Visualization analysis of the characteristics of COVID-19 clinical trial

Objectives: This article points out the characteristics and trends of COVID-19 clinical trials through data collection, translation, mining and visualization to help in clinical trial design. Method: The registered data of COVID-19 clinical trials are gathered from the Chinese Clinical Trial Registry and ClinicalTrials.gov website transformed by Python, further demonstrated by visual tools. Results: As of 24:00 on March 28, 2020, totally 732 trial registration records have been retrieved. Overall, there are 406 (55.46%) interventional studies and 271 (37.02%) observational studies. Among interventional studies, 38.93% are randomized parallel trials, 55 (13.55%) trials considered time condition for clinical recovery as the primary endpoint, and 46 (11.33%) trials through clinical parameters and laboratory index as the primary endpoint. In the selection of intervention measures, chemical or biological agents was under the responsibility of 43.60%, of which antivirals accounted for 14.53%, antimalarials accounted for 8.87%, and 98 cases (24.14%) of studies involving Traditional Chinese Medicine or Integrated Medicine. In addition, joint network analysis of antivirals to explore the combination of drugs is further conducted. Conclusions: By Mining characteristic information of topical COVID-19 clinical trial registration, this article deserves further trial design ideas for researchers to enhance the effects.


Introduction
Since the discovery of pneumonia cases caused by COVID-19 in December 2019, the pandemic spreads rapidly, and medical researchers and clinicians have concentrated on this without hesitation.
On January 31, 2020, World Health Organization (WHO) declared in Geneva that COVID-19 was a Public Health Emergency of International Concern (PHEIC). 1 In order to test the effects of intervention, international researchers have been done a large number of related medical scientific research projects. Among them, clinical trials played an important role in the prevention and control of the COVID-19 in order to verify the efficacy of clinical treatment of drugs and form the necessary evidence. 2 As more and more medical institutions conduct clinical trials, there are also some researchers who research and evaluate them. Numerous experts and researchers analyze the data, but short of profound consideration. In view of this, this article further 3 explores the research characteristics and problems of COVID-19 clinical trial registration by processing and assessing the relevant data from the Chinese Clinical Trial Registration Center database and ClinicalTrials.gov website.

Materials And Methods
Data acquisition strategy 533 trial registration data were retrieved from the Chinese Clinical Trial Registry (ChiCTR) platform using "COVID-19" as the keyword. The source code of the pages was acquired, and rules for the collection of unstructured data were developed using programming methods. 3 Clinical trial registration fields obtained include: registration number, registration time, sample size, study type, study design, blinding, primary indicator, intervention and randomized method. In the ClinicalTrials.gov website, 199 relevant trial data were retrieved with the keyword "COVID-19" and all available field data were downloaded. All of the above data were included in 732 registrations as of March 28.

Data processing
This article uses Pandas, Openpyxl libraries and regular expressions in Python to normalize the data and replace the missing values, transform them into profitable discrete structures and construct data tables for subsequent analysis. 4 In order to prevent bias in the data screening and conversion process, the two authors of this article divided the data and compared them separately. The accuracy of the results is then ensured by consulting with the relevant medical staff on the disputed data. In a word, overall registrations in March are higher than the previous two months. 4 By analyzing the trend of overall registrations (Figure 2), it can be observed that the number was a generally uptrend over time, which has declined since March 17 and increased significantly after the 24. In addition, it was predicted that the coming period, registrations would go up continuously because of the activity of vaccine and unlisted drug clinical trials.  (Table 1). Ethics committee approval is a prerequisite for clinical trial registration, 5 but the timeliness of COVID-19 has resulted in some trials not being registered in time.

