupations fornearly 90% of the shift workers while thesame occupational groups had only indicatedday work for 4% of the sample, thus partlyvalidating the information. But despite ofthis, we had no detailed information onexposure and especially not on the“concentration” or “dose” of the exposure.The same is true for the NWECS (III). Shiftwork was defined as having answered yes toone of the possible answers: two shift, threeshift with running shifts, three shifts withchanging shifts, irregular placement in day orweek after schedule, fixed evening, fixednight, early morning, others. This was thencombined into night work including fixednight and three-shift work, evening includingfixed evening and two-shift work and otherirregular working hours. This way the informationis prevalent and covers a wide varietyof work schedules. Two-shift work can alsobe covering two times 12-hour shift.41

Shift work and heart diseaseIn population V we examined different typesof shift definition on the association withbiomarkers. Differences occurred betweenshift work including and excluding nightwork, the latter associated to biomarkers forIHD while the exposure to evening shiftwork was associated to glycated hemoglobin,a stress marker, but only marginally to lipids.This suggests that the night component ofshift work might have potential impact on themagnitude of the risk.Measuring ergonomic criteriasAs stated above the ergonomic criteriasshould be accounted for in the shift workexposure assessment in order to be able totake into account that different shift workschedules might confer different risks.DiscussionIn publication IV and V the problem ofexposure classification was elaborated. As aconsequence of the very individual schedulesused at the hospital wards a way to compareboth the new and old schedule for the individualand to compare schedules betweensubjects was needed. A further problem wasthe fact that the ergonomic principles encompassvery different features of a schedule,and it is problematic to compare for instancea schedule with only three consecutive nightsin a row but a backward rotation with aforeward rotating schedule with four nights.The RRPA was used for this purpose. Theprogram evaluates criteria that are not entirelythe same as the 14 criterias from theBEST-group (see p. 19). Although the numberof shifts in a row is included in theperiodicity-score, and the overall number ofnight shifts mirrors in the opportunity to restat night (ON)-score, we further counted themaximal number of night shifts in a row andthe number of night shifts in the schedule.The same was done with evening shifts.Among the changes made, the reduction ofnumber of different shifts is not identifieddirectly in the RRPA, and we also countedthe number of different shifts in the schedule.m t w t f s s m t w t f s s m t w t f s s m t w t f s sD d d d d d d d d d d d d d d d d d d d dN n n n n n n n n n n n n n nM n n n n n n n n n n n n n na a a d d d d d x n n n n n n d dd d a n n n d d d a a d d d n nb y n n n d d d d d d d d d d d d n nn d d d d n n n n d d d d d d d dTable 6. Five schedules with different characteristics (see text). d=day, a=afternoon, n=night,x=day shift followed by night shift, y=short day shift (until noon) followed by night shift42

Shift work and heart diseaseDiscussionscoreday shift(D)7+7 night shift(N)5+2 night shift(M)rotating shift: preintervention (a)rotating shift:post intervention(b)RE 10.0 10.0 10.0 7.9 7.2PE 10.0 1.4 3.7 6.6 6.9LS 8.3 8.0 8.0 7.9 8.2LW 8.3 8.5 8.8 8.6 8.5ON 10.0 4.9 5.1 7.8 8.2PR 8.3 7.5 8.3 4.2 5.1OH 4.5 10.0 10.0 7.4 7.0OE 10.0 10.0 10.0 8.7 10.0OW 10.0 5.5 5.5 5.5 4.7Table 7: RRPA scores for the schedules shown in table 6For comparison and in order to establish acouple of waypoints the RRPA scores forshift schedules from the wards are presented(table 6 and 7). The first schedule is forcomparison, a discontinous day shift (normalday shift, Monday-Friday 8.00-15.24, markedD in the table), followed by a permanent7/7 night shift schedule with 7 consecutivenight shifts and 7 days off (28 hours a week,196 night hours 22-06 in 8 weeks, markedN). The ergonomic better 5/2 night shift withthe same number of hours is also presented,marked M.Most of the schedules in the NARFE projectwere irregular, one of these schedules selectedat random from the intervention ward Bwhere the largest change was found, is presented((a): pre intervention, March andApril 1997, and (b): post intervention, Marchand April 1998).The schedules ran for four weeks, but eightweeks was used in calculations to diminishthe impact of holiday and vacations (and oneward went from 4 to 12 weeks of scheduling).For comparison (the constancy scoreand the PR (predictability score) depends onthe number of weeks) the first three shiftsschedules are also defined as 8 week schedules(althought they would normally be runningin a 4 week schedule, and in reality are1 and 2 week schedules repeating themselves).It can be seen from the RRPA-figures (table7) that the day-only shift is better than all ofthe other shifts presented, that the 7+7 nightespecially lowers the physical figures (PE andON), and raises OH, that measures the a-mount of time off in weekdays. The proposedchange of the permanent night shift witha split of the nights in 5+2 consecutive nightsgives the predicted change with better valuesin the physical (PE, LW and ON (due to theaccumulation of fatigue)), but also a slightlybetter PR-score, which is brought about by achange in the constancy score rather than thebasic score. This is due to the 5/2 systemgiving days off every week which the programevaluates as an advantage in agreementwith the literature 180 .43

Shift work and heart diseaseThe rotating shift (a) is less regular (RE) andpredictable (PR) than both the permanentday and night shift. The ON and PE scoresare in between the day and night shift, as thenumber of night shifts is lower.The change (b) brings lower scores of regularitybut higher on predictability and a higherON and PE score due to having morenight shifts in 8 weeks. The change in predictabilityis probably due to the change from 3to 2 shift types. The OW falls due to fewerSundays off (5 in a, 4 in b), and the OE is 10because there is no evening shifts in (b).The RRPA scores from the interventionexample shows the problem in comparingscores: The changes that included a reductionin the number of types of shifts fromthree to two at the same time gave morenight shifts in the eight week period, whichaffects the PE and ON score negatively. Soa positive change in one score is followed bya negative change in another.The different criterias are not independentand a factor analysis with varimax rotation ofthe schedules at baseline in publication Vshowed that the combined RRPA-scores andthe extra figures taken from the scheduleloads on four different partly independentfactors measuring:! nights (PE-, ON-, OE-score, maximumand total number of night shifts, maximumand total number of evening shifts),! number of non-day time hours (OH-,OW-scores, number of shifts, number ofsundays),! a regularity score (RE-, LS-score, PRscore,number of shifts) and! a working week length score (LW, numberof offences against the 11 hour ruleand number of times with less than 36hours between blocks of shifts).DiscussionAs the scores were highly correlated, especiallyPR and RE, it was necessary to chooseamong the scores.The RRPA was used in two manners. Inpublication IV it was used to validate whetherchanges were in fact made at the wardsagreing upon this. It showed that regularityrose in the intervention group introducingfour principles, but that periodicity as ameasure of circadian disruption did notchange. This is in accordance with the factthat the number of night shifts in a row wasnot changed either.The RRPA was also used to measure changesin scores by dividing the change intobetter score, no change or worse score regardlessof whether the change was a resultof the intervention or not.In publication V we used RRPA to characterizethe shift schedule in order to measurefor instance the regularity or periodicity ofshifts. In this context one problem is that it ishard to translate a RRPA figure into a meaningfullconstruct and that the scores are notnecessarily measures on a ratio-interval scale,so we chosed to use them as rank scales andused the scores divided into tertiles in alinear regression model.In order to demonstrate the possible positiveeffects on shift work exposure assessmentusing the RRPA-approach, I have used theRRPA on the publications on shift work,biomarker measurements, and IHD in whichthe authors gave information on the actualshift schedule, that is the sequence of shifts,the placement of weekends (or if a schedulewas completely described by the sequenceand an indication of the schedule length), andthe start and stop hours of the shifts. Thisinformation is placed in table 5 (at the end ofthe chapter (p. 54)), while the RRPA-figuresis found in table 9, also at the end of thechapter.44

Shift work and heart diseaseThis approach to evaluate exposure assessmentin the literature reveals two features. Inthe intervention studies where two schedulesare being compared, there are changes inmore than one RRPA score and often inopposite directions. In the clockwisecounter-clockwiserotation experiment 18 thebetter periodicity is accompanied by muchlower opportunity for weekend recreation (aswas also the reason for the police officers toprefer the counter-clockwise rotation). In thestudy dividing the night shifts in a row 199 thenegative effect of better periodicity andpredictability is a lower score on regularity.Secondly the schedules have very largedifferences in their RRPA scoring, with forinstance regularity ranging from 6 to 8.5 suggestingthat the shift exposure in the literatureis very different. Again, if ergonomiccriterias are related to the risk of shift workthis would be possible explanations for thedifferent results. If for instance regularity isrelated to biomarkers (as suggested by publicationIV and V) then when using very differentschedules the exposure might be verydifferent and it would argue that studiesshould give more information on the schedules.For instance Angersbach 119 and Knutsson 208used shift workers from several differentschedules. They internally differed withregard to RRPA scores and in relation topublication V might have very differentexposures. It would have been interesting toknow whether the inclusion of schedulecharacteristics led to differences in risk.Compared to the baseline RRPA scores inthe NARFE intervention (IV,V) the literaturehas ranges on both regularity, predictabilityand periodicity, that covers the median baselinefigures and are thus comparable. Theload in week figure is higher than in theliterature which mirrors that part timeemployment was a feature of the hospitalDiscussionwards. The “opportunity for rest at night”and “evening recreation” figures were alsohigher in the NARFE project suggesting thatthe number of night shifts was lower than inindustry, and thus in the majority of theliterature. Finally, most literature studies hadlower figures on “opportunities for weekendrecreation”. One characteristic of the schedulesin Danish hospitals are that weekendshave high priority and most often with bothsaturday and sunday off which is possiblebecause staffing are lower at weekends. Thiswill on the other hand tend to give moreirregular schedules.To summarize, a valid exposure measurementshould include information on type ofshift work, schedule characteristics (throughthe inclusion of the schedule in the materialsection of the paper or for instance with theRRPA scores), “concentration” of relevantfeatures, and the “dose” in terms of forinstance years. Whether it would be possibleto concentrate all of this information in onefigure is doubtfull but will rely on futureinformation on the relative importance ofdifferent features of the schedule.One possible place to start would be toinclude hours worked at night in a referenceperiod. In relation to the discussion on thereference group an alternative would be toselect different shift schedule populationswith homogene RRPA measures and comparethem for the importance of differentcomponents. This would also mean thatsmaller differences in relation to possibleconfounding between shift work and referencegroup would be expected.Measurement of biomarkersAnother problem relating to information iscircadian rhytmicity of biomarkers. It hasbeen shown that also lipids have a circadianrhytmicity (see p. 15). It is well known that45

Shift work and heart diseasethe circadian change due to night work isalways a change in acrophase and a somewhatlower amplitude 129 . This means that ameasured difference between day and shiftworkers or in different shifts could be theresult of measuring at different places on therhythm (figure 2).Cesana 1990 217 ?Cesana 1985 157Costa 1990 218De Backer 1987 219Kecklund 1994 199In the morningDiscussionMorning after a night off-dutyIn morning shiftIn the morning, after fastingKnutsson 1989 220Knutsson 1988 208Knutsson 1990 147Morning after a nights sleep,fastningMorning after a nights sleep,fastningMorning after a nights sleep,fastningLasfargues1996 221 After fasting overnight. Indiscussion night shiftworkers may not be fasting -> after nightshift?Figure 5. Circadian rhythm for triglycerides.Per cent deviation fromarithmic mean. From Rivera-Coll etal., 1994Whether the acrophase would be advancedor delayed in the night worker is a matter ofthe timing of sleep, but in a circadian rhythmwith the highest value early in the night anda steep decline in concentration during themorning hours (as triglyceride) 140 the resultof a phase delay of just 4 hours would lead tothe measurement of maybe 25% highervalues at 8 am (see figure 5).Table 8. Sampling of blood in studies on shiftwork and biomarkersFirst author, yearBursey 1990 216Bøggild 1999 (II)Blood sampling characteristicsgiven in the publicationDuring mornings, for shiftand day workers alikeIn the morning, after fastingLennernäs1994 146Nakamura 1997 86Netterstrøm ?1996 222In a day shift (~ fromschedule it means > 72hour)On morning shift, after onenight fastingOrth-Gomér1983 18 After a nights sleepPeternel 1990 223Romon 1992 224Thelle 1976 225 ?Theorell 1976 226Last day on morning shift,last day on night shiftBetween 7 and 8, morningshift for shift workers, 12hour fastingDay time, 1. and 3 week(from figure)The other lipids have the lowest values in themorning, and a much lower variation, so theywould probably be less affected.The literature in general does not providedetailed information on the timing of bloodsampling in relation to working hours,i46

