Baseline results of a living systematic review for COVID-19 clinical trial registrations

Background: Since the coronavirus disease 2019 (COVID-19) outbreak was first reported in December 2019, many independent trials have been planned that aim to answer similar questions. Tools allowing researchers to review studies already underway can facilitate collaboration, cooperation and harmonisation. The Infectious Diseases Data Observatory (IDDO) has undertaken a living systematic review (LSR) to provide an open, accessible and frequently updated resource summarising characteristics of COVID-19 study registrations. Methods: Review of all eligible trial records identified by systematic searches as of 3 April 2020 and initial synthesis of clinical study characteristics were conducted. In partnership with Exaptive, an open access, cloud-based knowledge graph has been created using the results. Results: There were 728 study registrations which met eligibility criteria and were still active. Median (25 th, 75 th percentile) sample size was 130 (60, 400) for all studies and 134 (70, 300) for RCTs. Eight lower middle and low income countries were represented among the planned recruitment sites. Overall 109 pharmacological interventions or advanced therapy medicinal products covering 23 drug categories were studied. Majority (57%, 62/109) of them were planned only in one study arm, either alone or in combination with other interventions. There were 49 distinct combinations studied with 90% (44/49) of them administered in only one or two study arms. The data and interactive platform are available at https://iddo.cognitive.city/. Conclusions: Baseline review highlighted that the majority of investigations in the first three months of the outbreak were small studies with unique treatment arms, likely to be unpowered to provide solid evidence. The continued work of this LSR will allow a more dependable overview of interventions tested, predict the likely strength of evidence generated, allow fast and informative filtering of relevant trials for specific user groups and provide the rapid guidance needed by investigators and funders to avoid duplication of efforts.


Introduction
The urgent need for immediate solutions to tackle public health emergencies of international concern undermines effective coordination and collaboration between researchers 1,2 . Since the coronavirus disease 2019 (COVID- 19) outbreak was reported in December 2019, a huge number of clinical studies have been initiated, firstly in China, South Korea, then subsequently in Europe, Japan, North America and Australia. The number of trials recorded in clinical trial registries is proliferating exponentially, with many independent research efforts designed to answer similar questions. Whilst multicentred trials can potentially improve generalisability of study results across widely divergent endemic settings, this requires comparable study designs and comparators. Conversely, independent trials with small sample sizes and underrepresentation of some patient populations limit the acquisition of substantive evidence to assess the effects and applicability of the many diagnostic methods, therapeutic and prophylactic strategies targeted to address the COVID-19 pandemic 3 . As the research landscape becomes increasingly saturated, better coordination in study design, dosing strategies and pooling data for subsequent metanalyses, are needed to ensure definitive answers can be translated into clinical practice as rapidly as possible. Tools that allow researchers to review the diversity of interventions already underway are critical in facilitating collaboration, harmonisation, partnerships and open access research.
Methods to distil the current research environment can help identify studies likely to be adequately powered to answer research questions, key research gaps that still need to be addressed, as well as duplication of efforts. Furthermore, as results accrue it becomes clear which scientific questions can only be addressed by pooling and standardising individual patient-level data across different trials. Since early 2000 there has been progressive demand from scientific journals and funders for investigators to register trials to monitor health research conducted on humans; over 21 registries are now available 4-8 . Although trial registries do not capture all research conducted for a specific theme, they provide an accepted source and proxy of the volume of research either proposed or underway 9, 10 . To support these efforts and to provide clarity regarding study design, including objective measures of quality and power to generate reliable evidence of the safety, efficacy, and utility of interventions, a living systematic review permitting regular and maintained search updates of registered COVID-19 clinical trials is underway 11 .
The primary objective of this living systematic review is to provide an open, accessible and frequently updated resource summarising the characteristics of COVID-19 clinical study registrations. Herein, we present the results of the baseline review of all eligible trial records identified by the systematic searches as of 3 April 2020 and initial synthesis of clinical study characteristics.

