Starting HIV treatment at high CD4 counts protects against both AIDS and non-AIDS events: overall and in subgroup analyses of START study

Simon Collins, HIV i-Base

IAS 2015 logo - top onlyThe international START study produced headline news at IAS 2015 that confirms benefits from starting antiretroviral treatment (ART) at CD4 counts above 500 cells/mm3.

The study also reported no upper CD4 threshold that was protective against AIDS-related events, even though the overall absolute risk of events in START was low.

START presentation by Jens Lundgren

Professor Jens Lundgren presenting ground-breaking results from the START study, with other key members of the study team, Abdel Babiker and Jim Neaton at lower right.


The START study is notable for reporting results 18 months ahead of schedule, following a recommendation by the study’s independent Data and Safety Monitoring Board (DSMB) in May that participants in the arm deferring ART until their CD4 count reached 350 cells/mm3 should be offered immediate treatment, and that follow-up should continue as planned for both arms.

The results were presented by Professor Jens Lundgren from University of Copenhagen on behalf of the START study team in two sessions at the conference: an opening plenary on the first day and the International AIDS Society members meeting later in the programme. [1, 2] The study was also simultaneously published online in the New England Journal of Medicine. [3]

This HTB report combines results from both IAS 2015 and NEJM paper.

Although preliminary findings were released on 27 May 2015 based on the dataset used for the DSMB decision, the expanded results cover three key areas:

  • The additional endpoints in the final dataset modifies the primary endpoint results by finding that the reduction in non-AIDS events now reaches statistical significance in favouring early ART. Back in May, this endpoint fell short of significance.
  • New details about the AIDS and non-AIDS events seen in each arm and still finding that AIDS events drove the primary results and occurred at high CD4 counts.
  • In subgroup analyses, early ART was consistently protective for all key baseline and demographic subgroups. It is important to stress that benefits from the study were not just for people with the highest risks.

Methods and baseline characteristics

From December 2009 to December 2013, START randomised 4685 HIV positive treatment-naive adults with CD4 counts >500 cells/mm3 to either an immediate (IMM) or deferred (DEF) ART, with the deferred group waiteing until their CD4 count reached 350 cells/mm3.

The combined primary endpoint included AIDS related and non-AIDS related complications including grade 4 events and deaths from any cause.

The study included 215 sites in 35 countries, equally divided between high and low/middle income countries. Baseline demographics have already been widely reported and published online [4, 5] and included approximately 27% women, 55% MSM, with a median age 36 years (IQR 29 to 44). Median CD4 and viral load were 651 cells/mm3 (IQR 584 to 765) and 12,700 copies/mL (IQR 3,000 to 43,000), respectively, with no significant differences between groups. At study entry, median time since HIV diagnosis was 1.0 years (IQR: 0.4 to 3.1).

Primary and key secondary endpoint results

Mean follow-up was 3.0 years (median 2.8; IQR 2.1 to 3.9) with 23% having greater than 4 years follow up. Endpoint results were available for 96% and 95% of the IMM and DEF groups respectively.

By 26 May 2015, 98% vs 48% of the IMM vs DEF participants had started ART (at median CD4 count of 651 vs 408 cells/mm3 respectively). Although ART in the study was provided free from a central repository, and included the choice of all or nearly all approved drugs, the majority of patients in both arms used tenofovir/FTC as background NRTIs (approximately 90%). Efavirenz was the most widely used third drug, by 73% and 51% of the IMM vs DEF arms respectively, with atazanavir/r, darunavir/r, rilpivirine and raltegravir making up the majority of other combinations. The study reported high rates of viral suppression with 98% vs 97% of those on treatment having <200 copies/mL at month 12.

