Development and Validation of a Novel Score to Predict Mortality in Stevens-Johnson Syndrome and Toxic Epidermal Necrolysis: CRISTEN

Abstract

Background: Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are life-threatening, severe mucocutaneous adverse reactions. Severity prediction at early onset is urgently required for treatment. However, previous prediction scores have been based on data of blood tests.

Objective: This study aimed to present a novel score that predicts mortality in patients with SJS/TEN in the early stages based on only clinical information.

Methods: We retrospectively evaluated 382 patients with SJS/TEN in a development study. A clinical risk score for TEN (CRISTEN) was created according to the association of potential risk factors with death. We calculated the sum of these risk factors using CRISTEN, and this was validated in a multinational survey of 416 patients and was compared with previous scoring systems.

Results: The significant risk factors for death in SJS/TEN comprised 10 items, including patients' age of ≥65 years, ≥10% body surface area involvement, the use of antibiotics as culprit drugs, the use of systemic corticosteroid therapy before the onset, and mucosal damage affecting the ocular, buccal, and genital mucosa. Renal impairment, diabetes, cardiovascular disease, malignant neoplasm, and bacterial infection were included as underlying diseases. The CRISTEN model showed good discrimination (area under the curve [AUC] = 0.884) and calibration. In the validation study, the AUC was 0.827, which was statistically comparable to those of previous systems.

Conclusion: A scoring system based on only clinical information was developed to predict mortality in SJS/TEN and was validated in an independent multinational study. CRISTEN may predict individual survival probabilities and direct the management and therapy of patients with SJS/TEN.

Keywords: Disease mortality; Scoring system; Stevens-Johnson syndrome; Toxic epidermal necrolysis.