| |
Are previous treatment interruptions associated with higher viral rebound rates in patients with viral suppression?
|
| |
| |
AIDS:Volume 22(3)30 January 2008p 349-356
The UK Collaborative HIV Cohort (UK CHIC) Study
From the Royal Free and University College Medical School, London, UK.
"...In conclusion, we have identified that there may be long-term effects of treatment interruption that are present even if viral suppression is subsequently achieved. Among those with a suppressed viral load on HAART, those who have previously interrupted therapy while having a raised viral load may be at an increased risk of subsequent viral rebound and should be monitored regularly. All patients starting HAART should be counselled about the potential negative long-term effects of interrupting therapy (both in terms of the risk of clinical events and long-term viral failure), even if the short-term effects appear to be minimal."
Abstract
Objective: We investigated whether previous treatment interruptions are associated with a raised risk of viral rebound in individuals who have attained virological suppression.
Methods: All patients achieving an undetectable viral load while on therapy were followed until viral rebound or the time of the last viral load. Poisson regression was used to describe the independent impact of treatment interruptions on rebound rates.
Results: A total of 12 977 patients from the United Kingdom Collaborative HIV Cohort (UK CHIC) Study achieved a viral load of less than 50 copies/ml. These patients contributed a total of 37 314 person-years of follow-up. The overall rebound rate was 8.07 (7.78, 8.36) per 100 person-years. In adjusted analyses, rates of viral rebound were up to 64% higher (rate ratio 1.64; 1.43, 1.88) in those who had previously interrupted therapy compared with those who had not. Patients who had interrupted at detectable viral loads had up to a 74% (1.74; 1.42, 2.14) higher chance of rebounding compared with those who had not interrupted with a detectable viral load. We found no evidence to suggest interrupting treatment at an undetectable viral load was associated with viral rebound.
Conclusion: Among patients with an undetectable viral load, having previously interrupted therapy while the viral load was detectable is associated with a raised risk of rebound.
Introduction
Despite findings of an increased risk of clinical progression in patients who interrupt HAART [1], patients continue to consider the option of temporarily discontinuing treatment, largely with the aim of minimizing the unwanted effects or inconvenience of HAART.
Although it has been shown that patients who have interrupted HAART have a raised risk of drug resistance and increased immunosuppression [2-10], many patients are able to suppress their viral load upon restarting therapy [11-13]. Presently lacking in the literature, however, is information on the risk of subsequent viral rebound in these patients. If resistant virus has emerged after treatment interruptions this may have been archived in latently infected cells and might only emerge at some later timepoint when such cells become activated [14]. The pre-HAART use of nucleoside mono or dual therapy has been shown to be associated with a raised risk of rebound in individuals with viral suppression [15-20], presumably as a result of such a mechanism. In patients with viral suppression to below detectable levels, other factors known to be associated with an increased rate of viral rebound include the use of particular antiretroviral drugs [20,21] and poor adherence to medication [22]. An increased duration of virological suppression has also been associated with lower rates of viral rebound [16,19], as has a lower number of previous regimens failed [23].
The aim of this analysis was to determine whether previous episodes of treatment interruptions are associated with a raised risk of viral rebound in individuals who have attained virological suppression (≦ 50 HIV-RNA copies/ml), after accounting for the factors mentioned above.
Discussion
Although previous studies have considered the short to medium-term effects of interrupting therapy [1-13], this is the first study to look at the long-term effects of treatment interruptions among those who subsequently achieve an undetectable viral load. In our study, a substantial minority of patients with an undetectable viral load whilst on HAART had previously interrupted therapy. Patients who had previously interrupted while their viral load was detectable (> 400 copies/ml) had up to a 74% higher chance of rebounding, even after controlling for the duration of viral suppression. In contrast, in main analyses we found no consistent evidence to suggest that interrupting therapy at a viral load of 400 copies/ml or less was associated with a raised risk of viral rebound. In subgroup analyses of those currently on a PI-based regimen, patients who interrupted therapy at an undetectable viral load did appear to have a small increased risk of viral rebound, although this may have been a chance finding.
We know of no previous studies that have addressed the question we studied. Other studies have, however, focused on rates of viral suppression after interrupting therapy [6,9]. Findings from those studies have demonstrated that the interruption of treatment at an undetectable viral load is generally associated with a lower rate of suppression compared with continuous therapy. It should be noted that our study focuses solely on those patients with a suppressed viral load while on therapy. In particular, patients who have previously interrupted therapy must have re-attained viral suppression on restarting therapy to be included in these analyses. We may thus have included a selected group of patients with better adherence who may be expected to have better long-term outcomes.
