Pilot study of prescription-event monitoring in Japan comparing troglitazone with alternative oral hypoglycemics

Objective: To examine adverse events reported in a pilot study of the presc ription-event monitoring in Japan (J-PEM) scheme, comparing troglitazone wi th other oral hypoglycemics. Methods: We used a cohort study with a concurrent control in which informat ion was gathered from both doctors and pharmacists. Crude event rates were calculated and compared between troglitazone (T) and alternative oral hypog lycemics (control drugs, C) using the likelihood ratio test. When the diffe rence was statistically significant, possible confounding mechanisms were e xamined using Poisson regression analysis. Results: Of 3115 patient codes registered, pharmacists were sent 2078 quest ionnaires and returned 1814 (87%), while doctors were sent 1858 questionnai res and returned 671 (36%). The difference in crude rates was statistically significant in 11 events (seven where T > C and four where C > T) reported by pharmacists and ten events (three where T > C and seven where C > T) re ported by doctors. Among those, in two events, "weight increased" (T > C) a nd "abnormal hepatic function" (T > C), significant differences were observ ed in data from both doctors and pharmacists. Regression analysis revealed that the difference in crude rates for "nausea" (T > C) was possibly due to an uneven distribution of genders and that for "weight increased" (T > C) was possibly due to an uneven distribution of compliance. Patients with hep atic function abnormalities associated with troglitazone could be divided i nto two subtypes: one with a slight increase in serum lactate dehydrogenase concentration only and the other with elevated serum alanine aminotransfer ase. Conclusions: Comparison of the event rates between troglitazone and control drugs, followed by regression analysis, revealed several features of adver se events associated with drugs, including possible confounding mechanisms. Troglitazone-induced hepatic function abnormalities may be divided into tw o subtypes.