Interesting, but from the abstract it looks too underpowered to allow firm conclusions.
Of the two stated primary measures, one was not associated with autism and the other was only very marginally significant (at p=0.05), so it is rather vulnerable to a reversal if the research is replicated.
I also wonder why they did simple case-control matching rather than propensity scores. The number of unobserved confounders you could imagine is high in an analysis like this.
It looks like this is too underpowered to get anything out of this. The 95% CI is 1.02-1.7. Considering that the great majority of statistical tests in studies like this have null results (but usually aren't reported), there is a very good, possibly more than 50% chance, that this is bogus.
I've basically stopped paying attention to studies like this. In replication studies in medicine and psychology, these kind of fishing expeditions have a very low chance of being successfully replicated.
Interesting, but given that DDT (of which DDE is a metabolite) was banned except for a few specific use cases, can we really do much about this now? Is this DDT/DDE from the few specific use cases were DDT is still legal (disease control of malaria) or is this from previous accumulations from the environment from when DDT was widely utilized?
It's probably the latter. From the first paragraph of the wikipedia page on DDE [1]:
Due to DDT’s massive prevalence in society and agriculture during the mid 20th century, DDT and DDE are still widely seen in animal tissue samples. DDE is [..] rarely excreted from the body, and concentrations tend to increase throughout life. The major exception is the excretion of DDE in breast milk, which transfers a substantial portion of the mother's DDE burden to the young animal or child
The breast milk thing makes me think: If a substantial portion of the DDT is removed from the mother through breast milk, what difference does it make for subsequent children? And could it be somehow exploited to help remove more of the risk?
Or, you should see a big difference in breast fed children vs those that had formula. I'm pretty sure that variable has been heavily researched without any significant difference.
I see your logic and it makes sense. My point was more that if we posit DDT in breast milk caused autism, we'd expect to see a difference in rates of breast vs bottle fed kids, which we don't. So I'm skeptical of it being a big player, but I'll admit I haven't looked into this area deeply.
Not sure if this is what you meant, but breast-feeding a child wouldn't increase the autism risk for that child -- the damage would have been done by then.
You would need to research the occurrence of autism among children of breast-fed mothers vs formula-fed (i.e. a cross-generational study). I'm not sure if that has been studied as extensively as you assume.
Breastmilk is the evolved complete nutrition for babies.
Various milk formulas are developed towards that goal.
You cant compare these with the assumption they do perform the same - it would be like using Soylent 2.0 as the control for adult dietary research.
It's important research because the chemical industry and their boosters are always trying to get it unbanned. Koch-funded libertarian mouthpiece Reason still runs near-constant attacks on Rachel Carson and Silent Spring, 55 years later.
Research isn't made important because it satisfies a political imperative, but because it advances knowledge in a reliable way.
This sort of narrowly significant result is just the sort of research that is vulnerable to p-hacking and other biases. I'll wait for some supporting research before accepting something so weak, no matter the politics.
I wonder what other pesticides are associated with autism?
You'd of course hope the FDA/EPA would research this [autism risk] before making a pesticide legal, but how could they? In actuality, they give a very high dose to rats/dags/rabbits for a short term and use that as their best approximation of safety[1].
Of the two stated primary measures, one was not associated with autism and the other was only very marginally significant (at p=0.05), so it is rather vulnerable to a reversal if the research is replicated.