Treatment interruptions

Mike Youle, MD, Royal Free HIV Clinic, London, for NATAP

Treatment interruption, stopping therapy pre-term and fixed period intermittent on-off therapy schedules have all been proposed, discussed and to some degree studied over the past couple of years.

A number of things about these strategies are clear, much more is less so. When treatment is stopped viral load rebounds usually to around the level pre-treatment and over the next few months CD4 cell counts decline at a variable rate that appears to have little predictability. Some small studies (none randomised) have hinted that this strategy is more successful in those people who were treated at seroconversion and that the further along in HIV disease you go the more risky interrupting therapy that has suppressed HIV, will be. Tantalizing data, first shown at the 8th European Conference on Clinical Aspects and Therapy in HIV in Athens in October 2001 appeared in the Giga-HAART study, from Christine Katlama’s group in Paris. This was the first controlled data to build on work by our group, that of Steve Deeks from UCSF and of Veronica Miller formerly of Frankfurt that showed that potential benefits existed in short term treatment interruptions in salvage patients who then recommenced a new regimen.

So this conference was always going to be a forum for synthesizing new data into the idea that interruption of antiviral suppression may be a double-edged sword. What you gain by lower toxicity and lack of pills may be offset by physical and laboratory changes showing that HIV was once more attacking the immune system. The fundamental question of course still stands – do you gain any delay in disease progression by treatment that is not available if you delay treatment until the risk of disease occurrence is higher? So the “treat now or treat later” discussion is intricately linked to the question of treatment interruption. An in-depth meeting convened by The Forum for Collaborative Research will be held later this month in Montreal to discuss the multifarious aspects of treatment interruption. In addition there is now a large clinical endpoint study called Optima that will examine 1300 patients who are triple class failures and who will be randomised to have a three to six month treatment interruption versus immediate switch to a salvage regimen (similar to the giga-HAART approach). Also they will be randomised to take a standard three drug optimised HAART regimen or a mega-HAART five or more drug regimen. This trial is now open to recruitment in the US through VA hospitals, across Canada and the UK and should provide some more definitive answer as to the utility of this approach in later stage disease.

More information can be obtained at the following website:

Does an interruption encourage risk for neurocognitive dysfunction?

Overall there were 26 presentations on the topic of STIs, including two late-breakers in Seattle. They fell into a number of subject groups. The first were those that examined the effects on the viral populations and immune function in different tissues when treatment was ceased. Andrew Leigh Brown from the University of Edinburgh worked with a group from UCSD led by Ellis to examine the dynamics of viral rebound in the cerebrospinal fluid (CSF) [1, 2].

In two of nine patients studied where sufficient data points were available (understandably, repeat lumbar punctures are not popular with patients), there was a difference between CSF and plasma viral kinetics. A rise in white cells (pleocytosis) in four subjects was seen following the viral burst (viral load increase) in the CSF, suggesting that this is a response rather than a cause of viral turnover in the central nervous system. This work also appears to shed light on the issue that reducing viraemia systemically may limit the entry of lymphocytes and linked virus into the CSF.

Further work to evaluate the source of virus when treatment is suspended came from the group of Lydia Ruiz and Martinez-Picardo in Barcelona, Spain. They extensively examined the env sequences for 12 subjects over up to four sequential structured treatment interruptions (SSTI) [3]. Genetic analysis was performed for portions of the envelope genome of the virus in DNA, plasma RNA and PBMCs. In some patients there was little shift in the genetic structure of emerging populations whilst in others there was a marked alteration. This may limit the ability of treatment interruptions to strengthen the HIV-specific cell-mediated immune responses since the exposed viral antigens keep changing during each SSTI. A single subject was more extensively studied and examination of plasma samples revealed that the HIV populations that emerged over successive interruptions probably reflect a combination of reactivation of archived populations and ongoing selection due to the effect of cytotoxic T-lymphocyte responses [4]. The role of neutralising antibodies in treatment interruptions was examined by a group of investigators working on the Swiss Spanish Intermittent Therapy Trial, where subjects had four cycles of two weeks off/eight weeks on therapy [5]. In 10 subjects at week two and 25 subjects after week 42 virus was isolated from PBMCs and examined for replicative capacity and for neutralising antibodies. Although neutralising antibody levels did not increase over the cycles of the study during the four to six treatment free periods, after the four cycles there was a significant increase in these antibodies (90% inhibition titres >1/1000). This occurred especially in the subgroup of subjects who had maintained control of viraemia throughout the study.

