Importance: The diagnostic properties of a graded provocation challenge (PC) among children presenting with a rash in the course of amoxicillin treatment are currently unknown.
Objective: To assess the accuracy and the negative predictive value of the PC in a cohort of children referred with suspected allergy to amoxicillin.
Design, setting, and participants: A cohort study was conducted between March 1, 2012, and April 1, 2015, at the allergy clinic of the Montreal Children's Hospital, Montreal, Quebec, Canada. All children referred with suspected allergy to amoxicillin were approached. In addition, 346 eligible children were followed up to assess reactions to subsequent use of amoxicillin at the time of illness in cases with negative PC results. Data were collected on clinical characteristics, suspected antibiotic exposure, personal and first-degree relatives' comorbidities, and history of atopy and management of the reaction. Univariate and multivariate logistic regressions were compared to determine factors associated with immediate and nonimmediate reactions to the PC.
Interventions: All children had a graded PC.
Main outcomes and measures: Reactions to the graded PC, the negative predictive value of the PC for nonimmediate reactions, and factors associated with immediate and nonimmediate reactions to the PC.
Results: A total of 818 children were assessed (median age, 1.7 years [interquartile range, 1.0-3.9 years]; 441 [53.9%] male). Among all participants, 770 (94.1%) tolerated the PC, 17 (2.1%) developed mild immediate reactions, and 31 (3.8%) developed nonimmediate reactions. The graded PC had a specificity of 100.0% (95% CI, 90.9%-100.0%), a negative predictive value of 89.1% (95% CI, 77.1%-95.5%), and a positive predictive value of 100.0% (95% CI, 86.3%-100.0%). Among all 346 participants eligible for annual follow-up, 250 (72.3%; 95% CI, 67.2%-76.8%) responded, 55 of whom received subsequent full treatment with amoxicillin; 49 of these 55 participants (89.1%) reported tolerance to subsequent full treatment with amoxicillin, while 6 (10.9%) developed nonimmediate cutaneous reactions. History of a reaction occurring within 5 minutes of exposure was associated with immediate reactions to the PC (adjusted odds ratio = 9.6; 95% CI, 1.5-64.0), while a rash that lasted longer than 7 days (adjusted odds ratio = 4.8; 95% CI, 1.4-16.4) and parental history of drug allergy (adjusted odds ratio = 3.0; 95% CI, 1.3-6.8) were associated with nonimmediate reactions to the PC.
Conclusions and relevance: Graded PCs provide an accurate and safe confirmatory test for skin-related reactions to amoxicillin. Further studies are required to assess factors associated with the PC outcome groups.