CROI News Feature: Experts divided over the future of treatment interruptions

Experts at the Thirteenth Conference on Retroviruses and Opportunistic Infections, held this week in Denver, were divided over the future of structured treatment interruptions during a lively panel discussion that followed the presentation of the results of the SMART study, reported separately here along with five other structured treatment interruption studies. Whilst the CD4-guided treatment interruption arm of the Trivacan trial in west Africa was also terminated early due to excess illness in those who interrupted therapy, two other CD4-guided treatment interruption studies - Staccato and ACTG 5170 - appeared to have more positive outcomes, as did two fixed-length treatment interruption trials, Window and PART.

The Trivacan trial - sponsored by the French National Agency for AIDS research (ANRS) - is one of two structured treatment interruption studies currently taking place in sub-Saharan Africa. The trial, based in the Ivory Coast capital of Abidjan, planned to compare two different structured treatment interruption strategies with continuous highly active antiretroviral therapy (HAART) in antiretroviral naive HIV-positive adults with CD4 counts between 150-350 cells/mm³.

Ninety-one percent of the 840 participants (77% of whom were women) received a HAART regimen comprising two nucleoside reverse transcriptase inhibitors (NRTIs) - AZT (zidovudine, Retrovir) and 3TC (lamivudine, Epivir) - alongside the non-nucleoside reverse transcriptase inhibitor (NNRTI), usually efavirenz (Sustiva); the remaining 9% receiving two NRTIs plus a protease inhibitor (PI).

After at least six months on HAART, and only when CD4 counts were above 350 cells/ mm³ and viral load was below 300 copies/ml, participants were randomised to three arms: a fixed-length treatment interruption arm of two months on, four months off therapy; a CD4-guided treatment interruption arm where treatment stopped at 350 cells/mm³ and was restarted CD4 counts fell to below 250 cells/mm³; and continuous HAART.

In October 2005, after the independent Data Safety Monitoring Board (DSMB) performed a review of the study, the DSMB recommended stopping the CD4-guided treatment interruption arm prematurely because serious illness - a primary endpoint of the study - was significantly more frequent compared with those in the fixed-length interruption and the continuous treatment arms.

Christine Danel from the University of Treichville, Abidjan, presented 19 month follow-up data from the 216 participants in the CD4-guided treatment interruption arm and the 110 participants in the continuous treatment arm, that showed a remarkably high number of World Health Organisation (WHO) stage three or four classifying events in the participants who underwent CD4-guided interruptions.

With an incidence rate of 17.6 per 100 person-years, compared with 6.7 per 100 person-years, incidence of serious illnesses was 2.6-times higher in the CD4-guided interruption arm compared with those on continuous treatment. Participants in the CD4-guided treatment interruption arm were sixteen-times more likely to be diagnosed with invasive bacterial infections (most commonly pneumococcus and salmonella), three-times more likely to have severe oral candidiasis ("thrush"), and 1.5-times more likely to be diagnosed with tuberculosis (TB) than those in the continuous treatment arm.

Of the 24 invasive bacterial infections diagnosed in the CD4-guided treatment interruption arm, fourteen occurred with documented bacteraemia, despite everyone in this arm being prophylactically prescribed the antibiotic co-trimoxazole (Bactrim/Septrin). During questions from the audience, Dr Danel said that 85% of the bacterial infections were co-trimoxazole-resistant.

Dr Danel also said that five deaths occurred during the study, with the risk of death being twice as high in the CD4-guided interruption arm compared with those on continuous treatment.

No data was presented on the fixed-length treatment interruption arm of two months on, four months off therapy, which is continuing. This is somewhat similar to the treatment interruption arm of the five-year DART (Development of Antiretroviral Therapy in Africa) study currently taking place in Uganda and Zimbabwe, which compares participants who have taken HAART for 48 or 72 weeks with CD4 counts over 350 cells/mm³ to either continuous therapy or to three months on, three months off therapy. The DART DSMC and Trial Steering Committees had previously assessed as safe a pilot study of 100 patients who underwent this treatment interruption strategy.

