| |
Risk of HAART interuptions in Africa and the USA: The risk of treatment versus the risk of HIV replication (treatment interruptions)
|
| |
| |
COMMENTARY
See article following Commentary
The Lancet June 17, 2006; 367:1955-1956
Steven G Deeks
San Francisco General Hospital, San Francisco, California 94110, USA
It has been about 25 years since the initial AIDS case reports and 10 years since the first reports suggesting that combination highly active antiretroviral therapy (HAART) can potently and perhaps indefinitely suppress HIV replication.1 During that time, several widely accepted conclusions about treatment have emerged. First, HAART works. When the right drugs are given to a motivated and highly adherent patient, HIV replication is suppressed indefinitely. Second, most if not all antiretroviral drugs have short-term and long-term side-effects. These side-effects include chronic gastrointestinal symptoms, low-level malaise, mood changes, lipid abnormalities, accelerated atherosclerosis, and peripheral fat-wasting (lipoatrophy). Third, life-long adherence to even a well-tolerated regimen can be challenging for most patients. Pill fatigue eventually sets in for many patients, leading to a lower quality of life and an increased risk of non-adherence and eventual treatment failure. Fourth, antiretroviral therapy and its monitoring are expensive. First-line regimens range from US$5000 to $10 000 a year in resource-rich countries and from $300 to over $1000 a year in resource-poor regions.2 Second-line regimens are even more expensive.
This tension between the benefits and risks of antiretroviral therapy has led to widespread enthusiasm for intermittent therapy. The idea is simple: give antiretroviral therapy until the CD4+ T-cell count recovers and then discontinue therapy until the T-cell count drops to low levels. One such strategy was tested by Christine Danel and colleagues in west Africa (ANRS 1269 or Trivacan)3 and is outlined in today's Lancet. 326 patients on antiretroviral therapy (<300 copies HIV RNA per mL) with low viral loads and high CD4+ T-cell counts (>350/mL) were randomised to continuous versus intermittent therapy.3 In the intermittent arm, treatment was interrupted at CD4+ T-cell counts above 350/mL and resumed when counts fell to below 250/mL. The study's data and safety monitoring board (DSMB) stopped the study early because of a higher risk of severe morbidities-particularly bacterial infections-in the intermittent treatment arm. Another DSMB recently halted the similarly designed SMART study.4 SMART, with 5472 patients, was done mainly in resource-rich countries, and found an increased risk of severe complications and death in those randomised to intermittent therapy.4 Although in each study the difference in outcomes was highly significant, the absolute risk of disease progression was still low in those randomised to intermittent therapy (17-6 events in Trivacan and 3-7 events in SMART, both per 100 patient years).
These studies have clear implications for those setting treatment guidelines and/or designing clinical trials. When possible, treatment should be provided indefinitely for patients who are doing well on therapy and who have a CD4 cell count below 350/mL. Importantly, these studies also provide indirect support for starting therapy earlier than the currently recommended threshold of 200/mL. Whether therapy can be administered intermittently at higher CD4 thresholds remains unclear but enthusiasm for that approach is likely to wane.5,6 One recent large randomised clinical study of starting and stopping at CD4 cell counts above 350/mL revealed only slightly increased risk of minor HIV-related complications in the interruption arm (Staccato; table).7
The implications for the individual management of antiretroviral therapy level are less clear. Because of the Trivacan and SMART studies, the most recent US Department of Health and Human Services guidelines state that treatment interruptions should be avoided in clinical practice and should only be done in a closely monitored clinical trial.8 However, this is not practical because many patients in clinical practice have significant side-effects, pill fatigue, or treatment failure. Presumably any study that specifically focuses on those with strong reasons to stop therapy might show a benefit (or at least a lack of harm) in going on and off therapy.
