Scientific Concept of the National Cohort (status ... - Nationale Kohorte

A.6 Planned statistical analyses and statistical power considerations
The study sizes and targeted age distributions for the overall cohort (N=200,000) and a
20% subcohort with intensified phenotyping (N=40,000) are motivated mainly by statistical
power calculations and criteria for the statistical detection of minimally relevant quantitative
associations of risk for frequent chronic diseases and premature mortality with their most
common causes (see Sect. A.6.4).
For the more frequent forms of cancer, estimates of expected incident events per disease
type were obtained by applying age- and sex-specific incidence values from German cancer
registries to the projected age distribution of the cohort (Table 6.3), adjusting for progressive
attrition of the cohort as a result of overall mortality. Likewise, for MI and diabetes
estimates of cumulative disease incidence were obtained, using age-specific incidence as
ascertained in the MONICA/KORA registry for MIs in Augsburg as well as in the populationbased
MONICA/KORA cohort study. For stroke, age-specific incidence from the populationbased
registries of the Erlangen Stroke Project and the Ludwigshafen Stroke Study were
used. As for the German population no incidence data for rheumatoid arthritis and COPD
were available, data on age-specific incidence in the UK were used (Table 6.4).
In these various estimations of expected cumulative disease incidence, simplifying assumptions
were made, in that no account was made of attrition of the cohort that will result from
subjects’ future requests for withdrawal from the study or losses to follow-up due to other
causes. These simplifications would be expected to cause some degree of overestimation
of cumulative cancer incidence. However, based on experience from ongoing studies, we
can anticipate low active withdrawal rates of study participants and very high case ascertainment
rates through record linkage with cancer registries or active follow-up for the other
chronic diseases. Thus, the above factors are unlikely to cause more than a 5–10% overes-
Table 6.3: Expected counts of incident cancer cases after 5, 10, 15, or 20 years of average
timation follow-up, of over for the the overall first 10- cohort to 15-year (N=200,000) follow-up or for the period. intensified subcohort (N=40,000).*
Table 6.3: Expected counts of incident cancer cases after 5, 10, 15, or 20 years of average followup,
for the overall cohort (N=200,000) or for the intensified subcohort (N=40,000).*
Expected cumulative incidence
at study level 1
(200,000 subjects)
177
Expected cumulative incidence
at study level 2
(40,000 subjects)
Disease
Average follow-up duration (years)
(+ expected corresponding calendar date)
5 yrs 10 yrs 15 yrs 20 yrs 5 yrs 10 yrs 15 yrs 20 yrs
(2022) (2027) (2032) (2037) (2022) (2027) (2032) (2037)
Any cancer 5,100 13,000 21,000 29,000 1,000 2,600 4,000 6,000
Breast 780 1,800 2,900 4,000 160 370 590 800
Prostate 720 1,900 3,200 4,600 140 380 640 910
Colon,
Rectum
670 1,800 3,100 4,500 130 360 620 890
Lung 560 1,400 2,400 3,400 110 290 480 680
Bladder 260 710 1,200 1,800 50 140 250 360
Kidney 190 500 850 1,200 40 100 170 240
non-
Hodgkin L.
140 340 580 820 30 70 120 160
Pancreas 120 330 580 830 20 70 120 170
Corpus
Uteri
120 320 540 770 20 60 110 200
Brain+CNS 90 200 330 450 20 40 70 90
Ovary 110 260 440 610 20 50 90 120
Calculations based on age-specific incidence from German Cancer Registry 809
A.6