In a previous post, I wrote about a woman who attended one of my readings who wanted to discuss why there are so many contradictory studies regarding labor and delivery. She suggested ethical and legal constraints get in the way of doing really good work. I talked to professionals who do such studies to try to learn more about this. While it is true that anyone conducting a trial that directly involves patients (in this case pregnant women) must past the muster of a hospital's ethics panel and must use standard legal waivers and obtain full consent of the patients who participate, there are more complicated reasons for study disparities. First, some basics. The 'best' kind of trial is so-called randomized. That means that researchers select a group of subjects, sign them up for the study, and then, assign them to receive a certain treatment on a "random" basis. For example, they will either get, or not get, an epidural during labor. This is extraordinarily difficult to do, as most women are not ambivalent about their choices for pain relief during labor, and would not accept such random allocation. (Nonetheless, randomized trials of epidural analgesia in labor have been done.) For ethical reasons, the researchers need to allow women to change their mind during the study. For example, if they are randomized to receive an epidural and find that labor is actually not so bad, then they can not be forced to receive the epidural. Likewise, if they are randomized to no epidural, and find that labor is very painful, they must be allowed to receive an epidural if they wish. These so-called "protocol non-compliant" patients make statistical analysis complex, and interpretation of the results difficult. Depending on the outcome being studied, these trials are often expensive and time-consuming. Of course rare events (let's use uterine rupture as an example) need an even larger number of women to study. The importance of randomization can not be stressed too highly. Non-randomized studies suffer because the subjects being compared may differ in meaningful ways that would influence the outcome being investigated, something statisticians call "selection bias". For example, any non-randomized sample of women in labor receiving epidurals will show that women with more complicated labors, longer labors, and more difficult labors will tend to get epidurals more than women having easy, smooth labors. Hence an analysis of any outcome in this type of cohort would be skewed by the nature of the patients. Likewise, if one tries to compare any type of obstetrical or birth outcome in a non-randomized sample of say, women who have midwifery care vs. physician care, or home birth vs. hospital birth, or vaginal vs. cesarean delivery - these studies will always suffer from a preponderance of healthier patients in the midwifery or home or vaginal groups, respectively. On the other hand, randomizing women to one or the other of these types of care is extraordinarily difficult, some may say impossible. The other kind of study is retrospective. For the purposes of childbirth, these studies will look at medical charts or birth certificate data to compare outcomes. The whole story may not be revealed because there is no interaction with the patient, and the charts may not reveal the entire story.
Bottom line: Look at the type of study to determine its weight.
The next issue has to do with where the study is published. Some journals or medical publications may be more or less "friendly" to studies that show a certain type of outcome. One possible approach to this dilemma would be to have editorial review panels of these publications consist of members of both the medical and non-medical community when childbirth-related manuscripts are considered. This would allow for some cross-pollination between the natural and obstetrical camps. That will never happen, of course. Birth is too political. And in the end, women are stuck in the middle, often unsure what is best for them.