|
Virological Response to Initial Antiretroviral Regimens Containing Abacavir or Tenofovir
- New UK CHIC Study BRIEF REPORT
|
|
|
The Journal of Infectious Diseases Sept 1 2009;200:710-714
Loveleen Bansi,1 Caroline Sabin,1 Richard Gilson,1 Brian Gazzard,2 Clifford Leen,11 Jane Anderson,3 David Dunn,4 Teresa Hill,1 Martin Fisher,12 Jonathan Ainsworth,5 Deenan Pillay,6 Margaret Johnson,7 John Walsh,8 Chloe Orkin,9 Philippa Easterbrook,10 Mark Gompels,13 and Andrew Phillips,1 on behalf of the UK Collaborative HIV Cohort Study
1University College Medical School, Royal Free Campus, 2Chelsea and Westminster National Health Service (NHS) Trust, 3Homerton University Hospital NHS Trust, 4Medical Research Council Clinical Trials Unit, 5North Middlesex University Hospital NHS Trust, 6Department of Virology, University College London, 7Royal Free NHS Trust, 8Imperial College Healthcare NHS Trust, 9Barts and The London NHS Trust, and 10King's College Hospital, London, 11Lothian University Hospitals NHS Trust, Edinburgh, 12Brighton and Sussex University Hospitals NHS Trust, Brighton, and 13North Bristol NHS Trust, Bristol, United Kingdom
Patients with a baseline viral load >100,000 copies/mL receiving abacavir (ABC) as part of the nucleoside-backbone component of their first highly active antiretroviral therapy (HAART) regimen have been reported to have a greater failure rate than those receiving tenofovir (TDF). We analyzed short-term outcomes of the use of HAART combinations that included ABC or TDF. The mean 2-8-week change in viral load was calculated using linear regression. In total, 1136 patients started ABC, and 412 started TDF. After adjustment for baseline viral load and other factors, there was no difference in the change in viral load between the patients who started ABC and those who started TDF (0.03 [95% confidence interval, -0.07 to 0.12]) log copies/mL; p=.59). Furthermore, there was no evidence that this effect differed according to baseline viral load (p=.88 for the interaction between pre-HAART viral load and nucleoside started). Likewise, there was no difference in rates of virological failure between the 2 drugs at 24-48 weeks after starting HAART.
In a recent clinical trial (AIDS Clinical Trials Group [ACTG] 5202) comparing highly active antiretroviral therapy (HAART) regimens that included abacavir (ABC) and lamivudine (3TC) as the nucleoside reverse-transcriptase inhibitor backbone with those that included tenofovir (TDF) and emtricitabine (FTC) in antiretroviral-naive patients, interim analyses revealed a significant difference between the 2 arms in relation to the time to virological failure (viral load >1000 copies/mL at or after 16 weeks and before 24 weeks or else HIV RNA level >200 copies/mL at or after 24 weeks). Patients with a baseline viral load >100,000 copies/mL receiving ABC-3TC had a greater failure rate than did those receiving TDF-FTC. As a result, the data and safety monitoring board for the trial recommended that patients in the high-viral-load arm should be unblinded to their nucleoside backbone [1], and the study team decided on management of patients randomized to receive ABC-TDF. The results for those with baseline viral loads <100,000 copies/mL have not yet been disclosed, and the trial remains blinded for this portion of the study. While the current treatment guidelines of the British HIV Association and the International AIDS Society continue to recommend that ABC be a part of initial therapy [2, 3], recent US Department of Health and Human Services guidelines no longer include ABC as part of a recommended first-line regimen [4], with the results from ACTG 5202 cited as part of the rationale. The other factor mentioned was reports of an increased risk of myocardial infarction associated with ABC [5].
Given the implications of the results of ACTG 5202, it is important to examine evidence from other sources to determine the consistency and magnitude of any differences in virological efficacy between nucleoside-backbone drugs. Here, we present an analysis of the short-term outcomes of the use of HAART combinations that included ABC or TDF in a large clinical cohort, with particular focus on the relative effect of the 2 combinations at different pretreatment viral loads. We also present an analysis of the medium-term outcomes of these regimens, focusing on virological failure at 24 weeks after the initiation of HAART.
Methods.
