Field of Science

Autism and low birthweight: study--and quote--limitations

Which headline would you choose?
The story's out there: autism and low birth weight linked. But the headlines and reportage vary considerably. The usual conflict of interest is at work here: If you're a science writer interested in accuracy about the research, your headline might look like this:

If you're a headline writer hoping to write eyeball magnets that unremittingly lead to mouse clicks, then the headline might look like this:

Or this, which takes it a step further

Which one would you click on, if you were (a) pregnant, (b) parent or relative of a low-birthweight baby, or (c) just generally curious? I think we all know the answer.

But the only headline--and indeed, the only story--that accurately reflects and presents the study's findings is the first one. The study itself, published in Pediatrics, is not open access, but you can find the abstract here.

Study background
The report in question is of a long-term (longitudinal in science lingo) study that followed low-birthweight infants from birth through age 21, all born between 1984 and 1989. The cohort was "periodically assessed" over the years. Screenings when the cohort members were 16 turned up almost 19% who screened positive for an autism spectrum disorder (ASD). Of these 117 teens, 40% were not available for assessment at age 21, the assessment used in the current study. The remaining 70 were assessed, and 16% (11 individuals) found to have ASD.

The researchers also had about 81% of their original cohort who did not screen positive for ASD at age 16. They "selected" 23.5% of them (119) for assessment at age 21. Three of these, who had not been identified as having an ASD at age 16, were identified as having one at age 21. The final total of individuals who screened positive for an ASD at age 21 was 14. From their data, the researchers extrapolated an ASD rate among the low-birthweight population of about 5%.

Is it five times or two times the general population?
First, I'll point out that the study doesn't focus on gestational period but on birthweight, targeting people who were born weighing 4.5 lbs or less. There is not an indication that prematurity was used as a parameter, and references to prematurity, while intuitive, are not an actual reflection of this study. Prematurity and autism have been linked already.

The Big Find of this study is supposed to be that the authors identified a rate of autism of 5%, about five times the 0.9% or so estimated for the general population regardless of birthweight. But the Big Issue with the study is that the authors assessed the targeted population of low-birthweight people using autism assessments that typically aren't administered willy-nilly to the general population. In other words, they took a number, 0.9%, derived from the general population of people who are not all specifically tested for autism and used that as a measure against a specific population of people all of whom were specifically tested for autism.

It's entirely possible--and a study from Korea bears this out--that estimates for the general population are low. The Korean study, which did involve directly testing a very large segment of the general population (about 55,000 children) identified an autism rate of 2.6%. In other words, based on that rate, autism among very low-birthweight children is about twice that of the general population. Instead of using this very recent Korean data, however, the authors use data from the Centers for Disease Control from 2006, presumably with the rationale that those data would best reflect the US-based cohort for their study.

The study's lead author has been quoted as saying that she thought the rate would be "about double," rather than five times. She may have been right about that.

Is it autism or ADHD or...?
A total of 14 people in this study screened positive for ASD, three of them who originally screened negative for it at age 21. What has gone largely unreported--and unheadlined and uncommented--is the fact that almost half of this group--six people--had "co-morbidities," other disorders such as ADHD or phobias. The symptom crossover among these diagnoses is significant.

A problem with quotes
The description of autism is always an issue in the news media. While we've certainly begun moving away from using words like "disease" or "tragic," descriptives still prevail that don't fit with the personal experiences of autistic people or family and friends of autistic people. It's hard in a news story to get across the nuance of a developmental difference. Indeed, it took longform writing for Amy Harmon to convey it in her excellent NYT piece about a man with mid-intensity autism. So, I understand when the news media resorts to shorthand to discuss ASDs. What I'm starting to wonder about is the researchers involved in ASD studies.

Conflict of interest statement: I am a non-neurotypical parent of non-neurotypical children, one of whom has a diagnosis of autism and another of whom has a diagnosis of "provisional" Asperger's. My personal experience with autism means that I take quotes like the following quite personally:
Those with mild autism spectrum disorders, such as people with Asperger's syndrome, "may only want to talk about one subject," Pinto-Martin said. "They're perfectly functional. They can go to college. They can hold a job."
Or this one:
Lower functioning people with autism spectrum disorders may have no social interactions at all. "You may see them sitting in a corner, rocking back and forth," Pinto-Martin said.
In autism communities, using phrases like "low-functioning" and "high-functioning" is very controversial, and they are considered to be loaded terms. High and low function depend entirely on context; for example, I'm probably pretty low functioning when it comes to communicating in Chinese, but my level of functioning in English is pretty high. How well I function with either form of communication depends entirely on who my interlocutor is.

