Read the Full Video Transcript

Christopher Wallis: Hello, and thank you for joining us for a UroToday Journal Club discussion. Today, we're discussing a recent publication entitled Mediators of Racial Disparity in the Use of Prostate Magnetic Resonance Imaging Among Patients With Prostate Cancer. I'm Chris Wallace, an assistant professor in the division of urology at University of Toronto. With me today is Zach Klaassen, assistant professor in the division of urology at the Medical College of Georgia. As you can see here, this recent work was published in JAMA Oncology and led by Dr. Leapman and colleagues.

Racial disparities in prostate cancer are well recognized in the United States. African American men in particular are more likely to be diagnosed with prostate cancer but are then less likely to receive guideline concordant treatment and are nearly twofold more likely to die of prostate cancer. It, through a whole wealth of research, has been found that socially driven differences in care and particularly in access to guideline concordant care, rather than biologic differences, likely underpin the differences in outcomes experienced by African American men and white men.

Most recently, there's been relatively large advancements in technological utilization in the evaluation of prostate cancer with an increasingly large role of upfront magnetic resonance imaging, and disparities in access to these emerging diagnostic technologies may widen gaps in prostate cancer diagnosis, treatment, and outcomes, and further exacerbate disparities. Previous work has suggested that there's lower use of prostate MRI among non-white patients, with other differences due to both insurance status and age.

And so in the present study, the authors sought to identify factors that underlie racial disparities in the use of prostate MRI among a population-based cohort in the US. To do so, they performed a retrospective cohort study using the SEER-Medicare linked dataset, and they used multiple mediation analysis to explore the existence of various intervening variables, termed mediators, in the causal pathway between race and utilization of MRI. The cohort included men age 66 years and older with clinically localized prostate cancer who were diagnosed between 2011 and 2015. Patients were excluded if there was no record of a prostate biopsy, an unknown month of diagnosis, diagnosis based on death certificates or autopsy, a prior cancer diagnosis preceding this one, death within six months of diagnosis, a subsequent second cancer diagnosis within 12 months, or any prior prostate cancer therapy.

The outcome of interest was receipt of prostate MRI within 13 months surrounding diagnosis, and they included both the month of diagnosis, as well as six months prior and six months after this period. They examined a host of variables that we will go through here, as they are potentially important mediators of this relationship. So race was defined as black, white, or other. The primary comparisons in the study were between patients categorized as black and those categorized as white. Secondary comparisons examined black versus other race or ethnicities. The authors further examined a number of patient-level characteristics, including age, medical comorbidity, disability, marital status, and dual Medicaid eligibility. They characterized cancer details using a modified CAPRA score, which omitted the number of positive cores, as this is not available in the dataset.

Finally, they considered small area socioeconomic factors and looked at three in particular. The first was a binary indicator for poverty, and to do so, they examined within the zip code for each patient whether 20% or more of individuals were living below the poverty line. Further, they examined the imaging rate within the hospital referral region that a patient lived, and this examined utilization of CT and bone scan among prostate cancer patients, and the authors operationalized this into quartiles. Finally, they used Index of Concentration at the Extremes for race to examine for residential segregation at the zip code level. Higher scores on the ICE are indicative of more segregation and less intermixing of races. This was operationalized by quintiles.

In terms of analysis, the authors first assessed the association between race, sociodemographics, clinical characteristics, and healthcare-related characteristics on MRI utilization using both unadjusted and multi-variable adjusted logistic regression models. They then used mediation analysis to assess the presence and relative contributions of factors that may mediate this association.

So you can see in the figure here to the right, the four domains were assessed, including clinical and pathologic characteristics, patient-level sociodemographics, spatial characteristics, and structural racism that may contribute to the effect of race on receipt of MRI.

This mediation analysis was performed using generalized method for mediation analysis with multiple additive regression trees, and the authors used 100 bootstrapped iterations to assess both direct and indirect effects of race on the receipt of prostate MRI. Now, I'm going to pass it over to Zach to walk us through the results of the study.

Zachary Klaassen: Thanks, Chris, for that great introduction. This is table one for the paper, looking at the clinical and sociodemographic characteristics of the study, which is stratified by receipt of MRI. And this is a large table, so I've broken it down into two separate slides. Looking at the table, we'll focus on the two columns on the far right. The right side is MRI patients, and to the left of that is patients that did not receive an MRI. And we can see that, overall, 9.5% of patients in this cohort of more than 35,000 patients had an MRI, with the most common age breakdown being 66 to 74 years of age, which encompassed about 75% to 80% of the patients.