Data from the UK Collaborative HIV Cohort (UK CHIC) Study were used for these analyses [6] (see the Appendix). All patients who started HAART on or after 1 January 2002 with ABC or TDF (but not both) plus 3TC or FTC and either a single nonnucleoside reverse-transcriptase inhibitor (NNRTI) or a protease inhibitor (PI) (boosted or otherwise) were eligible for the analyses. Multivariate logistic regression was used to determine factors independently associated with starting ABC (as opposed to TDF), and linear regression was used to identify factors associated with the mean 2-8-week change in viral load after the initiation of HAART. In these analyses, patients were required to have a viral load measurement before starting HAART and a subsequent measurement in a 2-8-week window after starting HAART. If >1 viral load measurement was recorded in the 2-8-week window, the first viral load measured was used in the analyses. The baseline viral load measurements were categorized into the following 4 divisions, chosen a priori: <30,000, 30,001-100,000, 100,001-300,000, and >300,000 copies/mL.
Pre-HAART viral load was the main explanatory variable considered, with adjustment for sex, ethnicity, risk group, year of starting HAART, time to viral load measurement, and type of third drug used in the initial HAART regimen (NNRTI or PI). Sensitivity analyses were performed in which we took into account left-censoring, which occurs because the mean change in viral load cannot be measured exactly for those who had viral loads below the limit of detection [7]. Patients who switched therapy before their 2-8-week viral load measurement were identified, but such patients were not excluded from these analyses. A switch was defined as either stopping an existing nucleoside or adding a nucleoside to the regimen, with the exception of switches between FTC and 3TC. To assess whether the relative effects of ABC and TDF differed for those with high viral loads, we formally incorporated an interaction term between the pre-HAART viral load (categorized as <100,000 or >100,000 copies/mL) and treatment arm (ABC or TDF).
All patients with a pre-HAART viral load and a viral load measured between 24 and 48 weeks after HAART initiation were included in the medium-term outcome analysis: virological failure at 24-48 weeks after starting HAART. For this analysis, patients were defined as having experienced virological failure if the first viral load measured in this window was >50 copies/mL. Treatment switches were ignored. Logistic regression was used to identify risk factors associated with virological failure, with adjustment for the above-mentioned variables.
Results.
In total, 1548 patients started an eligible regimen containing either ABC or TDF on or after 1 January 2002 and had at least 1 viral load measurement during the 2-8-week period after starting HAART (measured at a median of 4.0 [interquartile range, 2.9-4.9] weeks after HAART initiation). Of these patients, 1136 (73.4%) started HAART with a TDF-containing regimen, and 412 (26.6%) started HAART with an ABC-containing regimen. Of those patients who started HAART with TDF, the number who had a pre-HAART viral load <30,000, 30,001-100,000, 100,001-300,000, and >300,000 copies/mL were 265 (23.3%), 289 (25.4%), 289 (25.4%), and 293 (25.8%), respectively. Of those who started HAART with ABC, the corresponding number of patients in each of these viral load categories were 129 (31.3%), 108 (26.2%), 101 (24.5%), and 74 (18.0%), respectively. The median pre-HAART viral load was 4.9 (95% confidence interval [CI], 4.3-5.3) log copies/mL among those starting ABC and was 5.0 (95% CI, 4.5-5.5) log copies/mL among those starting TDF. Compared with patients starting ABC, those starting TDF were more likely to be of white ethnicity (59.4% vs 45.6%), to be men who have sex with men (60.7% vs 42.7%), and to use a PI as the third drug in their initial HAART regimen (30.6% vs 20.9%). Patients who started HAART with ABC were more likely to start their treatment during or after 2006 (62.4% of those who started HAART with ABC and 26.0% of those who started HAART with TDF did so in 2006 or 2007). In multivariate logistic regression models, all of the factors indicated above were found to be significantly associated with initial HAART regimen.
Overall, patients starting ABC experienced a mean reduction in viral load of 2.08 (95% CI, 2.16-1.99) log copies/mL over the first 2-8 weeks of HAART, compared with a mean reduction of 2.14 (95% CI, 2.19-2.09) log copies/mL among those starting TDF. The difference was not statistically significant (mean difference, 0.07 [95% CI, -0.03 to 0.17] log copies/mL). A strong linear trend was seen between the observed change in viral load over the first 2-8 weeks of HAART and the pre-HAART viral load in both treatment arms.
Pre-HAART viral load remained a highly significant predictor of the mean 2-8-week change in viral load (table 1) after adjustment for other covariates. In particular, the mean reduction in viral load at 2-8 weeks was 0.28 log copies/mL greater among those with pre-HAART viral loads >300,000 copies/mL than among those with pre-HAART viral loads of 30,001-100,000 copies/mL (p<.001). The nucleoside started (ABC or TDF) was not significantly associated with the mean 2-8-week change in viral load (the mean reduction in viral load was 0.03 [95% CI, -0.07 to 0.12] log copies/mL lower among patients starting ABC than among those starting TDF; p=.59) (figure 1). The result of a test for interaction between pre-HAART viral load and nucleoside started (ABC vs TDF) was nonsignificant (p=.88).