The final dataset included a total of 140 primary endpoint events: 42 (1.8%) in the IMM arm vs 96 (4.1%) in the DEF arm, equivalent to rates of 0.60 vs 1.38 per 100 patient years, respectively. The hazard ratio (HR) for the composite primary endpoint was 0.43 (95% CI: 0.30 to 0.62), significantly in favour of the IMM group, p<0.001. Hazard ratios for other key secondary endpoints also significantly favoured the IMM group: 0.28 (95%CI: 0.15 to 0.50) for serious AIDS-related events (p<0.001) and 0.61 (95%CI: 0.38 to 0.97, p=0.04) for serious non-AIDS-related. There was no significant difference between groups for all cause mortality: HR 0.58 (95%CI: 0.28-1.17, p=0.13). See Table 1.

Table 1: Hazard ratio (HR) of primary and key secondary endpoints
IMM (N = 2326) DEF (N = 2359) HR (95% CI) P value
no. rate/100 PY no. rate/100 PY
Composite primary endpoint 42 0.60 96 1.38 0.43 (0.30 to 0.62) <0.001
Secondary end points
Serious AIDS events 14 0.20 50 0.72 0.28 (0.15 to 0.50) <0.001
Serious non-AIDS events 29 0.42 47 0.67 0.61 (0.38 to 0.97) 0.04
Death from any cause 12 0.17 21 0.30 0.58 (0.28 to 1.17) 0.13 NS

Key: IMM=immediate arm. DEF=deferred arm, HR=Hazard Ratio, NS=Non Significant.

Clinical endpoints

An unexpected outcome in START is the degree to which AIDS events were more common at high CD4 counts than non-AIDS events. Throughout the study, the greatest impact of early ART was expected to be reduced inflammation-related events. Also, consistent with the planned study design, only 4% of follow-up time in the deferred arm occurred at a CD4 count <350 cells/mm3 and accounted for only 5 primary events.

The most common events were cardiovascular disease (29% vs 15%), non-AIDS cancers (21% vs 19%) and tuberculosis (14% vs 20%) in the IMM vs DEF groups respectively.

Endpoints that were significantly reduced in the IMM group included TB (HR 0.29; 95%CI: 0.12 to 0.73), p=0.008) and Kaposi’s Sarcoma (HR 0.09; 95%CI: (0.01 to 0.71, p=0.02) but not malignant lymphoma (p=0.07), non-AIDS cancers (p=0.09), cardiovascular disease (p=0.65), Grade 4 events (p=0.97), unscheduled hospitalisation (p=0.28) and combined Grade 4 event, unscheduled hospitalisation, or death from any cause (p=0.25). See Table 2.

Table 2: Other important clinical secondary endpoints
IMM (N = 2326) DEF (N = 2359) HR (95% CI) P value
no. rate/100 PY no. rate/100 PY
Tuberculosis 6 0.09 20 0.28 0.29 (0.12 to 0.73) 0.008
Kaposi’s sarcoma 1 0.01 11 0.16 0.09 (0.01 to 0.71) 0.02
Malignant lymphoma 3 0.04 10 0.14 0.30 (0.08 to 1.10) 0.07
Non-AIDS cancers 9 0.13 18 0.26 0.50 (0.22 to 1.11) 0.09
Cardiovascular disease 12 0.17 14 0.20 0.84 (0.39 to 1.81) 0.65
Grade 4 events 73 1.06 73 1.05 1.01 (0.73 to 1.39) 0.97
Unscheduled hospitalisation § 262 4.02 287 4.40 0.91 (0.77 to 1.08) 0.28
Grade 4 event, unscheduled hospitalisation, or death from any cause 283 4.36 311 4.78 0.91 (0.77 to 1.07) 0.25

§ This category excludes hospitalisations for AIDS-related illnesses.

Clinical events by geographical region

Earlier treatment had better outcomes in both high and low/middle income countries, although there were differences in the type of events by geographical region. Most of the TB cases (16/20) were in Africa and most of the cancers (22/27) and cardiovascular events (19/26) occurred in Australia, Europe, Israel and the US.

