Napravnik, Sonia PhD*; Edwards, David MPH†; Stewart, Paul PhD‡; Stalzer, Brant BSc*; Matteson, Elizabeth*; Eron, Joseph J Jr MD*
Reduction in plasma HIV RNA below levels of detection is a primary marker of antiretroviral therapy success, generally leading to immune reconstitution, improved clinical outcomes, and limited emergence of mutations conferring drug resistance.1,2 However, many patients in routine clinical care experience sustained detectable viral replication while on potent combination antiretroviral therapy.3,4 Clinical management in this situation depends on the patient's HIV disease progression, therapeutic history, and suspected cause of the ongoing viremia, including suboptimal adherence, reduced pharmacologic drug levels, or diminished antiretroviral drug susceptibility.
Switching to at least 2 antiretrovirals to which the patient's HIV variants remain susceptible is generally recommended when drug resistance is suspected as the cause of detectable HIV replication.1,5 However, exposing a patient to new drugs risks experience with new agents the patient may not tolerate, emergence of resistance to additional agents possibly conferring cross-resistance, and exhaustion of future therapeutic options. Moreover, given the extensive antiretroviral therapy experience of some patients and the substantial degree of cross-resistance within classes of antiretroviral agents, 2 drugs to which a patient's HIV variants remain susceptible may not exist.
Because maintaining a patient on a nonsuppressive antiretroviral combination confers some immunologic and virologic benefit,6 remaining on a suboptimal regimen may be an attractive alternative to switching in some patients, especially if treatment options are limited. However, this approach risks the emergence of HIV variants with polymorphisms conferring additional drug resistance, further limiting therapeutic efficacy and jeopardizing subsequent treatment opportunities. In this study we estimated the likelihood, and predictors, of the emergence of incident mutations associated with reduced antiretroviral drug susceptibility, among patients maintained on a potent combination therapy with persistent detectable plasma HIV-1 RNA.
All HIV-infected patients 18 years of age or older and receiving primary HIV care at the University of North Carolina HIV Clinic are approached for their willingness to participate in an observational and prospective clinical cohort study. This study was established in January 2000 and has continuing enrollment. Written informed consent is obtained from all participants, with <2% of patients approached declining participation. Clinical and demographic data are collected at enrollment and at subsequent 6-month intervals, through standardized medical records abstractions.
For this study we identified all patients participating in the cohort study (n = 1605) with results of 2 genotypic antiretroviral resistance tests obtained more than 30 days apart (n = 242). Time at risk was defined as the interval between the 1st (baseline) and 2nd (follow-up) genotypes. Patients were excluded if they had any change in their therapy in the 30 days prior to baseline or during the time at risk (n = 144).
Demographic and clinical characteristics were obtained through existing institutional databases and periodic medical chart reviews. HIV RNA levels were measured using standard or ultrasensitive quantitative reverse transcriptase polymerase chain reaction (quantification limit 400 and 50 HIV RNA copies/mL, respectively; Amplicor HIV Monitor Test; Roche Diagnostic Systems, Inc., Branchburg, NJ). All sequencing was performed using commercially available assays, with 95% of assays using HIV Genosure or Genosure Plus (LabCorp., Research Triangle Park, NC). The Genosure Plus also includes the VirtualPhenotype (Virco, Mechelin, Belgium), which relies on a large database of resistance tests from HIV-1 clinical isolates in which both the genotype and phenotype are known.7 Based on the pattern of mutations a composite is calculated as the average fold-change of 50% inhibitory concentration (IC50) of the pool of variants with matching genotypes and phenotypes in the database.
