Limitations of Genetic Findings That Are Not in Hardy-Weinberg Equilibrium

Abstract

We thank Drs. Body and Schwinn for their critical remarks on our study.1To clarify, our double-blinded, placebo-controlled trial was primarily designed to test the hypothesis that the perioperative administration of bisoprolol would reduce the incidence of cardiovascular complications in patients with or at risk for coronary artery disease undergoing surgery with spinal nerve block. However, because single nucleotide polymorphisms of the β-adrenergic receptor genes may act as disease modifiers,2biologically important nonsynonymous coding variants of the β-adrenergic receptor were determined in our study population. This analysis showed that carriers of at least one Gly allele of the β1-adrenergic receptor polymorphism Arg389Gly experienced a higher number of adverse events than Arg homozygotes.1Drs. Body and Schwinn raised concerns with respect to genotyping errors based on the observed deviation from Hardy-Weinberg equilibrium for this particular polymorphism in our trial. Genotyping error is indeed one of the many possible sources of Hardy-Weinberg disequilibrium. To exclude this possibility in our trial, we have carefully regenotyped all patients for the Arg389Gly polymorphism using internal controls for wild-type, heterozygous, and mutant genotypes for each amplification process. As reported in our publication, from the 189 patients, who consented to genotyping, 186 could be unequivocally identified and confirmed by regenotyping by three independent individuals. In three patients (notably without a primary outcome), genotyping was not possible, and these patients were excluded from genotype-related outcome analysis. Our genotyping platform was further meticulously validated by bidirectional sequencing of DNA samples for the Arg389Gly polymorphism from 12 randomly selected patients of this particular study and from many other patients not related to this study. Bidirectional sequencing is regarded as the standard of genotyping and as by far more reliable than any other genotyping platform.Although testing for Hardy-Weinberg equilibrium is used as some quality-control measure, particularly in case–control gene association studies, it cannot be used to detect genotyping error.3,4Genotyping errors are generally small and do not generate sufficient deviations from Hardy-Weinberg equilibrium to be detected. In the case of a reduced number of observed heterozygous patients, as may occur in the presence of poor amplification of one of the alleles, large samples sizes are necessary to detect deviations from Hardy-Weinberg equilibrium. For example, with the Gly389 allele of the Arg389Gly polymorphism with a reported allele frequency of 0.27 and an error rate of 0.05, more than 8,000 patients would be necessary to detect deviation from Hardy-Weinberg with a power of 0.80 at an α level of 0.05. Increasing the error rate to 0.15 reduces the sample size to the still considerably high patient number of 944. Therefore, testing for Hardy-Weinberg equilibrium is an unreliable tool to identify genotyping errors. Conversely, the presence of Hardy-Weinberg equilibrium does not rule out that genotyping errors might have occurred. Hence, it seems unlikely that genotyping error is the source of the Hardy-Weinberg disequilibrium observed in our study. Because approximately 10% of all genotype–phenotype association studies show deviation from Hardy-Weinberg equilibrium, the results of our trial cannot be considered “abnormal.”5Rather, as outlined in the discussion of our findings,1a selection bias (population stratification) may have occurred because of inclusion and exclusion criteria of this randomized trial. A mortality bias (different survival of marker genes) due to varying genetic and environmental background (e.g. , response to cardiovascular medication) in this elderly study population at the end of life expectancy may have also caused this disequilibrium. Of note, Hardy-Weinberg disequilibrium that is caused by most interesting biologic phenomena typically results in excess homozygosity, as observed in our study.1However, we agree with Drs. Body and Schwinn that violation of Hardy-Weinberg equilibrium in our study population implies a selected rather than a random sample, invalidating direct comparisons with other populations. Therefore, we share their view that our results should be regarded with caution. Our findings should be confirmed in future prospective larger-scale clinical trials, specifically designed and adequately powered to detect genotype-specific differences in cardiovascular outcome in patients with or at risk of coronary artery disease.*University Hospital Zurich, Zurich, Switzerland. michael.zaugg@usz.ch