Agreement and Correlation Between Different Topical Corticosteroid Potency Classification Systems

Question  How can investigators meaningfully and transparently classify topical corticosteroid (TCS) potency for pharmacoepidemiologic research?

Findings  In this classification study, a comprehensive list of 232 TCS formulations was classified into 3 systems. There was low-to-moderate agreement but strong correlation between the potency classification systems.

Meaning  The method used to classify TCS potency in pharmacoepidemiologic research could alter the results and interpretation of studies; investigators need to transparently report their TCS potency classifications and consider alternative classifications in sensitivity analyses.

Importance  Topical corticosteroids (TCSs) are available in multiple potencies that alter their effectiveness and safety. Pharmacoepidemiologic studies on TCSs are hampered by the absence of a universal potency classification system, limiting comparisons across studies, robust exposure classification, and clinical interpretation.

Objective  To classify TCSs into 3 commonly used potency classification systems and evaluate the agreement and correlation between the 3 systems.

Design, Setting, and Participants  In this classification study, a comprehensive list of TCS formulations was compiled using sources identified in the literature, the Ontario Drug Benefit Formulary, a recent Cochrane review on the use of TCSs in people with eczema, and the Anatomical Therapeutic Classification (ATC) of the World Health Organization from August 11, 2021, to January 6, 2022. Topical corticosteroid potency classifications were assigned and compared using the 7-category US classification system, a 4-category classification from a recent Cochrane review largely based on the UK formulary, and the 4-category ATC classification. To facilitate comparisons across systems, the 7-category US system was consolidated into 4 categories.

Main Outcomes and Measures  Cohen weighted κ (κ_w) and Spearman rank correlation coefficients (r) were computed to examine agreement and correlation between the classification systems.

Results  A total of 232 unique TCS formulations (ATC, n = 231; US classification, n = 232; Cochrane review, n = 89) were included. Overall, there was low-to-moderate agreement but strong correlation between the classification systems. The US classification had weak agreement with the ATC system (κ_w, 0.53; 95% CI, 0.45-0.60) and moderate agreement with the Cochrane review classification (κ_w, 0.60; 95% CI, 0.48-0.73); there was weak agreement between the ATC and Cochrane review classifications (κ_w, 0.58; 95% CI, 0.46-0.71). The US classification strongly correlated with the ATC system (r, 0.77; 95% CI, 0.71-0.82) and Cochrane review classification (r, 0.74; 95% CI, 0.62-0.82). There was also a strong correlation between the Cochrane review and ATC classifications (r, 0.71; 95% CI, 0.58-0.80).

Conclusions and Relevance  This classification study used multiple resources to classify 232 TCS formulations into 3 potency classifications. Because these systems are often incongruent, they may yield different results in pharmacoepidemiologic studies; investigators need to be transparent in their classification approach and consider alternative potency definitions in sensitivity analyses.