We challenge the conclusions of the metaanalysis by Wax et al, which reported that planned home births had higher neonatal mortality rates than hospital births and were therefore less safe. The metaanalysis includes poor quality studies, has a high risk of methods bias, and does not meet the Journal’s requirement to comply with metaanalysis of observational studies in epidemiology guidelines. For example:

• 1
  

  The outcome of a metaanalysis is highly dependent on which studies are included and excluded. In this case, there is no list of citations and of which studies were excluded and why.

  
  
• 2
  

  The quality judgment for each individual study should have been reported. For example, the study that was based on routine data for Washington State contributed the largest numbers of neonatal deaths but was at high risk of misclassifying unplanned home births as planned home births because this information was not recorded in the dataset. This study has other methods problems.

  
  
• 3
  

  The assessment of confounding was inadequate. The authors reported that the sensitivity analysis by quality did not change the findings but gave no details.

  
  
• 4
  

  All relevant available studies should have been included, and contact should have been made with authors where necessary. Funnel plots show that the decision to exclude studies that had not been published in peer-reviewed journals contributes to publication bias. This could explain the lack of heterogeneity that was reported. If the authors had chosen a random-effects model, this would have been more appropriate because of the high clinical heterogeneity in the included studies.

  
  
• 5
  

  There is no graphic summarizing individual study estimates with overall estimates. The authors have not reported which individual studies contributed to which metaanalyses.

We identified 8 studies that had data on overall neonatal mortality rates, not 7. We also identified several different definitions of neonatal death in the included studies. Some studies used the same definition as the authors, but others did not. If Wax et al had contacted the authors of the very large Dutch study and included their neonatal mortality data, then no difference in neonatal mortality rates would have been evident.

It is of particular concern that this study was published in this present form when it does not meet the criteria for publication set out by the Journal itself. We believe that the American Journal of Obstetricians and Gynecologists should withdraw this publication in view of the failure of the peer review process to pick up these fundamental and fatal flaws.