The week 2-8 viral load was <50 copies/mL in 222 (14.3%) of the 1548 patients; sensitivity analyses that accounted for the left-censoring of the viral loads reached conclusions similar to those of the main analyses. Of the 1548 patients included in the above analyses, 20 (1.3%) made a switch in treatment before their 2-8-week viral load measurement; these switches occurred among 11 patients (1.0%) who had started TDF and among 9 patients (2.2%) who had started ABC.
A total of 1492 patients had a viral load measured between 24 and 48 weeks after HAART initiation. Among these patients, 60 (17.9%) of those starting ABC and 213 (18.4%) of those starting TDF experienced virological failure. No association was found between initial HAART regimen and virological failure in univariate analyses (odds ratio for those starting ABC- vs TDF-containing regimens, 0.96 [95% CI, 0.70-1.32]; p=.81) or in analyses adjusting for potential confounders (odds ratio, 1.02 [95% CI, 0.72-1.45]; p=.91). The result of a test for interaction between baseline viral load and initial HAART regimen was also nonsignificant (p=94).
Discussion.
In these analyses, which are based on a large observational cohort, we found that baseline viral load at HAART initiation was strongly associated with the change in viral load 2-8 weeks after starting HAART; however, no association was found between nucleoside backbone (ie, ABC vs TDF) and the change in viral load. Furthermore, no interaction was detected between nucleoside backbone and baseline viral load in relation to the 2-8-week change in viral load. Analysis of outcomes at weeks 24-48 also revealed no association between the nucleoside backbone and virological failure.
In our study, we adopted inclusion criteria similar to those in ACTG 5202 [1]. Patients were >16 years old and ART naive before initiating HAART. Although we did not restrict our analyses to those who were receiving either efavirenz or atazanavir-ritonavir as the third drug in their initial HAART combination, we did select only those patients who were using exactly 2 nucleosides (ABC or TDF with either FTC or 3TC) and either 1 NNRTI or a PI (boosted or otherwise). However, our results did not appear to be consistent with the interim analyses from this trial. In particular, our finding that the interaction between pre-HAART viral load and nucleoside backbone was not significant suggested that there was no evidence that patients with high pre-HAART viral loads taking ABC experienced a poorer virological response than did those with similarly high pre-HAART viral loads taking TDF. Statistical tests of interaction are known to have low power, and it is possible that we may have failed to detect an interaction despite the large size of our study. However, our estimates of the mean viral load changes among those with high viral loads did not suggest a difference in outcomes in this group. Furthermore, our analyses of virological failure at 24-48 weeks after HAART initiation also showed a nonsignificant interaction between initial HAART regimen and baseline viral load. If ABC has meaningfully lower antiviral potency than TDF, then we would expect to have seen a difference in outcomes between the drug regimens. A possible explanation for the differences seen between our analyses and the ACTG trial is that patients in the trial were randomized and blinded to their initial HAART regimen. Our analyses are from a cohort study, and, hence, we cannot rule out any physician or patient biases in choosing initial HAART regimens.
A likely explanation for the association seen between pre-HAART viral load and the mean 2-8-week change in viral load is that the full extent of the reduction in viral load cannot be measured among those with low pre-HAART viral loads. For example, patients with pre-HAART viral loads 30,000 copies/mL are more likely to have achieved undetectable viral loads and, hence, have a higher percentage of censored observations than those with higher pre-HAART viral loads. We did perform sensitivity analyses in which we adjusted for this censoring using Kaplan-Meier methods [7]. The results from these sensitivity analyses were similar to those of our main analyses, but these methods will not fully account for this issue when censoring is heavy. Regression to the mean could also play a role.
Our results are from an observational cohort study, so patients initiating ABC or TDF are not balanced with respect to possible confounding factors. Although we were able to control for measured potential confounders, there could be others we have not measured that remain unadjusted for. In particular, we do not yet collect data on HLA B5701 testing, which is likely to influence the initial choice of HAART regimen. Our analyses focused on short-term changes in viral load in order to minimize the impact of treatment switches. Although it has been shown that short-term changes are associated with longer-term outcomes [8], a direct correlation between the 2 may not always exist. In conclusion, our findings indicate that patients with high pre-HAART viral loads who initiate HAART with ABC have short-term viral load reductions comparable to those who initiate HAART with TDF, and this should be taken into consideration when deciding which nucleosides should be used as part of a patient's initial HAART regimen.
![table1.gif](../images/081309/table1.gif)
Potential conflicts of interest: none reported.
Financial support: UK Medical Research Council (grants G0000199 and G0600337).
The views expressed in this article are those of the researchers and are not necessarily those of the UK Medical Research Council.
|
|
|
|
|
|
|