In settings in which they can be afforded, the use of antiretroviral drugs in combinations of three or more has had a major impact on the incidence of AIDS diseases and mortality [1–6]. These drugs exert their influence on the risk of these outcomes by reducing HIV-1 replication, and thus allow the regeneration or maintenance of CD4 lymphocyte-mediated immune responses against opportunistic pathogens [6–11]. The achievement and maintenance of prolonged viral load suppression is therefore the key to the long-term efficacy of therapy . Previous studies have suggested that the aim of initial therapy should be to achieve a plasma viral load of less than 50 copies/ml by approximately 16–24 weeks, as this is likely to lead to more prolonged suppression than if viral suppression is less profound [12–16], and this is reflected in current treatment guidelines . There is relatively little information, however, on the long-term durability of viral suppression [18–20], particularly in individuals who have achieved a viral load of less than 50 copies/ml within 24 weeks. Nor is there a clear understanding of the reasons for the loss of suppression (i.e. viral rebound), particularly in unselected clinic populations, as opposed to participants in clinical trials [21,22]. An important distinction, often not considered, is whether rebound represents an intrinsic failure of therapy; i.e. occurs while therapy is believed to be being taken (viral breakthrough), or is caused by a complete interruption of therapy because of its toxic effects, the inconvenience of taking the drugs or some intercurrent condition.
In order to address these issues, we have identified all individuals within the Frankfurt HIV Clinic Cohort who have achieved a viral load of less than 50 copies/ml within 24 weeks of the initiation of antiretroviral therapy. These patients were followed to assess the incidence of viral rebound over the ensuing 3.3 years and to assess the extent to which such a rebound could be explained simply by the interruption of therapy.
Patients and methods
We followed intensely, with 4–8 weekly viral load measurements, all 336 antiretroviral-naive patients attending the Goethe Universitat Clinic, with viral loads greater than 500 copies/ml starting antiretroviral therapy. These patients began treatment with a regimen containing three or more protease inhibitors (PI) (except when saquinavir hard gel was the only PI, as a result of the reduced potency of this drug and well documented poorer virological responses [23–25]), non-nucleoside reverse transcriptase inhibitors, or abacavir, including at least two (other) nucleoside analogue reverse transcriptase inhibitors. They all had viral loads of less than 50 copies/ml, measured using the Roche Ultra-sensitive assay (lower limit of quantification 50 copies/ml) within 24 weeks of therapy initiation. This assay was used routinely after January 1998. In order to include more patients and to achieve a greater length of follow-up, we retrospectively measured the viral loads, when stored plasma was available, of patients for whom viral loads greater than 500 copies/ml had been achieved by 16 weeks. This resulted in the identification of 94 patients who achieved viral loads less than 50 copies/ml before January 1998. The subsequent outcome was not used to select the stored plasma samples to measure with the Ultra-sensitive assay, so this approach should not have biased the results. This is a closely monitored cohort of patients, in whom adherence is thought to be generally good, but all patients identified on the basis of the above criterion were included in the follow-up, regardless of their perceived level of adherence.
Viral rebound was defined as either: (i) two consecutive measurements greater than 500 copies/ml (the date of rebound was the date of the first of these); or (ii) a viral load greater than 50 copies/ml followed by the initiation of at least two new drugs (the date of rebound was the date of the viral load value being greater than 50 copies/ml). The case notes were re-examined for each identified case of viral rebound, for which there was no record on the computerized patient database of therapy having been interrupted at the time of rebound. In many cases there was found to have been an interruption of therapy, often because of an intercurrent condition such as hepatitis, tuberculosis or gastric ulcer. Occurrences of viral rebound were classified according to whether they were known to have occurred after complete therapy interruption (interruption-associated viral rebound) or to have apparently occurred in the presence of continued therapy (viral breakthrough).
