Of the retrieved articles, we screened 3647 and excluded 3622 mostly for not describing drug interventions for COVID-19 (Figure 1). Out of 25 full-text articles considered for potential inclusion, we excluded seven studies and in September 2022, we additionally excluded one retracted article and its corresponding data (see Figure 1) (49). All the eligible articles were in English and we did not identify any additional unique study references from the retrieved systematic review.
Study characteristics
Table 1 shows the data we analyzed from 17 RCTs which were published between 2020 and 2022; 53% (n=9) of the RCTs were published in 2020, 18% (n=3) were published in 2021, 24% (n=4) were published in 2022, while 6% (n=1) was a preprint. The median impact factor (IF) of 15 publications was 58 (10-79) regardless of the publication year, 18 (8-79) in 2022, 44 (14-88) in 2021, and 68 (57-75) in 2020. The preprint (50) and one published article (51) had no IF. The total number of participants was 17935. Regardless of treatment, there were 452 (3%) participants that identified as American Indian, 829 (5%) as Asian, 642 (4%) as Black, 5 (0.03%) as Native Hawaiian, 1556 (8%) as Hispanic/Latinx, 3973 (22%) as White, 534 (3%) as Other (424 [24%] reported in RCTs that were NIH/OMB race categories and 203 [1%] reported in RCTs that were not NIH/OMB categories including participants reported as North African [n=110, 54%], Sub-Saharan African [n=47, 23%], and Arab [n=46, 23%]), and 289 (2%) as multi-race. Race and ethnicity data were not reported for all participants, consequently 44% of the participants in the trials had no reported race due to missing data. One RCT (52) generally categorized self-reported ethnicities that were not Hispanic or Latino as Not Hispanic or Latino (n=293), which precluded this non-specific reporting from our analysis. The specific reporting of race and ethnicity from each study is shown in Table 1.
There were 15 (88%) parallel-group RCTs (9020 participants), 10 (59%) were double-blind (6618 participants), and 9 (53%) were phase 3 RCTs (7467 participants) [Tables 2 and 3]. All of the RCTs reported both male and female participants in RCTS with a median follow-up time of 29 days (IQR 28-45) in trials that reported a follow-up time (50-63). There were 11 (65%) RCTs with at least one US site (50, 52, 56-61, 64-66) [Supplementary Table 1]. RCTs conducted outside of the US (n=5, 29%) were mostly in European countries (Supplementary Tables 1-3). The majority (n=16, 94%) was registered on ClinicalTrials.gov (63, 67). None of the RCTs explicitly reported a requirement to speak or understand English written trial information.
Interventions
The drug interventions used in the RCTs varied (Table 1). The majority of the trials used remdesivir alone (51, 53, 61-66). The majority (n=14, 82%) of RCTs administered interventions once daily (50, 52, 53). Overall, the median duration of interventions was 10 (IQR 5-10) days. None of the studies mentioned patients' satisfaction with the intervention or comparator.
The median number of participants lost to follow-up treated with an intervention was 6 (IQR 0-18), while a median of 2 (IQR 0-18) participants were lost to follow-up for the control group (51, 53, 54, 64, 65). Placebo was the most common (n=10, 59%) comparator for RCTs (50, 51, 57-60, 62, 64-66) [Table 1], followed by standard care [n=3, 18%]. There were 8 (47%) RCTs with at least one industry-funder (50, 56-58, 61, 64, 65). Most (n=13, 76%) of the RCTs took place in an inpatient hospital setting (51-56, 59-63, 65, 66). Mostly severe COVID-19 requiring hospitalization not in conjunction with other COVID-19 disease severities was described in 6 (35%) RCTs (52, 54, 56, 62, 63, 65). Comorbid diseases were described in 13 (76%) RCTs, where the most common out of these was diabetes types I or II in 12 (92%) [Supplementary Table 4].
Outcomes
Predominantly, the main outcome measure was clinical status (Table 1) measured in 3 RCTs (17%) (51, 61, 65) on an ordinal scale (68) where one of these RCTs used a point 7-point ordinal scale (61) that included (1) death; (2) hospitalized, requiring invasive mechanical ventilation or extracorporeal membrane oxygenation; (3) hospitalized, requiring noninvasive ventilation or use of high-flow oxygen devices; (4) hospitalized, requiring low-flow supplemental oxygen; (5) hospitalized, not requiring supplemental oxygen but requiring ongoing medical care (related or not to COVID-19); (6) hospitalized, not requiring supplemental oxygen or ongoing medical care; and (7) not hospitalized . One RCT used a 4-point ordinal scale that excluded death, patients hospitalized not requiring supplemental oxygen, or non-hospitalized patients items from the abovementioned 7-point scale (51). The same RCT also used a 6-point scale that excluded the non-hospitalized patients not requiring supplemental oxygen item (51). The next most predominant outcomes were hospitalization (n=3, 18%) (52, 53, 62) and mortality (2 [12%] for in-hospital) (54, 63), 1 [6%] for all-cause mortality (56), and 1 [6%] for both all-cause and in-hospital mortality (55).
