Adaption and National Validation of a Tool for Predicting Mortality from Other Causes Among Men with Nonmetastatic Prostate Cancer

To avoid over- and under-treatment of prostate cancer, providers should balance the risk of death from prostate cancer against other causes (non-prostate cancer mortality). Less than a quarter of urologists and radiation oncologists use a tool to estimate non-prostate cancer mortality and, without standard tools, provider risk estimates are often inaccurate and may be biased. Specifically, overreliance on age can cause treatment to be withheld from older healthy men and given to younger men in poor health.

Available tools for non-prostate cancer mortality are not well adapted to estimating risk among men with non-metastatic prostate cancer in a busy clinical setting. We have developed and validated an electronic health record (her) compatible prediction tool in a general population of 6.5 million people. The tool combines laboratory data and body mass index from the Veterans Aging Cohort Study Index with age and components of the Charlson Comorbidity Index (VACS-CCI). We recalibrated it to predict non-prostate cancer mortality specifically among men with non-metastatic disease.

Using the National Veterans Administration Healthcare System EHR database, we created an observational cohort of men with biopsy confirmed non-metastatic prostate cancer from 2001-2018 divided by year of diagnosis into development (2001-6, 2008-18) and validation (2007) sets. Accuracy of mortality prediction was assessed using calibration curves and C-statistics in development, validation, and combined, overall and by age (<65, 65+ yrs.), race (White, Black), Hispanic ethnicity, and treatment groups.

Among 107,370 individuals, we observed 24,977 deaths, 86% of which were not due to prostate cancer). Median age was 65 years, 4,947 were Black, and 5,010 were Hispanic. VACS-CCI was better at predicting non-prostate cancer mortality than age alone, or age and CCI. This was true in younger and older men, Black and White men, and men of Hispanic ethnicity.

Our study supports the implementation of VACS-CCI. First, we validated a non-prostate cancer mortality tool that only includes data directly available within the VA EHR, enabling wide implementation. Second, VACS-CCI relies on standard statistical functional forms and employs a manageable number of predictor variables.

When we stratified men by their VACS-CCI estimated risk of ten-year mortality which includes age in its estimate, age remained strongly associated with receipt of surgery. Younger men with poor prognoses were much more likely than older men with excellent prognoses to undergo surgery. Without a standardized and validated tool, providers and patients may be overly influenced by age. But unless these tools are readily available in the clinic setting, they are unlikely to be used. Our findings VACS-CCI is ready for EHR-assisted implementation as a prognostic tool for non-PCM among men with non-metastatic prostate cancer treated within the VHA. The use of VACS-CCI might mitigate inaccuracy and unconscious bias in provider estimation.

Written by: Amy C. Justice, MD, PhD¹ & Janet P. Tate, MPH, ScD²

VA Connecticut Healthcare, West Haven, CT, USA; Pain Research, Informatics, Multimorbidities, Education (PRIME) Center, VA Connecticut Healthcare System, West Haven, CT, USA; Department of Medicine, Yale School of Medicine, New Haven, CT, USA; School of Public Health, Yale University, New Haven, CT, USA.
VA Connecticut Healthcare, West Haven, CT, USA; Department of Medicine, Yale School of Medicine, New Haven, CT, USA.

Read the Abstract

Login