The following is a summary of “IFN-γ ELISpot-enabled machine learning for culprit drug identification in nonimmediate drug hypersensitivity,” published in the January 2024 issue of Allergy & Immunology by Chongpison, et al.

Diagnosing drug-induced allergies, particularly nonimmediate phenotypes, presents challenges, often resulting in incorrect classifications with undesirable consequences. For a study, researchers sought to evaluate the diagnostic utility of IFN-γ ELISpot and clinical parameters in predicting drug-induced nonimmediate hypersensitivity using machine learning techniques.

The study enrolled 393 patients, using positive patch tests or drug provocation tests (DPT) to define drug hypersensitivity. Various clinical factors were considered in developing random forest (RF), and logistic regression (LR) models, and their performances were compared against an IFN-γ ELISpot-only model.

Most 102 patients who underwent 164 DPTs had severe cutaneous adverse reactions (35/102, 34.3%) and maculopapular exanthems (33/102, 32.4%). The most common suspected drugs were antituberculosis (46/164, 28.1%) and β-lactams (42/164, 25.6%). The mean age of patients with DPT was 52.7 (20.8) years. IFN-γ ELISpot, fixed drug eruption, Naranjo categories, and nonsteroidal anti-inflammatory drugs emerged as the essential features in all models. The RF and LR models demonstrated higher discriminating abilities. An IFN-γ ELISpot cutoff value of 16.0 spot-forming cells/106 PBMCs achieved 94.8% specificity and 57.1% sensitivity. Optimal cutoff values for RF and LR models could be chosen based on clinical needs to achieve either high specificity (0.41 for 96.1% specificity and 0.52 for 97.4% specificity, respectively) or high sensitivity (0.26 for 78.6% sensitivity and 0.37 for 71.4% sensitivity, respectively).

The IFN-γ ELISpot assay proved valuable in identifying culprit drugs, whether used individually or in prediction models. The performances of RF and LR models were comparable, and further validation with additional test datasets containing DPTs would be beneficial.

Reference: jacionline.org/article/S0091-6749(23)01108-9/abstract