When to start HIV treatment: cohort studies disagree on how early

Two major analyses of the risk of death or AIDS-related illness in people who started treatment at different CD4 counts have produced conflicting evidence about the benefit of starting treatment substantially earlier than current guidelines recommend, the Sixteenth Conference on Retroviruses and Opportunistic Infections heard on Monday.

A North American cohort study found that people who started treatment with a CD4 count below 500 cells/mm³ had a 60% higher risk of death than people who started treatment above this level, but a collaborative study by European and North American cohorts failed to find any extra benefit of starting above 400 cells/mm³. The second study found that while there was a clear benefit to starting treatment at a CD4 count in the range 350-450 cells/mm³ when compared with a lower count, starting treatment at CD4 count in the range 450-550 cells/mm³ did not provide a further benefit in terms of reducing the risk of AIDS or death.

Treatment guidelines in the developed world all concur that antiretroviral treatment should start around the time the CD4 cell count falls to 350 cells/mm³, although some groups of people, such as those with high viral load or with HIV/hepatitis C co-infection, should consider starting treatment before this point is reached.

However some physicians now believe that treatment should start considerably earlier, when the CD4 count is above 500 cells/mm³, in order to minimise the time during which individuals with HIV are immunosuppressed (the normal CD4 count in healthy adults is in the range of 700-1200 cells/mm³). Immunosuppression might increase the risk of some non-AIDS-defining cancers, which occur more frequently in people with HIV, and might also exacerbate the damage caused by hepatitis C infection, which appears to be more virulent and damaging to the liver in people with lower CD4 cell counts.

At last year’s ICAAC meeting US researchers presented an analysis of US and Canadian patient cohorts which showed a significant advantage to starting treatment before the CD4 count fell below 350 cells/mm³ (a 70% higher risk of death was identified in those who started treatment at a CD4 count below 350 cells/mm³ when compared to above 350 cells/mm³), but that study was not designed to determine whether starting treatment with a CD4 count above 500 cells/mm³ conferred additional benefit.

In order to answer that question Mari Kitahata and colleagues from the North American AIDS Cohort Collaboration looked again at the 9174-patient cohort, and identified 2620 patients who had started treatment at a CD4 count above 500 cells/mm³, and compared their risk of death to those who started treatment later.

Their observational study attempted to mimic the conditions of a randomised trial by only including people who had been diagnosed with HIV at a CD4 count above 500 cells/mm³, in order to take into account the potential biasing effect of missing people who might die with a CD4 count below 500 cells/mm³ before starting treatment. Some previous studies have only looked at people who actually started treatment, and would thus have missed those most likely to die.

They found a 60% higher risk of death for those who deferred treatment (relative hazard 1.60, p

After six years 10% of those who deferred treatment had died, and 15% by eight years, indicating that although the absolute risk of death was small, it was not negligible.

However, this study is still not able to answer the question of whether there is a significant difference between starting treatment in the CD4 count range of 350 to 500 cells/mm³, or starting at a CD4 count above 500 cells/mm³.

A second analysis, using data from 21,247 patients in seven cohorts, yielding 68,256 person-years of follow-up was carried out by the When to Start Consortium, led by Jonathan Sterne at Bristol University in the United Kingdom. That group compared the effects of deferring treatment across a range of CD4 cell bands below 550 cells/mm³, and found that while there was no significant difference in the risk of AIDS or death between those who started in the range 451-550 cells/mm³ and those who started in the range 351 to 450 cells/mm³ (HR 0.99, 95% CI 0.76 – 1.29), nor in the range 401-500 cells/mm³ compared to 301-400 cells/mm³ (HR 1.09, 95% CI 0.85 – 1.38), a significant difference was apparent in the range 351-450 cells/mm³ when compared to 251-350 cells/mm³ (HR 1.28, 95% CI 1.04 – 1.57).

This study also attempted to deal with the problem of deaths that might not be detected by a straightforward comparison of immediate versus deferred treatment, by estimating the rate of events seen at the CD4 counts being studied prior to the adoption of antiretroviral therapy, and then using the observed trends to fill in gaps where follow-up data were missing due to late diagnosis.

It should be noted that the two analyses are not strictly comparable. The North American analysis measured only the risk of death, while the When to Start Consortium measured the risk of death or AIDS. Neither analysis was able to capture the risk of non-AIDS-defining serious illnesses, such as non-AIDS-defining cancers or cardiovascular events, which were detected at a higher frequency in individuals with CD4 counts below 350 cells/mm³ off treatment in the SMART treatment interruption study. Determining the relative risk of these serious illnesses is likely to be a factor in measuring the risk/benefit of earlier treatment, and in particular in determining whether the number needed to treat in order to avoid one event is sufficient to be cost-effective.

The findings sparked a lively debate at the conference about the feasibility of conducting a large 'when to start' trial. In particular, there was concern that the observational analyses might not be able to capture the effect of confounding factors.

As Andrew Zolopa of Stanford University noted, referring to the North American analysis: “Patients who choose to defer [antiretroviral treatment] have different health-seeking behaviours than those who start treatment earlier, who probably wear seat-belts, don’t smoke, etc.”.

However, said Mari Kitahata, the North American cohort analyses had attempted to determine the size of unknown confounding effect that would be necessary in order to affect the result. Any confounding factor would have to reduce the risk of death fourfold to change the result of the analysis, she said – a far greater effect than detected in this study.

Professor Doug Richman of the University of California San Diego questioned whether a 'when to start' trial was worth the expense. “Rather than spend millions on a trial, given that most people aren’t diagnosed until much later, why not use all that money to identify people who have the higher risk?”, he asked.