Reporting of race and ethnicity from RCTs
Study characteristics of the trials
Almost one-third (n=4, 24%) of trials did not report at least one race and ethnicity as demographic variables (Table 1). The median follow-up time of these RCTs was 28 days (IQR 16-74). Of the RCTs that did not report at least one race and ethnicity as demographic variables, 3 (75%) were funded by non-industry entities (55, 62, 63), while 1 (25%) did not describe a funding source (51). Two (50%) of the RCTs that did not report race and ethnicity as demographic variables were open-label (55, 63) and three (75%) had parallel group assignment (62, 69) intended for treatment or prevention (n=1, 25%) (63). The relative risk (63) or mean change from baseline (62, 69) both were the predominant (both n=2, 50%) measures of effect in RCTs that did not report race and ethnicity as demographic variables. The impact factor (one RCT was published in a journal without an impact factor (51) of the journals that did not report race and ethnicity variables was 60 (IQR 25-79).
One (6%) RCT registered on both ClinicalTrials.gov and the International Standard Randomised Controlled Trial Number registry (63) and one (6%) RCT did not report a registration number (51); neither RCT reported race or ethnicity as demographic variables. Two (50%) RCTs (55, 63) that did not report race had at least one in-hospital mortality outcome. Of the four RCTs that did not report race or ethnicity, remdesivir was the only studied intervention compared with either placebo (n=2, 50%) (62, 69), standard care (n=1, 25%) (52), or nothing (n=1, 25%) (Table 1) (63). Of the four studies that did not report race, the RCTs were mostly conducted in European countries (Supplementary Table 2) (70). Of the 13 (76%) RCTs that reported at least one race and ethnicity as demographic variables (50, 52-54, 56-62, 64-66), the median follow-up time was 29 days (IQR 28-45). The majority (n=8, 62%) was industry-funded (50, 56-58, 60, 61, 64, 65), 3 (23%) RCTs had non-industry funding (54, 59, 66), and 2 (15%) did not report a funding source (52, 53). RCTs reporting race and ethnicity mostly assessed the effectiveness of interventions with the hazard ratio (n=4, 31% (52, 56, 64, 65) [Table 4]. The RCTs reporting race were published in journals (one RCT was a preprint without an impact factor (50), with a median IF=56 (IQR 10-79). All of the RCTs that reported at least one race and ethnicity demographic variable were registered on ClinicalTrials.gov.
Hospitalization was the predominant (n=3, 23%) outcome in RCTs (52, 53, 62) reporting at least one race and ethnicity as demographic variables. Predominantly, remdesivir was the intervention in 5 (38%) RCTs (53, 61, 64-66) with 2 (15%) using standard care as a comparator (53, 61), and 3 (20%) using placebo (64-66). All of the industry-funded RCTs described at least one race and ethnicity as demographic variables (Table 2).
Of the studies that did report at least one race or ethnicity as demographic variables, the median intervention duration was 8 (IQR 5-10) days. Three (23%) of the RCTs that reported race and ethnicity did not describe an intervention duration (50, 52, 53).
Reporting of specific race and ethnicity from RCTs
American Indian people mostly participated in trials that used baricitinib (Table 5). Asian people mostly participated in RCTs involving remdesivir and standard care, while participants that identified as Black, Latinx, and White were mostly enrolled in RCTs that studied remdesivir. Participants that identified as multiracial were enrolled most often in trials that studied molnupiravir. The inclusion of different races and ethnicity varied by funder type (Table 3) and outcome (Table 4). The median intervention and follow-up durations were similar across the reported races and Latinx ethnicity (Supplementary Table 5). Of the underrepresented races in the RCTs, American Indian participants were in 4 (31%) RCTs where at least one site was in the US (54, 56, 58, 60, 66). Ten (77%) RCTs that included individuals self-reporting as Asian were conducted at 9 (62%) sites with least one in the US (50, 56-61, 65, 66), while 1 (8%) RCT was conducted in Canada (54). Participants that identified as Black were in 11 RCTs (85%), where 9 had at least one US site (50, 56-61, 65, 66) and 1 (15%) (54) was solely conducted in Canada. Participants that identified as Latinx were in 9 (53%) RCTs, where 8 (47%) had at least one US site (50, 52, 57, 58, 60, 61, 64, 66) and 1 (8%) RCT (51) was in Canada. One RCT (8%) (60) that had at least one US site reported participants’ race as Native Hawaiian or Pacific Islander. Participants that identified as White were in 11 RCTs (65%), where 9 (69%) RCTs had at least one US site (50, 56-61, 65, 66) and 2 (12%) outside the US; 1 in Canada (54) and 1 in several European countries (53). Out of the 4 (31%) RCTs that included multi-race participants, 3 had at least one US site (50, 57, 58). All reported races and ethnicity were reported predominantly in RCTs that were phase 3, double-blind, parallel, and placebo-controlled (Table 3). Native American, Native Hawaiian, Latinx, and multi-race participants were predominantly in industry-funded RCTs. Asian, Black, White, and other participants were mostly enrolled in non-industry RCTs (Table 3).
Outcomes of the studies that reported race and ethnicity
Half of the reported races were in RCTs that assessed the primary outcome with the odds ratio (Table 4). Regarding outcomes, Native American and Hawaiian participants were mostly assessed for progression to high-flow oxygen, non-invasive ventilation (Table 4). Time to recovery was assessed predominantly in Black and White participants as well as in participants that identified as other, while hospitalization or death was mostly assessed in Asian, Latinx, and multi-race participants.