Is Prostate Imaging Reporting and Data System equally predictive across racial groups? A narrative review of multiparametric magnetic resonance imaging performance in Black versus White men

Abstract

Background

Multiparametric magnetic resonance imaging (mpMRI) has become integral to prostate cancer (PCa) screening, improving detection and reducing unnecessary biopsies. This study investigated radiological disparities in mpMRI between Black and White PCa patients to improve diagnostic equity.

Materials and methods

A PubMed literature search was conducted using keywords related to PCa, MRI, and racial disparities. Articles discussing mpMRI outcomes in White and Black PCa patients were selected.

Results

From 13 studies, two reported higher Prostate Imaging Reporting and Data System (PI-RADS) 4 and 5 scores among Black patients, while three showed higher proportions in White patients without increased clinically significant prostate cancer (csPCa). csPCa is defined as International Society of Urological Pathology (ISUP) grade group ≥2. The csPCa detection rate in Black patients was equal to or higher than in White patients, with some studies showing more PI-RADS 3 lesions among Black patients with csPCa. Key findings include the following: (1) Development bias: PI-RADS may not account for racial variations; (2) inconsistent lesion detection across racial groups; (3) lower PI-RADS scores may mask clinically relevant cancer in Black patients; and (4) inadequate representation of the disease in Black men. Anatomical variations may affect lesion visibility, requiring dedicated imaging pathology studies.

Conclusions

While valuable for PCa detection, current mpMRI applications may contribute to racial disparities and biopsies should not be omitted in black patients with negative mpMRI or PI-RADS 3 lesions. Addressing these disparities requires inclusive research and consideration of racial variations in diagnostic pathways.

1. Introduction

Multiparametric magnetic resonance imaging (mpMRI) has become an invaluable tool for visualizing lesions and stratifying risks using standardized criteria such as the Prostate Imaging Reporting and Data System (PI-RADS).¹ In recent years, MRI has been increasingly incorporated into prostate cancer (PCa) screening guidelines. For instance, the Canadian Urological Association endorsed the 2021 Cancer Care Ontario guidelines, which recommend mpMRI as a part of the diagnostic process for men with elevated prostate-specific antigen (PSA) levels or other risk factors.² Similarly, the American Urological Association acknowledges PSA as the primary screening tool but recognizes MRI as a valuable adjunct to guide biopsy decisions while acknowledging that prebiopsy mpMRI may assist in identifying clinically significant prostate cancer (csPCa; grade group ≥2).³ A recent systematic review published in JAMA Oncology further supports MRI’s effectiveness in detecting csPCa while minimizing unnecessary biopsies.⁴ Epidemiological evidence consistently shows that Black men face a higher incidence of developing PCa, often at a younger age and with more advanced disease than White men.⁵ These disparities underscore the need to investigate whether differences also exist in imaging outcomes. Chatterjee et al.⁶ suggested that Black men may exhibit differences in prostate tissue composition, such as a greater proportion of epithelium and less lumen space in cancerous tissue, which could affect the detection of abnormalities on MRI. This study aimed to explore the potential radiological disparities in MRI between Black and White patients with PCa. Such insights may improve diagnostic accuracy, reduce inequities, refine clinical guidelines, and support personalized approaches to PCa care.

2. Material and methods

2.1. Literature search

We searched the PubMed database using advanced features. The publication date filter has been adjusted over the last 10 years. Filters were applied to include only human studies published in English. A search was conducted using the following keywords: (“prostate cancer” OR “prostate carcinoma” OR “prostatic neoplasm”) AND (“MRI” OR “imaging” OR “magnetic resonance imaging” OR MRI-∗) AND (“dispariti∗” OR “sociodemographic” OR “race” OR “racial” OR “White men” OR “African American” OR “Afro-American” OR “non-hispanic White American” or “Black men”).

2.2. Selection process

Using PubMed’s advanced search features, 216 articles were initially identified (Fig. 1). Eligibility was assessed according to predefined inclusion criteria: (1) studies reporting data on both Black and White patients with PCa and (2) studies comparing MRI radiological (e.g., PI-RADS scores) and histological outcomes (e.g., csPCa detection rates) between these groups. Of the 216 identified articles, 203 were excluded after title and abstract screening. When relevance could not be determined from the abstract alone, a full-text review was performed. Articles that did not address the research question or that failed to meet the selection criteria were excluded. Consequently, 13 articles were included in this study. Data abstraction was conducted by a single reviewer using a standardized approach.

