We conducted simulation model analyses to identify factors that would determine the cost-effectiveness of UGT1A1 pharmacogenetic testing to inform prescribing of first-line protease inhibitor-containing ART in the United States. In a hypothetical cohort of patients who would otherwise initiate atazanavir-containing ART at CD4 500/µl, the cost-effectiveness of UGT1A1 testing was most sensitive to UGT1A1 assay cost and whether atazanavir-associated hyperbilirubinemia caused patients to be lost to follow-up only to return with substantially more advanced HIV disease. At low UGT1A1 assay cost, the cost-effectiveness of UGT1A1 testing was also sensitive to the population frequency of UGT1A1*28/*28. However, if atazanavir-containing ART and darunavir-containing ART differed in either cost or efficacy, UGT1A1 testing was not cost-effective under any scenario. That is, the benefits of higher efficacy or lower cost outweighed the benefits of testing. These findings highlight the areas in which model-based analyses can inform priorities for translating pharmacogenetics into HIV clinical care.
Previous model-based analyses have shown that HLA*B-5701
testing to prevent abacavir hypersensitivity reactions is cost-effective [6
], but HLA*B-5701
testing is far more sensitive and specific for abacavir hypersensitivity than UGT1A1
testing is for atazanavir discontinuation. Moreover, abacavir hypersensitivity is potentially life-threatening, while atazanavir-associated hyperbilirubinemia is a relatively mild adverse event with minimal impact on quality of life. Our finding, that UGT1A1
testing is not cost-effective if an alternative drug that does not require pharmacogenetic testing is either more effective or less expensive, is consistent with what has been reported for HLA*B-5701
Testing for UGT1A1
was not cost-effective at the current stand-alone test cost of $107 unless hyperbilirubinemia resulted in loss to follow-up that could be prevented by UGT1A1
testing. Loss to follow-up represents a critical gap in HIV care in the US [35
], where currently an estimated one-third of HIV-infected individuals linked to care are not retained in care [37
]. Future evaluations of pharmacogenetic tests for HIV regimen selection should consider the clinical implications of test results for retention in care, as well as the feasibility of genetic test panels that substantially reduce the cost attributable to each marker tested. Electronic health records increase the feasibility of integrating such large scale pharmacogenetic testing with clinical care [38
Our analysis is subject to several limitations. We used as model inputs reported results of a retrospective analysis of 121 patients initiating ART in the Swiss HIV Cohort Study, among whom UGT1A1
homozygosity was associated with atazanavir discontinuation rates of 62.5%, 23.8%, and 14.6% among patients homozygous for *28/*28
, heterozygous for *1/*28
, and homozygous for *1/*1
, respectively [4
]. Such data may not be generalizable to the U.S. or other populations. When we varied genetic risk to reflect variation in UGT1A1*28
frequency by ancestry, the cost-effectiveness ratios varied considerably at a hypothetical $10 assay cost and similar variation was observed if we assumed lower discontinuation rates. Furthermore, a recent analysis of data from a clinical trial conducted in the U.S. suggested that UGT1A1*28*28
was associated with atazanavir discontinuation in Hispanics, but not in whites or blacks [39
]. In the base case, we assumed atazanavir and darunavir had equivalent efficacy and cost. A clinical trial comparing atazanavir-containing and darunavir-containing first-line ART is ongoing [40
], and drug costs may vary depending on manufacturer discounts. Our assumption that clinicians or patients might prefer to initiate atazanavir was not captured in base case QALYs or costs. The range of quality-of-life effects of hyperbilirubinemia that we considered may not capture the full spectrum of clinical situations. For example, icterus may more adversely affect quality of life in certain groups (e.g. adolescents, or individuals in jobs where their appearance affects their ability to work). We did not consider the potential future benefit of UGT1A1
testing to inform prescribing of other non-HIV drugs.
To determine the cost-effectiveness of UGT1A1 and other pharmacogenetic testing to guide ART regimen selection, it will be important to better understand the associations between genotype, adverse events, and loss to follow-up. Cost-effectiveness of initial ART, however, will remain primarily driven by regimen efficacy and cost.