Abstract 6496: Exposure phenotype risk scores (E-PRS) and prostate cancer aggressiveness in the Michigan Genomics Initiative (MGI)

Abstract

Abstract Background: Exposure phenotype risk scores (E-PRS) use genetic variation as a natural experiment to examine the causal relationship between risk or protective factors and diseases. A recent review summarized 76 studies using this approach, finding that several factors (alcohol consumption, BMI, telomere length, hormones) likely cause cancer. However, this article also highlighted the need for larger studies incorporating more, newly discovered associated, variants in the E-PRS and investigating specific cancer types vs. all cancers and cancer subtypes. Methods: We used 21 published E-PRS that were derived from summary statistics of large genome-wide association studies using PRSCS. We evaluated 15 continuous and 6 binary E-PRS for factors that are known or hypothesized to contribute to cancer risk or progression and their association with prostate cancer aggressiveness and death. Our sample included 1784 unrelated, genotyped prostate cancer patients from the Michigan Genomics Initiative (MGI). Conditional logistic regression was used to estimate these 21 E-PRS's odds ratios (OR) and their 95% confidence intervals (CI) for high- vs. low-stage, high- vs. low-grade, and aggressive (high-stage or -grade) vs. non-aggressive (low-stage and -grade) prostate cancer. Cox regression was used to evaluate the hazard ratios (HR) and their 95% CI of prostate cancer deaths by E-PRS. Lasso and Ridge regressions were used to reduce model complexity and prevent overfitting for the final E-PRS model. All multivariable models were adjusted for age, genotyping batch, recruiting study, and genetic PCs. Results: For high- vs. low-stage prostate cancer, we detected statistically significant associations with E-PRS for type 2 diabetes (per standard deviation (SD) change OR=0.86, 95% CI=0.77-0.97) and estimated glomerular filtration rate (eGFR) (per SD change OR=1.19, 95% CI=1.00-1.41). We also observed a suggestive inverse association with sleep apnea E-PRS. We observed no statistically significant associations for high- vs. low-grade prostate cancer, but there was a possible inverse association with the diastolic blood pressure (DBP) E-PRS. We noticed no statistically significant associations when comparing aggressive vs. non-aggressive prostate cancer; however, suggestive inverse associations with E-PRS for DBP and smoking were found. For prostate cancer death, we observed a statistically significant decreased risk of death with higher BMI E-PRS (per SD HR=0.29, 95% CI: 0.09 - 0.89). We also observed a suggestive inverse association with the eGFR E-PRS. Similar results were found in Lasso and Ridge regression. Conclusions: Several E-PRS were significant predictors of aggressive or fatal prostate cancer. In the future, we will test significant E-PRS from this analysis to determine whether they predict aggressiveness or mortality in newly recruited MGI prostate cancer cases in 2022. Citation Format: Xinman Zhang, Lars G. Fritsche, Bhramar Mukherjee, Alison M. Mondul. Exposure phenotype risk scores (E-PRS) and prostate cancer aggressiveness in the Michigan Genomics Initiative (MGI) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6496.