The cost-effectiveness of HLA-B*5701 genetic screening to guide initial antiretroviral therapy for HIV
[EPIDEMIOLOGY AND SOCIAL: CONCISE COMMUNICATIONS]
AIDS:Volume 22(15)1 October 2008p 2025-2033
Schackman, Bruce Ra; Scott, Callie Ab; Walensky, Rochelle Pb,c,d; Losina, Elenab,c,e,f; Freedberg, Kenneth Ab,c,f; Sax, Paul Ec, aDepartment of Public Health, Weill Cornell Medical College, New York, New York bDivisions of General Medicine and Infectious Diseases and the Partners AIDS Research Center, Massachusetts General Hospital, USA cDivision of AIDS and Center for AIDS Research, Harvard Medical School, USA dDivision of Infectious Diseases, USA eDepartment of Orthopedic Surgery, Brigham and Women's Hospital, USA fDepartments of Biostatistics and Epidemiology, Boston University School of Public Health, Boston, Massachusetts, USA.
Objective: To evaluate the clinical impact and cost-effectiveness of HLA-B*5701 testing to guide selection of first-line HIV regimens in the United States.
Design: Cost-effectiveness analysis using a simulation model of HIV disease. The prevalence of HLA-B*5701 and the probabilities of confirmed and unconfirmed severe systemic hypersensitivity reaction among patients taking abacavir testing HLA-B*5701 positive and negative were from the Prospective Randomized Evaluation of DNA Screening in a Clinical Trial study. The monthly costs of abacavir-based and tenofovir-based regimens were $1135 and $1139, respectively; similar virologic efficacy was assumed and this assumption was varied in sensitivity analysis.
Patients: Simulated cohort of patients initiating HIV therapy.
Interventions: The interventions are first-line abacavir, lamivudine, and efavirenz without pretreatment HLA-B*5701 testing; the same regimen with HLA-B*5701 testing; and first-line tenofovir, emtricitabine, and efavirenz.
Main outcome measures: Quality-adjusted life years and lifetime medical costs discounted at 3% per annum, cost-effectiveness ratios ($/QALY).
Results: Abacavir-based treatment without HLA-B*5701 testing resulted in a projected 30.93 years life expectancy, 16.23 discounted quality-adjusted life years, and $472 200 discounted lifetime cost per person. HLA-B*5701 testing added 0.04 quality-adjusted months at an incremental cost of $110, resulting in a cost-effectiveness ratio of $36 700/QALY compared with no testing. Initiating treatment with a tenofovir-based regimen increased costs without improving quality-adjusted life expectancy. HLA-B*5701 testing remained the preferred strategy only if abacavir-based treatment had equal efficacy and cost less per month than tenofovir-based treatment. Results were also sensitive to the cost of HLA-B*5701 testing and the prevalence of HLA-B*5701.
Conclusion: Pharmacogenetic testing for HLA-B*5701 is cost-effective only if abacavir-based treatment is as effective and costs less than tenofovir-based treatment.
Abacavir is a nucleoside reverse transcriptase inhibitor (NRTI) used as part of combination antiretroviral therapy (ART) for HIV. It has proven efficacy in first-line ART regimens, with few drug long-term toxicities observed . In a small proportion of patients, however, treatment with abacavir may be associated with systemic hypersensitivity reaction (HSR), a multiorgan system process that can be severe enough to cause hospitalization or death . Both retrospective and prospective studies [3,4] have demonstrated a strong association between the presence of the HLA-B*5701 allele and the risk of HSR in patients taking abacavir. Furthermore, in a randomized, prospective clinical trial, patients screened for HLA-B*5701 before abacavir therapy experienced dramatically reduced rates of both clinically diagnosed and immunologically confirmed HSR .
HIV treatment guidelines issued by the United States Department of Health and Human Services in February 2008 recommend abacavir as a preferred component of initial ART only for patients who test negative for HLA-B*5701 . Given the safety and efficacy of abacavir-based treatment in the absence of HSR, our goal was to evaluate the long-term clinical impact and cost-effectiveness of pretreatment HLA-B*5701 testing to guide initial HIV therapy.