The fundamental challenge raised by Trivacan and SMART is how to translate data from broad heterogeneous study populations to an individual. For patients who are doing well on a stable regimen, the Trivacan and SMART data clearly indicate that uncontrolled HIV replication is more harmful than modern treatment regimens, and that well-tolerated drugs should be continued indefinitely. This finding is not surprising and was widely accepted even before these studies were done (at least it pertains to regions where treatment is widely available). At what point the harm associated with treatment in an individual outweighs the harm associated with uncontrolled HIV replication requires a careful understanding of treatment-related side-effects, HIV pathogenesis, and the consequences of interrupting therapy. The Trivacan and SMART studies provide a quantitative risk assessment of the consequences of stopping therapy and are therefore important studies for all treating clinicians. The task at hand now is to ensure that all patients-including those in resource-constrained regions, such as West Africa-have access to well-trained health-care providers who can readily adopt emerging data into any decision regarding when (if ever) to stop antiretroviral therapy.
I have received research support or honoraria from Boehringer Ingelheim, Bristol-Myers Squibb, GlaxoSmithKline, Pfizer, Roche, Tibotec, and Trimeris.
CD4-guided structured antiretroviral treatment interruption strategy in HIV-infected adults in west Africa (Trivacan ANRS 1269 trial): a randomised trial
The Lancet 2006; 367:1981-1989
Christine Danel a, Raoul Moh a, Albert Minga a, Amani Anzian a, Olivier Ba-Gomis a, Constance Kanga a, Gustave Nzunetu b, Delphine Gabillard d, Francois Rouet c, Souleymane Sorho a, Marie-Laure Chaix e, Serge Eholie b, Herve Menan c, Delphine Sauvageot a, Emmanuel Bissagnene b, Roger Salamon d and Xavier Anglaret d , for the Trivacan ANRS 1269 trial group
".....In conclusion, our data show that a CD4-guided strategies using CD4 interruption and re-treatment thresholds of 350 per mm3 and 250 per mm3 should not be recommended in adults with the same baseline characteristics as our patients. They also provide good evidence for the importance of severe common infections in adults on HAART at intermediate levels of immunosuppression in Cote d'Ivoire. In view of these high rates of early severe morbidity, the need for trials to assess the benefits of earlier HAART initiation in HAART-naive patients deserves further discussion....."
Summary
Background
Structured treatment interruptions of highly-active antiretroviral therapy (HAART) might be particularly relevant for sub-Saharan Africa, where cost-saving strategies could help to increase the number of patients on HAART. We did a randomised trial of structured treatment interruption in Abidjan, Cote d'Ivoire.
Methods
HIV-infected adults were randomised to receive continuous HAART (CT), CD4-guided HAART (CD4GT) with interruption and reintroduction thresholds at 350 and 250 cells per mm3, respectively, or 2-months-off, 4-months-on HAART. Primary endpoints were death and severe morbidity (any WHO stage 3 or 4 events and any events leading to death) at month 24. We report data from the CT and CD4GT groups until Oct 31, 2005, when the data safety monitoring board recommended to prematurely stop the CD4GT arm. Analyses were intention-to-treat. This study is registered at ClinicalTrials.gov, number NCT00158405.
Results
326 adults (median CD4 count nadir 272 per mm3) were randomised to the CT or CD4GT groups and followed up for median of 20 months. Incidence of mortality (per 100 person-years) was not different between groups (CT 0-6, CD4GT 1-2; p=0-57). Incidence of severe morbidity (per 100 person-years) was higher in the CDG4T group (17-6) than in the CT group (6-7; p=0-001). The most frequent severe events were invasive bacterial diseases. 79% of severe morbidity episodes occurred in patients with CD4 count 200-500 per mm3.
Conclusion
Patients on CD4GT had severe morbidity rates 2-5-fold higher than those on CT. This difference was mainly due to high rates of common diseases in patients with CD4 count 200-500 per mm3. This CD4-guided structured treatment interruption strategy should not be recommended in Abidjan.
Introduction
In patients infected with HIV, highly active antiretroviral therapy (HAART) is a lifelong intervention. Interrupting successful HAART during limited periods of time has been conceptualised with the term structured treatment interruption as a potential solution to improve long-term observance and drug tolerance.1-3 These potential benefits would be equally important in industrialised and non-industrialised countries. Additionally, in resource-limited settings, structured treatment interruption might be seen as especially relevant cost-saving strategies that could contribute to increasing the number of patients on HAART. Interrupting HAART also entails important potential risks, including viral rebound, decrease in number of CD4-positive T lymphocytes (CD4),4-9 and increased resistance to drugs.10-13 The benefits and risks of structured treatment interruption thus need to be carefully weighed through controlled trials.14
Structured treatment interruption can be guided by CD4 measurements, with treatment interruption and reintroduction being decided on the basis of predetermined thresholds. It may also consist of fixed periods of time on and off treatment. One important challenge for such strategies is to prevent patients from entering an at-risk zone of immune suppression, which would increase the risk of severe morbidity. The patients' characteristics and the context of access to care could affect the efficacy of structured treatment interruption.
