Analysis of clinical and methodological characteristics of early COVID-19 treatment clinical trials: so much work, so many lost opportunities

Since December 2019, SARS-CoV-2 has caused a global outbreak of a respiratory illness termed coronavirus disease (COVID-19). COVID-19 ranges from mild, self-limiting respiratory tract illness to severe progressive pneumonia, multiorgan failure, and death [1, 2]. To date, there are no therapeutic agents specifically designed for the treatment of COVID-19. Fierce medical research is currently underway, however, there are historical reasons that led us to question if the global research community is maximizing the expected benefit from these efforts [3]. To shed some light on this question we decided to assess the clinical and methodological characteristics of treatment clinical trials set forth as an early response to the COVID-19 pandemic.

Previous analyses of COVID-19-related studies have been published focusing on scientific articles on COVID-19, and registered studies beyond clinical trials [4]. We aim to focus on clinical trial records, providing a deeper insight into the methodological characteristics that help to assess the capability of clinical research in a pandemic context. COVID-related clinical trials have already been recognized to lack features that optimize their scientific value [5]. We intend to add a further perspective providing a follow-up of the early-phase research and review the output of early COVID-19 treatment trials.

We found 693 records of clinical trial protocols (supplementary material), 648 (93.5%) on COVID-19, and 45 (6.5%) on COVID-related conditions (notably pulmonary rehabilitation or exercise and mental health among health care workers).

Among COVID-19 trials, 572 (88.3%) evaluated treatment only (Table 1), 68 (10.5%) evaluated prophylaxis only, and eight (1.2%) evaluated both. Our results focus on the 580 trials evaluating COVID-19 treatment.

On January 30th, 2020, COVID-19 was declared a public health emergency of international concern and on March 11th, it was declared a pandemic. As of our initial search date, April 19th, 1,603,209 cases had been reported worldwide, 54,225 patients were in a serious or critical state, and 169,750 had died [13]. These numbers continued to increase hourly as the world raced to mitigate the impact of the disease. Our results suggest that a considerable clinical research effort was mobilized early on against COVID-19. The hundreds of trials conducted worldwide offer some hope. To maximise the efficacy of research in infectious disease epidemics, research must be fast, flexible, and integrated with the frontline response [14].

Adaptive clinical trials may be particularly useful in the current pandemic [15, 16] as they enable sample sizes and allocation ratios to be refined, treatments or doses abandoned, and focus moved towards patients with a higher likelihood of benefit. A few large trials, notably, the RECOVERY trial and the WHO Solidarity trial, are underway and use an adaptive design. Currently, preliminary results are available confirming the usefulness of this type of trial design in providing informative evidence, namely showing positive results for dexamethasone [17] and lack of benefit for hydroxychloroquine [18] and lopinavir/ritonavir [10]. Another example is the Adaptive COVID-19 Treatment Trial, (ACTT) [19], sponsored by the National Institute of Allergy and Infectious Diseases (NIAID) that led remdesivir to receive a conditional marketing authorisation by the European Medicines Agency (EMA) and an initial Emergency Use Authorization (EUA) followed by formal approval by the US Food and Drug Administration (FDA). With these few exceptions, the research we found seems largely insufficient to provide clear answers for core outcomes in the shortest possible time. However, it is important to consider that although limitations were present from the onset, several protocols were later revised based on the evolution of the pandemic and increasing knowledge of the disease. The overall impact of protocol modifications is beyond the objectives of the present study. Nevertheless, based on the final results of several studies, most limitations present from the onset seem to have remained and affect the final publication.

Our results suggest that patients at the largest risk of death due to COVID-19 were not being prioritised in clinical trials. This is an important missed opportunity since the high baseline risk means that smaller treatment effects would be more easily detected [3]. Therefore, participants with a higher likelihood of benefit should not only be the focus of clinical research from an ethical point of view but would be the most efficient population to study in order to identify which treatments are worthwhile pursuing and which are not. Although the focus on high-risk patients would have been an important strategy to adopt in early trials, in later trials the inclusion of non-selective patients would be essential to ensure the external validity of results.

While most treatment trials use a randomised controlled design, we found that 92 (15.9%) were single-arm trials. These trials are a matter of concern, namely as CFR changes considerably over time, and trials using historical controls will likely lead to more false-positive findings [20]. Also of concern is the fact that most treatment trials (58.3%) were not multicentric. It is well documented that evidence from single-centre trials is more prone to bias compared with multicentre trials, and tend to provide larger treatment effects [21]. This is particularly clear in the critical care setting, where many positive single-centre trials have been contradicted by subsequent multicentre trials [22]. Furthermore, few RCTs conducted in intensive care units and using mortality as a primary endpoint show a beneficial impact of the intervention on the survival of critically ill patients [23].

Another essential aspect of trial design is the choice of endpoints. While we recognize the importance of surrogate endpoints, which allow faster results to be obtained when compared with core clinical endpoints [24], the lack of hard and more pragmatic endpoints such as death and length of hospitalisation are causes for concern. To eschew these more clinically relevant endpoints is a methodological mistake that is hard to understand. High-quality information, as straightforward and generalisable as possible, will be key to moving forward. For research to permit informed clinical decision-making, this will have to change, and trials must use uniform disease-related definitions. Of note, several core outcome sets have been defined since the trial protocols were first registered [25]. However, these were published after the first wave of the pandemic in the northern hemisphere, and therefore beyond the scope of time during which these might have been most useful. Nevertheless, many treatment trials are a priori already deemed not to detect statistically significant and clinically relevant results. The median target sample size of assessed COVID-19 pharmacotherapy trials is 100 participants. This is manifestly insufficient to detect anything but an extremely large treatment effect. For example, regarding trials with mortality as a primary endpoint, only three trials were powered to detect a difference of 50% or more between treatment groups (assuming a baseline risk of 2%). Even assuming a baseline risk of 10%, only 38 trials were adequately powered to detect a difference of 50% or more between treatment groups. In fact, half of the trials evaluating mortality have a sample size of 112 participants or fewer and are only able to identify a treatment effect on mortality over 90% (irrespectively of the baseline risk), which is unrealistic (supplementary material). Although the results of underpowered trials can be meta-analysed, thereby increasing the power to detect a treatment effect, this approach remains inefficient in terms of time and overall resources. The results of systematic reviews and meta-analysis can be threatened by underpowered trials due to publication bias and related types of small-study effects [26]. On the other hand, the planned trial duration is also relatively short (median of 184 days for overall treatment trials), given the uncertainty regarding the natural history of COVID-19.

