In this setting, we investigated whether trends in the incidence of metastatic PCa from 2005 to 2021 differed by race/ethnicity. Furthermore, we sought to assess whether any observed changes in the incidence of metastatic disease coincided with race-based changes in the rate of prostate-specific antigen (PSA) screening. We hypothesized that there would be rising rates of metastatic disease and declining odds of screening across our study period and that the trends would differ based on patient race/ethnicity.

We used data from the Surveillance, Epidemiology, and End Results (SEER) database (Research Limited Field Data, 22 Registries) [14] from 2005 to 2021 to study trends in metastatic PCa incidence rates by race/ethnicity. Additionally, we analyzed Behavioral Risk Factor Surveillance System (BRFSS) survey data [15] from 2012 to 2020 to examine national PSA screening trends by race/ethnicity. These were the earliest years for which consistent recent screening data is available. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for cross-sectional studies [16]. This study used only de-identified data and was approved and exempted from review by the Brigham and Women's Institutional Review Board.

SEER is a population-based cancer registry maintained by the National Cancer Institute in the USA. The SEER Research Limited-Field Data provides cancer incidence rates and covers about 47.9% of the US population [17]. We queried the SEER database (Research Limited Field Data, 22 Registries, Nov 2022 Sub (2000-2020)) for annual age-adjusted incidence rates of metastatic PCa in non-Hispanic Black (NHB) and non-Hispanic White (NHW) men between 2005 and 2021. The study cohort was limited to NHB and NHW men of any age diagnosed with metastatic PCa. The SEER Combined Summary Stage was used to identify patients diagnosed with metastatic PCa at initial presentation. This was done to focus on advanced disease which could plausibly be related to lack of screening rather than disease which progresses after diagnosis and treatment. Age-adjusted incidence rates (per 100,000 individuals) of metastatic PCa were extracted from the SEER database using the SEER*Stat 8.3.4 software.

The SEER data are not at the patient-level, but instead are at the national level. In particular, the SEER data are in the form of annual age-adjusted estimated incidence rates of metastatic PCa for NHW and NHB men from 2005 to 2021, excluding men of other or unknown races and ethnicities. Patients who were initially diagnosed with localized PCa and subsequently developed distant metastases were not included in our analysis. A log-linear regression model was then fit to 34 national age-adjusted incidence rates, with main effects for race and categorical calendar year, and their interaction. With year as a categorical covariate, this model is completely non-parametric in that we are not assuming any trend in time for NHW or NHB. The parameters of the log-linear regression model were calculated using optimal weighted least squares [18] (weighted by the inverse of the estimated variance for each rate) with a log-link. This model allows us to test if the annual incidence rate ratio changes over time by testing if the interaction in the model is significant. The exponential of linear combinations of the main effects and interaction terms from the log-linear regression model are used to estimate annual rate ratios of NHB versus NHW men (these are often called 'marginal effects').

BRFSS is an annual, nationwide telephone-based survey conducted by the US Center for Disease Control and Prevention. It collects data from adult US residents on health-related risk behaviors, chronic health conditions and the use of preventive services. With over 400,000 interviews conducted annually across all 50 states, it is the largest continuous health survey in the world. We analyzed individual-level BRFSS survey data from 2012, 2014, 2016, 2018 and 2020, as the PSA screening question was included only in biennial surveys. We excluded survey data before 2012 because a different statistical weighting method was used in earlier years. Data from 2022 was also excluded because the PSA screening question was only asked in a limited number of states. Our BRFSS study cohort included NHW and NHB men aged 40-69 years, selected based on current AUA guidelines [19]. The AUA recommends offering PSA screening to Black men beginning at age 40-45 and to the general population between 50 and 69. Men with a prior diagnosis of PCa were excluded from the analysis. The primary outcome was receiving a PSA screening within 2 years prior to the survey, identified using the BRFSS variable 'psatime'. The primary explanatory variables of interest were self-reported race (NHB, NHW) and survey year (2012, 2014, 2016, 2018 and 2020). Additional covariates included were health insurance, self-reported health status, highest level of education, marital status and median household income (MHI).

We calculated descriptive summary statistics for the study cohort at two levels: the individual survey participant level and the population level. Population-level estimates were derived using survey weights, which accounted for the complex sampling design. We calculated the weighted proportions of NHW and NHB men who received PSA screening within 2 years for each survey year. A multivariable logistic regression analysis of predictors of PSA screening from 2012 to 2020 was conducted. The primary explanatory variables of interest included in the model were race/ethnicity and survey year. The variable survey year was specified as a categorical variable, with 2012 as the reference year, so each year's odds ratio represents the change in screening odds compared to 2012. The model was adjusted for health insurance, self-reported health status, highest level of education, marital status and MHI.

Both primary explanatory variables of interest were significant predictors of PSA screening. A second multivariable logistic regression analysis was performed incorporating an interaction term between race and survey year. A subgroup analysis was conducted to calculate the odds ratio (OR) for PSA screening between NHW and NHB men for each survey year separately. This analysis adjusted for health insurance, health status, highest level of education, marital status and MHI.

Statistical analysis of SEER and BRFSS data were conducted using the Stata 17.0 BE-Basic Edition (Stata Corp, College Station, TX). Two-sided significance tests were applied, with p values less than 0.05 considered statistically significant.