Shift work and heart diseasenformation in studies from publication I isgathered in table 8.It seems as if most studies demanded onenight off which means the possibility ofhaving only 24 hour between the end of thelast night shift and blood sampling.Lennernäs 146 had at least 72 hours betweenthe last of five night shifts and blood sampling.On the other hand, Lasfargues 221 whileclaiming a overnight fast in the discussionindicates that the night shift workers mighthave been eating (and thus awake) and hisfinding of a higher triglyceride level of shiftworkers could in theory be a result of this. Ina recent study 161 sampling was done in themorning, after either day work, a day off ornight shift, thus also having the possibility ofmeasuring the desynchronized circadianrhythm.In normal persons the biological half-life ofLDL-cholesterol is 3-5 days and for HDLcholesterol5-7 days 74 . Fibrinogen has abiological half-life of 3-4 days 227 . It hastherefore been suggested 74 that the acutechanges (within hours) in relation to acutestressors are probably reflecting changes inplasma volume that is also subject to circadianrhytmicity 131 more than the amount oflipoprotein.In publication IV and V we used an approachof having at least three days after a night shiftbefore measuring. Although it is not knownhow long the circadian rhythms of lipids needto return to day orientation, most measuredrhythms have a quick return 127 . The alternativeapproach would be to measure severaltimes at the same day in order to establishthe circadian rhythm before finding the relevantvalue. This would also allow for studyingthe acute effect of shift work on lipids.Some 133-136 but not all 137-139 studies with 24hour measurements of blood pressure havesuggested that there are differences in bloodDiscussionpressure between day and night shifts, onestudy suggesting a longer high plateau ofsystolic blood pressure when working in thenight compared to day work (20 hours versus14) although the mean pressure did notchange 133 . By analogy, this could in theorymean that the lipid influx into the wessel wallwas longer lasting and could change theequilibrium of atherosclerosis 76 .Different biomarkers may have different ratesat which they respond, remain elevated andreturn to normal 228 . This has not recievedmuch attention but it would not be surprisingif metabolism in some instances were capableof down regulating the response, so forinstance a stress biomarker was lowered 75 .This might especially be a problem in thefibrinolytic/hemostatic system with its numerousfeed-back systems 83 and could be anexplanation on the essential negative resultsin publication IV and V.For the stress biomarkers we chosed glycatedhemoglobin, immunoglobulin A and prolactin.For IgA most measures have been donein saliva and it might have been more resonableto use IgG 229 or IgM as markers of thesystem, although the use of these as biomarkersof stress have had less attention in theliterature 110 . In theory especially IgM wouldbe more robust than IgA. The use of prolactinas a marker was troubled by very largevariations and it did not behave as expected.We did not include the traditional stressbiomarkers (epinephrine, cortisol). This waschosen as the use of these is in general notwell suited for field studies due to the variability,requiring several measurements 103 , andit will as a measure too close to the beginningof the cascade give the possiblity of afeedback tending to maintain values within anarrow physiological range 104 .In relation to IHD the use of lipids and lipoproteinsas biomarkers are well validated 51 .The choice of factor II-VII-X and t-PAI as47

Shift work and heart diseasemarkers of the hemostatic system are debatable,but were chosen on the basis of bothbeing shown to be risk factors for diseaseand to the actual possibilities at the localDepartment of Biochemistry (besides PAI allanalyses were part of the routine possibilities).This was done in order to obtain themost unproblematic analyses as laboratoryvariations should be the smallest possible.The biomarkers are used as surrogate endpoints for actual disease and this pose at leasttwo problems. The use is only valid if thebiomarkers are actually causally related tothe endpoint 230 (IHD and stress). The otherpoint is related to the discussion above; ifsome feed back in the system maintain equilibriumthe use of biomarkers are not a validsurrogate of the diseases, unless the risk ofdisease also change when the biomarker isheld at normal concentrations. On the otherhand, the alternative would be large scaletrials running for years and with numerousproblems in relation to statistical power anddistant outcomes 230 besides the problemsrelating to selection and changes in exposureover the years.Information from questionnairesMost information in both the CMS, NWECSand NARFE II projects was obtained fromquestionnaires. This poses potential problemsespecially when trying to measure constructsas disease, stress, sleep quality, and socialconflicts. The building of scales can easily bemisleading 231 and scales should be subjectedto psychometric validation. We chosed mostof the scales for the questionnaires in NAR-FE II from literature governed by theoreticalconsiderations. As we conducted an intervention,the scales were chosen so that thefloor and ceeling properties made it likelythat changes would be possible to detect(thus for instance leaving out most of theDiscussionShort Form-36). Most of the scales camefrom the Standard Shiftwork Index 201 whichis a set of questionnaires selected for shiftwork research, but as especially the stressoutcomes are not part of the SSI others wereselected for specific purposes. Where possiblewe obtained the Danish versions, butfor the SSI and others we translated themfrom English to Danish and back and discussedthe translations. In some instances wewere not able to find existing scales (forinstance on adaptation to shift work) and inthese instances we made up our own scales.This was done by first selecting the field forwhich the scale should cover, then we wereindependently setting up questions and discussingthe proposals, before agreeing on theselection of questions for a scale. The resultingquestionnaire was tested in a small pilotstudy in a telephone company.The scales were subjected to psychometrictesting in the first questionnaire (just beforethe intervention). For all scales the Cronbachalpha, item-scale correlations, and factoranalyses were conducted. For scales takenfrom the literature the factor analysis wasused to test whether we found the sameconstructs as was given with the scale, forour own scales it was used to describe theinternal properties of the scale. For most ofthe scales from the literature there wereacceptable properties with Cronbach alpharanging from 0.5 to 0.9 and in most instanceswith acceptable item-scale correlations. Ascale on dietary habits was not working as ascale. For our own scales a couple had verylow Cronbach alpha’s and were obviouslynot working as scales.But although the internal properties of thescales were in most instances reliable andvalid, the external validity of the scales werenot tested as this would demand a goldenstandard. For both symptom scores andstress this might be a problem.48

Shift work and heart diseaseFor stress the measurement is of subjectivefeelings, a “snapshot” 232 of a transactionbetween environment and the individual. Asstress in the Lazarus’ mode include an appraisalof threat, challange or harm, it seemshard to obtain valid information through aquestionnaire. Instead we used Setterlindsstress symptom questionnaires 233 whichwould be a measure of the results of a stressprocess. In general no changes were foundon symptom scores in relation to the intervention,and one might wonder whether thisresult really is valid; The changes were regardedas positive from the large majority ofparticipants, and it might be possible thatthey would be experiencing less stress. Itcould simply be that the questionnaire wasnot capable of detecting smaller changes instress perception not leading to lower resultingsymptoms.Dietary questions were included in the diary,but the response rate was low. We insteadused a single question from the questionnaireas a marker of the use of fat, but this onlygave information on baseline dietary habitsand was not related to changes in biomarkers.The literature suggested that the distributionof carbohydrates over the night shiftis related to cholesterol 146 .Intervention and biomarkersThe intervention in publication IV led tochanges in biomarkers. We were able tocontrol for some but not all lifestyle factors.The information used to adjust were frombaseline data, and it might be postulated thatthe effect of the intervention was workingthrough changes in these factors. Formerstudies 147 have not found lifestyle habits tochange at such speed[14] and in our studyDiscussionthe use of tobacco, alcohol and exercise didnot change during the period either (IV).Goldenhar and Schulte 186 reviewed interventionresearch in occupational health andsafety between 1988-93, and among 36studies they concluded that the magnitude ofstudies lacked theory, were small and did nothave the intensity in intervention to suspecta change to occur. In our intervention studywe had a theoretical background governingthe suggested changes in shift scheduling.The changes were seen in the biomarkerswith the smallest variance, and the reason fornot finding changes in some of the stressbiomarkers or in the markers for hemostasismight be that the study was too small.Whether the intervention was large enough,is not easily answered. It is clear from theattempts to validate the changes obtained,that no change was seen in the periodicityscore and in the number of consecutive nightshifts. It turned out when the schedules werecollected just prior to the intervention, thatmany already had 3-4 night shifts in a row atboth the intervention and intervention controlwards and some also had two shift work.There were changes in the regularity scoreand in the number of participants with achange from three to two types of shifts, butthe changes were small.It is a paradox that the control group had thelargest negative change in schedule characteristics(in the paired analyses). It could bea result of the changes putting restrains onthe possibilities for scheduling also for thecontrols at the intervention wards, but thechange was seen in both the internal (at theintervention wards) and in the external controls(at the intervention control wards).Another more possible explanation was thatthe frames for scheduling were changed[14] Although van Amelsvoort (see note 7) foundsmoking shift workers to smoke more over a year (appended july 2000)49

Shift work and heart diseasethrough new agreements on overtime work.Eventually some of the statistical differencesare the result of a larger control group.We chosed to analyse the data in a modified“intention to treat” way dividing participantsin groups after their potential change, knowingthat some of them did not have theirschedules changed. This was the result of themain interest of the study, to examine whetherthe introduction of principles could led tochanges. The reason for this was that thewards had planned to schedule after theprinciples, but were not necessarily able todo so (especially on two of the wards) due tostaff shortage et cetera. This approach willineventibly lead to lower power of the study.Evidence of a relationsshipIn determining whether shift work raises therisk of IHD there must be a plausible causalconnection between the exposure and therisk. In publication I the model describedprior by Knutsson 16 was elaborated to incorporatechanges in biochemical risk factors(biomarkers) (see figure 1) and the literaturewas reviewed in line with the model. Thegeneral conclusion of the review was thatprevious research has focused on life stylechanges and changes in markers of atherosclerosis,while other possible pathways beingmismatch of circadian rhythms and disturbedsociotemporal patterns have only indirectlybeen described.The results in the literature were not consistent.In six of 15 studies shift workers smokedmore, in one study less than day workers.In two of 10 studies shift workersdrank more alcohol while one found a lowerprevalence. In one of five study shift workerswere exercising more and in one less. Weightor BMI were higher in two of eight studies.Diet has been studied in six studies and ingeneral there have been found no differencesDiscussionwith regard to intake of lipids while shiftworkers seemed to snack more, mostly sowith higher carbohydrate intake. Theselifestyle changes seems to be pointing towardsmoking and dietary habits as important inexplaining the relationship between shiftwork and IHD. It also points to the importantbut neglected influence of culture inlifestyle habits; It might well be, that shiftworkers in some social classes or countriesadapt to shift work by smoking while they donot in others.The relatively large amount of literature onlifestyle does not mean that the other possiblepathways could not be potentially asimportant. With regard to the mismatch ofcircadian rhytmicity MI and AP have beenshown to have higher incidence in the morningand a peak in the evening 234 has beenrelated to awakening after day sleep. In thecausal chain the metabolic relationship betweenpatterns of nutrition and the circadianrhytmicity of biomarkers could be of importance.The sleep changes related to shiftworking have been investigated 33 and sleepapnoe syndrome has been related toIHD 235,236 but it has not been shown whethersleep quality or sleep length are related toIHD, and whether sleep apnoe are moreprevalent among shift workers. There aregaps in the knowledge at these points.The finding in publication IV that anintervention in shift scheduling aiming at lessstrainfull shift work leads to lowerbiomarkers with the changes being highest inthe group changing the most, and inpublication V that among a relativelyhomogenous group of night working nursesand nurses aides the number of hours workedbetween 10 pm and 6 am were correlated toconcentrations of HDL-cholesterol andtriglycerides suggests that the causal effect isreal. It is in line with the study on clockwiseand counter-clockwise rotation 18 and on50