Some of our concerns have been increased by the early termination of a clinical trial studying remdesivir, conducted in 10 hospitals in Wuhan, China [27]. This trial was to enrol 453 participants but as the disease in the area was brought under control the number of eligible patients became too small, and recruitment was stopped at 236 participants. This led to reduced power in the trial, of only 58%, while it was intended to be 80%. This trial failed to demonstrate any difference in time to clinical improvement with remdesivir (HR 1.23 [95% CI 0.87–1.75]), while, later on, the ACTT trial, an adaptive, larger, and adequately powered trial succeeded in showing the usefulness of remdesivir in hospitalized patients (rate ratio for recovery 1.32 [95% CI 1.12–1.55]) [19].

We found that only 18.9% of treatment trials were industry-funded. Although trials undertaken on the initiative of investigators are important, the absence of experienced industry-supported trials means that there is considerable room for improvement and effort by many of these multinational corporations. Although we are aware of important efforts by the industry to maintain pharmacotherapies available during the crisis, these entities have the resources and experience available to provide an important contribution in conducting high-quality, high-output clinical research.

In mid-March 2020, the majority of COVID-19 cases worldwide were no longer from continental China, and as of April 19th, 2020, only around 4% of cases were from China [13]. These figures are in clear contrast with the fact that so many treatment trials (47.6%) were conducted exclusively in China. Previous research has suggested that a large proportion of clinical trial data submitted to support new drug registrations in China may be considered to be incomplete or substandard [28]. Therefore, despite the large push in research, we question if the upcoming flood of data is of high enough quality to produce clear answers to critical questions at early stages and the necessary recruitment capability. We are also worried that considerable efforts were employed in the 92 treatment trials and combined planned target samples of 18,892 participants studying the effect of TCM on COVID-19. We would urge that these resources be used more wisely.

The main pharmacotherapies under investigation were chloroquine/hydroxychloroquine, antivirals, notably lopinavir/ritonavir, and monoclonal antibodies. We found no substantive methodological differences between trials evaluating these interventions. However, there are differences in patients’ clinical characteristics, with monoclonal antibody trials allowing the inclusion of more patients with several and critical illness, as well as participants with relevant comorbidities. Antiviral trials include relatively fewer patients with several and critical illness, and largely excluded participants with COVID-relevant comorbidities.

As of September 14th, 2020, 6 months after the pandemic was declared, only a minority of these early trials made results available to the public. Additionally, according to planned completion dates, 251 trials should have been completed by this date, while only 33 accomplished this commitment. This corresponds to much lower figures than anticipated, considering the trials’ planned follow-up. Moreover, the information on the recruitment status of the trials often seems to not be updated in the registries, rendering it difficult to interpret the real state of research on this topic. The current available overall results and status from these early 580 treatment trials reinforce our initial worries about the overall inadequacy of these trials to provide clinically relevant conclusions.

Our study has several limitations. First, the data presented focus exclusively on registered trials. Similarly, phase 1 trials may be underrepresented. Second, there is a significant amount of missing or unsubmitted data for certain data fields, which limits the completeness of the analyses and thus the interpretability of the results presented. Third, given the rush to conduct more research, there may be trials underway that had not yet been registered, an important issue given the common nature of retrospective registration [7]. Fourthly, the quality of the available records was largely poor, with inconsistencies and errors throughout. We think that most of the issues have been resolved, though we cannot be certain that nothing was missed.

With the hundreds of trials enrolling thousands of people currently underway, a more efficient and useful approach would be for research bodies such as the WHO, NIH, Inserm, etc. to create a coordinated research response to face the pandemic. The EMA has called for similar efforts [29]. We understand that there are political, ethical, administrative, contractual, regulatory, logistic, economic, and societal factors that may hinder research, though these difficulties should be overcome in times of global crisis. Persisting on the path of isolated investigations will likely only lead to futile trials and more death on a global scale. Given the large numbers of people with COVID-19, and the recent push for more real-world evidence, we consider that there is an urgent need for a global high-quality COVID-19 patient registry, which could be used to detect large beneficial effects [30] and provide relevant evidence for health care decision-making [31].

Initiatives such as the WHO R&D Blueprint aim to tackle the challenge of generating new evidence during disease outbreaks. We believe that clinical research must be integrated as an essential element of a coordinated international response to epidemics. As those are exceptionally difficult contexts for clinical research, tools such as adaptive protocols that could feasibly be integrated into clinical practice, as well as global research networks and platforms, may be of great help to produce informative research. Due to the unpredictable features of new outbreaks, continued enrolment throughout different locations should be advocated, this would enable the inclusion of sufficient numbers of participants as well as the combination of research efforts [32].

The current treatment and prophylaxis options are few and built upon very scarce and fragile data. With the proper forward planning, critical questions could have been answered earlier. As clinical investigators, we have the obligation to adjust and improve the research being conducted. The world was not ready to react with the appropriate research to a pandemic. We believe that the scientific community, the pharmaceutical industry, and research agencies could have done better.