Survival bias and drug interaction can attenuate cross-sectional case-control comparisons of genes with health outcomes. An example of the kinesin-like protein 6 (KIF6) Trp719Arg polymorphism and coronary heart disease

Abstract

Background: Case-control studies typically exclude fatal endpoints from the case set, which we hypothesize will substantially underestimate risk if survival is genotype-dependent. The loss of fatal cases is particularly nontrivial for studies of coronary heart disease (CHD) because of significantly reduced survival (34% one-year fatality following a coronary attack). A case in point is the KIF6 Trp719Arg polymorphism (rs20455). Whereas six prospective studies have shown that carriers of the KIF6 Trp719Arg risk allele have 20% to 50% greater CHD risk than non-carriers, several cross-sectional case-control studies failed to show that carrier status is related to CHD. Computer simulations were therefore employed to assess the impact of the loss of fatal events on gene associations in cross-sectional case-control studies, using KIF6 Trp719Arg as an example.

Results: Ten replicates of 1,000,000 observations each were generated reflecting Canadian demographics. Cardiovascular disease (CVD) risks were assigned by the Framingham equation and events distributed among KIF6 Trp719Arg genotypes according to published prospective studies. Logistic regression analysis was used to estimate odds ratios between KIF6 genotypes. Results were examined for 33%, 41.5%, and 50% fatality rates for incident CVD.In the absence of any difference in percent fatalities between genotypes, the odds ratios (carriers vs. noncarriers) were unaffected by survival bias, otherwise the odds ratios were increasingly attenuated as the disparity between fatality rates increased between genotypes. Additional simulations demonstrated that statin usage, shown in four clinical trials to substantially reduce the excess CHD risk in the KIF6 719Arg variant, should also attenuate the KIF6 719Arg odds ratio in case-control studies.

Conclusions: These computer simulations show that exclusions of prior CHD fatalities attenuate odds ratios of case-control studies in proportion to the difference in the percent fatalities between genotypes. Disproportionate CHD survival for KIF6 Trip719Arg carriers is suggested by their 50% greater risk for recurrent myocardial infarction. This, and the attenuation of KIF6 719Arg carrier risk with statin use, may explain the genotype's weak association with CHD in cross-sectional case-control studies. The results may be relevant to the underestimation of risk in cross-sectional case-control studies of other genetic CHD-risk factors affecting survival.

Figure 1

Simulated effects of genotype-differences in fatalities on the odds ratio for CVD in KIF6 719Arg carriers vs. noncarriers. KIF6 719Arg carriers are assumed to represent 59% of the population and to increase CVD risk by 50% (consistent with WOSCOPS and CARE trials [6]) and by 25% (consistent with ARIC and WHS [11,13]) relative to noncarriers. X-axis represents the percent difference in fatality between carriers and noncarriers of the KIF6 allele (see methods).

Figure 2

Simulated effects of statin use on the odds ratio for CHD in KIF6 719Arg carriers vs. noncarriers. Results are presented where KIF6 719Arg carriers have 50% or 25% increased risk of CHD compared with noncarriers if there is no statin use, and 0% increased risk of CHD compared with noncarriers if statins are used. Odds ratios (vertical axis) are presented for increasing percentage of statin use (horizontal axis).

Figure 3

Simulated effects of statin use on the odds ratio for CHD in KIF6 719Arg carriers vs. noncarriers. for varying effects of statin on KIF6 719Arg carriers risk. Results are presented where KIF6 719Arg carriers have 50% increased risk of CHD compared with noncarriers if there is no statin use, and where statin use eliminates 20%, 40%, 60%, 80% and 100% of the excess risk associated with KIF6 719Arg carrier status (Figure 2 assumed 100% of the excess risk was eliminated by statin use). Odds ratios (vertical axis) are presented for increasing percentage of statin use (horizontal axis).