Genetics
Is There a “PRS Paradox” In Prostate Cancer Survival Studies? – Investigating Potential Bias in Genetic Studies of Prostate Cancer Survival Anqi Wang* Anqi Wang Konrad H. Stopsack Lorelei A. Mucci Anna Plym
Background: Polygenic Risk Score (PRS) has been shown to robustly stratify the overall prostate cancer risk, with a stronger association with aggressive versus non-aggressive diseases. In contrast, case-only studies indicate that higher PRS is associated with a lower to neutral risk of prostate cancer death, implying potential biases. Two underlying biases should be considered: 1) collider-stratification, which occurs when conditioning on case status and may skew the distribution of unmeasured confounders; and 2) lead time, resulting from potential earlier detection of cancer in men with higher PRS due to effects on prostate specific antigen (PSA) levels.
Methods: Simulations with a hypothetical cohort of 100,000 cancer-free men were used to evaluate the aforementioned biases. Absolute risks of cancer incidence and mortality by PRS were incorporated in the simulations, derived from prior prostate cancer GWAS studies.
Results: In the presence of collider bias, the Relative Risk (RR) for the PRS on the risk of prostate cancer death tends to be underestimated in case-only studies. We show that in instances of a null observed association (RR=1.0), the true RR could be at least 1.3, assuming no other sources of bias. The magnitude of collider bias is positively correlated with the RRs of unmeasured confounders on prostate cancer incidence and mortality (Figure). In the context of lead time bias, our simulation demonstrates that earlier diagnosis in men with higher PRS artificially extends perceived survival time. For example, detection 2.5 years earlier than average could reduce a true RR of 1.1 to an observed RR of 1.0.
Conclusion: Our simulations show that the observed inverse association of PRS with prostate cancer survival may be attributable to bias. Given its potential role in determining prostate cancer prognosis, it is crucial to distinguish the true PRS impact on prostate cancer prognosis in case-only studies.