Risk of HIV progression in therapy interruptions in EuroSIDA patient group

Next were those studies that evaluated the effects of treatment interruption of three months or more on the progression of disease in patients who either interrupted or did not interrupt therapy. The largest of these came form the EuroSIDA cohort that represents large clinical centres across Europe and is a long established study that has been prospectively followed for many years [6]. This study received much attention and discussion by attendees at the Conference.

8,530 patients are followed in EuroSIDA. Jens Lundgren from Copenhagen presented data on 565 (16%) subjects who interrupted therapy in a cohort of 5385 patients representing 15,312 person years of follow-up after commencing HAART up to the first interruption or last follow-up visit. There were 10,637 person years of follow-up from the start of HAART to the first new AIDS event/death or last follow-up visit, to assess the risk of disease progression.

Characteristics of the patients when they started HAART: 80% were male; average age 37; transmission risk: 48% MSM, 24% heterosexual; HIV viral load 25,000; CD4+ count 200; nadir (lowest ever) CD4+ count 138; 31% previously had AIDS; on average patients started HAART in March 1997 (October 1996 through December 1998). Over the course of 60 months the number of patients interrupting therapy increased. By the 60 month time point a total of 20% of patients had interrupted therapy.

What were the risk factors for patients who interrupted therapy?

More females than men interrupted therapy (p=0.02). More IVDUs interrupted therapy than MSM (P<0.001). Patients from South (p<0.001) or Eastern Europe (p=0.005) tended to interrupt less frequently than patients from Central or Northern Europe. Patients tended to interrupt therapy the longer they were on therapy (p<0.001), with a 30% increase per calendar year. Patients with higher viral load on HAART tended to interrupt therapy more often (p<0.001). If a patient previously had AIDS that did not seem to indicate a propensity to interrupt therapy. Patients with higher CD4+ tended to interrupt therapy (p=0.07).

At the time of interruption the median CD4+ cell count was 228 and viral load was 4000. Patients lost about 30 cells on average over the first three months following the interruption. Viral load increased on average about 1 log during the first three months.

What is the pattern of disease progression observed (patients taking interruptions of three months or more)?

There were 408 clinical events for patients while on HAART and 37 for patients off HAART. The death rate was 25% for the patients on HAART and was about 37% for patients off HAART. NHL was mentioned as occurring more often for patients on HAART. For patients off HAART pulmonary TB, PCP, and oesophageal candidiasis occurred. At the time of these events the average CD4+ count was 111 in the on HAART group (23-239), and 23% of the events occurred when CD4+ were above 250. Average viral load was 8000 (400 to 126,000). The average CD4+ count off HAART when events occurred was 52 (12-120). The maximum CD4+ count at which an event occurred was 250. Average viral load was 158,000 (20,000 to 500,000).

What was the incidence of clinical disease progression (interruptions of 3 months or more)?

