Impact of Selection Bias on Estimation of Subsequent Event Risk

Abstract

Background Studies of recurrent or subsequent disease events may be susceptible to bias due to selection of subjects who both experience and survive the primary indexing event. Currently, the magnitude of any selection bias, particularly for subsequent time-to-event analysis in genetic association studies, is unknown. Methods and Results We used empirically inspired simulation studies to explore the impact of selection bias on the marginal hazard ratio (HR) for risk of subsequent events among those with established coronary heart disease (CHD). The extent of selection bias was determined by the magnitudes of genetic and non-genetic effects on the indexing (first) CHD event. Unless the genetic HR was unrealistically large (> 1.6 per allele) and assuming the sum of all non-genetic HRs was less then 10, bias was usually less than 10% (downward towards the null). Despite the low bias, the probability that a confidence interval included the true effect decreased (undercoverage) with increasing sample size due to increasing precision. Importantly, false positive rates were not affected by selection bias. Conclusions In most empirical settings, selection bias is expected to have a limited impact on genetic effect estimates of subsequent event risk. Nevertheless, due to undercoverage increasing with sample size, most confidence intervals will be over precise (not wide enough). When there is no effect modification by history of CHD, the false positive rates of association tests will be close to nominal.

Adjustment for index event bias in genome-wide association studies of subsequent events

“…Selection into the study depends both on having an initial diagnosis and on surviving initial coronary events, so factors that influence survival may also confound analysis of prognosis. In a simulation study 15 , we showed that index event bias could be small in the GWAS context. Here we have confirmed this in the case of Crohn's disease, but have…”

“…Such approaches are difficult in genetic studies, because there may be a substantial polygenic confounder that cannot be modelled directly nor easily captured by a propensity score 11 . Recently, the implications of index event bias for genetic studies have been discussed in the contexts of association discovery 15 and Mendelian randomisation 2 . Although the magnitude of bias appears small in currently typical settings, it is unclear how GWAS will be affected as meta-analyses grow to greater orders of magnitude and polygenic analyses combine effects over thousands of variants.…”

Phenome-wide association analysis of LDL-cholesterol lowering genetic variants in PCSK9

“…Proteins influencing the risk of a first event may also influence the risk of subsequent events, as observed in the case of the target of statin drugs that are effective in both primary and secondary prevention (1). For this and other reasons (34)(35)(36), examination of the effects of PCSK9 variants on the risk of subsequent CHD events in patients with established coronary atherosclerosis is the subject of a separate analysis led by the GENIUS-CHD consortium (36).…”

Biases in GWAS – the dog that did not bark

“…Genetic randomization on genotype at conception means many years' worth of selective survival could occur before recruitment [4]. The possibility of bias in GWAS arising from survival on genotype, has been considered and is thought, from simulation, to have little effect on estimates of associations [5] before the age of 75 years [6].…”