No Difference? The case of cesarean surgery and postpartum infection

When a study compares one practice with another and reports “no difference” in outcomes, is that the end of the story? Not necessarily. “No significant difference” can mean any of these:

  • there really is no difference
  • the study was too small to find the difference that really is there (it lacked “statistical power”)
  • the outcome was measured or reported in a way that obscured a difference that really is there
  • the statistical test yielded a “false negative result,” failing to find a difference that really is there (this type of error occurs in up to 1 out of 20 statistical results)

So in 3 out of our 4 possibilities, the difference really is there, we just can’t see it. How do we know when “no difference” means “no difference” and when “no difference” means “keep looking”?

To understand this problem, and how to best get around it, let’s have a look at postpartum maternal morbidity after cesarean surgery versus vaginal birth.

Perhaps you’ve heard that there’s “no difference” in health outcomes between cesarean surgery and vaginal birth. This is what women have been told ever since a so-called “State-of-the-Science Conference” report was issued by the National Institutes of Health in 2006.  After evaluating the body of literature comparing planned vaginal birth and planned cesarean section in healthy women, the NIH reported, among other findings, that the evidence that infection is more likely with planned cesarean is “weak”. The report said that, although observational studies find a link between cesarean and infection, the only randomized controlled trial (RCT) of planned cesarean versus planned vaginal birth showed “no significant difference” in infection rates.

It's basically 50/50, right?

It's basically 50/50, right?

In contrast, a new population-based study from Denmark involving over 32,000 women giving birth between 2001 and 2005 reported that women who gave birth by cesarean were nearly 5-times more likely than women who gave birth vaginally to experience a wound infection, urinary tract infection, or blood stream infection. Most of the difference was attributed to wound infection, which occurred in 5.6% of women having unscheduled cesareans in labor, 3.9% of women having scheduled cesareans, and only 0.08%  of women giving birth vaginally. In the United States, where up to 500,000 cesareans may be safely preventable each year,  these data suggest that more than 20,000 postpartum wound infections could be avoided annually along with the excess cesareans.

So what is the disconnect between the “state of the science,” which tells us evidence is “weak” and this new study, which paints a very different picture?

First, the NIH conference, and the AHRQ-sponsored systematic review underpinning the conference, compared planned routes of delivery while the Danish study compared actual routes of delivery. The reviewers acknowledge that studies comparing planned vaginal birth and planned elective cesarean delivery in women with no pregnancy complications are scarce and randomized controlled trials are absent. Rather than conduct a systematic review without any randomized controlled trials, the reviewers included the Term Breech Trial (which I have previously posted about), even though the results of the review would affect recommendations made to women carrying head-down babies.  But the Term Breech Trial’s results  are not applicable to women carrying head-down babies, especially when infection is the outcome of interest. Why? Because 57% of women randomized to give birth vaginally actually had cesareans, and most of these presumably underwent the cesareans in labor, which increases infection risk (along with risks of hemorrhage and other complications). What’s more, breech vaginal births are much more likely than vertex vaginal births to involve episiotomies, instrumental delivery, or both, which increase the likelihood of infection. So in the Term Breech Trial, the vaginal birth group included far more women with cesareans, episiotomies, and instrumental deliveries than we would expect to see in a similar group of women planning vaginal birth of head-down babies. Despite these limitations that make the trial useless for evaluating the true risks of infection associated with each birth route, the AHRQ reviewers rated the Term Breech Trial as the highest quality evidence comparing planned vaginal with planned cesarean birth, trumping observational studies. In the process, the message to women – and clinicians for that matter – transformed from “Planning a cesarean is risky because you expose yourself to excess infection risk” to something more like, “there could be an excess risk of infection with cesarean, but it’s probably small and, hey, for all we know there may be no real difference after all.”

Another reason behind the disconnect is that the Term Breech Trial (again, the “highest quality” evidence in the AHRQ review underpinning the NIH conference) only measured infections occurring prior to hospital discharge, a very common cut-off in both RCTs and observational studies. The new Danish study provides striking evidence that this way of measuring infection is grossly inadequate – more than three-quarters (77%) of infections occurred after hospital discharge. In other words, they would have been missed if the researchers had stopped counting the number of infections as soon as the woman left the hospital. The researchers were able to capture these infections because Denmark has a national database of all births, which was linked to a national database of all clinical laboratory results of infectious diseases, a national database of all antibiotics prescribed, and a national database of all hospital readmissions. (The United States, by the way, has none of these, rendering this type of study literally impossible to conduct here.) In randomized controlled trials, following participants beyond the initial hospital stay is logistically difficult and very costly. Even when follow-up is intended, substantial numbers of new mothers may not respond to surveys or return (to the same provider/facility or at all) for postpartum care. As a result, properly constructed national databases provide an important source – sometimes the only source – of evidence on long-term outcomes.

So which “evidence” do we tell women? The “state of the science” or our less rigid but almost certainly more reliable assessment of the observational studies and our common sense? Of course cesarean can lead to infection – it’s surgery after all. And yes, even with optimal care, some women who plan vaginal births will need to have cesareans and some of those women will get infections. If we’re serious about helping women avoid serious complications like infections, we must:

Women experience many physical and mental health problems in the postpartum period, some of which may be safely prevented with a different approach to care given in pregnancy, labor, and birth or better education provided prenatally. We do not know the factors contributing to postpartum health problems because they are shamefully understudied. But one treasure trove of postpartum data in the United States provides a powerful foundation for addressing and researching postpartum health outcomes. Childbirth Connection’s 2008 report, New Mothers Speak Out, compiles the postpartum data and new mother testimonials from both the 2006 Listening to Mothers II Survey and a follow-up survey conducted with participants six months later.  All women’s health professionals and advocates should spend time with this report so we can begin to give postpartum concerns their due attention.

Hannah ME, Hannah WJ, Hewson SA, Hodnett ED, Saigal S, & Willan AR (2000). Planned caesarean section versus planned vaginal birth for breech presentation at term: a randomised multicentre trial. Term Breech Trial Collaborative Group. Lancet, 356 (9239), 1375-83 PMID: 11052579

Leth RA, Møller JK, Thomsen RW, Uldbjerg N, & Nørgaard M (2009). Risk of selected postpartum infections after cesarean section compared with vaginal birth: A five-year cohort study of 32,468 women. Acta obstetricia et gynecologica Scandinavica, 1-8 PMID: 19642043

Visco AG, Viswanathan M, Lohr KN, Wechter ME, Gartlehner G, Wu JM, Palmieri R, Funk MJ, Lux L, Swinson T, & Hartmann K (2006). Cesarean delivery on maternal request: maternal and neonatal outcomes. Obstetrics and gynecology, 108 (6), 1517-29 PMID: 17138788

Different Methods for Different Questions , , , ,