10.12.2012 Views

Understanding the Fundamentals of Epidemiology an evolving text

Understanding the Fundamentals of Epidemiology an evolving text

Understanding the Fundamentals of Epidemiology an evolving text

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

(I1 - I0)/I0 = (0.01200-0.00800 / 0.00800) = 0.50. On this basis, <strong>the</strong> association involving vitamin C<br />

is subst<strong>an</strong>tially greater th<strong>an</strong> that involving vitamin E. This relative difference measure is <strong>of</strong>ten called<br />

<strong>the</strong> excess risk (or "excess rate", since <strong>the</strong> data are rates, not proportions). If we add 1.0 to <strong>the</strong><br />

excess risk or rate, we obtain <strong>an</strong> even simpler relative measure, I1/I0, which is variously termed<br />

relative risk, risk ratio, rate ratio, cumulative incidence ratio, incidence density ratio, or, for<br />

prevalences, prevalence ratio.<br />

Here are two real-life examples that contrast relative <strong>an</strong>d absolute measures <strong>of</strong> association. The first<br />

is based on data from a follow-up study by M<strong>an</strong>n et al. (presented in a seminar at UNC by Bruce<br />

Stadel):<br />

Relative versus Absolute Measures <strong>of</strong> Association Incidence <strong>of</strong> myocardial infarction (MI)<br />

in oral contraceptive (OC) users per 100,000 women-years<br />

Age (years) Cigarettes/day<br />

Oral<br />

contraceptive<br />

users<br />

Non-users RR** AR***<br />

30-39 0-14 6 2 3 4<br />

15 + 30 11 3 19<br />

40-44 0-14 47 12 4 35<br />

Notes:<br />

15 + 246 61 4 185<br />

* RR=relative risk (rate ratio)<br />

** AR=attributable risk (rate difference, absolute difference)<br />

In this table, <strong>the</strong> incidence <strong>of</strong> MI is clearly greater for OC users, since in each age-smoking stratum<br />

<strong>the</strong> OC users have a higher incidence (ID) th<strong>an</strong> do <strong>the</strong> nonusers. Moreover, <strong>the</strong> ratio <strong>of</strong> <strong>the</strong> two<br />

incidences (<strong>the</strong> RR) is nearly const<strong>an</strong>t across strata, a desirable property for a summary measure,<br />

whereas <strong>the</strong> rate difference (AR) varies widely. According to Breslow <strong>an</strong>d Day, <strong>the</strong> rate ratio tends<br />

to be more stable across strata, supporting its desirability as a measure <strong>of</strong> association. Not all<br />

qu<strong>an</strong>titative epidemiologists agree with <strong>the</strong>m.<br />

The second example comes from a follow-up study <strong>of</strong> lung c<strong>an</strong>cer <strong>an</strong>d coronary artery disease in<br />

relation to cigarette smoking:<br />

_____________________________________________________________________________________________<br />

www.epidemiolog.net, © Victor J. Schoenbach 2000 7. Relating risk factors to health - 167<br />

rev. 7/25/1999, 9/4/1999, 12/22/1999, 10/2/2000, 10/9/2000, 5/8/2001

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!