Is Our Science Representative? A Systematic Review of Racial and Ethnic Diversity in Orthopaedic Clinical Trials from 2000 to 2020

Abstract

Background: A lack of racial and ethnic representation in clinical trials may limit the generalizability of the orthopaedic evidence base as it applies to patients in underrepresented minority populations and perpetuate existing disparities in use, complications, or functional outcomes. Although some commentators have implied the need for mandatory race or ethnicity reporting across all orthopaedic trials, the usefulness of race or ethnic reporting likely depends on the specific topic, prior evidence of disparities, and individualized study hypotheses.

Questions/purposes: In a systematic review, we asked: (1) What proportion of orthopaedic clinical trials report race or ethnicity data, and of studies that do, how many report data regarding social covariates or genomic testing? (2) What trends and associations exist for racial and ethnic reporting among these trials between 2000 and 2020? (3) What is the racial or ethnic representation of United States trial participants compared with that reported in the United States Census?

Methods: We performed a systematic review of randomized controlled trials with human participants published in three leading general-interest orthopaedic journals that focus on clinical research: The Journal of Bone and Joint Surgery, American Volume; Clinical Orthopaedics and Related Research; and Osteoarthritis and Cartilage. We searched the PubMed and Embase databases using the following inclusion criteria: English-language studies, human studies, randomized controlled trials, publication date from 2000 to 2020, and published in Clinical Orthopaedics and Related Research; The Journal of Bone and Joint Surgery, American Volume; or Osteoarthritis and Cartilage. Primary outcome measures included whether studies reported participant race or ethnicity, other social covariates (insurance status, housing or homelessness, education and literacy, transportation, income and employment, and food security and nutrition), and genomic testing. The secondary outcome measure was the racial and ethnic categorical distribution of the trial participants included in the studies reporting race or ethnicity. From our search, 1043 randomized controlled trials with 184,643 enrolled patients met the inclusion criteria. Among these studies, 21% (223 of 1043) had a small (< 50) sample size, 56% (581 of 1043) had a medium (50 to 200) sample size, and 23% (239 of 1043) had a large (> 200) sample size. Fourteen percent (141 of 1043) were based in the Northeast United States, 9.2% (96 of 1043) were in the Midwest, 4.7% (49 of 1043) were in the West, 7.2% (75 of 1043) were in the South, and 65% (682 of 1043) were outside the United States. We calculated the overall proportion of studies meeting the inclusion criteria that reported race or ethnicity. Then among the subset of studies reporting race or ethnicity, we determined the overall rate and distribution of social covariates and genomic testing reporting. We calculated the proportion of studies reporting race or ethnicity that also reported a difference in outcome by race or ethnicity. We calculated the proportion of studies reporting race or ethnicity by each year in the study period. We also calculated the proportions and 95% CIs of individual patients in each racial or ethnic category of the studies meeting the inclusion criteria.

Results: During the study period (2000 to 2020), 8.5% (89 of 1043) of studies reported race or ethnicity. Of the trials reporting this factor, 4.5% (four of 89) reported insurance status, 15% (13 of 89) reported income, 4.5% (four of 89) reported housing or homelessness, 18% (16 of 89) reported education and literacy, 0% (0 of 89) reported transportation, and 2.2% (two of 89) reported food security or nutrition of trial participants. Seventy-eight percent (69 of 89) of trials reported no social covariates, while 22% (20 of 89) reported at least one. However, 0% (0 of 89) of trials reported genomic testing. Additionally, 5.6% (five of 89) of these trials reported a difference in outcomes by race or ethnicity. The proportion of studies reporting race or ethnicity increased, on average, by 0.6% annually (95% CI 0.2% to 1.0%; p = 0.02). After controlling for potentially confounding variables such as funding source, we found that studies with an increased sample size were more likely to report data by race or ethnicity; location in North America overall, Europe, Asia, and Australia or New Zealand (compared with the Northeast United States) were less likely to; and specialty-topic studies (compared with general orthopaedics research) were less likely to. Our sample of United States trials contained 18.9% more white participants than that reported in the United States Census (95% CI 18.4% to 19.4%; p < 0.001), 5.0% fewer Black participants (95% CI 4.6% to 5.3%; p < 0.001), 17.0% fewer Hispanic participants (95% CI 16.8% to 17.1%; p < 0.001), 5.3% fewer Asian participants (95% CI 5.2% to 5.4%; p < 0.001), and 7.5% more participants from other groups (95% CI 7.2% to 7.9%; p < 0.001).

Conclusion: Reporting of race or ethnicity data in orthopaedic clinical trials is low compared with other medical fields, although the proportion of diseases warranting this reporting might be lower in orthopaedics.

Clinical relevance: Investigators should initiate discussions about race and ethnicity reporting in the early stages of clinical trial development by surveying available published evidence for relevant health disparities, social determinants, and, when warranted, genomic risk factors. The decision to include or exclude race and ethnicity data in study protocols should be based on specific hypotheses, necessary statistical power, and an appreciation for unmeasured confounding. Future studies should evaluate cost-efficient mechanisms for obtaining baseline social covariate data and investigate researcher perspectives on current administrative workflows and decision-making algorithms for race and ethnicity reporting.

Fig. 1

This flow diagram demonstrates the search strategy and inclusion criteria. Criteria included English-language studies, human studies, RCTs, publication date from 2000 to 2020, and studies published in Clinical Orthopaedics and Related Research; The Journal of Bone and Joint Surgery, American Volume; or Osteoarthritis and Cartilage. All results that were not full-text articles, not RCTs, and without patients were excluded.

Fig. 2

This graph shows the annual trend in the proportion of orthopaedic clinical trials reporting race or ethnicity from 2000 to 2020.

Fig. 3

This graph shows the sample’s racial or ethnic representation of RCTs in the United State compared with the racial and ethnic representation reported in the 2019 United States Census.