Study Types and Methods
Summary statistics for the study types are in Table 2. There are 406 interventional trials. The main research methods are Randomized Parallel Assignment (38.93%), Non-randomized Assignment 5 (5.33%), and Single Group Assignment (6.83%). Parallel control designs are observed experimentally at the same time, which is beneficial to eliminate errors caused by factors such as time and conditions. Since there is not any specific drug for COVID-19 and it has a longer survival period than other epidemics, most of interventional trials designed by parallel assignment. Among the 271 observational trials, Sequential Assignment (18.72%), Cohort (6.96%), and Factorial Assignment (4.51%) were preferable owing to a largish sample size. In addition, there were 28 Diagnosis, 10 Epidemiological and 17 other trials.  Blinding is one of the important tools for reducing subjects' perceptions of treatment allocation schemes between groups, reducing bias and improving the scientific and validity of trials. 6 In this statistic, as can be seen from Table 3, 268 studies were blinded and 464 (26.33%) were unspecified.
In addition, the subgroups of the study are counted in this article. 356 trials involved two groups; 285 trials involved a single group. Fifty-five and 22 trials involved groups three and four, with the remaining groups used less frequently. During COVID-19, most trials were divided into two groups for controlled studies. This grouping allows differences between groups to be efficiently assessed for "potent drugs" to deal with the virus.

Sample size
Sample size is an important aspect of clinical trial design, directly related to the reliability, reproducibility, and efficiency of the study. It was noted that the sample size of clinical trials was mainly concentrated in the range of 0-299, with a larger proportion of interventional and observational trials (Figure 3). Whereas, the number of observational trials increased with the expansion of the sample size interval in the larger sample size interval.

Primary endpoint of Interventional study
When designing a clinical trial, one of the most challenging and critical issues is how to select the primary endpoint used to assess efficacy. In view of the reliable evidence for benefits and risks should be provided, the primary endpoint is preferably a outcome measure that clearly reflects the benefit to the patients. 7 Among 406 interventional studies, time to clinical recovery accounted for 13.5%, clinical parameters and laboratory index for 11.33%, the change of pneumonia severity for 11.08%, questionnaire or scales for 9.85%, virus negative conversion rate of time for 8.13 %, cure rate for 7.39%, and the mortality for 6.16% (Figure 4).

Interventions
This article counted the interventions of 406 interventional studies. Details are shown in Table 4.   In terms of antiviral drug selection, most trials have chosen to combine LPV/r with interferon, a similar design approach that was used in clinical trials for the treatment of Middle East Respiratory Syndrome (MERS). 12 Whereas in critically ill COVID-19 adult hospitalized patients, study team did not observe a significant therapeutic effect of LPV/r compared to standard treatment. 13 Since novel coronaviruses

Combination of antiviral drugs and other drugs
are not yet available as specific drugs, the present study has visualized the use of antivirals in combination with other drugs to find multiple drug combination options, such as Favipiravir with Tocilzumab, Darunavir with Cobicistat, etc., to provide more ideas for drug combination design in relevant clinical trials.
Also, antimalarials (such as Chloroquine and Hydroxychloroquine) may be helpful for viral therapy. 14 Cell therapy and clinical trials related to plasma in recovered individuals are also ongoing.
Considering the fact that frontline health care workers also face greater psychological pressure under long, high-intensity, demanding and high-risk working conditions, some researchers mainly used health questionnaires to understand their psychological changes and the reasons for them, in order to establish timely mind-set adjustment programs in case of emergency epidemics. In general, there are more treatment options for COVID-19 than during SARS, 15 with a wider range of options and considerations for subjects.
Innovation of this article is that this article used data visualization to mine and analyze the registration characteristics and current progress of COVID-19 clinical trials, and finally to find potential antivirals combinations for investigators through a joint network. However, most of the trials are inconclusive, and we will be following them up continuously to examine the validity and reliability of their design in depth.

Conclusion
In this article, trial registration data from ChiCTR database and ClinicalTrials.gov website were collected and processed, followed by research and Visualization analysis of the 732 COVID-19 clinical trial registration features included. This article provides clinical investigators with recent trial design information and new ideas for early validation of efficacy, providing data support to guide the next phase of virus control efforts.

Declarations
Ethics approval and consent to participate  Statistics of primary endpoint Relationship between antivirals and other drugs