Kristof is correct to note:
Frankly, these are difficult issues for journalists to write about. Evidence is technical, fragmentary and conflicting, and there’s a danger of sensationalizing risks.
But he falls into these very traps when suggesting that toxins play a major role in autism. Let me pick on two sentences. Try this one:
There are genetic components to autism (identical twins are more likely to share autism than fraternal twins), but genetics explains only about one-quarter of autism cases.
Kristof doesn't note that identical twins both are autistic ninety percent or more of the time (conditional on one of the twins being autistic), yet the concordance is much lower for fraternal twins. That militates in favor of genetic explanations, although the mechanics of transmission are poorly understood. It's wrong to cite genetics as explaining one-quarter of autism cases or to imply that genetics do not explain three-quarters. There are recent studies which look for correlated genes across autistics and find less than overwhelming results and perhaps this is what he has in mind. More accurately, there is a common problem with finding "simple" genetic markers for traits which are very likely or even certain to be genetic. The degree of correlation across genetic patterns we can find should not be taken as a measure of how many autistic cases — or any other condition — can be explained by genetics. By the way, here is one paper with a plausible genetic model of autism.
Kristof also writes: "Of children born to women who took valproic acid early in pregnancy, 11 percent were autistic."
Probably he is referring to Moore et.al. (2000), "A Clinical Study of 57 Children with Fetal Anticonvulsant Syndrome." A total of four (supposedly) autistic children were observed to produce this conclusion. What happened is that some mothers took a potentially dangerous substance during pregnancy, many of their children had problems — of a variety of kinds – and some of these problems ended up resembling some features of autism or at least were interpreted as such. It's unlikely those were four autistic children in the classic sense. The paper also gives no real information on its standard of diagnosis for autism or what it means by autistic traits. It's common that papers like this find some problems in children and simply call those children "autistic," then leaping to false overall conclusions.
There's also a paper on using valproic acid to treat autism. One possibility is that the mothers taking valproic acid already were more likely to have autistic children; more likely our entire body of knowledge on valproic acid and autism doesn't offer real information.
Cross-sectional studies, spanning decades of age groups, suggest a roughly constant rate of autism, even when environmental toxins are changing considerably over those lengthy time periods. Plenty of other studies relate autism clusters successfully to non-toxin factors, such as parental education or supply-side services or standards of diagnosis.
There are likely well over 50 million autistics in the world and most of them have not had significant exposure to the cited toxins. While there are some plausible heterogeneities within autism, it is necessary to ask whether "genes *or* toxins" is one of those and probably it is not.
Epigenetic factors have not been ruled out in autism but the most careful discussions recognize that the relevant epigenetic factors — if indeed any are important – are unknown and also need not fit our usual intuitions about what is harmful in terms of direct dosages. A different way to approach the question is to ask which environmental features raise the rate of mutation. That way the genetic and epigenetic explanations are at least potentially consistent.
I'm not defending the feeding of "toxins" to children, but on examination I think virtually all of the major specific claims in this Op-Ed — at least those about autism — are wrong.
Addendum: David Bernstein scores some telling points.