Combination antiretroviral therapy (ART) generally includes two nucleoside analogue reverse transcriptase inhibitors (NRTIs) and one nonnucleoside analogue RTI (NNRTI), one protease inhibitor or, rarely, a third NRTI. Its efficacy is estimated by intention-to-treat (ITT) analyses of virological outcomes in randomized trials of previously untreated adults. Guideline panels mostly recommend regimens after sequential review of individual, randomized trials rather than on formal systematic review [1–5].
Three overviews have explored which drugs might form the most effective ART. Two meta-analyses compared only NNRTI and protease inhibitor-based therapy but included trials of ART-experienced patients [6,7]. One systematic review of 53 48-week, randomized trials or cohorts through July 2004 concluded that two NRTIs with an NNRTI or a ritonavir-boosted protease inhibitor were superior to triple-NRTI or unboosted protease inhibitor-based therapy; efficacy was greater in those with higher baseline CD4 cell counts . These variables, however, explained only 35% of variability in efficacy, suggesting unevaluated factors also affect initial ART efficacy.
The Antiretroviral Cohort Collaboration found independent factors for progression over 5 years to AIDS or death with initial ART were increasing age, HIV acquisition by injection drug use, a prior AIDS-defining illness, a pretreatment HIV viral load above 100 000 copies/ml plasma, and lower CD4 cell count . The third drug type was not significant, even though much protease inhibitor therapy in these cohorts was unboosted. Only 10 factors were analysed, however, as predictors of efficacy.
Better knowledge regarding efficacy and reasons for stopping would assist drug development and guideline recommendations. We analysed all randomized trials and prospective cohorts evaluating combination ART in previously untreated adults. We hypothesized that unevaluated factors encompassing study design and quality, eligibility criteria, participant characteristics and pill-dosing characteristics also affect ART efficacy and tolerability; the latter has not been systematically reviewed.
Data sources and search strategy
We followed QUORUM guidelines regarding search strategy, study selection (except that we also included cohort studies), data abstraction, analysis and presentation.
We searched PubMed, Current Controlled Trials and the Cochrane Clinical Trials Register to December 2007. For each licensed or investigational antiretroviral drug, the search strategy was ‘[drug] and (HIV or antiretroviral) and (cohort or randomized trial)’, restricted by age greater than 13 years and publication since January 1996. We also searched the 2006 and 2007 conference websites of the Conference on Retroviruses and Opportunistic Infections, the International AIDS Society and the Interscience Conference on Antimicrobial Agents and Chemotherapy, from which webcasts and posters of unpublished studies, including as yet unlicensed drugs, were reviewed. We also reviewed studies cited in the above treatment guidelines [1–5], in drug labels and other available documents of the Food and Drug Administration (including superseded drug formulations), in the above-mentioned meta-analyses and systematic review, and in original study protocols where available. Study entry criteria were cross-checked against ClinicalTrials.gov. Study sponsors and authors were approached for unreported eligibility criteria and patient characteristics. No pharmaceutical company was involved in the decision to perform this study or in its analysis plan.
We included all randomized trials and all prospective, cohort studies of at least 24-week duration in previously untreated adult or adolescent populations that evaluated at least one regimen that is or has been recommended (as a ‘preferred’ or ‘alternative’ regimen) in key guidelines published up until December 2007. We also included recent studies that found similar efficacy outcomes to recommended regimens. Key characteristics of included studies are shown in supplementary Table S1. Ritonavir at a boosting dose was not considered to be an antiretroviral drug.
We excluded studies (supplementary Table S2) that did not present ITT data or specify the type of ITT analysis, as well as studies of children and of primary HIV infection. We also excluded treatment groups that only evaluated dual-NRTI regimens, four-drug regimens, protease inhibitor only therapy, alternating regimens and nonrecommended, triple-NRTI regimens (e.g. tenofovir-abacavir-lamivudine ); any group that included more than one-third drug; and therapy that was directly observed. We excluded prospective, cohort studies if data for each regimen were not presented separately. No retrospective study was included. If a study had both ineligible and eligible treatment groups (e.g. a two-drug control group and a three-drug investigational group), only data from the eligible treatment group were included. Data from one randomized trial group were excluded, as they appeared also to have been published subsequently as part of a larger cohort that we included [10,11].
