Perhaps you’ve seen headlines or even bothered to read past them in the recent reports linking autism severity and ultrasounds. As always with anything that’sbeen linked to autism, it’s best to look closely at the study itself to answer nagging questions like, “Oh my God, I had an ultrasound during pregnancy and what about autism severity?!” So below, a few possibly clarifying and potentially edifying observations.
The authors report finding a mathematical link between having a diagnostic ultrasound in the first trimester of pregnancy and increased severity of some autism symptoms in the resulting offspring, especially among children who are born with specific genetic differences that are associated with autism risk.
Ultrasounds have not been shown to cause autism
First, this is a mathematical link. This paper, by Sarah Jane Webb and colleagues and published in Autism Research, does not show that ultrasounds cause autism.
This paper’s findings relate only to first-trimester diagnostic ultrasounds
Second, these results are about diagnostic ultrasound in the first trimester. Women having these typically present to health care because of ominous or odd symptoms. The current indications for having a first-trimester diagnostic ultrasound include suspected ectopic pregnancy, pelvic pain, suspected multiple gestation, assessment for fetal anomalies and vaginal bleeding. Diagnostic ultrasounds in the first trimester are not routine. They mean someone thinks there’s a problem with the pregnancy or the dating of the pregnancy is unclear.
Dating a pregnancy using ultrasound is most accurate earlier in gestation and can prevent later problems, including unnecessary interventions that are based on incorrect dating.
Standard clinical practice, according to current information from the American Congress of Obstetricians and Gynecologists, is a second trimester ultrasound done at 16 to 20 weeks, which also happens to be the time when ultrasound can detect fetal sex. In this autism-ultrasound study, the authors do not report associations of their endpoints with second- or third-trimester ultrasounds.
Previous, rigorous studies have found no link between autism and ultrasound
Numerous studies—including some with an optimal randomized design—have found no link between ultrasound in any trimester and a host of outcomes, from heart disease to neurological differences.
In the introduction to their paper, Webb and colleagues describe two such studies:
Stoch et al. … did not find a relation between child ASD diagnosis and randomized prenatal exposure to a single second trimester ultrasound versus multiple second and third trimester scans (at 18, 24, 34, 38 weeks); nor to levels of adult autism traits in a primarily “neurotypical” population. Grether et al. … analyzed antenatal ultrasound exposure (primarily 2ndtrimester exposure) as a risk factor for ASD using medical data from Kaiser Permanente of Northern California Health Care System; there was no difference in number of exposures during the entire gestation or by trimester in 393 controls and 362 autism cases. Critical to the triple hit hypothesis and not addressed by previous reports, ultrasound as an exogenous stressor would have the most significant impact during the 1sttrimester and only in those with specific genetic risk factors.
In Grether et al., more than 25% of their population had first-trimester scans.
The “triple-hit hypothesis” is that three factors—environment, genetics and timing—interact as the determinants of autism. Given that those factors determine just about everything in development and that autism is a developmental condition with a large genetic component, that’s not exactly a groundbreaking idea. After all, no gene (or DNA sequence) is an island.
But that “triple hit” reference is the opening the authors needed for their current study rationale: Sure, ultrasounds haven’t been linked generally to autism, but what about ultrasounds (hit 1) at just the wrong time (hit 2) in pregnancies involving embryos/fetuses that have some genetic (hit 3) sensitivity or susceptibility?
According to this hypothesis, one might expect to see outcomes build as each candidate “hit” is added in. Let’s see how that works out.
Ultrasound is linked to a a less severe outcome on one measure
Let’s start by turning to the paper itself. The authors discuss much of the below, either offering rationales or acknowledging them as limitations.
The paper contains three main-text tables, each one reflecting the results of separate analyses. Table 1 takes the population as a whole (668 with no first-trimester ultrasound; 1,081 with a first-trimester ultrasound). Webb and co-authors looked at how first-trimester ultrasound relates to the results of nine evaluations, three each for (1) adaptive/cognitive, (2) social affective and (3) repetitive behaviors. Among each of the three measures for each of these three categories, some are parent reported and some clinician derived.
For adaptive/cognitive behaviors (eg, IQ measures), the authors found no significant link between autism and first-trimester ultrasound with scores on any of the three measurement instruments. For social affective behaviors, only a clinician-reported measure was linked to first-trimester ultrasound, which was associated with better scores for this outcome. The other two instrument scores showed no significant link.
Finally, for repetitive behaviors, again, only one measure—this one parent reported—was significantly linked to first-trimester ultrasound. The other two were not.