In accordance with guidelines proposed by the International AIDS Society-USA,8 we defined major protease inhibitor (PI) mutations as amino acid changes at protease gene positions 30, 33, 46, 48, 50, 82, 84, and 90 and minor mutations at 10, 20, 24, 32, 36, 47, 53, 54, 63, 71, 73, 77, and 88. Mutations associated with reduced susceptibility to nucleoside (tide) reverse transcriptase inhibitors (NRTIs) were defined as mutations at reverse transcriptase gene positions 41, 44, 65, 67, 69, 70, 74, 115, 118, 184, 210, 215, and 219, including the 151 complex (62, 75, 77, 116, and 151). Nonnucleoside reverse transcriptase inhibitor (NNRTI) mutations included amino acid substitutions in RT gene positions 100, 103, 106, 108, 181, 188, 190, 225, 230, and 236.
Mutations conferring reduced susceptibility to each drug class (PI, NRTI, and NNRTI) were described at both baseline and follow-up. Acquired mutations were defined as mutations appearing at follow-up that were not detected at baseline. In our primary analyses we included both major and minor PI mutations and assessed the effect of excluding minor PI mutations in sensitivity analyses. For calculating the proportion of patients with resistance to each drug class, we included in the denominators only patients who were receiving an antiretroviral agent in that drug class as part of their current therapy. For comparison of categorical variables we used the Fisher exact test, and for continuous measures the Wilcoxon rank sum test, with 2-sided P values reported in all cases.
Incidence rates (and associated 95% CIs) of acquiring mutations were estimated as the number of new mutations divided by person-time at risk. To estimate incidence rate ratios, and associated 95% CIs, for the number of acquired mutations per person-year, we relied on Poisson log-linear regression models. Model fit was assessed by the deviance χ2 test, and we calculated likelihood ratio tests of fitting alternate distributions, including the negative binomial.
We considered a priori possible predictors of resistance evolution including demographic factors such as age, sex, and race. Several measures of HIV RNA levels were also assessed, including peak, baseline, and follow-up HIV RNA. Additionally, we calculated the average HIV RNA during the time at risk and fit linear regression models to estimate HIV RNA slope during the time at risk. Similarly for CD4 cell counts, we considered nadir, baseline, and follow-up values and the average value and change during the time at risk. Antiretroviral therapy-related measures included time from therapy initiation, time on the current therapy, number of prior regimens, number of prior antiretroviral agents, type of current therapy (PI, NNRTI, PI and NNRTI, or NRTI only), and number of drug resistance mutations identified at baseline.
All predictors with P values <0.10 in the bivariate analyses were entered into a full model. In building the final predictive model, variables with likelihood ratio P values >0.05 were removed from the full model by using a backward stepwise elimination procedure. For models that included an average or change in HIV RNA level, or CD4 cell count, we used robust standard errors.9,10 To examine more thoroughly the effect of continuous measures on the likelihood of acquiring mutations, we assessed several variable specifications in the log-linear Poisson regression models, including continuous, categorical, and lower and upper tail restricted quadratic splines. Likelihood ratio tests were used to contrast nested models, and the simplest specification that adequately represented the data was used in the predictive models.11
For assessing antiretroviral drug susceptibility, we entered all mutations identified at the baseline and follow-up genotypes into an online and publicly available HIV drug resistance database, which provides a rule-based interpretation of drug resistance.12 The algorithm produces a susceptibility score to each drug and we relied on a score indicating intermediate- or high-level resistance in defining drug resistance. Additionally a VirtualPhenotype was available for 75 patients at both baseline and follow-up. Using this restricted sample, we defined drug resistance if the average fold change in IC50 was above the normal susceptibility range, as defined by the assay. The number of drugs a patient remained susceptible to was summed for each antiretroviral drug class based on both the rule-based interpretation and the predicted phenotype.
All statistical analyses were conducted with the SAS statistical package (version 9.1, SAS Institute, Inc., Cary, NC). This study was approved by the University of North Carolina Chapel Hill Committee on the Protection of the Rights of Human Subjects.