The PART study, presented by Lucia Palmisano from the Institute Superiore di Sanità in Rome, Italy, also concluded non-inferiority of their complex version of fixed-legnth structured treatment interruptions compared with continuous therapy, based on a primary endpoint of the proportion of participants with greater than 500 CD4 cells/mm³ after two years' follow-up.

No major illness or death was seen in this study, probably due to both the high CD4 nadirs (above 400 cells/mm³) and high CD4s at entry (around 700 cells/mm³). The 273 participants had previously been on first-line therapy for an average of two years with no prior failure, 70% of whom were on NNRTI-based HAART, the remainder on unboosted PI-based HAART.

However, the emergence of resistance in the treatment interruption arm led to a very high dropout rate: almost five-times as many participants randomised to the treatment interruption arm left the study compared with those on continuous therapy. Indeed, of the 136 participants randomised to interrupt therapy, 38 (30%) developed mutations, leading to a 2.6-fold increased risk of virological failure when they resumed therapy, compared with the participants whose HIV did not develop resistance.

Confusingly, Dr Palmisano reported that the emergence of mutations was independently predicted not just by using an unboosted PI regimen (50% of those on PIs developed mutations compared with 20% on NNRTIs; p= 0.04) by also by the presence of archived mutations at baseline (p = 0.001). During questioning, Dr Palmisano admitted that since everyone was supposed to be on first-line therapy, they had no idea how this could have occurred.

Resistance was also a problem in the US quasi-CD4-guided treatment interruption study, ACTG 5170, presented by Daniel Skiest of Baystate Medical Centre, in Springfield, Massachusetts. This was a single-arm prospective study that mainly enrolled male patients in their early 40s who had begun therapy during the 'hit hard, hit early' treatment strategy of the late 1990s: median time on therapy was 4.5 years, median CD4 nadir was 436 cells/mm³ and more than 90% had over 500 CD4 cells (median 833 cells/mm³) at study entry.

Here, participants who wanted to take a treatment interruption stopped therapy and were followed for two years. HAART was restarted at the discretion of the individual's doctor although the study investigators "strongly recommended" restarting when CD4 counts fell below 250 cells/mm³.

There was an "even split" over participants taking NNRTI- and PI-based HAART, but more resistance was seen in those participants who interrupted NNRTI-based HAART. This is probably due to their ceasing the NNRTI component just two days prior to the NRTI component, which was the recommended method of ceasing NNRTIs at the time.

In contrast, participants stopping NNRTIs in the Window trial (which saw no difference in resistance mutations in either arm), did so with a seven day washout period for efavirenz. The lower number of mutations seen in participants on NNRTIs in the interruption arm in the PART trial may also have been due to the three day washout period for nevirapine and the six day washout period for efavirenz in that study.

Comparison within and between fixed-length treatment interruptions and within and between CD4-guided treatment interruptions are fraught with difficulties, since each study has different stopping and restarting criteria; has been carried out in vastly different populations; and has various, and often confusing, primary and secondary endpoints.

Although the DSMB of the Trivacan trial has allowed its fixed-length interruption arm to continue, and it appears likely that the fixed-length interruption arm of DART will continue, the risk and benefits of treatment interruptions in sub-Saharan Africa (where issues around costs of drugs and the practicalities and costs of frequent CD4 monitoring vie for importance with concerns over reducing long-term side-effects, and improving adherence and quality of life) may be assessed differently in the North.

Some previous studies of fixed-length treatment interruptions, using different study populations and interruption lengths, have been seen as risky. A two months on, one month off strategy from the United States' National Institute of Health (NIH) was terminated early when three participants developed drug resistance and failed to re-suppress virus. And the week on, week off arm of the Staccato study was terminated in 2003 due to excessive virologic failure in that arm.

However, the ANRS' Window study - two months on, two months off - and the Italian PART study - a complex design of three months on therapy, peppered with five interruptions of increasing length from one to three months, always followed by another three months on therapy - both presented upbeat conclusions in Denver.

