Abstract
Background/Aims Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are rare, life-threatening cutaneous reactions that may occur following exposure to certain medications, including antiepileptic drugs (AEDs). No reliable incidence estimates of SJS/TEN for US populations have been calculated since the introduction of ICD-9 codes specific for SJS/TEN in October 2008. This study seeks to: 1) investigate the incidence of SJS/TEN using retrospective case identification, 2) test the positive predictive values of ICD-9 codes prior-/post-October 2008, and 3) pilot the feasibility of conducting a population-based pharmacogenomic study.
Methods With input from dermatologists, pharmacoepidemiologists, and pharmacogeneticists at FDA, HealthCore, and 11 HMORN sites, we validated ICD-9 codes associated with SJS/TEN in 3 cohorts: 1) patients with an inpatient code of 695.1x between 01/01/01–12/31/08; 2) codes 695.12–695.15 from 08/01/08–08/31/12; and 3) codes 279.5, 279.51, 279.53, 695.8, 695.81, 693.0x, 694.8x, 694.4x or diagnosis codes 692.9, E85x.x-E858.9, 693.8, 692.89, 695.89, 695.1x, 692.3x, 695.0 in an inpatient setting plus specific drug exposure between 01/01/01–08/31/12. We extracted a standardized clinical and laboratory dataset. A sample of 265 potential SJS/ TEN cases at 5 sites was selected for adjudication by medical chart review. Statistical models will be created to estimate the total number of SJS/TEN cases at all sites based on the positive predictive value of the case identification algorithm. To pilot the feasibility of a pharmacogenomic study, we estimated the number of living cases available, contacted potential cases, and tested 100 random DNA samples from the Marshfield Clinic Personalized Medicine Research Project for HLA-A*3101/HLA-B*1502 alleles to establish a control source.
Results A total of 54,049 patients was identified from electronic records: 3,058 in cohort 1; 1,733 in cohort 2; and 49,258 in cohort 3. Potential cases (n = 9) were identified and will be invited to participate in a genetic study. Among the 100 samples genotyped, 5 were positive for HLA-A*3101 and none were positive for HLA-B*1502, consistent with known frequencies.
Conclusions We identified a large number of potential SJS/TEN cases using code-based algorithms that will be used to determine the incidence of these events and to plan a larger follow-up study of specific AEDs and genomic factors.
