Risk of extensive virological failure to the three original ART drug classes over long-term follow-up from the start of therapy in patients with HIV infection: an observational cohort study
The Lancet Dec 8, 2007; 370:1923-1928
Prof Andrew N Phillips PhD a , Clifford Leen MD b, Alan Wilson b, Jane Anderson FRCP c, David Dunn PhD d, Achim Schwenk FRCP e, Chloe Orkin MRCP f, Teresa Hill PhD a, Martin Fisher FRCP g, John Walsh MBBS h, Prof Deenan Pillay FRCPath i, Loveleen Bansi MSc a, Brian Gazzard FRCP j, Philippa Easterbrook FRCP k, Richard Gilson MD l, Prof Margaret Johnson FRCP m and Prof Caroline A Sabin PhD a, for the UK Collaborative HIV Cohort (CHIC) Study
".....Our study has shown that the rate at which patients starting antiretroviral therapy have extensive virological failure of the three original main antiretroviral drug classes seems to be slow, especially in those with CD4-cell counts greater than 200 cells per μL at the start of therapy. Furthermore, there is evidence that the rate has decreased over time as drugs and experience with their use have improved. Even in patients who have such extensive virological failure, the death rate remains fairly low (about 10% at 5 years, even in those starting treatment with a CD4-cell count less than 200 cells per μL). This finding suggests that drugs in these classes can sustain virological efficacy for a substantial length of time, which is especially encouraging for developing countries where drug availability will probably be restricted to the three original classes for some time....
.... A large proportion of patients with extensive triple-class failure had at least one subsequent viral load less than 50 copies per mL. Since few patients used the entry inhibitor enfuvirtide and none had used an integrase inhibitor5 or CCR5 antagonist,6 this result suggests that there remains some antiviral activity of drugs within the main three drug classes, even when extensive triple-class failure has arisen. The proportion achieving a viral load less than 50 copies per mL could well increase further in developed countries as new drugs and new classes are made available, especially in view of the promising results of trials in patients with heavy antiretroviral exposure....
.... The underlying reason for development of resistance and virological failure seems to be suboptimum adherence...
.... Estimates similar to ours of the rate at which extensive triple-class failure occurs are relevant for developing countries where such drugs are likely to be the only ones available for some time..."
The long-term durability of viral-load suppression provided by the three original antiretroviral drugs is not well characterised. We estimated the proportion of patients who had extensive triple-class failure during long-term follow-up and examined characteristics associated with an increased rate of failure.
7916 patients who started antiretroviral therapy with three or more drugs were followed up from the time that therapy started until the last viral-load measure. Extensive triple-class virological failure was defined by failure of three subclasses of nucleoside reverse transcriptase inhibitors, a non-nucleoside reverse transcriptase inhibitor, and a ritonavir-boosted protease inhibitor.
167 patients developed extensive triple-class failure during 27 441 person-years of follow-up. The Kaplan-Meier estimate for the cumulative risk of extensive triple-class failure was 9·2% by 10 years (95% CI 5·0-13·4). There was evidence that this rate has decreased over time (adjusted hazard ratio 0·86 [0·77-0·96] per year more recent; p=0·006).
Of the 167 patients with extensive triple-class failure, 101 (60%) subsequently had at least one viral load less than 50 copies per mL.
The risk of death by 5 years from the time of extensive triple-class failure was 10·6% (2·4-18·8, nine deaths).
We have shown that extensive virological failure of the three main classes of drugs occurs slowly in routine clinical practice. This finding has implications for the planning of treatment programmes in developing countries, where additional drugs outside these classes are unlikely to be available for some time.