Shift work and heart diseasedividing the night shifts in 3+4 199 both leadingto changes in risk factors for IHD, andwith a study from Japan 200 showing that theabandoning of night shift in a car factory ledto higher HDL and lower triglycerides.What was linking the changes are not clear.It could be a result of shift work leading tocircadian disruption, to stress, or to dietaryhabits.Shift work and stressIn the literature stress has been described inshift workers as an important pathway toIHD, and it has been suggested that newermodels of shift work interference with healthhas borrowed from the general literature onstress theory 123 .We did measure stress biomarkers and stresssymptoms (publication IV). Stress symptomscores only changed in the interventiongroup introducing 1-3 principles. In generalstress biomarkers did not change, except fora small but significant decrease in glycatedhemoglobin in relation to the intervention(IV). Glycated hemoglobin was also higheramong shift workers at baseline and associatedto the hours worked at night among nightshift workers (V). Prolactin behaved oppositeof what was initially predicted. It has beenshown to be higher when participants arepassive 112 . The interpretation of this is thatthe intervention group should be more passivethrough the six month follow up. Thiswould mean that they were more active justbefore the intervention was made, and attime 1 their values were in fact significantlylower than the controls.The approaches to measure stress in shiftworkers have not been theoretically wellfounded. On one hand, shift work has beenlabeled as a stress inducing factor (see p. 16)related to sleep disturbances and social lifeDiscussionconflicts, and on the other several studieshave recently been published looking atstress in shift workers with the argument thatshift work is leading to IHD and that stressmight be one possible explanation. But themodels used seems not to be appropriate.The Karasek-Theorell model measures skilldiscretion and decision authority which areorganizational merits of the jobs. The questionsinclude learning and development ofskills, and conflicting demands and fast workto measure the amount of strain 93 . None ofthe questions adress the working time dimensionsand although publication III suggeststhat shift work is associated with Karasek-Theorell types of job strain, it might be thatthe jobs are just more strained - not that shiftworkers are worse off than day workers ifworking in strained job. In publication IV wealso included questions of the Karasek-Theorellmodel. In this context (all nurses andnurses aides) there were no differences betweenday and shift workers which was asanticipated.The Efford-Reward-Imbalance approachviews the imbalance in certain people of theefforts related to work itself with the percievedreward recieved. In this combination theERI could be used to measure stress in shiftwork, regarding the extra effords connectedto the working hours as an extra strain andthe rewards as for instance extra payment forodd working hours, time off at other hours etcetera as possible rewards for shift work.The existing questionnaire is however notrelating to these aspects of work and as suchERI is not measuring working time relatedstress properly.The importance of creating a theoreticalapproach covering shift work stressors - andways of measuring it - is large. As stress isseen as a major contributor to shift workrelated disease, part of the explanation on theapparent heterogenic results could be that51

Shift work and heart diseasesome people percieve their shift work asstressfull while others do not. This wouldlead to a better understanding of the effectsand to potentials for prevention.Is shift work a a risk factor for IHD?The ultimate question is whether shift workis a risk factor for IHD. The immediateanswer is that the literature suggest an associationwith a relative risk of 1.4 being themost reasonable guess of the strength. Butthen, the discussion points at several methodologicalproblems and the question is maybemore problematic than at first glance. In lineof the previous discussion it could be rephrasedto:Is it a causal risk factor?It might be that confounding makes shiftwork a proxy for other risk factors some ofwhich have been controlled for in the literature,but both differences in social class andwork environment might theoretically beresponsible for the results (publication II andIII).Among several proposals for criterias suggestedto indicate a causal relation (as opposedto a statistical association), Hill (expanding areport from the Surgeon General AdvisoryCommitee on smoking and health) suggestednine aspects to be relevant to judge betweenthe two situations 237 . Rothman and Greenlandhave warned on using them as checklists,giving examples of situations where thearguments does not hold 238 . With Rothmanand Greenlands warnings in mind the case ofshift work and IHD is put up against the Hillcriterias:DiscussionStrength of the association: The literaturereview (I) suggested - in line with previousreviews - a risk of 1.4 to be the most plausible.In the metaanalysis 115 argumenting thatthe aggregated studies were from another“population” of studies we identified twogroups of studies, the “positive” with ahomogene meta-estimate of 1.31 (1.17-1.45). The previous discussion on shift workas an exposure would however questionswhether these four studies are examining thesame relationship and the result must thereforebe carefully interpreted. Even so, this isnot a very strong relationship but although astrong association is normally regarded as asign of causal relationship, Rothman andGreenland cites several examples of exceptionsfrom this rule going in both directions.Nevertheless, a weak association gives thepossibility of a unaccounted modest confounderto be responsible for the association.Consistency: The studies (I) were very heterogeneic,some finding no associationsothers up to a risk of roughly 2. The metaanalysisshowed two subgroups of studies,one essentially positive, one negative. Thiscould suggest that shift work in some instanceswere related to IHD while in others therewere no risk attributable to shift work. Icomment more on this in the next section.Specificity: The original idea of Hill was thata cause could only lead to a single effect.Shift work is besides IHD related to sleepproblems, gastrointestinal disease and socialdisruption. Rothman and Greenland givesseveral examples showing that this aspect isentirely invalid.Temporality: That the exposure is presentbefore the effect is a sine qua non in judginga causal inference. This means that shift workshould lead to IHD in workers that were freefrom disease at the entry. This has been thecase in most of the cohort studies.Biologic gradient: Knutsson 7 demonstrated adose-response up until 20 years of shiftwork. Two other studies found more modest52

Shift work and heart diseasedose-response curves, but as discussed underthe exposure section (p. 38) a possible explanationcould be a misclassification and dilutionof exposure. Rothman and Greenlandshows that a dose-response is not necessarilya sign of a causal effect, a confounder, that ishigher for the most exposed could likewiseproduce this result. Social class should notbe able to produce this result, but if harmfullworking conditions were more prevalentamong shift workers (had a higher OR) backin time and had been approaching the prevalencein day workers with time, this couldtheoretically give rise to the dose-responsecurve described. Publication III is crosssectionalso the results of the age stratificationare not able to show whether this is true;it is however not entirely unrealistic that itcould be the case.Plausability and coherence: Originally theywere seen as two different aspects, but as thedistinction between them is hard they areevaluated as one. The model and discussionon a causal web in publication I suggest thatshift work could be plaing through knownrisk factors. However, it should be rememberedthat it is entirely based on beliefs 239 itmight both be that the model is entirelywrong, or that the suggested findings are notrelated to shift work. The demonstration ofan effect of intervention on biomarkers inpublication IV and of a relation betweennumber of hours worked at night amongnight shift workers and biomarkers in publicationV gives plausability to the claim.Experimental evidence: There are no directanimal experiments on shift work and IHD.An experiment with hamsters that weregenetically susceptible for the developmentof cardiomyopathia showed that a weeklyreversal of the light-dark cycle of the hamstersled to shorter lifespans 240 . A perhapsmore relevant observation is that the changesin three schedules (IV) 18,199 or the starting orDiscussionstopping of night work 147,200 in all five instancesled to changes in biomarkers of IHD inthe predicted direction. The changes werehowever not very homogenously with somerelating to triglycerides, others to HDLcholesterolor total cholesterol, and for allstudies with inadequate controlling for dietarychanges in parallel with the interventions.Analogy: There are no tempting analogies inthe literature stress being a possible exception,but as pointed out by Rothman andGreenland analogies might be handicappedby “the intentive imagination of scientists,who can find analogies everywhere” (p. 27)In summary, the evidence for a causal relationbetween shift work and IHD is perhapson balance unsettled but mostly tilting infavour of a relation. The experimental evidencein five studies and the sharp rise witha longer exposure would favour the relationship,while the relatively low strength and thelack of consistency would be counting against.The evidence at the same time is heavilydependent on few studies with the mostcrucial 7 building on only 43 IHD cases. Thelarger epidemiological studies (including II)are not methodological as good and the areastill needs good epidemiological studies inorder to settle the question. The crucialproblems are exposure assessment, control ofselection, controlling relevant confounders,especially social class and work environment,and comparing with a more relevant controlgroup.Is it a causal risk factor for all?Taking the balance in favour of a causal riskfactor the metaanalysis 115 suggests that shiftwork would only be a risk factor in somesituations. Thus, it might be that shift workwere only a riskfactor for some of the exposedworkers or for exposed workers in cer-53

Shift work and heart diseasetain situations.If stress in the Lazarus sense is involved shiftwork might only be a risk for some people.Some shift workers regard their shift work asnot being particular problematic, they sleepwell, are comfortable with their workinghours, and would not want it any other wayaround. It has been used to propose that shiftwork should be reserved for those choosingit for themselves. It might be that if shiftwork were not percieved as stressfull thenthe risk of IHD would not be higher forthem. It might explain the heterogeneity ofthe epidemiological literature but it has notbeen examined in relation to IHD or biomarkers.Shift work could also be a risk factor onlywhen interacting with other known riskfactors. In a Finnish study 241 shift work hada superadditive effect on BMI and smoking,leading to a relative risk of 2.7 (1.8-4.1) forshift work and smoking (the risk for nonsmoking shift workers 1.6 and for day workingsmokers being 1.3, both compared withnon smoking day workers) and for shift workand overweight 2.3 (1.5-3.6) with the basalrisks being 1.2 and 1.3. This might suggestDiscussionthat shift work would be a risk factor especiallyfor those being obese or smokers. In anyet unpublished study[15] from the SHEEP-VIP case-referent study we repeated theanalyses and found that while smoking andBMI did not act superadditively in this study,WHR had a significantly superadditive effectin women and physical inactivity was superadditivein both genders. This question isthus also still unanswered.Another type of explanation of a selectiverisk would be related to the place and time ofthe study. It would as previously describedbe possible that some shift workers wereforced to work nights as other jobs were notavailable for them. Thus, the societal backgroundcould force some reluctant workersto a work pattern including night work aginsttheir wishes. These shift workers could intheory have a higher risk while self selectedand well adapted had no additional risk. Thishas not been investigated either.[15] Bøggild, Knutsson, Nilsson, Hallquist, Reuterwall.Combined effects of shift work and lifestylefactors on risk of myocardial infarction in the SHE-EP/VIP study. Unpublished54

Shift work and heart diseaseDiscussionTable 5. Exposure information in studies on shift work, risk factors, biomarkers and IHDFirst author, yearAlfredsson 1985 6Alfredsson 1982 5Angersbach 1980 119Baumgart 1989 137Bursey 1990 216Bøggild 1999 (II)Cesana 1985 157Cesana 1990 217Shift work exposure definitionAggregated if reported by 50% in other study: irregular working hoursAggregated if reported by 50% in other study: continously changing day andnight work schedule> ½ year on either (90%) (1): DN--DN-/-DN--DN/--DN--D/N--DN–/ or(10%) (2): DN-DN--/N-DN-DN/-DN-D--/ (D: 6.00-18.30; N: 18-6.30)Eight hour (three shift), weekly rotation with weekends off> five year in shift. Three shift, eight hour.Prevalent, irregular working hours, shift work, often night workWeekly rotating shiftRotating and permanent night workersChau 1989 133 DDAANN-- (D: 4.00-12.00; A: 12.00-20.00; N: 20.00-04.00)Costa 1990 218De Backer 1987 219Fouriaud 1984 213Fredén 1984 242Gordon 1986 212Kawachi 1995 116Kecklund 1994 199Knutsson 1986 7Knutsson 1988 208Not statedThree shift systemWork site physician: shift work, irregular schedulesNot statedDoes your job involve a variable work shift. That is, do you work the day shiftsometime and the night shift at other time> 1 year of > 3 nights a month (rotating)Old: AAAAAAA/--DDD--/DD--FFF/FFFF---/---NNNN/NNN----/New: AAAANNN/--DDD--/DD--AAA/NNNN---/FFFFFFF/-------/(D:7.15-15.45; F: 7.00-15.00; A: 15.00-23.00; N: 23.00-7.00)> 6 month: Before 1975: (1) DDAAAAD*/--DDDD-/NN--NNNN*/AANN---/,after 1975: (2) NNNN---/DD-AAAD*/--DDDD-/AAA-NNNN*/(D: 6.00, A: 14.00, N: 22.00, *=12 hour shift)(3): NNNN---/DD-AAAD/--DDDD-/AA-NNN-/-------/(4): DDDDAAA/A-NNNN-/-------/(5): DD-AAAD/-------/AAA-NNN/--DDDD-/NNNN---/(D: 6.00, A: 14.00, N: 22.00)Knutsson 1989 220 As Knutsson 1988Knutsson 1990 147 As Knutsson 1988 (3)Knutsson 1999 117Did you undertake shiftwork (during the last 5 years), if yes: type of schedule.(and) when was the major part of your work hours scheduled (6-18, 18-22,22-6)?55

