MacDorman revises risk of C-section downwardMacDorman et al. appear to have taken to heart the serious criticsm of their Septemenber 2006 paper that purported to show that Cesarean section increases the risk of neonatal death almost 3 fold. The data was flawed and incomplete, and the method of statistical analysis was not the best method. In a new paper, Neonatal Mortality for Primary Cesarean and Vaginal Births to Low-RiskWomen: Application of an "Intention-to-Treat" Model, in the March 2008 issue of Birth, MacDorman et al. have analyzed the same flawed data with different techniques designed to remove some of the errors from the dataset, and modified their original claim substantially.
First, some background:
In September 2006, MacDorman et al. published (and heavily publicized with interviews to the lay press) a paper that purported to show that Cesarean section increases the risk of neonatal death almost 3 fold. MacDorman et al. compared outcomes of C-sections with "no indicated risk" (a blank space on the risk section of the birth certificate) with outcomes from vaginal deliveries with "no indicated risk" and found that the neonatal death rate was higher in the C-section group. However, MacDorman neglected to mention that it is well known that the risk section of the birth certificate is often left blank even when there are serious risk factors and complications. Indeed, in 50% or more of serious risk factors (heart disease, kidney disease, etc.) the space is left blank. So their assumption that "no indicated risk" means no risk is completely unjustified. (See Disingenuous and Why try to trick people?)
In this month's issue of Birth, MacDorman et al. use the same dataset, known to be flawed and incomplete, but applied a better form of analysis, Neonatal Mortality for Primary Cesarean and Vaginal Births to Low-RiskWomen: Application of an "Intention-to-Treat" Model. According to the March 2006 NIH State-of-the-Science Conference on "Cesarean Delivery on Maternal Request", a study comparing vaginal delivery to C-section should employ:
an "intention-to-treat" methodology, a methodology commonly used in medical research... [E]mergency cesarean sections performed after a woman was in labor would be combined with vaginal births to create a "planned vaginal delivery" category since the original intention of the physician and the mother in both cases was presumably to deliver the infant vaginally. The "planned cesarean delivery" group would include only those deliveries where a cesarean section was performed without labor.Using this new, more accurate statistical analysis, MacDorman et al. went back and analyzed their incomplete, flawed dataset.
...In the most conservative model, the adjusted odds ratio for neonatal mortality was 1.6 (95% CI 1.35–2.11) for cesareans with no labor complications or procedures, compared with planned vaginal deliveries. Conclusions: The finding that cesarean deliveries with no labor complications or procedures remained at a 69 percent higher risk of neonatal mortality than planned vaginal deliveries is important, given the rapid increase in the number of primary cesarean deliveries without a reported medical indication.So MacDorman et al. have adjusted their claim downward by a substantially amount. Instead of their original claim that C-section increases the risk of neonatal death by 200%, they now find that C-section increases the rate of neonatal death by only 69%. Moreover, they make no effort to determine if the difference is statistically significant.
The new paper is valuable because it implicitly acknowledges that the dataset is flawed and attempts to apply statistical corrections. Applying improved statistical analysis reveals that two thirds of the previously claimed increase in neonatal death from C-section was actually the result of flawed data. The data is still flawed and the new analysis does not completely correct those flaws. Put another way, if only 0.002% of the birth certificates were missing risk data (2 out of every 100,000 birth certificates), there would be no difference in mortality in the two groups at all. In addition, the authors do not claim that the differences between the two groups are even statistically significant.
As I have said all along, MacDorman et al. have not, and still do not, demonstrate that C-section increases the risk of neonatal death.