We assessed the efficacy of a CD4-guided strategy of structured treatment interruption, in HIV-infected adults in Abidjan, the economic capital of Cote d'Ivoire, west Africa.
Among HIV-infected adults with a median CD4 count nadir of 272 per mm3 and a CD4 count of greater than 350 per mm3 at first interruption, who had previously successfully received continuous HAART for 6-18 months, the CD4-guided strategy led to a 2-5-fold higher incidence rate of severe morbidity than the continuous treatment strategy.
In our opinion, there are two main reasons for this finding. First, in the CD4-guided treatment group, the CD4 interruption and reintroduction thresholds maintained patients in the 250-500 per mm3 CD4 range, and the most frequent causes of WHO stage 3 or 4 morbidity were invasive bacterial diseases, followed by oral candidiasis and tuberculosis. Significant rates of bacterial diseases and tuberculosis have previously been reported in HIV-infected adults with more than 200 per mm3 CD4 in various sub-Saharan African settings.18,22-24 Therefore, when designing our trial, we could have chosen higher CD4 thresholds.
Second, in a CD4-guided strategy, both the CD4 interruption-reintroduction thresholds and the population should be targeted considering the risk of a fall in CD4 count to an unacceptable zone between two measurements. In our trial, the risk of a CD4 count below 200 per mm3 was associated with a low CD4 nadir, a low CD4 count at first interruption,25-27 and a low haemoglobin concentration at first interruption; the risk of severe morbidity was associated with a high pre-HAART viral load and a low BMI at first interruption. A trial comparing a continuous treatment with a CD4-guided strategy using the same CD4 threshold in a population with higher CD4 nadir, higher CD4 at first interruption, higher haemoglobin concentration at first interruption, lower pre-HAART viral load, and higher pre-HAART BMI might have given different results.
Previous studies with structured treatment interruption have been reported, all of them in industrialised countries.2-12,28-32 At the time when we stopped the CD4-guided arm, four other trials of structured treatment interruptions were about to terminate: two assessing fixed-period strategies (WINDOW and DART trials),33,34 one assessing a CD4-guided strategy (SMART trial),35 and one assessing both (STACCATO trial).36,37 Our trial, unlike the other two trials that included a CD4-guided group, assessed a CD4-guided strategy in Africa. The STACCATO trial, done in Switzerland, Australia, and Thailand, used a higher CD4 re-treatment threshold (350 per mm3), and terminated on schedule without observing significant differences in term of severe morbidity.37 The SMART trial, mainly done in the USA and in Europe, used the same CD4 thresholds as ours, and was prematurely stopped in January, 2006, because of a higher rate of morbidity and mortality in the CD4-guided treatment group. The incidence and pattern of morbidity in this trial were consistent with the range of HIV-morbidity in industrialised countries. The overall morbidity rate was five-fold lower than ours, and was mostly due to oesophageal candidiasis, recurrent pneumonia, Pneumocystis jirovecii pneumonia, and Kaposi's sarcoma.35 Compared with these data, our morbidity data provide evidence that, for a given intervention, results obtained in HIV-infected patients in industrialised countries are not always comparable to those obtained in resource-limited settings.
In Africa, co-trimoxazole prophylaxis is given earlier than in industrialised countries, to target bacterial diseases and malaria.38 In our trial, all patients were prescribed co-trimoxazole, and bacterial diseases remained the first cause of severe morbidity. Although patients from the CD4-guided arm might be suspected to have been less compliant with co-trimoxazole, there are three arguments against this suggestion: the time on co-trimoxazole was not different between the CT and CD4GT groups; 88% of isolated bacterial pathogens were resistant to the drug; the incidence of bacterial diseases in our study was consistent with a previous estimate of bacterial morbidity incidence in adults on co-trimoxazole in Cote d'Ivoire,18 and lower than the rate previously reported in the placebo arm of a trial of early co-trimoxazole prophylaxis in the same setting.39 The fact that the rates of severe morbidity in our study were similar to those that could be expected from the CD4 level in Cote d'Ivoire also contradicts the hypothesis that the CD4-guided strategy could have induced immunological impairment beyond what can be measured by CD4 count.
