HIV viraemia may explain increased risk of cardiovascular disease, death and other serious events in people interrupting treatment in the SMART trial: new study to randomise people with CD4 counts >500 to start immediate treatment or defer to <350 cells/mm3

Simon Collins, HIV i-Base

Several presentations at the meeting included new results from the SMART study. This included higher rates of bacterial pneumonia, and higher risks and events for patients, coinfected with hepatitis B or C, who interrupted treatment.

Early results relating to the hypothesis that HIV viraemia and inflammatory responses could explain the increase risk of cardiovascular events seen in the treatment interruption arm, were also presented.

While this large randomised study unexpectedly found an early difference in favour of continuous treatment that resulted in early termination of the study, there is now a considerable bank of samples that may explain the unexpected results.

In a symposium looking at the benefits and risks from earlier treatment (ie starting at a higher CD4 count) [1], Jim Neaton, from the INSIGHT trial network responsible for the SMART trial, presented results from retrospective analysis of inflammatory markers (hs C-reactive protein (hs-CRP), IL-6, Serum amyloid A, Serum amyloid P) and coagulation markers (D-dimer, PA1-1, Prothrombin fragment 1+2 [F1.2]) from patients enrolled in SMART. [2]

The first results presented, on D-dimer, were some of the most interesting tentative new data shown at the IAS meeting.

D-dimer is a coagulation marker that has been linked to cardiovascular disease in the general population and that has been previously reported to be higher in HIV-positive patients and linked to more advanced HIV disease. In this analysis they similarly found that baseline D-dimer levels were higher in SMART patients compared to the general population: 34% patients had levels about the upper limit of the normal range (0 – 0.3 or 0.5 ug/mL). Secondly, levels were significantly higher in patients not on treatment at baseline compared to those already on treatment: mean 0.69 ug/mL (+ 0.95) vs 0.48 ug/mL (+ 0.78), p=0.02.

The link to HIV viraemia was strengthened when looking at the changes in D-dimer between week 0 to week 4, by baseline use of treatment and subsequent randomisation. Patients not on HAART at baseline, randomised to the continuous treatment arm, had a mean drop of –0.22 ug/mL, compared to little change if they remained off treatment (p=0.052). Patients who were receiving HAART at baseline, and who were randomised to interrupt treatment, saw D-dimer increase by a mean of +0.10, compared to -0.03 reduction in those who stayed on treatment (p<0.001). When the results of patients on HAART at baseline were analysed by whether their viral load was <400 copies/mL, the results were even more pronounced: D-dimer increased by +0.13 uh/mL in patients randomised to interrupt treatment (p<0.001) and was stable in all others.

Dr Neaton then reported that there was a strong positive correlation between change in D-dimer and viral load: with increases in D-dimer of 0.0, +0.04, +0.11 and +0.28 ug/mL in patients whose viral load at week 4 was <400 c/mL, 400-10,000 c/mL, 10,000-50,000 c/mL and >50,000 c/mL respectively (p=0.0005 for trend).

Finally, to look at the clinical implications of these findings, the researchers looked at baseline D-dimer levels in a nested case control study of 85 deaths and 170 case controls from SMART without cardiovascular disease (mean 1.07 vs 0.37 ug/mL respectively, p=0.0001), and at the odds ratio of death by D-dimer quartile.

Patients with baseline levels in the upper quartile (>0.63 ug/mL) were at 13 times the risk of death compared to those with baseline levels in the lower quartile (<0.18 ug/mL), OR 13.37, p<0.0001. The trend for increased risk correlating with increased D-dimer levels was also highly statistically significant for the trend (p<0.0001).

The adjusted odds ratio associated with a 0.15 ug/mL increase was significant when looking at all deaths, major cardiovascular events and AIDS events, see Table 1.

Table 1: Adjusted odds ratios associated with a 0.15 µg/mL increase in D-dimer

Event No. of events ADj. OR 95% CI p-value
All deaths 74 1.23 1.07-1.42 0.004
Major CVD 59 1.12 1.01-1.24 0.04
AIDS 75 1.40 1.19-1.66 0.0001

The presentation concluded that these biomarker data provided additional evidence for looking at benefits of earlier treatment in the proposed START trial.

