Secondary Logo

Journal Logo


The prevalence of antiretroviral drug resistance in the United States

Richman, Douglas Da,b; Morton, Sally Cc; Wrin, Terrid; Hellmann, Nicholasd; Berry, Sandrac; Shapiro, Martin Fc,e; Bozzette, Samuel Aa,b,c

Author Information
doi: 10.1097/01.aids.0000131310.52526.c7
  • Free



Drug-resistant HIV evolves as virus replicates in the presence of the selective pressure of drug treatment [1]. Resistance to a drug diminishes the efficacy of that drug and often of members of the same drug class as well, thus diminishing the probability of identifying an effective subsequent treatment regimen [2]. By diminishing the efficacy of antiretroviral therapy, HIV drug resistance has negative implications both for treatment of individuals, for whom effective therapy has been shown to reduce morbidity and mortality, and for public health, since effective therapy can reduce transmissibility. Transmitted drug-resistant virus also impairs the response to treatment in the newly infected patient [3].

Since the initial description of resistant virus during the phase II clinical trial of the first antiretroviral drug, zidovudine [4], drug resistance and testing for it have become a routine part of antiretroviral drug development and clinical management, particularly of patients with virologic failure [1,2,5]. Although the prevalence of resistance has been reported in selected cohorts of limited size and geographic representation, the true prevalence of HIV drug resistance has not been described in any large population [6]. The objective of this study was to estimate the prevalence of antiretroviral drug resistance in a large, well-characterized study population representing adults receiving care for HIV infection throughout the contiguous United States.


The study sample is a subset of the nationally representative HIV Cost and Service Utilization Study (HCSUS) cohort, which represents the 231 000 adults under care for HIV in the contiguous United States at the start of the era of highly active antiretroviral therapy (HAART) in January and February 1996 [7]. We located and contacted for interview 1919 patients from the HCSUS cohort, who were receiving care in urban clinics during January and February of 1996 and were alive in 1998. The patients were asked to provide an anonymous blood sample. Interviewers met the patients at the blood draw center to provide the laboratory with the patient's study ID number for matching with other study data. Interviewers also collected information about current and past antiretroviral therapies, HIV disease status, CD4 cell count and viral load history. Blood samples were shipped to Quest Diagnostics (San Juan Capistrano, California, USA), centrifuged to separate cells and plasma, which were divided into aliquots and frozen at −70°C. Blood was successfully obtained, and CD4 and viral load information could be determined for 1797 patients, representing the 208 900 adults under care who survived from early 1996 to late 1998 (Fig. 1). This study focuses on results of drug susceptibility assays performed on 1099 blood specimens representing the 132 500 (63.4% of survivors) adults in care with ≥ 500 HIV RNA copies/ml plasma (`viremic’ subset of the cohort population). The study procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional or regional) and with the Helsinki Declaration of 1975, as revised in 1983.

Fig. 1. The baseline HIV Cost and Service Utilization Study cohort consisted of 2864 randomly sampled members representing approximately 231 400 HIV-infected individuals who received care in early 1996 in the contiguous United States [7].:
We focus on two subsamples of this cohort. The first consists of the 1797 who survived until blood specimens were drawn in 1998, and who represent approximately 208 900 ‘survivors'. The second consists of the 1099 who survived and had a viral load ≥ 500 copies HIV RNA/ml plasma and who represent the ‘detectable viral load’ or ‘viremic’ subpopulation of approximately 132 500 individuals. The box representing this population on whom drug resistance assays were performed is shaded. *778 subjects are accounted for by analytic weights [10].

Quantitative CD4+ T-cell determinations were performed on blood samples by flow cytometric analysis by Quest Diagnostics. Aliquots of plasma were assayed for levels of HIV RNA by the Gen-Probe HIV-1 viral load assay [8]. Drug susceptibility assays were performed by the ViroLogic PhenoSense HIV assay which amplifies a gag–pol amplicon from plasma HIV RNA and inserts it into a resistance test vector [9]. Drug susceptibility is expressed as a ratio of the 50% inhibitory drug concentration (IC50) of the patient's plasma virus in comparison with a standard reference HIV-1 strain, NL4-3. Drug susceptibility was measured against the 15 antiretroviral drugs approved by the FDA as of early 2001. Drug resistance was defined for each drug by the IC50 ratio associated with a significantly decreased clinical response to treatment with the drug in clinical trials (abacavir, didanosine, stavudine, and lopinavir) or, when clinically defined resistance levels were not available, by the greater of either the upper 95% confidence interval (CI) for reproducibility of IC50 ratios from repeated testing of clinical virus isolates or the upper 95% CI for drug susceptibility from > 1400 wild-type, patient-derived virus isolates (number of tested isolates ranged from 1430 to 1515 per drug). Patient virus : reference virus IC50 ratios above the following values were considered to be indicative of drug resistance: abacavir 4.5; didanosine 1.7; lamivudine 1.8; stavudine 1.7; zalcitabine 1.7; zidovudine 2.3; delavirdine 4.7; efavirenz 2.2; nevirapine 3.4; amprenavir 1.9; indinavir 1.9; lopinavir 10; nelfinavir 3.0; ritonavir 2.3; and saquinavir 1.9.

