Summary of STI studies: SMART shows clear increased risk from interrupting treatment at any CD4 count

Simon Collins, HIV i-Base

Results from research into treatment interruptions studies have been debated and reported at most HIV conferences for at least the last five years. The initial hope that intermittent periods of viremia might boost immunological responses to HIV quickly failed to realise that early promise, and recent studies have focused on impact of reduced drug exposure on toxicity, quality of life, and cost. Single treatment interruptions in heavily treatment-experienced patients, in order to shift to wild-type virus and a more successful response to resensitised drugs, have also not shown long-term benefit.

Cohort studies showing that serious symptomatic HIV-related illnesses are generally less common at CD4 counts over 200 cells/mm3, and that responses achieved to HAART were comparable to patients who started at higher CD4 levels, led to recommendations for starting treatment in the UK, prior to or around this level. Evidence from several studies, showing that treatment interruptions were generally safe in people who started treatment at CD4 counts that were higher than referred to in current guidelines, led to recognising that a treatment interruption could be considered in people who started at CD4 counts >200 cells/mm3 and >350 cells/mm3 in the UK and US treatment guidelines respectively. Interruptions during chronic infection in patients with a lower CD4 nadir have always been more controversial.

However, the SMART study – planned to last for 9 years, the largest trial on any aspect of HIV to date and certainly dwarfing any other interruption study to date – was stopped in January 2006 after only 2 years follow-up. [1] The summary results presented at CROI, will increase the caution and potentially reverse the view on the safety of interrupting treatment at any CD4 count during chronic infection. [6]

Over half the presentations in the treatment strategy session in Denver related to treatment interruptions. Most of these studies are summarised below and each provided information on different aspects of this approach.

Researchers from each study, including SMART, very clearly restricted their conclusions to the exact intervention studied. It was repeatedly stated that no conclusions could be drawn on other strategies, and certainly not – heaven forbid – on the implications of when to start treatment, although the SMART study does raise questions about the protective nature of HAART that makes this question important again. Further research into factors that could predict increased risk or safety during an interruption, particularly the duration of the period off-treatment, and the rules used for restarting therapy, are still warranted. Shorter periods off-treatment ‘may’ be safer – the difference between the two arms in SMART did not show until after 4-6 months off-treatment, for example. Whether a shorter period off-treatment would increase quality of life, or increase resistance, both need to be considered.

Interruptions for the majority of patients in all studies (>95% in SMART), over the period of follow-up, did not result in an AIDS defining illness. But CD4-mediated treatment interruption more than doubled the risk of serious events, including death, compared to continuous treatment. The rate was almost four-fold more risky for patients who entered the study with viral load <50 copies/mL. However, for an individual experiencing severe difficulties or HAART-related toxicity, on treatment that is not currently maximally suppressive, a relatively low absolute risk for a shorter interruption may be acceptable, if future analyses show that they have other low risk factors.

Perhaps more optimistically, the original motive for interrupting treatment as a strategy, was driven by the burdens of the earliest treatment regimens: high pill count, more frequent dosing, low use of ritonavir-boosting, poorly managed lipodystrophy and other toxicities. The development of easier, more tolerable regimens not only reduces the incentive of stopping treatment for many patients, but also lessens some of the ethical difficulties around running a large randomised trial looking at optimal time to start treatment.

If treatment was one inexpensive pill, once a day, with no side effects, and no risk of resistance, then it is likely that patients would start treatment on diagnosis, or certainly at much higher CD4 counts than current guidelines recommend. Given the choice, reducing ongoing HIV replication and viral diversification associated with viraemia >50 copies/mL, hopefully in all important compartment sites, has to be a good thing. But everything is a balance between risks and benefits, and for a significant number of patients, treatment is still far from this ideal.

ACTG 5170

ACTG 5170 prospectively followed 167 patients for 96 weeks who stopped treatment, after having previously been stable on >/=2 drug combinations for at least 6 months. [2] ARV treatment was restarted based on patient or doctor recommendation. The primary endpoint was time to CDC Category B or C event or death or CD4 count <250 cells/mm3.

