Tag Archives: research bias

The longer-term effects of CS for breech in Denmark

Screen Shot 2014-08-15 at 06.30.06Something is rotten …

Well, I suppose a backlash was inevitable. Due to the campaigning of women and the willingness of a significant number of health care providers to provide women with a real choice when it comes to breech childbirth, the argument for re-skilling to better support breech births has been gaining momentum. But this week saw the publication of two studies arguing this is not such a good idea.

First it was a Dutch group (Vlemmix et al 2014) who made the argument that at 1.3/1000 compared to nil, vaginal breech birth (VBB) results in ten times the mortality of planned CS. They overlooked the need to match the word ‘planned’ with an intention-to-treat analysis, and their own research (Vlemmix et al 2013) suggesting that neonatal mortality was doubled in pregnancies following an elective CS compared to those where a VBB was planned in the first pregnancy (2.5/1000 vs 1.3/1000). (Read my response here.)

When baby's head has descended into the pelvis, the pubic bones are directly behind the occiput

When baby’s head has descended into the pelvis, the pubic bones are directly behind the occiput

Now a Danish study asks, “Can Caesarean section improve child and maternal health? The case of breech babies” (Jensen and Wust, 2014). Wait, did I read that right? Are we asking whether CS can improve maternal health? Really?

Unlike their Dutch counterparts, Jensen and Wust have decided: “In our estimation sample, we observe only few infant deaths for breech babies. Thus we do not consider this very rare outcome in the proceeding analysis.” One country decides this outcome is a deal breaker, another feels it is so rare that it is not necessary to consider it.

Jensen and Wust present a lot of beautiful graphs and calculations showing that there was a noticeable improvement in Apgar scores and a reduction in visits to the GP, but no significant change in serious morbidity (ill-health) or hospitalisations in the first three years of life, following the sudden increase in elective CS for breech associated with the publication of the Term Breech Trial in 2000.

Although I am in favour of attempting to calculate the longer-term effects of such sweeping changes, I am concerned about what they did and did not choose to speculate upon.

Their most statistically significant finding was an increase in prolonged maternal hospitalisation following an elective CS. This has noticeable financial implications, which they calculated, but they do not address the increase in costs and risks in future births. The financial blind eye is concerning, but their conclusion that CS does not affect the health of the mother, without considering future births, is even more disconcerting.

On the other hand, they speculate that the additional expenditure for elective CS is balanced by costs savings as a result of significant reductions in cerebral palsy and subsequent care needed. They did not have information on actual rates of CP in this population. Instead, they reference a 2001 study by Krebs, which did NOT note a relationship between CP and mode of delivery, and found in 20,000 breech births a total of 4 serious long-term disabilities and 18 minor disabilities possibly related to low Apgar scores. They ignored Krebs previous research (1999) indicating that an increase in CP for breech-presenting babies was NOT associated with mode of delivery. They also ignored a recent systematic review and meta-analysis (O’Callaghan and MacLennan 2013) demonstrating that CS does NOT reduce the risk of CP for breech-presenting babies.

It’s important to get this right. Low Apgars are definitely associated with increased CP and other problems in head-down babies, possibly because the birth itself less often causes minor asphyxia, and therefore the cause is often an underlying fragility. In Krebs’ work, he found the association was mostly with small for gestational age infants; this corresponds with other breech research which consistently associates smaller babies with poorer outcomes. But because many breech babies have lower Apgars as a consequence of the way they are born, Apgars are not such a clear indicator of future risk in this population. They were not in the Term Breech Trial (Whyte et al 2004), and that is the best evidence we have to go on at the moment.

The main outcomes shown in this research are an increase in overall Apgar scores and a decrease in visits to the GP in the first year, less significantly in the second year, and not significantly different in the third year of life. They found no increase in significant neonatal morbidity or hospitalisations in the first 3 years of life. Although it was not originally one of their primary outcomes, the GP visits could be significant. It could also be that in the wake of a major change in practice suggesting that breech-born babies are at increased risk, everyone’s a bit nervous in the first couple of years? We need more information regarding these babies’ actual health problems to understand and use this data, because it conflicts with a Finnish study which found the opposite was the case at 7 years (Ulander et al 2004).

