Curses--winner's and otherwise--in genetic epidemiology

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Initial genetic association studies often overestimate marker allele effects due to stringent significance thresholds, a phenomenon known as the 'winner's curse'.
• Genome-wide association studies (GWAS) are particularly susceptible, as markers must meet high significance levels for replication.
• Overfitting occurs when the same dataset is used for both model selection and accuracy estimation, leading to inflated predictive accuracy.

Discussion:

• Researchers may overstate findings by focusing on relative risks when sensitivity and specificity are more relevant measures.
• The 'winner's curse' and overfitting represent quantitative biases that can distort genetic research findings.
• Qualitative overstatement of findings can also mislead interpretation, even without quantitative biases.

Key Insights:

• The 'winner's curse' leads to exaggerated effect sizes for genetic markers in initial discovery studies.
• Overfitting inflates the estimated predictive accuracy of genetic risk models.
• Awareness of these biases is crucial for both researchers and readers of genetic association studies.

Outlook:

• Epidemiologists must be vigilant in identifying and mitigating these biases in their own work and in published literature.
• Developing robust statistical methods to correct for winner's curse and overfitting is essential for reliable genetic epidemiology.
• Promoting transparent reporting of methods and results will help minimize overstatement and improve the reproducibility of genetic findings.