Protocol
The living systematic review protocol was prospectively designed and published elsewhere 12 , containing extensive details of the methodology and rationale for use of the living method.

Eligibility criteria
All clinical trial registrations either being planned, or currently underway to diagnose, treat or prevent COVID- 19 were eligible for inclusion in the baseline results of this review. This included patients known to be infected with COVID-19 as well as healthy volunteers, healthcare workers or other patient populations where health related outcomes were assessed in the context of COVID-19. The review was not limited by outcome, language or intervention given the desire to capture all clinical trial registrations planning to or currently evaluating any COVID-19 diagnostic, prevention or treatment modality. Eligibility was not restricted to clinical trials only; no limitations on study design were applied so that all research and observational studies would be incorporated, including retrospective studies. All trial registry records included in the review are hereafter referred to as clinical 'study' registrations.

Information sources and search strategy
Formal searches for clinical trial registrations were conducted as detailed in the protocol publication 12 up until 23 March 2020. Due to heavy traffic on the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) which aggregates records from 17 country and regional trial registries 4 , on 24 March 2020 this central information source was no longer accessible to anyone outside of the WHO. A download of all COVID-19 trials from the ICTRP database was made available to us on 1 April 2020, and the latest export publicly available for download on 3 April which was derived using the search terms ((COVID-19) OR (novel coronavirus) OR (2019-ncov)) is the last search update used for inclusion in the baseline results of this review. The source registry records which supplemented all included studies identified through the WHO ICTRP, were accessed live on the dates of data extraction for each record and not from an archived version.

Search and update schedule
This article presents the baseline results of the systematic review. Searches for this living systematic review are performed weekly and updates to the review following data extraction of newly identified and included records will be made available via the IDDO COVID-19 Living Systematic Review Website as completed 13 . Searches and updates will continue for the foreseeable future during 2020. This article will be updated every 6 months. The rationale for the search schedule is provided in the study protocol 12 .

Study records, data items and outcomes
The process for study selection and data for all eligible trial records were identified and extracted as per the published protocol in a standardised, pre-piloted REDCap database 12,14,15 . This included automated imports of WHO ICTRP data using Trifacta® Wrangler (Trifacta Inc. USA), blinded screening of trial records for eligibility by at least two reviewers [BM, SR], and data entry of additional manual variables completed by one reviewer with each variable cross-checked for quality control by a second reviewer. No records were excluded at the title level, each record was reviewed in its entirety either as reported in WHO ICTRP or the source registry before assessment of eligibility. All data items including their classifications, assumptions, rules and definitions for extraction are presented in the variable and data dictionaries available as extended data 13 . These data, along with all of the results data, analysis scripts and supplementary tables and figures have been deposited and are available from an open access repository or from the IDDO COVID-19 Living Systematic Review Website 13 .
Definitions of the study designs manually assigned to records in this review were as per the Cochrane Consumers & Communication Review Group Study Design Guide 16 . Interventional studies were categorised as 'randomised control trials (RCT)'; 'quasi-randomised trials'; or 'non-randomised interventional studies'. For observational and epidemiological research one of the following study designs were selected: 'cohort-study'; 'case report'; 'case series'; 'case-control'; 'cross-sectional'; or 'prognostic'. For any other study designs, including health services research, behavioural and social science studies, the category of 'other' was attributed. A final study design of 'diagnostic test accuracy' was assigned where the primary outcome was assessment of diagnostic methods. The World Bank list of economies (June 2019) was used for income classification of countries 17 .
The taxonomy of interventions was adapted from the initial categories designed by The CEBM Oxford COVID-19 Evidence Service 18 and extended for any new treatments identified. At the highest level, interventions within each study arm were categorised as: "Pharmacological interventions"; "Traditional Chinese Medicine"; "Vaccine prevention"; "Vaccine treatment"; "Advanced therapy medicinal products (ATMP)" [defined as a gene therapy medicinal product; a somatic cell therapy medicinal product; or a tissue engineered product 19 ]"; "Behavioural intervention" [including social science studies], "Placebo", "No Intervention" [where explicitly reported or zero details provided], "Standard of care " [where standard of care, routine treatment/therapy, best supportive care, general treatment, conventional treatment, standard treatment or related synonyms were explicitly reported], "Diagnostic intervention"; or "Others " [including nutritional supplements and enteral feeds; physiotherapy and exercise; physical therapies, such as renal replacement therapy; psychotherapies]. Pharmacological interventions and ATMP categories were further sub-classified in alignment with the CEBM taxonomy (see variable dictionary for further details) 13,18 . Without available expertise and a lack of standards in the reporting, further examination of "Traditional Chinese Medicine" interventions was not undertaken.
The baseline data for this analysis were not supplemented by further details from the investigators regarding ambiguous registration details or unreported trial features.