Almost one-third of the 98 patients who met study inclusion criteria were women (29%), 34% were white, 64% African American, and 2% Hispanic, with a median age of 37 years (interquartile range [IQR], 33-44). One-quarter were men who have sex with men (26%), and 11% had a history of injection drug use. At baseline the majority of patients had considerable antiretroviral experience. The median number of prior regimens was 3 (IQR, 1-5), and the current therapy was the 1st for 15% of patients, 2nd for 11%, 3rd for 18%, and for 55% of patients at least their 4th regimen. The median time from therapy initiation to baseline was 4 years (IQR, 2-6). The median number of antiretroviral agents patients had been exposed to by baseline was 6 (IQR, 4-9), with 99% having previously received at least 1 NRTI, 88% a PI, and 63% an NNRTI. The median time patients received their current therapy prior to baseline was 8.9 months (IQR, 4.1-16.1 months). Almost all patients (98%) were taking an NRTI as part of their current therapy, with the majority of regimens being PI based (55%), although 14% were NNRTI based, 14% PI and NNRTI based, and 16% of patients received only NRTIs. Two patients were taking Enfuvirtide.
Immunologic and Virologic Characteristics
The median time between the baseline and follow-up genotypic resistance tests was 9.3 months (IQR, 4.2-15.5 months). Median CD4 cell counts were substantially higher at current therapy initiation, baseline, and follow-up, compared with the median nadir value (99 CD4 cells/μL; IQR, 24-224 cells/μL), and median HIV RNA levels were considerably lower, in comparison with the median peak value (5.2 log10 HIV RNA copies/mL; IQR, 4.7-5.6 log10 HIV RNA copies/mL; P < 0.001 for all contrasts). CD4 cell counts and HIV RNA levels were similar at current therapy initiation and baseline (P = 0.11 and 0.15, respectively). During the time at risk, HIV RNA levels were observed to increase, with CD4 cell counts remaining essentially stable (Table 1).
Antiretroviral Drug Resistance Mutations
In all cases sequences were subtype B, and at baseline 88% (n = 86) had at least 1 mutation associated with reduced susceptibility to antiretroviral drugs, with a median of 3 mutations observed per patient (IQR, 1-5) (Table 1, Fig. 1). Among patients receiving PIs at baseline, 79% (54/68 patients) had evidence of at least 1 major or minor mutation to PIs, and 37% (25/68 patients) had evidence of at least 1 major mutation to PIs. Among the 96 patients receiving an NRTI at baseline, 69% (66/96 patients) had at least 1 NRTI mutation. Although a larger proportion of patients had at least 1 mutation at follow-up than at baseline, these differences were not statistically significant for PIs or NRTIs. However, a significant increase in NNRTI-associated mutations was observed among patients receiving an NNRTI, with 57% (16/28 patients) having at least 1 NNRTI-associated mutation at baseline, and 86% (24/28 patients) at follow-up (P < 0.001).
Fifty-nine patients (60%) acquired at least 1 incident resistance mutation during their time at risk, with a median number of new mutations of 1 (IQR, 0-2) (Table 2, Fig. 1). Similar relative frequencies of acquired mutations were observed at follow-up, to those evident at baseline, in each of the drug classes (Fig. 2). The M184V mutation was the most commonly observed NRTI-associated mutation, and 42% of patients who were receiving lamivudine without evidence of the M184V at baseline acquired this mutation during follow-up. Among 12 patients who were receiving an NNRTI and had no evidence of an NNRTI-associated mutation, 8 (67%) acquired at least 1 NNRTI-associated mutation during follow-up. Just over half of the patients (51%) with a time at risk for <6 months (n = 39) acquired at least 1 mutation, in comparison with 59% of patients with a time at risk from 6-12 months (n = 22), and 70% of patients with a time at risk >12 months (n = 37). The overall incidence rate was 1.61 mutations per person-year at risk (95% CI: 1.36-1.90). The overall incidence rate was lower when we excluded minor PI mutations (incident rate = 1.08; 95% CI: 0.87-1.32).
