Skip to Main Content


Prostate Cancer Cohort

Cohort Description

The goal of the Prostate Cancer cohort is to develop a prognostic model for men diagnosed with localized disease that will predict those likely to develop metastatic cancer or die from prostate cancer, those likely to die from something else, and those likely to survive living with prostate cancer for many years. This prediction will help patients and providers make more informed treatment decisions.

Using the VA’s Oncology Registry, we identified 222,000 veterans diagnosed with prostate cancer between 2001 and 2017. The registry contains information not typically found in electronic health record (EHR) data including histology, TNM stage of disease, and Gleason score. We limited to those with biopsy confirmed disease that was classified as adenocarcinoma, leaving 207,000. Five percent had disease that had spread beyond the prostate (N1 or M1), and another 5% had unknown stage based on information in the registry. To get a more homogeneous sample we excluded 26,000 who had another cancer documented in the registry and those with a Gleason score <6, leaving a potential sample of 156,000. We were unable to find a value for PSA at diagnosis for 4%, and no record of a biopsy performed in the VA for 15%. Because of uncertainty that the registry captured the initial diagnosis we chose to exclude these men. For the same reason we excluded men with PSA <1 at diagnosis because this suggested they had begun treatment outside the VA. Additionally, those with PSA >100 at diagnosis were excluded because of high potential for occult metastases. Our final analytic sample comprised nearly 119,000 veterans.

In addition to data captured in the oncology registry, we have available full EHR data going back to 2000 and continuing to the present. This includes diagnoses, procedures, laboratory results and medications, as well as BMI, smoking status and alcohol use patterns. Cause of death, from death certificates, is currently available to 2017 and is being updated to 2018.

Population Characteristics

The median year of diagnosis was 2010, with median follow-up of 6 years, but up to 16 years for those diagnosed in 2001. Median age at diagnosis was 65; 59% were white and 27% were black. Most had tumor stage 1 (71%) with Gleason score of 6 or 7 (83%). As of the end of 2017, 22,000 had died of any cause and 3,200 from prostate cancer. Some more characteristics can be found here.