Xiao R, Boehnke M. Quantifying and correcting for the winner's curse in genetic association studies. Genet Epidemiol 33: 453-462

Department of Biostatistics and Center for Statistical Genetics, University of Michigan, USA.
Genetic Epidemiology (Impact Factor: 2.6). 07/2009; 33(5):453-62. DOI: 10.1002/gepi.20398
Source: PubMed


Genetic association studies are a powerful tool to detect genetic variants that predispose to human disease. Once an associated variant is identified, investigators are also interested in estimating the effect of the identified variant on disease risk. Estimates of the genetic effect based on new association findings tend to be upwardly biased due to a phenomenon known as the "winner's curse." Overestimation of genetic effect size in initial studies may cause follow-up studies to be underpowered and so to fail. In this paper, we quantify the impact of the winner's curse on the allele frequency difference and odds ratio estimators for one- and two-stage case-control association studies. We then propose an ascertainment-corrected maximum likelihood method to reduce the bias of these estimators. We show that overestimation of the genetic effect by the uncorrected estimator decreases as the power of the association study increases and that the ascertainment-corrected method reduces absolute bias and mean square error unless power to detect association is high.

6 Reads
  • Source
    • "Thus, even conservative estimates of the sample size needed for tests of replication based on the lower bound of the original effect size estimate may be too small. Statistical methods have been developed to improve sample size estimates under conditions of expected shrinking effect sizes, however (Xiao and Boehnke 2009). "

    Journal of Abnormal Child Psychology 04/2013; 41(4). DOI:10.1007/s10802-013-9741-0 · 3.09 Impact Factor
  • Source
    • "Therefore, detection of an association would require a larger sample size, regardless of disease prevalence [87]. Most candidate gene studies for IS had small cohorts, so it is difficult to discern whether associations were not replicated in other cohorts (same or different ethnic background) as a consequence of genetic heterogeneity or ascertainment bias (e.g., “winner’s curse”) [88] (Fig. 2). Based on estimations from Hattersley and McCarthy [89], a study needs thousands of individuals to detect a common variant of a risk allele with a low-to-moderate effect. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Idiopathic scoliosis is a complex developmental syndrome defined by an abnormal structural curvature of the spine. High treatment costs, chronic pain/discomfort, and the need for monitoring at-risk individuals contribute to the global healthcare burden of this musculoskeletal disease. Although many studies have endeavored to identify underlying genes, little progress has been made in understanding the etiopathogenesis. The objective of this comprehensive review was to summarize genetic associations/linkages with idiopathic scoliosis, as well as explore the strengths and weaknesses of each study, such that it may serve as a guide for the design and interpretation of future genetic studies in scoliosis. We searched PubMed and Human Genome Epidemiology (HuGE) Navigator using the search terms "gene and scoliosis". Linkage or association studies published in English and available full-text were further analyzed as regards results, experimental design, and statistical approach. We identified and analyzed 50 studies matching our criteria. These consisted of 34 candidate gene studies (6 linkage, 28 association) and 16 genome-wide studies [14 pedigree-based linkage, 2 genome-wide association studies (GWAS)]. Findings involved genes related to connective tissue structure, bone formation/metabolism, melatonin signaling pathways, puberty and growth, and axon guidance pathways. Variability in results between studies suggested ethnic and/or genetic heterogeneity. The major difficulty in idiopathic scoliosis research is phenotypic and genetic heterogeneity. Genetic research was overrepresented by underpowered studies. The use of biological endophenotypes, as well as restricted clinical definitions, may help to partition variation and increase the power of studies to detect or confirm an effect.
    European Spine Journal 06/2012; 21(10):1905-19. DOI:10.1007/s00586-012-2389-6 · 2.07 Impact Factor
  • Source
    • "While they are relatively easy to perform, interpretation of the results of successful association studies is neither straightforward nor always replicable. In addition, overestimation of genetic effect size in initial studies due to "winner's curse" may cause follow-up studies to be underpowered and so to fail [25]. Results of the current study and others do not replicate previously identified hypertension-related SNPs in GWAS. "
    [Show abstract] [Hide abstract]
    ABSTRACT: A recent genome wide association study in 1017 African Americans identified several single nucleotide polymorphisms that reached genome-wide significance for systolic blood pressure. We attempted to replicate these findings in an independent sample of 2474 unrelated African Americans in the Milwaukee metropolitan area; 53% were women and 47% were hypertensives. We evaluated sixteen top associated SNPs from the above genome wide association study for hypertension as a binary trait or blood pressure as a continuous trait. In addition, we evaluated eight single nucleotide polymorphisms located in two genes (STK-39 and CDH-13) found to be associated with systolic and diastolic blood pressures by other genome wide association studies in European and Amish populations. TaqMan MGB-based chemistry with fluorescent probes was used for genotyping. We had an adequate sample size (80% power) to detect an effect size of 1.2-2.0 for all the single nucleotide polymorphisms for hypertension as a binary trait, and 1% variance in blood pressure as a continuous trait. Quantitative trait analyses were performed both by excluding and also by including subjects on anti-hypertensive therapy (after adjustments were made for anti-hypertensive medications). For all 24 SNPs, no statistically significant differences were noted in the minor allele frequencies between cases and controls. One SNP (rs2146204) showed borderline association (p = 0.006) with hypertension status using recessive model and systolic blood pressure (p = 0.02), but was not significant after adjusting for multiple comparisons. In quantitative trait analyses, among normotensives only, rs12748299 was associated with SBP (p = 0.002). In addition, several nominally significant associations were noted with SBP and DBP among normotensives but none were statistically significant. This study highlights the importance of replication to confirm the validity of genome wide association study results.
    BMC Medical Genetics 04/2012; 13(1):27. DOI:10.1186/1471-2350-13-27 · 2.08 Impact Factor
Show more


6 Reads
Available from