Table 2.
Quality Assessment of Included Studies
| No. | Author/year | Selection bias | Design | Confounders | Blinding | Data collection methods | Withdrawal and drop outs | Intervention integrity | Analysis appropriate to question | Overall assessment |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Aboud (2009) | Participants are likely to be representative of the target population; 72% of screened individuals agreed to participate. MODERATE | Randomized Controlled Trial, doubleblinded. Randomization described (separate publication). STRONG | There were important differences between groups (HIV status); this was accounted for in the design and analysis. STRONG | Double blinded, placebo controlled. STRONG | Data collection tools were valid; Data collection tools were reliable. STRONG | Reason for loss to follow up provided (premature closure of study). WEAK | More than 80% of participants received the intervention of interest. Consistency of interventions was measured during frequent follow up visits. Unclear if participants received an unintended intervention. MODERATE | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; Can’t tell whether the intervention allocation is based on intention to treat. MODERATE | MODERATE |
| 2 | Fawzi (2004) | Participants are likely to be representative of the target population; can’t tell what percentage of screened individuals agreed to participate. MODERATE | Randomized controlled Trial placebo controlled, randomization method described. STRONG | Unlikely that there were major differences between groups prior to the intervention; baseline characteristics between intervention and placebo group similar. STRONG | Outcome assessors and participants unaware of intervention and exposure status. STRONG | Data collection described. Outcomes were reviewed and confirmed by study investigators. STRONG | Numbers lost to follow up not reported. Instead compliance to treatment reported. WEAK | 79% of participants received the intervention of interest for the duration of the study. MODERATE | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; Analysis based on intention to treat. STRONG | MODERATE |
| 3 | Jamieson (2012) | Participants are likely to be representative of the target population; 79% of screened individuals agreed to participate. MODERATE | RCT; Study described as randomized; Randomization method described; Randomization method appropriate. STRONG | No important differences between groups prior to interventions; Most confounders were controlled. STRONG | Outcome assessors unaware of intervention and exposure status; Can’t tell whether study | Data collection tools valid; Data collection tools were reliable; Internal checks in place to ensure consistency of data. STRONG | Reasons for loss to follow up not described. 23% loss to follow up at 24 weeks; By 48 weeks 20% loss to follow up in the maternal | More than 80% of participants received the intervention of interest. The consistency of the interventions was measured during frequent follow-up visits. Can’t | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; | STRONG |
| participants were aware of the study question. MODERATE | ART group; 20% in the infant nevirapine group and 19% in the control group. MODERATE | tell whether participants received an unintended intervention. STRONG | Can’t tell whether the intervention allocation is based on intention to treat. MODERATE | |||||||
| 4 | Kesho Bora (2012) | Participants are likely to be representative of the target population; can’t tell what percentage of screened individuals agreed to participate. MODERATE | Randomized Controlled Trial: Randomization method described. STRONG | No important differences between groups prior to interventions; Most confounders were controlled. STRONG | Outcome assessors and participants were not blinded. WEAK | Standardized case report forms used; data collection tools valid; Data collection tools were reliable; Internal checks in place to ensure consistency of data. STRONG | Reasons for loss to follow up not described. Loss to follow up at 18 months after delivery were 92.1% and 89.0% in the triple ARV arm and AZT/sdNVP arm, respectively; at 24 months they were 86.1% and 84.7%, respectively. STRONG | More than 80% of participants received the intervention of interest. The consistency of the interventions was measured during frequent follow-up visits. Can’t tell whether participants received an unintended intervention. STRONG | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; Analysis based on intention to treat. STRONG | MODERATE |
| 5 | Kupka (2008) | Participants are likely to be representative of the target population; can’t tell what percentage of screened individuals agreed to participate. MODERATE | Randomized controlled trial, placebo and doubleblinded; randomization described. STRONG. | Unlikely that there were major differences between groups prior to the intervention; baseline characteristics between intervention and placebo group similar; STRONG | Outcome assessors unaware of intervention and exposure status; Can’t tell whether study participants were aware of the study question. MODERATE | Data collection described. Outcomes were reviewed and confirmed by study investigators. STRONG | Numbers lost to follow up reported; 22% lost to follow up in placebo group, 25% lost to follow-up in the intervention group (CD4, CD8. CD3). MODERATE | More than 80% of participants received the intervention of interest. The consistency of the interventions was measured during frequent follow-up visits. Can’t tell whether participants received an unintended intervention. STRONG | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; Analysis based on intention to treat. STRONG | STRONG |
| 6 | Li (2014) | Participants likely to be representative of target population; can’t tell about agreement to participate. MODERATE | Prospective cohort study. Nonrandomized. MODERATE | Not applicable | Not applicable | Data extracted from clinical records; comprehensive data quality controls in place such as double data entry systems; data collection tools were valid; data collection tools were reliable. STRONG | Not applicable | Likely more than 80% of women received the allocated intervention. The consistency of the intervention was measured. Can’t tell whether subjects received an unintended intervention. MODERATE | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; STRONG | STRONG |
