The impact of mass deworming programmes on schooling and economic development: an appraisal of long-term studies

Background: Documents from advocacy and fund-raising organizations for child mass deworming programmes in low- and middle-income countries cite unpublished economic studies claiming long-term effects on health, schooling and economic development. Methods: To summarize and appraise these studies, we searched for and included all long-term follow-up studies based on cluster-randomized trials included in a 2015 Cochrane review on deworming. We used Cochrane methods to assess risk of bias, and appraised the credibility of the main findings. Where necessary we contacted study authors for clarifications. Results: We identified three studies (Baird 2016, Ozier 2016 and Croke 2014) evaluating effects more than 9 years after cluster-randomized trials in Kenya and Uganda. Baird and Croke evaluate short additional exposures to deworming programmes in settings where all children were dewormed multiple times. Ozier evaluates potential spin-off effects to infants living in areas with school-based deworming. None of the studies used pre-planned protocols nor blinded the analysis to treatment allocation. Baird 2016 has been presented online in six iterations. The work is at high risk of reporting bias and selective reporting, and there are substantive changes between versions. The main cited effects on secondary school attendance and job sector allocation are from post hoc subgroup analyses, which the study was not powered to assess. The study did not find any evidence of effect on nutritional status, cognitive tests or school grades achieved, but these are not reported in the abstracts. Ozier 2016 has been presented online in four iterations, without substantive differences between versions. Higher cognitive test scores were associated with deworming, but the appropriate analysis was underpowered to reliably detect these effects. The size of the stated effect seems inconsistent with the short and indirect nature of the exposure to deworming, and a causal pathway for this effect is unclear. Croke 2014 uses a data set unrelated to the base trial, to report improvements in English and maths test scores. The analysis is at high risk of attrition bias, due to loss of clusters, and is substantially underpowered to assess these effects. Conclusions: In the context of reliable epidemiological methods, all three studies are at risk of substantial methodological bias. They therefore help in generating hypotheses, but should not be considered to provide reliable evidence of effects.