New data shed light on poor outcome in SMART study

This article is more than 16 years old. Click here for more recent articles on this topic

In the recently discontinued SMART study, opportunistic illnesses and death occurred more often in participants with higher CD4 cell counts and lower viral loads, according to data presented at the Sixteenth International AIDS Conference in Toronto on August 16th. But these factors could only explain part of the unexpectedly poor outcomes in the treatment interruption arm.

The SMART study included 5,472 participants in more than 30 countries, 95% of whom had prior treatment experience. At enrollment, patients were required to have a CD4 cell count above 350 cells/mm3. At baseline, the average age was 44 years, about three-quarters were men, and 29% were black. The median CD4 cell count at study entry was 598 cells/mm3, the median nadir (lowest ever) CD4 cell count was 251 cells/mm3, and 71% had viral load below 400 copies/ml.

Participants were randomly assigned to either the “drug conservation” arm (interrupting antiretroviral therapy when CD4 cell counts reached 350 cells/mm3 and resuming therapy when they fell to 250 cells/mm3) or the “viral suppression arm” (remaining on continuous treatment with the aim of maximum suppression of viral load).


treatment interruption

Taking a planned break from HIV treatment, sometimes known as a ‘drugs holiday’. As this has been shown to lead to worse outcomes, treatment interruptions are not recommended. 

opportunistic infection (OI)

An infection that occurs more frequently or is more severe in people with weakened immune systems, such as people with low CD4 counts, than in people with healthy immune systems. Opportunistic infections common in people with advanced HIV disease include Pneumocystis jiroveci pneumonia; Kaposi sarcoma; cryptosporidiosis; histoplasmosis; other parasitic, viral, and fungal infections; and some types of cancer. 

hazard ratio

Comparing one group with another, expresses differences in the risk of something happening. A hazard ratio above 1 means the risk is higher in the group of interest; a hazard ratio below 1 means the risk is lower. Similar to ‘relative risk’.


Expresses the risk that, during one very short moment in time, a person will experience an event, given that they have not already done so.


Lowest of a series of measurements. For example, an individual’s CD4 nadir is their lowest ever measured CD4 count.

As reported at the Conference on Retroviruses and Opportunistic Illnesses in February, the study was terminated after it became clear that patients in the treatment interruption arm were 2.6 times more likely to experience HIV disease progression or death.

Dr Jens Lundgren presented the results of an analysis exploring whether the higher risk of fatal and non-fatal opportunistic infections and non-opportunistic infections-related death in the treatment interruption arm could be explained by so-called “proximal” CD4 cell counts and viral load levels measured closest to the onset of illness or the time of death.

During the follow-up period, there were 95 total fatal or non-fatal opportunistic infections and 72 total deaths. In the treatment interruption arm, there were 120 opportunistic infections or deaths, compared with 47 in the continuous therapy arm (3.3 vs 1.3 per 100 person years, respectively).

The median proximal CD4 cell count was 343 cells/mm3 in the treatment interruption arm compared with 540 cells/mm3 in the continuous therapy arm. Whilst, as expected, the rate of opportunistic infections or death was higher for participants with lower CD4 cell counts and higher viral loads in both arms, the risk of adverse outcomes was comparatively greater in the interruption arm for patients with higher CD4 cell counts. Stated another way, the risk of opportunistic infections or death in the two arms was similar among patients with CD4 counts below 350 cells/mm3, but significantly greater in the treatment interruption arm among patients with CD4 counts of 350 cells/mm3 or higher (p

SMART was intended to look at outcomes in patients who spent more total time at lower CD4 cell counts and higher viral loads due to periodic treatment breaks. Participants in the treatment interruption arm spent a total of 32% of follow-up time at CD4 cell counts below 350 cells/mm3, compared with just 7% for the continuous therapy arm; patients in both arms spent very little time at CD4 cell counts below 200 cells/mm3

While lower CD4 cell counts in the treatment interruption arm accounted for some of the observed difference in the rates of opportunistic infections and death in the two arms, CD4 cell levels could only provide a “partial explanation,” the researchers said. Some “residual excess risk” was noted in the treatment interruption arm among patients with proximal CD4 cell counts of 350 cells/mm3 or higher. In this subgroup, the overall median viral load was 10,000 copies/ml, compared with less than 400 copies/ml in the continuous therapy patients with similar CD4 cell counts. The researchers suggested that this excess risk in the treatment interruption arm implies there is some “impairment of immune function not reflected in peripheral blood CD4 count.”

Until this “missing link” is identified, Dr Lundgren recommended that interruption of antiretroviral therapy should not be undertaken except in randomised clinical trials with adequate statistical power to assess the risks and benefits of this strategy.

Risk factors for opportunistic infections and death

A second study, presented by Dr Wafa El-Sadr, looked at risk factors for the poor outcomes in the SMART treatment interrupt arm, analysing hazard ratios (comparative rates of adverse events) in various subgroups.

The overall hazard ratio was 2.6, meaning participants in the treatment interruption arm were 2.6 times more likely to experience opportunistic infections or death.

Whilst the proximal CD4 cell measurement closest to the onset of illness or time of death was correlated with adverse events, as described above, nadir CD4 cell counts was not a significant predictor of outcome, in contrast with some past research. Baseline viral load did have a significant effect. Among patients with HIV RNA levels of 400 copies/ml or less, the rate of opportunistic infections or death was 3.2 in the treatment interruption arm, compared with 0.8 in the continuous therapy arm (p

Hazard ratios were similar across a wide range of subgroups, and there was no significant effect related to age, sex, injecting drug use history, sexual risk category (e.g. men who have sex with men), prior AIDS diagnosis, CD4 percentage, hepatitis B or C co-infection, or type or duration of antiretroviral therapy.

However, the researchers did see an effect of race/ethnicity. The hazard ratio was somewhat greater for blacks compared with white or Latino patients, and this was largely driven by a higher rate of non-opportunistic infection-related deaths.

Dr El-Sadr emphasised that in every subgroup examined, the rate of adverse outcomes was lower in the continuous therapy arm; no subgroup appeared to benefit from treatment interruption. However, some subgroups had “particularly inferior outcomes” using episodic treatment. In general, patients who were doing best on potent HIV therapy before interrupting therapy, with lower viral loads and higher CD4 cell counts, experienced the most inferior outcomes using the treatment interruption strategy.

Dr El-Sadr agreed with Dr Lundgren that it was “both smart and wise” to avoid episodic antiretroviral therapy until more can be learned about the reasons for the poor outcomes in the SMART study.


Lundgren JD et al. Progression of HIV-related disease or death (POD) in the randomised SMART study: why was the risk of POD greater in the CD4-guided ((re)-initiate ART at CD4 Sixteenth International AIDS Conference, Toronto, abstract WEAB0203, 2006

El-Sadr W et al. Inferior clinical outcomes with episodic CD4-guided antiretroviral therapy aimed at drug conservation in SMART study: consistency of finding in all patient groups. Sixteenth International AIDS Conference, Toronto, abstract WEAB0204, 2006