When future needs for new drugs and health-care resources in developed and developing countries are being planned, the likely long-term durability of the clinical benefits of antiretroviral therapy in routine clinical settings is important to understand. Antiretroviral drugs exert their clinical benefits through suppression of HIV replication, so the main underlying factor determining the durability of the clinical benefit of antiretroviral therapy is the rate at which patients have virological failure (ie, failure to suppress viral load or loss of viral-load suppression) of the three original main classes of drugs (nucleoside reverse transcriptase inhibitors [NRTIs], non-nucleoside reverse transcriptase inhibitors [NNRTIs], and protease inhibitors [PIs]). Triple-class failure has generally been defined by virological failure of at least one drug in each of the three main classes.1-3 However, many patients who meet this definition will have other options for achieving virological suppression from within these three clases, especially if they have not had virological failure of a ritonavir-boosted PI, so more extensive failure of the nucleoside and PI classes should be considered.
Since there are three new classes of antiretrovirals in developed countries-fusion inhibitors, integrase inhibitors, and CC chemokine receptor 5 antagonists4-6-even patients with more extensive triple class failure will have further options for constructing a regimen to suppress viral load. Nonetheless, failure of the original three classes represents a milestone in a patient's treatment history and suggests a risk that drug options could become completely exhausted. In developing countries, access to drugs will probably be restricted to the three original classes for some time, so understanding the durability of the antiviral efficacy of these classes is crucial for long-term planning.
We aimed to estimate the proportion of patients who had extensive triple-class failure by up to 10 years and to examine characteristics associated with earlier failure.
The UK Collaborative HIV Cohort (UK CHIC) Study is an observational cohort of some of the largest HIV clinical centres in the UK. The dataset used for this analysis included information from ten centres. The creation and design of the UK CHIC Study has been previously described.7,8 The data consisted of information obtained as part of routine clinical care and included: patients' demographics, antiretroviral history, laboratory findings, AIDS-defining events, and deaths. All data provided by the participating centres were thoroughly checked and any inconsistencies resolved when possible. We included all patients starting antiretroviral therapy with three or more drugs and for whom a CD4-cell count and viral load were available before the start of antiretroviral treatment (or at least 1 year from start of therapy if patient started in 1996 or earlier).
Virological failure of a drug was defined as having occurred if a viral load above 400 copies per mL was measured in an individual, despite more than 4 months (122 days) of continuous use of the drug. We defined extensive failure of the nucleoside class as virological failure of at least one drug from each of the following subclasses: zidovudine, stavudine; lamivudine, emtricitabine; and didanosine, tenofovir, abacavir. Extensive failure of the NNRTI class was defined as virological failure of efavirenz or nevirapine, and extensive failure of the PI class as virological failure of at least one ritonavir-boosted PI. After virological failure had occurred, any new regimen including at least one drug that had not previously failed was defined as a new line of antiretroviral therapy. A line of antiretroviral therapy was defined as having failed when any drug within it had virologically failed.
Extensive triple-class failure was defined by extensive failure of all three classes. The risk of failure was compared with that by the more common, looser definition, which defines triple-class failure as virological failure of at least one drug in each of the three classes and, for the PI class, the drugs that failed do not need to include a ritonavir-boosted PI.
In sensitivity analyses, we also considered separately two variations in the standard definition of virological failure. The first was that the required time to have been on the drug was reduced from 122 to 91 days (ie, 4 to 3 months). The second was to consider the next viral-load value before a drug was deemed to have virologically failed. If this next viral load was less than 400 copies per mL (measured within the next 122 days) and the patient was on the same drug, the drug was not defined as having failed.
The Kaplan-Meier method was used to estimate the cumulative risk of extensive triple-class failure, and 95% CIs were calculated by Greenwood's formula. Cox models were used to assess factors that were associated with a more rapid rate of failure. Covariates considered for the multivariable models were sex, age, HIV-exposure group (homosexual men, heterosexual, other), calendar year when antiretroviral therapy started, CD4-cell count at start of treatment, viral load at start of treatment, AIDS before therapy commenced, and the initial regimen (two NRTI plus NNRTI, two NRTI plus PI (not boosted with ritonavir), two NRTI plus ritonavir-boosted PI, three NRTI, or other regimen type). Proportional hazards assumptions were investigated for every covariate by fitting the interaction with log of the survival time. All p values are two-sided. Analyses were undertaken with SAS (version 9.1).
