Treatment interruptions: cycles, pauses, are just plain stopping?

Dr David Margolis, University of Texas, for

Interruption of antiretroviral therapy occurs every day as part of clinical practice, and as part of everyday life for many with HIV infection.

Over the past several years prescribed interruptions of therapy have been proposed to accomplish a variety of goals: to improve the immune response via auto-vaccination with HIV, to diminish the proportion of circulating drug-resistant HIV and improve the response to salvage therapy, or to reduce the exposure to antiretrovirals and ameliorate drug-related toxicities.

A great deal of data was presented at the 10th CROI around this issue, and well summarised by Guenthard for those with the stamina to stay and listen to the meeting’s final symposium. Little support was found for an auto-vaccination effect of STI, as cytotoxic T cell responses (CTL) returned upon interruption, but without the appearance of a new, broader, or more potent response. Further, several studies showed that viral load rebounds were not significantly different after STIs, suggesting a lack of improved immune control. In the special case of HAART given upon acute HIV infection or shortly thereafter, several small studies of cycled or prolonged interruption [1, 2, 3, 4] came to differing conclusions as to the benefit of therapy in acute infection and its interruption. These differences may reflect small or disparate samples.

There are many potential drawbacks to treatment interruptions. Resistance, particularly to drug with low genetic barriers (M184V resistance for 3TC, and K103N resistance for efavirenz) has developed during STIs. CD4 declines, relapse of an acute retroviral syndrome, and opportunistic infections have all been reported during STIs. On the other hand, some patients have remained stable following long interruptions. There is clearly short-term amelioration of some drug-related toxicities following treatment interruption, long-term cumulative benefits of cycling interruptions have not yet been demonstrated. Clearly the devil is in the detail as to when therapy is interrupted, and how.

A study of intermittent therapy following the one week on/one week off model was initially piloted by the NIAID Clinical Center group [5]. It was performed in Thailand in a group of patients with a history of dual NRTI therapy followed by treatment with two NRTIs and a PI in several clinical studies. Three cohorts of such patients, approximately 25 patients per group, were then randomised to continued HAART therapy interruption until CD4 counts dropped below 350 cells/mm3, or one week on/one week off intermittent therapy. At entry CD4 counts were in the 500-700 cells/mm3 range; 75% of the subjects were men. At 48 weeks of study, 13% of the subjects in the CD4-guided interruption arm had suffered CD4 declines to < 350 cells/mm3, but all but one patient in the other arms had CD4 counts > 350 cells/mm3. However, in the week on/off arm seven patients had virological failure, one had CD4 < 350 cells/mm3, and two were lost to follow-up. Only 35% of patients in the on/off arm had HIV RNA < 50 copies/ml, while 96% of the subjects in the continuous arm were < 50 copies/ml. Of nine patients from whom genotypes were available in the on/off arm, four had resistance mutations (three in RT and one in PR). The Swiss Cohort also found evidence of significant viral replication within one week of an interruption. [6]

Overall the one week-on-one-week-off strategy appeared to be an unmitigated disaster. No evidence of decreased toxicity or improved immune control was reported, at the cost of a striking incidence of virological failure and the appearance of resistance when compared to continuous therapy. The CD4-guided arm appeared to have done well, but the median CD4 count at reinitiation of therapy was 536 cells/ul, suggesting that therapy was re-initiated after small declines in CD4.

Further support for the use of prolonged interruptions without cycling to spare toxicity or cost was provided by a preliminary report of CD4-guided therapy from the Barcelona group [7]. Of 120 patients with a median CD4 count of cells/mm3and HIV RNA < 50 copies/ml for a median of 40 months, half were randomised to interrupt therapy. Therapy was restarted if HIV RNA > 100,000 copies/ml or CD4 count dropped < 350/ul. Forty-four per cent of subjects stayed off therapy for 48 weeks, and almost all of the 56% who restarted did so because of viral rebound > 100,000/ml at a median of eight weeks after stopping. Those who restarted also had a steeper CD4 decline slope and a lower nadir CD4 cell count. Those who remained off therapy for 48 weeks lost 33 CD4 cells/mm3 per month.

Stefano Vella reported on 56 week results of longer cycled interruptions from an Italian cohort [8]. One hundred and thirty-seven patients continued therapy, while 136 underwent one month on/three months off cycles. Two-thirds of subjects used NNRTI-based HAART; median CD4 at entry was 691 cells/mm3. Due to the long half life of NNRTIs, in those on NNRTIs the NNRTI was stopped several days before the NRTIs to reduce the chance of NNRTI resistance. While patients on cycled therapy generally suppressed to < 400 copies/ml on reinitiation, a disturbing trend towards increased resistance mutations, particularly NNRTI mutations, in patients undergoing cycling gives pause. No metabolic data was yet available, and given that there is therefore no data to show the benefit of this strategy, such cycling should be left to the clinical investigator for now.