No differences were observed in incidence rates by sex (P = 0.18); race (P = 0.52); age (P = 0.51); number of prior antiretroviral drugs received (P = 0.27); number of prior antiretroviral regimens (P = 0.85); type of current therapy (P = 0.28); or any of the CD4 cell count measures considered, eg, nadir CD4 cell count and average CD4 cell count during time at risk (P = 0.31 and 0.44, respectively). Years of prior therapy were associated with the development of new mutations in bivariate analyses (P = 0.004) but did not remain predictive in multivariable analyses (P = 0.06).
The number of mutations at baseline, average HIV RNA level, and HIV RNA slope during the time at risk remained independent predictors in the final model (Table 3). Additional HIV RNA measures were predictive in bivariate analyses but did not remain predictive in the full model. For example, for baseline HIV RNA level, the P value in bivariate analyses was 0.047, and it was 0.64 in multivariable analyses.
Patients with decreasing or stable HIV RNA levels (≤0.2 change in log10 HIV RNA copies/mL per month) had a lower incidence rate per person-year of acquiring mutations than patients with increasing HIV RNA levels (ie, >0.2 change in log10 HIV RNA copies/mL per month) (1.50, 95% CI: 1.25-1.80 vs. 2.90, 95% CI: 1.95-4.33, respectively) (Table 3). This association remained in adjusted analyses, in which patients with a >0.2 increase in log10 HIV RNA copies/mL per month were at greater risk of acquiring mutations than patients with either decreasing or stable HIV RNA levels (incidence rate ratio 1.93, 95% CI: 1.23-3.01). We did not observe differences in the types of antiretroviral drug resistance mutations acquired according to whether patients experienced decreasing, stable, or increasing HIV RNA levels. In all cases, patients with increasing HIV RNA levels were more likely to acquire NRTI, NNRTI, major PI, and minor PI mutations in comparison with patients with stable or decreasing HIV RNA levels.
The risk of acquiring mutations was greatest among patients with average HIV RNA levels between 3-4 HIV RNA log10 copies/mL and lower among patients with average HIV RNA levels <3 or >4 HIV RNA log10 copies/mL (unadjusted incidence rates 2.34, 1.11, and 0.98, respectively). In adjusted analyses, patients with average HIV RNA levels of 3-4 log10 copies/mL were at twice the risk of acquiring mutations in comparison with patients with average HIV RNA levels >4 log10 copies/mL (incidence rate ratio 2.11, 95% CI: 1.43-3.09). However, patients with average HIV RNA levels <3 log10 copies/mL were at a similar risk of acquiring mutations compared with patients with average HIV RNA levels >4 log10 copies/mL (incidence rate ratio 0.84, 95% CI: 0.39-1.79).
Patients with no mutations at baseline had the greatest incidence rate per person-year of acquiring mutations in comparison with patients with 1-3 or >3 mutations at baseline (incidence rates 3.11, 0.99, and 1.86, respectively) (Table 3). Among the 12 patients with no baseline mutations, 5 were on an NNRTI, 6 on a PI, and 1 patient received only NRTIs. All patients received at least 1 NRTI, and 7 acquired at least 1 new NRTI mutation (58%). Seven patients were taking lamivudine, and of these, 4 (57%) acquired the M184V during follow-up. Of the 5 patients receiving an NNRTI, all were taking efavirenz and 4 acquired a new NNRTI mutation (80%), with 1 patient acquiring K103N, 1 patient G190A, 1 patient K103N and P225H, and 1 patient K103N, V108I, and G190A. Of the 6 patients on a PI, 3 (50%) did not acquire a PI mutation during follow-up and received amprenavir/ritonavir, indinavir/ritonavir, or lopinavir/ritonavir. One patient taking nelfinavir acquired M36I, and 2 patients taking amprenavir/ritonavir both acquired major PI mutations V82A and I84V, as well as minor PI mutations, including L10I/F and M46I/L.