For some medications, pretreatment genetic screening can significantly improve drug safety . In HIV therapy, important new discoveries are emerging that describe genetic associations with severe adverse drug events . Although genetic screening can reduce the incidence of toxicity, additional expenditures are required to conduct the genetic test. Cost-effectiveness analysis can assist decision-makers in evaluating the value of conducting pharmacogenetic tests.
We conducted a cost-effectiveness analysis of pharmacogenetic testing in HIV-infected patients in the United States, taking into account the implications for initial regimen selection and subsequent treatment options. We found that HLA-B*5701 testing to guide selection of a first-line ART regimen is cost-effective, with a cost-effectiveness ratio below the commonly accepted thresholds in the United States of $50 000-$100 000/QALY. The results were critically dependent on the comparable efficacy and lower cost of abacavir-based treatment compared with tenofovir-based treatment.
We found that the economic value of testing also depends on the cost of the test and the prevalence of HLA-B*5701, although a higher test cost could be offset by a reduction in the price of abacavir-based treatment. In non-US settings, treatment and testing costs will vary depending on the site and prevalence will vary depending on the race/ethnicity characteristics of the population. The findings reported here are consistent with an analysis conducted in the United Kingdom before the PREDICT-1 results became available, which found that the incremental cost of testing per abacavir hypersensitivity reaction avoided was sensitive to prevalence and medication costs (when test costs were not varied) . In the United States, cost considerations may be particularly important for public programs such as the AIDS Drug Assistance Programs funded by the Ryan White CARE Act. These state-administered programs are major payers for HIV drugs, but do not generally pay for laboratory tests, and in some states tight budget limits have already led to cost-containment strategies.
The present analysis has several limitations, and follow-up cost-effectiveness analyses will be valuable as uncertainties regarding therapeutic options and future genetic tests are resolved. We project long-term survival based on sequential lines of ART of varying efficacy based on data available from clinical trials; the actual regimens to be chosen in the future remain unknown, especially in light of recently approved agents with novel mechanisms of action. We did not take into account the possibility of increased risks of cardiovascular events associated with abacavir reported recently , and we did not consider alternatives to HLA-B*5701 testing that may be less expensive but have similar test performance characteristics in predicting abacavir hypersensitivity reactions. A less expensive test may result in a more attractive cost-effectiveness ratio, depending on the test characteristics.
The assumption about the reduction in the rate of HSR diagnoses after testing may be conservative, because it was derived from the PREDICT-1 trial in which clinicians were blinded to whether patients experiencing symptoms had tested HLA-B*5701 negative or had been assigned to the control arm. Although we found that the cost-effectiveness results were only moderately sensitive to this assumption, our model does not fully reflect the benefits to providers and patients of avoiding such a diagnosis, including fewer physician-patient contacts and fewer worries about a potentially life-threatening adverse event.
Most importantly, we found that HLA-B*5701 testing was only cost-effective under the assumption of equal efficacy between abacavir-based therapy and tenofovir-based therapy. In one randomized, prospective study of first-line ART , abacavir/lamivudine was found to be noninferior to tenofovir/emtricitabine when both were used in combination with ritonavir-boosted lopinavir. Our findings were unchanged when we restricted the analysis to patients with HIV RNA less than 100 000 copies/ml, based on interim results from a second blinded study that is still underway . However, if this study's results ultimately demonstrate a lower efficacy overall of abacavir/lamivudine regardless of baseline HIV RNA, then starting with a tenofovir-based treatment is more cost effective, regardless of whether HLA-B*5701 testing is performed.
On the basis of the currently available data, use of genetic testing for HLA-B*5701 to guide selection of initial treatment for HIV in the United States is effective and is cost-effective using a threshold of $50 000/QALY. However, the cost-effectiveness ratio is highly dependent on the comparable efficacy of abacavir-based treatment to tenofovir-based treatment, on the relative costs of the drugs, and on the cost of the HLA-B*5701 test itself. With the many highly effective options for initial HIV treatment now available in the United States, pharmaceutical manufacturers and test providers need to ensure that the cost of this innovation is commensurate with the value that it provides.
Initiating abacavir-based treatment in all patients without testing for HLA-B*5701 was the least expensive first-line strategy among all alternatives considered. With this strategy, 2.7% of patients developed confirmed HSR (1.6% of patients developed mild HSR, 1.1% developed severe HSR, and 0.02% developed fatal HSR). An additional 5.1% had an unconfirmed HSR diagnosis and were switched from abacavir to tenofovir-based treatment. Patients initiating this strategy had a life expectancy of 371.17 months (30.93 years), a discounted, quality-adjusted life expectancy of 194.75 months (16.23 QALYs), and an average discounted lifetime cost of $472 200 (Table 3).