We abstracted data for each group pertaining to study design and quality, eligibility criteria, antiretroviral regimens, participant characteristics and treatment outcomes.
Measures of study design and quality included description as randomized, use of placebo, study phase, sponsor, year of commencement, recruitment by region, time to primary endpoint, type of ITT analysis, journal and year of publication. Masking conditions were not included, as these were rarely reported. For studies or study groups terminated prematurely (but still with at least 6 months of follow-up), study duration was estimated as the mean follow-up for the study or study group. If not stated, we assumed that an academic study had industry support if one of the authors was from a pharmaceutical manufacturer of one of the drugs used in the trial, and classified a study as solely industry-sponsored if a pharmaceutical company employee was the first or last author. As a marker of journal quality, we recorded journal impact factors from ISI 2006 (Thomson Scientific, Philadelphia, Pennsylvania, USA).
In each treatment group, the primary endpoint was the proportion of patients with plasma HIV viral load undetectable by ITT analysis, in which detectable plasma HIV RNA and missing HIV RNA data at the primary endpoint study visit were classified as failures. ITT data were reported by at least one of three approaches. The ITT missing-equals-failure (ITTM=F) analysis considers undetectable viral load at study conclusion as treatment success, regardless of prior, transient viraemia or of any change to assigned treatment. The noncompleter equals failure (ITTNC=F) approach also regards any treatment change as failure. The time-to-loss-of-virological-failure (TLOVR) analysis additionally regards transient viraemia above a defined threshold as treatment failure and has been the recommended analysis since 2002 . For this analysis, we abstracted TLOVR data preferentially, followed by ITTNC=F then ITTM=F data.
Regarding eligibility criteria, we recorded each study's lowest permitted age, CD4 cell count, HIV viral load, haemoglobin, highest permitted serum hepatic transaminase level, and whether participants were excluded for prior AIDS-defining illness, presumed HIV acquisition by injection drug use, or for genotypic resistance at baseline. We assumed that patients with any haemoglobin level, any serum hepatic transaminases, prior AIDS or who acquired HIV by injection drug use were eligible unless specifically stated as ineligible.
For each antiretroviral regimen, we recorded drugs initiated, pills per day (including for placebo), doses per day and dosing relative to food. Unless stated otherwise, dosing characteristics were assumed to be in accordance with the FDA product label at the time of study initiation or in accordance with a subsequent label if an investigational formulation was used.
We recorded the following baseline characteristics for each study group: age, sex, race, risk factor(s) for HIV infection, history of AIDS-defining illness, CD4 cell count, HIV viral load, weight, and serostatus for chronic hepatitis B infection (surface antigen) and hepatitis C infection (antibody). European, North American, Hispanic and South American participants were classified as ‘white’, except that blacks from any region were classified as ‘black’. Asians, Pacific Islanders, and indigenous Australians and indigenous Canadians were classified as ‘other’.
Outcome data recorded were the proportion of participants in each group with viral load undetectable by ITT and the proportions who ceased assigned therapy for adverse events, patient choice, loss to follow-up, virological failure or ‘other reason’ (including unspecified reason, pregnancy, HIV disease progression and death). Whenever possible, these data were abstracted from a CONSORT flow diagram. We preferentially recorded ITT data for a viral load of 50 copies/ml plasma. Regardless of the viral load assay used, each undetectable level was considered to be equivalent, although the vast majority of studies used a Roche assay.
We recorded the incidence of any grade adverse events and of grade 2 or greater adverse events. Whenever specified, adverse events were generally reported using the AIDS Clinical Trials Group scale. Adverse events abstracted were nausea, diarrhoea, headache, rash (which included hypersensitivity reactions) and fatigue, as these were the most commonly reported. We also abstracted the combined incidence of all grade 2 or greater adverse events.
