Use of a Data Mining Algorithm in Retrospective Analysis of

Use of a Data Mining Algorithm in Retrospective Analysis of Adverse Events

By Dr. Jacinta Aniagolu-Johnson

Objective:

The aim of this paper is to illustrate the usage of the MGPS algorithm through a retrospective analysis of the post-marketing data within the AERS data source for a couple of a withdrawn drug product – troglitazone (Rezulin).

Background: To review the functionality of MGPS in a real-world environment, the MGPS algorithm was applied retrospectively to the AERS data source to generate signal ratings (EBGM/EB05/EB95) for troglitazone that was approved treatment of type 2 diabetes mellitus in March 1997. Rare circumstances of reversible jaundice had been reported in the managed scientific trials that supported acceptance, but by October 1997, 35 situations of idiosyncratic hepatocellular damage (including 1 transplant and 1 death) have been reported to the FDA.

Strategies: Using the MGPS algorithm, a two dimensional evaluation was performed in a cleaned, de-duplicated and public discharge edition of the FDA AERS data source (through 4Q02) predicated on the next specifications: MedDRA PT-coded occasions; stratification by gender and age group; sub-setting by FDA season (year reports had been received by FDA); cumulative and discrete datasets by FDA receipt season; threshold minimum: EBGM 05 of > 1.0 for assessment; and drug search requirements by essential ingredient or generic name

Results: For the function term hepatic failing, an EB05 of > 2.0 was determined and maintained from 1998 (post-drug approval season) to withdrawal year (2000) and post-approval years (2001 and 2002); season of exception was 1997. For the function term hepatocellular harm, an EB05 of > 2.0 was determined and maintained from 1997 (medication approval season) to withdrawal year (2000) and post-approval years (2001 and 2002).

Conclusions: Overall, by particular FDA season(s) a statistical significant EB05 (> 2.0) was determined for the specified MedDRA PT keyphrases that the drug items were withdrawn from the marketplace. EB05 statistical need for >2.0 was reached and sustained (over the FDA years of post-approval advertising), either earlier or simultaneously that a regulatory actions (LC/DDL) was taken.

The same pattern was established for specific event terms including ALT Increased and Bloodstream Bilirubin Increased, expect that post-withdrawal year (in 2002) an EB05 of