Synthesis of results
Descriptive statistics were used to present the extracted data. Categorical variables were summarised with proportions and frequencies; continuous variables were summarised with totals, means, quartiles, minimums and maximums. Summary statistics and figures were produced using Stata 14.2 (StataCorp, College Station, TX, USA) and R software (version 3.6.3, The R Foundation for Statistical Computing, Vienna, Austria).
In partnership with Exaptive, an open access, cloud-based knowledge graph has been created using the baseline systematic review results 20,21 . Leveraging Exaptive's virtual collaboration platform, the "Cognitive City", built upon a graph-based architecture at its core, uses the REDCap database from the living systematic review to allow stakeholders to query the data generated. Ongoing search updates are scheduled to be incorporated into this open, online, searchable tool as they become available.

Risk of bias
The living systematic review of registered trials does not address specific clinical questions. Baseline results are limited to the initial data extraction, descriptive analysis of trial characteristics and visualisation of clinical trial registry records. Accordingly, assessment of meta-biases or the strength of the body of evidence represented by included records are not relevant. In a first effort to distil the hierarchy of evidence being generated for COVID-19 interventions, registration details have been used to make a preliminary assessment of the strength and potential bias in relation to the design and intended execution within individual studies using the Oxford Centre for Evidence-Based Medicine 2011 Levels of Evidence 22 .

Study selection
A total of 819 unique trial registry records were screened and assessed by at least two independent reviewers against the prespecified eligibility criteria. The PRISMA flow diagram and summary of the literature screening is provided in Figure 1. Of the 819 unique records, 29 records were excluded because they did not include COVID-19 patients, and were not conducted in the context of COVID-19. Of the 790 records that met eligibility, 62 clinical trials were cancelled after registration and therefore excluded from further analysis.

Study characteristics
The most common registered study design identified was randomised control trial (RCT), with 294 studies identified ( Figure 2A). Other study designs frequently identified were: case series (N=109), non-randomised interventional studies (N=105), and cohort studies (N=98). There were relatively few studies classified as diagnostic test accuracy (N=26), case-control (N=23), cross-sectional (N=21), prognostic (N=17) or quasi-randomised (N=5). In total 29 (4%) studies did not contain sufficient information in the registry to determine the study design. A total of 435 studies involved a treatment intervention, 56 involved a diagnostic intervention and 38 involved a preventative intervention.
Most studies identified had proposed relatively small sample sizes; the median (25 th , 75 th percentile) sample size for all studies was 130 (60, 400). Reported sample sizes were generally low across all study designs, with median (25 th , 75 th percentile) sample sizes of 134 (70, 300) for RCTs and 60 (24, 120) for non-randomised interventional studies ( Figure 2B). Cohort studies tended to have larger sample sizes with median (25 th , 75 th percentile) 245 (100, 800). Of the 728 included studies, 207 plan to assess a pharmacological intervention, 132 plan to assess a traditional Chinese medicine, 48 plan to assess an advanced therapy medicinal product; 39 plan to assess a diagnostic intervention and 28 trials plan to assess a behavioural intervention. There were six studies identified in this search planning to assess a preventative vaccine and three studies planning to assess a vaccine treatment. Reviewers considered 317 (44%) of studies to be level 2 on the Oxford CEBM levels of evidence scale, 188 (26%) level 3; 216 (30%) level 4 and 7 (1%) level 5. At the time of data extraction, 54% (392) of identified studies were registered as currently recruiting participants.
Planned location of studies. The majority of registered studies (74%, 527/728) involved a single study site with only 25% (185/728) of studies planning to recruit participants from multiple sites in the same country and 2% (16/728) of studies planning to recruit participants from multiple countries. The vast majority of registered studies (80%, 581/728) planned to recruit participants from China ( Figure 3). Only four studies included in the review included sites located in low income or lower middle income countries, with only 17% (8/47) of lower middle income countries represented in the studies and 0% (0/31) of low income countries. Conversely, 35% (28/80) of high income countries and 30% (14/47) of upper middleincome countries were represented by proposed recruitment sites.