| 7 | Liotta (2013) | Participants likely to be representative of target population; can’t tell about agreement to participate. MODERATE | Retrospective observational cohort study. Nonrandomized. MODERATE | Important confounders existed prior to the intervention; Confounders addressed during analysis. MODERATE | Not applicable | Data collected from electronic medical records in a controlled study setting. Data collection tools were valid; Data collection tools were reliable. STRONG | Reasons for loss to follow up not described. Predelivery loss to follow up was 9.8% in women initiating ART during prenatal care and 3.9% in women initiating ART for their own health. Postdelivery abandonment of care was 7.8% in women initiating ART during prenatal care and 5.4% in women initiating ART for their own health. STRONG | Likely more than 80% of women received the allocated intervention. The consistency of the intervention was measured. Unlikely that subjects received an unintended intervention. STRONG | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate. STRONG | STRONG |
| 8 | Marazzi (2009) | Participants likely to be representative of target population; can’t tell about agreement to participate. MODERATE | Prospective observational cohort (Mother infant pairs); no control group. MODERATE | Not applicable | Not applicable | Data collected from electronic medical records in a controlled study setting. Data collection tools were valid; Data collection tools were reliable. STRONG | Reasons for loss to follow up not provided. 8% of pairs at 6 months and 17% of pairs at 12 months were lost to follow up. STRONG | Likely more than 80% of women received the allocated intervention. The consistency of the intervention was measured. Unlikely that subjects received an unintended intervention. STRONG | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate. STRONG | STRONG |
| 9 | Marazzi (2011) | Participants likely to be representative of target population; can’t tell about agreement to participate. MODERATE | Retrospective observational cohort. MODERATE | The control group (68 persons who did not receive ART) may be significantly different from others who did receive ART. Reasons for nonprovision of ART not disclosed however enrollment was based on the provision of ART to all participants. WEAK | Not applicable | Data collected from electronic medical records in a controlled study setting. Data collection tools were valid; Data collection tools were reliable. STRONG | Not applicable | Likely more than 80% of women received the allocated intervention. The consistency of the intervention was measured. Unlikely that subjects received an unintended intervention. STRONG | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; STRONG. | MODERATE |
| 10 | Matthews | Participants likely to be representative of target population; | Prospective cohort. No randomization. MODERATE | It is unlikely that there were major differences between the two comparison groups (baseline characteristics were compared); | Not applicable | Study occurred in a controlled setting. Data most likely extracted | Reasons for loss to follow up not described; loss to follow up was low (3 and 7% at years | All participant received the appropriate intervention; the consistency of intervention was | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate. STRONG | STRONG |
| can’t tell about agreement to participate. MODERATE | important confounders such as age, CD4 cell count, viral load, and number of children were incorporated. STRONG | from medical records. Data validity and reliability not described. MODERATE | 1 and 5 and did not differ among women with and without pregnancy (7 vs 8%). STRONG | measured (and steps taken to address loss to follow up); can’t tell whether unintended interventions were received. STRONG | ||||||
| 11 | Melekhin (2010) | Participants likely to be representative of target population; can’t tell about agreement to participate. MODERATE | Retrospective cohort; no randomization. MODERATE | No important differences between groups prior to the intervention; Groups were compared through baseline characteristics and outcome characteristics. STRONG | Not applicable | Data collection described. Outcomes were reviewed and confirmed by study investigators. STRONG | Withdrawals and drop outs not discussed with the exception of median follow up. 8% loss to follow up in the discontinuation group. STRONG | All participant received the appropriate intervention; the consistency of intervention was measured (and steps taken to address loss to follow up); can’t tell whether unintended interventions were received. STRONG | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate. STRONG | STRONG |
| 12 | Nunn (2011) | Yes 1; Agreement to participate 64%. MODERATE | Double blind randomized placebo controlled trial; Method of randomization | No important differences between groups. STRONG | Outcome assessors unaware of intervention and exposure status; | Data collection tools were valid; | 44% of randomized women lost to follow up. | High level of loss to follow up 44%; Regular follow up visits for participants during the study. | Unit of allocation individual; Unit of analysis individual; | MODERATE |
| described; Method appropriate. STRONG | Can’t tell whether study participants were aware of the study question. MODERATE | Data collection tools were reliable. STRONG | Reasons for loss to follow up provided. WEAK | Can’t tell whether subjects received an unintended intervention. MODERATE to WEAK based on high loss to follow up | Statistical methods appropriate; Analysis based on intention to treat. STRONG | |||||
| 13 | Onen (2008) | Participants likely to be representative of target population; can’t tell about agreement to participate. MODERATE | Retrospective cohort. MODERATE | Can’t tell whether there were confounders identified prior to the intervention; multivariate analysis conducted for potential confounders. MODERATE | Not applicable | Data extraction most likely done from medical records. Can’t tell whether data collection tools were valid; Can’t tell whether data collection tools were reliable. WEAK | Not applicable | All participants received the interventions as scheduled; consistency of the intervention was measured (efforts made to trace missing data); can’t tell whether any unintended interventions were received. STRONG | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate. STRONG | MODERATE |