Usually, interim analyses are done at a more stringent α level, to adjust α for the final analysis so that the overall α level will be preserved. In our study, no adjustment was used, and the overall α level for the study was inflated. However, the p values for primary outcome comparisons are such that adjusted or non-adjusted testing would lead to the same statistical conclusion.
Because of the premature ending of our CD4-guided arm, we were not able to make conclusions on the other endpoints that were to be assessed at 24 months. Although our trial had three arms, it should now be seen as two different trials. After the CD4-guided arm was stopped on the basis of results of the interim analyses on primary outcomes, the comparison of the CD4-guided arm with the continuous treatment arm was complete, leading to a clear conclusion: that this CD4-guided strategy should not be recommended in Cote d'Ivoire. The comparison of the 2-months-off, 4-months-on arm with the continuous treatment arm is still incomplete. At the same time as the data safety monitoring board members recommended to stop the CD4-guided strategy, they also recommended that the 2-months-off/4-months-on arm should be continued until the last randomised patient reached 24 months after being randomised, as planned in the protocol. At the end of this second comparison, if the analyses on primary outcomes show that the 2-months-off/4-months-on arm is non-inferior to the continuous treatment arm, the decision to recommend the 2-months-off/4-months-on strategy in Cote d'Ivoire will have to be discussed on the basis of results of secondary-outcome analyses.
In conclusion, our data show that a CD4-guided strategies using CD4 interruption and re-treatment thresholds of 350 per mm3 and 250 per mm3 should not be recommended in adults with the same baseline characteristics as our patients. They also provide good evidence for the importance of severe common infections in adults on HAART at intermediate levels of immunosuppression in Cote d'Ivoire. In view of these high rates of early severe morbidity, the need for trials to assess the benefits of earlier HAART initiation in HAART-naive patients deserves further discussion.
Results
Figure 1 shows the flow of participants through each stage. 840 patients were included in the pre-randomisation phase. The most frequent reasons for not being included was by far a CD4 count and CD4% outside of the bracket (n=417), followed by unwillingness to participate (n=169), haemoglobin less than 75 g/L (n=21), not returning to the study centre until after the inclusion phase had closed (n=14), and pregnancy (n=12). At inclusion, 316 patients had a CD4% between 12-5% and 20% and CD4 count between 150 and 350 per mm3. The remaining patients met either only the CD4 % (n=138) or only the CD4 count (n=386) criteria.
Between Aug 27, 2003, and March 31, 2005, 651 patients were randomised (figure 1). 189 enrolled patients were not randomised; of those in whom at least one randomisation criterion was absent until month 18, reasons included CD4 count less than 350 per mm3 (n=115), detectable last viral load (n=34), severe HIV disease (n=19), pregnancy (n=2), ongoing HAART regimen including abacavir or nevirapine (n=6), past history of HAART regimen modification during the pre-randomisation phase (n=7), and not returning to the study centre (n=3). All patients received the allocated intervention.
Table 1 shows the main characteristics of patients at randomisation and the main follow-up characteristics after randomisation. As of Oct 31, 2005, the incidence of loss to follow-up was 0-6 per 100 person-years (PY) in the CT group and 0.9 per 100 PY in the CD4GT group (incidence rate ratio [IRR] 0-65, 95% CI 0-01-8-11, p=0-77).
The probability of survival was not statistically different between the groups. The incidence of mortality was 0-6 per 100 PY (95% CI 0-02-3-7) in the CT group and 1-2 per 100 PY (95% CI 0-3-3-2) in the CD4GT arm (IRR 0-48, 95% CI 0-01-4-91, p=0-57).