Viral rebound rates were calculated as the number of individuals with viral rebound divided by the total person-years of follow-up after the first viral load value of less than 50 copies/ml. For those patients with rebound the follow-up time was measured as the time to the date of rebound, whereas for those without rebound the follow-up time was measured to the date of the last viral load measurement. Kaplan–Meier estimates were used to assess the risk of viral rebound according to the time since viral load suppression. We also performed an analysis in which those patients with a last viral load over one year before the cut-off date for this analysis (October 2000) were treated as having experienced viral rebound at the time of the last observed viral load. When calculating the rate of viral breakthrough, follow-up for those with interruption-associated viral rebound was right-censored at the time of rebound. Changes in therapy during follow-up, which by definition occurred before viral rebound and were usually caused by toxicity problems with a particular drug, were ignored. The association between time with viral load suppression and the risk of viral rebound was assessed using Poisson regression . We assessed factors associated with the rate of viral rebound in a Cox model .
Details of the patient characteristics and of the regimens started are shown in Table 1. The median age (interquartile range) of the 336 individuals studied was 37 years (31–44). Eighty-five (25%) were women. The presumed mode of infection was 15% injection drug use; 31% heterosexual sex; 46% male sex with another man; 8% other. At the start of therapy the median viral load was 5.3 log copies/ml (range 2.8–6.7) and the CD4 cell count was 219/mm3 (range 1–895); 15% had a previous AIDS diagnosis. The calendar year of starting therapy ranged from May 1996 to July 2000, with a median of May 1998 and an inter-quartile range of August 1997–April 1999. For 20% of patients the initial regimen contained abacavir, 19% nevirapine, 16% efavirenz, 32% indinavir, 10% ritonavir and 23% nelfinavir. All but three of the 35 individuals on ritonavir were using this with another PI. Of those patients under follow-up at one year (n = 229), 14% were on abacavir, 21% on nevirapine, 24% on efavirenz, 20% on indinavir, 4% on ritonavir, and 24% were on nelfinavir. At 2 years (n = 124 under follow-up), the corresponding percentages were 6% on abacavir, 23% on nevirapine, 15% on efavirenz, 23% on indinavir, 2% on ritonavir, and 31% on nelfinavir.
There were a total of 543.1 person-years of observation in these 336 individuals by this time after the viral loads had declined below 50 copies/ml (25% followed for over 2.5 years), during which time 4321 viral load measurements were carried out; an average of one per 6.5 weeks. Overall, 61 patients experienced viral rebound during this time; one per 8.9 person-years of follow-up. There was an estimated 25.3% [95% confidence interval (CI) 18.9–31.7] risk of viral rebound by 3.3 years from the first viral load of less than 50 copies/ml (Fig. 1). All but two of the rebounds were caused by two consecutive values of over 500 copies/ml, rather than a single value of over 50 copies/ml with a change in at least two drugs (see Patients and methods section). When those patients with a last viral load over one year before the cut-off date for this analysis (October 2000) were treated as having experienced viral rebound at the time of the last observed viral load, there was an estimated 31.5% risk of rebound by 3.3 years.
For 47 of the 61 observed occurrences of viral rebound, there was a documented complete interruption of all antiretroviral therapy, so only 14 represented viral breakthrough of therapy. The rate of viral breakthrough was thus one per 38.8 years, with an estimated 5.2% (95% CI 2.4–8.0%) risk by 3.3 years. The rate of viral breakthrough decreased over time; one per 24.0 person-years (12 with viral breakthrough in 288.2 person-years) within the first year after a viral load of less than 50 copies/ml was achieved, compared with one per 127.5 person-years after this (two with viral breakthrough in 254.9 person-years;P = 0.01).
We also performed an analysis excluding those patients for whom frozen plasma samples were used retrospectively to ascertain that a value of less than 50 copies/ml had been reached, and a similar rate of viral rebound was found (27.6% at 2.6 years for viral rebound and 6.8% for viral breakthrough).
According to current guidelines, the initial aim when starting antiretroviral therapy is to achieve a viral load of less than 50 copies/ml by 16–24 weeks . We aimed to evaluate the long-term durability of viral suppression in individuals from a routine clinic cohort in whom such an initial response was achieved, distinguishing between the loss of suppression caused by therapy interruption and genuine viral breakthrough. Although we found a 25.3% risk of viral rebound by 3.3 years after initial suppression, most was caused by therapy interruption rather than failure of the virological efficacy of therapy. There was only a 5.2% risk of viral breakthrough, despite continued therapy by this time. In addition, not all interruption of therapy was necessarily documented, so even these rates may represent an underestimate of the failure of the virological efficacy of therapy. We studied drug-naive individuals, because this is the context in which triple (and more drugs) will be used in the future. The durability of suppression has been shown to be shorter in individuals with previous nucleoside reverse transcriptase inhibitor experience before starting a triple regimen in ours and other clinic cohorts [even in those initially achieving a viral load less than 50 copies/ml (data not shown)] [4,23,27], and this is likely to be related to the pre-existing selection of resistant viral strains. Our findings are important because they indicate that, when used in drug-naive individuals and if suppression to viral loads of less than 50 copies/ml is initially achieved, these regimens only rarely actually fail to maintain virological suppression, so long as they are taken. In our analysis there were too few individuals with viral breakthrough (n = 14) to assess the factors associated with this endpoint.
