Advocates for evidence based care feel our ears perk up when we hear about a new Cochrane systematic review. Cochrane Review = Evidence! Right? Indeed, systematic reviews represent the top of the "evidence pyramid" and Cochrane systematic reviews are the gold standard for their rigor and transparency. A Cochrane review can even conclusively settle important clinical controversies.

But sometimes Cochrane reviews leave us with more questions than answers.

Last week, the Cochrane Library released a systematic review evaluating the effectiveness of intrapartum antibiotics for known maternal group B streptococcal (GBS) colonization. And it's a hot mess.

The four included trials that compared IV antibiotics with no treatment in labor collectively had only 500 participants, which we automatically know is far too small to find statistically significant differences in a condition that affects 1 in 2000 newborns, and results in death or long-term complications even less frequently. But small sample sizes were the least of the problems here. The reviewers noted several other problems with the trials:

• In one study, researchers tracked their findings and halted the trial as soon as a significant difference was found (favoring treatment with antibiotics). This is a blatant form of bias - it is like flipping a penny until you get heads 5% more often than you got tails. If you keep flipping long enough (or stop flipping soon enough) you'll be able to find that 5% difference simply by chance.
• In the same study, researchers changed to a different statistical test that allowed them to achieve statistical significance with their data, when the originally planned (and more appropriate) test would have produced a nonsignificant finding.
• None of the studies used placebos, so women, care providers, and hospital staff knew which women received antibiotics and which did not. This may have altered treatment of the women or the babies, possibly in ways that would make no antibiotics appear safer (for instance, avoiding or delaying membrane rupture in a woman who is GBS+ but not getting antibiotics).
• One study excluded women who developed fevers in labor. GBS colonization can cause maternal fever and newborn sepsis, so excluding these cases makes no sense.
• Some women included in the studies were likely GBS negative because methods used to determine GBS status were inadequate
• Outcomes were poorly defined.
• Data on a substantial proportion of women and babies were missing.
• Groups were mysteriously differently sized.
• Need I go on?

The Cochrane reviewers, in my opinion, did a respectable job with what they had, but what they had was garbage and as the saying goes, "Garbage in, garbage out." You can't make reliable conclusions out of a bunch of bad research, even if you're a Cochrane reviewer.

So what were the findings?

Three trials, which were around 20 years old, compared ampicillin or penicillin to no treatment and found no clear differences in newborn deaths although the occurrence of early GBS infection in the newborn was reduced with antibiotics.

More, better research is needed, but the Cochrane reviewers are not optimistic:

Ideally the effectiveness of intrapartum antibiotics to GBS colonized women to reduce neonatal GBS infections should be studied in adequately sized double blind controlled trials. The opportunities to conduct such trials have likely been lost as practice guidelines have been introduced in many jurisdictions. (p. 11)

Source: Centers for Disease Control and Prevention

http://www.cdc.gov/mmwr/preview/
mmwrhtml/rr5111a1.htm

In the meantime, women should be aware that other evidence, albeit not from randomized controlled trials, suggests that antibiotic treatment reduces deaths from early onset GBS disease in newborns. According to the Centers for Disease Control and Prevention, a steady decline in GBS disease has been seen in individual institutions, in the whole U.S. population, and in other countries as antibiotic use has risen. But these population-level data cannot tell us whether antibiotics or some other factor caused the decline.

What other advice can we share with women?

1. Be aware that antibiotics are not harmless. Severe allergic reactions are possible, and antibiotic use in labor can result in thrush (candida infection) which causes painful breastfeeding and sometimes early weaning. We do not know other possible harmful effects because they have never been studied adequately or at all.
2. No study confirms the effect of labor practices on GBS infection in newborns, but here we can use our common sense. Care providers should avoid or minimize sweeping/stripping membranes before labor, breaking the bag of waters, vaginal exams, and other internal procedures, especially those that break the baby's skin and can be a route for infection. These include internal fetal scalp electrodes for fetal heart rate monitoring and fetal blood sampling.
3. Keep mothers and babies skin-to-skin after birth. This exposes the baby
   to beneficial bacteria on the mother's skin, facilitates early breastfeeding, and lowers the likelihood that the baby will exhibit signs or symptoms that
   mimic infection, such as low temperature or low blood sugar, which could cause the need for blood tests or spinal taps to rule out infection.

Citation: Ohlsson A, & Shah VS (2009). Intrapartum antibiotics for known maternal Group B streptococcal colonization. Cochrane database of systematic reviews (Online) (3) PMID: 19588432

Tags