Reacción a "Reactions to study associating induced childbirth with poorer academic performance at age 12"

This article has many strong points that advise a thoughtful review, which should be agreed by a large and heterogeneous group, impossible in a few days. 

1. They provide links to the protocol and the statistical analysis plan, which would make it possible to check whether the study is free of selective reporting bias (presenting the analysis with the results that suit them). 
2. They claim to analyse all births. If true, they would be free of the suspicion of selection bias - having eliminated cases for a variable that is a simultaneous consequence of the intervention and their cognitive development. 
3. They are aware of a possible confounding of residual effects - the differences come from other variables that are impossible to adjust for, perhaps because they cannot be determined with sufficient precision to be able to adjust for them. 

That said, it is worth qualifying: 

1. They say they adjust for variables, such as maternal age or maternal education level, without detailing their functional relationship with the response, or how it was specified, or whether it was done ex post, again potentially falling into selective reporting bias. 
2. They acknowledge that they must interpret the results at the level of association, not causation ("our findings are limited to associations and do not necessarily imply causality"). But they consistently interpret it at the causal level (or seem to induce causality in the reader: "reduced long-term intellectual development in the offspring", "These longterm effects of induction of labor should be incorporated in counseling and decision making"). 
3. The observed effect is minuscule: 0.05-SD means that it is one twentieth of the natural dispersion between cases in school performance. Such a (hypothetical) low effect will have a low impact. And because it is so low, it is much more likely to be explained by artefacts such as those discussed above. 
4. Like most observational studies, the calculations they present (such as uncertainty intervals) are based on the random sample assumption, which they do not meet.