Accounting for variations in prostate biopsy recommendation and acceptance

From the SWOG Statistical Center, Division of Public Health Sciences

Detectable prostate cancer is pervasive in older men.  Currently, men over the age of 55 with a normal digital rectal exam who have undergone a 6-core prostate biopsy with low prostate-specific antigen (PSA) have a 15% risk of prostate cancer. Dr. Catherine Tangen, a Full Member in the Division of Public Health Sciences and the SWOG Statistical Center, adds that “factors unrelated to cancer risk can impact which men receive a prostate biopsy, and because prostate cancer is so ubiquitous in aging men and usually asymptomatic, factors that increase the likelihood of a biopsy will appear to be a risk factor.” Factors that determine a man’s likelihood to undergo a recommended biopsy include age, marriage status, and body mass index (BMI). If an analysis is constructed that mimics a uniform strategy of cancer screening and biopsy and the magnitude or direction of the risk factor association with prostate cancer changes, this suggest the presence of detection bias in the observed data. 

In 2015, a study conducted by Dr. Wendy Barrington and colleagues evaluated race and BMI in the Selenium and Vitamin E Cancer Prevention Trial (SELECT) which assessed effectiveness of dietary vitamin E and selenium for cancer prevention. The study reported a positive linear trend in African Americans for BMI and the incidence of prostate cancer (p=0.03) and no trend in Non-Hispanic White Americans (p=.63) (1).  In the current study, Tangen et al, re-examined the relationship between race, BMI and prostate cancer risk in SELECT.  The objective of the study was to evaluate whether accounting for differential biopsy referral and acceptance rates altered the associations between race, BMI, and prostate cancer.  “This is a methods paper to show a strategy for reducing differential detection bias and secondly, we use BMI and race as an application of our method”. The paper is published in Cancer Epidemiology.

In the SELECT trial design, 35,533 men were recruited from 2001 -2004. African American men, 50 years or older, were eligible for the study; all other races were recruited at 55 years of age or older. The men were recruited from 427 study sites in the US, Canada, and Puerto Rico. All men in the study were randomized to dietary selenium, vitamin E, a combination of both, or neither in a double-blind construct.  The inclusion criteria consisted of men with a PSA less than 4 ng/ml, a normal digital rectal exam screening, and no prior diagnosis of prostate cancer.

A total of 22,673 Non-Hispanic White (median age = 63 years) and 3,398 African American (median age = 58 years) men from SELECT were evaluated for the analysis.  Based on descriptive statistics, African American men had higher BMI (p< 0.001) and more co-morbid conditions (e.g. hypertension, diabetes).  The probability of prostate cancer for men without biopsies was estimated using the Prostate Cancer Prevention Trial (PCPT) risk calculator.  Risk factors entered in the PCPT risk calculator included age, PSA, digital rectal exam status within the prior year, family history of prostate cancer, prior negative biopsy, and race. Dr. Tangen explains the impact of the study’s findings: “We introduced a statistical method to adjust for differential prostate biopsy thereby reducing bias in estimates of prostate cancer risk factors.  This is important because biased conclusions regarding risk factors and potential preventive agents can lead to spending large amounts of time and resources following the wrong scientific path.”

Graphical representation of Incidence Density Ratio (IDR) Estimates for Prostate Cancer Incidence, Stratifying on Race and BMI Categories, Lowest Non-Hispanic White BMI Group (W1) is the Reference Group
Incidence Density Ratio (IDR) Estimates for Prostate Cancer Incidence, Stratifying on Race and BMI Categories, Lowest Non-Hispanic White BMI Group (W1) is the Reference Group Image from Dr. Catherine Tangen

Figure 1 shows the trends for the unadjusted and biopsy bias adjusted incidence density ratios (IDRs, similar to hazard ratios) for race and BMI group.  African American men in BMI groups A1 – A5 have a higher risk of prostate cancer in comparison to the lowest Non-Hispanic White BMI group (IDRs: 1.12, 1.51, 1.45, 1.46, 1.60) and the BMI linear trend is statistically significant.  The bias adjusted BMI IDRs are reduced closer to 1.0 and no longer show a linear trend (IDRs: 0.99, 1.32, 1.21, 1.25, 1.30).  For Non-Hispanic White men, the biopsy bias adjustment has a much smaller impact on the IDRs and continues to show no association between BMI and prostate cancer.

The results of the article suggest that without controlling for biopsy recommendation and acceptance, the risk of prostate cancer is over-estimated in African American men with higher BMI. Within this new approach, there isn’t clear evidence of a linear trend as seen in Barrington et al (1). However, African American men with a BMI ≥ 25.0 have a 21 – 32% increased risk of prostate cancer compared to Non-Hispanic White men with a BMI < 25.0. There may still be merit in interventions that target BMI in African American men since there was no increased risk in the lowest BMI group (IDR=0.99).  Dr. Tangen summarized the results, “Our modeling indicates that if all men in the SELECT cohort received biopsies uniformly based on their screening values, African American men would still have a higher risk of prostate cancer compared to non-Hispanic White men but the observed positive association between larger BMI values and increased prostate cancer risk would not be so pronounced.”

Now that the SWOG Statistical Center has  contributed an approach to analyze the incidence of prostate cancer with reduced detection bias, Dr. Tangen details her group’s next step “ will examine associations between other risk factors that may also fall under this bias scenario such as diabetes, statin use, aspirin use, fatty acids, vitamin D and see if associations with prostate cancer change when accounting for biopsy bias. As a research community, we need to push for more detailed screening and biopsy information for our prostate cancer cohorts so these types of adjustments can be evaluated.”


This work is supported by a U01 Cancer Cohort grant from the National Cancer Institute.

Fred Hutch/UW Cancer Consortium members Catherine Tangen and Wendy Barrington contributed to this research.

Tangen, C.M., Schenk, J., Till, C., Goodman, P.J., Barrington, W., Lucia, M.S. and Thompson, I.M., 2019. Variations in prostate biopsy recommendation and acceptance confound evaluation of risk factors for prostate cancer: Examining race and BMI. Cancer epidemiology63, p.101619