On the stated association between labour epidural analgesia and risk of autism spectrum disorder in offspring
- PDF / 255,142 Bytes
- 2 Pages / 595.276 x 790.866 pts Page_size
- 26 Downloads / 174 Views
CORRESPONDENCE
On the stated association between labour epidural analgesia and risk of autism spectrum disorder in offspring Bryan A. Glezerson, MD, FRCPC MPH, FRCPC
. Vatsal Trivedi, MD, FRCPC . Daniel I. McIsaac, MD,
Received: 27 October 2020 / Revised: 7 November 2020 / Accepted: 9 November 2020 Ó Canadian Anesthesiologists’ Society 2020
To the Editor, We read with interest the retrospective study from Qiu et al. in the journal JAMA Pediatrics, examining the association between labour epidural analgesia (LEA) and the risk of autism spectrum disorder (ASD) in offspring.1 A collective concern about the clinical and socio-cultural implications of the report has been expressed by physician groups around the world, including in Canada.2 We welcome the opportunity to discuss certain challenges in the interpretation of this study. Confidence that any nonrandomized, retrospective study of treatment can produce a causal estimate requires accurate ascertainment of exposure and outcome, and robust control for sources of bias. We respectfully suggest that both misclassification of the outcome and inadequate control for indication bias may have led to an effect estimate that reflects unmeasured bias instead of true causal effects.
Bryan A. Glezerson and Vatsal Trivedi contributed equally to the conception and writing of this manuscript. This letter is accompanied by an editorial. Please see Can J Anesth 2021; this issue. B. A. Glezerson, MD, FRCPC (&) Krembil Brain Institute, Toronto Western Hospital, University of Toronto, Toronto, ON, Canada e-mail: [email protected] V. Trivedi, MD, FRCPC Interdepartmental Division of Critical Care Medicine, University of Toronto, Toronto, ON, Canada D. I. McIsaac, MD, MPH, FRCPC Department of Anesthesiology and Pain Medicine, University of Ottawa and The Ottawa Hospital, Ottawa, ON, Canada
Non-randomized studies of treatment are at risk of indication bias. In the Qiu et al. study, even after adjustment, limitations in available data and analytic choices suggest that bias likely remains. As little is known with certainty about the causal risk factors for ASD, many relevant variables likely were not measured. Since both groups differed significantly across measured variables at baseline, it is likely that unmeasured variables were similarly unbalanced. While propensity-based techniques led to balance in measured variables, these techniques cannot address limitations in variable coding. Many variables were dichotomous, including a composite of ‘‘history of comorbidity’’.1 This will not reflect differences in disease severity,3 or the likelihood of unique associations of different comorbidities with the outcome. While Qiu et al. calculate that a single unmeasured variable would at least require a 1.76-fold relative association with exposure (i.e., LEA) and outcome (i.e., ASD) to nullify their findings, we submit that the multi-dimensional limitations of the data set and analysis create far more concern that the effect estimate is biased away from the null, compared with any rea
Data Loading...