Reasons for the loss of viral suppression despite continued therapy could include the appearance and selection of viral strains with reduced sensitivity, failure to maintain sufficient drug levels, the development of intracellular resistance as a result of glycoprotein-P activity, impairment in phosphorylation of some drugs, or some so far unrecognized factor. Little is known about the role of many of these, and it is possible that their influence could become greater beyond the span of this study, and thus preclude the long-term virological responses suggested by our analysis. Descamps et al.  studied viral rebound occurring in eight patients on a zidovudine, lamivudine, indinavir triple drug regimen in the Trilege trial. Results suggested that low drug levels, most likely caused by poor adherence, was a major factor in causing these rebounds. Havlir et al.  studied 17 patients with viral rebound on triple therapy in ACTG343, and found evidence that those with rebound were more likely to have very low drug levels. Furthermore, Fischl et al.  showed that out of 50 individuals in the US Department of Corrections who were given directly observed three- or four-drug therapy as their first antiretroviral regimen, 95% had viral loads remaining under 400 copies/ml at 80 weeks in an intent-to-treat group (treatment change/missing ≥ 400 copies/ml) compared with 75% in a control group having self-administered therapy. These findings seem to be broadly in keeping with our own results.
Most previous analyses of rates of viral rebound after initial viral suppression have not attempted to distinguish between rebound that relates to the interruption of therapy and that which is genuine breakthrough. Indeed, unless precise details of transient therapy interruptions are recorded in case notes, such an exercise is not possible. Furthermore, few long-term studies have restricted attention to those achieving less than 50 copies/ml. Gulick et al.  studied the 3 year outcomes of individuals on zidovudine, lamivudine and indinavir, and found that 21 out of 33 had viral loads less than 500 copies/ml by this time. Nine patients experienced virological failure (two consecutive viral load values greater than 500 copies/ml). Six of these failures occurred in the first year. Levy et al.  studied the rate of confirmed viral rebound greater than 50 copies/ml on efavirenz and indinavir-based highly active antiretroviral therapy regimens. Incidences of viral rebound caused by therapy interruption were not treated as viral rebound. Of those who achieved viral loads of less than 50 copies/ml, it appears that 84% on efavirenz and 70% on indinavir had not achieved viral rebound by 2 years. Some authors (including us) have previously studied individuals with viral loads of 50–500 copies/ml on two consecutive occasions, and found that subsequent viral load values of less than 50 copies/ml without changes in therapy is common [29,30], so the 50 copies/ml definition of viral rebound may be too sensitive.
The risk of generation, by mutation, of new resistant strains relates to the frequency with which new productively infected cells arise (by ongoing new infection of cells or activation of latently infected cells) in the whole body. It is difficult to know how a given plasma level of virus translates into an amount of viral replication throughout the whole body , although one study  suggested that those with less than 50 copies/ml in plasma may be consistent with 250 000 lymph node cells actively producing new virus . Nevertheless, it seems from our results that the frequency with which new productively infected cells arise in patients with plasma viral loads of less than 50 copies/ml by 24 weeks of therapy must be sufficiently small to mean that viruses with reduced drug sensitivity are unlikely to arise.
There are several potential reasons for the decreasing rate of viral rebound that we observed. There could be some kind of selection effect, whereby those patients who are most adherent to therapy, those who experience the least drug toxicity, those who achieve most consistently high drug levels, those who have the fewest pre-existing mutations associated with drug resistance, or those who have some other biological advantage are gradually selected out. Another possible explanation is that the declining rate reflects a declining rate of the appearance of new productively infected cells, perhaps as the pool of latently infected cells becomes reduced in size. However, the rate of decline in the pool of latently infected cells has been found to be very small , so this seems unlikely.
