Some of you may be familiar with the recent press around the study ‘Intrapartum risk factors for levator trauma‘ that appears to have discovered that epidurals may reduce long-term pelvic floor damage compared to an unmedicated vaginal birth.
I am on a maternity care discussion group listserv, and this study was discussed by a number of different health care practitioners.
Remi Ejiwunmi, a midwife practicing in Mississauga, Ontario, speculated (quoted with permission):
As I read the summary caption, linking epidural use to pelvic floor integrity, my first inclination was that the benefits are likely more tied to delayed active pushing (i.e. waiting for the presenting part to descend onto the pelvic floor to provide an urge to push versus purple pushing at onset of full dilation) than the epidural… and further that maybe the benefits they see in Australia have more to do with their approach to managing women with epidurals than the epidurals themselves. I would wager that a woman with an epidural who was instructed to push because “it was time” and she was fully dilated, would likely not see the kinds of benefits that are described in the article.
Directed pushing is known to harm the pelvic floor, whereas women who push when they feel the urge are at less risk of damage. Questions of when and how these women were pushing aside, does this study really prove anything at all?
Dr. Joshua Steinberg, a physician practicing in New York state, had this to say about the evidence in this study (quoted with permission).
This is not high level evidence to guide care. It is suggestive and helps one speculate and produce hypotheses, but it does not answer a question and should not guide care. This was not a study of what risk factors are associated with later-life organ prolapse and impaired continence. It was a study of what risk factors are associated with injuries to pelvic floor muscles shortly after birth, which is not an actual outcome that patients care about. Worse than that, it wasn’t even a study of actual muscle injury after childbirth, but sonographic images which are correlated with pelvic floor muscle injuries, which are correlated with later life actual clinical outcomes. If you read the first paragraph, they use the words “may partially explain” and “may play a role” and “suggesting” to try to make the links of childbirth, pelvic floor muscle injuries, ultrasonic evaluation of those injuries, and later life stuff that actual patients would actually care about.
The findings in this study are DOE’s, or Disease Oriented Evidence. This evidence tells you (if the sonographic stuff is accurate) what happens to pelvic floor muscles, but that’s disease oriented, not patient oriented. Patient oriented evidence that matters would be an actual study on pregnancy and delivery factors and actual pelvic organ prolapse and continence issues 10 or 20 or 30 years later (or 4 months later like this study).
There are two other issues. This was an exploratory study, not a controlled study. There was no one group of women prospectively who were compared to another group of women. It was just 400+ women followed forward and then looked for associations backwards in time after outcomes were known. They found who had sonographic muscle abnormalities and then looked backwards to see what factors were associated. They found that some women had forceps delivery and the others did not. So they created two imaginary groups of women and compared them to each other after the fact to see if there was a difference. They looked at episiotomy, birth weight, length of second stage, head circumference, and so forth, each time creating new theoretical groups in order to see whether comparing those groups was correlated with differing outcomes. That’s not cheating. It’s a perfectly appropriate way of looking at data to form hypotheses about what correlates with what. But it’s not proof, it’s merely suggestion. It’s also called a fishing expedition! Wade into a pond of data and see if you can catch something, anything, that appears statistically significant.
The third thing is statistical significance. Our standard and quite arbitrary definition is that if something has 5% or less likelihood of being caused by chance alone, we consider it a positive association. Well, if you say that 1 out of 20 (5%) chance of happening at random is your threshold and you go fishing around for 20 or 30 or 40 factors, pure random chance alone dictates that you are going to find SOMETHING that correlates with your outcome. We don’t know how many issues these researchers looked at. They itemized 11 different factors in the results section on pages 4 and 5, but perhaps they looked at 30 factors. Perhaps some were eye color (blue vs. brown vs. black), hair length (long vs. short), handedness (left vs. right), etc. If you make enough comparisons, even silly ones, and set the threshold at 5% (p value <= 0.05), then approximately every 20 comparisons you will find something that correlates, such as epidural use (in only one but not both outcomes, by the way).
So let’s see. We have non patient-oriented outcomes that only might suggest real later outcomes, we have sonographic measurement which only might suggest real muscle injury, and we had a fishing expedition to find anything that might correlate with these non patient-oriented outcomes. That’s pretty damn weak evidence!
It’s far from useless evidence. Now we should take this possible impact of epidurals and test it. It would take a long time to follow current patients forward for the next 20 years to see what real clinical symptoms they do or don’t have. But perhaps there are really large healthcare systems with really good data from 1990 deliveries and 2010 health status, such as HMO’s or nationalized healthcare registries where researchers could look back at delivery data and then follow the women forward to see what their outcomes are. That would be a real cohort study. It wouldn’t be the best evidence, a Randomly Controlled Trial (RCT), but since we aren’t going to get an RCT with the size and length of followup to ideally answer the question, we can look for the next levels of evidence lower than that.
So it seems like all the media hype around this study is just that: hype. As with any study that appears to find that an intervention in a healthy mom and baby is better for the mom and baby than a physiological birth, we really need to be cautious about jumping to conclusions, and we also need to examine the types of care and management of the birth process that the mom and baby experienced, as these can have a huge impact on many different outcomes.