These patients had strong immune systems (median 833 cells/mm3), and had started treatment early by today’s standard. They had a high CD4 nadir (median 436, IQR 375 to 510) and were generally on stable treatment (>75 % had viral load <400 copies/mL). The median time on ART was around four years and the median number of cumulative adverse reactions was 3 (range, 3 to 9)

Most people stayed off treatment for the whole trial – the median length of ART interruption was 96 weeks. CD4 cell decline following stopping treatment, occurred in 2 phases:  a drop of around -200 cells/mm3 in the first 8 weeks, and then decreases of 1.74 cells/mm3 per week, from week 8 through 96. At 96 weeks, 17 patients had confirmed CD4 count <250 cells/mm3 and 46 subjects had restarted ART. One patient had acute retroviral rebound syndrome and 4 subjects had thrombocytopenia (platelet count <55,000). A possible or definite CDC category B or C diagnosis occurred in 3 and 2 subjects, respectively (all with CD4 >350 cells/mm3).

Five patients died, but although no death was related to HIV/AIDS by the researchers, this included cardiovascular events, which could arguably be considered now as potentially HIV or ARV-related. In multivariate analyses only CD4 nadir was predictive of shorter time to a primary endpoint (p=0.049).


The Staccato trial randomised 430 patients (from Thailand, Switzerland and Australia) who had CD4 counts >350 cells/mm3, and viral load <50 copies/mL 1:2 to either continuous therapy (n=146) or scheduled treatment interruption (STI) (n=284). Treatment restarted before falling below 350 cells/mm3. The median time on randomised treatment was 22 months. At the end of the study, both groups were again treated continuously for 12 to 24 weeks. [3]

The percentage of patients restarting treatment was 53% at 6 months, 64% at 12 months, and 74% at 24 months. After starting treatment, the percentage of patients with viral <50 copies/mL was 91.8% in continuous therapy, compared to 90.3% in STI arm (NS difference of 1.6%, 95%CI –4.0 to +7.3%).

At the end of randomised treatment, median CD4 counts were 374 cells/mm3 in the STI arm (with 60.5% patients >350 cells/mm3), and 601 cells/mm3 in the continuous therapy arm (96.2% >350, p <0.002). After re-treatment, median CD4 counts rose in STI from 374 to 459 cells/mm3 after 12 weeks, with 85.9% >350 cells/mm3, compared with 96.9% in continuous therapy (p <0.01).

After stopping treatment, 17 patients (5.8%) had symptoms of acute retroviral syndrome. Diarrhoea (p = 0.04) and neuropathy (p = 0.03) were more frequent in continuous therapy, whereas oral and vulvo-vaginal candidiasis (p = 0.03) and thrombocytopenia (p = 0.06) were more frequent in STI.


The ANRS WINDOW trial randomised 403 adults with a nadir CD4 >100 cells/mm3, plasma HIV RNA viral load <200 copies/mL, and CD4 >450 cells/mm3 for >6 months while on HAART to either an 8-weeks-off/8-weeks-on intermittent therapy; or continuous therapy for 96 weeks. Primary outcome was time to CD4 <300 cells/mm3. [4]

Median and nadir CD4 counts were 741 and 280 cells/mm3 respectively. 7 (4%) and 12 (6%) patients in the intermittent and continuous arm were lost to follow-up. In the intermittent arm, 2 patients died (deaths were not related to HIV). No AIDS-defining event was recorded.

By intent-to-treat analysis 7 (3.6%) vs 3 (1.5%) reached primary endpoint <300 cells/mm3 in the STI and CT arms respectively. At week 96, the proportion of patients, intermittent vs continuous, with CD4 >450 cells/mm3 and with viral load =400 copies/mL was 75% vs 92% (p <0.0001) and 81% vs 90% (p=0.02).


The Trivacan trial radomised 110 patients to continuous treatment (CT) and 216 to CD4-guided treatment interruptions. A third arm, looking at fixed interruption of 2-months off/4-months in around 330 patients is continuing, but was not reported in this analysis. [5]

Primary endpoints were death and serious morbidity (any WHO stage 3 or 4 classifying event). The STI arm was stopped in October 2005 after a DSMB review of an interim analysis.

Patients (77% women) were all on HAART (91% NNRTI-based, 9% PI-based), with median CD4 460 cells/mm3. Median follow-up was 20 months and 4 patients were lost to follow-up.