We included a total of 7916 patients who started antiretroviral therapy with three or more drugs. Table 1 shows baseline characteristics of the patients. The median CD4-cell count at start of antiretroviral therapy was low because a high proportion of patients had presented late for care (as shown by the fact that almost a fifth of patients had a previous diagnosis of AIDS; table 1). Regimens containing an NNRTI were the most common, followed by those containing a PI, with use of a single PI slightly more common than use of a ritonavir-boosted PI (table 1). The proportion of patients being followed up who were on antiretroviral therapy at the start of the year was 91% (5715/6313) for year 1; in years 2-10, respectively, it was 89% (4438/4989), 89% (3433/3876), 89% (2608/2935), 89% (1878/2099), 89% (1257/1405), 91% (657/721), 89% (203/227), 100% (28/28), and 100% (six of six).
Table 1. Characteristics of patients who started antiretroviral therapy with three or more drugs. 61% on initial regimen of NNRTI, 11% PI/r, 15% PI, 4% triple nukes.
During 27 441 person years of follow-up, extensive triple-class failure developed in 167 patients. 129 (77%) of those who developed extensive triple-class failure had previously had a viral load of less than 50 copies per mL. At the time of extensive triple-class failure, 150 (90%) patients had had virological failure of seven drugs or more and 32 (19%) had had failure of nine drugs or more. 97 (58%) had had failure of two lines of antiretroviral therapy, 57 (34%) three lines, and 13 (8%) four or more.
The figure shows the Kaplan-Meier estimates of the cumulative risk of extensive triple-class failure after treatment was started. There was an estimated 3·5% (95% CI 2·9-4·1) risk of extensive triple-class failure by 5 years, and 9·2% (5·0-13·4) by 10 years from start of antiretroviral therapy. This finding compares with an 18% risk by 10 years when we applied the looser definition of triple-class failure. The 5-year cumulative risk of extensive triple-class failure was similar to the overall estimate of 3·8% when restricted to those who started antiretroviral therapy during or after 2000.
Table 2 shows Kaplan-Meier estimates of the risk of extensive failure of the classes separately, according to the number of years after a drug of that class was started. 8-year cumulative risk of extensive failure was greater for the NNRTI and PI classes than for the NRTI class. For patients starting their first ritonavir-boosted PI regimen, we assessed the association between number of regimens previously failed and rate of failure of that class. After adjustment for age, sex, risk group, calendar year, CD4-cell count, viral load, and presence of AIDS at start of antiretroviral treatment, patients who had previously had failure with one regimen had a 1·53-times (95% 1·27-1·85; p<0·0001) greater rate of virological failure than did those who had never previously had virological failure of any regimen before the start of the boosted PI. For those who had previously had failure of two or more regimens, the hazard ratio was 2·05 (1·50-2·79). The equivalent hazard ratio for patients starting an NNRTI regimen who had had failure of one regimen was 2·47 (2·00-3·06), and for those who had had failure of two regimens 3·11 (1·89-5·11).
Table 3 shows analysis of baseline predictors of extensive triple-class failure. In univariable analysis, being in the heterosexual risk group, low CD4-cell count at start of antiretroviral therapy, younger age, having started treatment a long time ago, and having had AIDS diagnosed before antiretroviral therapy started, were risk factors that were all significantly associated with a raised rate of extensive triple-class failure. In multivariable analysis, low CD4-cell count, younger age, being in the heterosexual risk group, and having started therapy a long time ago remained independently associated with a raised rate. Sex, baseline viral load, and initial regimen were not significantly associated with the rate of extensive triple-class failure, although there was some evidence to suggest that the lower hazard ratio for starting with a ritonavir-boosted regimen than with an NNRTI regimen increased with longer follow-up (p=0·004 test for interaction with log survival time). No other covariate showed a change in hazard ratio with increased follow-up (ie, no evidence to suggest that the proportional hazards assumption did not hold). When we considered only patients who started antiretroviral therapy with CD4-cell counts of 200 cells per μL or more, the cumulative risk of extensive triple-class failure by 10 years from start of therapy was 5·5% (95% CI 3·5-7·5), compared with 12·1% (5·1-19·1) for those with pretreatment CD4-cell counts less than 200 cells per μL.
