Methods
Trial sample
We included interventional SARS-CoV-2 treatment efficacy trials registered on ClinicalTrials.gov with a trial site in the United States (USA) and a start date between 2020–01-01 and 2020–06-30. We excluded trials evaluating interventions aimed at the prevention of SARS-CoV-2. Prevention trials typically enroll healthy volunteers, which was not the population of interest in our study. We included all eligible trials in our cohort; therefore, no sample size calculation was performed. Data were downloaded from the web front-end of ClinicalTrials.gov on 2020–12-01 and again on 2021–01-04, allowing us to evaluate patient-participant recruitment results at the 6-month mark from date of trial start (see [11] for further elaboration of our methods).
Data curation
We evaluated patient-participant recruitment for each trial in our cohort. Pre-specified criteria for unsuccessful recruitment were: (i) trial was “terminated” or “suspended” for a reason unrelated to efficacy, safety or progression of science (an automated extraction of the “why_stopped” element from ClinicalTrials.gov trial registration records was performed, enabling this categorization); (ii) trial was “completed” or “active, not recruiting” with a final enrollment less than 85% of the anticipated enrollment cited in the registration record at trial start, thus reflecting a substantial loss of statistical power for the primary outcome [4]; or, (iii) trial was “recruiting” or “enrolling by invitation” and the recruitment period had been extended to at least twice as long as anticipated, based on the planned recruitment length at trial start.
We aggregated trial recruitment goals on a state-by-state basis, and then compared this to the number of contemporaneous active SARS-CoV-2 cases in each state between 2020–01-01 and 2020–06-30. This involved (i) estimating the goal patient-participant enrollment per state for each trial (dividing planned enrollment per trial by the total number of sites per trial and multiplying this by the number of trial sites per USA state); (ii) aggregating total patient-participant enrollment goals per state; (iii) estimating the number of active SARS-CoV-2 cases per state between 2020–01-01 and 2020–06-30 using data downloaded from usafacts.org; and, (iv) calculating the proportion of SARS-CoV-2 cases that would need to be recruited to fulfill aggregate trial enrollment goals by state. Multisite trials with at least 1 site in the USA were included in our analysis. Patient-participant enrollment numbers were prorated up to 2020–06-30 for trials not achieving primary outcome completion by this date.
Outcomes and statistical analysis
We report the proportion of trials with successful patient-participant recruitment, as well as the proportion of SARS-CoV-2 cases required to fulfill trial recruitment goals by state. For all states with two or more active trials during our 6-month time period, we performed a simple and multivariable regression analysis evaluating the relationship between the proportion of trial sites within each USA state with unsuccessful patient-participant recruitment and: (i) the proportion of cases required to reach state recruitment goals; (ii) state population based on data from the US Census [12]; and, (iii) number of trial sites per state. We hypothesized that increased state recruitment goals would result in an increase in unsuccessful patient-participant recruitment. To investigate this further, a robust regression was performed using the M-estimation with Huber weights [13]. This specific type of regression analysis was chosen due to the continuous nature of the dependent variable and our detection of influential observations based on the Cook distance. The results of regression analysis were presented as beta coefficient (b) with 95% confidence interval (95% CI). Robust regressions were performed using the robustbase package [14], R version 4.1.0 (2021) [15]. We defined p < 0.05 as statistically significant.