Eligible populations.
Discerning whether at-risk populations were eligible for inclusion in the study was not always possible from the registry information (Supplementary Table 1 -see extended data 13 ). Of the risk groups investigated, 365 (50%) studies explicitly reported planned exclusion of pregnant women, 9 (1.2%) studies planned inclusion of pregnant women, and for the remaining 354 (49%) studies there was insufficient information to classify the study. A total of 110 (15%) studies planned to include children (<18 years), in 491 (67%) studies children were not eligible for inclusion and in 127 (17%) studies there was insufficient information on age eligibility. Most study registrations contained insufficient information to classify the eligibility of patients with: diabetes mellitus (94% unknown); HIV (84% unknown); immunocompromised condition (82%); chronic lung disease or moderate to severe asthma (82%); hypertension (96%); serious heart conditions (77%); severe obesity (99%).
Number of patients planned to be enrolled were available for 63% of study arms (244/385) with median (IQR, Range) of 40 (20-80, 4 -10000) participants with highest sample sizes planned in RCTs. Irrespective of study design the most commonly reported intervention was antivirals, while the largest sample size was planned for antimalarials ( Figure 4).

Pharmacological and ATMP Interventions assessed in RCTs
Of the 178 randomised control trials, there were a total of 293 arms assessing 80 pharmacological and ATMP interventions ( Table 1). Among RCTs, the most commonly used drug categories were antivirals (134 arms in 70 trials), antimalarials (64 arms in 44 trials), ATMP (Advanced Therapy Medicinal Products) (38 arms in 29 trials) and monoclonal antibodies (22 arms in 16 trials). Among antivirals assessed in RCTs, ritonavir featured in the most arms (56 arms in 33 trials with an overall planned sample size of at least 2,905 patients) followed by lopinavir (50 arms in 32 trials with an overall planned sample size of at least 3,130 patients) and interferons (36 arms in 24 trials, with an overall planned sample size of at least 2,142 patients). The two antimalarials assessed were hydroxychloroquine (46 arms in 33 trials, planned sample size of 13,549 patients) and chloroquine (19 arms in 17 trials, planned sample size of 11,216 patients). Of the ATMPs assessed in RCTs, the most frequently assessed were convalescent plasma treatment (7 arms in 7 trials, planned sample size of 345),   Reducing the fragmentation of results via graph-based data structures Robust extraction of the results described above was hindered by the incompatible and inconsistent data fields, categorisation, and naming conventions across the many source trial registries. This lack of harmonisation limits the fidelity of results pooled across these sources. To overcome this and ensure reliable results, we implemented manual review of source registry records, further data extraction, and data harmonisation. This process of harmonisation has further enabled mapping of the results to a graph-based data structure that links data across key variables.
Data have been linked to support structured comparison of data, making results easier to visualise and interpret. Using the example of drug naming conventions, Figure 5a shows the highly fragmented outputs of raw drug names extracted from WHO ICTRP, compared to the harmonised network enabled by the standardised categorisation of the IDDO COVID systematic review data in Figures 5b and 5c.
Multiple interactive dashboards are built on top of the graph database to provide a simple interface to query the repository of clinical study data. More advanced interactive network visualizations allow for ad-hoc exploration of the underlying data. Data can be searched and filtered to exploit patterns based on the open meta-data generated by the systematic review, including type of interventions, treatment arms, study design, sample size, country, registry source, and strength of evidence, to find relevant trials that fit their specific criteria. The data and interactive platform are available at https://iddo.cognitive.city/ 21 .

