Nigel Hawkes: A bad case of bias against Caesareans

Behind the numbers

The World Health Organisation disapproves of over-medicalising birth, and it's probably right. Too many Caesareans are done for no good reason.

But it isn't going to advance its case by publishing evidence as unconvincing as that in a recent Lancet paper. The 23 authors concluded that a mother who opts for a Caesarean without a good medical reason is 2.7 times as likely to suffer death or complications as a mother who completes a normal birth.

That's enough to put the fear of God into the "too posh to push" brigade, which I suspect was the authors' intention. Unfortunately, the conclusion is in no way supported by the evidence presented. The team, from the WHO's global survey on maternal and perinatal health research group, looked at medical records from 107,950 births in Asia. Of these, just 1,515 were Caesareans chosen in advance of birth with no medical indication to justify them. Almost all of them were in China.

So how many women died? None. How many suffered complications? Eight: five needed treatment in an intensive care unit (ICU), and three needed a blood transfusion. The risks for women who completed a normal birth were significantly higher. One in a thousand died, five times as many required a blood transfusion, and twice as many were admitted to an ICU.

Overall, if deaths and complications are added up to make a "Maternal mortality and morbidity index", risks to mothers in the Caesarean group were 60 per cent lower than in the normal birth group.

So how, from this, does the team conclude that risks to mothers who have Caesareans are actually 2.7 times greater? There's a hefty difference between 60 per cent smaller and 270 per cent greater, but statistical manipulation is a powerful tool.

The team achieves this transformation by correcting for a range of risk factors. If the Caesarean mothers were an especially low-risk group, then it is possible that such a correction could narrow the gap or even eliminate it. The adjustments in this case included such things as age, education, previous medical disorders, high blood pressure, and so on.

It's easy enough to work out how much lower the risks must have been to make such a huge correction to the raw data. The women who had Caesareans must have had only one seventh as much baseline risk as those who had a natural birth in order to make it work – and that's highly implausible.

Nowhere in this paper does the team comment on the remarkable transformation of the figures. But they do say: "The most important finding of the survey is the increased risk of maternal mortality and severe morbidity which was analysed as a composite outcome in women who undergo Caesarean section with no medical indication."

Did none of the 23 think this an odd conclusion to have reached? Did no one check the arithmetic in the tables, which are full of errors? The Lancet is a distinguished journal – were its referees asleep?

Connoisseurs of the bending over backwards that medical authors do to get the "right" answer call it White Hat bias, after old-style Westerns where the hero always wore a white hat. This is a classic example. Policy says Caesareans are bad, so the evidence had better prove it.

And what about the babies? Those born by elective Caesarean without medical indications were seven times less likely to suffer death or complications (raw data) or less than half as likely if you believe the corrected data. Understandably, the authors don't make much of this.

Nigel Hawkes is director of Straight Statistics