Shift work and heart diseaseDiscussionFirst author, yearKoller 1983 243Lang 1988 244Lasfargues 1996 221Shift work exposure definitionANN----/-DDDDDN/N--AAN-/MAANN--/MAANN-D*/ANN----/-DDDDD*N/N--AANN/N--AAN-/DAANN-D*/ANN----/-DDDDD*-/-DDDDD*N/N--AAN-/DAANN-D*/ANN---D*/D:6.50-13.50; A:13.50-20.50; N:20.50-6.50; D*6.50-20.50Work site physician, employees, visits: shift workNight workersLennernäs 1994 146DDDDD--/AAAAA--/NNNN---/(D: 5.30-14.00, A:14.00-22.30; N:22.30-5.30)McNamee 1996 120 >1 month, one week forward rotationMichel-Briand Shift work1980 245Nakamura 1997 86Netterstrøm 1996 222> 3 years, DDDD-AAAA-NNNN--(D: 8.15-15.15; A:15.15-22.30; N: 22.30-8.15)Not statedOrth-Gomér 1983 18 Counter clockwise: 22.00-07.00; 18.00-02.00;14.00-22.00;10.00-18.00;07.00-14.00; - ; - /, Clockwise: 7.00-14.00; 10.00-18.00; 14.00-22.00; 18.00-02.00; 22.00-07.00; - ; - /Peacock 1983 246Old system: NNN*AAMMM------, New system: nn-dd---M: 08.00-16.00, A: 16.00-24.00, N: 24.00-08.00. N*: 24.00-08.00 and 16.00-24.00, d: 07.00-19.00, n:19.00-07.00Peternel 1990 223 Three shift rotation, 7 days between shifting. 6-14, 14-22, 22-6Romon 1992 224 MMAANN---/ 6-14, 14-22, 22-6Steenland 1996 121 Prevalent night shiftSundberg 1987 139 MMMM and NNNN ? M: 7-15, N: 21.30-7.30Taylor 1972 4 Three shift, weekly rotation; three shift, faster rotation; alternating day andnight; double days; rotating 12-hour shift; regular night work.Tenkanen 1997 118 Mostly MMMM-AAAA-NNNN-----, 20 weeks schedule (?), or MMAANN----.Exposure relies on questionnaire data: part-time, two shift, three shift,irregular night workThelle 1976 225Not statedTheorell 1976 226 Three weeks of fixed night shift (?) 22.30-7.30, three weeks of day 7.30-16.30Thiis-Evensen Not stated1949 256

Shift work and heart diseaseDiscussionFirst author, yearTüchsen 1993 177Åkerstedt 1987 209Morikawa 1999 158Murata 1999 161Aanonsen 1964 3Shift work exposure definitionAggregated if reported by 20% in other study. Night and morning work(bakers), late evening (taxi operators, transport, hotels, restaurants andcafés, cooks, waiters), 24-hour work (fishermen, traffic staff, shipmaster,railway staff, bus, coach, road transport, rescue services, customs, excise,police), other irregular work (dataprocessing, doormen, ticket-checkers,marshals, security staff, drivers, production workers, wage-earners notelsewhere classified).Aggregated if reported by 20% in other study: Night work.2/3: DDDDD--/NNNNN--/AAAAA–-, 1/3 DDDD-NNNN-AAAA- or DDD-NNN-AAA- (D: 8.00-, A: 16.30- N: 00.15- or 6.30, 13.00 and 21.30)Four shifts, 1 day off, forward rotationNNNNNN-/MMMMMM-/AAAAAA- or MM-----/AAAAAAM*/-MMMMMA/NNNNNNA/M-MM---/AAAAAAA/-MMMMM-/NNNNNNA/M---MM-/AAAAAAM*/-MMMMM-/NNNN-NNA/ (M*: morning and night shift same day)57

Shift work and heart diseaseDiscussionTable 9. RRPA scores and extracted information from studies of shift work, biomarkersand IHD, giving full information on shift schedule (schedules are shown in table 5)Angersbach (1)Angersbach (2)ChauKecklund (old)Kecklund (new)Knutsson (1)Knutsson (2)Regularity (comb.) 7.0 6.9 7.9 8.5 8.1 8.0 8.4 8.5 8.3 8.2 7.8 8.8 8.4 5.9 5.9 8.0 8.1 7.5Periodicity 8.3 8.0 7.7 5.4 6.8 6.4 6.2 6.3 6.3 6.3 7.2 6.2 5.6 8.0 8.8 7.2 7.6 7.7Load shift 4.7 4.7 8.0 7.9 7.9 7.6 7.6 8.0 8.0 8.0 7.3 7.9 7.8 7.9 7.9 7.1 5.1 8.0Load week 7.4 6.6 7.7 8.2 8.2 7.6 8.3 8.4 8.2 8.4 7.5 8.1 7.6 8.0 8.0 7.9 7.6 8.1Opport rest night 7.5 7.1 7.7 8.1 8.2 7.5 8.0 8.0 8.1 8.0 7.6 8.0 7.4 8.4 8.4 7.8 7.7 7.9Predictability 8.9 8.8 8.1 4.6 5.1 5.0 3.3 5.6 7.2 5.6 2.2 7.5 4.6 8.4 8.4 6.0 7.2 6.9Opport household 7.5 6.7 7.6 7.1 7.1 7.6 8.0 8.0 7.8 8.0 7.6 6.7 7.6 6.9 6.9 7.7 7.6 7.9actOpport evening 7.6 6.8 7.6 8.1 8.2 7.5 8.0 8.0 8.5 8.0 7.5 6.6 7.5 6.4 6.4 6.9 7.6 7.8recr.Opport weekend 1.5 5.0 0.8 5.5 5.5 4.0 4.4 4.4 4.0 4.9 3.1 10.0 2.7 10.0 3.0 5.5 3.0 3.3recr.worktime/w 43.8 50 42 35.1 35.1 42 33.6 32 32 32 41.6 37.7 42 40 40 36 42 37.8operating time 172 148 168 168 168 168 168 168 112 168 168 113 168 40 40 56 168 168hour 22-06 56 48 112 56 56 56 56 56 32 56 224 33 224 12 12 24 112 112< 11 hours off 0 0 0 0 0 2 1 1 0 1 0 0 0 0 0 1 0 0< 36 hours off 0 5 0 0 0 0 0 0 0 0 6 0 0 0 0 0 7 0free sunday in rota 1 2 1 3 3 2 3 3 1 3 5 3 4 1 1 1 3 3max nbr of night/row 1 1 2 7 4 3 3 3 4 4 3 4 4 . . 3 2 2nbr of night shifts in 7 6 16 7 7 7 7 7 4 7 28 4 28 . . 3 14 14rotamax nbr of. . 2 7 4 4 3 3 4 3 2 5 4 . . 2 . 2evening/rownbr of evening . . 16 7 7 6 6 6 4 6 20 5 28 . . 2 . 14shifts in rotashift length 12 12 8 8 8 8/128/12 8 8 8 8/12 8 8/10 8 8 8 12 8number of shifts 2 2 3 4 4 5 5 3 3 3 4 3 3 5 5 4 2 3length of rota 4 3 8 6 6 4 5 5 3 5 16 3 16 1 1 2 8 9hour 22-06/week 14 16 14 9.3 9.3 14 11.211.210.611.2 14 11 14 12 12 12 14 12.4Knutsson (3)Knutsson (4)Knutsson (5)KollerLennernäsNakamuraOrth-gomér (ccr)Orth-gomér (cr)Peacock (old)Peacock (new)Romon58

Shift work and heart diseaseConclusion and perspectivesConclusion and perspectivesIn conclusion! the studies suggests that although socialclass and work environment differs betweenshift and day workers shift workcould be a causal risk factor for IHDespecially as intervention towards ergonomicallybetter schedules change biomarkersof IHD and as different shift characteristicsas the amount of night work arerelated to biomarkers.! By changing shift schedules in accordancewith ergonomic criterias introducingregularity and predictability the lipid biomarkersof IHD were changed in thepredicted direction.! Shift work, night work, and especially thenumber of hours worked between 22 and6 among night workers, were associatedto lipid biomarkers and glycated hemoglobin.Regularity and perodicity were alsoassociated as predicted, regardless ofwhether exposure was taken as the resultof the intervention or regardless of this.! In general, except for glycated hemoglobinneither the stress biomarkers norquestionnaire data were associated withshift work but an unsolved problem is thatthe stress of shift work is not very theoreticalunderstood.Shift work as a risk factor for IHDOn balance shift work seems to be a causalrisk factor for IHD. Some comments shouldhowever be given.With regard to the causal chains that must beconnecting shift work with IHD the literaturehas been concentrating on lifestyle factors,and although they did not have impact in theintervention study it might be that smokingand especially diet could be responsible forthe association. This would not mean thatshift work was not leading to IHD, but merelythat the timing of eating was acting as amediating factor. It would, however, beimportant to know in relation to preventionstrategies. The results of the interventionstudies suggests that causality via diet is lessprobable.In contrast, not much has been done toevaluate other possible causal pathwaysespecially relating to circadian disruption andstress. For the latter, both sleeping and domesticproblems have been described atnumerous occasions and it is surprisinglywhat little focus has been on these agendas inrelation to risk of IHD or biomarkers.The population impact of shift work on IHDwas evaluated by Olsen and Kristensen 72 suggestinga attributable risk of 7% for bothmen and women. This might be to high. Isuggested the risk to encompass only part ofthe exposed population although the precisemechanisms is not known, and as the riskmight be related only to shift workencompassing certain characteristics as forinstance night work the number of exposedpeople might also be lower than assumed inthe calculations by Olsen and Kristensen. Inpopulation III, 3.6% of the employees hadnight work, 4.8% evening or two shift work,and 10.4% other types of irregular workinghours. This could mean that the exposedgroup could be below 5% (in relation to riskof IHD).For the exercise, the exposed population isset at 5%, 50% of whom are regarded assensitive and the result of the metaanalysis 115is used as the relative risk (1.3). This leads toan attributable risk of (1.3 - 1) * 0.5 * 0.05/((1.3 - 1) *((0.5 * 0.05 ) + 1)) or 2.4%.Besides the unsecurity of the figures in the59

Shift work and heart diseasecalculations aetiological fractions has theproblem of assuming that the causes ofdisease are unrelated which is probably notthe case for IHD.Relating to the biomarkers a 15% reductionin for instance total/HDL-cholesterol followingan ergonomically better schedule (IV)would have potential impact on the exposedshift working population as the distributionof cholesterol in the population would shift 247 .Even so, the potential benefit of preventioncould be larger as other negative effects ofshift work are also considered to bepreventable by changing the risk of IHDalthough nothing is known on the schedulecharacteristics and for instancegastrointestinal disease.Future researchThe research perspectives were described inpublication I but a few highlights are relevantin relation to the discussion in this thesis.There is still a need for well conductedepidemiological studies. They wouldhowever need to be better planned if resultsshould be less open for debate. First andforemost the populations should be muchbetter defined in terms of exposure and especiallythorough selection of the control groupshould be given high priority. It might notnecessarily be day workers but could be forinstance two shift workers or workers with aless demanding schedule (taking advantageof for instance the RRPA or other ways ofdescribing the schedule) in order to secure areference group that is more closelyresembling the shift workers (except on theshift work).Controlling should be done on age, gender,social class, and work environment factors.The possiblity to control also for personalityfactors as hostility should be considered. TheConclusion and perspectiveseffects of sleep deprivation, social disruption,and shift work related stress in relation toIHD should be given priority in order to fillthe gaps in knowledge on the web ofcausation.Information on lifestyle factors should betreated as both confounders and mediatingvariables and multivariate models should bepresented both with and without these.Information on biomarkers should normallybe considered mediating factors, especiallylipids and blood pressure that could bemediating a higher risk.For future reanalysis as much information onthe shift work schedule and the dose of shiftwork should be included in the “material andmethods” section of the publication,preferable the schedules worked in.The biomarker studies should preferable bemore precise in their measuring, eitherensuring that participants are at dayorientation or that they measure severalsamples a day in order to obtain informationon the circadian phase. New markers mightbe useful, especially in relation tohemostasis 54,87 and stress 110 . The descriptionof timing and kinetics of markers should begiven priority.Further work on definition of shift workexposure assessment should be made eventuallyin terms of “integrated lifetime hoursof night shift work”.Preventive perspectivesHow could the information from publicationIV and V be used in prevention?The results show that the shift schedule isassociated with biomarker levels. Schedulesthat follows the ergonomic criterias lowersbiomarkers thus potentially leading to lowerr60

Shift work and heart diseaseisk of IHD. In the NARFE II study a moreregular and predictable schedule led to betterbiomarkers. The schedules were very irregularfrom the outset, and as the wards wereself selected it might not be a random sampleof shift systems in Danish hospitals. Otherschedules might be more regular and itwould be relevant to test changes in othersettings.In a study on 402 German industrial andservice sector schedules Gissel and Knauth 248found that many of the schedules could beimproved from a ergonomically point ofview. It seems plausible that this situationwould not be very different in Denmark,although information on Danish shift schedulesare not available.The other schedule characteristics of relevanceis that the number of night shift hourswere associated with biomarkers amongnight workers. This suggests that the lowestrisk would be obtained by spreading out thenight shift hours on more people in order tohave the lowest change in distribution. Ishould note though, that we had only 6 fixednight shift workers in the analysis, and thisarea would be very relevant to persue.Conclusion and perspectivesOn the other hand, changes in society callsfor more flexibility, both on behalf of theemployers (possibility of having peopleworking at short notice, in schedules changingwith market and production 249 , leanproduction, production on demand et cetera)and on behalf of the employees who wishesfor flexible shift scheduling in order to obtainautonomous organisation of social anddomestic life 249 . The labour market maybecome more segregated with people havinglong term contracts with individual flexiblityand short term employees having to workwhen work is needed. This might give rise tomore pronounced social differences in workinghour conditions.The result of publication IV suggest thatalthough flexibility is judged highly positiveit should be followed by regularity. This willnecessarily mean that employers, cooperationcommittees and health and safety committees,as well as the employees themselves willhave to put up rules for the scheduling ofboth shift and other flexible types of workschedules in order to maintain regularity inthe growing demand for flexibility.61

