JMIR Publications

ABSTRACT

Background:

With the improvement of drug evaluation system in China, an increasing number of clinical trials have been launched in Chinese hospitals. However the traditional clinical trial quality management models largely rely on human monitoring and counting, which can be time consuming and is likely to generate errors and biases. There is an urgent need to upgrade and improve the efficiency and accuracy of clinical trial quality monitoring system in hospital based research institutions in China.

Objective:

The objective of this study is to develop a digital monitoring platform that allows real-time monitoring, readout and warning of risk points throughout the entire lifecycle of clinical trials, on the basis of historical clinical trial quality control (QC) findings.

Methods:

Leveraging the Risk-Based Quality Monitoring (RBQM) mindset, we built a digital dynamic monitoring platform using big data analysis and automatic quantitative technology. Data from clinical trial QC reports generated during 2019-2023 in Beijing University Cancer Hospital, China, were used for training of automated classification tool, establishment of warning thresholds and validation of threshold values. Quality findings obtained during early stage, interim-stage and conclusion-stage QC of clinical trials were rated by 3 severity grades (minor, major, critical) and classified into 5 categories (with 4 taxonomy levels under each category). Text from QC reports were processed using an automated natural language processing (NLP) tool. All QC reports were grouped into 2 clusters using hierachical clustering analysis (HCA). QC findings from the relatively high-risk cluster (reports that were more likely to have major/critical findings as determined by experienced QC analysts) were used to determine warning threshold values for the monitoring platform, ie. a lowest number of findings was set to be the threshold value for each specific study stage + each level-3 taxonomy + each severity grade.

Results:

The most frequently reported Level-3 taxonomies in QC reports of 2019-2022 were Standard Procedure and Process, Safety Event Reporting, and Source Data Collection and/or Recording. A total of 189 warning threshold values were established based on data from 1380 QC reports during 2019 2022 to cover 3 severity grades, 21 Level 3 taxonomies and 3 rounds of QC. The warning thresholds were then applied to 211 QC reports generated in 2023, and 19.9% of those reports triggered warnings. Similar patterns of QC findings, including most frequently noted Level-3 QC findings, were observed in reports generated in 2023 compared to those obtained during 2019 2022.

Conclusions:

Application of this tool into clinical practice will enable automated monitoring and readout of risk points throughout all stages of clinical trials, accurately identify the most relevant trial procedure and/or function line, and notify quality management personnel in a real-time manner to take prompt actions and dynamically prevent recurrence of quality issues.