Looking at the racial breakdown among men that had MRI, 6.7% were black and 85.7% were white, whereas of those that did not have an MRI, 10.4% were black and 82.1% were white. Moving down to the middle of the slide, percent of region with high school degrees or less, in terms of less than 30%, 52.5% had an MRI and 34.1% did not have an MRI. Percent of the census tracked below the poverty level, less than 5% of the poverty level, 33.9% had an MRI compared to 23.3% did not have an MRI. And then moving down to the resident location in terms of metropolitan, 91.7% MRI compared to 83.2% that did not have an MRI.

And moving on to geographical location, the most common region in this study that had MRI was the Northeast at 31.7%. And we can see here, looking at diagnosis here in the middle of the slide, the patients continued to have more MRIs as we moved through the time of the study, from comparing 2015, 33.8%, compared to 2011, 14.2%. Looking specifically at the SEER registry, the areas in the SEER registry that had the most MRIs accounted for included greater California, 22.3%, New Jersey, 24.1%, and Los Angeles, 11.1%.

This figure looks at the racial disparity between black and white patients in the use of prostate MRI, and as Chris already highlighted in the methods, by breaking down these factors, the total indirect effect of receipt of MRI accounted for 81% of the disparities between black and white patients receiving MRI. Specifically, the highest was SEER registry at 24% of the accountability for disparity, followed by the ICE quintile at 19%, high poverty area at 19%, and dual Medicaid eligibility at 11%.

Again, when we compare the racial disparity between black and other races, we see very similar outcomes, as opposed to black and white patients, with a total indirect effect of 92%, again with SEER registry comprising 54% of the disparity, the ICE quintile 22% of the disparity, and high poverty area 15% of the disparity.

This figure looks at the proportion of patients with prostate cancer receiving a prostate MRI around the time of their diagnosis by SEER registry region, stratified by race. And so for this figure, the black patients are in dark blue, the white patients are in sort of medium color blue, and the other patients are in lighter blue. Some highlights of this figure are, if we look at the disparities between black and white, certainly we see this disparity in the Los Angeles registry, the Connecticut registry, and the New Jersey registry, as well as, with less of a denominator, the Louisiana registry. Interestingly, we see very little disparity in several registries as well, including the greater California registry and the Atlanta, Georgia, registry.

This table looks at the mediation analysis estimations of racial disparity and prostate cancer for MRI receipt. And this is essentially just a tabular form of the table I showed previously, again highlighting that direct effects related to disparity accounted for 19% in the black versus white, whereas the indirect effects accounted for 81%. And when looking at black versus other patients, again the direct effects had an 8% involvement for disparity, and the indirects had a 92% involvement in the disparity of receipt of MRI.

There are several important discussion points from this study, and I'll highlight this in the next couple of slides. We found in this study that black men with localized prostate cancer were significantly less likely to receive a prostate MRI in the period surrounding their diagnosis. There was identified mediators that accounted for 81% of the observed racial disparity in prostate MRI use between black and white patients, and specifically geospatial factors mediated the greatest share of racial disparity in prostate MRI use, accounting for almost one quarter of the black/white disparity being attributed to differences in the use of MRI between the SEER registry regions. Furthermore, residents in high poverty areas where prostate MRI use was less common was associated with 19% of the effect of race, and the ICE measures, which is a representation of the extent to which an area's residents are spatially concentrated by race, which is a proxy for structural racism, explain 19% of the racial disparity in receipt of MRI.

So these results can help to inform focused efforts to improve equitable access and quality of diagnostic cancer imaging. First, greater access to the prostate MRI has been championed by the AUA, the American College of Radiology, and patient advocacy groups, to improve quality and precision of prostate cancer care. However, unfortunately, tangible plans are undefined amongst these groups.

Secondly, addressing sources of geographic and structural variation is a strategic initial step for improving equity in prostate cancer imaging. And finally healthcare institutions, payers, and professional organizations can undertake multi-level initiatives to help improve this, for example tracking and responding to patterns of MRI use among all patients in their geographic region.

So, in conclusion, in this population-based cohort study of US adults using mediation analysis, sociodemographic factors, including geographic variation, poverty, and structural racism explained most of the observed racial disparity in prostate MRI use among black and white Medicare beneficiaries with a new diagnosis of prostate cancer. This research suggests that a targeted action to address race-based differences in prostate MRI use should be incorporated and focus on spatial factors, as well as upstream social determinants of health. In addition, this work highlights the feasibility of identifying quantifying processes that underlie racial disparity and that can be leveraged in other diseases and treatments.

Thank you very much for your attention. We hope you enjoyed this UroToday Journal Club discussion of this paper published in JAMA Oncology.