Twenty-year trends in the racial and ethnic composition of big randomised controlled trials for diabetes




The generalizability of study results and health equality may be jeopardized if Black, Indigenous, and People of Color (BIPOC) enrolment are not sufficiently representative. The purpose of this study was to investigate trends in BIPOC group enrolment in diabetic randomized controlled trials (RCTs) and the relationships between trial characteristics and high BIPOC group membership.


In order to find big diabetic RCTs with a sample size of around 400 participants published between 2000 and 2020, we conducted a systematic literature search. We evaluated the temporal patterns in the trials’ enrolment of racial and ethnic groupings. In order to examine the association between trial characteristics and the high enrolment as measured by the median enrollment rate, logistic and linear regression analyses were performed.


For analysis, 405 RCTs in total were included. The median enrollment rate for BIPOC groups was 24.0%, with the Black group having a rate of 6.4%, the Hispanic group of 11.2%, the Asian group of 8.5%, and the other BIPOC groups of 3.0%, respectively. Between 20.1 and 28.4% of BIPOC enrollees in diabetic RCTs over the past 20 years have shown an increase in trend (P for trend = 0.041). There was a discernible upward trend in Asian group enrollment. We discovered that weekly or daily intervention frequency (OR = 0.48, 95% CI: 0.26, 0.91) and duration of intervention > 6.5 months (OR = 0.59, 95% CI: 0.37, 0.95) were significantly related to decreased odds of high-enrollment, while type 2 diabetes was associated with high enrollment of BIPOC groups (OR = 1.44, 95% CI: 1.04, 1.99).


Over the past 20 years, it has been observed that BIPOC enrolment has increased in significant diabetic RCTs; various trial variables may be significantly related to BIPOC enrollment. These results might emphasize the value of including BIPOC groups and offer guidance for the planning and execution of the next clinical trials for diabetes.


As an example, diabetes prevalence was significantly higher in Black, Indigenous, and People of Color (BIPOC) groups compared with White participants in the USA and UK [4,5,6,7,8,9]. The prevalence of diabetes also differs significantly by racial and ethnic groupings [1,2,3]. Due to their socioeconomic background, health insurance, level of education, religious convictions, and language problems, among other factors, the BIPOC participants generally had a lower rate of diabetes diagnosis and required greater attention for the quality of treatment and care [9,10,11,12]. While the majority of participants in clinical trials, particularly those conducted in western nations, were White groups [13], only a small percentage of newly approved drugs revealed differences in drug exposure and response across racial and ethnic groups, and only a few of these cases led to recommendations for treatment tailored to particular racial and ethnic groups [14]. As a result, a lack of diversity in BIPOC enrollment may jeopardize the applicability of study findings and health equity [15].

Clinical research requires adequate enrollment and analysis plans for racial and ethnic groups, according to the National Institutes of Health (NIH) [16, 17]. The appropriate enrollment could be calculated using the participation to prevalence ratio, which is the percentage of BIPOC among trial participants divided by the percentage of BIPOC in the overall diabetes population [8]. The current practise in many randomised controlled trials (RCTs)—including those for cancer [18,19,20], stroke [21], systemic lupus erythematosus [22], cardiovascular disease [23,24,25,26,27], obesity [28], acute pain [29], coronavirus disease 2019 [30], and vaccine [31]—remains suboptimal despite the significance of representative racial and ethnic enrollment. Similar to this, a recent study found that BIPOC groups were underrepresented in 62% and 78%, respectively, of the diabetic RCTs done in the US and UK [8]. However, there was a lack of data in diabetic RCTs about the present situation and historical trend of BIPOC groups’ overall enrolment.

So, with a sample size of about 400 participants, the purpose of this study was to assess the current situation and temporal trend in the participation of BIPOC groups in significant diabetic RCTs. We also looked into any connections between trial-related variables and high enrollment of BIPOC groups. This study’s research protocol was submitted to PROSPERO for registration (International Prospective Register of Systematic Reviews, CRD42021229100).


Search technique

The study’s methods were described in detail in other publications [8]. We used the terms “diabetes mellitus” and “randomized controlled trials” to thoroughly search the following electronic databases: the Cochrane Library, MEDLINE (through PubMed), and EMBASE. Between January 1st, 2000, and December 31st, 2020, the search was conducted. Google Scholar, ClinicalTrials, and the Clinical Trials Registry Platform of the World Health Organization were also searched. We also looked for more potential research in the reference lists of the highlighted papers. Multiple pre-searches were used to develop the search techniques, which are displayed in Supplemental Table 1 for MEDLINE.