Data from all relevant studies were entered into two, identical databases by one author (A.C.). Each database was then compared and discrepancies resolved by comparison with the primary study report.
Quantitative data synthesis
Design, quality, eligibility, participant and therapy characteristics were described as proportions (for categorical variables) and means (for continuous variables including proportions) with study group used as the unit for analysis weighted by the number of participants per group. The main outcomes were the proportions of patients with an undetectable viral load and that ceased therapy due to an adverse event, predictors of which were determined using study size-weighted linear regression. Multivariable analysis using forward, stepwise, selection procedures was used to determine association with the outcomes of interest . All variables with P less than 0.05 in univariate analysis were assessed in building the multivariable models. Whenever a variable that was described as both an entry criterion and a baseline measure (e.g. eligibility for minimum CD4 cell count and mean, baseline CD4 cell count) was significant in univariate analysis, only the baseline measure was assessed in the multivariable model. Serostatus for viral hepatitis was not assessed in the multivariable models due to the large amount of missing data. No imputations were conducted for missing data. Multivariable analysis only includes records with no missing data. Heterogeneity was assessed using the using the F-test and the Wald test for categorical variables.
Three sensitivity analyses of variables associated with both endpoints were conducted. First, we applied the multivariable model to only the randomized, controlled trials to address whether inclusion of cohorts might be biased to those reporting only positive outcomes. For similar reasons, we also performed an analysis that excluded unpublished studies. Lastly, we evaluated only studies with week-48 outcome data (including longer studies with available week-48 data) to allow for better comparison with the previously reported systematic review . Publication bias was also assessed using a funnel plot .
Study size-weighted Pearson's correlation was used to determine the association between reasons for cessation and types of adverse events. Statistical significance was determined using a two-sided alpha of 0.05. Bubble plots of the association between adverse events and time of study commencement were constructed and weighted in direct proportion to the number of study participants. No adjustment was made for multiple comparisons.
Analyses were performed with SAS version 9.0 (SAS Institute Inc., Cary, North Carolina, USA) and STATA version 10 (STATA Corp., College Station, Texas, USA).
Study groups and participants
The search strategy yielded 1050 articles, of which 931 were not reviewed, as they were substudies of primary reports, review articles, editorials, retrospective studies, groups with mixed antiretroviral regimens, guidelines or citation of included trials in another paper. A further 35 papers were reviewed but were excluded, as they did not fulfil our inclusion criteria. Five papers fulfilled inclusion criteria, but did not report adequate data. Therefore, we analysed 64 randomized trials [39 (61%) with placebo] and 15 prospective, cohort studies, which included 23 067 participants in 143 groups (Table 1). Across study groups, mean age was 36.0 years, 26.4% were women, 35.6% were nonwhite, 12.5% were from Asia or Africa, and the mean CD4 cell count was 286 cells/μl. Per study regimen, the average number of pills prescribed per day was 8.1 in 2.2 doses (Table 2).
After a mean follow-up of 14.3 months, mean treatment efficacy was 59.1% [95% confidence interval (CI) 56.8–61.4]. There was no evidence of publication bias as the distribution of treatment success rate was symmetrical and not related to study size. Parameters independently associated with studies reporting a higher rate of treatment success (Table 3) were the third drug class used (favouring ritonavir-boosted protease inhibitor or NNRTI; P < 0.0001), dosing relative to food (favouring dosing without regard to food and then dosing only with food; P = 0.001), type of dual-NRTI backbone (favouring didanosine or tenofovir with emtricitabine or lamivudine; P = 0.002), nonwhite race (P = 0.002), lower baseline CD4 cell count (P = 0.014), shorter study duration (P = 0.017) and exclusion of patients based on haemoglobin (P = 0.0006). This model accounted for 79% of the variability in the rates of treatment success. The dual-NRTI combination of didanosine with emtricitabine or lamivudine (adjusted mean proportion HIV RNA undetectable 69 and 67%, respectively) was more strongly associated with higher group efficacy than the more frequently recommended and more widely prescribed combination of zidovudine–lamivudine [weighted mean proportion HIV RNA undetectable 56%; difference 15.1% (95% CI 2.7–27.5%)].