Deaths and serious clinical events occurred at least twice as frequently in the STI compared to the CT arm. 5 patients died (incidence 0.6/100 person-years in CT arm, 1.2/100 person-years in the CGT arm, relative risk [RR] 0.48, 95%CI 0.01 to 4.91), and 60 patients had experienced 84 serious morbidity events (CT 6.7/100 person-years, CGT 15.2/100 person-years, RR 0.44, 95%CI 0.21 to 0.87).

The study abstract listed clinical events: 42 oro-pharyngeal candidiasis (CT 2.3/100 person-years, CGT 6.4/100 person-years, RR 0.36, 95%CI 0.09 to 1.08), 24 invasive bacterial events (CT 0.6/100 person-years, CGT 6.7/100 person-years, RR 0.08, 95%CI 0.01 to 0.52), 16 tuberculosis (CT 2.3/100 person-years, CGT 3.6/100 person-years, RR 0.65, 95%CI 0.15 to 2.14), and 2 others. Of the 24 bacterial events, 14 occurred with documented bacteremia (all in the CD4 guided treatment interruption arm) and 10 occurred without bacteremia but with hospital admission. In the 286 patients who reached month 12 after randomization (CT 96, CD4 guided treatment interruption 190), the median month 12 CD4 count was 546 cells/mm3 (IQR 452 to 645 cells/mm3) in the CT arm and 333/mm3 (IQR 286 to 404 cells/mm3) in the treatment interruption arm.


The SMART trial randomised almost 5,500 patients to either CD4 guided treatment interruptions (STI), stopping when >350 cells/mm3 and restarting at 250 cells/mm3, or continuous treatment (CT). This was an international study with sites in 33 countries. This study was planned to run for nine years but was stopped on 10 January 2006, after an interim analysis and DSMB recommendation showed that not only were more deaths and AIDS-related events occurring in the STI arm, but that the serious ‘non-AIDS’ events (cardiovascular, hepatic, renal), were also occurring more frequently in the STI arm. One of the rationales behind SMART was that continuous treatment arm would lead to more serious drug-related events compared to intermittent treatment. [6]

The majority of patients (95%) were ART-experienced with a median of 6 years prior ART use. The mean age was 46 years; 27% were women. Median baseline CD4 count and CD4 nadir were 598 cells/mm3 and 253 cells/mm3 respectively.

At baseline, 70% had HIV RNA <400 copies/mL and 24% had received a prior AIDS diagnosis. Median follow-up was 10 months (IQR 4 to 23), with 6139 person-years of follow-up at this analysis; only 2.1% of patients were lost to follow-up.

In the STI and CT arms, ART was used 33% and 93% of the follow-up time and patients spent 938 (31.7%) and 236 (8.1%) patient-years at CD4 <350 cells/mm3, 276 (9.3%) and 59 (2.0%) patient-years at CD4 <250 cells/mm3, respectively. Confirmed clinical endpoints are shown in Table 1. The relative risk of disease progression or death (STI/CT) was greater in the first 2 years of follow-up (RR = 2.7, 95%CI 1.8 to 4.2) compared to afterward (RR = 1.1, 95%CI 0.6 to 2.2).

The study results were unchanged when the data were analysed by other criteria and category, including gender, race and geographical region. Most problematic was the finding that neither proximal CD4 count (the CD4 count before the event), nor nadir CD4 count, was predictive of events during an STI. This prevented the study from amending the protocol for interruptions, enabling data from each strategy to be safely studied and greater follow-up data to be collected.

Although median follow-up in SMART was 10 months (mean 14 months), some patients have been in the study for over 4 years. Median duration of treatment interruption in the presentation was given as 18 months because of calculated estimates as the vast majority of people in the STI arm of the study had not restarted treatment when the study was stopped in January.

Further follow-up data will continue to be collected as per study protocol. Additional results, including the quality of life sub-study results collected for over 1200 patients, will be presented in future meetings, hopefully for the IAS meeting in Toronto this Summer.

Patients in the STI arm who are currently off-treatment are being advised of the ‘prudence’ of restarting treatment in the light of these results.