Bruno Marchou of Purpan Hospital in Toulouse, France presented two-year Window study data on 391 treatment-experienced individuals (80% of whom were men, and the median time on prior HAART was five years) with a median CD4 count at baseline of 745 cells/mm³ and a viral load below 200 copies/ml who were randomised to continuous or interrupted treatment. To qualify for the study, participants' CD4 cell counts must never have fallen below 100 cells/mm³ (the median CD4 nadir was, in fact, 280 cells/mm³), and it excluded anyone taking abacavir (Ziagen) or nevirapine (Viramune, or those coinfected with hepatitis B virus who were also taking 3TC or tenofovir (Viread).

After six cycles of two months on, two months off therapy, using the same antiretroviral combination that had been interrupted (evenly split between efavirenz- or PI-based HAART) the study concluded that their version of structured treatment interruption was non-inferior to continuous therapy. The study's primary endpoint was defined as the number of participants whose CD4s dropped below 300 cells/mm³. However, the Window investigators designed the study so that a 70% difference between the arms would mean inferiority, and their intent-to-treat analysis found there was 'only' a 56% difference (3.6% vs. 1.5% dropped below 300 cells/mm³ in the interruption versus continuous treatment arms, respectively).

Dr Marchou did report, however, that those who interrupted therapy had a median loss of 155 CD4 cells by the end of the study, compared with no change in the continuous treatment arms, which was statistically significant. He also reported three cases of acute retroviral syndrome, and thrush was more common (10 vs. 6 cases, respectively) in the interruption versus continuous treatment arms.

Dr Marchou also reported that thrombocytopenia (low platelet count) - something also seen in the Staccato and ACTG 5170 trials - was a "real but infrequent concern" which developed in nine individuals in the interruption arm and two in the continuous treatment arm (although both of the individuals in the continuous treatment arm had interrupted their therapy at the time of diagnosis). This resulted in five of the nine in the interruption arm switching to a continuous regimen. Although two deaths (liver failure due to alcohol-related cirrhosis, and murder or suicide) also occurred in the interruption arm, they were not deemed to be related to either HIV or the study.

The surprising results of the SMART study have opened up a new paradigm of debate regarding the safely of structured treatment interruptions. Where previously there was concern that people with low nadir CD4s may be more prone to what is traditionally thought of as HIV-related disease progression, SMART appears to have unearthed the possibility that anyone who interrupts treatment - regardless of CD4 nadir - is at risk of what might be termed the 'new generation' of HIV-related disease progression: heart attack, stroke, coronary artery disease requiring surgery and kidney or liver disease. Although preliminary analysis suggests this may be linked to the most recent CD4 count prior to the event happening, it's too early to be certain this is the case.

This was uppermost on the minds of the six panellists who debated what SMART - and these other studies - might mean for the future of treatment interruptions. Professor Patrick Yeni, of Hôpital Bichat in Paris, France, argued that it was SMART's and Trivacan's low CD4 thresholds for restarting therapy - 250 cells/mm³ - that was "associated with an unacceptable high risk. CD4s need to be kept more or less in the normal range," he said.

When Dr Jintanat Ananworanich of HIV-NAT in Bangkok, Thailand, presented data from the 284 participants of the Staccato study that had CD4-guided interruption (broadly similar to the results presented in 2004 based on the first 74 participants), which found "minor manifestations of HIV infection" (including oral and vaginal thrush, low platelets and acute retroviral syndrome) but no AIDS-defining illnesses in the treatment interruption arm, she pointed out that Staccato's CD4 threshold for restarting therapy was 350 cells/mm³.

Dr Diane Havlir of the University of California-San Francisco, and principal investigator of ACTG 5170, agreed that Staccato "really limited the exposure to CD4 at low levels by keeping CD4 levels high with their 350 cell cut-off. I think CD4s are clearly the biggest predictor [of safety] - whether it's nadir or how long you're at a high level."