Fig. 1.

Screening and evaluation process for inclusion in this study.

2.3. Data collection process

Data were extracted from the included reports using a standardized data extraction form. For each eligible article, information was collected on study characteristics (authors, year of publication, country, journal, and sample size), participant characteristics (age, race, and PSA levels), and imaging/histological parameters (PI-RADS scores and csPCa detection rate). One of the authors extracted the data. The primary outcome was the comparison of MRI radiological parameters between Black and White patients with PCa, specifically PI-RADS score distribution regarding csPCa detection rate. csPCa is defined histologically as PCa with a Gleason score ≥3 + 4 (ISUP grade group ≥2). All the outcomes were extracted, as reported in the included studies.

We conducted a heterogeneity test on PSA values, which yielded an I² of 48% (indicating moderate heterogeneity), with a Q-test P value of 0.07. The analysis was limited by missing interquartile range data in many studies, meaning that the I² calculation could only be performed on seven of the 13 included articles. This suggests that if all studies were integrated, heterogeneity would likely reach statistical significance at the 5% level.

Considering this heterogeneity, we opted not to conduct a formal statistical analysis, such as calculating the I² statistic, because our review was narrative in nature, rather than a systematic review or meta-analysis. Such analyses typically require adherence to rigorous methodological standards, including the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, which were not applied in this project.

Furthermore, the studies varied substantially in design, patient characteristics, and outcome definitions, rendering statistical pooling inappropriate and potentially misleading. Instead, we adopted a descriptive approach to qualitatively highlight variability across studies.

3. Results

Median PSA values were 6.5 ng/mL (range: 5.3–8.6) in Black patients versus 7.0 ng/mL (range: 6.2–9.3) in White patients, with an average PSA difference of +0.56 ng/mL in favor of White patients. Since higher PSA levels are associated with an increased risk of aggressive PCa, such variability could mask differences in the prevalence of high PI-RADS scores.

To assess whether Black patients were at a risk of misdiagnosis based on the current mpMRI PI-RADS scoring system, we first compared the proportion of patients with high PI-RADS scores across both populations (Table 1). Two studies (Deebajah⁷ and Meza⁸) reported a higher proportion of Black patients with PI-RADS 4 (+16%) and PI-RADS 5 (+7.4%), respectively. Notably, these are the only studies that included comparable numbers of Black and White participants, which may have limited selection bias. In contrast, most other studies enrolled relatively few Black patients, introducing greater variability in this group that could obscure potential differences.

Table 1.

Summary of PI-RADS, csPCa differences, and PSA by study and race

Author	Yr	% Black PCa	% White PCa	% Black PI 3	% White PI 3	% Black PI 4	% White PI 4	% Black PI 5	% White PI 5	csPCa diff	PSA (ng/mL) Black	PSA ng/mL White
Chatterjee26	2014-2021	32.4	67.6	NA	NA	NA	NA	NA	NA	0	8.5	8.6
Zabihollahy10	2010-2023	33.33	66.67	8.2	12.7	37.8	47.6	NA	NA	0	6.8	6.5
Henning14	2014-2017	33.5	66.5	32.8	39.1	32.8	31.3	34.4	29.5	0	9.3	8.1
Deebajah7	2015-2017	50	50	48	40.7	20.3	4.1	12.2	8.9	1 (12%)	6.6	5.4
Meza8	2014-2021	50	50	31.5	37.0	50	51.9	18.5	11.1	1 (9.5%)	6.5	5.3
Koller9	2017-2020	32.5	67.5	25.9	11.3	35.5	43.8	16.3	35	0	6.3	6.4
Kongnyuy27	2007-2015	50	50	NA	NA	NA	NA	NA	NA	NA	7.8	7.2
Walton15	2015-2018	28	68	53	54	47	46	NA	NA	NA	NA	NA
Bloom13	2007-2017	16	84	NA	NA	NA	NA	NA	NA	1 (10%)	NA	NA
Falagario11	2013-2019	23	77	43	22	35	45	22	32	0	6.2	6.1
Shin28	2012-2015	17.7	82.3	NA	NA	NA	NA	NA	NA	0	7.1	6.9
Tamada29	2011-2016	13	87	NA	NA	NA	NA	NA	NA	NA	NA	NA
Mahran16	2014-2017	21.7	78.3	44.6	45.9	31.6	28	23.7	26	NA	7	6.0

csPCa, clinically significant prostate cancer; PCa, prostate cancer; PI-RADS, Prostate Imaging Reporting and Data System; PSA, prostate-specific antigen.