In adjusted analyses, patients with no mutations at baseline were at >3 times the risk of acquiring a mutation in comparison with patients with 1-3 mutations at baseline (incidence rate ratio 3.15, 95% CI: 1.98-5.03). Patients with >3 mutations at baseline were also at increased risk of acquiring mutations during follow-up in contrast to patients with 1-3 mutations at baseline (adjusted incidence rate ratio 1.74, 95% CI: 1.16-2.60).
We assessed independent predictors of acquiring mutations in additional multivariable models. We fit models where we restricted the type of mutations by excluding minor PI mutations. These models estimated the overall incidence rate ratios for acquiring at least 1 NNRTI, NRTI, or major PI mutation. We obtained analogous results to those including all mutations presented here. Additionally we fit separate models among patients who were taking a PI vs. those taking an NNRTI. In each case we observed similar relationships with baseline number of mutations, average HIV RNA, and HIV RNA slope as described here. The only difference observed was in models including only patients taking an NNRTI, in which with a sample size of only 28 patients the associations did not reach statistical significance.
Antiretroviral Drug Susceptibility
Based on the genotypic rule-based interpretation among the patients who received an NRTI (n = 96), PI (n = 68), or NNRTI (n = 28), 23%, 15%, and 32% lost susceptibility to at least 1 antiretroviral drug in that class, respectively. However, the majority of patients maintained susceptibility to most NRTI and PI agents (Fig. 3). No significant differences were observed in the predicted number of active NRTIs, or PIs, at baseline, in contrast to follow-up (P = 0.33 and 0.41, respectively). However, among patients receiving an NNRTI, the number of drugs patients remained susceptible to decreased from baseline to follow-up (P = 0.046). Analogous results were obtained based on the predicted phenotype.
In our sample of 98 HIV-infected patients with relatively substantial antiretroviral therapy experience, 88% (95% CI: 80%-94%) had evidence of at least 1 mutation at baseline, and 93% (95% CI: 86%-97%) a median 9.3 months later (IQR 4.2-15.5). High prevalence estimates of decreased susceptibility to antiretroviral agents among patients receiving routine clinical care have been reported by others.13,14 Consistent with prior research,15,16 60% of our patients acquired at least 1 mutation while being maintained on a stable antiretroviral regimen with ongoing detectable HIV replication, for an incidence rate of 1.61 mutations per person-year (95% CI: 1.36-1.90).
Two HIV RNA measures remained independently predictive of acquiring mutations in our analyses: average HIV RNA and HIV RNA slope during the time at risk. The greatest risk for acquiring drug resistance mutations occurred among patients whose average viremia was 3.0-4.0 log10 copies/mL. The relatively low rate of resistance evolution we found among patients at low levels of HIV replication is consistent with very low rates observed in individuals with HIV RNA levels <50 copies/mL.2,17 Occurrence of new mutations is a function of replication rate, and at HIV RNA levels <1000 copies/mL, ongoing continuous rounds of replication are likely to be low. Conversely, at higher replication rates new mutations are more likely to occur. However, these mutations will become predominant in the population only if they convey an advantage in the face of selective pressure. Because sufficiently high drug levels must be present to exert positive selective pressure on viral variants with decreased drug susceptibility, the observed low rate of resistance evolution at high average HIV RNA level is likely a function of diminished therapeutic drug levels or limited adherence.18,19
The rate of resistance evolution was directly proportional to increasing HIV RNA slope. This is consistent with a simple Darwinian selection process, because viral quasispecies with ≥1 mutation conferring reduced drug susceptibility have a selective advantage in the presence of drug pressure and will rapidly outgrow other quasispecies, resulting in an increase in HIV RNA. Three prior studies did not find a correlation between baseline degree of viral replication and evolution of drug-resistant mutations.15,16,20 Consistent with these studies, we also found that the HIV RNA level measured at the 1st genotype test only marginally predicted acquiring mutations in bivariate analyses, with no association evident in multivariable analyses.