Dr Wafaa El-Sadr of Columbia University, New York, who presented the SMART study's findings, recommended caution on concluding that keeping CD4 counts above was 350 cells/mm³ is safe. "No other studies have shown that treatment interruptions are safe, even at higher CD4 counts," she argued. "To assume that interrupting treatment is safer at higher thresholds is maybe premature because most of the studies [suggesting it may be safe] have included limited numbers of participants."

Dr Havlir agreed that what makes SMART's findings so credible is the large number of participants. "What is surprising and intriguing are the cardiovascular events, which I think were really brought out with the randomised study," she said, "and makes one wonder about the findings we had in 5170."

Five deaths occurred in ACTG 5170, but with no control arm, the investigators initially reported that the deaths were unrelated to HIV or the study. However, at least three of the five deaths were related to cardiovascular disease and when Dr Skiest was summarising the study's findings, he said there was "some concern over the deaths [in the study], especially if they are cardiovascular disease-related: is there something going with viral replication and inflammation that could lead to increased coronary events?" he asked, before adding "but that is highly speculative."

The panel were equally perplexed on this point. Professor Bernard Hirschel of Geneva University Hospital in Switzerland, and a member of the Staccato Study Group, suggested "a positive spin: the SMART results reopen the window of opportunity to re-question the pathogenesis of certain [adverse events]. If you look at side-effects, and then you see more 'side-effects' in a group that has less drugs, then the whole question of what we are observing is up in the air," he told the audience. Could HIV itself explain at least some of the excess cardiovascular events that were reported in Denver by the D:A:D study.

Following the publicity surrounding the termination of the SMART study last month, New York-based treatment advocates, the Treatment Action Group (TAG), pointed out in a press release that "it is important to emphasize that the cessation of SMART does not necessarily mean that all treatment strategies involving interruptions of antiretroviral therapy are dangerous, just that the specific approach employed by the SMART study design was less successful at preventing clinical events than continuous treatment." TAG's Richard Jefferys told the Washington Post this week: "Our concern is that this will be considered the end of any studies of treatment strategies. We think that would be premature," he said. "The assumption that people are just going to be on therapy for the rest of their lives is not a practical assumption."

Dr Stefano Vella, president of the International AIDS Society, and one of the PART investigators, told the audience that based on the data he'd heard, treatment interruptions should now be considered "a harm-reduction strategy" rather than a treatment strategy.

Professor Hirschel added that "the issue of stopping therapy will not go away. There will always be people who want to stop therapy. If we take the SMART results at face value they say, 'once you start HAART you can never stop'. But what we really don't know is, when do we have to start HAART? And I think one of the consequences of what we've heard today is the renewed, or added, interest to set up a trial which compares early versus late antiretroviral therapy [in the treatment-naive]."

Dr Sharon Walmsley of the University of Toronto felt that we should be careful not to over-interpret the treatment interruption data heard today. "This does not in any way influence when we should start therapy," she argued. "Stopping therapy may not be the same thing as starting therapy. This should not alter the guidelines as when to initiate antiretroviral therapy."

All of the panellists agreed with the panel's chair, Dr Scott Hammer, of New York's Columbia University, who recommended: "Until these results are digested CD4-guided strategies should be reserved for clinical trials only."

Dr El-Sadr confirmed that follow-up in SMART will continue, and Dr Hammer reminded the audience that the analysis of results presented so far were done within a very short period of time. "We're very interested in seeing whether the increased risks associated with treatment interruptions can be reversed by going back on therapy longer-term," Dr el Sadr, said, adding "but these are very preliminary ideas."

They also agreed that further research needs to be done, although some favoured small-scale studies and others, "SMART 2 with higher CD4 thresholds."

Dr Havlir also felt that therapeutic vaccine studies that include treatment interruptions as part of the protocol ought to continue. "I don't think this should categorically halt all current studies looking at therapeutic vaccines because those are types of advances that really may help the field," she said.