The probability of remaining free of severe diseases was significantly lower in the CD4GT than in the CT group (figure 2). Table 2 shows the number and rates of first severe episodes. There were 98 episodes of severe morbidity in 63 patients. At the onset of the 98 severe morbidity episodes, the last available CD4 count was less than 200 per mm3 in eight episodes (8%), 200-500 per mm3 in 77 episodes (79%) and greater than 500 per mm3 in 13 episodes (13%). The most common causes of severe morbidity were bacterial diseases. A pathogen was isolated in 28 of the 32 episodes of bacterial diseases. Among the 28 pathogens, 17 were isolated at least in one blood culture (all in the CD4GT group) and 11 only in samples other than blood (CT one, CD4GT ten). The 28 strains were Escherichia coli (CT one, CD4GT ten), non-typhi salmonella (0, eight), Streptococcus pneumoniae (0, eight) and Shigella sp (0, one). Most (25, 89%) of these strains were resistant to co-trimoxazole. The mean (SD) last available CD4 count at the onset of severe events was 314 per mm3 (108) for bacterial diseases and 386 per mm3 (117) for tuberculosis.
When restricting the analyses to patients on a non-nucleoside reverse transcriptase inhibitor (NNRTI)-based regimen at baseline, the results were unchanged (data not shown). In a multivariate analysis restricted to the CD4GT group, two factors were significantly associated with an increased risk of severe morbidity: a pre-HAART concentration of HIV RNA in plasma of greater than 5 log10 copies per mL (p=0-03), and BMI of less than 20-5 kg/m2 at baseline (p=0-04; webtable 1).
Two other causes of morbidity were significantly different between groups: acute fever with no focus (CT 21-1 per 100 PY, CD4GT 41-3 per 100 PY; IRR 0-55, 95%CI 0-35-0-85, p=0-004), and papular prurigo (CT 7-8 per 100 PY, CD4GT 14-6 per 100 PY; 0-54, 0-26-1-00, p=0-04). The differences in term of malaria (CT 4-1 per 100 PY, CD4GT 6-0 per 100 PY; 0-68, 0-24-1-67, p=0-38) and herpes zoster (CT 4-7 per 100 PY, CD4GT 6-7 per 100 PY; 0-70, 0-27-1-64, p=0-41) did not reach significance.
Figure 3 shows the mean CD4 count at each point. Overall, patients from the CT group spent 1%, 10%, 32%, and 57% of their follow-up time in the CD4 less than 200 per mm3, 200-349 per mm3, 350-499 per mm3, and 500 per mm3 or greater strata, respectively. In the CD4GT group, these proportions were 5%, 45%, 37%, and 13%, respectively.
In a multivariate analysis restricted to the CD4GT arm, a CD4 count of less than 200 per mm3 was significantly associated with a low CD4 nadir (p=0-001), a low CD4 count at randomisation (p=0-004), and a low haemoglobin concentration at randomisation (p=0-01; webtable 2).
Among patients from the CD4GT group, 64 (30%) never restarted HAART because their CD4 count never fell below 250 per mm3, 111 (51%) once restarted HAART and continued following the structured treatment interruption strategy until the CD4GT arm was stopped, and 41 (19%) were prescribed continuous HAART with no further interruption because of a desired pregnancy (n=1), severe morbidity event (n=38), and rapid fall of CD4 count less than 200 per mm3 (n=2). Overall, the median time from randomisation to first HAART reintroduction was 9-1 months (95% CI 7-2-10-3). The 64 patients who never restarted HAART statistically differed from the 152 other patients for the following characteristics: at baseline, they were more frequently at WHO stage 1 or 2 (79% vs 54%, p=0-0004), they had a higher mean CD4 count (349 vs 248 per mm3, p<0-0001), and a lower mean concentration of HIV RNA in plasma (4-4 vs 5-0 log10 copies per mL, p=0-0002); at randomisation, they were more frequently at WHO stage 1 or 2 (64% vs 44%, p=0-007), had a higher mean BMI (24-7 kg/m2vs 22-7 kg/m2, p=0-003), and a higher mean CD4 count (595 vs 463 per mm3, p<0-0001).