After accounting for average level of viremia and change in HIV RNA during the time at risk, the number of mutations detected at baseline remained an independent predictor of acquiring mutations during follow-up. The highest incidence rate was observed among patients with no mutations at baseline, possibly reflecting the emergence of minority viral populations not detected at baseline. Given the relatively rapid reversion to wild-type as the dominant viral population once selective drug pressure is withdrawn, it is reasonable to assume that some patients with no detected mutations at baseline may have been harboring minority variants with resistance mutations. Therefore it is possible that among patients with <5 mutations, the risk of acquiring new mutations decreases with decreasing number of existing mutations. Because viral quasispecies harboring drug resistance mutations may have diminished pathogenesis, and most have diminished replicative capacity,6,21 the dominant HIV variants may reflect a balance between increasing selective advantage with additional mutations, and preservation of replicative capacity. With multiple mutations in reverse transcriptase or protease, there may also be structural constraints on the accumulation of additional mutations. This maintenance equilibrium may explain the ceiling effect we observed in the incidence rate of acquiring additional mutations with a greater number of preexisting mutations.
Although 60% of patients acquired mutations during follow-up, the median number of acquired mutations was 1 (IQR 0-2). One mutation may lead to high-level resistance, such as the M184V resulting in lamivudine resistance; however, for many NRTIs or PIs, serial accumulation of multiple mutations is required.22 Therefore when we considered the number of antiretroviral agents to which patients maintained susceptibility during follow-up, we did not observe any significant difference for either the NRTIs or the PIs. Because a single mutation may cause resistance to the NNRTIs, which also have a substantial degree of cross-resistance,22 patients were susceptible to fewer NNRTIs at follow-up, in contrast to baseline.
Our study has several limitations, including reliance on clinically obtained resistance testing and inability to evaluate resistance to fusion inhibitors.23 Commercially available resistance assays have limited sensitivity in detecting minority viral populations,24 which may be heterogeneous.25 Depending on whether the dominant viral populations sampled harbored a greater or lesser degree of antiretroviral drug resistance, our results may be either underestimating or overestimating the evolution of drug resistance. Because viral quasispecies with reduced susceptibility to antiretroviral agents are outgrown by viruses with greater replicative capacity once selective drug pressure is withdrawn, it is possible we are underestimating the magnitude of resistance among patients who were less adherent. Lack of an accurate adherence measure is a limitation of our study.
Predicting the degree of resistance and its clinical relevance of any set of mutations to a specific antiretroviral agent, or combination therapy, is complicated by several factors, including the potential for interactions within and between drug classes.8,26,27 Therefore we augmented our analyses of antiretroviral susceptibility based on a genotypic rule-based algorithm, with predicted phenotypic results. Because the degree of reduced susceptibility correlated with diminished clinical response is not well defined for most antiretroviral agents, we assessed several available cut-offs.28 In all cases our analyses were consistent with the genotype rule-based results.
Current treatment guidelines recommend switching patients to at least 2 active antiretroviral agents at the earliest evidence of detectable viremia related to decreased drug susceptibility.1,5 This approach is appealing if a new regimen is available, well tolerated, and succeeds in limiting HIV replication to undetectable levels, thereby minimizing the probability of acquiring additional resistance. The aggressiveness with which to approach virologic failure, however, must be balanced by the risks associated with exposing patients to additional antiretroviral agents. Our findings demonstrate a relatively slow rate of resistance evolution in patients with HIV-1 subtype B, especially among individuals with multiple mutations, who have stable HIV RNA levels in plasma over time, and who maintain HIV RNA levels <1000 copies/mL. These data suggest that maintaining specific patients on a failing regimen results in relatively slow resistance evolution with limited reduction in antiretroviral drug susceptibility. However, this must be tempered by the specific regimen and prior resistance profile of the patient, because acquiring even a single mutation may cause resistance to all NNRTIs and the M184V mutation leads to resistance to lamivudine and emtricitabine.