The rate (number of new AIDS events or deaths per 100 person years of follow up) for patients who interrupted therapy was 20.3 (37/182) vs 3.9 (408/1054). Patients with less than 50 CD4+ had a dramatically worse rate of disease progression to AIDS or death (80% [14/18] vs 35.5% [117/331), if they were off HAART. This suggests that being on HAART had a benefit regardless of CD4+ count. For patients with 50-199 CD4+ cells on their latest blood test, the rate of disease progression was also greater for patients off HAART (32.3% [18/56] vs 7.3% [163/2241]). Again, this suggests an independent benefit from being on HAART regardless of CD4+ count. Interestingly, for patients with more than 200 CD4+ there did not appear to be a difference in disease progression whether they interrupted or stayed on HAART (2.8 rate for interrupters vs 1.6 rate when staying on HAART). This suggests that an interruption may be safer if CD4+ are higher. This question is a bone of contention. Some doctors feel it is relatively safe to take an interruption if CD4+ are high enough. They feel if CD4+ are too low a patient should be placed on prophylaxis medication for opportunistic infections. Other doctors feel an interruption is risky even if CD4+ are higher and the uncertainties about what may happen during an interruption concern them. As seen in the Ellis study, viral load in the CSF increases significantly during an interruption. It is assumed that viral load increases in many parts of the body and reservoirs after an interruption. The long-term effects of re-populating reservoirs with HIV and the gyrations in viral load in reservoirs are unknown. Will this wear out the immune system and reduce its long-term ability to control HIV? No one knows the answer to this question.

Lundgren reported that age, being off HAART, and having had previous AIDS diagnosis was associated with a hazard for experiencing an AIDS event or death. The risk of death/AIDS was increased five-fold if a patient was off HAART, and was the greatest risk factor for AIDS/death. A lower CD4 count and a higher viral load were also associated with risk for developing AIDS/death whether the patient was on or off HAART. But the risk was greater if the patient was off HAART.

Lungren mentioned the limitations of his study. The interruptions were not structured therapy interruptions. There may be some biased reasons why some patients interrupted therapy. For example, a patient may have been terminally ill. The implications from taking interruptions of less than three months were not studied. He summarised that the risk of interruptions in this study were closely linked to the latest CD4+ count. Interruption is risky if you have a low CD4 count. Interruption does not appear unsafe if CD4s do not fall below 200 to 250. HAART appears to have a benefit not explained by its effect on CD4+ and viral load. However, this benefit disappears once you stop drugs so Lundgren raised concern about interrupting therapy in patients with low CD4+ and high viral load.

Royal Free Cohort

Lampe and co-workers followed all 237 people from the Royal Free Hospital Clinic who were naive to antiretrovirals when they started a HAART regimen and who reached a viral load < 400 copies/mL (within 32 weeks) or who reached a viral load < 50 copies/mL without viral rebound [7]. Twenty one percent of subjects were female; main HIV exposures were homosexual (60%) and heterosexual (34%) sex. The median age at start of HAART was 35 years. HAART was started on a median date of Dec 98 (IQR Sep 96 – Nov 00). Baseline median (IQR) viral load was 5.3 (4.8 – 5.7) log copies/mL and CD4 count 193 (75 – 302). Nucleoside drugs in the initial regimen were zidovudine/3TC 126 (53%), stavudine/3TC 76 (32%) and other 35 (15%). Other drugs were nevirapine 58 (25%), efavirenz 44 (19%), indinavir 56 (24%), ritonavir 21 (9%), nelfinavir 50 (21%) and ritonavir boosted PI 29 (12%).

Median time from start of HAART to first measured value <50 copies/ml was 180 days. 1342 viral load measures were made over a total 347 person-years of follow-up after the first viral load < 50 cps/ml – a median of one measure per 13.5 weeks (median over all patients of 3.8 per year). Maximum follow-up was 4.4 years. Overall 13 people (5.4%) experienced viral rebound on therapy (rate 0.037 per person-year) representing one person with viral rebound per 26.7 person-years of follow-up. There was a (non-significant) trend towards lower rebound rate with increasing time from start of HAART.