Patients spent a mean of 19-4 months (SD 4-7) on HAART in the CT group, compared with 5-9 months (5-5) on HAART and 13-3 months (6-3) off HAART in the CD4GT group. Mean times on and off cotrimoxazole prophylaxis were 16-8 months (6-7) and 2-4 months (4-9), respectively, in the CT group, and 17-5 months (6-0) and 1-7 months (3-6) in the CD4GT group, respectively (p=0-54 for time on and 0-57 for time off, for comparison between the two groups).
The mean monthly proportion of patients on HAART who declared having missed at least one antiretroviral drug intake during the previous 4 days was 11-1% in the CT and 11-3% in the CD4GT group (p=0-85). This percentage did not change significantly over time.
Overall 293 patients (CT 96, CD4GT 197) reached month 12. Among them, 195 (CT 96, CD4GT 99) remained on HAART during at least 2 months of HAART at any time during the first 12 months and were thus taken into account for the description of the virological outcomes at month 12. Of these 195 patients, 24 (13%; CT seven [7%], CD4GT 17 [17%]) had a detectable viral load (>300 copies per mL), including nine (5%; CT three [3%], CD4GT six [6%]) without evidence of drug resistance mutation and 15 (8%; CT four [4%], CD4GT 11 [11%], p=0-07) with a resistant virus to at least one drug of their previous HAART regimen or regimens.
The viral loads of the nine patients with a wild type virus were 2-80, 2-96 and 4-24 log10 copies per mL in the CT group and 2-84, 2-94, 3-48, 3-81, 3-86 and 4-39 log10 copies per mL in the CD4GT group. Of the 15 patients with at least one resistance mutation, six were resistant to NNRTIs alone (CT 0, CD4GT six), eight to two drugs (CT four, CD4GT four), mostly to NNRTIs and lamivudine, and one (CD4GT) to three drugs: zidovudine, lamivudine, and NNRTIs.
The mean rate of unscheduled consultations was 159 per 100 PY in the CT group and 233 per 100 PY in the CD4GT group (p<0-0001). The mean number of days in hospital was 36 per 100 PY in the CT group and 68 per 100 PY in the CD4GT group (p<0-0001).
At month 12, there was no significant difference between the proportion of patients with arterial blood pressure 140/90 mm Hg or greater (CT 16%, CD4GT 12%, p=0.41), signs or symptoms of lipodystrophy (9%, 4%, p=0-15), LDL cholesterolaemia 3-4 mmol/L or greater (31%, 23%, p=0-18), and triglyceridaemia 2-3 mM or greater (3%, 2%, p=0-40). The incidence of grade 3-4 drug side-effects was 19-1 per 100 PY in the CT group and 17-3 per 100 PY in the CD4GT group (p=0-67). The main grade 3-4 side-effect was by far neutropenia (CT 19-6 per 100 PY, CD4GT 14-1 per 100 PY). During follow-up, the incidence rate of drug regimen modification was 1-7 per 100 PY in the CT group and 1-5 per 100 PY in the CD4GT group (p=0-82).
Methods
Patients and setting
We did a prospective, randomised, open-label, multicentre, noninferiority trial, to compare continuous HAART with two structured treatment interruption strategies. The study protocol was approved by the ethics committees of the Ministry of Health of Cote d'Ivoire and by the institutional review board of the French National Agency for Research on AIDS and viral hepatitis (ANRS, Paris, France).
Volunteers were recruited in five outpatient HIV-clinics. Between Dec 26, 2002, and April 28, 2004, all HIV-infected adults were candidates for inclusion if they had either a CD4 count between 150 per mm3 and 350 per mm3 or percentage of CD4 (CD4%) between 12-5% and 20%. They were not included if any of the following were present: age younger than 18 years, any past history of antiretroviral treatment (with the exception of short course antiretroviral treatment for the prevention of mother-to-child transmission of HIV), pregnancy, severe renal or hepatic failure, or severe haematological disorder, residence outside of Abidjan, unwillingness to participate, Karnofsky score less than 50, and severe psychiatric disease. All patients gave written informed consent.
