Health

…

Summary of assessments
We used the Navigation Guide methodology to rate studies based on several metrics. The risk of bias within each study was assessed using the GRADE approach to grade study characteristics that can introduce systematic errors in the magnitude or direction of the results. We rated each study for risk of bias, including participant recruitment/selection, blinding during the study, exposure assessment methods, outcome assessment methods, methods to address incomplete data, selective outcome reporting, and conflict of interest. We ranked each study on each parameter: 1 indicated low risk of bias, 2 indicated probably low risk of bias, 3 indicated probably high risk of bias, and 4 indicated high risk of bias. We calculated an average bias score for each study. For the ‘blinding during the study’ domain, observational studies were rated as high risk of bias (score of 4) when knowledge of the outcome could influence exposure reporting. For instance, retrospective studies relying on maternal self-reports of acetaminophen use collected after a child’s neurodevelopmental disorder diagnosis were rated high risk due to potential recall bias. Prospective designs or biomarker-based assessments mitigated this bias in higher-quality studies.
Deviations from scoring—such as inconsistencies in study methodology, incomplete data reporting, or challenges in applying bias criteria—were addressed through a structured process. During the study selection and data extraction phase, studies were triaged by title, abstract, and full text; two reviewers (AB and DP) independently assigned a score for each Navigation Guide category. Any deviations, such as studies with atypical designs or potential biases, were flagged for further evaluation. To handle these deviations, we conducted sensitivity analyses to assess their impact on the overall findings. Specifically, we performed two analyses: (1) excluding the lowest-scoring studies to evaluate their influence on the results, and (2) re-weighting confounding domains to address potential bias over- or underestimation.

…