The rationale behind this new study continued in the following presentation by Fred Gordin, from VA Medical Centre and George Washington University. [2]

Gordin outlined some of the accumulating data that report that HIV is not clinically latent at CD4 levels above current guidelines for initiating therapy. Both the UK-CHIC and CASCADE cohorts show increasing event rates of AIDS events and death as CD4 counts drop to 500-650 cells/mm3, 350 – 500 cells/mm3 and 200-350 cells/mm3, and that risk is higher in treatment naive patients in each strata compared to those who have started treatment. Numerous studies have also reported significantly higher rates of non-AIDS cancers compared to HIV-negative population.

The SMART study was closed early because both AIDS rates ‘and’ serious non-AIDS events (cardiovascular, renal, hepatic) that prior to the trial were assumed to be related to HAART-related toxicity – occurred at higher rates in the treatment interruption arm.

An analysis of the subset of approximately 500 patients in the SMART study who were treatment naive or on a treatment interruption at baseline also provided a source of looking at patients randomised to starting at earlier vs deferred treatment.

All serious endpoints occurred at a significantly higher rate in the deferred (treatment interruption) arm compared to patients who started treatment and remained on continuous treatment (early treatment), see Table 2.

Table 2: Event rate in patients not on treatment at randomisation to SMART trial

Event Treatment interruption (deferred Tx) Continuous treatment (early Rx) HR deferred vs early [95% CI] p-value
N rate N rate
OD or death 15 4.8 4 1.1 4.4 [1.5, 13.2] 0.009
OD fatal or non-fatal 11 3.5 3 0.8 4.4 [1.2, 15.8] 0.02
Serious non-AIDS 12 3.9 2 0.5 7.1 [1.6, 31.5] 0.01
Composite 21* 7.0 5 1.3 5.1 [1.9, 13.5] 0.001

Gordin then outlined the proposal for the START study (Strategic Timing of AntiRetroviral Treatment), an international study that recently received US funding for a pilot phase and that is based on the SMART network of trial sites. 1200 HIV-positive patients with CD4 counts >500 copies/mL will be randomised to either immediate treatment or deferred treatment defined as CD4 count <350 copies/mL. The primary endpoint is serious AIDS or non-AIDS events or death. The full study is estimated to need and additional 1800 patients but this will be dependent on results of the pilot phase.

For the study to enroll and run successfully, side effects will need to be proactively managed and adherence stressed and supported – if patients are not to develop early resistance and reduced treatment options – both of which are secondary endpoints in the study. However, several studies have also shown that side effects occur less frequently when treatment is started earlier, and lower baseline viral load may result in higher rates of viral suppression that is seen when treatment is started later.

The session prompted a long discussion (also available on the webcast) that addressed other aspects of the symposium: choice as a research priority; funding and cost of earlier treatment; current issue of late presentation in most settings (usually with CD4 count already <350 cells/mm3); and potential disparity between treatment access and guidelines in resource-rich and resource-poor settings should earlier treatment prove beneficial.

While many of these practical points are important, they do not detract from the scientific question of when to start treatment, that has still never been addressed in a large randomised trial.


The START trial is a critical study that has the potential to change not only clinical management and challenge many current assumptions about risks and benefits or treatment as significantly as the SMART study last year.

With millions of patients now accessing ARV treatment, following 3,000 patients closely in a trial, half of whom start treatment a few years earlier than they would based on current guidelines, has the potential to not only reduce the current difference between life expectancy for an HIV-positive person (with access to HAART) and that of an HIV-negative person still further, but also to drive less tangible social factors including more positive approaches to earlier testing.

The pilot phase of the START trial is due to open in the US in the next few month, and international sites, including the UK will follow, dependent on national and local ethics approval. [4]


  1. Treatment of early HIV disease. 4th IAS conference, Sydney, 2007. Symposium MOSY2. Available online with link to webcast:
  2. Neaton J. Changing patterns of morbidity and mortality in HIV disease. 4th IAS conference, Sydney, 2007. Abstract MOSY202. See above link to symposium MOSY2.
  3. Gordin F. Is an early treatment study important now? 4th IAS conference, Sydney, 2007. Abstract MOSY205. See above link to symposium MOSY2.

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