It is a potentially devastating indictment of the rising C-section rate. Most midwifery and “natural” childbirth websites claim that elective C-section triples the rate of neonatal mortality. Mainstream web sites like Feministing.com, and newspapers like The New York Times have repeated the claim. There’s just one problem. It’s not true.

The claim originated with the paper Infant and Neonatal Mortality for Primary Cesarean and Vaginal Births to Women with “No Indicated Risk,” United States, 1998–2001 Birth Cohorts, MacDorman et al, Birth Volume 33 Page 175, September 2006. According to the authors:

Neonatal mortality rates were higher among infants delivered by cesarean section (1.77 per 1,000 live births) than for those delivered vaginally (0.62). The magnitude of this difference was reduced only moderately on statistical adjustment for demographic and medical factors, and when deaths due to congenital malformations and events with Apgar scores less than 4 were excluded. The cesarean/vaginal mortality differential was widespread, and not confined to a few causes of death. Conclusions: Understanding the causes of these differentials is important, given the rapid growth in the number of primary cesareans without a reported medical indication.

The implication, of course, is C-sections done without a medical indication raises the risk of neonatal death by a factor of three. The entire study hinges on a critical detail. Are women with “no indicated risk” really women who have no risk factors? The answer is a resounding no.

Since birth certificates are such an important source for research information, they have been repeatedly studied for accuracy. Birth certificates are highly accurate for administrative data like parents’ names or numerical data like weight or Apgar scores. It is well known, however, that they are highly inaccurate when it comes to listing complications.

How Well Do Birth Certificates Describe the Pregnancies They Report? The Washington State Experience with Low-Risk Pregnancies, Dobie et al report:

Conclusions: Because birth certificates significantly underestimated the complications of pregnancies, number of interventions, number of procedures, and prenatal visits, use of these data for health policy development or resource allocation should be tempered with caution.

The reporting of pre-existing maternal medical conditions and complications of pregnancy on birth certificates and in hospital discharge data,  M. Lydon-Rochelle,  et al. found:

Results Birth certificate and hospital discharge data combined had substantially higher true-positive fractions than did birth certificate data alone for cardiac disease (54% vs 29%), acute or chronic lung disease (24% vs 10%), gestational diabetes mellitus (93% vs 64%), established diabetes mellitus (97% vs 52%), active genital herpes (77% vs 38%), chronic hypertension (70% vs 47%), pregnancy-induced hypertension (74% vs 49%), renal disease (13% vs 2%), and placenta previa (70% vs 33%)… Conclusion In Washington, most medical conditions and complications of pregnancy that affect mothers are substantially underreported on birth certificates,…

In other words, for virtually every serious pregnancy complication, that information was missing from the birth certificate in more than half the cases.

Even a cursory look at the data showed that the authors assumptions were entirely unfounded. Women in the group characterized as planned C-sections for “no medical indication” had birth certificates that indicated that they had been in labor for hours before the C-section. Although the indications had been absent, it was clear that there must have been indications for the C-section.

In response to pointed criticism in the Letters to the Editor, the authors who had originally looked at births from the 1998-2001 cohort, now looked at births from the 1999-2002 cohort, performing the same analysis but applying an intention to treat methodology. The paper entitled Neonatal Mortality for Primary Cesarean and Vaginal Births to Low-Risk Women: Application of an “Intention-to-Treat” Model was published in February 2008. As the authors explained:

… 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.

This analysis led to very different results:

… 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 now instead of claiming that C-sections increase the risk by a factor of 3, they are claiming that C-sections increase the risk of neonatal death by only half that amount. But the authors still do not address the primary flaw of the study. They really have no idea which C-sections were indicated and which were not. The difference is critical. If only 0.002% of the remaining birth certificates were missing risk data, there would be no difference in mortality in the two groups at all. Based on what we know about the reliability of birth certificate data, there is reason to believe that far more than 0.002% of birth certificates lack the relevant data.

The bottom line is that MacDorman and colleagues never showed that C-section increased the risk of neonatal death by any amount.. They demonstrated an entirely different principle: garbage in, garbage out. When you apply statistical analysis to erroneous data, you reach unsubstantiated, erroneous conclusions.