Systematic Review Finds Eating and Drinking in Labor to Be Benign

A Lamaze International internal e-newsletter points us to a new systematic review, meaning a study of studies on a particular issue, of nutritive oral intake vs. ice chips or water alone. In the interest of full disclosure, the review has its weaknesses, but nonetheless, it adds to the body of evidence that we shouldn’t be withholding oral intake in labor—at least not from healthy laboring women. For that reason, it’s worth going over the review, and it provides an opportunity to bolster its conclusions with data from other sources.

Frankly, expectations weren’t high when I saw that two of the four reviewers also conducted a systematic review on routine induction at 40 weeks that I had deconstructed for Science & Sensibility. Low expectations were justified. As with the 40-week induction review, not all the included trials were published studies. One of the 10 was only an abstract (summary) and another merely data from a “personal communication.” Quality systematic reviews exclude abstracts because they don’t provide enough information to evaluate the study, which would also be true of “personal communications.” Other weaknesses had to do with the analysis, but let’s start by summarizing the review.

The review included 10 trials totaling 3,982 laboring women with no obstetric or medical complications that would increase their likelihood of cesarean. Women were randomly allocated to be allowed oral intake, which the reviewers called the “intervention” group—although, of course, the real “intervention” is restricting access to food and drink—or were limited to the usual ice chips, water, or sips of water. What women were allowed varied among studies: a carbohydrate drink (5 trials), items from a low-residue diet (3 trials), honey date syrup (1 trial), or unrestricted food intake (1 trial). When data were pooled, no differences were found in rates of cesarean (25% vs. 24%), instrumental vaginal delivery (26% vs. 27%), labor augmentation (strengthening labor using IV oxytocin) (53% vs. 54%), epidural analgesia (66% in both groups), ketoacidosis (ketones in the blood, a starvation symptom) (22% vs. 23%), vomiting in labor (38% vs. 37%), vomiting under general anesthesia (0 in both groups), pulmonary aspiration (inhaling stomach contents) under general anesthesia (0 in both groups), 5-minute Apgar score < 7 (a measure of condition at birth) (1% vs. 2%), or newborn admission to intensive care (5% in both groups). The only benefit was an average 16-minute reduction in labor duration.

So, what are the problems?

• These are appallingly high medical intervention rates for healthy women with no obstetric complications. A 5% admission rate into neonatal intensive care seems shockingly high as well. Surely the highly-medicalized management of these women would swamp any clinically significant effects of nutritive intake on labor progress.
• None of the studies considered advantages from the participants’ perspective. A survey by Penny Simkin (1986) decades ago found that women reported feeling stressed by being deprived of oral intake, especially liquids, in labor.
• All trials were analyzed according to “intent-to-treat,” meaning that women were kept with their assigned groups regardless of actual treatment. “Intent-to-treat” analyses reduce bias, which is the point of random allocation, but when large percentages of participants don’t receive the intended treatment, it can make it falsely appear that there is no difference between groups. We don’t know to what degree that was the case in these trials, but it could well have been substantial. For example, two-thirds of the women in the oral intake group had epidurals and half had labor augmented. Were these women still considered low risk enough to be allowed oral intake?
• The number of participants even in the aggregate is much too small to establish whether allowing oral intake poses a risk of vomiting under general anesthesia, pulmonary aspiration, and consequent death, the fear that led to banning oral intake decades ago. To address this, the reviewers do a statistical projection based on 0 women of 1329 in the oral intake group either vomiting under general anesthesia or getting stomach contents into their lungs and conclude that as many as 1 in 360 women could vomit under anesthesia and 1 in 375 aspirate. Their projection, however, is meaningless because none of the trials reported how many women underwent general anesthesia, and it is likely to be not many because most cesareans are done under regional anesthesia. Fortunately, we have better data by which to gauge the risk. We have three U.S. studies totaling 78,000 women who ate and drank freely in labor among whom not one case of aspiration occurred (Greulich 1994; Haire 1991; Rooks 1992). Also, as of 2007, U.K. policy has permitted women at low risk for cesarean to eat and drink in labor. Despite this, the U.K. maternal mortality rate resulting from anesthesia complications is 1 per million, and we don’t even know whether pulmonary aspiration was the cause.

We have broader-based problems embedded in the premises of medical-model research as well. First, as we saw above, medical-model thinkers have a narrow concept of what constitutes “benefits.” “Soft outcomes,” such as women feeling less discomfort or feeling more normal and healthy, don’t make the list. (To their credit, the reviewers cite Simkin’s research and raise the issue of possible negative psychological consequences of forbidding women oral intake in the discussion section, but they immediately undercut the acknowledgement by reminding readers that the study was too small to determine safety.) Second, randomized controlled trials test new treatments against current management, and only if the new treatment proves superior is there motivation for change. Because the review didn’t find substantive benefits by medical-model standards and couldn’t rule out the possibility of serious harm, its findings don’t justify changing current policies. This is the reverse of what optimal care principles—the least use of medical intervention that produces the best outcomes given the woman’s individual case—would dictate. Researchers should start with physiologic care as the default and have to prove the superiority of routine interventions or restrictions before imposing them, not the other way around. Unfortunately, though, so long as soft outcomes such as comfort and confidence are considered last when they’re considered at all, and basic physiologic care is deemed an “intervention,” we are pretty much stuck.

The Take-Away

All the above being said, the bottom line is we have a new systematic review, with a meta-analysis (analysis of data pooled from multiple studies), published in a prestigious U.S. obstetric journal, that concludes that a policy of allowing oral intake had no adverse effects and some, albeit meager, benefit. Those elements automatically give the review credibility with most obstetricians, making it leverage for lobbying for the right for low-risk women, at least, to eat and drink at will in labor.

###