Our results show that the rebound rate increases as the number of previous treatment interruptions with detectable viral load increases. Our findings may be a consequence of the higher risk of resistance evolution in patients who had previously interrupted HAART [5]. Resistance mutations may have been archived during the period when drug levels were low immediately after interruption, as well as in the period before interruption when the viral load was detectable. Yerly et al. [8] reported that the M184V/I mutation is frequently selected during repeated treatment interruptions, consistent with results from Schweighardt et al., [10] who concluded that repeated treatment interruptions of certain antiretroviral drug regimens may lead to the development of drug resistance. It is also possible that a previous detectable viral load while on therapy, with the consequent evolution of resistance, is the main reason for the association, rather than the interruption itself. These hypotheses would be consistent with our findings of a higher rate of viral rebound in those who had previously interrupted an NNRTI with a detectable viral load and then restarted the same class of drug, which appeared to be larger that the corresponding level of increased risk for those who had interrupted a PI among those on a PI, and which may be a consequence of the longer half-life of the NNRTI class.
We would expect to see a lower rate of archived resistance in patients who interrupt their therapy when their viral load levels are undetectable as a result of the reduced opportunity for the replication of virus in the presence of suboptimal drug levels. Our findings of a lack of association between the interruption of therapy at undetectable viral loads and subsequent viral rebound are consistent with this.
Our findings might also be explained by confounding by adherence, in that patients who have previously interrupted HAART may be those who are generally less adherent to HAART and thus are more likely to experience viral rebound, either because of the emergence of resistance or the presence of suboptimal drug levels. If so, then the explanation for the lack of a relationship between treatment interruptions at undetectable viral loads and viral rebound may be that those patients who interrupted treatment at undetectable viral loads and then managed to resuppress their viral load on restarting treatment may generally have better adherence levels than those who interrupt at a detectable viral load and may therefore be more likely to maintain viral suppression.
Our analyses focused on the relationship between previous treatment interruptions and viral rebound rates. All other factors found to be independent predictors of viral rebound, in particular, the number of regimens previously failed and the duration of viral suppression, are consistent with findings presented in an earlier paper from this group [23]. The independent factors associated with viral rebound are also consistent with findings in other studies, in particular, current drug regimen [21], lower baseline CD4 cell counts [25], the use of monotherapy [16,17,19,21,26-28], younger age [29] and black ethnicity [16,21].
Multivariable regression is the standard approach to determining the independent effects of covariates when there are many potential confounders [30]. It should, however, be borne in mind that issues such as unmeasured confounding, measurement error in covariates, undetected model misspecification and indeed sampling error can lead to misleading results [31].
It is possible that the risk of viral rebound was higher among individuals with treatment interruptions as they had failed more regimens at baseline. Our analyses had, however, adjusted for this factor and further analyses were performed in which we stratified by the number of regimens failed as opposed to adjusting for this factor. Conclusions were very similar to that of the primary analyses.
Our study does have the limitations that are inherent in an observational cohort. In particular, we cannot control for unmeasured confounding such as adherence, which may play a key part in both the decision to interrupt treatment as well as the risk of subsequent viral rebound. Furthermore, our findings of a non-significant relationship between the number of treatment interruptions taken while the viral load was undetectable and viral rebound, should be interpreted cautiously, given the relatively small number of such interruptions.
In conclusion, we have identified that there may be long-term effects of treatment interruption that are present even if viral suppression is subsequently achieved. Among those with a suppressed viral load on HAART, those who have previously interrupted therapy while having a raised viral load may be at an increased risk of subsequent viral rebound and should be monitored regularly. All patients starting HAART should be counselled about the potential negative long-term effects of interrupting therapy (both in terms of the risk of clinical events and long-term viral failure), even if the short-term effects appear to be minimal.