Shift work and heart diseaseAbstractAbstractWork situated outside normal day workinghours is the condition for many employees.These shift working hours have been relatedto heart disease for the last 15 years but withstudies not finding a higher risk. Severalmethodological problems have been described,and it was the aim of this thesis to examinewhether shift work could be regardedas a causal risk factor for ischemic heartdisease (IHD).The epidemiological literature have beenextensively reviewed, and possible causallinks discussed. The problem is attackedfrom two points: Two publications discussesthe effect of social class and work environmentfactors, both risk factors for IHD. Bothconfounders are heavily screwed betweenday and shift workers, and have potentialimpact on the measured relationship betweenshift work and IHD. The second point comprisedof a intervention in shift schedulingaiming at more ergonomic schedules. Theintervention at four hospital wards succededin changing the schedules, and in parallelwith this biochemical risk factors (biomarkers)changed towards lower values. Thebaseline figures were utilised to examine theeffect of different characteristics of shiftwork and showed that especially night workand among night workers the amount ofnight work (given as hours between 10 pmand 6 am in an eight week period) was associatedto higher biomarkers of IHD, especiallytriglycerides and HDL-cholesterol. Stressbiomarkers were in general neither changingnor associated to characteristics, with theexception of glycated hemoglobin that washigher among shift workers.The thesis discusses confounding and exposurecharacteristics and concludes, that shiftwork is probably a causal risk factor raisingthe risk by 30-40%, but with the possibilitythat shift work is interacting with unknownfactors so that in some instances the risk isnot raised.The demonstration that shift scheduling caninfluence biomarkers of IHD suggests thatshift work associated IHD may be preventedby using knowledge on circadian rhytmicityand social preferences, as suggested byseveral lists of ergonomic scheduling. Thepublications in this thesis suggests, thatespecially regularity of schedules and theamount of night shift work is relevant forprevention.62

Shift work and heart diseaseResuméResuméArbejde udenfor normal dagtid er hverdagfor mange ansatte. Disse skiftarbejdstider harværet relateret til udvikling af hjertesygdom.Litteraturen er imidlertid behæftet med mangemetodemæssige problemer, og det varmålet med denne afhandling at undersøge omskiftarbejde kunne betragtes er en kausalårsag til iskæmisk hjertesygdom (IHS)Den epidemiologiske litteratur er grundigtgennemgået og mulige kausale årsager diskuteret.Afhandlingen angriberproblemstillingen fra to sider: I to publikationerundersøges effekten af forskelle i socialklasse og forekomsten af andre arbejdsmiljøfaktorermellem dag og skiftarbejdere pårisikoen. Begge potentielle konfoundere varskævt fordelt mellem grupperne og kunnevære af betydning. Det andet angrebspunktbestod af en kontrolleret intervention i planlægningenaf skift arbejde med henblik påindførelse af mere ergonomisk planlagtearbejdstidsystemer. Interventionen på firesygehusafdelinger medførte arbejdstidssystemerder var mere regelmæssige og parallelthermed ændredes kolesterol og lipoprotein(biokemiske risikofaktorer (biomarkører) forIHS) i retning af lavere værdier.Ved analyse af tværsnitsdata fra indgangen tilinterventionsundersøgelsen fandtes sammenhængmellem karakteristika ved skiftsystemetog biomarkører. Der fandtes sammenhængmellem skift arbejde og højere biomarkører(særligt triglycerid og HDL-kolesterol), ogsærligt natarbejde og blandt natarbejderneomfanget af natarbejde (udtrykt som antaltimer mellem 22 og 6 på otte ugers vagtplan)og højere biomarkører for IHS.Bortset fra glykeret hæmoglobin var biomarkørerfor stress ikke associeret til interventioneneller til karakteristika ved arbejdstidssystemet.Afhandlingen diskuterer konfounding ogeksponeringskarakterisering for skiftarbejde,og konkluderer at skiftarbejde formentlig eren kausal risikofaktor med en øget risiko på30-40%. Der er dog mulighed for at skiftarbejdeinteragerer med andre ukendte risikofaktorer,og at risici kun er øget i visse situationer.Påvisningen af at skiftsystemets udformninghar betydning for niveauet af biomarkører forIHS medfører en anbefaling af at en skiftassocieretrisiko for IHS kan forebygges gennemergonomiske principper for skiftsystemetsopbygning. Disse principper, der bl.a.bygger på viden om circadiane rytmer ogsociale ønsker findes i flere versioner. Iartiklerne der hører til denne afhandling erdet særligt betydningen af regelmæssighed ogomfanget af natarbejde, der er relevant forforebyggelse.63

Shift work and heart diseaseReferencesReferences1. Harrington JM. Shift work and health - A criticalreview of the literature. London: Her Majesty'sStationary Office, 1978.2. Thiis-Evensen E. Skiftarbeid og helse. Porsgrunn:Andreas Jakobsens boktrykkeri, 1949.3. Aanonsen A. Shift work and health. København:Scandinavian University Books, Munksgaard,1964.4. Taylor PJ, Pocock SJ. Mortality of shift and dayworkers 1956-68. Br J Ind Med 1972;29:201-7.5. Alfredsson L, Karasek R, Theorell T. Myocardialinfarction risk and psychosocial work environment:an analysis of the male swedish work force. SocSci Med 1982;16:463-7.6. Alfredsson L, Spetz C, Theorell T. Type of occupationand near-future hospitalization for myocardialinfarction and some other diagnoses. Int JEpidemiol 1985;14:378-88.7. Knutsson A, Åkerstedt T, Jonsson BG, Orth-Gomér K. Increased risk of ischaemic heart diseasein shift workers. Lancet 1986;2(8498):89-92.8. Åkerstedt T, Knutsson A, Alfredsson L, TheorellT. Shift work and cardiovascular disease. Scand JWork Environ Health 1984;10:409-14.9. Akerstedt T, Knutsson A. Shiftwork. In: Levy BS,Wegman DH (eds). Occupational health,recognizing and preventing work-related disease.Boston: Little Brown & Co, 1994; p. 407-17.10. Costa G. The impact of shift and night work onhealth. Appl Ergon 1996;27:9-16.11. Waterhouse JM, Folkard S, Minors DS. Shiftwork,health and safety: an overview of the scientificliterature 1978-1990. HMSO, 1993.12. Harrington JM. Shift work and health - a criticalreview of the literature on working hours. AnnAcad Med Singapore 1994;23:699-705.13. Kristensen TS. Cardiovascular diseases and thework environment: a critical review of theepidemiologic literature on nonchemical factors.Scand J Work Environ Health 1989;15:165-79.14. Kristensen TS. Cardiovascular diseases and thework environment. In: Cheremisinoff PN (eds).Encyclopedia of environmental control technology.Houston: Gulf Publishing Company, 1994; p. 217-43.15. Theorell T. The psycho-social environment, stress,and coronary heart disease. In: Marmot M, ElliottP (eds). Coronary heart disease epidemiology: fromaetiology to public health. Oxford: OxfordUniversity Press, 1992; p. 256-73.16. Knutsson A. Shift work and coronary heart disease.Scand J Soc Med 1989;44 suppl.:1-36.17. Wagner U. Schichtarbeit und ischämischeherzkrankheit. Z Arztl Fortbild (Jena)1989;83:229-33.18. Orth-Gomér K. Intervention on coronary riskfactors by adapting a shift work schedule tobiologic rhythmicity. Psychosom Med 1983;45:407-15.19. International Labour Office. Conditions of workdigest. Geneva: ILO, 1986.20. Jeppesen HJ. Skiftarbejde - et kendt og fortrængtproblem. København: Specialarbejderforbundet iDanmark, 1990.21. Nielsen PB. Nyhedsbrev: Nå der bliver vendt op ogned på dag og nat. København: Arbejdstilsynet,1998.22. Nord-Larsen M, Ørhede E, Nielsen J, Burr H.Lønmodtagernes arbejdsmiljø 1990. København:Arbejdsmiljøfondet, 1992.23. Åkerstedt T. Arbetstid, hälsa och säkerhet: ensammanfattning. Stockholm: Stress ResearchReports no. 274, 1997.24. US Congress,Office of Technology assessment.Biological rhythms: implications for the worker.OTA-BA-463. Washington DC: US Governmentprinting office, 1991.25. Wedderburn A. Statistics and news. Dublin:Bulletin of European Studies on Time (BEST).European Foundation for the Improvement ofLiving and Working Conditions, 1996.26. Daus C, Sanders DN, Campbell D. Consequencesof alternative work schedules. In: Cooper CL,64

Shift work and heart diseaseRobertson IT (eds). International Review of Industrialand Organizational Psychology, vol. 13. JohnWiley & Sons Ltd, 1998; p. 185-223.27. Rosa RR, Colligan MJ, Lewis P. Extended workdays:Effects of 8-hour and 12-hour rotating shiftschedules on performance, subjective alertness,sleep patterns and psychosocial variables. WorkStress 1989;3:21-3.28. Smith L, Folkard S, Tucker P, Macdonald I. Workshift duration: a review comparing eight hour and12 hour shift systems. Occup Environ Med1998;55:217-29.29. Spurgeon A, Harrington JM, Cooper CL. Healthand safety problems associated with long workinghours: a review of the current position. OccupEnviron Med 1997;54:367-75.30. Wallace M, Greenwood KM. Twelve-hour shifts.Work Stress 1995;9:105-8.31. Wedderburn A. Compressed working time. Dublin:Bulletin of European Studies on Time (BEST).European Foundation for the Improvement ofLiving and Working Conditions, 1996.32. Colligan MJ, Rosa RR. Shiftwork effects on socialand family life. In: Scott AJ, Ladou J (eds). OccupationalMedicine: State of the Art Reviews.Philadelphia: Hanley & Belfus, Inc., 1990; p. 315-22.33. Tepas DI, Carvalhais AB. Sleep patterns ofshiftworkers. In: Scott AJ, Ladou J (eds).Occupational Medicine: State of the art reviews.Philadelphia: Hanley & Belfus, Inc., 1990; p. 199-208.34. Åkerstedt T. Sleepiness as a consequence of shiftwork. Sleep 1988;11:17-34.35. Åkerstedt T. Psychological andpsychophysiological effects of shift work. Scand JWork Environ Health 1990;16:67-73.36. Smith L, Folkard S, Poole C. Increased injuries onnight shift. Lancet 1994;344:1137-9.37. Bisanti L, Olsen J, Basso O, Thonneau P, KarmausW. Shift work and subfecundity: a Europeanmulticenter study. JOEM 1996;38:352-8.38. Bodin L, Axelsson G, Ahlborg G. The associationof shift work and nitrous oxide exposure inpregnancy with birth weight and gestational age.Epidemiology 1999;10:429-36.References39. Nurminen T. Shift work and reproductive health.Scand J Work Environ Health 1998;24 (suppl3):28-34.40. Cole RJ, Loving RT, Kripke DF. Psychiatricaspects of shiftwork. In: Scott AJ, Ladou J (eds).Occupational Medicine: State of the art reviews.Philadelphia: Hanley & Belfus, Inc., 1990; p. 301-14.41. Kripke DF, Drennan MD, Elliott JA. The complexcircadian pacemaker in affective disorders. In:Touitou Y, Haus E (eds). Biologic rhythms inclinical and laboratory medicine. Berlin: Springer-Verlag, 1992; p. 265-76.42. Moore JG. Chronobiology of the gastrointestinalsystem. In: Touitou Y, Haus E (eds). Biologicrhythm in clinical and laboratory medicine. Berlin:Springer Verlag, 1992; p. 410-7.43. Tüchsen F, Jeppesen HJ, Bach E. Employmentstatus, non-daytime work and gastric ulcer in men.Int J Epidemiol 1994;23:365-70.44. Vener KJ, Szabo S, Moore JG. The effect of shiftwork on gastrointestinal (GI) function: a review.Chronobiologica 1989;16:421-39.45. Anderson JR. Muir's textbook of pathology.London: Edward Arnold, 1985.46. Sundhedsministeriet. Hjertesygdom i Danmark.København: Sundhedsministeriet, 1994.47. Tüchsen F, Endahl LA. Increasing inequality inischaemic heart disease morbidity among employedmen in Denmark 1981-1993: the need for a newpreventive policy. Int J Epidemiol 1999;28:640-4.48. Marmot M. Coronary heart disease: rise and fall ofa modern epidemic. In: Marmot M, Elliott P (eds).Coronary heart disease epidemiology: fromaetiology to public health. Oxford: OxfordUniversity Press, 1992; p. 3-19.49. Osler M, Sørensen T, Sørensen S, Rostgaard K,Jensen G, Iversen L, Kristensen TS, Madsen M.Trends in mortality, incidence and case fatality ofischaemic heart disease in Denmark, 1982-92. IntJ Epidemiol 1996;25:1154-61.50. Nieto FJ. Cardiovascular disease and risk factorepidemiology: a look back at the epidemic of the20th century (editorial). Am J Public Health1999;89:292-4.51. WHO Scientific group on cardiovascular disease65