People seem to have a perception about Asperger's that the news media--and shows like Glee--probably perpetuate. It comes across as almost quaint, a quirky little thing to have that's adorably absent-minded professory, reminiscent of the Leary family in The Accidental Tourist, perhaps with a soup├žon of genius mixed in. It's not.

Asperger's carries with it deficits in social communication that can lead and have led to emotional and physical torture of Aspies in childhood and, despite the assertion in the quote above about jobs, difficulties in later life both in higher education and in job seeking. Their cognition is about average, on average--Aspies are not all walking geniuses in mismatched socks--yet underlying that averageness in all social interaction contexts is a hidden, exhausting struggle for control of the senses and the hard work of trying to achieve some semblance of "norm" in the interaction. That perceived "function" for an Aspie represents the intensely difficult work of overcoming a disability to accommodate the expectations of society.

Regarding the quote about "lower-functioning" autistics, I say it again: Behavior is communication. Rocking back and forth is a communication. It self soothes, from which anyone should be able to infer that the rocker is feeling anxiety. But that quote, that vision of "lower-functioning" autistics sitting in a corner, never interacting socially? I invite people to the reality of the intensely autistic by visiting here and here, as a start. Intensely autistic people do communicate, sometimes even in ways that non-autistic people with no experience outside their own norm can understand.

The quotee in that article is the lead author of the study, Jennifer Pinto-Martin. She also references premature babies in the piece, although the paper's focus was low birthweight, not specifically prematurity (extreme prematurity has already been associated with autism). In the article, she speculates that the medical advances that keep premature infants alive may be "contributing to the increased prevalence of autism." Of course, it could also be a million other things, from the causative agent in an early birth to agents we've tried to use to prevent early birth. What is unclear is whether or not she refers to the prevalence reported in the study, or if she's referring to an overall increase, the explanations for which remain controversial.

Pinto-Martin says that the next step is to follow up on a hypothesis that brain injury at birth may play a role in an autism diagnosis. The MSNBC piece is a little confusing because it states that brain ultrasounds are "now common" in premature babies, but it's unclear if they were common for this particular birth cohort, born in the 1980s. According to Pinto-Martin, her group will now check records of brain hemorrhage or injury data in the cohort to look for associations.

I close with two quotes, the first from an outside expert, Craig Erickson of the Indiana School of Medicine, quoted in the MSNBC piece. Referencing the longitudinal nature of the study, he says, "By [21 years of age], you really know if people have autism or not." Most parents notice differences before a child is two years old. It is odd that three of the cohort were not identified as having an ASD at age 16 but were identified as having one at age 21.

The final quote is one from Dorothy Bishop, professor of developmental neuropsychology at Oxford University. The quote first appeared in the BBC News, but I found it here. Bishop notes:
The association looks real, but nevertheless, most low birthweight children don't have autism, and most children with autism don't have low birth weight.
[Image credit: U.S. Navy, via Flickr.]


  1. Haven't read the paper, but it does seem as though, from what you're saying, they kept screening at different ages. Given that none of these measures are perfectly reliable, if you repeatedly screen the same people, you'll inevitably eventually get a higher rate of autism than if you only screen once.

    The main issue, as you say, is that comparing prevalence rates across studies with different methodologies is not particularly informative because different methods give different prevalences.

    We've had the same thing with the California autism twins study, where the headline was that non-identical twins showed much higher concordance than previous studies looking at sibling concordance - leading everyone to get very excited about prenatal environment as the cause of autism. Then a couple of weeks later, another study came out suggesting that sibling concordance was actually a lot higher than previously estimated too.

    Point being, these numbers are interesting from a practical point of view perhaps, but without a control group assessed using the same measures, they're not very informative about causes.

    So basically, I've just reiterated all of your points!

  2. There are a lot of extrapolations you make based on opinion rather than fact not having actually read the entire article. I have discussed this with the lead author and received the entire text before even the abstract was published.