Treatment interruption before intensification

The Barcelona group also presented a prospective randomised study of subjects with multiple exposure to antiretrovirals who were randomised to a three month treatment interruption or an immediate switch to a new salvage regimen that consisted of Lopinavir/ritonavir plus saquinavir soft-gel, 3TC, ddI and abacavir [8]. In the 22 subjects who had the treatment interruption (Group A) 13 (64%) showed a reversion to wild-type virus during this period, perhaps raising the question as to whether the period off treatment was long enough. The outcome in this group were compared to the 24 subjects from Group B who had an immediate switch to the salvage regimen. They showed no significant differences. Within three months after restarting therapy the CD4+ count in this group was about the same as the patients who did not take a break in therapy. Longer-term follow-up in greater numbers of patients may however be required to show such differences. The other major Catalan research group of Jose Gatell evaluated the use of mycophenalate mofetil (MMF) and the cytostatic drug hydroxyurea in groups of patients who had started HAART close to seroconversion [9, 10]. These were small studies 15 and 20 patients respectively but did show that blunting of viral load rebound occurred under both agents that appeared not to be related to decrease in cell turnover or from increase in apoptosis. Further larger studies would be valuable for each of these agents perhaps combined with another approach.

Therapeutic vaccination

Therapeutic vaccination to maintain the benefits accrued during antiretroviral therapy has become an area of intense interest with several studies currently reaching a degree of fruition. For instance, the Quest study is currently evaluating the utility of a canarypox vaccine, Remune, or both prior to cessation of HAART in subjects who have been treated just after seroconversion. The group of Franco Lori and Juliana Lisziewicz presented data evaluating the utility of a novel topical DNA immunisation (DermaVir) in preventing viral rebound during treatment interruptions in macaques [11]. Ten animals with late-stage disease were studied and randomised to continuous HAART, STI-HAART (three weeks on three weeks off) and after six cycles were also given DermaVir (a composition of HIV DNA plus Langerhans cell stimulant called polyethylenimine-mannose designed to stimulate recognition of the DNA by the animals’ immune system). All the macaques on HAART alone were dead by month seven whereas only one on STI-HAART had succumbed. Of the three monkeys who were then treated with DermaVir, median viral load decrease during each subsequent interruption occurred until control at <200copies/mL was achieved. Clearly if this were possible to replicate in humans it would be highly significant and could open the door to combination HAART and immunotherapy combination studies.

Other studies also seemed to suggest this trend with different vaccinations. An NIH study using NYVAC-SIV in macaques showed similar blunting of viral rebound in immunised animals versus controls [12]. A pilot study nested within a Remune protocol evaluated the rise in viral RNA in subjects who stopped therapy for 6 weeks after suppressing to undetectable for at least 6 months [314-W]. In the 20 Remune patients the slope of the rise was 0.16 log10 copies/mL versus 0.21 in the eight control subjects who were only treated with the adjuvant IFA (p<0.05). The difference between the post-interruption viral load level between the groups was not significant but the numbers of patients in the study precluded this being properly evaluated.

So many approaches are being taken to the treatment interruption scenario, using HAART, immunotherapy, cytostatic treatments and immune enhancing agents. This once again suggests that a combination of different interventions may be the best way to continue to support the inherent immune response to HIV.

Additional information on STIs reported at the Conference

Douek and colleagues from the NIH Vaccine Research Center added a cautionary note about therapy interruptions [13]. They identified populations of T cells responding (ie. producing interferon gamma) to either HIV antigens or (as a control) CMV antigens, and sorted these cells by flow cytometry in patients during a treatment interruption. Then using a novel PCR technique, they compared the amount of proviral HIV DNA within the two cell populations. The proportion of either population that was infected, as evidenced by the presence of HIV DNA within them, was low. However, HIV-specific memory CD4 cells were significantly more likely to be infected. This could be looked at as bad news, or no news. In responding to sites of HIV replication, like a fire fighter these cells place themselves at risk. Douek concluded that this phenomenon of HIV specifically infecting the very cells that respond to it adds a cautionary note to the practice of STIs.

