“This study investigates patterns of autism diagnosis in Denmark. It assesses whether autism rates are higher amongst individuals whose mothers worked in particular professions, either at any time in the mother's life or specifically around the time of the birth. The study found that mothers who worked in certain professions were more likely to have a child with autism spectrum disorder, with the sharpest difference seen for those who had worked in a military or defence role. This study is not an experiment - it was not conducted under controlled conditions, and so it is not necessarily the case that the mother's profession is the causal factor driving differences in autism risk. Additionally, autism is not universally diagnosed, and so some portion of the observed difference in risk may relate to differences in diagnosis rates.

“The study missed an opportunity by not considering occupations of fathers. This would have been valuable in helping to distinguish correlation from causation, particularly in differentiating a true causal effect of the exposure during pregnancy from an unrelated correlative factor. The reason is that correlative factors are likely to affect the distributions of mother's and father's employment similarly, and so if a stronger statistical association is seen for mother's employment profession, that would strengthen the evidence that the observed correlation is reflective of a causal effect. As it is, this finding should be seen as suggestive evidence, not definitive evidence.”

“The research paper (and the press release) both point out that previous studies investigating possible associations between some maternal occupations and autistic spectrum disorder (ASD) in children had indicated there might potentially be associations with some occupations, but that the results were not consistent. I’m not confident that this new study has moved knowledge very far from that position. It has some dubious data issues and statistical aspects, and some of it is too speculative about what might be causing differences in ASD diagnosis rates.

“The study involved, in all, over 110,000 mother-child pairs, which sounds impressively large. However, the figure that really counts in terms of the ability of the study to find associations is the number of cases of ASD, which was 1,702. The rest of the 110,000 were so-called controls, that is, mothers and children where the child was not diagnosed with ASD. They are there as a baseline for comparison, but the statistical uncertainty comes very predominantly from the ASD cases (and their mothers’ occupations).

“If only one comparison were being made, then 1,702 cases would probably be very adequate to give good statistical estimates. But in fact the comparisons reported in the paper are for individual occupation groups, or subsets of occupational groups, and the number of ASD cases in each of those groups is very much smaller (single digits in some cases). Thus the level of statistical uncertainty in each of the comparisons is often very large.

“This high level of uncertainty can mean that associations that really exist cannot be picked up in the study, because there’s just not enough data. It also means that the sizes of associations that appear to exist, from the study data, might not be known at all precisely enough.

“I think there may well be serious issues of definition, which matter. The data come from a 40-year period of births in Denmark, 1973 to 2012. During that period, definitions of ASD changed, criteria for diagnosing the condition changed substantially, and in fact the term ASD did not formally exist in classifications of diseases and disorders in the earlier parts of that period. But the statistical analysis in this study seems not to take into account the date of the child’s birth.

“There are also issues of definition for the occupational classes. It appears that, if a mother changed jobs from one category to another, then they would be included separately under all the job categories they worked in over the whole period involved (from 1 year pre-conception to infancy). So a mother who took jobs in several categories would be included in the data for each of those categories, even if they worked in one of them for only a short time.

“The research paper also points out an issue about maternity leave. Under certain circumstances, a mother could be employed in a job but never be present at work during periods of maternity leave. That would probably make it less likely that job-related factors are involved

“All this does rather muddy the picture, I’d say.

“The researchers explain their statistical evidence (which is quite complicated) for saying that there is a possible association between the mother having worked in ground transportation, public administration, judicial occupations, or the military and defence, and the child having an ASD diagnosis. But they then go on to give reasons why working in these specific occupational groups could actually lead to an ASD diagnosis. I think this is extremely dubious for several reasons.

“This is an observational study, and no observational study can provide really solid evidence that the exposure (working in one of those occupations) is actually the cause of the higher rate of ASD diagnosis. There will be many other differences between mothers who worked in a certain occupation, and mothers who did not, apart from the job they worked in, and these other differences may be the real cause of the difference in ASD diagnosis rates. The researchers did make some statistical adjustments to allow for other differences, but one can’t be sure that everything relevant has been taken into account.

“That’s why it’s misleading for the press release to say things like “Working in military or defence occupations before or during pregnancy increased the odds of a child being diagnosed with ASD by 59%, the study found.” (It says several other similar things.) This reads to me to be saying that the higher odds of ASD diagnosis if the mother worked in the military caused the higher odds of ASD diagnosis. We just can’t tell from this study what is causing what in that sense – it could be working in the military, but the cause could be something completely different (as the press release does eventually state).

“But I think there’s potentially a greater problem when the researchers start trying to explain their findings in terms of exposure to pollutants, or in terms of working in high-stress occupations. Though it’s plausible that some people in the military, or working some types of transport, have higher exposure to some pollutants, the researchers didn’t have data on which actual jobs and tasks the mothers were doing. People doing administrative or office work for the military or in ground transport might not be more exposed to pollutants than anyone else.

“And when it gets to jobs in public administration, where again nothing was known about which particular jobs or task were involved, the researchers don’t explain at all why they think that involves more stress than other job categories. Is this based on any empirical evidence, and if so, why don’t they say what that evidence is? This all looks very speculative to me, and even has echoes of past theories of causes of ASD, now generally rejected, which blamed how mothers related to their children.”

“This study relies on information gathered from existing databases such as the Danish Pension Fund Registry and the Danish National Patient Registry. These sources may be incomplete and/or lacking in detail. For example, the patient registry does not include diagnoses made in primary care, potentially missing many cases of autism. It is also well-recognised that autism is under-diagnosed in women and girls: in this study over 70% of the people with autism were male. Again, this means that potentially relevant information about autism cases in Denmark has been missed.

“The information from the Danish Pension Fund Registry does not provide enough detail for the authors to be able to pinpoint the occupations that are potentially linked with autism. For example, they conclude that mothers who were employed before conception up to infancy in 'ground transportation' may have an increased risk of having a child with autism,  but this is a broad category and will likely include many occupations in which there is very little exposure to exhaust fumes or particulate matter. In a similar fashion they suggest that roles in the judicial sector may increase risk due to the high-stress nature of these jobs - again this broad category will contain variety of roles with varying amounts of occupational stress.

“This is an interesting study and the authors’ conservative conclusion that future studies should investigate specific exposures is appropriate.”