Random control selection for conducting high-throughput adverse drug events screening using large-scale longitudinal...

Area of Science:

• Pharmacoinformatics
• Computational Biology
• Health Informatics

Background:

• Large-scale longitudinal health data enables the detection of novel adverse drug event (ADE) signals.
• Existing control selection methods, like dynamic/super control selection, are computationally intensive due to repeated individual evaluations.

Purpose of the Study:

• To introduce and evaluate a novel random control selection approach for case-control pharmacoinformatics studies.
• To assess the efficiency and performance of the random control selection method compared to existing approaches.

Main Methods:

• A case-control design was employed using high-throughput pharmacoinformatics on longitudinal health data.
• A random control selection approach was developed, matching ADE case index dates with randomly generated control index dates.
• Performance was evaluated against dynamic/super control selection methods using two large datasets and a gold standard of 399 drug-ADE pairs.

Main Results:

• The random control selection approach demonstrated comparable F-scores (0.586-0.600) to dynamic/super control selection (0.545-0.562).
• The random control selection method was approximately 1000 times faster in preparing case-control datasets.
• This new approach requires individuals to be evaluated only once, drastically reducing computational load.

Conclusions:

• The random control selection approach is an efficient and effective method for detecting ADE signals in large-scale pharmacoinformatics studies.
• This method significantly reduces computational time and resources required for preparing case-control datasets.
• It offers a viable alternative for utilizing longitudinal health data in ADE signal detection.