Shift work and heart diseaserisk factors. Cardiovascular disease risk factors:new areas for research. Geneve: World HealthOrganisation, 1994.52. Pyörälä K, De Backer G, Graham I, Poole-WilsonP, Wood D. Prevention of coronary heart disease inclinical practise: recommendations of the task forceof the european society of cardiology, europeanatherosclerosis society and european society ofhypertension. Europ Heart J 1994;15:1300-31.53. Epstein FH, Kannel B, Klimov N, Meade TW.Relation of risk factors to clinical events. In:Olsson AG (eds). Atherosclerosis: biology andclinical science. Edinburgh: Churchill Livingstone,1987; p. 381-8.54. Pahor M, Elam MB, Garrison RJ, Kritchevsky SB,Applegate WB. Emerging noninvasive biochemicalmeasures to predict cardiovascular risk. ArchIntern Med 1999;159:237-45.55. Hopkins PN, Williams RR. A survey of 246suggested coronary risk factors. Atheroscler1981;40:1-52.56. Sytkowski PA, D'Agostino RB, Belanger A, KannelWB. Sex and time trends in cardiovascular diseaseincidence and mortality: the Framingham HeartStudy, 1950-1989. Am J Epidemiol 1996;143:338-50.57. Sundhedsstyrelsen. Forebyggelse af iskæmiskhjertesygdom i Danmark. København:Sundhedsstyrelsen, 1994.58. Wohlfarth T. Socioeconomic inequality andpsychopathology: are socioeconomic status andsocial class interchangeable? Soc Sci Med1997;45:399-410.59. Svalastoga K. Prestige, class and mobility.Copenhagen: Munksgaard, 1959.60. Hansen EJ. Danskernes levekår 1986 sammenholdtmed 1976. København: Hans Reitzels forlag, 1986.61. González MA, Artalejo FR, Calero JdR.Relationship between socioeconomic status andischaemic heart disease in cohort and case-controlstudies: 1960-1993. Int J Epidemiol 1998;27:350-8.62. Kaplan GA, Keil JE. Socioeconomic factors andcardiovascular disease: a review of the literature.Circulation 1993;88:1973-98.63. Nilsson PM, Møller L, Östergren PO. Social classand cardiovascular disease - an update. Scand JSoc Med 1995;23:3-8.References64. Marmot MG, Smith GD, Stansfeld S, Patel C,North F, Head J, White I, Brunner E, Feeney A.Health inequalities among British civil servants:the Whitehall II study. Lancet 1991;337:1387-93.65. Smith GD, Blane D, Bartley M. Explanations forsocio-economic differentials in mortality: evidencefrom Britain and elsewhere. European Journal ofPublic Health 1994;4:131-44.66. Blane D, Bartley M, Smith GD. Disease aetiologyand materialist explanations of socioeconomicmortality differentials. European Journal of PublicHealth 1997;7:385-91.67. Smith GD. Down at heart - the meaning andimplications of social inequalities in cardiovasculardisease. Journal of the Royal College of Physiciansof London 1997;31:414-24.68. Marmot M, Theorell T. Social class andcardiovascular disease: the contribution of work.International Journal of Health Services1988;18:659-74.69. Netterstrøm B, Kristensen TS, Møller L, Jensen G,Schnohr P. Angina pectoris, job strain, and socialstatus: a cross-sectional study of employed urbancitizens. Int J Behav Med 1998;5:312-22.70. Kristensen TS. Cardiovascular diseases and thework environment: a critical review of theepidemiologic literature on chemical factors. ScandJ Work Environ Health 1989;15:245-64.71. Nurminen M, Hernberg S. Effects of interventionon the cardiovascular mortality of workers exposedto carbon disulphide: a 15 year follow up. Br J IndMed 1985;42:32-5.72. Olsen O, Kristensen TS. Impact of workenvironment on cardiovascular diseases inDenmark. J Epidemiol Community Health1991;45:4-10.73. Goldstein MG, Niaura R. Psychological factorsaffecting physical condition: cardiovascular diseaseliterature review. Part I: coronary artery diseaseand sudden death. Psychosomatics 1992;33:134-55.74. Niaura R, Stoney CM, Herbert PN. Lipids inpsychological research: the last decade. BiolPsychol 1992;34:1-43.75. Hulka BS. Overview of biological markers. In:Hulka BS, Wilcosky TC, Griffith JD (eds).66

Shift work and heart diseaseBiological markers in epidemiology. New York:Oxford University Press, 1990; p. 3-15.76. Manson JE, Tosteson H, Satterfield S, Hebert P,O'Connor GT, Buring JE, Hennekens CH. Theprimary prevention of myocardial infarction. NEngl J Med 1992;326:1406-16.77. Thelle DS. The relative importance of blood lipidsand other atherosclerosis risk factors. Eur Heart J1992;13 (suppl. 3):29-33.78. Anonymous. Triglyceride, high density lipoprotein,and coronary heart disease. National institutes ofhealth, 1992.79. Marmot M. Work and other factors influencingcoronary health and sickness absence. Work Stress1994;8:191-291.80. Evans K, Laker MF. Intra-individual factorsaffecting lipid, lipoprotein and apolipoproteinmeasurement: a review. Ann Clin Biochem1995;32:261-80.81. Law MR, Wald NJ, Thompson SG. By how muchand how quickly does reduction in serumcholesterol concentration lower risk of ischaemicheart disease? BMJ 1994;308:367-72.82. Singer DE, Nathan DM, Anderson KM, WilsonPW, Evans JC. Association of HbA1c withprevalent cardiovascular disease in the originalcohort of the Framingham Heart Study. Diabetes1992;41:202-8.83. Jespersen J. Pathophysiology and clinical aspectsof fibrinolysis and inhibition of coagulation. DanMed Bull 1988;35:1-33.84. Møller LF, Jespersen J, Kristensen TS. Plasmafibrinogenog hjerte-kar-sygdom. Ugeskr Læger1991;153:2343-5.85. Møller L, Kristensen TS. Plasma fibrinogen andischemic heart disease risk factors. ArteriosclerThrombz 1991;11:344-50.86. Nakamura K, Shimai S, Kikuchi S, Tominaga K,Takahashi H, Takana M, Nakano S, Motohashi Y,Nakadaira H, Yamamoto M. Shift work and riskfactors for coronary heart disease in Japanese bluecollarworkers: serum lipids and anthropometriccharacteristics. Occup Med 1997;47:142-6.87. Nordøy A. Haemostatic factors in coronary heartdisease. J Intern Med 1993;233:377-83.References88. Meade TW, Ruddock V, Stirling Y, Chakrabarti R,Miller GJ. Fibrinolytic activity, clotting factors,and long-term incidence of ischaemic heart diseasein the Northwick Park Heart Study. Lancet1993;342:1076-9.89. Segal A. Hypertriglyceridaemia and vascular risk -report of a meeting of physicians and scientists,University College London Medical School. Lancet1993;342:781-7.90. Jespersen J, Munkvad S, Gram JB. The fibrinolysisand coagulation systems in ischaemic heart disease- risk markers and their relation to metabolicdysfunction of the arterial intima. Dan Med Bull1993;40:495-502.91. Lazarus RS, Folkman S. Coping and adaption. In:Dayle-Gentry W (eds). Handbook of BehavionalMedicine. New York: Guilford Press, 1984; p. 282-325.92. Arnold J, Robertson IT, Cooper CL. Work, stressand psychological well-being. In: Arnold J,Robertson IT, Cooper CL (eds). Work psychology -understanding human behaviour in the workplace.Pitman publishing, 1993; p. 284-305.93. Karasek R, Theorell T. Healthy work: stress,productivity, and the reconstruction of workinglife. Basic Books, 1990.94. Kristensen TS. Job stress and cardiovasculardisease: a theoretic critical review. J Occup HealthPsychol 1996;1:246-60.95. Schnall PL, Landsbergis PA, Baker D. Job strainand cardiovascular disease. Ann Rev Public Health1994;15:381-411.96. Theorell T, Karasek R. Current issues relating topsychosocial job strain and cardiovascular diseaseresearch. J Occup Health Psychol 1996;1:9-26.97. Bosma H, Marmot MG, Hemingway H, NicholsonAC, Brunner E, Stansfeld SA. Low job control andrisk of coronary heart disease in Whitehall II(prospective cohort) study. BMJ 1998;314:55898. Siegrist J, Klein D, Matchinger H. Occupationalstress, coronary risk factors and cardiovascularresponsiveness. In: Weiner H (eds). Frontiers ofstress research. Toronto: Hans Huber Publishers,1989; p. 323-35.99. Siegrist J. Adverse health effects of higheffort/low-rewardconditions. Journal ofoccupational health psychology 1996;1:27-41.67

Shift work and heart disease100. Siegrist J, Peter R. Job stressors and copingcharacteristics in work-related disease: issues ofvalidity. Work Stress 1994;8:130-40.101. Peter R, Alfredsson L, Hammar N, Siegrist J,Theorell T, Westerholm P. High effort, lowreward, and cardiovascular risk factors inemployed swedish men and women: baselineresults from the WOLF study. J EpidemiolCommunity Health 1998;52:540-7.102. Brindley DN, McCann BS, Niaura R, Stoney CM,Suarez EC. Stress and lipoprotein metabolism:modulators and mechanisms. Metabolism 1993;42(1 suppl):3-15.103. Björntorp P. Stress and cardiovascular disease.Acta Physiol Scand 1997;640:144-8.104. Kelly S, Hertzman C, Daniels M. Searching for thebiological pathways between stress and health.Ann Rev Public Health 1997;18:437-62.105. Rosenman RH. Psychological influences on thevariability of plasma cholesterol. Homeostasis1993;34:129-36.106. Orth-Gomér K, Eriksson I, Moser V, Theorell T,Fredlund P. Lipid lowering through work stressmanagement. Int J Behav Med 1994;1:204-14.107. Friedman M, Rosenman RH, Carroll V. Changes inthe serum cholesterol and blood clotting time inmen subjected to cyclic variation of occupationalstress. Circulation 1958;17:852-61.108. Netterstrøm B, Danborg L, Olesen H. Glycatedhemoglobin (HbA1C) as a measure of jobstress.Behav Med 1988;14:13-6.109. Netterstrøm B, Sjøl A. Glycated haemoglobin(HbA1C) as an indicator of job strain. Stress Med1991;7:113-8.110. Hertzman C, Kelly S. Measuring stress with ablood test: a discussion document. CIAR workingpaper no. 30. Toronto: The Canadian Institute forAdvanced Research, 1994.111. Theorell T, Orth-Gomér K, Moser V, Unden AL,Eriksson I. Endocrine markers during a jobintervention. Work Stress 1995;9:67-76.112. Theorell T. Prolactin - a hormone that mirrorspassiveness in crisis situations. Integr PhysiolBehav Sci 1992;27:32-8.113. Mann H, Cremer P, Nagel D, Labrot B, Strok J.ReferencesCase-control study regarding myocardial infarction,shiftwork and occupational exposure to noise.Epidemiology in occupational health, 9.international symposium. Cincinnati: InternationalCommision on Occupational Health, 1992; p. 157.114. Rafnsson V, Gunnarsdóttir H. Mortality study offertiliser manufactures in Iceland. Br J Ind Med1990;47:721-5.115. Bøggild H, Knutsson A. Shift work and heartdisease: meta analysis of the epidemiologicalliterature. In: Hornberger S, Knauth P, Costa G,Folkard S (eds). Shiftwork in the 21st century.Challanges for research and practics. Frankfurt:Peter Lang, 2000; (in press).116. Kawachi I, Colditz GA, Stampfer MJ, Willett WC,Manson JE, Speizer FE, Hennekens CH.Prospective study of shift work and risk of coronaryheart disease in women. Circulation 1995;92:3178-82.117. Knutsson A, Hallqvist J, Reuterwall C, Theorell T,Åkerstedt T. Shift work and myocardial infarction:a case-control study. Occup Environ Med1999;56:46-50.118. Tenkanen L, Sjöblom T, Kalimo R, Alikoski T,Härmä M. Shift work, occupation and coronaryheart disease over 6 years of follow up in theHelsinki Heart Study. Scand J Work EnvironHealth 1997;23:257-65.119. Angersbach D, Knauth P, Loskant H, KarvonenMJ, Undeutsch K, Rutenfranz J. A retrospectivecohort study comparing complaints and diseases inday and shift workers. Int Arch Occup EnvironHealth 1980;45:127-40.120. McNamee R, Binks K, Jones S, Faulkner D,Slovak A, Cherry NM. Shiftwork and mortalityfrom ischaemic heart disease. Occup Environ Med1996;53:367-73.121. Steenland K, Fine L. Shift work, shift change, andrisk of death from heart disease at work. Am J IndMed 1996;29:278-81.122. Folkard S. Editorial. Ergonomics 1993;36(1-3):1-2.123. Taylor E, Briner RB, Folkard S. Models ofshiftwork and health: an examination of theinfluence of stress on shiftwork theory. Humanfactors 1997;39:67-82.124. Hastings M. The brain, circadian rhythms, andclock genes. BMJ 1998;317:1704-7.68