Results
A total of 25 260 patients were included in the UK CHIC dataset at the time of analysis. Of these, 15 920 (63.0%) had received HAART at some timepoint over follow-up. A total of 12 977 patients (81.5%) attained a viral load of 50 copies/ml or less at least once whilst on HAART; patients contributed a total of 15 409 unique viral suppression episodes to the analysis (median one episode, range one to five per person). Of the 15 409 viral suppression episodes, 3348 (21.7%) were in individuals who had interrupted therapy before the start of the episode. The characteristics of the patients included in the analysis at the start of each viral suppression episode, stratified according to whether or not the patient had previously interrupted therapy, are shown in Table 1. In general, patients who had interrupted therapy were more likely to be of white ethnicity (63.3 versus 59.2%, P < 0.001), to have attained viral suppression in later calendar years (56.1 versus 42.5% from 2002 onwards, P < 0.001) and to have failed at least one previous regimen before viral suppression (71.5 versus 35.1%) compared with those patients who had not previously interrupted therapy.
Of the 4957 presuppression-episode treatment interruptions that occurred (Table 2), 46.3% were interruptions of PI-based regimens and 34.7% of NNRTI-based regimens. Of the 1719 NNRTI interruptions, 40% ended with the patient restarting the same NNRTI, compared with 26% of PI interruptions, which ended with the patient restarting the same PI. The median duration of each treatment interruption was 4.4 (interquartile range 1.9, 10.1) months and treatment interruptions were initiated at a median CD4 cell count of 256 (139, 410) cells/mm3. For a quarter (25.6%) of all interruptions the most recently measured viral load at the time of interruption was undetectable.
Patients were followed for a total of 37 314 person-years. In line with the characteristics of patients in the cohort, most person time was contributed by men [30 053 (80.5%) years], those of white ethnicity [23 906 (64.1%) years] and those from a homo/bisexual risk group [24 352 (65.3%) years]. In total, 3012 viral rebound events occurred over follow-up, giving an overall rebound rate of 8.07 [95% confidence interval (CI) 7.78, 8.36] per 100 person-years. The rate of viral rebound increased with the number of previous treatment interruptions (Fig. 1), after adjusting for potential confounders.
In an unadjusted Poisson regression model, patients who had interrupted treatment only once before viral suppression had a 63% increase (RR 1.63; 95% CI 1.49, 1.79) in the rate of viral rebound compared with those patients who had no previous treatment interruptions. The rate was more than twice as high (2.38; 2.10, 2.70) among patients who had interrupted therapy on two or more occasions. After adjusting for known potential confounding factors, an increased number of previous treatment interruptions continued to be significantly associated with viral rebound: rates of viral rebound were increased by 28% (1.28; 1.16, 1.88) and 64% (1.64; 1.43, 1.88) in patients who had one and two or more previous interruptions, respectively, compared with those who had not previously interrupted therapy (Table 3, model 1). In line with previous analyses of this cohort, patients had a 31% (1.31; 1.25, 1.38) higher rate of rebound per additional regimen failed and rebound rates decreased by 20% (0.80; 0.77, 0.84) for every one-year increase in the duration of viral suppression. Other independent factors of viral rebound included earlier calendar year, lower CD4 cell count, previous use of mono/dual therapy, younger age and black-African ethnicity. The current use of zidovudine/lamivudine as the NRTI backbone and the use of efavirenz as the 'third' drug in the regimen were associated with a lower rate of viral rebound.
In further analyses, we counted treatment interruptions made when the most recent viral load was detectable separately from the number made when the most recent viral load was undetectable. An increased number of treatment interruptions taken while the viral load was detectable was associated with a higher rate of viral rebound (1.74; 1.42, 2.14 for those who had previously interrupted more than two times at detectable viral loads compared with those patients with no previous interruptions at detectable viral loads), whereas the number of treatment interruptions taken while the viral load was undetectable was not significantly associated with subsequent rebound (Table 3, model 2). A greater number of previous treatment interruptions appeared to be significantly associated with a higher rate of viral rebound, regardless of the regimen interrupted (Table 3, model 3).
The results of multivariable analyses that accounted for both the viral load at the time of treatment interruption and the type of regimen interrupted confirmed that the viral load at the time of interruption was a more important determinant of subsequent viral rebound than the type of regimen interrupted (data not shown). It was, however, felt that these analyses would be confounded by the regimen that patients subsequently restarted. Sub-group analyses were thus performed separately among those who were receiving a NNRTI-based regimen at the time of viral suppression and those who were receiving a PI-based regimen and focused on previous treatment interruptions taken while the viral load was detectable (Table 4). Among those receiving a NNRTI-based regimen, those who had previously interrupted at least one NNRTI regimen at a detectable viral load had a near twofold increased rate (adjusted RR 1.93; 1.46, 2.58) of viral rebound compared with those patients who had never interrupted therapy. The rate of rebound among patients who had interrupted only non-NNRTI-based regimens at detectable viral load levels was only 44% higher (1.44; 1.14, 1.82) compared with that in patients who had never previously interrupted therapy. In contrast, among patients who were currently on a PI-based regimen, those who had previously interrupted a PI-based regimen with a detectable viral load had only a 44% increased rate (1.44; 1.21, 1.72) of viral rebound compared with those patients who had never interrupted therapy. Patients in this group who had interrupted only non-PI-based regimens at a detectable viral load had a rate of viral rebound that was 32% (1.32; 1.06, 1.63) higher than that among patients who had never interrupted therapy.