The included studies exhibited significant heterogeneity. This variability may be partially attributed to differences in the racial composition of study populations, such as varying proportions of Black versus White patients as well as discrepancies in the reported rates of csPCa. The csPCa column in Table 1 highlights differences in csPCa detection rates between Black and White patients across studies. In most cases, the difference is minimal or absent, suggesting similar detection rates. However, a few studies (e.g., Deebajah⁷ and Meza⁸) reported slightly higher csPCa rates in Black patients, raising questions about PI-RADS sensitivity. Observed heterogeneity supports cautious interpretation of mpMRI findings when study populations include racially diverse patients.

PI-RADS 3 lesion findings: Koller et al.⁹ reported a higher proportion of PI-RADS 3 lesions in Black patients (25.9%) than in White patients (11.3%). In a secondary analysis of the PREVENT study, Black men with PI-RADS 3 lesions had an overall PCa detection rate of 70% versus 43% in White men (P = 0.004), and a csPCa detection rate of 60% versus 27% (P = 0.003).

4. Discussion

Our review aimed to investigate potential radiological disparities in MRI outcomes, such as PI-RADS score distribution, between Black and White patients with PCa. The included studies exhibited significant heterogeneity. These results suggest potential differences in the diagnostic performance of the mpMRI PI-RADS scoring system between Black and White patients with PCa. In other studies, with comparable numbers of participants from both racial groups (Deebajah⁷ and Meza⁸), a higher proportion of Black patients were assigned PI-RADS 4 or PI-RADS 5 scores, which could reflect either a greater burden of suspicious lesions or differences in imaging interpretation. In contrast, other studies (Zabihollaly,¹⁰ Koller,⁹ and Falagario¹¹) observed a higher frequency of PI-RADS 4/5 lesions in White patients, without a corresponding increase in csPCa detection, than in Black patients. This lack of correlation raises the possibility that PI-RADS scoring may underestimate the malignant potential of lesions in Black patients, potentially leading to an underdiagnosis or delayed intervention.

Our findings also indicate that PI-RADS 3 lesions may hold greater significance in Black patients. Since PI-RADS 3 lesions often require more conservative management, this pattern could have clinical implications for biopsy decision-making and patient outcomes. Consequently, clinicians who rely solely on PI-RADS cutoffs may risk underestimating significant disease in Black patients, potentially leading to delays in biopsy or more aggressive diagnostic procedures when scores are only marginally elevated.¹²

From a practical standpoint, these findings suggest that biopsy decision-making may need to be individualized for Black patients. For example, a Black man with elevated PSA and a PI-RADS 3 lesion may warrant a lower threshold for biopsy than a White patient with the same imaging findings, given the higher likelihood of clinically significant cancer in this group.

Furthermore, a study conducted by Bloom¹³ reported an elevated rate of csPCa among Black patients, although it should be noted that mpMRI data were not available. In contrast, three studies (Henning,¹⁴ Walton,¹⁵ and Mahran¹⁶) found no significant differences in the PI-RADS distribution between the two populations under investigation. It is noteworthy that two of these studies also documented the most substantial differences in PSA levels, with increases of +1.2 ng/mL and +1 ng/mL; the third study did not report PSA levels. These findings raise concerns regarding the comparability of their PI-RADS data.