Procedures
The trial comprised two phases. In the pre-randomisation phase15,16 all patients received a continuous antiretroviral treatment. The first-line regimen was zidovudine plus lamivudine (Duovir, Cipla, Mumbai, India) plus: (1) preferably efavirenz 600 mg once a day (Stocrin, Merck Sharp and Dohme), or (2) indinavir 800 mg and ritonavir 100 mg twice a day (Crixivan, Merk Sharp and Dohme, Haarlem, Netherlands; and Norvir, Abbott, Zwolle, Netherlands) for HIV-2 infected patients, women refusing contraception, and women with a history of nevirapine prophylaxis.
At any time between 6 months and 18 months in the pre-randomisation phase, patients were randomised into one of three arms whenever they met the following criteria: CD4 count greater than 350 per mm3, plasma HIV-1 RNA less than 300 copies per mL, absence of ongoing morbidity episode and of pregnancy, ongoing HAART regimen excluding abacavir and nevirapine, and no history of HAART regimen modification involving more than one drug during the pre-randomisation phase such that further alternative regimens would not be available in the eventuality of treatment failure after randomisation.
The groups were: continuous therapy (CT; one in six patients); CD4-guided therapy (CD4GT) (two in six patients), with treatment interruption whenever the CD4 count rose above 350 per mm3 and re-treatment whenever the CD4 count fell below 250 per mm3; and a 2-months-off, 4-months-on treatment arm (three in six patients). In the CD4GT arm, a CD4 threshold of 250 per mm3 was chosen for reintroduction of HAART before CD4 count fell below the clinically significant threshold of 200 per mm3. At the time when the trial started, HIV-mortality and most causes of HIV-severe morbidity were thought to be very rare in patients with more than 200 CD4 per mm3.17
The aim of the trial was to test the hypothesis that each structured treatment interruption strategy was non-inferior to the continuous treatment strategy. Three primary endpoints were defined a priori: incidence of mortality; incidence of severe morbidity, which combined any event leading to death and any WHO clinical stage 3 or 4 event;18 and proportion of patients with a CD4 count of greater than 350 per mm3 at month 24. Severe morbidity and mortality were hypothesised a priori to be rare and were not used for sample size calculation.
The secondary endpoints were viral and immunological efficacy, adherence, proportion of patients with resistance to at least one antiretroviral drug, the cost of care, and the incidence of ANRS grade 3 or 4 adverse events.
All reports of potential clinical outcomes were referred for independent review to an event documentation committee. Bacterial diseases with at least one blood culture positive for non-typhi salmonella were classified as WHO clinical stage 4. Diseases with at least one blood culture positive for other clinically significant pathogens were classified as WHO clinical stage 3. When blood cultures were negative, the following bacterial diseases were regarded as severe and therefore classified as WHO clinical stage 3: pneumonia, pleurisy, enteritis with stool cultures positive for non-typhi salmonella or Shigella sp, salpyngitis, pyelonephritis, prostatitis, orchi-epididymitis, meningitis, endocarditis, pyomyositis, pericarditis, or deep abscess.
A computer generated sequentially numbered blocked randomisation list, stratified by centres, was drawn up by an independent statistician, and then included in a software that allowed the trial coordinator to access the next available number but not the entire list. When a patient was to be randomised, the study centre physician phoned the trial coordinator in Abidjan who verified in the centralised database that randomisation criteria were met, and then assigned the next available number to the patient. Participants and trial teams were not blinded to the intervention.
Patients were asked to return every month to their study clinic, which they could also access freely in the interval. Standardised questionnaires were used to record characteristics at baseline and at monthly visits. Questionnaires included, among other things: at each monthly visit, self-reported adherence to treatment during the previous 4 days, body-mass index (BMI), and arterial blood pressure; and every 6 months, self-reported and physician-detected signs of lipodystrophy, and hip and waist perimeters. Symptoms were managed following standardised algorithms. Transport, consultations, investigations, hospital admission, and drugs were free of charge, whether patients dropped out or stayed in the study. The Cote d'Ivoire Ministry of Health undertook to finance subsided HAART for all participants after study completion.