Conflicting evidence was presented with regard to the benefit of interruptions in the setting of salvage therapy. A large and careful study, CPCRA 064, studied patients failing therapy with long drug experience and multidrug resistance [9]. Salvage therapy was guided by genotypic and phenotypic resistance tests done at study entry, and patients randomised to immediate salvage or salvage therapy after a four month interruption. This population with advanced disease had a mean CD4 count of 180 cells/mm3, 26% had CD4 counts below 50 cells/mm3, a mean baseline viral load of 100,000, and exposure to five NRTIs, 4.2 PIs, and 1.5 NNRTIs. Forty-eight per cent had three class drug resistance on testing. There was no difference in the number of active drugs prescribed to the two groups (2.7 vs. 2.8).

STI in this study was inferior to immediate salvage therapy. There were 22 primary endpoints (progression of disease or death) in the STI group and 12 in the no-STI group (hazard ratio = 2.57, 95% CI = 1.2, 5.5, p = 0.01). Events in the STI arm included seven esophageal candidiasis, four PCP, three cryptosporidiosis, two lymphomas, and one CMV. Mean difference in CD4 favoured the no-STI arm by 85 cells/mm3 (p < 0.001) for months one to four (STI phase), 47 cells (p < 0.001) for months five to eight, and 31 cells/mm3 (p = 0.11) for months 12-20. The study closed to accrual as recommended by the DSMB based on data before full accrual.

In stark contrast, this strategy appeared to be successful in a study performed by Katalama and collaborators in France [10]. Patients with multiple failures of therapy and very advanced HIV disease (HIV VL > 50,000cps/ml and CD4cells < 200 cells/mm3) were randomised to either immediate therapy or therapy after eight weeks of STI. Therapy consisted of three to four NRTI and one NNRTI ± hydroxyurea (500 mg bid) and ritonavir (400 mg bid) and amprenavir (600 mg bid) or lopinavir and a third PI (indinavir 400 mg bid or saquinavir 600 mg bid or nelfinavir 1,250 mg bid). Seventy patients were randomised, 68 started study drugs, and 63 were evaluated at weeks 12 and 24, and 64 at week 48. At baseline, median plasma HIV RNA was 5.3 log copies/ml (200,000), CD4 27/_l, duration of ARV therapy was 6.6 years with a median of 11 antiretroviral drugs. By ITT missing equal failure analysis, the percentage of patients with HIV VL decrease > 1 log from baseline was 26% at week 12, 24% at week 24 if immediately salvaged, versus 62% at week 12 and 50% at week 24 in the STI group (p = 0.007 and p = 0.043, respectively). Median decrease in HIV RNA from baseline was -0.37 at week 12, -0.29 at week 24, and -0.37 at week 48 in the immediate group versus -1.91, -1.08, and -0.79 in the STI group (p = 0.008 at week 12, p = 0.013 at week 24). Percentage of patients with HIV RNA < 400 cp/ml was 15% at week 12, 12% at week 24 in immediate salvage vs 38% and 32% in the STI group (p = 0.053 and p = 0.077, respectively). Median increase in CD4 cell count from baseline was +7 cells/mm3 at week 24 and week 48 in immediate salvage vs +51 cells/mm3 and +69 cells/mm3 in the STI group. Two subjects died in each arm. Twenty-two per cent and 47% of patients were still on treatment with more than six drugs at week 48. Three major factors were associated with virologic success: treatment interruption with reversion of resistance, adequate drug concentration, and the use of lopinavir.

It was difficult to reconcile the success of this extraordinarily intense regimen with the poor outcome of CPCRA 064. Possible explanations include the longer STI in CPCRA 064, and unique, ineffable characteristics of the French patient population. Many investigators at CROI felt that it was likely that the CPCRA experience was more likely to reflect the success of this strategy in general clinical practice.

The other side of interruptions: when to continue failing therapy and why?

Several presentations addressed the continued immunological and clinical benefits of therapy that is continued despite loss of complete (that is, for the time being, <50 copies/ml) suppression of viral replication. A synthesis of the insights into the benefits of antiretroviral therapy, the still-emerging long-term risks of therapy, and the risks and benefits of cycled or intermittent therapy is probably the most important immediately clinically applicable insight to be gained from the 10th CROI.

