Data from 7 studies with a concurrent control group were included in the meta-analysis. There were 5 studies of neonates (n=31,041), 6 studies of infants (n=19,191) and 7 studies of pre-school children (n=89,242). Two studies were of a before-and-after design, while an additional 4 concurrent studies introduced the intervention into control areas and reported before-and-after data. Data from one additional unpublished trial could not be obtained.
The validity scores ranged from 70 to 109, the intervention characteristic scores ranged from 32 to 49, and the cointervention scores ranged from 11 to 45. The reviewers felt that although all the studies had methodological flaws, none were so great as to invalidate the results.
Neonates.
The intervention significantly reduced total neonatal mortality. The RR was 0.73 (95% CI: 0.65, 0.82) and the RD was 27% (95% CI: 18, 35). No significant heterogeneity was detected (P=0.895). The meta-regression showed no significant association with any of the covariates.
The intervention significantly reduced pneumonia mortality. The RR was 0.58 (95% CI: 0.43, 0.78) and the RD was 42% (95% CI: 22, 57). No significant heterogeneity was detected (P=0.711). The meta-regression showed no significant association with any of the covariates.
Infants.
Significant heterogeneity was detected in the meta-analysis of total mortality (P=0.053). After excluding one study conducted in the Philippines, no significant heterogeneity was detected (P=0.17).
The intervention significantly reduced total infant mortality. The RR for homogeneous studies was 0.80 (95% CI: 0.72, 0.89) and the RD was 20% (95% CI: 11, 28). The meta-regression showed no significant association with any of the covariates. Studies with a higher methods score showed a lower effect, while studies with a higher intensity of intervention showed a greater effect.
The intervention significantly reduced pneumonia mortality. The RR was 0.64 (95% CI: 0.52, 0.80) and the RD was 36% (95% CI: 20, 48). Homogeneity was borderline (P=0.088). The effect size was greater in the first year of the intervention (P=0.07). In the second year, the effect size was greater with advancing years (P=0.02) and greater with an increased health service score (P=0.04).
Children aged 0 to 4 years.
Significant heterogeneity was detected in the meta-analysis of total mortality (P<0.001). The intervention significantly reduced total infant mortality. The RR was 0.76 (95% CI: 0.67, 0.86) and the RD was 24% (95% CI: 14, 33). Only the year was significantly associated with the effect size (P=0.04).
Significant heterogeneity was detected in the meta-analysis of pneumonia mortality (P=0.005). The intervention significantly reduced pneumonia mortality. The RR was 0.64 (95% CI: 0.51, 0.80) and the RD was 36% (95% CI: 20, 49). Only the year of the intervention was significantly associated with the effect size (P=0.05).
Funnel plots for each of the six outcomes showed no evidence of publication bias.
The results were similar after excluding each study year in turn. The results from trials with a concurrent intervention group were similar to those from before-and-after studies.