There was a 25.3% risk of viral rebound by 3.3 years, which was mostly caused by therapy interruption. However, the true rate of interruption-associated viral rebound is likely to be higher than this, because in this analysis we right-censored follow-up at the time of the last viral load in all cases in which rebound was not observed (and thus implicitly assumed that the rebound rate in those who had not been seen was the same as in those who continued to be seen). When we assumed that all patients not seen for at least one year before the cut-off date for this analysis (October 2000) experienced viral rebound at the time of the last viral load, there was a 31.5% risk of viral rebound.
We took a longitudinal approach to the analysis, rather than the series of cross-sectional assessments (e.g. percentage with viral load below 50 copies/ml at 4 weekly intervals) that tend to be made in the analysis of many randomized trials. The longitudinal approach has advantages when analysing observational cohort data, in that it deals naturally with the fact that follow-up times and exact times of viral load measurements vary between individuals, there is rarely complete follow-up, and the previous experience of an individual (e.g. whether they have ever had viral rebound since starting a given regimen rather than just whether they are experiencing rebound at a given point in time) can be incorporated.
If viral suppression to less than 50 copies/ml is achieved within 24 weeks, the intrinsic effectiveness of highly active antiretroviral therapy in naive patients appears to be such that viral suppression can be maintained for several years in patients not interrupting therapy. Given that the complete failure of all treatment options in this group is likely to take much longer, the major challenge is to develop regimens that can be taken consistently and safely for decades.
1. Egger M, Hirschel B, Francioli P. et al
. Impact of new antiretroviral combination therapies in HIV infected patients in Switzerland: prospective multicentre study. BMJ 1997, 315: 1194–1199.
2. Palella FJ, Delaney KM, Moorman AC. et al
. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. N Engl J Med 1998, 338: 853–860.
3. Mocroft A, Vella S, Benfield TL. et al
. Changing patterns of mortality across Europe in patients infected with HIV-1.EuroSIDA Study Group.
Lancet 1998, 352: 1725–1730.
4. Ledergerber B, Egger M, Opravil M. et al
. Clinical progression and virological failure on highly active antiretroviral therapy in HIV-1 patients: a prospective cohort study. Lancet 1999, 353: 863–868.
5. Ledergerber B, Egger M, Erard V. et al
. AIDS-related opportunistic illnesses after initiation of potent antiretroviral therapy. JAMA 1999, 282: 2220–2226.
6. Mocroft A, Katlama C, Johnson AM. et al
. AIDS across Europe, 1994–98: the EuroSIDA study. Lancet 2000, 356: 291–296.
7. Staszewski S, Miller V, Sabin CA. et al
. Determinants of sustainable CD4 lymphocyte count increases in response to antiretroviral therapy. AIDS 1999, 13: 951–956.
8. Weverling GJ, Mocroft A, Ledergerber B. et al
. Discontinuation of Pneumocystis carinii
pneumonia prophylaxis after start of highly active antiretroviral therapy in HIV-1 infection. Lancet 1999, 353: 1293–1298.
9. Furrer H, Egger M, Opravil M. et al
. Discontinuation of primary prophylaxis against Pneumocystis carinii
pneumonia in HIV-1 infected adults treated with combination antiretroviral therapy. N Engl J Med 1999, 340: 1301–1306.
10. El-Sadr WM, Burman WJ, Grant LB. et al
. Discontinuation of prophylaxis against Mycobacterium avium
complex disease in HIV-infected patients who have a response to antiretroviral therapy. N Engl J Med 2000, 342: 1085–1092.
11. Miller V, Staszewski S, Nisius G, Cozzi Lepri A, Sabin CA, Phillips AN. Risk of new AIDS diseases in people on triple therapy. Lancet 1999, 353: 463.463.
12. Kempf DJ, Rode RA, Xu Y. et al
. The duration of viral suppression during protease inhibitor therapy for HIV-1 infection is predicted by plasma HIV-1 RNA at the nadir. AIDS 1998, 12: F9–F14.
13. Raboud JM, Montaner JS, Conway B. et al
. Suppression of plasma viral load below 20 copies/ml is required to achieve a long-term response to therapy. AIDS 1998, 12: 1619–1624.