The paper findings are internally inconsistent
So for the first two hits—timing (first trimester) and ultrasound—the results are underwhelming and not internally consistent for direction (better social affective score, worse repetitive behavior score), parent vs clinician report or category of symptom.
Table 2 adds in a “hit” in the form of genetic variations that have been linked to autism. These variations are different numbers of sequence repeats at specific areas on a chromosome, known known as copy number variations or CNVs. Having more or fewer of these repeats in some cases has been associated with increased autism risk, although not everyone with these changes is autistic and by no stretch do all autistic people carry one or more of these CNVs. In this study, about 7-8% of the population carried relevant CNVs.
Here again, the authors looked at the same three categories, using the same set of measurement tools as before, but they compared only those autistic children who carried the CNVs and whose mothers didn’t (49 children) or did (84 children) have a first-trimester ultrasound.
In this case, Webb and colleagues got a single significant hit among the nine comparisons: a significantly lower nonverbal IQ among the children whose mothers had the ultrasounds. Nonverbal IQ was not among the significant differences highlighted in the Table 1 analysis. But when you cut your population down by 92% and add in a variable for the comparison, all kinds of things can happen.
Finally, their Table 3 adds in another “hit:” being male. Here, they looked only at autistic boys who carried the CNVs and whose mothers did (73 boys) or didn’t (38 boys) have first-trimester ultrasounds.
The authors report two significant results. One is a significantly different nonverbal IQ, lower in association with ultrasound. Given that 111 of 133 (83.5%) of their population for Table 2 were boys, one would expect some similar results between tables 2 and 3, so that’s no surprise.
With the added “hit” of being male, though, the nonverbal IQ values went up in both the ultrasound and no-ultrasound groups compared to scores for CNV boys and girls together. That doesn’t suggest that maleness increased the “dose response.”
The other significant result is for clinician-reported repetitive behaviors, with higher scores in association with ultrasound. But it’s not with the same scale that showed significance for the overall population, which was parent-reported.
Overall, the authors found different associations across the three analyses that don’t seem to suggest a pattern or a dose-response effect of going from ultrasound only (Table 1), to ultrasound plus CNVs (Table 2) to ultrasound+CNV+male (Table 3). Not even the same scales, categories or type of observer showed a pattern or a dose response of “hits.”
Ultrasound history is based on recall
The data about first-trimester ultrasounds come from women recalling when and why they had ultrasounds at a time when their children are four to 18-years-old. The first trimester also spans a time when critical windows for environmental influences can be extremely precise, as in the thalidomide example below. Both the recall and the breadth of change in the first trimester mean that this factor has a lot of squishiness to it.
The children included in this analysis had a wide variety of genetic syndromes, not just autism
Finally, here’s where those specific genetic differences linked to autism come in. If you have the patience to look at the supplementary data tables for this study, you’ll find information about those CNVs that are treated in these analyses as a uniform factor.
They are anything but uniform. For example, 11 girls with CNVs had a first-trimester ultrasound and 11 girls with CNVs did not. Between these two groups, only two girls have a CNV in common. The rest are distributed across several chromosomes, with a mix of duplications and deletions.
There were 111 boys with CNVs in the analysis, 38 without first-trimester ultrasound and 73 with one. Their CNV type and distribution are quite heterogeneous and involve sizes that can range across three orders of magnitude. The authors compared relative proportions of duplications and deletions and average CNV sizes in their groups and found no significant differences, but that seems largely immaterial against the greater question of how powerfully each individual variant influences relevant traits.
Treating them as a single factor assumes that duplications and deletions across almost all chromosomes and of any size have the same “hit” in the context of autism and are simply interchangeable. Biologically, that’s not tenable. And many of these CNVs are associated with rare but known syndromes, each with a specific group of traits that range from craniofacial manifestations to intellectual disability.
This variability suggests that the gene variants that underlie these syndromes have widely different effects (and possibly confer widely different susceptibilities to environmental influences). Autism is just one of many possible but not inevitable manifestations of these broader conditions.
These different CNVs with their different effects surely influenced results for each of the measurement tools used for this analysis. Indeed, something has to explain why the significant (and even nonsignificant) outcomes varied so much, even directionally, from table to table. A uniform factor like an ultrasound “hit” in the first trimester wouldn’t be the explanation for these internally inconsistent results, but variability in effects of these CNVs certainly could be.
So are first-trimester diagnostic ultrasounds a risk factor for autism or autism severity?
Despite what the headlines say, it’s not obvious from these results that ultrasounds are a risk factor for increased autism symptom severity. One analysis shows an association with reduced severity, there’s an absence of a dose response or increasing effect with each additional “hit,” and the many different CNVs are treated as interchangeable factors when they are not.