| 14 | Shapiro (2013) | Participants are likely to be representative of the target population; 58% of screened individuals agreed to participate. MODERATE to WEAK | Randomized Controlled Trial; randomization process not described. STRONG | Likely there were important differences between groups (CD4 cell counts); this was accounted for in the design and analysis. STRONG | Blinding not described; no reference about the awareness of participants to the research question. MODERATE | Can’t tell whether data collection tools were valid; Can’t tell whether data collection tools were reliable. No reference to data management. WEAK | Reasons for loss to follow up not described. At 6 months less than 10% of participants were lost to follow up across all groups; at 24 months less than 80% of women had completed the full | Almost 100% of participants received the allocated intervention; consistency of the interventions was measured; can’t tell whether subjects received an unintended intervention. STRONG | Unit of allocation individual; Unit of analysis individual; statistical methods appropriate; Can’t tell whether the intervention allocation is based on intention to treat. MODERATE | MODERATE |
| protocol in the randomization arms while more than 80% had completed the protocol in the control arm. However, vital status was known for almost 100% of participants across all arms. MODERATE | ||||||||||
| 15 | Songok (2003) | Yes 1; Almost 100% of participants agreed to participate. STRONG | Method not stated. WEAK | Likely there were important differences between groups (44% did not accept intervention treatment); can’t tell about control methods used for confounders. WEAK | Can’t tell whether assessors were blinded to intervention; can’t tell whether collection tools were shown to be reliable. MODERATE | No reference to data collection systems. Can’t tell whether data collection tools were valid; Can’t tell whether data collection tools were reliable. WEAK | Withdrawals reported with reasons for children; apparently 30 percent of children did not complete the study; loss to follow up for mothers not reported. WEAK | Less than 60% of participants received the intervention of interest; the consistency of the intervention was measured; can’t tell whether the participants received an unintended intervention. WEAK | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; likely not an intention to treat analytic approach (surviving children with unknown HIV status excluded from the analysis). WEAK | WEAK |
| 16 | Thomas (2011) | Participants are likely to be representative of the target population; 87% of screened | Open label, single arm clinical/intervention trial. STRONG | Not applicable no control group | Not applicable | Can’t tell whether data collection tools were valid; Can’t tell whether data | Reasons for loss to follow up partially provided; 20% of infants lost to | 84% of participants adhered to treatment; consistency of the intervention | Unit of allocation individual; Unit of analysis individual; | MODERATE |
| individuals agreed to participate. STRONG | collection tools were reliable. No reference to data management systems. WEAK | follow up; 21% of mothers were lost to follow up. MODERATE | was measured; can’t tell whether the unintended interventions occurred. STRONG | Statistical methods appropriate; analysis based on intention to treat. STRONG | ||||||
| 17 | Tuomala (2005) | Can’t tell whether participants are representative of the target population; Can’t tell what percentage of individuals agreed to participate. WEAK | Prospective cohort. MODERATE | Important differences occurred in terms of the type of ART provided. Variables controlled by analysis. STRONG | Not applicable | Utilized medical records. Data quality checks in place. Data collection tools were valid; Data collection tools were reliable. STRONG | Not applicable | Likely more than 80% of women received the allocated intervention. The consistency of the intervention was monitored and changes in treatment regimens acknowledged. MODERATE | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; STRONG | MODERATE |
| 18 | Westreich (2013) | Participants likely to be representative of target population; likely that the majority of participants agreed to participate. MODERATE | Prospective cohort. No randomization. MODERATE | Important differences between groups identified; potential confounders identified; most confounders controlled for in the analysis. STRONG | Not applicable | Used clinical records to extract data. Data collection tools were valid; Data collection tools were reliable. STRONG | Loss to follow up defined for data collection and analysis purposes in the methods section. Not applicable | Likely more than 80% of women received the allocated intervention. The consistency of the intervention was monitored. MODERATE | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; STRONG | STRONG |
| 19 | Zvandasara (2006) | Participants likely to be representative of target population; Can’t tell that the majority of participants agreed to participate. MODERATE | RCT. Randomization described. Method described. Method appropriate. STRONG | No important differences between groups. STRONG | Outcome assessors unaware of intervention and exposure status; Can’t tell whether study participants were aware of the study question. MODERATE | Self-reported data by participants of utilization of health services. Serum values for Vitamin A and HIV tests were conducted at laboratories that participated in external quality assurance programs. Unable to comment on data collection tools for validity and reliability. MODERATE | Duration of follow up curtailed due to economic reasons. 11% of HIV negative participants and 15% of HIV positive participants were lost to follow up at the end of 12 months. 84% of HIV negative participants and 65% of HIV positive participants were lost to follow up at the end of 12 months. MODERATE | 100% of participants received the intervention of interest. The provision of interventions to participants is described. The administration of the intervention was onceoff. STRONG | Unit of allocation individual; Unit of analysis individual; Statistical methods appropriate; Can’t tell whether the intervention allocation is based on intention to treat. MODERATE | STRONG |