We constructed analytic weights accounting for sampling and attrition to adjust the sample to represent the reference population [10]. To adjust standard errors and statistical tests for the differential weighting and complex sample design, we used the linearization method [11] implemented in the statistical package Stata [12]. For patient subgroups defined by covariates chosen a priori, such as key demographics and antiretroviral use history, we report estimated population sizes and weighted proportions. For each covariate, we present a chi-squared test of association, and a pairwise test for each subgroup versus a reference subgroup. We estimate adjusted odds ratios from a multivariate logistic regression model predicting drug resistance as a function of the covariates, and for each covariate present an F-test and a Wald pairwise t-test for each subgroup versus the reference subgroup.


Drug resistance to one or more drugs was detected in virus from specimens representing an estimated 101 100 patients or 76.3% (95% CI, 73.0–79.2%) of the 132 500 surviving adults with more than 500 copies/ml HIV RNA plasma (Table 1). Among the viremic patients, the estimated prevalence of resistance to one or more drugs within each of the three drug classes ranged from 71.4% (95% CI, 67.6–74.9%) for nucleoside reverse transcriptase inhibitors (NRTI) to 40.5% (95% CI, 36.8–44.2%) for protease inhibitors and 25.2% (95% CI, 21.9–28.8%) for non-nucleoside reverse transcriptase inhibitors (Fig. 2). Lamivudine, an NRTI, was the single drug with the highest estimated prevalence of resistance (67.8%; 95% CI, 64.6–70.9%). Resistance to more than one class of drug, termed multiple drug resistance, was detected in an estimated 47.7% (95% CI, 43.6–51.8%) of the viremic population (63 200 patients). Resistance to all three drug classes was detected in an estimated 13.1% (95% CI, 10.6–16.1%) (17 300 patients). Details of resistance prevalence by drug, drug classes, and demographic group is provided in Table 2, and in more detail in Table 3.

Fig. 2. Prevalence of estimated HIV drug resistance in the represented populations.:
*Represents 63% of total study population. PI, protease inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor.
Table 1:
Demographics of the study population.
Table 2:
Characteristics of population represented by 1099 study specimens: the 132 500 American adults in care who survived to autumn of 1998 with Plasma HIV RNA > 500 copies/ml, and the proportion with any drug resistance.
Table 3:
Estimated levels of resistance (%) among American adults with plasma HIV RNA ≥ 500 copies/ml.

Resistance was much more prevalent among patients who were very early adopters of HAART or who were taking nucleoside analogs at the start of the HAART era (1996) compared with patients who had not taken antiretroviral therapy up to that time. For example, 2 years later, resistance to any drug was estimated to be present in 87, 82, and 43% of these subpopulations, respectively, and resistance to all three drug classes was present in 27, 11, and 2%.

An estimated 88% of viremic survivors taking antiretroviral therapy when blood was collected had detectable resistance to one or more drugs compared with 30% of those not currently taking therapy (P = 0.001). A significantly higher prevalence of resistance was also associated with advanced disease stage [odds ratio (OR), 2.95; 95% CI, 1.04–8.35], lower current viral load (OR, 1.57; 95% CI, 1.18–2.09), and lowest self-reported CD4+ T-cell count (OR, 11.07; 95% CI, 4.24–28.89), but not current CD4+ T-cell count (OR, 1.24; 95% CI, 0.59–2.60). A significantly higher resistance prevalence was also associated with male sex (OR, 1.63; 95% CI, 1.16–2.28), being a man who has sex with men (OR, > 1.25 with 95% CI, > 1 for all other risk groups), insurance coverage (OR, 1.87; 95% CI, 1.28–2.73), and more education (OR for college or more, 1.72, 95% CI, 1.20–2.48). However, among all these predictors in univariate analysis, only lowest reported CD4 count (OR, 7.51; 95% CI, 1.93–29.17) and current viral load (OR, 2.91; 95% CI, 1.93–4.39) were demonstrated to be persistent independent predictors of resistance in a multiple logistic regression. (Table 4).