Steve Deeks provided complex and provocative insights during his symposium discussion of ‘When to Switch’ [11], and important details were available in his poster with Bob Grant and the ViroLogic group [12]. Deeks himself expressed concerns after the presentation that his studies, meant to be an experiment to explore the relative immunological benefits of different components of antiviral therapy, will be misinterpreted as a prescription for inappropriate use of HAART.

It was hypothesised that among treated patients with multi-drug resistant HIV, interruption of all drugs from a single therapeutic class (‘partial treatment interruptions’) might (1) maintain partial viral suppression and its associated immunologic benefit, (2) prevent overgrowth of wild-type HIV, (3) delay viral evolution, and (4) reduce drug-toxicity and drug costs. Twenty subjects were studied in this non-randomised, open-label pilot study. Subjects actually chose which drug class they would interrupt. The median baseline viral load was 3.9 log [8,000] copies/ml (IQR 3.6 – 4.5) and the median CD4 T cell count was 336 cells/mm3.

Fifteen volunteers interrupted all PIs and continued all NRTIs; in these subjects viremia and CD4+ T cell counts were stable over 24 weeks. As expected after interruption of PIs, fasting triglycerides and non-HDL cholesterol improved. Genotypic and phenotypic resistance remained stable in all patients interrupting PI therapy through week 16-24; however, PI mutations waned and replicative capacity and viremia increased in two patients after week 24. The conclusion from this finding in this brief, small, uncontrolled study should be the insight that continued PI therapy in the presence of MDR HIV contributes to the control of viral replication, and may confer a potential clinical benefit. This is not surprising as many studies have suggested this in the past, but this experiment offers strong evidence that patients are benefiting from PIs despite resistance and viremia. The conclusion should not be that sub-optimal nucleoside therapy is acceptable in patients with MDR virus.

Five other volunteers interrupted NRTI therapy and continued PI therapy. In contrast, immediate and sustained increases in viremia (+0.03 log copies/week, P < 0.001) were observed. This is not surprising to virologists, as many studies have suggested the maintenance of RT mutations results in a virus that grows more slowly, and that this may confer a clinical benefit. Of note, three of five subjects interrupting NRTI therapy exhibited a delayed loss of M184V, which was temporally associated with a rise in viremia. This finding should provide satisfaction to Mark Wainberg and others, as it is the first clear and direct validation of their contention made in the 1980s that the maintenance of the 3TC resistance mutation confers a clinical benefit.

Deeks et al concluded that interrupting PI therapy in patients with multi-drug resistant HIV is associated with stable viremia, reduced toxicity, and halted accumulation of drug resistance (which may preserve future PI options). However this effect may not hold for more than 24 weeks, as illustrated in a handful of their cases. On the other hand, clinical benefits may be gained by a drug holiday of a few months. In contrast, interruption of NRTI therapy was associated with rapid rises in viremia, indicating that NRTIs may have continued antiviral effects against drug resistant HIV. Partial treatment interruptions may be appropriate for maintaining partial virologic responses in persons with limited treatment options. The critical caveat to these conclusions is the specific patient population in whom this clinical strategy might be considered. The patients in this study all had preserved CD4 counts (366 cells/mm3) and many PR and RT mutations. Such a manoeuvre is unlikely to work without a virus significantly impaired by MDR mutations and without a partially intact immune response. However, the findings of this group should give us pause for careful consideration in making therapy changes or interruptions, and identifies important questions to be explored in careful clinical trials.

Another important data set with regard to the decision to interrupt or continue therapy was presented by the PLATO collaboration [13]. This is a prospective observational study following patients from 13 cohorts in Europe, North America, and Australia. Data from 2,448 patients with 3-class drug failure was obtained, and CD4 counts were used to determine CD4 slopes during treatment failure. CD4 counts were excluded from the analysis if the concurrent viral load was not ‘stable’. That is, if the VL blipped up and subsequent VLs returned to the prior ‘stable’ level, the CD4 count at the time of the blip would not be analysed. It appeared that if the VL ‘reset’ to a stably higher level, CD4 counts would be used in this analysis. 2,596 CD4 counts were obtained to obtain CD4 slope during failure from 628 patients in whom baseline pretreatment CD4 counts were known. These patients were 90% male, 60% MSM, CD4 median 170/ul, mean VL 4.5 logs (32,000), and on any therapy for a median of 4.1 years, on HAART for a median of 2.2 years, and exposed to a median of eight antiretrovirals. The take-home message was that for any VL at failure, the CD4 slope was greater if the subject was off antiretrovirals, and that despite failure the mean CD4 slope per year was still positive (CD4 gain) if the VL was less than 4.0 logs (10,000). Further, even when failing with VLs of 4.5 logs or greater, the CD4 slope was roughly 20 cells/year, less than the 70 cells/yr seen in untreated infection. Further, when examining the delta viral load from setpoint (eg. 5 logs prior to therapy – 4 logs at failure = 1 log delta VL), a delta VL of 2 logs or more predicted a positive CD4 cell slope (CD4 gain).

