Design: Nationwide registry cohort with entropy balancing
Population: 18,635 adults with psoriasis on systemic therapy, 118,018 person-years
Primary outcome: Tuberculosis, meningitis, candidiasis and other fungal infection
Effect: Candidiasis IRR 2.67 (1.33 to 5.36) to 4.65 (3.46 to 6.24) for IL-17 vs other classes

This BADBIR cohort included 40,930 treatment episodes in 18,635 adults with psoriasis, 2007 to 2024, followed for 118,018 person-years. Treatment groups were balanced with entropy weighting and compared with doubly robust Poisson models.

Tuberculosis (22 cases, 0.19 per 1000 person-years) and meningitis (29 cases, 0.25) were rare and occurred mostly with TNF-α inhibitors, particularly adalimumab. Meningitis was mainly viral and tuberculosis mainly pulmonary. Fungal infection ran at 7.53 per 1000 person-years. IL-17 inhibitors alone raised candidiasis risk, with incidence rate ratios from 2.67 (1.33 to 5.36) against apremilast to 4.65 (3.46 to 6.24) against the IL-12/23 inhibitor. Individual IL-17 agents did not differ, and other fungal infections did not differ by class.

For Indian practice, where latent TB is common, this supports favouring IL-23 or IL-17 inhibitors over TNF inhibitors when TB risk is high, and IL-23 over IL-17 when candidiasis is a concern.

Screen for latent TB before every biologic, and repeat where exposure continues.
Where TB risk is high, prefer IL-23 or IL-17 inhibitors over TNF-α inhibitors.
Where recurrent candidiasis is a concern, prefer an IL-23 inhibitor.
Switching between IL-17 agents will not lower candidiasis risk.
Record infections in the biologic register or clinic database.

Why it matters

It gives class-level infection data from real practice that trial populations were too small to show.

The statistics, in plain English

An incidence rate ratio of 4.65 means candidiasis occurred at nearly five times the rate on IL-17 inhibitors as on the IL-12/23 inhibitor. The TB and meningitis counts are so small that comparisons between classes for those outcomes are unreliable. This is observational; weighting balances measured differences only.