The overall question that remains to be answered is whether HIV replication can be controlled by novel strategies that pairs cycles of antiviral therapy with cycles of immunotherapy, or whether slow but relentless infection of responding antiviral cells eventually erodes the immune response. Research so far has shown no evidence that therapy interruptions in persons chronically infected with HIV result in improved control of HIV when off HAART. Current studies continue to examine whether a drug holiday before initiation of salvage therapy improves response, whether early treatment within the first few months of therapy allows treatment of limited duration with better control of HIV replication, or whether immunotherapies such as HIV vaccines given during HAART allow HAART interruption with better control of viral replication. This study illustrates at the single-cell level the cost of unrestrained viral replication resulting from an interruption.

Two additional studies add cautionary notes about treatment interruptions. If a patient takes repeated interruptions (obviously) over time their average viral load will increase and their average CD4 count will decline. The studies below take a broad view cautioning about risks associated with this.

Total HIV viral load over time predicts death and AIDS events

In an interesting study from the Swiss HIV Cohort Study [Eggar, 14], researchers looked at the amount of viral load a patient has over time and how that might predict disease progression. Eggar examined the area under the viral load curve (AUC) as a measure of average viral burden and assessed its value as a predictor of clinical progression. They also examined cumulative drug exposure. A total of 2,324 patients who started HAART between 1st September 1995 and 30th November 1998, who had a CD4 count and viral load measurement within three months before starting HAART and at least one follow-up visit were included. The median viral burden over time was 2.80 log (630 copies/mL) copies (interquartile range 2.67 to 3.37).

The risk of progression to a new AIDS event or death was strongly related to viral burden over time. Kaplan-Meier probabilities at five years were:

  • 7.2% for patients with viral burdens of less than 3.0 log copies/mL per year
  • 20.6% for viral burdens of 3.0 log to 3.99 log
  • 38.7% for viral burdens of 4.0 log to 4.99 log
  • 79.1% for patients with viral burdens of 5 log or greater

Baseline CD4 cell count, a history of IVDU, and older age also predicted disease progression, but baseline viral load, clinical stage, treatment history, and sex did not. Drug exposure increased with increasing viral burden.

Viral burden over time was a stronger determinant of clinical progression than viral response in the first year after initiating HAART. Kaplan-Meier probabilities at five years:

  • 11.0% for patients who achieved and maintained undetectable viral loads
  • 15.9% for patients who achieved undetectable concentrations but had a viral rebound
  • 51.5% for patients who never achieved undetectable concentrations
    Probability of Progression to New AIDS Event or Death, using Kaplan-Meier curves:
  • 22 times greater if average viral load was 100,000 or higher compare to 0 to 2.9 log (1000 copies)
  • 6.5 times if average viral load was 10,000 to 100,000, compared to 0 to 2.9 log
  • 3.3 times greater if average viral load was 1,000 to 10,000, compared to 0-2.9 log

Long term effect of therapy interruptions on average CD4+ counts and viral load

Over 2000 patients in the Swiss HIV Cohort study were analysed over four years to see how interruptions affected their CD4+ count and viral load over time [15]. They looked at patients who either took or did not take interruptions, after starting HAART in 1996/1997. After four years, researchers found that 64% had <400 copies/mL of viral load, but only 50% on patients who interrupted therapy had undetectable viral load compared to 82% of patients who did not interrupt therapy. Average CD4+ increased from 200 before therapy to 357. But again, CD4+ increased only to 300 in patients who interrupted therapy compared to 420 for patients who did not interrupt therapy. After four years patients who interrupted therapy were more likely to have less than 200 CD4+ and less likely to have greater than 500 CD4+, compared to patients who did not interrupt therapy. 47% of patients who did not interrupt therapy vs 24% who interrupted therapy had greater than 500 CD4s. CD4 counts remained below 200 in 6% of patients who did not interrupt therapy vs 29% of patients who interrupted therapy. These differences were all statistically significant.