Criteria for inclusion and removal

We considered type 1 and type 2 diabetes RCTs with more than two individuals of different racial and ethnic backgrounds. Other requirements for inclusion were being over 18 and having a sample size of at least 400 people, as trials with smaller sample sizes were more likely to be single-center, early-stage research. We only included the trials reporting main outcomes with baseline data for the entire population. There were some publications from different stages of a trial, different subgroups, or exploratory analyses. Because gestational diabetes may be a temporary form and theoretically return to normal under glucose control, we omitted trials that specifically addressed this condition [32, 33]. Trials that pre-defined particular BIPOC was also disqualified; for example, a trial that only enrolled Black people was ineligible. Additionally, we eliminated duplicates, abstracts from conferences, comments, and letters, studies published in languages other than English, studies without full-text access, and studies without data extraction.

Studying choice

Before choosing possibly pertinent papers, two reviewers (J Zhang and Y Wang) separately screened and assessed the titles and abstracts of the search results after removing duplicates using software and manual inspection. The whole texts were then scrutinized by the two reviewers, who then made the final trial selection. Before screening the literature, a pilot test was carried out to make sure that every reviewer understood the process and inclusion and exclusion criteria. Conflicts about the choice of papers were resolved by speaking with a third reviewer (G Li). A flow diagram (Supplemental Fig. 1) showed the complete study selection process [34].


The status and temporal trend of overall BIPOC enrolment from the included RCTs were the main outcomes. The secondary results included trends in the enrolment of particular BIPOC groups, such as Black, Asian, Hispanic, and other BIPOC groups.

The median enrollment rate was utilized to divide the included RCTs into high-enrollment trials and reference trials in order to further investigate the association between trial variables (detailed below) and high enrollment of BIPOC individuals.

Extraction of data

Using a standardized extraction sheet, two reviewers (J Zhang and Y Wang) separately collected the following information from the trials that were included: (I) percentage of enrollment by race and ethnicity (all BIPOC, Black, Asian, Hispanic, and other BIPOC groups); (II) publication information (first and corresponding authors, year of publication); (III) study specifics (type of RCTs, type of diabetes, type of complication, sample size, age, sex, research goal, country or region of trial coordination office, outpatient enrollment, source of funding); and (IV) intervention and follow-up specifics (type of intervention, frequency and duration of intervention, type of follow-up, frequency and duration of follow-up). Discussions were held with a third reviewer to resolve disagreements (G Li).

Year of publication, sample size, type of diabetes, enrolment site, kind of RCTs, funding source, intervention, and follow-up information were trial factors of interest for analysis in this study. The results of the eligibility requirements or enrollment process included patient age, sex, as well as race, and ethnicity. The age and gender of the patient were therefore excluded from analyses to investigate the connection with an enrolment of BIPOC.

Analyses of statistics

We discussed categorical variables with counts and percentages as well as continuous variables with a median and interquartile range (IQR). The kernel-weighted local polynomial smoothing curve was used to depict the BIPOC enrollment percentages with an ascending publication year. The enrollment of BIPOC participants was examined over time using the Jonckheere-Terpstra proportion trend test to determine whether there was a significant trend.

In order to investigate the relationship between trial characteristics and high enrolment of BIPOC groups, we utilized univariate and multivariable logistic regression analyses. Considering the exploratory character of our analysis, all trial factors were incorporated into the multivariable logistic model to maintain all potential components in the model. Prior to including the continuous trial components in the model, we used median values to dichotomize them in order to increase clarity and interpretability. To quantify the association, odds ratios (ORs) and the related 95% confidence intervals (CIs) were used. With beta coefficients (s) and 95% confidence intervals (CIs) displayed for the relationship, we also conducted a sensitivity study using linear regression analyses to evaluate the correlations between trial parameters and the continuous BIPOC enrollment rate.

STATA software (version 16.0) was used for all data analysis, and a two-sided P value of less than 0.05 was regarded as statistically significant.


In the initial literature search, we found 18,278 records, of which 1463 were retrieved for full-text analysis. Over two-thirds of the 405 RCTs included in our analyses were published within the last ten years (2010–2020). The 405 trials included data on 84.4% of the Black group, 72.8% of the Asian group, 27.2% of the Hispanic group, and 89.1% of the other BIPOC groups. The factors of the included RCTs are listed in Table 1. 19.3% of the included RCTs were conducted in a single country, while 80.7% were conducted in numerous nations. For the multi-country RCTs, Supplemental Table 2 lists the nation or region where the trial coordinating office is located, with 23.2% in North America and 14.7% in Europe. 85% of the trials had type 2 diabetes as their primary focus. The included trials’ median sample size was 716 (IQR: 527–1246), their female representation was 46% (IQR: 39.9–50.1%), and their average age was 58. (IQR: 55.0–60.8). Glycemic control (31%), management (43%), and diabetic complications (26%) were the trial’s primary objectives. Drug intervention was studied in 86% of the trials, and the average length of the intervention was 6.5 months (IQR: 6.0–13.5). The majority of the trials (74.9%) were funded by the pharmaceutical business and had face-to-face follow-ups, with a median follow-up time of 12 months (IQR: 6.0-18.0).