Table 1: Summary of results from SMART study

Endpoint STI n/rate* CT n/rate RR (STI/CT) 95%CI (p-value)
Prog of disease or death 93 (3.1) 44 (1.4) 2.15 1.50-3.08 (p<0.0001)
Serious AIDS events 16 (0.5) 3 (0.1) 5.82 1.68-20.16 (0.01)
Fatal/non-fatal: MI, stroke, CAD requiring surgery, renal, & liver disease 59 (2.0) 37 (1.2) 1.62 1.07-2.44 (0.02)
Death 47 (1.5) 29 (0.9) 1.63 1.02-2.58 (0.04)

* per 100 person-years

Table 2: Treatment interruption studies at CROI

Study ACTG 5071 Staccato Window (ANRS) Trivacan (ANRS) SMART
CROI abstract 101 102 104 105LB 106LB
Country US Thailand, Swiss, Australia France Cote ‘d’Ivoire US, S. America, Europe, Australia (33 countries)
N 167 TI 284 TI; 146 CT 197 TI; 194 CT 216 TI; 110 CT ~ 5,500
% Women 17% ~ 50%- 60% ~ 20% 80% women 27% women
Study design & STI strategy Single arm. STI for 96 wk. Med f/u 96 wk STI, and med 46 wk f/u after restarting ARVs Randomised. Stop for 96 wks Randomised. 8 wks on/8 wks off ARV-naive, randomised to CT, CD4-guided TI or fixed TI (4 month on/2 month off – data not presented) Randomised. CD4 guided, <250 / >350 Median f/u 10 months, mean f/u 14 months.
Median CD4 at entry 833 ~ 500 741 ~ 460 598
Median CD4 nadir (R/IQR) 436 (R 183-1020) ~ 260 (IQR 200-340) 280 (IQR 200-370) ~ 270 253
Median time on Rx before STI 4.5 yrs (2.3-6.0) 15 months (IQR 7-34) 5.2 years 7 months 6 years
Median time off Rx 24 months 22 months 92% at 96wk 19 months 18 months
Resistance Not presented Low Low – similar in each arm 5% vs 11% to = 1 ARV (NS, p=0.13) Not presented
Study conclusion ‘TI generally safe’ By 96wk 46 pts restarted ARVs. 5 deaths (4 poss CVD, 1 sepsis). HR 1.95 for time to <250 or event Side effects more frequent in CT, more minor manifestations of HIV infection in STI. 2 deaths (1 each arm) ‘TI appeared clinically and immunologically safe’ 2 deaths in TI arm (1 alcohol related cirrhosis; 1 violent) CD4 guided STI showed >2 x risk of serious events compared to CT. STI arm stopped. (6.7 vs 17.6 /100 pt yr – RR 2.6) TI showed twice risk (RR2.5) of death and AIDS events compared to CT. RR 3.8 for pts RNA <400 c/mL at baseline

TI = Treatment Interruption; CT = Continuous Treatment; ~ = approximate, often range between different groups


One lesson to learn from CD4 guided treatment interruptions is to monitor patients closely (e.g. every 4 weeks) and catch them when their CD4 count is still well above 250 cells/mm3. 12% of patients in the treatment interruption arm in SMART were below 250 cells/mm3 as routine visits were planned only every 12 weeks after the first 6 months. The higher incidence of cardiovascular events in SMART is in contrast with the findings from the D:A:D study.


Unless stated otherwise, all references are to the Programme and Abstracts from the 13th Conference on Retroviruses and Opportunistic Infections, 5-8 February 2006, Denver, Colorado.

  1. SMART Study halted due to safety concerns with significantly more AIDS events in the treatment interruption arm. HIV Treatment Bulletin January/February 2006.
  2. Skiest D, Havlir D, Coombs R et al. Predictors of HIV disease progression in patients who stop ART with CD4 cell counts >350 cells/mm3. Abstract 101.
  3. Ananworanich J, Gayet-Ageron A, Le Braz M et al. CD4-guided scheduled treatments interruptions compared to continuous therapy: results of the Staccato trial. Abstract 102.
  4. Marchou B, Tangre P, Charreau I et al. Structured treatment interruptions in HIV-infected patients with high CD4 cell counts and virologic suppression: results of a prospective, randomised, open-label trial (Window – ANRS 106). Abstract 104.
  5. Danel C, Moh R, Sorho S et al. CD4-guided strategy arm stopped in a randomised structured treatment interruption trial in West-African adults: ANRS 1269 Trivacan trial. Abstract 105LB.
  6. El-Sadr W, Neaton J for SMART team. Episodic CD4-Guided Use of ART Is Inferior to Continuous Therapy: Results of the SMART Study. Abstract 106LB.

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