So Jensen and Wust have given us more research indicating an increase in short-term morbidity (with mortality being rare and insignificant?) but no difference in significant neonatal morbidity and the need for increased medical care after 2 years. This matches the information from the Term Breech Trial. Despite the authors insistence they have uncovered evidence that CS is best for all breech babies, and that it is ultimately cheaper (based on their non-evidence-based speculation on future costs of CP only), it actually adds to the literature confirming no concrete evidence of a difference in long-term outcomes for breech babies. Just goes to show you how you can take a group of statistics and put just about any spin on it you like, especially if your mathematical ability makes your analysis fairly impenetrable to most people. (Good thing I live with an actuary.)

Screen Shot 2014-08-15 at 06.30.06Turning now to the elephant in the room

Being born vaginally may be more risky for some babies than being born by CS. Most of the evidence seems to indicate that, in the short-term at least, using standard lithotomy delivery practices, this is the case. On the other hand, most of the long-term evidence does not indicate lasting effects.

What concerns me about literature like this, which makes predictions about what would be saved or not, financially or physically, with this approach or that – is that women, as long as they are human, will continue to have their own unique approach, and they should. That is what being human is about. Many will want to deliver their breech babies by CS, and they should have access to that care, even if it means a greater financial burden. And many will want to give birth vaginally, even in awareness that the rare outcome of neonatal mortality is more likely to happen to them, even in the awareness that if something goes wrong, they will need to live with it for the rest of their lives. We will always have death, and handicapped children that require our grief, our love and our devotion. This cannot be eradicated. Women deserve to be able to make this very personal decision without being made to feel criminal.

Instead of continuing to do research which tells us what we already know, we should invest in research exploring modern management strategies which are showing promise in reducing risk to babies born vaginally, so that women who live in countries where there ought to be a choice actually have one, and women who live in countries where CS is either inaccessible or a real danger to their health have the best chance of going home with a healthy baby. We should stop trying to have the last word on how breech babies should be born, let women decide how to balance the complex array of risks and benefits in their own lives and families, and develop our skills at being ‘with woman’ and her breech.

Shawn

How the consent process introduces bias into RCTs

Part 1: Why I remain sceptical of RCTs concerning obstetric interventions in normal labour and birth

Another blog post in response to a Twitter debate .. this time concerning various RCTs currently evaluating IOL vs expectant management. We’ve been discussing three trials:

The questions are valid. We know stillbirth is increased, especially in certain populations, the longer a pregnancy continues. In order to make an informed decision, many women will want to know the most likely outcomes and effects of opting in or out of proposed interventions. RCTs are considered the most unbiased way of settling these issues, unsullied by the biases of women or health professionals.

The problem is, these interventions are eventually applied to a population that is, due to being human, inherently biased. Some women feel a strong preference in one direction, some in another; and some want their doctor to decide for them. Women need to consent to be randomised into RCTs, and women who are most averse to the proposed intervention simply decline consent. Therefore the population recruited becomes slightly biased towards a preference for the intervention being investigated.

Is this important? Does it matter? I don’t know. Recent research by Wu et al suggests that women with a strong preference for vaginal birth were more likely to have a vaginal birth. RCTs cannot tell us the effect of women’s preferences on the outcomes they measure. Yet in theory their results are used to offer women an option they will almost certainly have an opinion about.

If I were contributing to the design of these trials, I would want to collect observational data alongside the main trial data. Things like:

  • Why do women consent or decline to participate in the RCT? Are the women who declined to participate due to a strong preference against induction more or less likely to have a normal birth? Are the outcomes for their babies significantly different than those in either arm of the trial?
  • What are the long-term outcomes? Especially in the over-35 population, an increase stillbirth rate may be due to inherent weaknesses in the baby. Significant long-term differences are often not detectable until 2 years of age.
  • And finally .. would they do the again? Would they recommend it to a friend?

These are questions best answered using quantitative techniques, but women may have different questions or priorities, which we will only discover using qualitative investigations.

See Part 2: Why midwives are sceptical.

Shawn