Systemic Medication Adjustment and Short-Term Relapse of Cytomegalovirus Retinitis After Allogeneic Hematopoietic Stem Cell Transplantation: A Research Communication

Background: Cytomegalovirus retinitis (CMVR) remains a sight-threatening complication after allogeneic hematopoietic stem cell transplantation (allo-HSCT). Systemic antiviral, immunosuppressive, and anti-infective regimens are often modified because of transplant-related complications, but the association between these modifications and ocular relapse remains unclear. Methods: We analyzed a prospective cohort of 44 patients with CMVR after allo-HSCT (65 affected eyes) who were followed for three months after discontinuation of local antiviral therapy. The primary analysis used Cox proportional hazards regression at the eye level with robust standard errors clustered by patient. Exact dates of the first systemic medication adjustment were retrieved, and additional time-dependent Cox and patient-level time-to-first-relapse Cox analyses were performed. Results: During follow-up, 24 eyes from 19 patients relapsed, and 18 patients underwent systemic medication adjustment. The first adjustment occurred a median of 25.5 days after discontinuation of local therapy (range, 7 to 72 days). Among the 13 adjusted patients who relapsed, all first adjustments preceded relapse, with a median adjustment-to-relapse interval of 15 days (range, 3 to 52 days). In the primary model, medication adjustment was associated with relapse (HR, 8.254; 95% CI, 2.656 to 25.645; p < 0.001), as was baseline aqueous IL-8 (HR, 1.190 per 100 pg/mL; 95% CI, 1.088 to 1.301; p < 0.001). In the time-dependent eye-level analysis, medication adjustment remained associated with relapses (HR, 11.360; 95% CI, 4.330 to 29.803; p < 0.001). A patient-level time-to-first-relapse Cox analysis showed a similar association (HR, 4.477; 95% CI, 1.688 to 11.878; p = 0.003). Conclusions: Systemic medication adjustment was associated with short-term CMVR relapse after allo-HSCT. Because treatment changes were clinically indicated and heterogeneous, the findings remain associative, and confounding by indication cannot be excluded.