The modified definition of virological failure, which needed a viral load of more then 400 copies per mL after only 91 days (3 months) of use of a drug rather than 122 days (4 months), gave an estimated cumulative risk of extensive triple-class failure by 10 years of 11·4% (7·0-15·8). If we instead applied the definition that considered also whether the next viral load was more than 400 copies per mL, the 10-year risk was 8·1% (4·1-12·1).
We also examined viral-load values after diagnosis of extensive triple-class failure. 101 (60%) patients had at least one viral load less than 50 copies per mL some time after diagnosis of extensive triple-class failure. 74 (44%) had at least two consecutive viral loads less than 50 copies per mL (40% cumulative probability of two consecutive viral loads <50 copies per mL by 1 year from extensive triple-class failure).
The fusion inhibitor, enfuvirtide, was the only drug other than the three original classes that was available in the period covered by our analysis. Of 7916 patients, 24 (<1%) used enfuvirtide at any point. Of the 167 patients with extensive triple-class failure, five (3%) used enfuvirtide.
239 of the 7916 patients died by 10 years from start of antiretroviral therapy (Kaplan-Meier estimate 6·2% [4·4-8·0]). The risk of death by 5 years from extensive triple-class failure was 10·6% (2·4-18·8; nine deaths). This latter figure was much the same when restricted to patients with CD4-cell counts less than 200 cells per μL at start of treatment.
Loss to follow-up, defined as the patient not being seen since mid-2004 despite not being known to have died, was 5·8% per year. To assess the likelihood that this loss resulted in bias in estimates of rates of triple-class virological failure, we assessed in a Cox model whether a time-updated variable indicating whether or not first-line failure had occurred predicted the risk of loss to follow-up. We noted a hazard ratio of 1·06 (p=0·33) of loss to follow-up during person time in which a previous regimen had virologically failed, compared with person time in which no such failure had occurred. This finding suggests that patients lost to follow-up were not more likely to have had a greater (or lower) risk of extensive triple-class failure.
Our study has shown that the rate at which patients starting antiretroviral therapy have extensive virological failure of the three original main antiretroviral drug classes seems to be slow, especially in those with CD4-cell counts greater than 200 cells per μL at the start of therapy. Furthermore, there is evidence that the rate has decreased over time as drugs and experience with their use have improved. Even in patients who have such extensive virological failure, the death rate remains fairly low (about 10% at 5 years, even in those starting treatment with a CD4-cell count less than 200 cells per μL). This finding suggests that drugs in these classes can sustain virological efficacy for a substantial length of time, which is especially encouraging for developing countries where drug availability will probably be restricted to the three original classes for some time.
A large proportion of patients with extensive triple-class failure had at least one subsequent viral load less than 50 copies per mL. Since few patients used the entry inhibitor enfuvirtide and none had used an integrase inhibitor5 or CCR5 antagonist,6 this result suggests that there remains some antiviral activity of drugs within the main three drug classes, even when extensive triple-class failure has arisen. The proportion achieving a viral load less than 50 copies per mL could well increase further in developed countries as new drugs and new classes are made available, especially in view of the promising results of trials in patients with heavy antiretroviral exposure.5,6,9,10
Over half the patients in our analysis started antiretroviral therapy with two NRTI plus an NNRTI, but we detected no difference in the rate of extensive triple-class failure according to the type of regimen started. We also investigated the risk of failure of the ritonavir-boosted PI and NNRTI classes from the time that the class was started. In both cases, there was the greatest risk of failure when there had been one or more previous virological failures. This finding is probably at least partly due to some resistance to the NRTI component having arisen with previous virological failure. Additionally, patients who have had previous virological failure of a regimen might on average have a greater predisposition to fail in future than might those who have not, perhaps since some patients have a greater tendency for suboptimal adherence than others.
