February 19, 2014

expert reaction to new study on risk factors for miscarriage

Researchers publishing in BJOG: An International Journal of Obstetrics and Gynaecology looked at observational data to identify the modifiable risk factors for miscarriage and to estimate the proportion which were preventable.

Prof Tom Bourne, Consultant Gynaecologist, Queen Charlotte’s and Chelsea Hospital, & Adjunct Professor, Imperial College London, said:

“This is a big study, but it does not really say anything new. There are also issues of recall bias, and the fact that they show an association rather than causation. Saying that by changing x or y a percentage of miscarriages could be prevented is quite a statement in the absence of an interventional trial (although they do present stats to model this). However, it adds to the view that alcohol in pregnancy is not a good idea, and we know miscarriage increases with age in any event. The association with lifting and night work has been reported before.”

Prof Patrick Wolfe, Professor of Statistics, University College London (UCL), said:

“This study does not establish a causal relationship between its reported risk factors and miscarriage. As the commentary that follows it notes, let’s not run before we can walk. The study has several statistical limitations, and so I caution that its conclusions may be subject to over-interpretation. It’s best to think of this study as identifying potential candidate risk factors that may be associated with miscarriages that occur later in pregnancy, rather than the last word on the subject overall.

“Two issues are well identified by the mini-commentary of Hemming that follows the study. First, the study does not demonstrate a cause-and-effect relationship (making the terms “prevention” and “preventable proportion” potentially confusing; as the authors note, “we cannot tell from this observational study whether the associations are causal”). Second, its results may be sensitive to recall bias (because the results include participants interviewed after their current or previous pregnancy ended in a miscarriage, thus potentially influencing the data they report for this study). How sensitive? We cannot say for sure, but Hemming suggests as a matter of good practice that the analysis might have instead reported results excluding those interviewed after miscarriage. 

“A third limitation, as the authors note, is that the majority of their participants were recruited “after the gestational age where miscarriage is most common”, meaning that their data only show about 1/3 of the expected total number of miscarriages from all pregnancies. How does this affect the conclusions? Again, we cannot say for sure.

“Finally, the study does not consider a totally representative sample of the population of all Danish women (see citation to Ref. 37). This is a common limitation, but one that is important to bear in mind because it means that we cannot simply say that its results (even with the 3 caveats above) apply in any way universally.”

‘Risk factors for miscarriage from a prevention perspective: a nationwide follow-up study’ by Nilsson et al.  published in BJOG on Wednesday 19 February