Results of the three sensitivity analyses of efficacy were similar to those of the primary analysis. For only randomized study groups, all parameters except study duration and race were significant. For published study groups, all parameters except type of NRTI backbone were significant. For week-48 data, all parameters except race were significant. The fit of these models (r2 = 0.76–0.79) was similar to that of the primary model (r2 = 0.79).
Individual adverse event data were reported in only about half of all studies, with only eight (6%) groups (four studies) reporting all adverse events in Table 4. Bearing in mind the limited data presented, the most common adverse events were diarrhoea (weighted mean = 29%), nausea (25%), headache (18%), rash (15%) and fatigue (13%). Adverse events of grade 2 or greater severity occurred in 21% of participants, an incidence that declined modestly over time (β = −0.04, P = 0.14) (Fig. 1a).
Adverse events were the most common cause of treatment cessation (9.0%), as compared to loss to follow-up (6.1%), patient choice (3.4%), virological failure (2.8%) and ‘other reasons’ (3.4%). Cessation for adverse events declined over time (Fig. 1b).
Studies reporting a high rate of cessation for adverse events were significantly more likely to be phase 2 or 3, academia-sponsored, shorter than 36 months, and that included participants who were older or were recruited in Europe or Australia (r2 = 0.70). The type of dual-NRTI backbone also significantly influenced the group's rate of cessation for adverse events. Sensitivity analyses based only on published studies and only on randomized trials yielded similar risk factors (data not shown) and model accuracy (r2 = 0.71–0.75).
The one adverse event rate significantly associated with lower group efficacy was a higher rate of grade 1 or greater nausea (r = −0.273; P = 0.028). The adverse event rates significantly associated with cessation for any reason were any grade nausea (r = 0.347; P = 0.005) and any grade fatigue (r = 0.413; P = 0.015). Cessation for ‘other reasons’ correlated significantly with a greater incidence of grade 1 or more fatigue (r = 0.415; P = 0.018), headache (r = 0.375; P = 0.009) and nausea (r = 0.325; P = 0.011).
By analysis of all data available through December 2007 and by inclusion of multiple, previously unstudied parameters, we have newly identified five factors (dosing relative to food, type of NRTI backbone, nonwhite race, exclusion of patients with low haemoglobin, shorter follow-up) that were independently associated with greater treatment success in studies of patients initiating ART.
The characteristics we identified explain 79% of the variability in study efficacy rate, as compared to 35% in a previous analysis that evaluated fewer characteristics . In contrast to the Antiretroviral Cohort Collaboration, we found greater efficacy with older age, lower baseline CD4 cell count, no independent association with baseline viral load, and no association with prior injection drug use or AIDS. This raises the possibility that factors apart from treatment success at 14 months of an initial regimen affect the rate of progression to AIDS or death at 5 years. Although it may seem self-evident that virological response may not necessarily equate to clinical response, licensing of antiretroviral drugs is based on virological response rates. As such, it is important to bear in mind that virological differences between regimens may well not yield similar clinical differences. Another possible reason for the difference observed is that the Antiretroviral Cohort Collaboration outcomes are based on analysis of individual patient data.
Initial therapy was more effective if participants with low haemoglobin levels were excluded. This is not surprising given that anaemia is a marker of more advanced HIV disease . What was surprising, however, was the finding that studies reporting higher efficacy tended to be those with lower average CD4 cell counts. The associations of lower average CD4 cell count, recruitment in Africa, and nonwhite race collectively suggest that patients in resource-limited settings, who often cannot access effective alternatives, may better adhere to initial ART than most white patients in resource-rich countries.
