Epidermal necrolysis (EN)--either Stevens-Johnson syndrome (SJS) or toxic EN (TEN)--is a severe drug reaction. We constructed and evaluated a specific algorithm, algorithm of drug causality for EN (ALDEN), in order to improve the individual assessment of drug causality in EN. ALDEN causality scores were compared with those from the French pharmacovigilance method in 100 cases and the case-control results of the EuroSCAR study. Scores attributed by ALDEN segregated widely. ALDEN pointed to a "probable" or "very probable" causality in 69/100 cases as compared to 23/100 with the French method (P < 0.001). It scored "very unlikely" causality for 64% of medications vs. none with the French method. Results of ALDEN scores were strongly correlated with those of the EuroSCAR case-control analysis for drugs associated with EN (r = 0.90, P < 0.0001), with probable causality being reported in 218/329 exposures. ALDEN excluded causality in 321 drugs that the case-control analysis had described as "probably not associated" and in 22/233 drugs that had been described as inconclusive exposures. Being more sensitive than a general method, ALDEN, which correlates well with case-control analysis results, can be considered a reference tool in SJS/TEN.