A Comparison of Bayesian and Frequentist Approaches to Incorporating External Information for the Prediction of Prostate Cancer Risk

Abstract

We present the most comprehensive comparison to date of the predictive benefit of genetics in addition to currently used clinical variables, using genotype data for 33 single-nucleotide polymorphisms (SNPs) in 1,547 Caucasian men from the placebo arm of the REduction by DUtasteride of prostate Cancer Events (REDUCE®) trial. Moreover, we conducted a detailed comparison of three techniques for incorporating genetics into clinical risk prediction. The first method was a standard logistic regression model, which included separate terms for the clinical covariates and for each of the genetic markers. This approach ignores a substantial amount of external information concerning effect sizes for these Genome Wide Association Study (GWAS)-replicated SNPs. The second and third methods investigated two possible approaches to incorporating meta-analysed external SNP effect estimates – one via a weighted PCa ‘risk’ score based solely on the meta analysis estimates, and the other incorporating both the current and prior data via informative priors in a Bayesian logistic regression model. All methods demonstrated a slight improvement in predictive performance upon incorporation of genetics. The two methods that incorporated external information showed the greatest receiver-operating-characteristic AUCs increase from 0.61 to 0.64. The value of our methods comparison is likely to lie in observations of performance similarities, rather than difference, between three approaches of very different resource requirements. The two methods that included external information performed best, but only marginally despite substantial differences in complexity.

Incorporation of Detailed Family History from the Swedish Family Cancer Database into the PCPT Risk Calculator

“…For example, the completion of multiple confirmatory genome-wide association studies identifying common and rare single nucleotide polymorphisms has promoted their incorporation into commonly used cancer risk prediction tools. 1–9 To date these markers have had only modest impacts on risk and they are often not widely used due to cost. 10 …”

A genetic-based approach to personalized prostate cancer screening and treatment

“…The authors also found that a model that incorporated SNPs had a better discriminative performance than family history, especially in men who were younger than 60 years old. Finally, other studies involving panels of 33 SNPs have documented similar utility and offer better discrimination between men with PC from those without than any other clinical variable 22–25 .…”

Multigene panels in prostate cancer risk assessment: a systematic review

“…The results were very similar, irrespective of the different methods of statistical analysis used. 49,58 The AUC for the 11-SNP panel was 65% 38 ; in a subset analysis limited to men without a reported family history of prostate cancer, the positive predictive value of the panel was 37%. 44 There were only single studies of the other panels, almost all of which reported panel development, with no information on internal or external validation.…”

“…57 The properties of an 11-SNP panel were investigated in two articles, 38,44 and of a 33-SNP panel in four; [49][50][51][52] one further article 58 compared the results of the application of alternative statistical analyses with the data described by Kader et al 51 The other 15 panels included between 2 and 35 SNPs, but each was investigated in a single study only; several of these considered family history and age in the risk prediction model. All but two evaluations were case-control (association) studies, and were heterogeneous in terms of the composition of each panel (specific SNPs and the number included), the inclusion of other risk factor data, the populations in which they were evaluated, and the metrics used to judge the performance of the panel as a "test. "…”