Didanosine with either lamivudine or emtricitabine may be an effective option as part of initial ART, and studies using these regimens appear more effective than those of the more frequently prescribed and recommended dual-NRTI combination of zidovudine–lamivudine. Didanosine is a nonpreferred option for initial ART in two guidelines [2,5], but is not mentioned by two other guidelines, which suggests that guideline development might be improved by incorporation of systematic data review.
For resource-limited settings, the World Health Organization guidelines recommend that didanosine only be prescribed as second-line therapy that also includes a ritonavir-boosted protease inhibitor, a recommendation that is not justified . Such combinations would necessitate dosing three-times daily both with and without food, a dosing schedule our analysis suggests is suboptimal. Didanosine's low cost and wide availability further support its use in resource-limited settings, although didanosine is not available as a co-formulated dual-NRTI and its nonenteric formulation requires dosing with food, which would make didanosine more complex for programmatic use in such settings.
Didanosine will not be appropriate for all patients. Didanosine can cause peripheral neuropathy, pancreatitis and steatohepatitis/lactic acidosis, which is more common in older women with more advanced HIV disease . However, didanosine rarely causes anaemia, renal impairment or hypersensitivity, and is an infrequent cause of nausea in its enteric formulation. A recent trial found it to be superior to zidovudine in treatment-naive adults, largely because of reduced gastrointestinal and haematological toxicity .
Adverse events were poorly reported in most trials. How well these were studied or analysed is unknown. The limited data suggest that treatment failure is not particularly driven by higher rates of grade 2 or greater adverse events. Only the rate of nausea and fatigue of any grade correlated significantly with the rate of treatment failure or with cessation for adverse events, which might be expected with a treatment that is permanent. Other unreported (and possibly unstudied) adverse events may have affected treatment outcomes or collection of adverse event data may not have been uniform. Unmeasured factors that could conceivably contribute to treatment cessation include adverse event duration and the availability or otherwise of alternative, less-toxic antiretroviral drugs. These findings imply that all grades of a broader range of adverse events should be studied and reported in future trials. As adverse event reporting was very incomplete, our findings regarding the relationships between adverse events and treatment outcomes should be interpreted with caution. Nevertheless, it is interesting that cessation for ‘other reasons’ correlated with several adverse events, suggesting that cessation for such reasons might reflect poor antiretroviral tolerability.
The available source material and studies have additional limitations that may have affected our findings. Fewer than 30% of participants were women (none was pregnant), only 10% had injection drug use as their major HIV risk factor (with extremely few active users) and only 12% were recruited in resource-limited settings. Adherence was rarely assessed and viral hepatitis serostatus infrequently reported. The mean follow-up to the primary endpoint was only 14 months, and so by and large does not take into account longer-term toxicities related to some types of ART such as lipodystrophy, osteoporosis, peripheral neuropathy, renal dysfunction and cardiovascular disease [18,19]. Also, some forms of ART may decrease the incidence of serious nonopportunistic conditions such as myocardial infarction, renal failure, cancer and cirrhosis [20,21]; these endpoints were also rarely reported.
Our analysis also has limitations. We analysed aggregate data rather than individual patient data in an ecological study design, with groups rather than individuals being the unit of study. The findings of this study should thus be interpreted cautiously as causality cannot be determined using this design. We analysed multiple parameters which raises the possibility of false-positive associations being identified. The additional parameters we identified, however, resulted in a substantial increase in our capacity to explain variability in treatment efficacy. We did not evaluate individual NNRTIs or boosted protease inhibitors, although individual trials comparing members within these two drug classes have not found major between-drug differences. The positive correlations between cessation of therapy for ‘other reasons’ and several adverse events may not be causal. Lastly, all abacavir studies were conducted prior to the routine use of HLA-B*5701 testing, which virtually eliminates abacavir hypersensitivity that otherwise occurs in about 5% of white adults . Abacavir efficacy may improve, therefore, by this amount in future studies. Nevertheless, preliminary data suggest that abacavir-lamivudine is less effective than tenofovir-emtricitabine in adults with plasma HIV viral load above 100 000 copies/ml plasma .