Previous research into multiclass failure has tended to focus on triple-class failure defined by failure of at least one drug from all three main classes.1-3 Most studies have included patients who started antiretroviral therapy with regimens that included one or two NRTIs, which were used before 1996.1-3 Patients from these studies had a substantially higher rate of triple-class failure than did those who we investigated. However, for prediction of future trends, study of the group who started antiretroviral therapy with at least three drugs is most relevant, because patients are now treated in this way. We therefore focused our study on this group of patients. In EuroSIDA, Mocroft and colleagues2 noted an 11% cumulative risk of triple-class failure by 6 years from the start of antiretroviral therapy in patients commencing with three or more drugs; the prevalence in this group was 5% and increased over time. For those who started therapy with single or dual NRTI, the corresponding rates were 21% and 16%. Lohse and co-workers3 reported a 7% cumulative risk by 7 years in patients who had started therapy with three or more drugs and overall (including those who started antiretroviral therapy with fewer than three drugs) prevalence was 7% and stable. A previous contribution from our study7 noted that around 15% of those who had been exposed to three drug classes had had triple-class failure.
Closely linked to the study of multiclass virological failure is the study of multiclass resistance.11 Most patients who have had virological failure of a drug will also have some evidence from resistance testing that their virus has some degree of resistance to that drug, although not in all cases.11,12 The underlying reason for development of resistance and virological failure seems to be suboptimum adherence.13-15 Furthermore, virological failure can sometimes arise because the level of adherence is not sufficient for virus suppression, even if no resistance is present. Virological rebound or failure can also occur despite full adherence to antiretroviral therapy and despite the fact that no resistance mutations can be identified,12 perhaps because there are mutations in other parts of the virus or other as yet unknown mechanisms. In routine clinical practice, sometimes a resistance test is not done at the time of virological failure. The cumulative risk of resistance to three drug classes has been estimated from the UK HIV drug resistance database and CHIC cohorts as 4% at 6 years.16 Harrigan and colleagues17 assessed risk of resistance to all three classes, including resistance to lamivudine plus at least one other NRTI and reported a cumulative risk of less than 2% by 2 years. Napravnik and co-workers18 have reported fairly high rates of resistance accumulation. An advantage of focusing on virological failure of three classes rather than documented resistance to three classes is that ascertainment of extensive virological failure is probably more complete than is ascertainment of extensive resistance.
Estimates similar to ours of the rate at which extensive triple-class failure occurs are relevant for developing countries where such drugs are likely to be the only ones available for some time. WHO recommended a first-line regimen which consists of an NNRTI plus two NRTIs and a second-line regimen that consists of (different) NRTIs plus a boosted protease inhibitor.19 There are, however, differences in how patients receiving therapy are being monitored in developing countries, especially in sub-Saharan Africa. There, viral-load monitoring is not available and arguably not needed because first-line regimens generally have to be maintained for as long as there is clinical benefit.19 Therefore, regimens are commonly used for some time after our definition of virological failure has been met. This approach is appropriate since there are benefits both clinically and in CD4-cell counts of antiretroviral therapies beyond the time that virological failure has occurred,20 and because second-line regimens are more expensive than first-line regimens. Because there is no evidence for substantive differences in adherence between patients in sub-Saharan Africa and in developed countries,21 our estimates for the time for occurrence of extensive triple-class virological failure probably underestimate the time that such regimens can exert clinical benefit in both developing countries and other settings. This notion is supported by the low death rate that we observed in patients with extensive triple-class failure.
A limitation of our analysis is that there were fairly few patients who had developed extensive triple-class failure, so power was restricted, and the period of observation of patients starting therapy with three or more drugs was still short compared with the decades that patients might have to remain on treatment. Further follow-up of larger numbers of patients is needed. Furthermore, we noted an appreciable rate of loss to follow-up. However, there was no evidence that patients who have had virological failure of at least one regimen had a higher (or lower) chance of being lost to follow-up than those who had not failed, reducing the likelihood of bias because of informative censoring in our estimates of the rate of extensive triple-class virological failure. Additionally, we need studies of viral load, CD4-cell count, and clinical outcomes of much larger numbers of patients with extensive triple-class failure to understand the prognosis of those reaching this stage in their treatment course.