乌鸦传媒

Two big studies hinted at small increases in risk of autism spectrum disorder (ASD) in children born to mothers who received epidural anesthesia during delivery, but their authors said it wasn't clear that there was a genuine association.

And an editorial in JAMA, where the studies were published, argued that even if it's real, it's too small to matter.

, conducted in the Canadian province of British Columbia, found a crude hazard ratio of 1.32 (95% CI 1.24-1.40) in an analysis of nearly 400,000 births, which declined to 1.09 (95% CI 1.00-1.15) when adjusted for a variety of covariates.

The other, , yielded a statistically significant association of similar magnitude prior to adjustments (HR 1.29, 95% CI 1.21-1.37), but it shrank to non-significance when other factors were accounted for (HR 1.05, 95% CI 0.98-1.11), despite including almost 500,000 births.

Their respective authors both cast doubt on the association being real.

Gillian E. Hanley, PhD, of Vancouver General Hospital in British Columbia, and colleagues called the apparent increase in ASD risk "small" and said that, while it did meet the threshold for statistical significance (barely), "the likelihood of residual confounding" was great enough to render the strength of evidence low.

And the Danish study authors, led by Anders Pretzmann Mikkelsen, MD, of Copenhagen University Hospital-Rigshospitalet, seized on the statistical nonsignificance of their adjusted analysis to argue against a genuine association.

In both studies, the absolute difference in ASD rates were small: 1.53% versus 1.26% in the Canadian data, and 1.52% versus 1.30% in Denmark, both before adjustments. As the study authors and the editorialists pointed out, that left plenty of room to doubt an effect.

But what if it is for real? The 95% confidence intervals for the adjusted analyses indicated that ASD risk increases in the 10%-15% range couldn't be ruled out. (though the literature overall is mixed) had also identified potential ASD risk with epidurals. Isn't that another reason to avoid epidurals, especially given other risks associated with them?

Not really, said Cynthia A. Wong, MD, and Hanna Stevens, MD, PhD, both of the University of Iowa in Iowa City, in the .

"For individual patients and their childbirth professionals, the benefits and risks of neuraxial labor analgesia, relative to the alternatives, should be carefully weighed," they wrote. The pair explained that epidurals are "unmatched by any other analgesic option" and their safety with respect to most outcomes is well established. That said, they added, epidurals do come with some documented risks, including intralabor fever, fetal cardiac rhythm abnormalities, and prolonged labor for some women.

Because of the recent studies, including those now being published in JAMA, "patients and clinicians will legitimately ask whether a risk of ASD should be included in this benefit-risk analysis," Wong and Stevens wrote. The answer, they said, is no.

"Based on these studies and the current knowledge of ASD, it appears that concern for ASD should not carry weight in this decision," they declared. "Current evidence does not justify considering risk of ASD when deciding whether to use (patients) or recommend (health care professionals) neuraxial labor analgesia."

Study Details

The two studies were similar in that they were retrospective analyses of medical records, but not identical. Hanley and colleagues drew on data for essentially all women in British Columbia delivering singleton infants from 2000 to 2014, a total of 388,254 births after excluding cesarean sections, preterm births, and children lost to follow-up before age 2 years. In this cohort, roughly 30% of infants were exposed to epidurals.

In the Danish study, Mikkelsen's group analyzed outcomes for 479,178 children born from 2006 to 2013. Fewer than one-quarter of these deliveries involved epidurals. Exclusions were narrower than in the Canadian study, including emigration, infant death, and postnatal diagnoses such as Down syndrome, totaling about 6,000 infants (vs more than 200,000 in British Columbia).

Adjustments in the Canadian study included year of birth, parental age, neighborhood income, community size, maternal complications of pregnancy, parity, smoking during pregnancy, maternal BMI, induction of labor, gestational age, infant sex, infant size relative to norms, and presence of congenital anomalies.

Because Mikkelsen and colleagues included a wider range of births, some factors that kept many infants out of the Canadian study ended up as adjustments in the Danish study. These included year of birth, parental age, gestational age, infant sex, firstborn or not, elective cesarean delivery, small (but not large) for gestational age, induction of labor, maternal complications of pregnancy (but fewer than in the Canadian study), obesity, smoking status (not restricted to pregnancy), family ASD history, family psychiatric diagnosis history, maternal education, maternal employment, region of Denmark, and past maternal medical-seeking behavior.

Another potentially important difference was that follow-up was somewhat longer in the Canadian study, about 9 years versus 7 on average.

The Danish study found that firstborn status, labor induction, region, delivery year, and infant sex were factors, besides epidural analgesia, correlating most strongly with subsequent ASD diagnosis. Hanley and colleagues found that most of the statistical attenuation in risk, relative to the unadjusted data, occurred when they factored in pregnancy complications, parity, smoking during pregnancy, and maternal BMI.

Both studies had the customary limitations of retrospective records analyses. In addition, while both British Columbia and Denmark have criteria for ASD diagnoses, there was no system to ensure that clinicians were following them in every case. And while both studies attempted to adjust for known and suspected covariates of ASD, the possibility of residual confounding remained substantial.