Numerous studies have shown the gradual accumulation of drug-resistance mutations when viremia is detectable. This process occurs more rapidly for single point mutations that confer high-level drug resistance, and when the level of viremia at failure is higher. However, decisions about when to change therapy should take into account the continued benefit of partial suppression of viral replication (delta viral load).


There is a reluctance to accept the findings from the GIGHAART study in the US, and this study was actually submitted, but not accepted, for last year’s Retrovirus conference. Some of the comments in this article still reflect this. However, the STI in the GIGHAART study not only selected a shorter STI period but reported success only when patients retained phenotypic drug sensitivity and optimal drug levels were achieved. Other multiple drug regimens have also produced greater sustained levels of viral load reduction than regular three or four drug rescue therapies.

Stopping antiretrovirals in patients with low CD4-cells may result in a higher risk of having an AIDS defining event and a longer period with lower CD4 than on a plain switch. Some cohort studies (e. g. EUROSIDA) have shown that ART may have a partially protective effect despite virological failure in patients with <100 CD4-cells. These facts should be discussed with a patient before interrupting treatment.

Reductive therapy to two nucleoside reported by Deeks may only be appropriate with a sufficiently protective CD4 count (in this study >300 cells/mm3. Individual patient histories may hold some of the reasons for this success, rather than the strategy itself. Maintenance dual therapy was originally suggested as the alternative arm to megaHAART in the OPTIMA study, though this was not adopted due to caution for accumulation of nucleoside mutations.


Unless stated otherwise, all references are to the Programme and Abstracts of the 10th Conference on Retroviruses and Opportunisitc Infections (CROI), 10–14 February 2003, Boston.

  1. Hoen B, Lacabaratz C, Fournier I et al. Structured treatment interruptions in acute HIV seroconverters: preliminary results of the multicenter prospective PRIMSTOP pilot trial (ANRS 100). Poster 512
  2. Lafeuillade A, Counillon E, Poggi C et al. Predictors of plasma HIV RNA control after discontinuation of HAART initiated at acute infection. Poster 513
  3. Goujard C, DeveauC, Sine M et al. Impact of therapeutic interruptions in patients treated during primary infection. Poster 517
  4. Hecht FM, Wang L, Collier A et al. Is HAART for primary/early HIV infection associated with improved outcomes after treatment discontinuation? Poster 519
  5. Ananworanich J, Cardiello P, Srasuebkul P et al. HIV-NAT 001.4: A prospective randomized trial of structured treatment interruption in patients with chronic HIV infection. Abstract 64
  6. Fischer M, Hafner R, Schneider C et al. HIV-plasma RNA rebounds within days during structured treatment interruptions. Poster 639
  7. Ruiz L, Gómez L, Domingo P et al. A multi-center, randomized controlled clinical trial of continuous vs intermittent HAART guided by CD4+ T-cell counts and plasma HIV-1 RNA levels. Abstract 65
  8. Vella S, Giuliano M, Palmisano L et al. ISS-PART: A prospective, randomized, multi-center clinical trial of intermittent therapy in HIV+ subjects with persistent suppression of viral replication. Abstract 66
  9. Lawrence J, Mayers D, Huppler Hullsiek K et al. CPCRA 064: A randomized trial examining structured treatment interruption for patients failing therapy with multi-drug resistant HIV. Abstract 67
  10. Katlama C, Dominguez S, Duvivier C et al. Long-term benefit of treatment interruption in salvage therapy (GIGHAART ANRS 097). Abstract 68).
  11. Deeks SG. When to switch antiretroviral therapy. Abstract 188
  12. Deeks SG, Martin JN, Hoh R et al. Continued reverse transcriptase inhibitor therapy is sufficient to maintain short-term partial suppression of multi-drug resistant viremia. Abstract 640
  13. Ledergerber B, Lundgren JD, Gregory P et al. Factors affecting CD4 count slope in patients with stable viral load following three class virologic failure: the PLATO collaboration. 146LB Oral Session 27.


Links to other websites are current at date of posting but not maintained.