Purpose
The “case‐population” design has been proposed for the surveillance of rare events like Stevens‐Johnson syndrome/toxic epidermal necrolysis (SJS/TEN), wherein a registry of cases is combined with sales data from the source population in order to estimate crude odds ratios (ORs). A major drawback of this method is the inability to distinguish between new and non‐new users of drugs, which for the study of hypersensitivity reactions is of utmost importance.
Methods
We have explored an approach in which the exposure to the drugs of interest in the source population is inferred from a primary health care database (BIFAP), which helped us to identify drug initiators among all users and additionally adjust for potential confounders. A total of 44 SJS/TEN cases from the Registry and 44 000 controls randomly sampled from BIFAP and matched with cases for index date were included. We estimated the adjusted ORs (AORs) and 95% confidence intervals (CI) of SJS/TEN associated with the new use of 13 drugs (for which we had at least two exposed cases) through a conditional logistic regression model.
Results
AORs (95% CI) were estimated for phenytoin, 4618 (434‐49112); cotrimoxazole, 1142 (163‐8015); allopurinol, 160 (36‐709); dexamethasone, 38 (1.33‐1077); ibuprofen, 33 (8.6‐124); lorazepam, 27 (5.8‐124); paracetamol, 13 (2.8‐62); levofloxacine, 12 (1.24‐120); amoxicillin, 6.9 (1.39‐35); pantoprazole, 6.5 (0.10‐420); metamizole, 6.3 (0.69‐57); amoxicillin clavulanic acid, 4.2 (0.53‐34); and omeprazole, 1.34 (0.06‐31). The inclusion of non‐new users dramatically decreased the AORs for all drugs.
Conclusions
The case‐population approach using a registry of cases and a primary health care database proved feasible and efficient for the active surveillance of SJS/TEN.