Associate Professor Department of Pharmacy Practice, Manipal College of Pharmaceutical Sciences, Manipal Academy of Higher Education Manipal, India
Background: Post marketing surveillance data is very important to visualize the real-world scenario. It provides the idea regarding the safety issues of the drug after release into the market. The FAERS database is used for spontaneous adverse event reporting. However, manual mining of data requires longtime and presently available tool such as openVigil also has many limitations.
Objectives: To develop a technology assisted data mining tool to estimate the safety profiles of the drugs using FDA-spontaneous adverse event reports system (FAERS) and to validate the same using the remdesivir data
Methods: We have developed a data-mining tool through python script, which connects through google-mob as an interface to reduce the time constraints of this process using data from the FAERS database. Remdesivir was taken as a model drug to validate the tool. Retrieval of raw data, deletion of duplicates and statistical analysis has been conducted by developed python script within a very less time (approximately 2 hours). For the validation of the developed tool, the same process has been done manually which takes much more time (approximately 5-6 months). We have conducted a retrospective post marketing pharmacovigilance study using FAERS for Remdesivir. The spontaneous reports reported from their first marketing approval to 31st December 2021 were retrieved. Duplicate reports were deleted based on the deduplication criteria suggested by FDA. A disproportionality analysis using 2 by 2 contingency table and Reporting odds ratio (ROR) has been conducted to find out the signals. The data obtained manually is comparable with data-mining tool
Results: A total of 12777 adverse event reports from 1212 unique adverse events (Pt’s) were extracted for remdesivir after deduplication. The most common adverse events which showed signals with high ROR were investigated for a relationship with literature findings. The major reactions were reported in an age group of 65-80 years (n=4499; 35%), followed by 45-64 (n=3824; 30%) years. The male domination was observed (n=7715; 60%) with a major contribution form United states (n=11129; 87%). The python script took two hours to complete the whole work. The same work has been performed manually and it took 5 months and 20 days. A total of 302 signals have been identified for remdesivir. The validation was successful and the data mined through the python script and manually were 100% matching in its all aspects. This confirms the 100% accuracy of the developed tool.
Conclusions: This tool will help to reduce the manpower and time resource required for the disproportionality analysis. The interface or application developed from this tool will be useful for the clinician and researchers to develop the safety data of various drugs.
