Vintzileos and Ananth recently expressed concerns about overinterpretation of findings from our recent study that may inappropriately influence delivery room practice and the potential role of misclassification bias concerning the indicator for autism diagnosis used in our study.

We observed an adverse association between autism diagnosis and labor induction/augmentation. We did not examine or implicate oxytocin, nor did we conclude that labor induction/augmentation or oxytocin causes autism. In our study and subsequent media interviews, we consistently proposed several possible explanations of the association, including the following: (1) underlying medical or obstetric conditions that make it more likely a mother will be induced/augmented; (2) events of labor and delivery that are associated with induction/augmentation; or (3) drugs and methods used in induction/augmentation. We emphasized the benefits to labor induction/augmentation, particularly among women with medical and obstetric conditions that increase the risk for stillbirth or neonatal morbidity, and we stressed that the findings should not change current obstetric practice.

The misclassification concern raised by Vintzileos and Ananth is rooted in a change in diagnostic criteria during the study years for a subcategory of autism spectrum disorder (ASD) called pervasive development disorder-not otherwise specified (PDD-NOS or atypical autism). The timing of this change resulted in an overdiagnosis of PDD-NOS between 1994 and 2000.

The implications on our findings are not straightforward because the definition of autism diagnosis in our study did not distinguish among subcategories of ASD (as discussed in the study limitations). If PDD-NOS was overdiagnosed, but the children classified as PDD-NOS fell elsewhere along the ASD spectrum, then the change in PDD-NOS does not affect our study. However, if the children misclassified as PDD-NOS did not in fact have ASD, then our study would overestimate autism prevalence for the indicated years. However, in model 6 from our paper, the estimated association between labor induction/augmentation and autism adjusted out the potential influence of any yearly trend in autism diagnoses, with the estimated odds ratios substantively unchanged from the previous specifications that did not account for birth year. Model 6 suggests that birth year does not confound the association between autism and labor induction/augmentation. Nevertheless, we limited the analysis to only children born in 1996-1998; however, the much smaller sample size introduced instability in the estimated coefficients.

We argue that further research is warranted to confirm the association within a larger dataset that could segment time periods and, if confirmed, to explore potential drivers of the association.