select search filters
roundups & rapid reactions
factsheets & briefing notes
before the headlines
Fiona fox's blog

expert reaction to study investigating the association between grandmothers smoking during pregnancy and autistic traits in grandchildren

A new study, published in Scientific Reports, investigates the association between a female’s maternal grandmother smoking during pregnancy and grandchild being at increased risk of developing autism.


Prof. Peter Hajek, Director of the Tobacco Dependence Research Unit, Queen Mary University of London (QMUL), said:

“Several large studies examined possible links between maternal smoking during pregnancy and autism and their meta-analysis shows no link, so this finding, if real, would be surprising. Some caution is needed with interpreting the finding though. Examining all the combinations of smoking status of parents and grandparents with the gender of the children and four different traits is quite likely to generate a chance finding of borderline significance, especially when using a series of different adjustments and not adjusting for multiple comparisons. Data on children who were actually diagnosed with autism (as opposed to having certain linked traits) are much more important, but this showed no significant association. When adjusted for a range of variables (whose selection is not well explained), a relationship between smoking and autism was detected, but this was significant only for grandsons and so not tallied with the ‘traits’ finding where only granddaughters were affected. The study proposes an interesting epigenetic effect, but more data are needed to decide whether it exists. ”


Prof. Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“The authors of this study point out that it is important that the associations they found are “confirmed in other studies”. I couldn’t agree more, but let’s pick apart what they mean by that.

“Well, perhaps the associations won’t be confirmed. That might simply be because they don’t exist generally, and might just be a coincidence of the patterns in this particular set of data. But another relevant point is that the associations are complicated and, in most cases, not very strong statistically. To have any real chance of detecting them, one needs a big set of data running across a long timescale. The ALSPAC study, that these researchers used, is one of only a few worldwide that provide the resource for such an investigation. Can it be done elsewhere? I’m not so sure. Even in this large long-term study, it’s not clear that there were enough cases of diagnosed autism, particularly in girls, to provide clear answers.

“The strongest associations in this study were not in relation to diagnosed autism itself, but with traits that had been linked to autism in previous studies. That’s a bit indirect, and of the four traits studied, they found evidence of a link to the maternal grandmother’s smoking during pregnancy in only two. There was a statistically significant association between the maternal grandmother’s smoking and a diagnosis of autism overall, but in statistical terms the evidence for that association isn’t particularly strong.

“The authors are careful to point out that they can’t know from these results whether the associations are causal, that is, whether the grandmother’s smoking actually caused the grandchildren’s chance of having one of the traits related to autism to increase. That’s because the study is observational, and it might be the case that some other factors, to do with family circumstances perhaps, are related to the grandmothers’ chances of smoking, and completely independently have some effect on the grandchildren. These researchers did allow for some such confounding factors statistically, and these statistical adjustments did have quite large effects on the strength of the relationships they found, in both directions (making some associations stronger, others weaker). That shows clearly how complicated the picture is. Yet statistical adjustment like this can never be perfect, and of course the researchers can’t adjust for things on which they do not have data.

“So a complicated picture, and yes, one that would benefit from other studies that might throw more light on what, if anything, is going on. But such further studies won’t be easy, and won’t be quick.”


Dr James Cusack, Director of Science, Autistica, said

“The study applies careful analyses to data from a world-leading birth cohort project. But conclusions are drawn from ‘reported or possible’ autism diagnoses and ‘predictive traits’.  Also, although confounding factors are taken into account, there may be others not considered in the analyses.

“Further research is needed to understand whether grandmother’s smoking during pregnancy is linked to autism in grandchildren. This study alone does not indicate a causal link.”


* ‘Grand-maternal smoking in pregnancy and grandchild’s autistic traits and diagnosed autism’ by Golding et al. published in Scientific Reports on Thursday 27th April.


Declared interests

Prof. Peter Hajek: “No conflicts of interest in this field.”

Prof. Kevin McConway: No conflicts of interest.

Dr James Cusack: “One of the paper authors, Dheeraj Raj, has worked with Autistica in an autism community priority setting workshop. Otherwise, no conflicts of interest to declare.”

in this section

filter RoundUps by year

search by tag