Polihexanide (PHMB) 0.08% versus currently used treatments for Acanthamoeba keratitis: indirect treatment comparisons

Acanthamoeba keratitis is a rare, severe corneal infection. Until the recent approval of polihexanide (PHMB) 0.08% by the European Medicines Agency, there were no licensed medical therapies and current treatments relied on off-label or compounded products. The purpose of this study is to estimate the relative efficacy of PHMB 0.08% compared with current treatments.

A patient-level indirect treatment comparison (ITC) compared data from a pivotal trial of PHMB 0.08% and a retrospective real-world study of current treatments: (1) any anti-amoebic treatment (AAT), (2) PHMB 0.02% plus a diamidine (propamidine or hexamidine) 0.1% and (3) chlorhexidine (CXL) 0.02% alone or in combination with a diamidine. The primary outcome was the clinical resolution rate (CRR) without surgery within 12 months. ITCs were implemented using propensity scoring analysis with overlap weighting and adjustment for covariates (age, sex, disease stage, treatment delay, prior use of corticosteroid or antiviral).

The CRR was 84.8% for PHMB 0.08% (n=66), 43.6% for any AAT (n=227), 55.0% for PHMB 0.02% plus a diamidine (n=111) and 40.0% for CXL 0.02% with or without a diamidine (n=35). In the unweighted analysis, the absolute difference (95% CI) in favour of PHMB 0.08% was 41.2% (28.8%, 51.2%; p<0.001) compared with any AAT, 29.9% (14.5%, 42.1%; p<0.001) compared with PHMB 0.02% plus a diamidine and 44.8% (23.9%, 62.3%; p<0.001) compared with CXL 0.02% with or without a diamidine. Similar results were observed in the weighted analyses.

These results suggest that PHMB 0.08% when delivered with the recommended protocol is significantly more effective than currently used treatments in achieving clinical resolution without surgery. The study limitations include differences in recruitment periods, diagnostic criteria and drug delivery methodology, as well as limitations of the ITC adjustment measures which can lead to residual confounding.