CD4 count (True Count technique on FACScan, Becton Dickinson, Aalst-Enembobegem, Belgium) was measured every 2 months. Plasma HIV-1 RNA (real-time PCR on Taq Man technology Abi Prism 7000, Applied Biosystems, Courtaboeuf, France, threshold of detectability 300 copies per mL19) was measured every 6 months, and, for the CD4GT group, at each new treatment interruption. Genotypic resistance tests were done as previously described20 for all detectable viral loads: for the CT group, in samples collected at months 12 and 24; for the CD4GT group, in samples obtained closest to these months. Only samples obtained after at least 2 months on treatment were taken into account. Resistance mutations were defined according to the 2005 International AIDS Society-USA list,21 and resistance patterns were interpreted according to the French ANRS algorithm. Serum liver enzymes and creatininaemia were measured every 2 months. Glycaemia, cholesterolaemia, and triglyceridaemia were measured every 6 months. Home visits were made for all patients who did not attend scheduled appointments. Patients from the CD4GT group were prescribed a continuous treatment with no further interruption at first severe morbidity event, pregnancy, or rapid fall of CD4 count to less than 200 per mm3.
Co-trimoxazole prophylaxis (800/160 mg daily) was systematically given monthly, except for patients with history of severe intolerance. Patients on antiretroviral treatment were provided monthly at the trial centre pharmacies with the quantity of pills needed for 35 days. Patients off treatment were prescribed not to take any antiretroviral drug during the following month; because efavirenz was known to have a longer half-life than nucleoside reverse transcriptase inhibitors (NRTIs), the patients who interrupted an efavirenz-based treatment were prescribed to stop efavirenz immediately, and to stop the two NRTIs 5 days later.
Final noninferiority analyses were planned to compare each group on structured treatment interruption and the CT group separately, using both the intention-to-treat and the per-protocol approaches. The number of patients was decided to be respectively two-fold higher higher in the CD4GT group and three-fold higher in the 2-months-off, 4-months-on group than in the CT group. This decision was based on an expected number of patients stopping the structured treatment interruption strategies, with consequences in terms of power for per-protocol analyses, and on expectation of higher variability in both structured treatment interruption groups. The sample size calculation was based on the third primary outcome (CD4 count), and on the intention-to-treat approach. The main hypothesis was that 70% of the patients would have a CD4 count greater than 350 per mm3 at month 24 in the CT group.
Each structured treatment interruption strategy was to be regarded as non-inferior to the CT strategy if the lower bound of the 95% CI for the difference between the structured treatment interruption group and the CT group was not lower than -15%. Assuming a one-side type-I error of 5%, the sample size of 105:210:315 patients randomised ensured a 80% power to demonstrate the non-inferiority of a structured treatment interruption strategy. Assuming that 70% of the patients included would meet the randomisation criteria after at least 6 months in the pre-randomisation phase, and that 5% of the patients could be lost to follow-up, the total sample size was calculated to be 840.
Statistical analysis
An interim superiority analysis of the first and second primary outcomes (mortality and severe morbidity) was to be made when 50% of randomised patients had reached month 24. It was decided a priori that a structured treatment interruption arm would be stopped if mortality or severe morbidity was significantly higher in this arm than in the CT arm, using the intention-to-treat approach.
On June 23, 2005, the independent data safety monitoring board recommended to bring forward the interim analysis, on the basis of an overall severe morbidity rate that was higher than expected. This analysis was reviewed on Oct 28, 2005 by the data safety monitoring board, who recommended that: the CD4GT arm should be stopped, because of a higher severe morbidity rate than the CT arm (p=0-001); and that the 2-months-off, 4-months-on and CT arms should be continued until trial termination. From Nov 1, 2005, patients included in the CD4GT arm were prescribed a continuous treatment. We present data collected until Oct 31, 2005, within the CD4GT and CT arms.
For mortality and severe morbidity, the CT and CD4GT groups were compared using time-to-first-event analyses, including Kaplan-Meier estimates, log-rank testing, and Poisson's regression. The association between patients' characteristics and occurrence of a first severe morbidity event or a first CD4 count below 200 per mm3 during follow-up was analysed with Cox's proportional hazard regression models. Patients were defined as lost to follow-up if: their last contact was before Oct 31, 2005, they were not known to be dead, and no further information on their vital status was available on March 31, 2006. All analyses were intention-to-treat. All reported p values are two-sided and are not adjusted for multiple testing. Statistical analyses were done with SAS 8.2 software.
Role of the funding source
The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
|
|
| |
| |
|
|
|