Race has been a well-studied epidemiological factor for PCa. Studies have consistently shown that Black men have a higher risk of developing PCa, often at a younger age,¹⁷ and with more advanced disease than White men.¹⁸ Disparities between races appear to arise from the interaction of many factors, such as structural racism, socioeconomic factors, environmental exposures, and distinct underlying genetics and disease biology. For example, Hage Chehade’s review indicates that Black patients are more likely to exhibit genomic alterations in Speckle-type BTB/POZ protein (SPOP) or B-type Raf kinase (BRAF), while demonstrating lower frequencies of alterations in the AKT/PI3K pathway, androgen receptor axis, and tumor suppressor genes.¹⁹

A study by Lillard Jr. et al.²⁰ suggests that racial disparities among Black men with PCa stem from cultural factors, including healthcare system mistrust, and financial constraints. Given these disparities, it is crucial to examine whether they extend to modern diagnostic approaches like mpMRI. mpMRI has emerged as a tool for lesion visualization and risk stratification through PI-RADS scoring. Recent work further highlights the growing interest in MRI-based risk stratification and the need for equitable diagnostic pathways across diverse populations.²¹

However, few studies have examined mpMRI radiologic feature differences between Black and White PCa patients. Additionally, equitable access to mpMRI as part of the diagnostic pathway must be addressed, given that structural barriers may limit its availability for historically underserved populations. In a study conducted by Sundaresan et al.,²² Black patients were consistently less likely to undergo mpMRI. Such delays in PCa imaging and diagnosis subsequently lead to delays in the definitive treatment.

It is also important to note that virtually no data exist regarding mpMRI or PI-RADS performance in Asian, Hispanic, or other racial/ethnic groups. The current evidence base is almost entirely limited to comparisons between Black and White men. The data for Asian men are particularly sparse but suggest opposite performance characteristics. In a multicenter study, Asian American men were less likely to be diagnosed with csPCa on targeted biopsy (37% vs. 48%, P < 0.001) and had lower detection rates for PI-RADS 3 lesions.²³ The PREVENT trial found that all PI-RADS 3 lesions in Asian men were negative for cancer and that Asian men had lower odds of overall PCa (OR: 0.79, P = 0.01) than White men.²⁴ Most Asian-specific data come from studies conducted in China, Taiwan, and Hong Kong rather than diverse U.S. populations. Data for Hispanic men are even more limited. One study found Hispanic men had similar odds of detecting csPCa compared to White men, but comprehensive validation studies are lacking.²⁵ This gap underscores the need for more inclusive imaging research to ensure that diagnostic tools perform equitably across diverse populations.

Overall, these findings highlight the need to further evaluate whether the current PI-RADS criteria, developed and validated predominantly in populations with fewer Black participants, adequately capture the spectrum of PCa presentation in diverse populations. Further research on this matter may lead to potential refinements to the scoring system to improve diagnostic equity.

Limitations include the narrative review design, heterogeneity in definitions, variable MRI protocols, and limited representativeness: several reports come from single institutions or limited geographic areas, limiting applicability to other healthcare systems and population subgroups.

5. Conclusion

Our practice-informing and hypothesis-generating results suggest that mpMRI and PI-RADS scoring may contribute to racial disparities in PCa diagnosis. Among Black men, our findings suggest that biopsies should be considered when mpMRI shows PI-RADS 3 lesions and even in the absence of target lesions.

To achieve diagnostic equity, better understanding of the biological factors influencing mpMRI appearance is necessary as well as ensuring that future mpMRI studies include adequate numbers of diverse patient populations. This may eventually result in refinements to PI-RADS criteria to better predict csPCa across different racial groups and improved diagnostic utility.

Author contributions

Conception and design of study: L.C., D.T., A.G., C.D.

Methodology, Statistical analyses and interpreted the results: L.C., A.G., C.D., D.T.

Management of database: L.C., A.G.

Writing—original draft: L.C., A.G., D.T.

Critical revision of manuscript for important intellectual content: all authors.

Supervison of project: D.T.

All authors reviewed and approved the final version of the manuscript.

The authors did not use generative AI tools for writing or analysis.

Data availability: All data were extracted from published sources; the extraction sheet is available upon request.

Ethical compliance

Ethics approval not required for narrative review.

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Data access statement

Research data supporting this publication are available from the authors.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

Paperpal was used to assist with grammar and language editing during the preparation of this manuscript. No generative AI tools were used for data analysis, interpretation, or content creation.

Funding

There are none to declare.

Conflicts of interest

All the authors declare no competing interests.