BACKGROUND: In treatment of human immunodeficiency virus (HIV), high levels of adherence to combination antiretroviral therapy (cART) are required to prevent failure of virologic suppression, development of drug resistance, and permanent loss of therapeutic options. No published research has assessed the association between cART prescription cost sharing and adherence to cART. OBJECTIVE: To analyze the association between cART prescription cost sharing and adherence to initial cART in commercially insured antiretroviral (ARV)-naïve patients with HIV. METHODS: This retrospective observational cohort study used 2002-2008 data from a large U.S. claims database of more than 56 million commercially insured individuals. Study subjects were patients aged 18 years or older who initiated cART during the period January 1, 2003, to December 31, 2007, had no ARV claims during the 6-month period prior to the initiation date, and had at least 1 ICD-9-CM diagnosis code for HIV infection (042, 795.71, V08) from 12 months before to 12 months after cART initiation. A minimum 12-month period of continuous enrollment after cART initiation was used to construct a patient-quarter repeated measures panel dataset in which each quarter of data that a patient contributed represented an observation. The evaluation period extended from cART initiation until the occurrence of 1 of the following events: addition of an ARV that was not part of the initial cART regimen, 30-day gap in possession of an ARV within the initiated cART regimen, hospitalization of 30 or more days, loss to follow-up due to study end (December 31, 2008), or disenrollment. The study's outcome was quarterly adherence to cART, defined as the number of days within the quarter that a patient possessed all components of the initial cART regimen. Each patient's cART cost-sharing amount was calculated per 30-day supply of the entire cART regimen. Adherence was dichotomized for analysis at the clinically meaningful thresholds of 95% and 78%. The dichotomized adherence outcomes were separately modeled using population-averaged generalized estimating equations (GEEs) with time-varying and time-constant covariates and an exchangeable working correlation structure. Independent variables included cost-sharing amount; sequential quarter number after cART initiation; interaction between cost-sharing amount and sequential quarter number (to capture any changes in the association of cost sharing with adherence that may occur over time after initiation of cART); and patient demographic, clinical, and insurance characteristics. For each sequential quarter after cART initiation, the GEE models were used to generate average predicted probabilities of adherence reaching each threshold (95% and 78%) at cost-sharing levels of $25, $75, and $144, which represented the 25th, 75th, and 90th percentiles of the cost-sharing distribution, respectively. RESULTS: The study sample included 19,199 patient-quarters and 3,731 patients: mean age 41.1 years; 83.2% male; mean (SD) duration of post-index period 5.1 (4.2) quarters; mean (SD) daily cART pill count 3.2 (2.2); mean (median) cost sharing per 30-day supply of the entire cART regimen $67 ($40). In the unadjusted analyses of patient-quarters, mean adherence ranged from 97.2% for cost-sharing levels within the 0-20th percentiles (from $0 to $20 per 30-day cART supply) to 94.0% for cost-sharing levels exceeding the 80th percentile (from $84 to $3,832 per 30-day cART supply). In the adjusted analyses for the second quarter (25th percentile of follow-up duration, n = 3,117 cases still under observation) at the cost-sharing levels of $25, $75, and $144, the predicted probabilities of at least 95% adherence were 0.782, 0.770, and 0.752, respectively, and the predicted probabilities of at least 78% adherence were 0.936, 0.931, and 0.924, respectively. The differences in the predicted probabilities of adherence grew over time. By the seventh quarter (the 75th percentile of follow-up duration, n = 1,096 cases still under observation), the predicted probabilities were 0.773, 0.746, and 0.707 for 95% adherence and 0.933, 0.922, and 0.904 for 78% adherence at cost-sharing levels of $25, $75, and $144, respectively. CONCLUSION: Increasing cART prescription cost sharing was associated with modestly decreased probability of maintaining clinically meaningful levels of cART adherence.