“Don’t have a Caesarean unless you must” said the headlines on stories about a new study from the World Health Organisation, published in The Lancet.

The study showed that women who opt to have a Caesarean without the medical need for one – the group traditionally described as “too posh to push” – have a nearly three times greater risk of death or complications than those who have a normal vaginal delivery.
 
Close examination of the actual figures suggests that this is an extraordinary conclusion to have drawn, in no way justified by the evidence. But as it comes from WHO and is published in The Lancet, it has every chance of being taken seriously. The National Institute for health and Clinical Excellence (NICE) has this week started reviewing its guidance on Caesareans and there are fears that it will put undue weight on this study.
 
The study was conducted in nine Asian countries – Cambodia, China, India, Japan, Nepal, Philippines, Sri Lanka, Thailand and Vietnam. The team studied the medical records of women admitted to hospitals during a three-month period in 2007-08 for hospitals with 6,000 or fewer deliveries per year, and two months for those with more than 6,000. There were 122 such hospitals and almost 108,000 birth records were analysed.
 
The records showed just 1,515 of these deliveries were by Caesarean section chosen in advance without medical indications to justify them. The vast majority (1,356) were in China; the other countries had very low numbers.
 
Of these 1,515 women, none died. Five were admitted to an intensive care unit (0.3 per cent) while three (0.2 per cent, though the figure published in the paper is 0.3 per cent, presumably a mathematical error) needed a blood transfusion – a total of eight out of 1,515. None required a hysterectomy.
 
In mothers who had vaginal deliveries, these risks were all greater: for spontaneous vaginal delivery 0.1 per cent of mothers died, 0.6 per cent were admitted to an ICU (the paper says 0.5 per cent, another miscalculation) and 1.0 per cent required a blood transfusion. A small number, 0.04 per cent, (though I make it 0.05 per cent) required a hysterectomy.
 
For assisted vaginal deliveries, where forceps or other techniques were needed, the risks were roughly tripled. Table 3 in The Lancet paper indicates that the risks of hysterectomy in this group were a worrying 3.3 per cent, though the actual numbers quoted, four hysterectomies in 3,465 cases, suggests the correct figure is 0.1 per cent. (Seldom have I seen a table so full of mathematical errors.)
 
Summing up these figures into a “maternal mortality and morbidity index” the authors conclude that spontaneous vaginal delivery has a risk of 1.6 per cent (1,215 cases out of 75,057 deliveries) and elective Caesarean without indications a risk of 0.6 per cent (nine out of 1,515 deliveries).
 
On this basis, the risk to the mother of electing for a Caesarean is less than half that of completing a spontaneous vaginal birth: 0.6 per cent against 1.6 per cent, a 60 per cent lower risk. Yet when this is corrected for confounding factors, the risk has suddenly become 270 per cent greater. If the risk to the mother having a vaginal birth is set at one, the risk to the mother having an elective Caesarean, the authors claim, is 2.7 (95 per cent confidence interval 1.4 to 5.5)
 
Bear in mind we are talking about just nine cases. To have transformed the risk so dramatically, these nine must have been extremely low-risk cases, yet all the authors had to go on were medical notes. They did not examine any of the women before or after they gave birth. Nor do they comment on the enormous discrepancy between the raw data and the corrected odds ratios, which is the least one might expect.
 
How low-risk would these nine women have to have been? The baseline risk of a spontaneous vaginal birth in this study is 1.6 per cent, so 1,515 × 1.6 per cent, or 24 events, would be expected if the risk of an elective Caesarean were the same. In fact 9 events were observed, but were said to be 2.7 times as many as expected, which means that only 3.3 events were expected out of 1,515 Caesareans, according to their results. 
  
That means the 1,515 must have been, on average, 3.3 ÷ 24, or one seventh of the risk of an average patient - that's the only way the results can be obtained.  And this seems completely implausible.
 
“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”, write the authors. “We conclude that Caesarean section should be done only when there is a medical indication to improve the outcome for the mother and the baby.”
 
Their data do not bear such a conclusion. The statistical analysis is almost certainly where the error arose. Did no referee raise the alarm?
 
One final point: the comparison they make is an unfair one. The proper comparison to have made would be between women who attempt a natural birth, and women who have an elective Caesarean.
Many of those who set out to have a spontaneous birth fail, for one reason or another, and require intervention. By choosing only those who succeed and ignoring the others, the authors are failing to make their comparison on an “intention to treat” basis.
 
WHO believes too many Caesareans are done without proper cause. But in interpreting these data, the authors appear to have bent over backwards to prove the point – a classic illustration of White Hat bias. The findings should be ignored.