CASCADE analysis of when to start treatment

Simon Collins, HIV i-Base

Joe Eron from University of North Carolina, presented an analysis from the CASCADE seroconverter study, looking at rates of AIDS events and deaths for treatment by baseline CD4 count. The presentation was given in the context of the debate about when to start treatment. [1]

The summary results supported current guidelines to start after CD4 count drops below 350 cells/mm3. The group also reported a significant benefit from starting at 350-500, but still emphasised that due to the difficulties of interpreting cohort data, the randomised START study was still essential to address this question. Of note, the analysis found no benefit from starting at CD4 counts over 500 cells/mm3.

CASCADE is an international collaboration of cohorts that includes people who have been diagnosed in primary HIV infection, with recent diagnoses confirmed by a recent (<1 year) negative test or PCR-positive. The cohort includes 52,268 person years of follow up (PYFU), median 4.7 years, from 9,455 people: 78% male. 56% MSM, 25% MSW, median age at diagnosis 30 years (IQR 25–37). From this group, 8.6% (n=812) developed AIDS and 5.8% died (n=544).

In this analysis, the group pooled observations from 161 monthly sub-cohorts from January 1996–May 2009, based on whether 3-drug HAART was started in each month or deferred. Primary endpoints included time to first AIDS diagnosis, all-cause mortality and last time alive, with secondary endpoints including time to death. Covariates with each monthly update included age, gender, CD4, viral load, injecting drug use, viral hepatitis, seroconversion illness and calendar year. The methodology including complicated weighting analyses to reduce lead-time bias included looking at the probability of starting or deferring as a factor of patient characteristics using these covariates.

The study was more strongly weighted in terms of both patient numbers and events to higher CD4 strata (see Table 1).

Table 1: Patients and events by CD4 strata

CD4 count n Initiated Events/deaths PYFU
0-49 183 107 102 664
50-199 1,521 832 353 6,934
200-349 4,459 1,792 732 22,106
350-499 5,527 1,005 815 29,653
500-799 5,162 615 696 28,631

Note: n, events and PYFU are not unique across CD4 strata

Adjusted hazard ratios (aHR), 3-year cumulative risk differences (RD), and numbers needed to treat (NNT) for 3 years to prevent one additional outcome were calculated. As with other studies, HAART was strongly associated with better outcomes for CD4 counts <200 cells/mm3. However, the study found no benefit to starting at 500-799 cells/mm3 and a relatively small decrease in risk among those with CD4 counts of 350-499 cells/mm3, see Table 2.

The absolute differences between the <350 and 350-500 groups were low over one year. The NNT at CD4 350-499 was 34 to prevent one AIDS event or death within three years and was 74 to prevent one death over the same period.

The presenter emphasised the importance of the data from the randomised START trial to be able to exclude confounders associated with cohort data (including non-AIDS events, comorbidity, concurrent medications, current IDU, depressions and social support). Similarly, viral response and treatment interruptions were not analysed to determine relative and absolute risk in patients on successful treatment.

Table 2: Adjusted relative and absolute effects of starting vs deferring HAART

aHR AIDS/death (95%CI) 3-Yr RD AIDS/death (95%CI) 3-Yr NNT
AIDS/death (95%CI)
aHR death alone (95%CI) 3-Yr RD death alone (95%CI) 3-Yr
NNT death alone (95%CI)
0-49 0.32 (0.17, 0.59) -30.0 (-45.1, -15.0) 3 (2, 7) 0.37 (0.14,
-18.2% (-32.0, -4.4) 6 (3, 23)
50-199 0.48 (0.31, 0.74) -15.0 (-19.7, -10.3) 7 (5,
0.55 (0.28, 1.07) -7.2% (-10.1, -4.4) 14 (10,23)
200-349 0.59 (0.43, 0.81) -4.8%
(-7.0, -2.6)
21 (14, 38) 0.71 (0.44, 1.15) -1.4% (-3.0, 0.3) 74 (33, 8)
350-499 0.75
(0.49, 1.14)
-2.9% (-5.0, -0.9) 34 (20, 115) 0.51 (0.33, 0.80) -1.4% (-2.2, -0.6) 71 (45, 165)
500-799 1.10 (0.67, 1.79) 0.3% (-3.7, 4.2) 8 (n/a) 1.02 (0.49, 2.12) -0.4% (-2.0, 1.2) 239
(49, 8)


This study was less powered than both the ART-CC and NA-ACCORD cohort studies that have reported conflicting results when looking at cohort data to inform the question of optimal CD4 count to start treatment. However the advantage of this study is that the group looked at absolute events rather that relative estimates. At 350-499 there is seems to be a low incidence of events anyway and the
benefits don’t appear till 3 years after initiation- which all needs to go into the decision weighing the risk:benefit ratio.

This is complicated in the CASCADE analysis as non-AIDS events are not recorded and there is sometimes incomplete information on the cause of death.

It was notable that the presentation referred to the need for data from the randomised START study to answer the question of when to start treatment at higher CD4 counts. [2]


  1. Eron J et al. HAART initiation and clinical outcomes: insights from the CASCADE cohort of HIV-1 seroconverters on ‘When to Start’. 18th IAS Conference, 18–23 July 2010, Vienna. Oral abstract late breaker THLBB201.
  2. Strategic Timing of AntiRetroviral Treatment (START) study.

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