SMART Stops Stopping
By Richard Jefferys
The Strategies for the Management of Antiretroviral Therapy (SMART) trial was designed to investigate whether antiretroviral therapy could be used on an as-needed basis rather than continuously.
The protocol involved interrupting therapy when the CD4 count was confirmed to have crested 350 cells/mL and restarting therapy when it fell below 250 cells/mL. The impetus for the trial came from the well-documented association between CD4 count and risk of clinical disease, and from a number of studies suggesting that people who had begun therapy too early could safely interrupt treatment for prolonged periods without obvious harmful consequences.
SMART was sponsored by the Community Programs for Clinical Research on AIDS (CPCRA) and led by investigators Wafaa El Sadr and Fred Gordin. The late, desperately missed AIDS activist Carlton Hogan -- who taught many fellow activists the occult art of statistics -- was also part of the protocol team. In a remarkable testament to the engagement and commitment of those involved, SMART began in 2002 and enrolled 5,472 participants in 33 countries, making it the largest treatment strategy trial ever undertaken.
On January 10th of this year, the disappointing news emerged that SMART's Data and Safety Monitoring Board (DSMB) had recommended that the trial stop due to roughly twice the number of progression events in the intermittent therapy arm, including deaths and serious complications associated with drug toxicity. The lead investigators concurred with the DSMB and promptly advised that all study participants restart antiretroviral therapy. Though initial confusion abounded as to precisely why the study was stopped, it later became clear that the DSMB had met and requested additional data in November and made the recommendation to terminate the study upon review of the new information in January.
In order to grasp the rationale behind the DSMB decision, it's necessary to understand the intent of the SMART protocol design. The hypothesis was that people receiving intermittent treatment in what was called the drug conservation (DC) arm would fare slightly better over the long term than those receiving continuous therapy in the virological suppression arm due to a reduction in serious ARV-related complications. SMART was statistically equipped with 80% power to detect 20% superiority of the drug conservation arm, as determined by the trial's primary clinical endpoints: death, progression, and serious complications. The study protocol makes it very clear that the design team expected the incidence of toxicity-related events to be lower in the drug conservation arm, and that, if both progression and toxicity events favored one arm over the other, this would be cause for study re-evaluation and, potentially, cessation. As the preliminary data at CROI revealed in February, this is exactly what happened.
The SMART Data
At the time the study was stopped, 2,720 participants were enrolled in the DC arm and 2,752 in the VS arm. The average age was 46; about a quarter of the total enrollees were women and roughly a third were black. Participants had been receiving antiretroviral therapy for an average of six years prior to joining the study. Baseline characteristics were similar between the arms, as seen below.
Drug Conservation (DC)
Virological Suppression (VS)
Median baseline CD4 count
Median CD4 nadir
Percent with viral load <400 copies/mL
Prior clinical AIDS
Years of prior antiretroviral therapy
Median follow up was 10.1 months in the DC arm and 9.4 in the VS arm. Participants in the conservation arm spent just 33% of their time on antiretroviral therapy compared to 93% for people in the VS arm.
The DSMB decision to stop the study was driven by divergent outcomes between the two arms in terms of disease progression, death, and serious complications. There were 117 disease progression events in the DC arm (3.7%), versus 47 (1.5%) in the VS arm. Crucially, there were also 68 serious complications in the DC arm versus 46 in the VS arm (2.1 vs. 1.4%). Serious complications included cirrhosis of the liver, coronary artery disease requiring surgery, heart attack, stroke, and kidney failure. Although the difference between the arms in terms of these complications was of borderline statistical significance, the fact that this difference also favored the virological suppression arm suggested to the DSMB that it was highly unlikely that the excess progression events seen in the drug conservation arm would ultimately be compensated by a reduction in serious complications (as the SMART protocol had projected).
Prognosis of disease or death
Serious disease progression
While the majority of progression events were relatively clinically innocuous (e.g. oral thrush), a significant number of more serious opportunistic infections also occurred, including bacterial pneumonias and lymphoma.
In terms of mortality, the most significant difference between the arms was a disproportionate number of deaths for "Other" and "Unknown" reasons in the DC arm. More details surrounding these deaths will hopefully become available once the study results are published (expected soon in The Lancet). However, the fact that the precise cause of death is unclear for these two categories does not necessarily undermine the study results. Since the association is statistically significant, it is extremely unlikely that the increased risk of death seen in the DC arm resulted from chance. This point regarding "all cause mortality" was made with typical acuity by Carlton Hogan in an article he wrote for the GMHC Treatment Issues newsletter ("Death as an Endpoint," September 2001).
As Wafaa El Sadr highlighted in her CROI presentation, the VS arm performed superior to the DC arm for just about every other possible endpoint. The only findings favoring the DC arm -- beyond the reduction in antiretroviral usage and associated costs -- were self-reported improvements in body shape changes and reduced use of lipid-lowering drugs. SMART also had a number of sub-studies including a fully enrolled investigation into the quality of life of the participants, for which data have yet to be presented. Quality of life is a critical parameter that informs the real world decisions that people make about starting and stopping therapy, so these data will be extremely valuable.