14. Pilcher CD, Miller WC, Beatty ZA, Eron JJ. Detectable HIV-1 RNA at levels below quantifiable limits by Amplicor HIV-1 Monitor is associated with virologic relapse on antiretroviral therapy. AIDS 1999, 13: 1337–1342.
15. Powderly WG, Saag MS, Chapman S, Yu G, Quart B, Clendeminn NJ. Predictors of optimal virological response to potent antiretroviral therapy. AIDS 1999, 13: 1873–1880.
16. Raboud JM, Rae S, Hogg RS. et al
. Suppression of plasma virus load below the detection limit of a human immunodeficiency virus kit is associated with longer virologic response than suppression below the limit of quantification. J Infect Dis 1999, 180: 1347–1350.
17. Department of Health and Human Services. Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents.
February 2001; http://www.hivatis.org.
18. Gulick RM, Mellors JW, Havlir D. et al
. 3-Year suppression of HIV viremia with indinavir, zidovudine, and lamivudine. Ann Intern Med 2000, 133: 35–39.
19. Levy R, Labriola D, Ruiz N. Low two-year risk of virologic failure with first regimen HAART.
In:8th Conference on Retroviruses and Opportunistic Infections
. Chicago, February 2001 [Abstract 325].
20. Murphy R, Santana J, Squires K, et al.START Observational study: longitudinal follow-up of virologic and immunologic responses in START I and START II patients.
In:8th Conference on Retroviruses and Opportunistic Infections
. Chicago, February 2001 [Abstract 314].
21. Descamps D, Flandre P, Calvez V. et al
. Mechanisms of virologic failure in previously untreated HIV-infected patients from a trial of induction-maintenance. JAMA 2000, 283: 205–211.
22. Havlir DV, Hellmann NS, Petropoulos CJ. et al
. Drug susceptibility in HIV infection after viral rebound
in patients receiving indinavir-containing regimens. JAMA 2000, 283: 229–234.
23. Paredes R, Mocroft A, Kirk O. et al
. Predictors of virological success and ensuing failure in HIV-positive patients starting highly active antiretroviral therapy in Europe: results from the EuroSIDA Study. Arch Intern Med 2000, 160: 1123–1132.
24. Wit FW, van Leeuwen R, Weverling GJ. et al
. Outcome and predictors of failure of highly active antiretroviral therapy: one-year follow-up of a cohort of human immunodeficiency virus type 1-infected persons. J Infect Dis 1999, 179: 790–798.
25. Grabar S, Pradier C, Le Corfec E. et al
. Factors associated with clinical and virological failure in patients receiving a triple therapy including a protease inhibitor. AIDS 2000, 14: 141–149.
26. Clayton D, Hills M. Statistical models in epidemiology.
Oxford: Oxford University Press; 1993.
27. Staszewski S, Miller V, Sabin CA. et al
. Virological response to protease inhibitor therapy in an HIV clinic cohort. AIDS 1999, 13: 367–373.
28. Fischl M, Castro F, Monroig R, Scerpella E, Thompson L, Rechtine D, Thoma D. Impact of directly observed therapy on long-term outcomes in HIV clinical trials.
In:8th Conference on Retroviruses and Opportunistic Infections
. Chicago, February 2001 [Abstract 528].
29. Grueb G, Cozzi-Lepri A, Ledergerber B, et al.Low level HIV viral rebound and blips in patients receiving potent antiretroviral therapy.
In:8th Conference on Retroviruses and Opportunistic Infections
. Chicago, February 2001 [Abstract 522].
30. Phillips AN. Blips and when to change.
Brighton, UK: British HIV Association; March 2001.
31. Phillips AN, Loveday C, Johnson MA. HIV suppression and risk of drug resistance mutations. AIDS 1998, 12: 1930.1930.
32. Koup, RA. Mechanisms of Immune Reconstitutio on Potent ART. Topics in HIV Medicine 2000, 8: 8–12.
33. Finzi D, Blankson J, Siliciano JD. et al
. Latent infection of CD4(+) T cells provides a mechanism for lifelong persistence of HIV-1, even in patients on effective combination therapy. Nat Med 1999, 5: 512–517.