08.12.2012 Views

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

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

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

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.

2000<br />

1000<br />

0<br />

0.05 0.1 0.15A.6 Planned 0.2 statistical 0.25 analyses 0.3 and 0.35 statistical 0.4power considerations<br />

0.45 0.5<br />

Allele frequency<br />

figure 6.5: Study size (number <strong>of</strong> incident cases <strong>of</strong> disease) required to detect minimal odds<br />

RGE 1.2 1.3 1.5 1.7 2 2.5<br />

ratios [MDOR] <strong>of</strong> various magnitudes, for gene–environment interaction with power 0.80, at significance<br />

level *assuming <strong>of</strong> 0.05; main nested effects case-control odds ratios <strong>of</strong> 1.2 study for genetic with 4 and controls 2.0 for per non-genetic case*, assuming risk factor main effects<br />

odds ratios <strong>of</strong> 1.2 for <strong>the</strong> alpha=0.05, genetic and prevalence 2.0 for <strong>of</strong> <strong>the</strong> non-genetic nongenetic risk factor risk 20% factor, plus a prevalence <strong>of</strong> <strong>the</strong><br />

nongenetic risk factor <strong>of</strong> 0.20<br />

N cases<br />

10000<br />

9000<br />

8000<br />

7000<br />

6000<br />

5000<br />

4000<br />

3000<br />

2000<br />

1000<br />

0<br />

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5<br />

Allele frequency<br />

RGE 1.2 1.3 1.5 1.7 2 2.5<br />

*assuming main effects odds ratios <strong>of</strong> 1.2 for genetic and 2.0 for non-genetic risk factor<br />

alpha=0.05, prevalence <strong>of</strong> non-genetic risk factor 20%<br />

A.6.4.4 Effects <strong>of</strong> random measurement errors and use <strong>of</strong> repeat<br />

measurements<br />

The degree <strong>of</strong> correlation between exposures in <strong>the</strong> two time periods, ρ m1,m2 , is an important<br />

determinant <strong>of</strong> <strong>the</strong> power <strong>of</strong> possible tests <strong>of</strong> “history-adjusted” exposure associations<br />

with disease risk. For example, when correlation ρ m1,m2 is as high as 0.80, a test whe<strong>the</strong>r<br />

risk is related to exposure in <strong>the</strong> second index period, adjusted for exposure at <strong>the</strong> earlier<br />

period (i.e., testing <strong>the</strong> null hypo<strong>the</strong>sis that β 2 = 0, conditionally on exposure level x 1 ) will<br />

have a high VIF (VIF = 2.8), and a corresponding reduction in statistical power. Likewise,<br />

a comparatively high correlation ρ m1,m2 will also limit possible power gains that could be<br />

achieved by basing <strong>the</strong> test <strong>of</strong> exposure–disease association on a weighted average <strong>of</strong><br />

<strong>the</strong> exposures (i.e., testing <strong>the</strong> null hypo<strong>the</strong>sis that β L = 0). For exposures (e.g., blood- or<br />

urine-based markers) that have a correlation <strong>of</strong> ρ m1,m2 = 0.30–0.70 between replicate measurements<br />

taken over time, a <strong>the</strong>oretical (maximal) average gain in effective sample size <strong>of</strong><br />

about 40–60% can be calculated, under <strong>the</strong> assumption that random errors <strong>of</strong> <strong>the</strong> replicate<br />

measurements are not correlated and that disease risk is equally associated with <strong>the</strong> exposure<br />

measurements at ei<strong>the</strong>r time point (see Annex C.2.4).<br />

figure 6.6 illustrates both relative and absolute gains in statistical power that can be obtained<br />

by (equally) averaging exposures from two time periods. This specific example focuses<br />

on kidney cancer, showing how statistical power will increase with increasing numbers<br />

<strong>of</strong> incident cases until <strong>the</strong> year 2032. The assumed strength <strong>of</strong> association is that <strong>of</strong><br />

185<br />

A.6

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

Saved successfully!

Ooh no, something went wrong!