Discussion
The COVID-19 clinical trial landscape is highly dynamic, changing on a daily basis. We present a comprehensive review of all registered trials up until 3 April 2020. The results highlight an unprecedented amount of research planned in a remarkably short period of time. While the mobilisation of the scientific community is admirable, this review illustrates a lack of coordination and duplication of effort that will ultimately delay    23 . These estimates were derived based on the distribution of outcomes in the population studied and will not apply to all settings but sample sizes of similar magnitude may be expected. A significant reduction in the sample size needed may be achieved through the adaptive approach to study design 24 , appropriate covariate adjustment 25,26 and the use of an ordinal rather than binary outcome 27 and a number of studies with such designs have been recently planned/started. While many studies are testing pharmacological treatment (n=204/728), a large number of trials (n=132/728) are testing traditional medicines, especially in China. A limited number of studies are testing diagnostic tools or are observing social parameters affected by the pandemic. Expectedly at this early stage of the research, very few vaccine trials were reported. The vast majority of studies are planning to recruit participants from high income and upper middle income countries, with only four studies planning to recruit participants from lower middle income and low middle income countries. This observed imbalance may reflect both challenges in funding mobilisation and the epidemiological pattern of the diseases of these regions. However, interpretation of this finding is nuanced as the number of trials from resource limited settings has increased over recent weeks. Similarly, focus areas and interventions for research are expected to change as research priorities evolve. This comprehensive analysis of data reported in trial registrations could help to identify research gaps and duplication, as well as guide funding priorities.
As with research on other novel interventions, vulnerable populations such as pregnant women, children or patients with co-morbidities are often excluded from trials and thus are under or unrepresented. Only nine studies indicated that pregnant women were eligible for inclusion. Reliable assessment of the inclusion of vulnerable populations among registered studies was hampered by the limited information included in study registries. The eligibility of pregnant women was unknown in roughly half of studies and the eligibility of the other at-risk groups was unable to be ascertained from the study registry in over 75% of registered studies.
Medical product quality These data provide vital evidence for understanding the rapidly changing landscape for what efficacious medical products will hopefully soon be available for the management of COVID-19 infection. As such evidence accrues, without global planning for production, distribution logistics, post-market surveillance, and actively detecting and acting on the inevitable substandard and falsified products, we risk a global crisis in access to and quality of anti-COVID medicines. Up to 7.8 billion people will be in dire need of a limited supply 28 . There has already been a deluge of public health problems from quackery to substandard and falsified products, such as chloroquine, masks, diagnostic tests, 'vaccines' and hand sanitiser (see IDDO Medicine Quality Monitoring Globe) 28,29 . A 'living' system will help identify the most promising medical product interventions to prompt crucial pro-active planning for coordinated production and logistics for equitable access, optimal and appropriate prevent, detect and respond monitoring of product quality 30 and trials of portable devices for empowering post-market surveillance 31 .

