Characterizing Response Patterns to Ranibizumab Therapy in Patients with Wet Age-Related Macular Degeneration: A Latent Class Growth Analysis

Abstract

Objectives: To identify and characterize response patterns in patients with wet age-related macular degeneration (AMD) over 24 month after ranibizumab therapy; to determine demographic and clinical predictors of response patterns at the initial time of treatment (baseline); and to quantify and compare resource utilization across response patterns at the end of treatment (month 24).Methods: We performed a secondary data analysis using existing data from a prospective, observational, multicenter, open-label trial of 0.5 mg of ranibizumab administered by intravitreal injection. Patients with wet AMD were followed over 24 +/- 3 months with intermediate data points at 6 +/- 2 months and 12 +/- 2 months, and a few data points at 2.5 +/- 1 month that coincided with the end of the loading phase. The primary outcome of interest was change in visual acuity (as measured by Early Treatment Diabetic Retinopathy Study [ETDRS] letters). Latent class growth analysis (LCGA) was used to examine treatment response variability in the data and to identify latent classes (unobserved groups) of response patterns. A multinomial logistic regression was specified to identify predictors of the response patterns. Variables related to resource utilization at the end of treatment were also examined across response patterns. Results: LCGA demonstrated a large variability in visual acuity change. We identified three clusters of patients for each response pattern. Patients in cluster 1 (partial responders, 56.68% of the total sample) had stable improvement in visual acuity that plateaued at month 3 and then gradually diminished over time. Patients in cluster 2 (optimal responders, 23.50%) showed progressive improvement in visual acuity that plateaued at month 6 and then remained stable over time. Patients in cluster 3 (non-responders, 19.82%) had the worst performing course in visual acuity and showed drastic decline for the first 12 months that tapered off. Multinomial logistic regression revealed significant differences across clusters in terms of age, baseline visual acuity, and certain lesion types. Resource utilization at the end of treatment also varied significantly across clusters, with non-responders on average receiving the lowest total number of ranibizumab per patient. Conclusions: LCGA identified three response patterns to ranibizumab among patients with wet AMD. The patterns were significantly associated with age, baseline visual acuity, and certain lesion types. Non-responders on average received the lowest total number of ranibizumab per patient. Identifying differential responders has obvious advantages for understanding treatment effects and may help provide a basis of classification for intervention.