Samenvatting
Background
Measles Vaccination (MV) has been investigated for having beneficial non-specific effects on health outcomes for 40 years. However, evidence comes from observational studies and few randomised trials, all of which were deemed at high or moderate risk of bias. Most studies derived treatment effects with no or insufficient adjustment for possible confounding factors.
Methods
Data was collected within the Pneumococcal Vaccine Trial in The Gambia. Propensity scores were generated from logistic regression containing potential confounders: demographic characteristics, health indicators, vaccine profile, father’s education, and geographical information. Hazard ratios from Cox regression models were adjusted for these covariates by PS matching and PS weighting. ATE and ATET were estimated along with hazard ratios for mortality as primary and radiological pneumonia and hospitalisation as secondary outcomes. Sensitivity analyses considered standard Cox regression analyses, the impact of missing data, and residual confounding. Results ATE and ATET estimations using both PS matching and PS weighting methods agreed
that the absolute risk of death was reduced by approx. 0.5–0.8% (95% CIs: 2% reduction to 0.8% increase) due to MV. The hazard of death was estimated to be lowered by 14–25% for those receiving MV by their first year vs. those who did not (95% CIs: 46% reduction to 20% increase). Regression adjustment and MI models yielded comparable effects of reduction. Comparable small effect sizes and wide CIs were obtained for the secondary outcomes. Small E-values indicated that residual confounding could explain away the observed associations.
Conclusions
Results obtained from diverse methods using different outcomes are remarkably
consistent by pointing in the same direction of modest beneficial non-specific effects of MV. As such, findings are in line with those from randomized trials. However, given the weak evidence for a modest treatment effect and the possibility of residual confounding, no causal link can be established between MV and non-specific beneficial health outcomes.
Measles Vaccination (MV) has been investigated for having beneficial non-specific effects on health outcomes for 40 years. However, evidence comes from observational studies and few randomised trials, all of which were deemed at high or moderate risk of bias. Most studies derived treatment effects with no or insufficient adjustment for possible confounding factors.
Methods
Data was collected within the Pneumococcal Vaccine Trial in The Gambia. Propensity scores were generated from logistic regression containing potential confounders: demographic characteristics, health indicators, vaccine profile, father’s education, and geographical information. Hazard ratios from Cox regression models were adjusted for these covariates by PS matching and PS weighting. ATE and ATET were estimated along with hazard ratios for mortality as primary and radiological pneumonia and hospitalisation as secondary outcomes. Sensitivity analyses considered standard Cox regression analyses, the impact of missing data, and residual confounding. Results ATE and ATET estimations using both PS matching and PS weighting methods agreed
that the absolute risk of death was reduced by approx. 0.5–0.8% (95% CIs: 2% reduction to 0.8% increase) due to MV. The hazard of death was estimated to be lowered by 14–25% for those receiving MV by their first year vs. those who did not (95% CIs: 46% reduction to 20% increase). Regression adjustment and MI models yielded comparable effects of reduction. Comparable small effect sizes and wide CIs were obtained for the secondary outcomes. Small E-values indicated that residual confounding could explain away the observed associations.
Conclusions
Results obtained from diverse methods using different outcomes are remarkably
consistent by pointing in the same direction of modest beneficial non-specific effects of MV. As such, findings are in line with those from randomized trials. However, given the weak evidence for a modest treatment effect and the possibility of residual confounding, no causal link can be established between MV and non-specific beneficial health outcomes.
| Originele taal-2 | English |
|---|---|
| Kwalificatie | Master of Science |
| Toekennende instantie |
|
| Begeleider(s)/adviseur |
|
| Datum van toekenning | 16-dec.-2020 |
| Status | Published - 2020 |
| Extern gepubliceerd | Ja |