Table 4:
Multivariate adjusted odds ratios.

As the cutoff criteria used to define resistance were not derived from treatment response criteria for all drugs, the study results may slightly over or under-estimate the true drug resistance prevalence, although the cutoffs utilized reflect the best current estimate of clinically significant criteria for impaired treatment responses due to drug resistance. When a much more conservative IC50 ratio of 10 is used to define resistance for all classes of drugs, despite the fact that with several drugs significantly impaired treatment responses have been documented with lower cutoff values, the prevalence of resistance to one or more drugs among the viremic population decreased slightly to 72.6%, whereas the prevalence of resistance to non-nucleoside reverse transcriptase inhibitors decreased to 21.8% and protease inhibitors to 27.4%. These small reductions in estimated prevalence of resistance did not substantially impact the results from analyses of drug resistance risk factors.


These first estimates of the prevalence of HIV drug resistance in adults across the United States have several clinical and public health implications, mostly deriving from the fact that suppression of circulating HIV is an important goal for improving patient outcomes and reducing transmission. We found that most adult Americans who received medical care for HIV infection at urban clinics at the start of the HAART era, including essentially all urban residents and over half of the small number of rural residents receiving HIV care, survived until late 1998. However, even when considering all patients including those with early disease and those not on therapy, most survivors had viremia with > 500 copies HIV RNA/ml plasma, and most of these viremic patients had drug-resistant virus. Clinicians and policymakers need to be aware that this large population of patients with viral loads above 500 copies/ml while on therapy are likely to have more limited treatment options and a diminished probability of complete suppression of viral replication as a treatment outcome.

The drug resistance rates reported do not reflect the level of resistance among those with very low viral loads in whom resistance was not measured, and should not be generalized to that population. Even if all patients with < 500 copies HIV RNA/ml plasma were assumed to harbor no drug-resistant virus, then an estimated 48% of all 208 900 surviving adults would have drug resistance. Nevertheless we know that suppression of viremia with potent combination therapy in patients with drug-resistant virus can be attained, but the resistant virus can be archived indefinitely in the latently infected cell reservoir [13–15]. Thus the true rate in the total population regardless of plasma HIV RNA falls between the two values. The patient population characterized in this study represents a large reservoir for potential transmission of drug-resistant virus, consistent with the reports of increasing rates of transmission of drug-resistant HIV in North America with resulting impaired treatment responses and heightened urgency to prevention efforts targeted at this group [3]. An additional public health concern is that patients infected with resistant virus may not come to medical attention or receive specific prevention messages for a considerable period of time. Only about one-half of the between 850 000 and 950 000 Americans infected with HIV get regular care, and that an estimated one-quarter are unaware of their infection [7,16].

As the data were generated from specimens taken 3 years into the era of potent combination antiretroviral therapy with protease and reverse transcriptase inhibitors, factors have been in place that could potentially impact the prevalence of drug resistance either higher or lower. The more prolonged and wider utilization of the non-nucleoside reverse transcriptase inhibitors and protease inhibitors may have increased resistance to these drug classes as well as multiple class resistance. On the other hand the diminishing practice of sequential therapy and the availability of more effective and better tolerated combination regimens, especially for patients without prolonged nucleoside treatment experience, have been shown to increase the likelihood of suppression of viremia with resulting prevention of acquired resistance [17,18]. Active surveillance efforts will be required to monitor the trends of drug resistance among HIV-infected populations in order to assess the evolution of resistance patterns and to define optimum HIV treatment and prevention strategies. Nevertheless, these data indicate the magnitude of drug resistance that can be selected in a decade for nucleoside reverse transcriptase inhibitors in only 1 to 2 years for non-nucleoside reverse transcriptase and protease inhibitors.

Clinical approaches to address this growing drug resistance problem include the routine use of drug resistance testing to manage patients, development of new drugs active against drug-resistant virus, and the more careful and effective use of these drugs by both health care providers and patients. Because of the high rates of replication and mutation of HIV, the extensive use of antiretroviral therapy provides one of the most dramatic examples of the impact of human intervention on evolution in an ecological system [19]. HIV drug resistance frighteningly recapitulates the history of antimicrobial drug resistance in bacteria, with a pernicious twist: HIV is not curable and drug-resistant variants are archived within each patient for life. In addition, the ability of HIV to avoid inhibition by antiretroviral therapy through accumulating mutations indicates that newer and more effective therapies will continue to be needed to control the pandemic.


We gratefully acknowledge many generous sources of support: Agency for Healthcare Research and Quality, Department of Veterans Affairs (San Diego VA Research Center for AIDS and HIV Infection), Health Services and Resource Administration, National Institutes of Health (Center for AIDS Research and R01 AI 29164), ViroLogic, Quest Diagnostics, Gen-Probe, Glaxo Wellcome, Bristol Myers Squibb, Merck.


