Abstract
Prostate-specific antigen (PSA) screening for prostate cancer remains controversial because it increases overdiagnosis and overtreatment of clinically insignificant tumors. Accounting for genetic determinants of constitutive, non-cancer-related PSA variation has potential to improve screening utility. In this study, we discovered 128 genome-wide significant associations (P < 5 × 10−8) in a multi-ancestry meta-analysis of 95,768 men and developed a PSA polygenic score (PGSPSA) that explains 9.61% of constitutive PSA variation. We found that, in men of European ancestry, using PGS-adjusted PSA would avoid up to 31% of negative prostate biopsies but also result in 12% fewer biopsies in patients with prostate cancer, mostly with Gleason score <7 tumors. Genetically adjusted PSA was more predictive of aggressive prostate cancer (odds ratio (OR) = 3.44, P = 6.2 × 10−14, area under the curve (AUC) = 0.755) than unadjusted PSA (OR = 3.31, P = 1.1 × 10−12, AUC = 0.738) in 106 cases and 23,667 controls. Compared to a prostate cancer PGS alone (AUC = 0.712), including genetically adjusted PSA improved detection of aggressive disease (AUC = 0.786, P = 7.2 × 10−4). Our findings highlight the potential utility of incorporating PGS for personalized biomarkers in prostate cancer screening.
Main
Prostate-specific antigen (PSA) is an enzyme produced by the prostate gland that degrades gel-forming seminal proteins to release motile sperm and is encoded by the KLK3 (kallikrein 3) gene1,2,3. As prostate epithelial tissue becomes disrupted by a tumor, greater PSA concentrations are released into circulation2,3. PSA levels can also rise due to prostatic inflammation, infection, benign prostatic hyperplasia, older age and increased prostate volume3,4,5. Increased body mass index is associated with lower PSA levels, but the underlying mechanisms remain unclear6,7. Low PSA levels, thus, do not rule out prostate cancer, and PSA elevation is not sufficient for a conclusive diagnosis8. Although PSA testing reduces deaths from prostate cancer9, between 20% and 60% of cancers detected using PSA testing are estimated to be overdiagnoses10,11,12. In addition, the long-term risk of lethal prostate cancer remains low, especially in men with PSA below the age-specific median13,14. As a result, clinical guidelines in the United States and globally advise against population-level PSA screening and promote a shared decision-making model15,16.
One avenue for refining PSA screening is to account for variability in PSA due to genetic factors. PSA is highly heritable, with 40 independent loci identified in the largest previous genome-wide association study (GWAS)17,18. The goal of genetically correcting PSA levels is to increase the relative variation in PSA attributable to prostate cancer, thereby improving their predictive value for disease detection. The first study to genetically correct PSA using just four variants reclassified 3% of participants to warranting biopsy and 3% to avoiding biopsy19. Incorporating additional genetic predictors has the potential to personalize PSA testing, reduce overdiagnosis-related morbidity and improve detection of lethal disease. To maximize the utility of this approach, it is critical to distinguish genetic variants that influence constitutive PSA levels from those affecting prostate tumor development. PSA and prostate cancer share many genetic loci17,19,20,21,22, but the extent to which this overlap reflects screening bias remains unclear, as GWASs of prostate cancer may capture signals for disease susceptibility and incidental detection due to benign PSA elevation.
Our study explores the genetic architecture of PSA levels in men without prostate cancer, with a view toward assessing whether genetic adjustment of PSA improves clinical decision-making related to prostate cancer diagnosis. It also provides a novel framework for the clinical translation of polygenic scores (PGSs) for non-causal cancer biomarkers.
Results
The study design of the Precision PSA study is illustrated in Fig. 1. Using data from five studies (Methods), we conducted genome-wide analyses of PSA levels ≤10 ng ml−1 in cis-gender men never diagnosed with prostate cancer. GWAS results were meta-analyzed within ancestry groups and then combined across populations for a total sample size of 95,768 individuals….