Shift work and heart disease125. Edmunds LN. Cellular and molecular aspects ofcircadian oscillators: models and mechanisms forbiological timekeeping. In: Touitou Y, Haus E(eds). Biologic rhythms in clinical and laboratorymedicine. Berlin: Springer-Verlag, 1992; p. 35-54.126. Arendt J. Melatonin and the pineal gland. In:Arendt J, Minors DS, Waterhouse JM (eds).Biological rhythms in clinical practice. London:Wright, 1989; p. 184-206.127. Knauth P, Rutenfranz J. Experimental shift workstudies of permanent night, and rapidly rotating,shift systems. Int Arch Occup Environ Health1976;37:125-37.128. Touitou Y, Haus E. Biologic rhythms in clinicaland laboratory medicine. Berlin: Springer-Verlag,1992.129. Haus E, Touitou Y. Principles of clinicalchronobiology. In: Touitou Y, Haus E (eds).Biologic rhythms in clinical and laboratorymedicine. Berlin: Springer-Verlag, 1992; p. 6-34.130. Lemmer B. Circadian rhythms in thecardiovascular system. In: Arendt J, Minors DS,Waterhouse JM (eds). Biological rhythms inclinical practice. London: Wright, 1989; p. 51-70.131. Lemmer B. Cardiovascular chronobiology andchronopharmacology. In: Touitou Y, Haus E (eds).Biologic rhythm in clinical and laboratorymedicine. Berlin: Springer Verlag, 1992; p. 418-27.132. Cornelissen G, Haus E, Halberg F. Chronobiologicblood pressure assessment from womb to tomb. In:Touitou Y, Haus E (eds). Biologic rhythms inclinical and laboratory medicine. Berlin: Springer-Verlag, 1992; p. 428-52.133. Chau NP, Mallion JM, Gaudemaris Rd, Ruche E,Siche JP, Pelen O, Mathern G. Twenty-four-hourambulatory blood pressure in shift workers.Circulation 1989;80:341-7.134. Goto T, Yokoyama K, Araki T, Miura T, Saitoh H,Saitoh M, atoh S. Identical blood pressure levelsand slower heart rates among nurses during nightwork and day work. Journal of HumanHypertension 1994;8:11-4.135. Yamasaki F, Schwartz JE, Gerber LM, Warren K,Pickering TG. Impact of shift work andrace/ethnicity on the diurnal rhythm of bloodpressure and catecholamines. Hypertension1998;32:417-23.References136. Kobayashi F, Watanabe T, Tanaka T, Nakagawa T.Effect of shift work and time of day on bloodpressure and heart rate in normotensive andhypertensive workers. Journal of OccupationalMedicine, Singapore 1992;4:58-63.137. Baumgart P, Walger P, Fuchs G, Eiff MV, VetterH, Rahn KH. Diurnal variations of blood pressurein shift workers during day and night shifts. IntArch Occup Environ Health 1989;61:463-6.138. Motohashi Y, Higuchi S, Maeda A, Liu Y, YuasaT, Motohashi K, Nakamura K. Alteration ofcircadian time structure of blood pressure causedby night shift schedule. Occup Med 1998;48:523-8.139. Sundberg S, Kohvakka A, Gordin A. Rapidreversal of circadian blood pressure rhythm in shiftworkers. J Hypertension 1988;6:393-6.140. Rivera-Coll A, Funtes-Arderiu L, Diez-Noguera A.Circadian rhythmic variation in serumconcentrations of clinically important lipids. ClinChem 1994;40:1549-53.141. Haus E, Touitou Y. Chronobiology in laboratorymedicine. In: Touitou Y, Haus E (eds). Biologicrhythm in clinical and laboratory medicine. Berlin:Springer Verlag, 1992; p. 673-708.142. Decousus H. Chronobiology in hemostastis. In:Touitou Y, Haus E (eds). Biologic rhythm inclinical and laboratory medicine. Berlin: SpringerVerlag, 1992; p. 555-65.143. Romon M, Le Fur C, Lebel P, Edmé J-L, FruchartJ-C, Dallongeville J. Circadian variation inpostprandial lipemia. Am J Clin Nutr 1997;65:934-40.144. Hampton SM, Morgan LM, lawrence N,Anastasiadou T, Norris F, Diacon S, Ribeiro D,Arendt J. Postprandial hormone and metabolicresponses in simulated shift work. J Endocrinol1996;151(2):259-67.145. Ribeiro D, Hampton SM, Morgan L, Deacon S,Arendt J. Altered postprandial hormone andmetabolic responses in a simulated shift workenvironment. J Endocrinol 1998;158:305-10.146. Lennernäs M, Åkerstedt T, Hambræus L.Nocturnal eating and serum cholesterol of threeshiftworkers. Scand J Work Environ Health1994;20:401-6.147. Knutsson A, Andersson H, Berglund U. Serumlipoproteins in day and shift workers: a prospectivestudy. Br J Ind Med 1990;47:132-4.69

Shift work and heart disease148. Rosa RR, Bonnet MH, Bootzin RR, Eastman CI,Monk TH, Penn PE, et al. Intervention factors forpromoting adjustment to nightwork and shiftwork.In: Scott AJ, Ladou J (eds). OccupationalMedicine: State of the art reviews. Philadelphia:Hanley & Belfus, Inc., 1990; p. 391-414.149. Monk TH. Coping with the stress of shift work.Work Stress 1988;2:169-72.150. Scott AJ, Ladou J. Shiftwork: effects on sleep andhealth with recommendations for medicalsurveillance and screening. In: Scott AJ, Ladou J(eds). Occupational Medicine: State of the artreviews. Philadelphia: Hanley & Belfus, Inc.,1990; p. 273-99.151. Singer B, Terborg J, Mayer S. Attitudinal,circadian, circumstantial, and subject selectionexplanations of shiftwork effects on health. JOccup Med 1994;36:66-9.152. Borjanovic S. The influence of shift work onvariability of physiological functions. Arbete ochhälsa 27, 1985.153. Frese M, Semmer N. Shiftwork, stress, andpsychosomatic complaints: a comparison betweenworkers in different shiftwork schedules, nonshiftworkers,and former shiftworkers. Ergonomics1986;29:99-114.154. Smolensky MH. The chronoepidemiology ofoccupational health and shift work. In: Reinberg A,Vieux N, Andlauer P (eds). Night and shift work;biological and social aspects. Oxford: Pergamonpress, 1980; p. 51-65.155. Knutsson A, Nilsson T. Job strain in shift anddaytime workers. Int J Occ Environ Health 1997;3(suppl 2):s78-s81156. Peter R, Alfredsson L, Knutsson A, Siegrist J,Westerholm P. Does a stressful psychosocial workenvironment mediate the effects of shift work oncardiovascular risk factors? Scand J Work EnvironHealth 1999;25:376-81.157. Cesana G, Panza G, Ferrario M, Zanettini R,Arnoldi M, Grieco A. Can glycosylated hemoglobinbe a job stress parameter? J Occup Med1985;27:357-60.158. Morikawa Y, Nakagawa H, Katsuyuki M, IshizakiM, Tabata M, Nishiijo M, Higashiguchi K, YoshitaK, Sagara T, Kido T, et al. Relationship betweenshift work and onset of hypertension in a cohort ofmanual workers. Scand J Work Environ Health1999;25:100-4.References159. Karlsson B, Knutsson A. Shiftwork and metabolicsyndrome in the Wolf-study. [Abstract] ShiftworkInternational Newsletter 1999;16:75160. Murata K, Yano E, Shinozaki T. Cardiovasculardysfunction due to shift work. JOEM 1999;41:748-53.161. Murata K, Yano E, Shinozaki T. Impact of shiftwork on cardiovascular functions in a 10-yearfollow-up study. Scand J Work Environ Health1999;25:272-7.162. Costa G, Lievore F, Casaletti G, Gaffuri E, FolkardS. Circadian characteristics influencinginterindividual differences in tolerance andadjustment to shiftwork. Ergonomics 1989;32:373-85.163. Nachreiner F. Individual and social determinants ofshiftwork tolerance. Scand J Work Environ Health1998;24 (suppl 3):35-42.164. Härmä M. Ageing, physical fitmess and shiftworktolerance. Appl Ergon 1996;27:25-9.165. Kristensen TS. Måling af stress ogstressbelastninger i epidemiologiske undersøgelser.Ugeskr Læger 1993;155:1115-20.166. Williams C. Social factors, work, stress andcardiovascular disease prevention. Brussels:European Heart Network, 1998.167. Morehouse RL. Shiftwork: the special challengesfor women. AAOHN Journal 1995;43:532-5.168. Wedderburn A. Women and nightwork. Bulletin ofeuropean shiftwork topics (BEST). Dublin:European foundation for the improvement of livingand working conditions, 1990.169. Folketinget. Lovbekendtgørelse nr 497 af 29. Juni1998: bekendtgørelse af lov om arbejdsmiljø.170. Rådet for de europæiske fællesskaber. Direktivetom visse aspekter i forbindelse medtilrettelæggelse af arbejdstiden (direktiv 93/104/EFaf 23. november 1993). Bruxelles: EF, 1993.171. Wilkinson RT. How fast should the night shiftrotate? Ergonomics 1992;35:1425-46.172. Folkard S. Is there a "best compromise" shiftsystem? Ergonomics 1992;35:1453-63.173. Wedderburn A. How fast should the night shiftrotate? A rejoinder. Ergonomics 1992;35:1447-51.70