Limitations and challenges
Although two individuals (at least one of whom was a medical doctor) reviewed each included trial record, extraction of manual variables was not blinded. Any changes made during extraction to the reported values such as study design categorisation were made on review of the totality of information provided within the entire registry record. While the rationale of the WHO International Standards for Clinical Trial Registries is undisputed 32 , our review highlights substantial heterogeneity, inconsistencies and perceived errors in the recorded information across source country and regional registries. For example, as different trial registries have different classification systems for study design, there were 76 unique descriptions of study design as collected by the WHO ICTRP, with 66 studies containing no information. To allow a common classification scheme for comparisons across registries, each study was manually reattributed to one of 11 study design categories. Inevitably there is a degree of subjectivity in this process leading to possible errors in classification within our database; however, this was minimised by ensuring that all categorical attributes were clearly and prospectively standardised, defined and documented prior to extraction. Other explanations for observed inconsistencies in the original reported registrations include language barriers preventing accurate completion of study registrations in English, human error inherent in urgent registration during the pandemic, poor understanding of the terminology, and intentionally vague or poor reporting of information.
While the current WHO International Standards for Clinical Trial Registries outlines minimum reporting guidance of what should be collected in registration records, consensus on how these details are collected limits the current utility of this information source alone. For this reason, caution should be raised in the interpretation of many initiatives currently presenting the raw WHO ICTRP registry data. Without harmonisation of data fields and common taxonomies and definitions, records across different source registries do not align.

Sustainability and future work
As a living systematic review, weekly search updates are being conducted and have already identified more than 2,000 additional clinical trials registered since the 3 April 2020, that are being conducted in the context of COVID-19 (as at 26 May 2020). Cancellations also add to the constant flux in the number of registrations, with almost 10% of registered studies being cancelled before commencement; the majority of these occurred in China where a decline of COVID-19 cases has resulted in few patients eligible for enrolment. This further emphasises the need for collaboration particularly within countries and regions as many isolated trials initiated will similarly encounter a lack of patients available for enrolment as the outbreak's infectivity changes over time. The WHO ICTRP was used as the primary source for this baseline review and while 3 source registries were also formally searched manually, some relevant study records from other registries published prior to our search cut-off date were not yet accounted for in the available WHO ICTRP export. As an example, while analysing the baseline results we were made aware of an additional 107 clinical trials in the Iranian Registry of Clinical Trials (IRCT) that were not reflected in WHO ICTRP and are therefore not included in this analysis 33,34 . Almost six months after COVID-19 was identified, the expansion of clinical research is far from slowing. Hence a living systematic review method is appropriate and sustained updates will be necessary for the foreseeable future. Ongoing search updates are scheduled to be incorporated into the open, online and searchable IDDO COVID-19 Cognitive City platform.
During completion of the first phase of this project, initiatives with parallel objectives were identified 35 . Of the few online clinical trial visualisation tools that are similarly going back to source registries and manually reviewing records for extraction, there are noticeable differences in approach and agenda, but also synergies across some variables. Standardisation and harmonisation of these comparable variables, matched with an ethos of open sharing of extended data from all projects could help to minimise the duplication of work across research organisations. Therefore, future work is planned to assess these synergies across tools and assess the feasibility and willingness for coordination and collaboration moving forward. Such an interchange could only enhance the utility of our collective efforts.
Furthermore, additional work planned, as requested by collaborators from the COVID-19 Clinical Research Coalition, will include a categorisation of trials assessing affordable and readily available interventions for deployment in low resource settings that are feasibly adaptable to such health care systems. At a later stage, this living review will allow prioritisation of research targets for individual patient data meta-analysis and will support other COVID-19 Clinical Research Coalition efforts 36 .
The continued work of this living systematic review will allow a more dependable overview of interventions tested, predict the strength of evidence likely to be generated, allow fast and informative filtering of relevant trials for specific user groups and provide the rapid guidance needed by investigators to avoid duplication of efforts.   Table 1) -Baseline_Table2.xls (Baseline results version of Table 2) -Baseline_Supplementary Table 1.pdf (Supplementary  table of eligible populations for inclusion) -Baseline_Supplementary_file_IDDO COVID19_DB_ VariableDictionary.pdf (Variable dictionary for data extraction of included records)R script for cross verification of study level summary.R (R script to generate a basic study level summary with unique number of studies by country and eligibility of risk populations) -R script for generation of Figure 4.R (R script to generate Figure 4) Perhaps article updates more frequently than every 6 months would be warranted, given the rapidly evolving state of the evidence.

