Safety of AT in pregnancy

Stimulated by Moon et al 2019.[1]

Photo by Suhyeon Choi on Unsplash.
I found this image when searching for pregnancy and Korea on Unsplash. I suspect that these women are not actually pregnant, but I strongly suspect that they are Korean, since the photographer is from South Korea herself. This blog is about a retrospective cohort, so I thought this image of the backs of 3 women was appropriate, if a little cryptic.

AT – Acupuncture Therapy

AT stands for acupuncture therapy, and is the abbreviation used by the authors of this paper.[1] This is another large retrospective observational study – they are coming thick and fast these days aren’t they! I was attracted to this paper because the safety of acupuncture in pregnancy is a topic that continues to be haunted by the spectre of ancient mystical ideas. Those ideas and the consequent uncertainty in the minds of practitioners who are swayed by them can only be set aside by the certainty of large prospective data sets where no one is lost to follow up.

But this is not a large prospective data set of course. I have already said it is retrospective. So, what is the problem with retrospective data compared with prospective data? The main problem is the risk that any associations you see may be confounded by things that you don’t see. In the prospective setting you have the opportunity to limit the influence of unseen confounders by random selection of the samples you subsequently compare. If you have the resources to include large numbers, then the risk of an uneven distribution of influences reduces. So why not just do that?

20 799 pregnant women

This trial sampled 20 799 pregnant women, and a prospective trial of that size would cost a fortune. So, whilst they are plagued with potential confounding, they can be huge with limited cost, and it is likely that real cause and effect associations, if they exist, will influence the observed ones measured.

Moon and colleagues used the Korean National Health Insurance Service cohort data, which includes details of roughly one million subscribers, or about 2% of the Korean population in 2002. I was a little disappointed to read this, as the retrospective observational studies from Taiwan had been able to sample some 23 million representing 99% of the population.

Anyway, the numbers are still fairly big, with 20 799 in the sample taken from the database between the years 2002 to 2013. Of this sample 1030 had acupuncture during their pregnancy. The authors then calculate the rate of preterm delivery and stillbirth in the two samples and various subdivisions based on age, income, multifetal gestation, high risk pregnancy etc.

There were no significant differences in terms of preterm delivery for the whole samples or any subdivision, and there were no stillbirths in the acupuncture cohort, and 7 in the no acupuncture cohort (a rate of 0.04%).

Whilst there were no significant differences, the ORs (odds ratios) were consistently above 1 for the acupuncture cohort, and on occasion the lower CI (confidence interval) got very close to 1. If the lower CI was above 1 then by definition the result would be statistically significant. I will come back to this point…

There were significant differences in the length of pregnancy, both for those having a term delivery or a preterm delivery. In both cases the acupuncture group had longer pregnancies – 3 weeks on average for term pregnancies and 2 weeks for preterm ones. These results were highly statistically significant. As were the differences in the number of visits to obstetricians – 15 in the acupuncture group and 13 in the no acupuncture control cohort.

So, what does all this mean? Well it is observational research, so we do not know the reasons. But since this is a blog, I am free to guess! And so are you…

Pause for guessing!

The authors suggest that longer pregnancies give more opportunity for women to develop symptoms for which acupuncture is sought. It would also give the women more time for extra consultations with their obstetricians. This seems very plausible and would imply that the acupuncture had nothing to do with the length of pregnancy, but the latter tended to select for the former thus overrepresenting longer gestations in the cohort having acupuncture.

What about the trend towards marginally higher rates of preterm delivery? Given that the figures show a very highly significant association between longer pregnancies (whether term or preterm) and acupuncture use, it seems highly unlikely that acupuncture is the reason for this non-significant trend in preterm delivery rates.

I cannot sign off without mentioning David Carr’s landmark paper in Acupuncture in Medicine in 2015.[2] David demonstrated from 15 prospective trials (n=823) in pregnancy, which used so called ‘forbidden points’, that there were no differences in the rates of preterm delivery or stillbirth in the groups receiving acupuncture.

acupuncture performed correctly during pregnancy does not have any risk of causing excess adverse pregnancy outcomes

In conclusion, this research adds more weight to the BMAS position that acupuncture performed correctly during pregnancy does not have any risk of causing excess adverse pregnancy outcomes.


1         Moon H, Kim M, Hwang D, et al. Safety of acupuncture during pregnancy: a retrospective cohort study in Korea. BJOG An Int J Obstet Gynaecol Published Online First: 19 September 2019. doi:10.1111/1471-0528.15925

2         Carr DJ. The safety of obstetric acupuncture: forbidden points revisited. Acupunct Med 2015;33:413–9. doi:10.1136/acupmed-2015-010936

Declaration of interests MC