Shift work and heart disease174. Rosa RR, Colligan MJ. Plain language aboutshiftwork. Cincinnati: US department of health andhuman services, Public health service, Centers fordisease control and prevention, National institutefor occupational safety and health, 1997.175. Barton J, Spelten E, Totterdell P, Smith L, FolkardS. Is there an optimum number of night shifts?Relationship between sleep, health and well-being.Work Stress 1995;9:109-23.176. Barton J. Chosing to work at night: a moderatinginfluence on individual tolerance to shift work. JAppl Psych 1994;79:449-54.177. Tüchsen F. Working hours and ischaemic heartdiseases in Danish men; A 4-year cohort study ofhospitalization. Int J Epidemiol 1993;22:215-21.178. Knauth P. Designing better shift systems. ApplErgon 1996;27:39-44.179. Knauth P. Design of shiftwork systems. In:Colquhoun WP, Costa G, Folkard S, Knauth P(eds). Shiftwork: problems and solutions. Frankfurtam Main: Peter Lang GmbH, 1996; p. 155-73.180. Knauth P. Changing schedules: shiftwork.Chronobiol Int 1997;14:159-71.181. Wedderburn A. Guidelines for shiftworkers.Bulletin of european shiftwork topics (BEST).Dublin: European foundation for the improvementof living and working conditions, 1991.182. Wedderburn A. Instruments for designing,implementing and assessing working timearrangements. Bulletin of european shiftworktopics (BEST). Dublin: European Foundation forthe Improvement of Living and WorkingConditions, 1994.183. Schönfelder E, Knauth P. A procedure to asessshift systems based on ergonomic criteria.Ergonomics 1993;36:65-76.184. Nachreiner F, Qin L, Grzech-Sukalo H, Hedden I.Computer-aided design of shift schedules.Ergonomics 1993;36:77-83.185. Jansen B, Kroon H. Rota-risk-profile-analysis.Work Stress 1995;9:245-55.186. Goldenhar LM, Schulte PA. Intervention researchin occupational health and safety. JOM1994;36:763-75.187. Lavie P, Tzischinsky O, Epstein R, Zomer J. Sleep-ReferencesWake Cycle in Shift Workers on a "Clockwise" and"Counter-Clockwise" Rotation System. IsraelJournal of Medical Sciences 1992;28:636-44.188. Phillips B, Magan L, Gerhardstein C, Cecil B.Shift work, sleep quality, and worker health: astudy of police officers. South Med J1991;84:1176-84, 1196.189. Williamson AM, Sanderson JW. Changing thespeed of shift rotation: a field study. Ergonomics1986;29:1085-96.190. Barton J, Folkard S, Smith L, Poole C. Effects onhealth of a change from a delaying to an advancingshift system. Occup Environ Med 1994;51:749-55.191. Hornberger S, Knauth P. Follow-up interventionstudy on effects of a change in shift schedule onshiftworkers in the chemical industry. Internationaljournal of industrial ergonomics 1998;21:249-57.192. Kecklund G, Åkerstedt T, Göranson B, SöderbergK. Omläggning av skiftschema: konsekvenser förvälbefinnande, hälsa, sömn/vakenhet ocharbetstrivsel. Resultatrapport 1: frågeformulär ochhälsoundersökning. Stockholm:Stressforskningsrapport nr. 234, 1992.193. Poole C, Evans GR, Spurgeon A, Bridges KW.Effects of a change in shift work on health. OccupMed 1992;42:193-9.194. Knauth P, Kieswetter E. A change from weekly toquicker shift rotations: a field study ofdiscontinuous three shift workers. Ergonomics1987;30:1311-21.195. Knauth P, Hornberger S. Changes from weeklybackwards to quicker forward rotating shiftsystems in the steel industry. International journalof industrial ergonomics 1998;21:267-73.196. Czeisler CA, Moore-Ede MC, Coleman RM.Rotating shift work schedules that disrupt sleep areimproved by applying circadian principles. Science1982;217:460-3.197. Kandolin I, Huida O. Individual flexibility: anessential prerequisite in arranging shift schedulesfor midwives. J Nursing Management 1996;4:213-7.198. Hornberger S, Knauth P. Effects of various types ofchange in shift schedules: a controlled longitudinalstudy. Work Stress 1995;9:124-33.199. Kecklund G, Åkerstedt T, Göranson B, Söderberg71

Shift work and heart diseaseK. Omläggning av skiftschema: konsekvenser förvälbefinnande, hälsa, sömn/vakenhet ocharbetstrivsel. Resultatrapport 2: frågeformulär,dagbok och hälsoundersökning. Stockholm:Stressforskningsrapport no 242, 1994.200. Kitahara K, Sung Kim IL, Yoshizumi K, et al.Survey on the health indices and life style relatedto work systems in manufacturing companies:effects of changed work system on health indices.[Abstract] Shiftwork International Newsletter1999;16:73201. Barton J, Spelten E, Totterdell P, Smith L, FolkardS, Costa G. The Standard Shiftwork Index: abattery of questionnaires for assessing shiftworkrelatedproblems. Work Stress 1995;9:4-30.202. Orth-Gomér K. Cardiovascular disease - factors ofimportance in shift workers. Arbete och hälsa1985;27:57-63.203. Knutsson A, Akerstedt T. The healthy-workereffect: self-selection among Swedish shift workers.Work Stress 1992;6:163-7.204. Thiis-Evensen E. Skiftarbeid og helse - II.Porsgrunn: Norsk Hydro, 1956.205. Åkerstedt T, Knutsson A. Shift work, disease, andepidemiology. In: Haider M, Koller M, Cervinka R(eds). Night and shiftwork: Longterm effects andtheir prevention. Frankfurt: Verlag Peter Lang,1986; p. 403-7.206. Rutenfranz J, Knauth P. Combined effects. In:Haider M, Koller M, Cervinka R (eds). Night andshiftwork: Longterm effects and their prevention.Frankfurt: Verlag Peter Lang, 1986; p. 319-26.207. Knutsson A, Nilsson T. Tobacco use and exposureto environmental tobacco smoke in relation tocertain work characteristics. Scand J Soc Med1998;26:183-9.208. Knutsson A, Åkerstedt T, Jonsson BG. Prevalenceof risk factors for coronary artery disease amongday and shift workers. Scand J Work EnvironHealth 1988;14:317-21.209. Åkerstedt T, Alfredsson L, Theorell T. Arbetstidoch sjukdom - en studie med aggregerade data.Stressforskningsrapport no. 190. Stockholm:Statens institut för psykosocial miljömedicin,1987.210. Jeppesen HJ, Bøggild H. Differences between fulltime and part time employed shift workers withfixed and rotating shifts. [Abstract] FirstReferencesinternational ICOH conference on psychosocialfactors at work, Copenhagen, Denmark august 24-26, 1998 1998;130211. Rothman KJ, Greenland S. Modern Epidemiology.Philadelphia: Lippincott-Raven Publishers, 1998.212. Gordon NP, Cleary PD, Parker CE, Czeisler CA.The prevalence and health impact of shiftwork. AmJ Public Health 1986;76:1225-8.213. Fouriaud C, Jacquinet-Salord MC, Degoulet P,Aimé F, Lang T, Laprunge J, Main J, OeconomosJ, Phalente J, Prades A. Influence ofsocioprofessional conditions on blood pressurelevels and hypertension control. Am J Epidemiol1984;120:72-86.214. Armstrong BK, White E, Saracci R. Principles ofexposure measurement in epidemiology. Oxford:Oxford University Press, 1992.215. Knutsson A, Jonsson BG, Åkerstedt T, Orth-Gomér K. Sjuklighet i ischemisk hjärtsjukdom hosen cohort dag- och skiftarbetare vid ett svensktpappersbruk - stressforskningsrapport nr. 179.Stockholm: Statens institut för psykosocialmiljömedicin, WHOs psykosociale center, 1985.216. Bursey RG. A cardiovascular study of shift workerswith respect to coronary artery disease risk factorprevalence. J Soc Occup Med 1990;40:65-7.217. Cesana G, Finotte S, De Vito G. CHD risk factorsprevalence in middle aged shift workers. In: CostaG, Cesana G, Kogi K, Wedderburn A (eds).Shiftwork: health, sleep and performance.Frankfurt am Main: Peter Lang, 1990; p. 362.218. Costa G, Betta A, Uber D, Alexopoulos C.Estimate of coronary risk in a group of italianshiftworkers. In: Costa G, Cesana G, Kogi K,Wedderburn A (eds). Shiftwork: health, sleep andperformance. Frankfurt am Main: Peter Lang,1990; p. 363-9.219. De Backer G, Kornitzer M, Dramaix M, Peeters H,Kittel F. Irregular working hours and lipid levelsin men. In: Schling G, Mörl H (eds). Expandinghorizons in atherosclerosis research. Berlin:Springer Verlag, 1987; p. 217-24.220. Knutsson A. Relationships between serumtriglycerides and g-glutamyltransferase among shiftand day workers. J Int Med 1989;226:337-9.221. Lasfargues G, Vol S, Cacès E, Clésiau HL,Lecomte P, Tichet J. Relation among night work,dietary habits, biological measures, and health72

Shift work and heart diseasestatus. Int J Behav Med 1996;3:123-34.222. Netterstrøm B, Kristensen TS, Møller L, Jensen G,Schnohr P. Psykisk arbejdsmiljø og helbred.København: Arbejdsmiljøfondet, 1996.223. Peternel P, Stegnar M, Salobir U, Salobir B, KeberD, Vene N. Shift work and circadian rhythm ofblood fibrinolytic parameters. Fibrinolysis 1990;4(suppl):113-5.224. Romon M, Nuttens MC, Fievet C, Pot P, Bard JM,Furon D. Increased triglyceride levels in shiftworkers. Am J Med 1992;93:259-62.225. Thelle DS, Forde OH, Try K, Lehmann EH. TheTromso heart study. Acta Med Scand1976;200:107-18.226. Theorell T, Åkerstedt T. Day and night work:changes in cholesterol, uric acid, glucose andpotassium in serum and in circadian patterns ofurinary catecholamine excretion. Acta Med Scand1976;200:47-53.227. Marckmann P. Diet, blood coagulation andfibrinolysis (disp). København: Lægeforeningensforlag, 1995.228. Kasl SV. Measuring job stressors and studying thehealth impact of the work environment: anepidemiologic commentary. J Occup HealthPsychol 1998;3:390-401.229. Theorell T. Notes on the use of biological markersin the study of long-term stress. In: Carlson JG,Serfert AR, Birbaumer N (eds). Clinical appliedpsychophysiology. New York: Plenum Press, 1994;p. 251-60.230. Temple R. Are surrogate markers adequate toassess cardiovascular disease drugs? JAMA1999;282:790-5.231. McDowell I, Newell C. The theoretical andtechnical foundations of health measurements. In:McDowell I, Newell C (eds). Measuring health: aguide to rating scales and questionnaires. NewYork: Oxford University Press, 1996; p. 10-46.232. Cox T, Ferguson E. Measurement of the subjectivework environment. Work Stress 1994;8:98-109.233. Setterlind S, Larsson G. The stress profile: apsychosocial approach to measuring stress. StressMed 1995;11:85-92.234. Muller JE, Stone PH, Turi ZG, Rutherford JD,ReferencesCzeisler CA, Parker C, Poole WK, Passamani E,Roberts R, Robertson T, et al. Circadian variationin the frequency of onset of acute myocardialinfarction. N Engl J Med 1985;313:1315-22.235. National Heart Lung and Blood Institute WorkingGroup on Sleep Apnea. Sleep apnea: is yourpatient at risk? Am Fam Physician 1996;53:247-53.236. Siegrist J. Sleep disturbances in pre-infarctionpatients. In: Armin T, Maseri A (eds). Predisposingconditions for acute ischemic syndromes.Darmstadt: Steinkopff Verlag, 1989; p. 42-50.237. Hill AB. The environment and disease: associationor causation? Proc R Soc Med 1965;56:295-300.238. Rothman KJ, Greenland S. Causation and causalinference. In: Rothman KJ, Greenland S (eds).Modern epidemiology. Philadelphia: Lippincott-Raven Publishers, 1998; p. 7-28.239. Marmot M. Diet and disease, and Durkheim andDasgupta, and deuteronomy. Epidemiology1998;9:676-80.240. Penev PD, Kolker DE, Zee PC, Turek FW. Chroniccircadian desynchronization decreases the survivalof animals with cardiomyopathic heart disease. AmJ Physiol 1998;275:H2334-7.241. Tenkanen L, Sjöblom T, Härmä M. Joint effect ofshift work and adverse life-style factors on the riskof coronary heart disease. Scand J Work EnvironHealth 1998;24:351-7.242. Fréden K, Olsson IL, Orth-Gomér K, Åkerstedt T.Effekten på sömn, vagbesvär och kardiovaskulärariskfaktorer av övergang från moturs till medursskiftrotation hos poliser. Stockholm: Stressforskningsrappporterno. 175, 1984.243. Koller M. Health risks related to shift work. IntArch Occup Environ Health 1983;53:59-75.244. Lang T, Pariente P, Salem G, Tap D. Social,professional conditions and arterial hypertension:an epidemiological study in Dakar, Senegal. JHypertens 1988;6:271-6.245. Michel-Briand C, Chopard JL, Guiot A, PaulmierM, Studer G. The pathological consequences ofshift work in retired workers. In: Reinberg A,Vieux N, Andlauer P (eds). Night and shift work:biological and social aspects. Oxford: PergamonPress, 1980; p. 399-407.246. Peacock B, Glube R, Miller M, Clune P. Police73

Shift work and heart diseaseofficers' response to 8 and 12 hour shift schedules.Ergonomics 1983;26:479-93.247. Rose G. Sick individuals and sick populations. IntJ Epidemiol 1985;14:32-8.Referencesjournal of industrial ergonomics 1998;21:233-42.249. Büssing A. Social tolerance of working timescheduling in nursing. Work Stress 1996;10:238-50.248. Gissel A, Knauth P. Assessment of shift systems inthe German industry and service sector: a computerapplication of the Besiak procedure. International74