Risk of Stevens-Johnson syndrome and toxic epidermal necrolysis in new users of antiepileptics
Mockenhaupt, M., Messenheimer, J., Tennis, P., & Schlingmann, J. (2005). Risk of Stevens-Johnson syndrome and toxic epidermal necrolysis in new users of antiepileptics. Neurology, 64(7), 1134-8.
BACKGROUND: Estimates of risk of Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) associated with some antiepileptic drugs (AEDs) have used denominators based on the number of prescriptions or daily doses. Because the risk of SJS is highest in new users of drugs, the use of denominators reflective of all users can lead to low estimates of risk associated with drugs. In this study, risk in new users is assessed. METHODS: Data on all hospitalized patients with SJS and TEN with use of carbamazepine (CBZ), lamotrigine (LTG), phenobarbital (PHB), phenytoin (PHT), or valproic acid (VPA) were obtained from the German Registry for Serious Cutaneous Reactions. For 1998-2001, the numbers of new users were estimated from number of dispensed prescriptions in Germany, the average prescribed doses, and the duration of use in the Mediplus database (IMS Health) Germany, and assumptions that relate new use to growth in national dispensings. To minimize the probability of underestimating risk in new users, conservative estimates of new use that were somewhat lower than predicted from national prescription data were used. RESULTS: More than 90% of SJS and TEN cases occurred in the first 63 days of AED use. Over the 4 years, increases in dispensing were 5% for CBZ, 65% for LTG, 6% for PHB, -16% for PHT, and 26% for VPA. Across a range of assumptions about frequency of incident use, the risk estimates vary between 1 and 10 per 10,000 new users for CBZ, LTG, PHT, and PHY and were consistently lower for VPA. CONCLUSION: Across a range of assumptions used, the risk of hospitalization for Stevens-Johnson syndrome or toxic epidermal necrolysis in new users is low for carbamazepine, lamotrigine, phenytoin, phenobarbital, and valproic acid. Because conservative incidence use fractions were used, it is likely that some risks were overestimated.