5.
Discussion around other existing databases with a similar focus on creating a living repository of clinical studies planned and underway may be added. One example is COVIDevidence.

6.
Perhaps a call for more standardization and consistency across study registrations is appropriate, given the observation that there was significant variability in data available.

7.
For eligibility criteria: consider explicitly specifying that you are interested in treatment studies for those with both suspected and confirmed SARS-CoV-2, primary and secondary prevention studies involving those without primary SARS-CoV-2 infection and those with a previous infection, and diagnosis studies across all populations.

8.
For the 29 records excluded for not including COVID-19 patients, this seems to run counter to the eligibility criteria specified: "All clinical trial registrations either being planned, or currently underway to diagnose, treat or prevent COVID-19 were eligible for inclusion in the baseline results of this review. This included patients known to be infected with COVID-19 as well as healthy volunteers, healthcare workers or other patient populations where health related outcomes were assessed in the context of COVID-19". This is likely semantics, but great if this could be clarified.

9.
This is not the main focus of the analysis, but it may be interesting to see the reasons for termination for the trials identified, and what interventions were being examined.

10.
For 'Eligible populations', on what basis were the given comorbidities selected? They are certainly relevant, but wonder if these are the ones most commonly reported?

11.
Great discussion re: many small studies likely being insufficiently powered for major outcomes of interest.

12.
Great discussion re: vulnerable populations. I think rather than their being underrepresented, patients with comorbidities are likely under-reported in study registrations.

13.
Great discussion re: sustainability and future work. 14.
There is a clear role for this living review in the evidence ecosystem along with other living evidence synthesis efforts around emerging evidence on treatment, prevention, diagnostics and prognostics. Perhaps mention regarding the role of this living review in the context of this broader evidence ecosystem would be helpful. 15.

Julia C. Cutts
Burnet Institute, Melbourne, Australia This living systematic review represents an important resource to enable researchers, clinicians, and other stakeholders to gain synthesized, up-to-date information on COVID-19 clinical trial registrations. Importantly, a living systematic review approach is highly justified in this context given the frequency with which new COVID-19 clinical trials are currently being registered. Important findings gleaned from this baseline results study include the identification of relatively small proposed samples sizes for most of the registered studies, the relative dearth of studies to be conducted in low or lower middle-income countries, the insufficiency of information regarding eligibility of study participants, the large number of studies testing traditional medicines, and the limited number of studies testing diagnostic tools or examining social parameters of the pandemic. As well as tabulating and graphing key characteristics of the registered COVID-19 clinical trials, the authors present engaging "Cognitive City" visualizations of the registered trials by intervention.
Conclusions drawn in the paper are supported by the results presented and the limitations and risks of bias are clearly addressed.
The paper is scientifically sound, but I would recommend the following minor changes to improve its clarity: Whilst detailed in the review protocol, I would suggest that the information sources and search strategy is briefly summarized in this manuscript. For example, whilst it is inferred from the flow diagram and the text in the "Information sources and search strategy" section that ICTRP was the primary source of information, I would suggest including a more explicit statement to this effect. Table 1: In this table the data are stratified by study design, but the current text formatting makes it difficult to discern the individual study design sections. For example, it is difficult to see where the data for "Case Series" ends and the data for "Cohort/Non-randomised Studies" begins. Perhaps reformat so that the study design headings are differentiated from the Drug names (e.g. using indenting or variation in bold or italics, as much as the journal's prescribed table formatting will allow). Alternatively data for each study design could be presented in separate tables, but this may be too cumbersome. Medical product quality section within Discussion, page 18: The final sentence (below) needs to be amended. The gist of it is understandable, but it is grammatically flawed (see underline) and would perhaps be easier to digest if broken up into two sentences. "A 'living' system will help identify the most promising medical product interventions to prompt crucial pro-active planning for coordinated production and logistics for equitable access, optimal and appropriate prevent, detect and respond monitoring of product quality and trials of portable devices for empowering post-market surveillance." Is the living method justified?