WHO/HIV/2013.33
© World Health Organization 2013
GRADE table: What ARV regimen t to start in children younger than 3 years old? Author(s): M. Penazzato Date: 2011-07-29 Question: Should an LPV/r-based combination ARV regimen versus NVP-based combination ARV regimen be used among infants living with HIV? Settings: Low- and middle-income countries Bibliography: Palumbo 2010, Violari 2012, EPPICC 2011, Achan 2012
Quality assessment
No of patients
Effect Quality Importance
No of studies
Design
Risk of bias
LPV/r-based Other Inconsistency Indirectness Imprecision combination considerations ARV regimen
NVP-based Relative combination (95% CI) ARV regimen
Absolute
Treatment discontinuation (follow-up 24 weeks; assessed by: any interruption of study drugs (virological failure, toxicity, TB medication, death)) 3 randomized trials no serious serious2 risk of bias1 no serious no serious indirectness3 imprecision none 45/251 (17.9%) 64/256 (25%) CRITICAL HR 0.56 101 fewer per ⊕⊕⊕Ο (0.41 to 1000 (from 56 MODERATE fewer to 139 0.75) fewer) 83 fewer per 1000 (from 46 fewer to 113 fewer) 119 fewer per 1000 (from 65 fewer to 164 fewer)
20%
4
30%4 Virological failure or death (follow-up 24 weeks; assessed by: 2 HIVRNA < 400 copies/ml or <1 log drop by 24 weeks) 3 randomized trials no serious serious2 risk of bias1 no serious no serious indirectness3 imprecision none 32/251 (12.7%) 61/256 (23.8%)
HR 0.54 102 fewer per CRITICAL ⊕⊕⊕Ο (0.38 to 1000 (from 47 MODERATE fewer to 140 0.78) fewer) 86 fewer per 1000 (from 40 fewer to 119 fewer) 125 fewer per
20%4 30%4
WHO/HIV/2013.33
© World Health Organization 2013
1000 (from 57 fewer to 173 fewer) Virological suppression (follow-up median 5 years; assessed by: <400 copies/ml at 12 months from treatment initiation) 1 no serious serious7 observational no serious no serious 5 6 studies risk of bias inconsistency indirectness none 30/42 (71.4%) 49/81 (60.5%) OR 1.39 75 more per ⊕ΟΟΟ (0.62 to 1000 (from VERY LOW 3.13) 118 fewer to 222 more) 82 more per 1000 (from 117 fewer to 258 more) 48 more per 1000 (from 87 fewer to 126 more) CRITICAL
50%
80% Immunological response (follow-up 48 weeks; measured by: Change in CD4% from baseline) 3 randomized trials no serious no serious serious3,9 risk of inconsistency8 bias1 serious10 none 251 256 -
mean difference 1.18 lower (2.86 lower to 0.5 higher)
⊕⊕ΟΟ LOW
CRITICAL
Growth (weight) (follow-up 48 weeks; measured by: change in weight z-score from baseline) 3 randomized trials serious1,11 no serious no serious inconsistency8 indirectness serious10 none 251 256 Mean difference 0.19 lower (0.61 lower to 0.23 higher) ⊕⊕ΟΟ LOW IMPORTANT
Growth (height) (follow-up 48 weeks; measured by: change in height z-score from baseline) 3 randomized trials serious1,11 no serious no serious inconsistency8 indirectness serious10 none 251 256 Mean difference 0.16 lower (0.32 lower to 0.01 higher) ⊕⊕ΟΟ LOW IMPORTANT
WHO/HIV/2013.33
© World Health Organization 2013
Drug-related adverse events (follow-up 48 weeks; assessed by: NIAID scake (grade 3 or 4)) 3 randomized trials no serious serious12 risk of bias no serious indirectness serious none 43/251 (17.1%) 50/256 (19.5%) RR 0.85 29 fewer per (0.58 to 1000 (from 82 1.27) fewer to 53 more) ⊕⊕ΟΟ LOW CRITICAL
7 fewer per 1000 (from 21 fewer to 13 more) 30 fewer per 1000 (from 84 20% fewer to 54 more) Switch to second-line therapy (follow-up median 5.9 years; assessed by: change three or more drugs for any reason or two drugs for treatment failure) 5% 1 observational no serious no serious no serious studies risk of bias inconsistency5 indirectness serious7 none 2/154 (1.3%)13 18/464 (3.9%)13 HR 0.26 29 fewer per IMPORTANT ⊕ΟΟΟ (0.06 to 1000 (from 36 VERY LOW 1.19) fewer to 7 more)
15 fewer per 1000 (from 19 fewer to 4 more) 73 fewer per 1000 (from 94 10% fewer to 18 more) 1 Two of the three trials were stopped earlier by the DSMB according to pre-specified criteria and, although unlikely. it is not clear whether this could have led to bias. 2 The PROMOTE trial (Achan 2012) introduces a significant degree of inconsistency, since it provides evidence of an opposite effect. However, this trial was a non-inferiority trial designed and powered to assess malaria outcomes rather than treatment response outcome, so this should be considered in looking at the overall measure of effect. 3 The study populations are different in terms of NNRTI exposure. Cohort 1 (Palumbo 2010) patients had all been previously exposed to NNTI, whereas in cohort 2 (Palumbo 2011), patients were all NNRTI unexposed. The treatment efficacy can therefore be generalized to both subgroups of the target population. 4 The range was considered by looking at NNRTI exposed and unexposed groups. 5 Only one study assessed this outcome (EPPICC 2011). 6 NNRTI-exposed and -unexposed children were included in this cohort study (30% had been exposed to maternal ARV). The cohorts included in this study were gathered from Europe. 7 Low power to detect small differences, since few children were started on an LPV-based regimen. 8 The PROMOTE trial (Achan 2012) introduces some degree of inconsistency, since it provides evidence of an opposite effect. However, this trial was a non-inferiority trial designed and powered to assess malaria outcomes rather than treatment response outcome, so this should be considered in looking at the overall measure of effect. 9 Patients enrolled in the trials were severely immunosuppressed, and it is unclear whether the same effect would be detected in infants starting treatment with good clinical and immune conditions. 2%
WHO/HIV/2013.33
© World Health Organization 2013
10 11
Trials were not design nor powered to establish a difference in immune response or growth parameters. Outcome measurement was unblinded, and potential biases arising from this cannot be ruled out. 12 Achan 2012 and Palumbo 2010 showed no difference between study arms, while in Violari 2012 there were more adverse events in the NVP arm. 13 Rates per 100 person-years of follow up.