1. Hirsch MS, Conway B, D'Aquila RT, Johnson VA, Brun-Vézinet F, Clotet B, et al.Antiretroviral drug resistance testing in adults with HIV infection.JAMA 1998; 279:1984–1991.
2. Hirsch MS, Brun-Vezinet F, D'Aquila RT, Hammer SM, Johnson VA, Kuritzkes DR, et al.Antiretroviral drug resistance testing in adult HIV-1 infection.JAMA 2000; 283:2417–2426.
3. Little SJ, Holte S, Routy JP, Daar ES, Markowitz M, Collier AC, et al.Antiretroviral drug susceptibility and response to initial therapy among recently HIV infected subjects in North America: a study from the Acute Infection and Early Disease Research Program (AIEDRP).N Engl J Med 2002; 347:385–394.
4. Larder BA, Darby G, Richman DD. HIV with reduced sensitivity to zidovudine (AZT) isolated during prolonged therapy.Science 1989; 243:1731–1734.
5. Hirsch MS, Brun-Vezinet F, Clotet B, Conway B, Kuritzkes DR, D'Aquila RT, et al.Antiretroviral drug resistance testing in adults infected with human immunodeficiency virus type 1: 2003 recommendations of an International AIDS SocietyUSA panel.Clin Infect Dis 2003; 37:113–128.
6. Little SJ. Transmission and prevalence of HIV resistance among treatment-naive subjects.Antiviral Ther 2000; 5:33–40.
7. Bozzette SA, Berry SH, Duan N, Frankel MR, Leibowitz AA, Lefkowitz D, et al.The care of HIV-infected adults in the United States.N Engl J Med 1998; 339:1897–1904.
8. Emery S, Bodrug S, Richardson BA, Giachetti C, Bott MA, Panteleeff D, et al.Evaluation of performance of the Gen-Probe human immunodeficiency virus type 1 viral load assay using primary subtype A, C, and D isolates from Kenya.J Clin Microbiol 2000; 38:2688–2695.
9. Petropoulos CJ, Parkin NT, Limoli KL, Lie YS, Wrin T, Huang W, et al.A novel phenotypic drug susceptibility assay for human immunodeficiency virus type 1.Antimicrob Agents Chemother 2000; 44:920–928.
10. Duan N, McCaffrey DF, Frankel MR, St Clair PA, Beckman R, Keesey JW, et al.HCSUS baseline methods technical report: weighting, imputation and variance estimation. 1999. MR-1060-AHCPR, RAND.(Report)
11. Kish L, Frankel MR. Inference from complex samples.J Royal Stat Soc B 1974; 36:1–37.
12. Stata Statistical Software. College Station: Stata Corporation, 2001.
13. Gulick RM, Mellors JW, Havlir D, Eron JJ, Gonzalez C, McMahon D, et al.Treatment with indinavir, zidovudine, and lamivudine in adults with human immunodeficiency virus infection and prior antiretroviral therapy.N Engl J Med 1997; 337:734–739.
14. Wong JK, Hezareh M, Günthard H, Havlir DV, Ignacio CC, Spina CA, et al.Recovery of replication-competent HIV despite prolonged suppression of plasma viremia.Science 1997; 278:1291–1294.
15. Finzi D, Hermankova M, Pierson T, Carruth LM, Chaisson RE, Quinn TC, et al.Identification of a reservoir for HIV-1 in patients on highly active antiretroviral therapy.Science 1997; 278:1295–1300.
16. Fleming P, Byers R, Sweeney P, Daniels D, Karon J, Janssen R. HIV prevalence in the United States, 2000.Ninth Conference on Retroviruses and Opportunistic Infections, Seattle, 2002 [abstract].
17. Robbins GK, DeGruttola V, Shafer RW, Smeaton LM, Snyder SW, Pettinelli C, et al.Comparison of sequential three drug regimens as initial therapy for HIV-1 infection.N Engl J Med 2003; 349:2293–2303.
18. Yeni PG, Hammer SM, Carpenter CCJ, Cooper DA, Fischl MA, Gatell JM, et al.Antiretroviral treatment for adult HIV-1 infection in 2002: updated recommendations of the International AIDS Society USA.JAMA 2002; 288:222–235.
19. Palumbi SR. Humans as the world's greatest evolutionary force.Science 2001; 293:1786–1790.

HIV; antiretroviral drugs; drug resistance; AIDS

© 2004 Lippincott Williams & Wilkins, Inc.