START analysis looks at who benefits most from immediate HIV treatment

Older people, those with high viral load or higher heart disease risk showed greatest benefit from immediate treatment
Jean-Michel Molina presenting at AIDS 2016. Photo by Liz Highleyman,
This article is more than 8 years old. Click here for more recent articles on this topic

Starting antiretroviral therapy (ART) soon after HIV diagnosis led to better outcomes than delayed treatment in all population sub-groups in the START trial, researchers reported at the 21st International AIDS Conference (AIDS 2016) last week in Durban, South Africa. But some people saw greater risk reductions, including those over age 50, those with a lower CD4:CD8 ratio and higher viral load, and those with cardiovascular risk factors.

Evidence has accumulated in recent years showing that starting HIV treatment promptly, rather than waiting for the CD4 T-cell count to fall below a certain threshold, is associated with reduced disease progression and death. Some of the most definitive evidence comes from the large START trial, which randomly assigned people with high CD4 counts to either immediate or deferred therapy.

At last year's International AIDS Society (IAS) Conference and in the 27 August, 2015, edition of the New England Journal of Medicine, START investigators reported that people who initiated treatment immediately after diagnosis had a significantly lower risk of illness and death than those who waited until their CD4 count fell below 350 cells/mm3.


absolute risk

The chance that a person will experience a specific event during a period of time. It is always between 0 and 1 (when expressed as a probability), or between 0 and 100 (when expressed as a percentage).


A molecule on the surface of some white blood cells. Some of these cells can kill other cells that are infected with foreign organisms.

middle income countries

The World Bank classifies countries according to their income: low, lower-middle, upper-middle and high. There are around 50 lower-middle income countries (mostly in Africa and Asia) and around 60 upper-middle income countries (in Africa, Eastern Europe, Asia, Latin America and the Caribbean).

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’.

number needed to treat

The average number of patients who need to be given a treatment in order to prevent one patient from having a bad outcome (e.g. AIDS or death).

At AIDS 2016 Jean-Michel Molina of the University of Paris Diderot presented further findings from the trial looking at benefits of early treatment in different subgroups based on demographics, disease status and other risk factors.

START enrolled 4685 previously untreated HIV-positive adults with CD4 counts above 500 cells/mm3; the median was about 650, but 11% had more than 900. Nearly three-quarters were men and the average age was 36 years. Participants came from 35 countries, including about half from low- and middle-income countries; about 45% were white, 30% were black, 14% were Hispanic/Latino and 8% were Asian. Median pre-treatment viral load was approximately 12,800 copies/ml, though 10% had over 100,000 copies/ml. About 6% had hepatitis B or C co-infection, 2% had high Framingham cardiovascular risk and nearly a third were smokers.

The investigators looked at a combined primary endpoint of serious AIDS events, serious non-AIDS events or death in pre-specified population subgroups (except for the CD4:CD8 ratio analysis, which was added later). They calculated event rates per 100 person-years (PY), absolute risk reduction (ARR) associated with immediate treatment and the number of people who needed to start immediate ART in a year to prevent a single event (number needed to treat, or NNT).

Overall, there were 42 of these events (0.60 per 100 PY) among people randomised to immediate treatment compared to 96 events (1.38 per 100 PY) in the deferred therapy group – a hazard ratio of 0.43, or 57% lower risk. Absolute risk reduction worked out to be 0.78 per 100 PY and the number needed to treat was 128.

Looking at demographic subgroups, all groups showed a relative risk reduction, or proportional change in risk with immediate ART. Hazard ratios were generally similar, ranging from 0.18 to 0.85, indicating that everyone benefitted from early treatment. But absolute risk reduction, considering actual numbers of events, revealed greater differences across groups.

Broken down by age, absolute risk reduction ranged from 0.49 for those under 30 years, to 0.66 for those age 30-49, to 2.24 for those age 50 and older. While 206 people under 30 and 151 people age 30 to 49 would need to be treated to prevent one event, this fell to just 45 for the over-50s – a statistically significant difference (p = 0.01).

Men and women had an absolute risk reduction of 0.74 and 0.93, respectively, and a number needed to treat of 136 and 108, respectively – not a significant difference (p = 0.63). Risk reduction was also statistically similar for racial/ethnic groups (ARR 0.70 for black people and 0.92 for white people; NTT 142 and 109, respectively, p = 0.69) and for high-income and low/middle-income countries (ARR 0.86 vs 0.69; NTT 116 vs 144, p = 0.62).

Analysing by baseline disease status, people with a CD4 count < 600 cells/mm3 had an absolute risk reduction of 1.10 compared to 0.68 for those with 600-800 and 0.51 for those with > 800. Numbers needed to treat were 91, 147 and 196, respectively, which was not a significant difference (p = 0.41). Categorising by CD4:CD8 ratio, however, did show a highly significant difference: absolute risk reduction was 1.67 for those with a ratio < 0.5, 0.47 for those with 0.5-0.8 and .040 for those with > 0.8 (NTT 60, 214 and 248, respectively, p = 0.005).

Looking at baseline viral load, absolute risk reduction was just 0.10 for those with < 3000 copies/ml, rising to 0.82 for those with 3000-50,000 copies/ml and 1.48 for those with > 50,000 copies/ml. Numbers needed to treat were 992, 122 and 67, respectively – the largest numerical difference, but just short of statistical significance (p = 0.06).

Finally, people with a low Framingham 10-year risk of heart disease had an absolute risk reduction of 0.36 compared to 0.90 for those with medium risk and 1.45 for those with high risk. Numbers needed to treat were 276, 111 and 69, which again did not reach the threshold for significance (p = 0.13).

"In asymptomatic ART-naive adults with >500 CD4+ cells/mm3, immediate ART was superior to deferred ART across all sub-groups with similar relative risk reduction," the researchers concluded, noting that higher absolute risk reduction and lower NNT were found for participants older than 50 years, with HIV RNA levels > 50,000 copies/ml, with CD4:CD8 ratios < 0.5 and with Framingham 10-year risk score > 10%.

The investigators did not analyse subgroups that had fewer than ten events in both study arms combined, but nevertheless cautioned that "NNT estimates can be unstable when event rates are small, and need to be interpreted with caution."

Based on these findings, they suggested that people in the groups that saw greater absolute risk reductions "might be prioritised for immediate access to ART" in settings where universal prompt treatment for everyone diagnosed with HIV is not yet feasible.

They emphasised that immediate treatment also reduces HIV transmission risk, which should be considered along with the number needed to treat when assessing the cost effectiveness of ART.

Prof Molina said that these results are from univariate analyses and the team is currently working on a multivariate analysis to shed light on how these factors interact with each other. They hope to use this data to develop a risk scoring system.


Molina J-M et al. Who benefited most from immediate treatment in START? A subgroup analysis. 21st International AIDS Conference, Durban, abstract THAB0201, 2016.

View the abstract on the conference website.

Download the presentation slides from the conference website.

Watch the webcast of this presentation on YouTube.