When to start HAART – a key research question with the least available data

Nathen Geffen, TAC

With the international START study about to open this month, several studies at CROI contributed results confirming the importance of the ‘when-to-start’ discussions.

US and Canadian cohort analysis starting HAART at CD4 >500 cells/mm3

Mari Kitahata presented a study at CROI on behalf of the North American AIDS Cohort Collaboration on Research and Design. [1]

This collaboration includes 60 geographical regions in the US and Canada and records demographic and other data from multiple observational cohorts in the IDEA database. The study compared all-cause mortality in patients who initiated HAART with CD4 counts > 500 cells/mm3 with those who deferred treatment until their CD4 counts were below 500. This group has previously reported that initiating HAART at 350-500 has improved survival over starting when CD4 counts are below 350.

Kitahata explained that the study attempted to mimic clinical trial conditions as closely as possible. State-of-the-art statistical techniques were used. The methods account for time varying confounders that are measured but, as with other observational studies, do not deal with unmeasured confounders.

The study included all patients with CD4 counts > 500 actively followed up between 1996 and 2006, but excluded patients already on HAART and with prior AIDS-defining illnesses. Any statistically significant differences in baseline demographics, were not considered clinically relevant. Median baseline CD4 and viral load was around 660 and 3.6 log in both the immediate and deferred groups.

The analysis included over 9,000 patients and over 28,000 patient-years of follow-up. Just over 2,600 patients (8,900 PYFU started immediate treatment (or within 1.5 years of their first CD4 count since 1996) and the remainder deferred HAART. Approximately 2,600 patients in the deferred group still had CD4 counts above 500 when the study ended.

Mortality was consistently worse in the deferred group from 1997 with a relative hazard of 1.6 [95%CI 1.3-1.9; p<0.001] in the deferred group. Older age was a significant independent predictor of mortality, (RH 1.6 for every 10 years of age [95%CI 1.5-1.7; p<0.001]. Surprisingly, controlling for HCV co-infection of history of IDU made little difference (RH 1.5 95%CI 1.2-1.9; p<0.01, and RH 1.6, 95%CI 1.1-2.2; p<0.01, respectively. Sex, race, baseline CD4 and viral load were not predictors of mortality.

The finding that deferring HAART was associated with higher mortality was robust when controlling for sex, race, cohort, CD4, viral load, HCV, history of IDU and calendar year. A sensitivity analysis showed that the cumulative effect of confounding factors would have to have a combined relative hazard of 4.0 as well as four times greater odds of deferring treatment to only accomplish reducing the relative hazard of mortality in the deferred arm from 1.6 to 1.3.

The MACS, SWISS and CASCADE cohorts analysis

Jonathan Sterne presented the results of a study conducted by the When to Start Consortium of HIV Cohort Studies. [2]

They combined data from seven cohorts (including MACS, the Swiss HIV cohort and CASCADE) and compared patients starting HAART across several CD4 ranges to determine when to start. Their results are not entirely consistent with Kitahata et al. They too used state-of-the-art statistical techniques, but with different methodology, to analyse an even larger number of person-years.

Their data included over 21,000 patients comprising over 68,000 person-years of follow-up. There were 5,356 AIDS events and 3,630 deaths. Patients presumed to be infected by IDU were excluded.

They compared hazard ratios for AIDS or death in patients who started HAART in adjacent CD4 ranges from when they started treatment. However, a naive comparison of hazard ratios between two sets of patients in adjacent CD4 ranges would be incorrect because this would not take into account the time taken to move from a higher CD4 range to a lower one.

For example, a naive comparison of the time to AIDS or death in a patient who starts treatment at 350 cells/mm3 versus one who starts at 250 would be wrong, because it does not account for the time the patient starting with a CD4 count of 250 took to move from 350 to 250. A correct calculation must take this “lead time” (to use Sterne’s phrasing) into account.

Furthermore, unseen events, such as a patient with a CD4 count of between 251 and 350 who dies without using treatment also need to be taken into account. Sterne accounted for unseen events and lead times by imputing from data in the pre-HAART era to the deferred CD4 ranges.

Table 1: Hazard ratios by CD4 count

Lower CD4 range Higher adjacent CD4 range Hazard ratio of higher vs lower range, adjusted for lead time and unseen events
351-450 451-550 0.99 (95%CI 0.76-1.29)
276-375 376-475 1.19 (95%CI 0.96-1.47)
251-350 351-450 1.28 (995%CI 1.04-1.57) **
0-100 101-200 3.35 (CI 2.9999-3.75) **

** indicates a statistically significant hazard ratio.

Adjusting for age at initiation, sex and risk group did not materially alter the naive hazard ratios.

Deferring HAART until a CD4 count below 250 was clearly associated with increased risk AIDS or death (see Table1). They also showed, albeit less profoundly, that delaying treatment until CD4 was below 350 was associated with an increased risk of AIDS or death. In contrast to Kitahata et al, they did not conclude that HAART should be started in patients with CD4 counts higher than 450 cells/mm3.

Sterne explained that their study did not account for serious non-AIDS events. He emphasised that a clinical trial was needed to eliminate the effects of confounding such as, he suggested, people who defer HAART possibly having poorer adherence.


Details of the soon-to–enrol Strategic Timing of Antiretroviral Treatment (START) trial were presented at an INSIGHT meeting shortly before CROI. [4]

This study will randomise ART-naive patients with a CD4 count > 500 cells/mm3 to either immediate HAART or to defer treatment until CD4 drops to below 350 or an AIDS event.

The pilot phase will include 900 patients (450 to each arm) and roll out to 4,000 patients, 2,000 in each arm. Enrolment for this international study is due to start in March 2009 with about 70 sites in 23 countries.

The trial will compare the event rate in the deferred arm versus the immediate arm and also estimate the fraction of non-AIDS events. There will be four sub-studies. These will be in genomics, neurology, informed consent and arterial stiffness. The primary funder is the Division of AIDS of NIAID. Several other NIH institutions are funding the trial as well as ANRS (France), BMBF (Germany), NEAT (European Network AIDS Trials) and the NHMRC (Australia). Drug supply for the study is provided by pharmaceutical company support.


The evidence that patients with access to non-d4T-based regimens benefit from starting at CD4 counts closer to 350 than 200 cells/mm3 is now widely integrated into national and even WHO guidelines.

A mass of observational data and a sub-study of the SMART trial also support the potential benefits of starting even earlier, both for patient heallth and reduced transmission. [4, 5]


Unless stated otherwise, all references are to the Programme and Abstracts of the 16th Conference on Retroviruses and Opportunistic Infections. 8-11 February 2009, Montreal. Oral abstracts are available as a webcast.

  1. Kitahata M et al. Initiating rather than deferring HAART at a CD4+ count >500 cells/mm3 Is associated with improved survival. 16th CROI, Oral abstract 71.
  2.  Sterne J et al. When should HIV-1-infected persons initiate ART? Collaborative analysis of HIV cohort studies. 16th CROI, Oral abstract 72LB.
  3. INSIGHT research group. Strategic Timing of AntiRetroviral Treatment Study.
  4. Walensky R et al. When to Start ART—A Policy Evaluation While Awaiting Trial Results: South Africa. 16th CROI, Abstract 596b
  5. Granich RM et al. Universal voluntary HIV testing with immediate antiretroviral therapy as a strategy for elimination of HIV transmission: a mathematical model. Lancet. 2009 Jan 3;373(9657):48-57.
  6. Hammer S et al. Recommendations for the Use of Antiretroviral Therapy. JAMA 2008.

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