July 18, 2024

expert reaction to study of prenatal dietary patterns and autism in children

A study published in JAMA Network Open looks at prenatal diet patterns and autism in children. 

Dr Rosa Hoekstra, Reader in Global Perspectives on Neurodevelopmental Disabilities, Institute of Psychiatry, Psychology and Neuroscience (IoPPN), KCL , said:

“The authors are more nuanced in their paper than in the summary provided. In the full paper they do not just report a statistical association between adherence to healthy maternal prenatal dietary pattern and a reduced likelihood of their child having an autism diagnosis (in the MoBa cohort) or fewer social communicative difficulties (in the ALSPAC cohort), but also a link between healthy maternal prenatal diet and increased restricted repetitive behaviours in the MoBa cohort children when they were 8 years old. These latter results, not reported in the summary, are in the opposite direction of the main headline findings.

“The authors report in their paper that they planned the statistical analyses in advance, but unfortunately they do not provide a link to the pre-specified statistical analysis plan. It is now established good practice to publish an analysis plan in advance, so readers can see the original plan and compare it with the final published research. Without access to this plan, it is difficult to establish whether all analytic decisions were planned out in advance. In a complex study like this lots of small statistical decisions are made, for example in deciding which covariates to include in the statistical model, how to exactly measure ‘low’, ‘medium’ or ‘high’ adherence to a healthy diet, etc. Each of these decisions can influence the outcome of the analyses presented, and if some of these decisions were not specified in advance but made later, it increases the risk that the findings reported here are spurious.

“The authors tried to capture ‘healthy diet’ through analysing data from a food intake questionnaire. Different questionnaires were used in the MoBa and ALSPAC cohorts and the data collected in these questionnaires behaved a bit differently in the two samples (which is unsurprising, as different questions were asked and the participants are from different countries with somewhat different dietary habits). Rather than looking at intake of one specific food group the authors wanted to create a measure that assesses healthy food intake overall. They tried to do this by using a method called factor analysis and extracting the loadings on one factor, which they conceptualised as measuring an overall ‘healthy prenatal dietary pattern’. The supplementary materials of the paper give some more detail on this. When looking at the factor loadings there are some slightly funny inconsistencies – wholemeal bread and granary bread loaded negatively in the ALSPAC data, so a person indicating they ate wholemeal bread would get a lower healthy diet score than those eating white bread in this sample. The authors nevertheless proceeded with this measure and divided all participants up into three equal groups, which they indicated as showing ‘low’ ‘medium’ or ‘high’ level of adherence to a healthy diet. When creating such groupings based on a newly developed factor score you want to have some confidence that what you are measuring is valid. The authors present in eTable 3 estimated nutrient intakes in both cohorts for these ‘low’, ‘median’ and ‘high’ adherence to healthy prenatal dietary patterns. I would have expected there to be a clear delineation and differences in nutrient intake between these groups, especially between the low vs high adherence groups. This is not the case: the ‘high’ adherence group has lower estimated fibre intake than the ‘low’ group, and the ‘high’ group has greater kilocalories than the ‘low’ group. (I note all groups in the ALSPAC data have remarkably low overall estimated kilocalories, with total energy intake well below the 2000 kilocalories per day standard for women.)  Based on the data presented in eTables 1, 2 and 3 I am not fully convinced that their measurement of a ‘healthy prenatal dietary pattern’ is valid, and that there grouping of ‘low’ medium’ and ‘high’ adherence to this diet is meaningful.

“A healthy diet is associated with many favourable health and wellbeing outcomes, but this paper does not provide evidence that it is an important factor associated with the likelihood of having an autistic child.”

‘Healthy Prenatal Dietary Pattern and Offspring Autism’ by Catherine Friel et al. was published in JAMA Network Open at 16:00 UK time on Thursday 18^th July.

DOI: 10.1001/jamanetworkopen.2024.22815

Declared interests

Dr Rosa Hoekstra: I have no conflicts of interest to disclose. My research is funded by the National Institute for Health and Care Research (NIHR).