Understanding the Results
The investigators were obviously dismayed by these results, although it is worth stressing that the vast majority of SMART participants (~95%) remained healthy regardless of the arm to which they were assigned. Efforts are now underway to try to understand the multiple factors that contributed to the study outcome. At CROI, investigators offered a glimpse at some of the work they have already completed with respect to this question.
The first and perhaps most obvious variable to consider is the lowest-ever CD4 count (CD4 nadir), but analyses did not reveal an association between this parameter and the events that occurred. Factors that did turn out to be associated with disease progression in the DC arm included the CD4 count immediately prior to the progression event (the proximal CD4 count), age (every 10 year increment in age was associated with an increased risk of progression) and a prior AIDS diagnosis. The same held true for the VS arm, which included the additional factor of viral load. Importantly, the investigators noted that these associations could not entirely explain the differential risk of disease progression between the two arms; even individuals with higher CD4 counts in the DC arm appeared more at risk (in the strata of 350-499 CD4s, for example, there were 28 events in the DC arm and 8 in the VS arm), suggesting that further analyses will be required to understand the SMART study outcome.
One strange but potentially critical finding which emerged from this work was that controlled viral load at study entry in the DC arm predicted an increased risk of disease progression and complications. Participants who entered the conservation arm with a viral load greater than 400 copies/mL did not experience more progression or complication events than those in the VS arm. The increased risk was confined to those who enrolled in the DC arm with viral loads <400 copies/mL and faced a nearly fourfold greater chance of an event than participants in the VS arm.
This finding left many people scratching their heads. The most widespread hypothesis -- albeit speculative -- is that the data may indicate that the inflammation that occurs as a result of viral load rebounds is not benign, as many people had surmised, but is in fact clinically very significant. In this scenario, the acute inflammatory response to interrupting antiretrovirals is more harmful than the low-level inflammation that accompanies persistently detectable viral load. It is also possible that the individuals with detectable viral loads had stronger HIV-specific immune responses, as Steve Deeks recently reported in his cohort of people who only partially control viral load on antiretroviral therapy (see Deeks et al, J Virol. 79;22:14169-78). Better HIV-specific immune responses could potentially serve to blunt the inflammation that accompanies a treatment interruption. Further, ongoing analyses of immune responses and known inflammatory markers may shed more light on these questions, but as it stands, it appears that the desperately unfashionable idea of using immune-based therapies (e.g. "immunosuppressant" drugs like cyclosporine and/or therapeutic HIV vaccines) in the context of treatment interruptions may be due for reevaluation.
Trials like SMART, which investigate treatment strategies using available drugs, tend to be the least sexy studies imaginable for the majority of HIV researchers (several of whom are known to have referred to it as "The Dumb Study"). Calls for a similar size study to investigate the question of when it is best to start highly active antiretroviral therapy have gone unheeded since 1996. In this context, it is important to stress that SMART was in no way a failure; it addressed a crucial "real world" question about the management of HIV disease and obtained an answer swiftly. Furthermore, the study represents a trove of data that will allow additional questions to be addressed. TAG and other community organizations have co-signed a statement from the Center for AIDS in Texas recommending that NIAID continue to support SMART as a large cohort study of treatment experienced individuals receiving antiretroviral therapy.
One prudent, immediate consequence of SMART's termination is that researchers are carefully evaluating ongoing studies of interrupted therapy. An international trial known as DART, which included an arm using fixed three-month treatment interruptions, recently announced that its DSMB recommended discontinuing this arm due to a higher incidence of clinical events compared to those seen during continuous therapy. A French study in Abidjan, Cote D'Ivoire, stopped its CD4-guided therapy arm a few months before SMART ended, also due to a higher rate of progression events. Conversely, follow up continues in the multi-national STACCATO trial, which includes a CD4-guided arm wherein therapy is reinitiated when the count dips below 350 cells/mL (as opposed to the 250 cells/mL cutoff used in the SMART and Abidjan trials). In an update on STACATTO at CROI, Jintanat Ananworanich noted that, based on the incidence rates seen in SMART, 17 events should occur in the CD4 guided arm of STACCATO. However, there have not been any AIDS-defining illnesses, and only a single death has occurred. In addition to the CD4 criteria, Ananworanich highlighted the fact that the average time spent on antiretrovirals differed between the studies, with STACCATO participants averaging just 15 months on therapy prior to enrollment.
The results of SMART and other relevant studies like STACCATO will need to be considered carefully when designing future clinical trials of intermittent therapy. Available data imply that higher, conservative CD4 count thresholds should be considered. The extensive treatment experience of the SMART study participants strongly suggests that any future exploration of therapy interruptions in a similar population should be undertaken with great caution. Conversely, it also suggests that it would be premature to generalize the SMART results to individuals who are less treatment experienced (a point supported by the STACCATO data). SMART clearly buttresses the rationale for studying whether immunological interventions can improve the safety and efficacy of intermittent antiretroviral therapy, so paradoxically, one valuable outcome of the study may be to reinvigorate this marginalized area of HIV research.