Many patients maintained on an incompletely suppressive regimen continue to derive immunologic, virologic, and clinical benefit, possibly due to the reduced replication capacity of mutant HIV variants, enhanced HIV-specific immune responses, and residual drug activity.3,6,29,30 Therefore, delaying switching of suboptimal regimens may be indicated in some patients. However, patients with HIV-1, with limited resistance, especially those with plasma HIV RNA >1000 copies/mL, are at risk for emergence of increasingly resistant virus. Further studies monitoring resistance evolution over time are needed, and combined analyses across observational clinical cohorts would strengthen our initial observations.
We greatly appreciate the support of all study staff members, HIV care providers, the infectious disease clinic staff, and particularly the patients who participated in this study.
1. Panel on Clinical Practices for the Treatment of HIV Infection. Department of Health and Human Services. Guidelines for the use of antiretroviral agents in HIV-1-infected adults and adolescents. Available at: http://www.aidsinfo.nih.gov
. Accessed December 2004.
2. Hermankova M, Ray SC, Ruff C, et al. HIV-1 drug resistance profiles in children and adults with viral load of <50 copies/mL receiving combination therapy. JAMA. 2001;286:196-207.
3. Ledergerber B, Egger M, Opravil M, et al. Clinical progression and virological failure on highly active antiretroviral therapy in HIV-1 patients: a prospective cohort study. Lancet. 1999;353:863-868.
4. Lucas GM, Chaisson RE, Moore RD. Highly active antiretroviral therapy in a large urban clinic: risk factors for virologic failure and adverse drug reactions. Ann Intern Med. 1999;131:81-87.
5. Yeni PG, Hammer SM, Hirsch MS, et al. Treatment for adult HIV infection: 2004 recommendations of the International AIDS Society-USA Panel. JAMA. 2004;292:251-265.
6. Deeks SG, Wrin T, Liegler T, et al. Virologic and immunologic consequences of discontinuing combination antiretroviral-drug therapy in HIV-infected patients with detectable viremia. N Engl J Med. 2001;344:472-480.
7. Hammer S, Peeters M, Harrigan R, et al., and the ACTG 372B/D Study Team. Virtual phenotype is predictive of treatment failure in treatment-experienced patients. Antivir Ther. 2001;6(Suppl 1):107.
8. Johnson VA, Brun-Vezinet F, Clotet B, et al. Update of the drug resistance mutations in HIV-1: 2004. Top HIV Med. 2004;12:119-124.
9. Huber PJ. The behavior of maximum likelihood estimates under nonstandard conditions. In: Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. Berkeley, CA: University of California Press; 1967:221-233.
10. White H. Maximum likelihood estimation of misspecified models. Econometrica. 1982;50:1-25.
11. Witte JS, Greenland S. A nested approach to evaluating dose-response and trend. Ann Epidemiol. 1997;7:188-193.
12. Rhee SY, Gonzales MJ, Kantor R, et al. Human immunodeficiency virus reverse transcriptase and protease sequence database. Nucleic Acids Res. 2003;31:298-303.
13. Richman DD, Morton SC, Wrin T, et al. The prevalence of antiretroviral drug resistance in the United States. AIDS. 2004;18:1393-1401.
14. Lucas GM, Gallant JE, Moore RD. Relationship between drug resistance and HIV-1 disease progression or death in patients undergoing resistance testing. AIDS. 2004;18:1539-1548.
15. Kantor R, Shafer RW, Follansbee S, et al. Evolution of resistance to drugs in HIV-1-infected patients failing antiretroviral therapy. AIDS. 2004;18:1503-1511.
16. Maggiolo F, Callegaro A, Quinzan G, et al. Evolution of HIV reverse transcriptase and protease mutations in patients on antiretroviral therapy [H-2057]. Paper presented at: 42nd Interscience Conference on Antimicrobial Agents and Chemotherapy; Sept. 27-30, 2002; San Diego, CA.