    First every child in the population area with a birth weight less than <2000 grams was identified from hospital records. All these cases were periodically followed up from birth to 21 years. All the cases at 16 who were diagnosed with an ASD were selected for followup. All other cases at 16 were adminstered an autism screening tool (Social Communication Questionnaire)and those with even the most liberal scores were followed up. 24% of those with negative scores were also selected for further study to insure not having any possible cases falling through the cracks.

    All the 31 diagnosed with an ASD were diagnosed using gold standard diagnostic tools (ADOS, ADI-R) adminstered by trained clinicians. Many of the individuals identifed could not be reached for further contact over the many years for many reasons including moving out of the three county area in central New Jersy. This is common in all epidemiolgy studies conducted over many years that attempt to followup on the original participants. The Dorothy Bishops comment you quoted 'The association looks real, but nevertheless, most low birthweight children don't have autism, and most children with autism don't have low birth weight.

    Of course most low weight babies don't have any developmental problems whether it is autism, mental retardation, learning disorders, cerebral palsy etc.

    It's a foolish analogy which implies that low birth babies are not at any risk for any developmental problem.

    Next time, before pontificating, try reading the actual paper.

  3. Nothing you've posted here is information I didn't already have, having the paper details in hand. Also, the final quote is not a "foolish analogy." It is a statement of fact. Nothing you've posted here counters the points I've made in the post, either, primarily about the comparative between a carefully monitored population and a general extrapolation from epidemiological data and the characterization of autistic people. Nothing in the post implies that low-birthweight babies are without risk.

  4. Also, I should add that I don't blog research papers without having the actual paper itself in hand.

  5. Furthermore, since you quoted Dorothy Bishop on the low weight baby you might be interested in her comments on the Korean study you biblically quoted. Comments that you will surely disagree with.

    Dorothy Bishop does a wonderful job acting as a sort of Devil's advocate when controversial studies are published giving her own unique and important take on the study.

    As far as the low birth weight study is concerned, unlike the Korean study, that study was a longitudional study with 3 million dollars of NIH funding and these childen were followed periodically from birth until 21 years of age.

    As far as being a 'foolish analogy'let me make another foolish analogy 'Not all low birth weight children have a developmental handicap and not all children with a developmental handicap are low birth weight babies'.

  6. You appear to be taking something personally here, but I can't help that. It wasn't a "biblical" quoting of the study but an observation that (a)it exists, (b) it involved targeted evaluation of a large population, and (c) it offers a different value for prevalence. I was aware of the critiques of the Korean study when it came out and have already seen the piece. I agree with her that diagnosis is extraordinarily messy, as the false-positives and false-negatives from the paper current under discussion can only emphasize.

    Golly, I'm totally impressed with the $3 million in NIH funding, especially given the clarity of the findings.

    Your final graf also serves only to emphasize the points from the post, so thank you.

  7. PS again. I have to correct you on another error you made in your analysis. You said 'Screenings when the cohort members were 16 turned up almost 19% who screened positive for an autism spectrum disorder (ASD)'.

    Nonsense. The turned up positive not for an ASD diagnosis but passing a treshold for an autism screening tool.
    Autism screening tools are not diagnostic tools and to claim almost 19% who screened positive for an autism spectrum disorder is flat out wrong.

  8. No Emily I have no personal agenda against you. But your followers need to understand you have a genetic determinist bias that shows in all your blog posts.

    Name one genetic variant that is specific to and 'causes' autism and I'll be glad to show you where you wrong.

  9. Fun game of gotcha you're trying to play. From the paper:
    "At 16 years (n _ 623), adolescents were screened for ASD using a wide net (previous professional diagnosis of an ASD or a score above a liberal cutoff on the Social Communication Questionnaire or the Autism Spectrum Symptoms Questionnaire). At 21 years (n _ 189), 60% of screen positives and 24% of screen negatives were assessed for diagnoses of ASD by the Autism Diagnostic Observation Schedule or the Autism Diagnostic Interview–Revised."

    And "The fractions of the 2 screening groups with ASD (14.3% in screen-positives and 2.5% in screen negatives) were weighted by fractions of screen-positives and screen-negatives among the adolescents (18.8% and 81.2%, respectively)."

    I am using their own terminology here.

  10. Also, if you continue to disregard the actual points of the piece, I won't have further response. Thanks.


Comments for this blog are closed.

Note: Only a member of this blog may post a comment.