Prostate cancer (PCa) is the second most common cancer in men and responsible for 375,000 deaths worldwide.(1) Although it presents an indolent clinical course, PCa still remains a major health burden with mortality rates expected to rise 1.05% by 2040.(2) PCa is generally asymptomatic in the early and later stages.(3, 4) Routine cancer screening can prevent future health complications by facilitating early detection and allowing for timely intervention. The most common screening methods for PCa are the digital rectal examination (DRE) and prostate-specific antigen (PSA) test. The largest conducted trial of DRE and PSA screening demonstrated the usefulness of screening with a subsequent risk reduction in PCa-related deaths of up to 49%.(5) However, there is controversy surrounding the effectiveness of PSA screening as false positive results, overdiagnosis, and overtreatment are associated with use of this screening tool.(6) In 2012, the United States Preventive Services Task Force issued a recommendation discouraging routine PCa screening in men regardless of risk factors, causing high-grade cases to increase by 11.3%.(4) Further efforts are warranted to improve current PCa initial screening approaches and methods.
Screening is generally recommended for men aged 55 and older, as the majority of PCa cases are diagnosed in older men. Although the average age of PCa diagnosis is 66, with the highest incidence seen in those older than 65,(7) more than 10% of cases occur in men 55 and younger(8) and current research indicates that younger men diagnosed with high-grade PCa have an overall poorer prognosis.(9) Developing an accurate screening tool to predict the risk of PCa for patients younger than the standard screening age would therefore allow for earlier identification of those younger patients at risk and potentially reduce the public health burden.
The high heritability of PCa(10) demonstrates that genetic factors play a considerable role in its development. Several genome-wide association studies have identified over 170 single nucleotide polymorphisms (SNPs) that are associated with an increased risk of PCa.(11) These genetic variants can be combined to determine an individual’s polygenic risk score (PRS), and PRSs have been demonstrated to have a large clinical utility potential for numerous diseases, including PCa.(12)
PCa is also associated with additional known risk factors, such as age and ethnicity,(13) that can be routinely entered into electronic health records. PRSs along with patient data may be used for earlier and more accurate predictions of PCa, leading to earlier interventions, increased survival, and reduced healthcare costs. We have developed and validated machine learning (ML) models to predict PCa diagnosis specifically in younger men (age ≤ 55) based on PRS and relevant patient data. This risk assessment screening method is not contingent on the use of PSA or DRE results.