  1. R. J. Ellis, S. Letendre, S. D. W. Frost et al – Dynamics of HIV Rebound in Cerebrospinal Fluid (CSF) after Interruption of Antiretroviral Therapy (ART). 9TH CROI Abstract 49
  2. R. W. Price, A. Nilsson, S. G. Deeks et al – Relationship of Cerebrospinal Fluid (CSF) White Blood Cells (WBCs) to CSF and Systemic HIV Infection: Cross-Sectional and Longitudinal Analysis. 9th CROI Abstract 62
  3. J. Martinez-Picado, S. D. W. Frost, S. Marfi et al – Viral Populations Emerging in Plasma during Sequential Structured Treatment Interruptions in Chronically HIV-Infected Individuals Are Genetically Distinct between Time-Points and Tissues. 9th CROI Abstract 50
  4. B. Joos, M. Fischer, A. Trkol1 et al – Long-Term Multicompartment Evolutionary Study of HIV-env in a Chronically HIV-Infected Patient before and during HAART Followed by 5 Structured Treatment Interruptions (STI). 9th CROI Abstract 531-M
  5. A. Trkola, H. Kuster, C. Ruprecht et al -The Role of Patient Virus Properties and Neutralizing Antibody Responses in Structured Treatment Interruption Therapy. 9th CROI Abstract 243-T
  6. J. D. Lundgren, S. Vella, L. Paddam et al – Interruption/Stopping Antiretroviral Therapy and the Risk of Clinical Disease: Results from the EuroSIDA Study. 9th CROI Abstract 48
  7. F. Lampe, M. A. Johnson, C. Loveday et al -Viral Rebound after Suppression with HAART: Experience from 237 People with Viral Load < 50 copies/mL Followed for up to 4.4 Years. 9th CROI Abstract 536-M
  8. L. Ruiz, E. Ribera, A. Bonjoch et al -Virological and Immunological Benefit of a Salvage Therapy that Includes Kaletra plus Fortovase Preceded or not by Antiretroviral Therapy Interruption (TI) in Advanced HIV-Infected Patients (6-Month-Follow-up). 9th CROI Abstract 421-W
  9. F. García, M. Plana, M. Brunet et al – Effect of Associating an Immunosuppressive Therapy (Mycophenolate Mofetil: MMF)+HAART during STI and Holding the MMF Drug after Definitive Interruption of HAART on Viral Replication. 9th CROI Abstract 534-M
  10. M. Plana, L. Lopalco, F. García et al -Effect of Associating a Cytostatic Drug+HAART and Holding the Cytostatic Drug after STI and a Definitive Interruption of HAART on HIV-1-Specific Immune Responses. 9th CROI Abstract 535-M
  11. J. Lisziewicz, J. Xu, J. Trocio et al – Control of Viral Load Rebound during Treatment Interruptions in Macaques with AIDS Induced by a Novel Topical DNA Immunization (DermaVir). 9th CROI Abstract 312-W
  12. E. Tryniszewska, M. G. Lewis, Z. Hel et al – Containment of Viral Rebound after Antiretroviral Therapy Suspension in Macaques Chronically Infected with SIV following Vaccination with NYVAC-SIV Recombinant Vaccines. 9th CROI Abstract 313-W
  13. D. Douek, J. Brenchley, M. Betts et al -HIV Preferentially Infects HIV-Specific CD4 T Cells. 9th CROI Abstract LB7
  14. M. Egger, B. Ledergerber, P. Grob et al – Viral Replication and Long-Term Clinical Progression in Patients Treated with Potent Antiretroviral Therapy. 9th CROI Abstract 471-M
  15. G. Kaufmann, L. Perrin, G. Pantaleo et al – CD4 T-Lymphocyte Recovery in Individuals with Advanced HIV-1 Infection Receiving Potent Antiretroviral Therapy for 4 Years: The Swiss HIV Cohort Study. 9th CROI Abstract LB8

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