With the universal testing strategy, no patients developed mild, severe, or fatal HSR, and 3.4% had an unconfirmed HSR diagnosis. Additionally, 5.7% were initially assigned to the tenofovir-based regimen due to a positive HLA-B*5701 test result. Universal testing strategy patients had an incremental gain of 0.04 discounted, quality-adjusted life months with an incremental discounted lifetime cost of $110. The cost-effectiveness ratio of universal testing, compared with no testing and initiating all patients on abacavir-based treatment, was $36 700/QALY.
Initiating patients on tenofovir-based treatment and substituting abacavir and lamivudine if treatment-limiting nephrotoxicity occurs, with or without HLA-B*5701 testing prior to making the substitution, resulted in a similar quality-adjusted life expectancy of 194.79 QALYs but was $230 more expensive than universal testing. Substituting the more toxic, less effective zidovudine-based regimen if treatment-limiting nephrotoxicity occurs resulted in 0.07 fewer quality-adjusted life months compared with universal testing. Hence, at current drug costs and assuming equal efficacy between abacavir-based and tenofovir-based treatment, universal testing is preferred to initiating patients on tenofovir-based treatment.
Sensitivity analyses varying efficacy of treatment
Universal testing remained preferred to strategies that involve starting patients on tenofovir-based treatment in terms of cost-effectiveness only if both treatments were assumed to be equally effective. If the proportion of patients with HIV RNA less than 400 copies/ml at 48 weeks was 1% lower on abacavir-based treatment than on tenofovir-based treatment, universal testing resulted in a lower discounted quality adjusted life expectany (16.22 QALYs) than starting patients on tenofovir-based treatment. It also resulted in a higher lifetime cost, because the lower cost of the abacavir-based regimen was offset by the need to switch sooner to a more expensive therapy.
Sensitivity analyses varying the cost of the HLA-B*5701 test
At half the primary analysis test cost ($34), the cost-effectiveness ratio for universal testing compared with no testing decreased from $36 700 to $25 300/QALY. At double the primary analysis test cost ($136), roughly equivalent to the lowest cost reported by hospital-based laboratories, the cost-effectiveness ratio for universal testing compared with no testing increased to $59 700/QALY, and universal testing had a cost-effectiveness ratio less than $100 000/QALY as long as the monthly cost of abacavir-based therapy was at least $3 less than tenofovir-based therapy (Table 4). At five times the primary analysis test cost ($340), roughly equivalent to the highest cost reported by a hospital-based laboratory, initiating tenofovir-based treatment, and substituting abacavir without HLA-B*5701 testing if treatment-limiting nephrotoxicity occurs, was the most cost-effective strategy compared with no testing. The incremental cost-effectiveness of universal testing compared with initiating tenofovir-based treatment was $480 000/QALY. When the monthly cost of abacavir-based therapy was $6 less than tenofovir-based therapy (instead of $4 per month less in the primary analysis), universal testing became preferred to initiating with tenofovir-based treatment and had a cost-effectiveness ratio of $128 300/QALY.
Sensitivity analyses varying population characteristics
Results were similar for patients with racial/ethnic characteristics comparable with patients initiating HIV treatment in the United States. The rank ordering of results did not change, but the cost-effectiveness ratio for universal testing compared with no testing increased to $45 200/QALY, reflecting the higher proportion of nonwhites tested and the lower prevalence of HLA-B*5701 in this population (4.1 versus 5.7% in the primary analysis). Results were also similar if the population was restricted to patients with HIV RNA less than 100 000 copies/ml at initiation of first-line ART (Table 4).
Results varied according to the prevalence of HLA-B*5701 in the population (Fig. 1). With the primary analysis test cost of $68, the cost-effectiveness ratio of universal testing remained below $100 000/QALY as long as the prevalence of HLA-B*5701 was greater than 1.4% and it remained below $50 000/QALY as long as the prevalence of HLA-B*5701 was greater than 3.6%, as compared with 5.7% in the primary analysis. At twice the primary analysis test cost, these thresholds became 2.9 and 7.4%.
