Background: Healthcare information databases are available in Japan for drug safety evaluations. Adverse drug reactions(ADRs) can be predicted from diagnosis codes, medications/procedures, or laboratory test results that are available from the databases.
Objective: The validity of prediction of statin-related ADRs from healthcare information database was evaluated by comparing with doctor's diagnoses and abnormal laboratory test results in patients receiving statin therapy.
Methods: A Drug Use-Results Survey database for statins was used. The predictability of well known ADRs caused by statins, such as rhabdomyolysis and liver disorder, from laboratory test results available from the database was evaluated. Abnormal results were observed; creatine kinase(CK) levels increased to 10 times the normal upper limit (1,500 IU/L or higher) and aspartate aminotransferase/alanine transaminase(AST/ALT) levels increased to 3 times the normal upper limit (120 IU/L/135 IU/L or higher). Positive predictive values(PPV) were calculated using doctor's diagnoses as the gold standard.
Results: Among 26,849 participants in post-marketing studies on statins (study period: 1989–2008), 64% were females and 81% were aged between 50 and 79 years. Four cases of rhabdomyolysis were reported (0.017%). CK levels were elevated in 20 patients and the PPV was 10.0%. Two hundred and four cases of liver disorder were reported (0.890%). AST/ALT levels were elevated in 98 patients and the PPV was 44.9%.
Conclusion: PPVs of statin-related ADRs from abnormal laboratory test results were relatively low and many false-negative or false-positive cases were observed. Both quantitative and qualitative information is required for conducting database analyses of drug safety profiles. Each ADR case should be described in detail, and the threshold of abnormal values should be examined to increase the validity of ADR predictions. (Jpn J Clin Pharmacol Ther 2013; 44(3): 193-200)