17. Martinez-Picado J, DePasquale MP, Kartsonis N, et al. Antiretroviral resistance during successful therapy of HIV type 1 infection. Proc Natl Acad Sci USA. 2000;97:10948-10953.
18. Bangsberg DR, Hecht FM, Charlebois ED, et al. Adherence to protease inhibitors, HIV-1 viral load, and development of drug resistance in an indigent population. AIDS. 2000;14:357-366.
19. Walsh JC, Pozniak AL, Nelson MR, et al. Virologic rebound on HAART in the context of low treatment adherence is associated with a low prevalence of antiretroviral drug resistance. J Acquir Immune Defic Syndr. 2002;30:278-287.
20. Kristiansen TB, Pedersen AG, Eugen-Olsen J, et al. The evolution of HIV in patients with insufficient viral suppression during HAART. Paper presented at: 42nd Interscience Conference on Antimicrobial Agents and Chemotherapy; Sept. 27-30, 2002; San Diego, CA. Abstract H-2058.
21. Barbour JD, Wrin T, Grant RM, et al. Evolution of phenotypic drug susceptibility and viral replication capacity during long-term virologic failure of protease inhibitor therapy in human immunodeficiency virus-infected adults. J Virol. 2002;76:11104-11112.
22. Hirsch MS, Brun-Vezinet F, Clotet B, et al. Antiretroviral drug resistance testing in adults infected with human immunodeficiency virus type 1: 2003 recommendations of an International AIDS Society-USA Panel. Clin Infect Dis. 2003;37:113-128.
23. Eron JJ, Gulick RM, Bartlett JA, et al. Short-term safety and antiretroviral activity of T-1249, a second-generation fusion inhibitor of HIV. J Infect Dis. 2004;189:1075-1083.
24. Hirsch MS, Brun-Vezinet F, D'Aquila RT, et al. Antiretroviral drug resistance testing in adult HIV-1 infection: recommendations of an International AIDS Society-USA Panel. JAMA. 2000;283:2417-2426.
25. Resch W, Parkin N, Stuelke EL, et al. A multiple-site-specific heteroduplex tracking assay as a tool for the study of viral population dynamics. Proc Natl Acad Sci USA. 2001;98:176-181.
26. Haubrich RH, Kemper CA, Hellmann NS, et al. The clinical relevance of non-nucleoside reverse transcriptase inhibitor hypersusceptibility: a prospective cohort analysis. AIDS. 2002;16:33-40.
27. Naeger LK, Margot NA, Miller MD. Increased drug susceptibility of HIV-1 reverse transcriptase mutants containing M184V and zidovudine-associated mutations: analysis of enzyme processivity, chain-terminator removal and viral replication. Antivir Ther. 2001;6:115-126.
28. Bacheler L, Winters B, Harrigan R, et al. Estimation of phenotypic clinical cutoffs for Virco(R)TYPE through meta-analyses of clinical trial and cohort data. Paper presented at: 44th Interscience Conference on Antimicrobial Agents and Chemotherapy; Oct. 30-Nov. 2, 2004; Washington, DC. Abstract H-1133.
29. Deeks SG, Hoh R, Grant RM, et al. CD4+ T cell kinetics and activation in human immunodeficiency virus-infected patients who remain viremic despite long-term treatment with protease inhibitor-based therapy. J Infect Dis. 2002;185:315-323.
30. Archer RH, Dykes C, Gerondelis P, et al. Mutants of human immunodeficiency virus type 1 (HIV-1) reverse transcriptase resistant to nonnucleoside reverse transcriptase inhibitors demonstrate altered rates of RNase H cleavage that correlate with HIV-1 replication fitness in cell culture. J Virol. 2000;74:8390-8401.
© 2005 Lippincott Williams & Wilkins, Inc.