In summary, we have identified new parameters associated with greater rate of treatment efficacy that might allow for more effective design and comparison of initial antiretroviral studies. It seems likely that most initial antiretroviral studies are too short and report insufficient adverse event data and that the development of antiretroviral treatment guidelines could be improved by ongoing systematic review of all available data. Didanosine is a largely ignored but effective, cheap and widely available option for initial ART, with immediate potential relevance to resource-limited settings.
The authors would like to thank Rachel Cameron (Bristol-Myers Squibb), Roy Gulick, Michael Kraft (Boehringer-Ingelheim) and Daren Thorburn (GlaxoSmithKline) for providing unpublished eligibility criteria and baseline patient data, to Matthew Law for review of the manuscript, and to Alexandra Calmy for helpful discussions.
A.C. conceived the study and abstracted the data. J.A. analysed the data. Both authors wrote the manuscript.
This study received no specific funding. A.C. has received research funding from Abbott, Merck and Roche; consultancy fees from Bristol-Myers Squibb, Gilead Sciences, GlaxoSmithKline, Merck and Roche; lecture sponsorships from Abbott, Boehringer-Ingelheim, Bristol-Myers Squibb, Gilead Sciences, GlaxoSmithKline, Merck and Roche; and has served on advisory boards for Abbott, Bristol-Myers Squibb, Gilead Sciences, GlaxoSmithKline, Merck and Roche. J.A. declares no conflict of interest. The National Centre in HIV Epidemiology and Clinical Research is supported by the Commonwealth Department of Health and Aging. A.C. is a recipient of a Practitioner Fellowship from the Australian National Health and Medical Research Council.
3. Hammer SM, Saag MS, Schechter M, Montaner JSG, Schooley RT, Jacobsen DM, et al, International AIDS Society – USA panel. Treatment for adult HIV infection: recommendations of the International AIDS Society – USA Panel. JAMA 2006; 296:827–843.
4. Gazzard B, Bernard AJ, Boffito M, Churchill D, Edwards S, Fisher N, et al, Writing Committee, British HIV Association. British HIV Association (BHIVA) guidelines for the treatment of HIV-infected adults with antiretroviral therapy. HIV Med 2006; 7:487–503.
5. European AIDS Clinical Society. Guidelines for the clinical management and treatment of HIV-infected adults in Europe
. Accessed on 5 July 2008. http://www.eacs.eu/guide/index.htm
6. Yazdanpanah Y, Sissoko D, Egger M, Mouton Y, Zwahlen M, Chêne G. Clinical efficacy of antiretroviral combination therapy based on protease inhibitors or nonnucleoside analogue reverse transcriptase inhibitors: indirect comparison of controlled trials. BMJ 2004; 328:249–255, doi: 10.1136/bmj.37995.435787.A6.
7. Chou R, Fu R, Huffman LH, Korthuis PT. Initial highly-active antiretroviral therapy with a protease inhibitor versus a nonnucleoside reverse transcriptase inhibitor: discrepancies between direct and indirect meta-analyses. Lancet 2006; 368:1503–1515.
8. Bartlett JA, Fath MJ, DeMasi R, Hermes A, Quinn J, Mondou E, et al. An updated, systematic overview of triple combination therapy in antiretroviral-naïve, HIV-infected adults. AIDS 2006; 20:2051–2064.
9. May M, Sterne JA, Sabin C, Costagliola D, Justice AC, Thiébaut R, Gill J, et al, Antiretroviral Therapy Cohort Collaboration. Antiretroviral Therapy (ART) Cohort Collaboration. Prognosis of HIV-1-infected patients up to 5 years after initiation of HAART: collaborative analysis of prospective studies. AIDS 2007; 21:1185–1197.