Methods
The source of data for our analyses was the United Kingdom Collaborative HIV Cohort (UK CHIC) Study. This is an observational cohort of HIV-infected individuals attending some of the largest HIV clinical centres in the United Kingdom. The dataset used for the current analysis includes information from 10 centres (Chelsea and Westminster, St Mary's NHS Trust, King's College Hospital, the Mortimer Market Centre, the Royal Free, St Bartholomew's and The Royal London Hospital, Brighton and Sussex University Hospital, Homerton Hospital NHS Trust, Edinburgh Hospital and North Middlesex University Hospital). The creation and design of the UK CHIC Study has been described previously [24]. Data collected include information on patient demographics, antiretroviral history, laboratory findings, AIDS-defining events and deaths. All data provided by the participating centres are thoroughly checked and any inconsistencies resolved.
These analyses follow on from previous analyses from the UK CHIC Study on predictors of viral rebound in individuals with viral suppression to 50 copies/ml or less [21,23]. All patients who achieved a viral load of 50 copies/ml or less while receiving HAART (defined as any treatment combination including three or more drugs) were eligible to enter the analyses. Person-time was calculated from the first time a patient's viral load fell to 50 copies/ml or less until viral rebound (defined as two consecutive values ≥ 400 copies/ml or one value of ≥ 400 copies/ml followed by the initiation of two or more new drugs). Patient follow-up was censored (i.e. the viral suppression episode ended and patients were removed from the risk set for the analysis) before this point if the patient discontinued or reduced HAART to less than three drugs, or on the date of the patient's last viral load if this was 400 copies/ml or less or was their first value greater than 400 copies/ml (i.e. without meeting the definition of viral rebound). Patients were eligible to re-enter the risk set and contribute a new suppression episode after viral rebound/censoring as a result of the discontinuation of HAART if they again managed to suppress their viral load to 50 copies/ml or less. Each patient could thus contribute multiple viral suppression episodes to the analysis. Rates of viral rebound were calculated by dividing the total number of events (viral rebounds) by the total person time.
A treatment interruption was defined as discontinuation of all therapy for at least 2 weeks after having started HAART and before the date of viral suppression for each particular period of observation.
Factors associated with viral rebound were identified using Poisson regression (using the GENMOD procedure in SAS). In addition to the total number of previous interruptions (initially classified as none, one, two, three, four and five or more), the following potential predictors were also considered: time with viral load suppression (per one year increase); number of previous regimens failed (per one regimen increase); current antiretroviral regimen (backbone nucleoside reverse transcriptase inhibitors (NRTI) and 'third' drug); calendar year; age; ethnicity (white, black African or other); sex, risk group (homo/bisexual, heterosexual, other); viral load at initiation of antiretroviral treatment; time since initiation of antiretroviral treatment and CD4 cell count at the start of the period of viral suppression. An interaction variable between the number of previous regimens failed and the duration of suppression was also included in the model, as suggested by our previous analyses [23]. In further analyses, the number of previous treatment interruptions was categorized as either none, one or two or more. We also took into account viral load at the time of interruption (i.e. the number of interruptions that had previously occurred while the viral load was detectable (> 400 copies/ml), undetectable or not known) and the type of regimen that was interrupted [protease inhibitor (PI)/non-nucleoside reverse transcriptase inhibitor (NNRTI)/NRTI only]. Furthermore, we also considered whether the impact of treatment interruptions differed in those who interrupted and then re-started a regimen containing the same or different class of drug.
As patients were eligible to enter the analyses on more than one occasion, rate ratios (RR) were also estimated using generalized estimating equations to fit univariable and multivariable Poisson regression models, allowing for multiple events in the same individuals. This present analysis is based on patients who received HIV care from January 1996 to December 2004.
|
|
| |
| |
|
|
|