Other sensitivity analyses
If HLA-B*5701 testing eliminated all unconfirmed HSR diagnoses, universal testing became more attractive, with a cost-effectiveness ratio of $33 500/QALY assuming the primary analysis test cost. Conversely, when testing provided no benefit in reducing the probability of unconfirmed HSR diagnoses, the cost-effectiveness ratio increased to $38 700/QALY (Table 4). Results were sensitive to the probability of confirmed HSR among HLA-B*5701 positive patients when test costs were higher: at twice the primary analysis test cost, the probability of confirmed HSR among HLA-B*5701 positive patients must be below 59.1% (versus 47.9% in the primary analysis) to be cost-effective at the $50 000/QALY threshold.
As long as the monthly cost of abacavir-based treatment was at least $2 lower than tenofovir-based treatment, the universal testing strategy had a lower discounted lifetime cost than tenofovir-based treatment. Varying the proportions of confirmed HSR that were mild, severe, and fatal within published ranges did not have a meaningful impact on cost-effectiveness findings nor did varying the cost or quality-of-life decrements associated with HSR or tenofovir-related nephrotoxicity. If patients unable to use abacavir or tenofovir substituted new regimens without any risk of additional toxicities, the cost-effectiveness of universal testing became $35 000/QALY compared with $36 700/QALY in the primary analysis.
We used the Cost-Effectiveness of Preventing AIDS Complications (CEPAC) model, a widely published simulation state transition model of HIV disease [7,8], along with data reported in the literature, to evaluate the clinical impact and cost-effectiveness of HLA-B*5701 testing to guide selection of first-line HIV treatment regimens in patients in the United States.
Upon entering the model, simulated patients initiated a first-line ART regimen consisting of a fixed-dose combination of abacavir and lamivudine with efavirenz or a fixed-dose combination of tenofovir, emtricitabine, and efavirenz. For patients initiating treatment that included abacavir, we compared 'universal testing' with 'no testing.' In the universal testing strategy, all patients were tested for HLA-B*5701 prior to ART initiation. Abacavir-based treatment was selected for patients testing HLA-B*5701 negative and tenofovir-based treatment was selected for patients testing HLA-B*5701 positive. In the no testing strategy, all patients were initiated on abacavir-based treatment. In either strategy, patients taking abacavir who developed a suspected HSR were treated in office-based or inpatient settings according to the severity of their symptoms, and were then switched to tenofovir-based treatment. Patients switched to tenofovir-based treatment who subsequently developed treatment-limiting tenofovir-associated nephrotoxicity were switched to a fixed-dose combination of zidovudine and lamivudine with efavirenz.
In the comparison strategy, all patients in the model initiated tenofovir-based treatment without initial HLA-B*5701 testing. For patients who developed treatment-limiting tenofovir-associated nephrotoxicity, we considered three alternatives to guide drug substitution: 'substitution HLA-B*5701 testing,' in which those testing HLA-B*5701 negative were switched to abacavir-based treatment and those testing HLA-B*5701 positive were switched to zidovudine-based treatment, substituting abacavir-based treatment without testing, and substituting zidovudine-based treatment without testing.
We used the CEPAC model to project the long-term outcomes of each possible combination of initial drug regimens, toxicities (mild, severe, or fatal suspected HSR; treatment-limiting tenofovir-associated nephrotoxicity; or no toxicity), and drug substitutions. Details regarding the CEPAC model structure, data, and assumptions have been previously described (see Appendix at http://www.aidsonline.com ). We then calculated the results for each strategy as the weighted average of these projected outcomes using probabilities derived from published studies as the basis for weighting. These probabilities included the incidence of immunologically confirmed HSR in patients taking abacavir with and without HLA-B*5701 testing, and the reduction in suspected HSR diagnoses observed with HLA-B*5701 testing. Tables 1 and 2 report baseline model inputs and ranges used in sensitivity analyses.
Results are reported as quality-adjusted life expectancies (QALYs) and lifetime direct medical costs in 2006 US dollars, both discounted to present value at an annual rate of 3%. All cost-effectiveness ratios are calculated on an incremental basis by ranking strategies from least to most expensive and comparing each strategy with the next most expensive strategy. Cost-effectiveness ratios are reported as cost per quality-adjusted life year ($/QALY).