10. Maggiolo F, Ripamonti D, Gregis G, Quinzan G, Callegaro A, Arici C, et al. Once-a-day therapy for HIV infection: a controlled, randomised study in antiretroviral-naïve HIV-1-infected patients. Antiviral Ther 2003; 8:339–346.
11. Maggiolo F, Migliorino M, Maserati R, Pan A, Rizzi M, Provettoni G, et al, Once Study Group. Virological and immunological responses to a once-a-day antiretroviral regimen with didanosine, lamivudine and efavirenz. Antiviral Ther 2002; 6:249–253.
12. U.S. Department of Health and Human Services Food and Drug Administration Center for Drug Evaluation and Research (CDER). Guidance for industry. Antiretroviral drugs using plasma HIV RNA measurements: clinical considerations for accelerated and traditional approval
. Accessed on 5 July 2008. http://www.fda.gov/cder/guidance/3647fnl.pdf
13. Building models stepwise. In: Practical statistics for medical research. Altman DG, Chapman and Hall, London; 1991. pp. 340–351.
14. Sterne JAC, Egger M, Davey Smith G. Investigating and dealing with publication and other biases in meta-analysis. BMJ 2001; 323:101–105.
15. Mocroft A, Ledergerber B, Zilmer K, Kirk O, Hirschel B, Viard JP, et al, EuroSIDA study group and the Swiss HIV Cohort Study. Short-term clinical disease progression in HIV-1-positive patients taking combination antiretroviral therapy: the EuroSIDA risk-score. AIDS 2007; 21:1867–1875.
16. Lactic Acidosis International Study Group. Risk factors for lactic acidosis and severe hyperlactataemia in HIV-1-infected adults exposed to antiretroviral therapy. AIDS 2007; 21:2455–2464.
17. Berenguer J, Gonzalez J, Ribera E, Domingo P, Miralles P, et al. Didanosine, lamivudine and efavirenz vs zidovudine and efavirenz, for initial treatment of HIV infection over 48 weeks: a randomised, noninferiority clinical trial, GESIDA. In: Abstracts of the 46th International Conference on Antimicrobial Agents and Chemotherapy, Chicago, USA, September, 2007, Abstract 383.
18. Carr A, Cooper DA. Adverse effects of antiretroviral therapy. Lancet 2000; 356:1423–1430.
19. Friis-Møller N, Sabin CA, Weber R, d'Arminio Monforte A, El-Sadr WM, Reiss P, et al, Data Collection on Adverse Events of Anti-HIV Drugs (DAD) Study Group. Combination antiretroviral therapy and the risk of myocardial infarction. N Engl J Med 2003; 349:1993–2003.
20. Strategies for Management of Anti-Retroviral Therapy Study Group. CD4+ count-guided interruption of antiretroviral treatment. N Engl J Med 2006; 355:2283–2296.
21. Baker J, Peng G, Rapkin J, Abrams D, Silverberg M, MacArthur RD, et al, Terry Beirn Community Programs for Clinical Research on AIDS (CPCRA). CD4+ count and risk of non-AIDS diseases following initial treatment for HIV infection. AIDS 2008; 22:841–848.
22. Mallal S, Phillips E, Carosi G, Molina J-M, Workman C, Tomazic J, et al, PREDICT-1 Study Team. HLA-B*5701 screening for hypersensitivity to abacavir. N Engl J Med 2008; 358:568–579.
23. Sax P, Tierney C, Collier A, Fischl M, Godfrey C, Jahed N, et al. ACTG 5202: shorter time to virologic failure (VF) with abacavir/lamivudine (ABC/3TC) than tenofovir/emtricitabine (TDF/FTC) as part of combination therapy in treatment-naïve subjects with screening HIV RNA ≥100,000 c/mL. In: Abstracts of the 17th International AIDS Conference, 3–8 August 2008, Mexico City, Mexico, Abstract THAB0303.
© 2009 Lippincott Williams & Wilkins, Inc.