Subgroup analysis consistently support earlier treatment

Another unexpected outcome from START was the consistency for the primary endpoint results in sub-group analysis for baseline demographics and other risk factors of serious events, all favouring the early treatment arm.

The expectation that events would only occur in the groups at highest risk and that lowest risk groups would be protected from events was not supported by the results.

This included analyses by age, sex, race, geographic regions, smoking status, cardiovascular risk or baseline CD4 and viral load. Even when 95%CI for the HR crossed 1.0 for several parameters (highest CD4 and CHD risk and lowest VL), none of the p-values for the interaction approached significance. See Table 3.

Table 3: Hazard rates (HR) for primary end point by subgroup

HR in favour of early ART p for interaction
Age 0.98
<35 0.47
>35 0.42
Sex 0.38
Male 0.47
Female 0.31
Race 0.65
Black * 0.57
White 0.40
Other 0.37
Region 0.55
High income 0.39
Low/middle income 0.48
Baseline CD4 0.71
< 600 0.28
600-800 0.50
> 800 * 0.56
Baseline viral load 0.25
< 5,000 * 0.66
5,000-30,000 0.38
>30,000 0.37
Smoker 0.93
Yes 0.43
No 0.44
10 year CHD Framingham risk 0.56
<0.8 * 0.46
0.8 to 3.6 0.39
> 3.6 0.50

* Although the individual 95% CI crossed 1.0 for these categories the p-value for trend for the subgroup was not statistically significant.

In conclusion, START results support the importance of providing ART for all HIV positive people.

They reinforce the need for further research to understand the pathogenesis of HIV in early infection and the urgency of achieving funding to ensure universal ART becomes a global reality.

Simon Collins is a member of the Community Advisory Board (CAB) for the START study.


START has produced a dataset that defines level of risk for not using ART, irrespective of any individual decision to start treatment. The level of confidence for doctors recommending early ART will now increase, even if the scale up issues for universal treatment when global coverage has not yet met this based on CD4 thresholds of 500 or even 350 cells/mm3.

Earlier treatment was already recommended in guidelines due to the impact that ART has on dramatically reducing the risk of onward transmission. However, START demonstrates the key missing evidence that this approach also produces clinical benefits for the person taking treatment.

These results are expected to change treatment guidelines globally. UK BHIVA guidelines have removed CD4 threshold as a criteria for starting ART in the 2015 draft. [6]

At IAS 2015, WHO announced that the updated 2015 guidelines will also recommend treatment for all HIV positive people. [7, 8]

Follow up for all participants continues until at least the end of 2016 but the unique nature of this group would also support a further extension.


Unless stated otherwise, references are to the Programme and Abstracts of the 8th IAS Conference on HIV Pathogenesis, Treatment and Prevention (IAS 2015), 19 – 22 July 2015, Vancouver.

  1. Lundgren J et al, The START study: design, conduct and main results. The Strategic Timing of AntiRetroviral Treatment (START) study: results and their implications. IAS 2015, Session MOSY03. Webcast and panel discussion after the presentation:
  2. IAS members meeting. IAS 2015, TUSS02. 14.00-15.30. Room 118-120.
  3. Lundgren J et al. Initiation of antiretroviral therapy in early asymptomatic infection. NEJM (20 July 2015). DOI: 10.1056/NEJMoa1506816.
  4. Annual DSMB open reports 2009 – 2015.
  5. HIV Medicine supplement. The START Trial Characteristics at Study Entry. HIV Medicine Special Issue: April 2015; 16(S1):1-146.
  6. British HIV Association. Treatment of HIV-1 positive adults with antiretroviral therapy. DRAFT guidelines for comment. (July 2015).
  7. Hirnschall G. WHO Global HIV Guidelines: How innovations in policy and implementation can pave the way to achieving 90-90-90. UN 90-90-90 Target workshop: lessons from the field.18 July 2015. Vancouver.
  8. Hirnschall G. Panel discussion following START presentation. IAS 2015, Session MOSY03. Webcast:

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