Premature Discontinuation of Atazanavir/ritonavir and UGT1A1 Variants: ACTG Protocol A5202
Heather Ribaudo1, E Daar2, C Tierney1, G Morse3, K Mollan1, P Sax4, M Fischl5, A Collier6, D Haas7, and ACTG
1Ctr for Biostatistics in AIDS Res, Harvard Sch of Publ Hlth, Boston, MA, US; 2Los Angeles Biomed Res Inst at Harbor-UCLA Med Ctr, Torrance, CA, US; 3Univ at Buffalo, State Univ of New York, US; 4Brigham and Women`s Hosp, Boston, MA, US; 5Univ of Miami Miller Sch of Med, FL, US; 6Univ of Washington, Seattle, US; and 7Vanderbilt Univ Sch of Med, Nashville, TN, US
Background: Atazanavir (ATV) can cause hyperbilirubinemia by competing with bilirubin for hepatic UGT1A1. Among HIV-infected patients prescribed ATV in observational cohorts, the UGT1A1*28 Gilbert variant 6(TA)>7(TA) rs8175347 has been reported to predict greater bilirubin elevations and increased likelihood of ATV discontinuation. We characterized associations between functional UGT1A1 variants, hyperbilirubinemia, and ATV/ritonavir (ATV/r) discontinuation in a large clinical trial.
Methods: A5202 randomized treatment-naïve patients to receive ATV/r (300mg/100mg) or efavirenz, with tenofovir/emtricitabine or abacavir/lamivudine. Genetic consent was obtained under protocol A5128; genotyping was by fragment analysis. The pre-specified primary endpoint for this analysis was time to ATV/r discontinuation for any reason. Genotype association was evaluated by exact Tarone test for trend and a Cox proportional hazards model adjusting for key baseline factors. Descriptive summaries included the proportion with total bilirubin >5x upper limit of normal (ULN) over the first 24 weeks of ATV/r. The positive predictive value (PPV) of *28/*28 genotype for predicting ATV/r discontinuation within 48 weeks was estimated. Throughout UGT1A1*36 5(TA) and *37 8(TA) were treated as *1 and *28, respectively.
Results: Among 925 ATV/r-treated A5202 participants, 646 consented and had UGT1A1 genotype available. *28/*28 was present in 8% of 291 white, 24% of 202 black, and 18% of 153 Hispanic participants. Total bilirubin was >5xULN over 24 weeks in 3% (*1/*1), 5% (*1/*28), and 20% (*28/*28). Among Hispanics, an association between the number of *28 alleles and ATV/r discontinuation was apparent (p = 0.005) and appeared driven by the *28/*28 genotype. Associations were not apparent among whites or blacks (p = 0.79, p = 0.46, respectively, Figure). Results were consistent in adjusted analyses. For a Hispanic subject with 28*/28*, the estimated probability of early ATV/r discontinuation (PPV) was 32% (95%CI 16% to 52%).
Conclusions: Among subjects randomized to ATV/r-containing regimens in A5202, UGT1A1 genotypes that are associated with hyperbilirubinemia were associated with ATV/r discontinuation in Hispanics but not whites or blacks. Despite this strong association in Hispanics, the *28/*28 genotype did not reliably predict ATV/r discontinuation. Discordance with prior data may reflect different thresholds for ATV/r discontinuation in different contexts, variable ATV systemic exposure, or other genetic variants.