SEARCH

SEARCH BY CITATION

To the Editor:

The increasing rate of cesarean delivery and the increase in interest in cesarean delivery by maternal request are issues of great significance in contemporary obstetrics, and the politicization of the issue is clear (1). However, it is critical—particularly in the face of intense political debate—that we objectively investigate the causes and safety of these interventions in order to provide our patients with improved care.

MacDorman et al attempt to address the risks of this increasing cesarean delivery rate, and the potential risks of an increased rate of cesarean delivery by maternal request, in their recent retrospective review of the United States linked birth-infant death database (2). They conclude that there is a greater than twofold increased risk of neonatal mortality in patients with cesarean deliveries for “no indicated risk” when compared with matched controls. We would like to emphasize that these results should be interpreted with great caution, if not skepticism.

The analysis of retrospective vital statistics data can be useful, particularly with rare outcomes (such as neonatal mortality). However, as acknowledged by MacDorman et al, they are far from ideal. Although we all strive for accuracy when filling out birth certificates, on a busy night or during a long shift we are susceptible to a failure to achieve perfection. Vital statistics data have been shown to provide accurate accounts of demographics and birthweight; however, procedures, complications, and maternal and neonatal conditions are usually underreported (3). The study by Lydon-Rochelle et al, for instance, showed that birth certificate data demonstrated low “true-positive fractions” for rather fundamental obstetric interventions, and concluded that “researchers should not rely on birth certificate data to detect maternal diagnoses and intrapartum procedures accurately” (4, p 460). For this reason, the statement “It would seem reasonable to expect any bias in the reporting of these items by method of delivery would favor overreporting of risks among cesarean deliveries” (2, p 181) is a bold assumption, a misinterpretation of the weakness of vital statistics data, and a weak foundation for the argument of MacDorman et al’s thesis. Furthermore, the technique of selecting out patients with risks as declared by the birth records may systematically select for patients with incomplete or inadequate reporting on such records. Possibly basing conclusions on a subset of patients with a higher proportion of incomplete records only makes the use of vital statistics databases more problematic.

Despite attempts to apply this study to the cesarean delivery by maternal request population, the cesarean delivery patients included in this study were probably poor representatives of cesarean delivery by maternal request patients. Given the rate of approximately 4 million births per year, the 311,927 cesarean deliveries included in the study represent approximately 2.5 percent of all deliveries in the United States. As Young points out in the editorial published concurrently with MacDorman et al’s paper, the estimates of cesarean delivery by maternal request rates (2.5%–18%) cited at the National Institutes of Health conference “… seemed highly unlikely. These estimates were primarily derived from vital records or hospital discharge data, and assumed that all cesareans with no documented medical indication were therefore by maternal request” (5, pp 171–172). Furthermore, the inclusion of cesarean patients who were in labor complicates the point even further, since most would agree that cesarean delivery during labor presents a much higher risk than an elective, planned cesarean delivery at term before the onset of labor.

The abstract of the article reports mortality risks of 1.77 versus 0.62 per 1,000 for cesarean delivery versus vaginal delivery, which is probably an overestimate. This finding is exclusive of the logistic regression analysis, which would have accounted for congenital abnormalities and events with Apgar scores less than 4, conditions that would be represented in much higher proportions in the cesarean delivery group and that would contribute to significantly higher mortality rates. Thus, to use this figure as the most representative figure of the paper is misleading.

It is critical that we keep in mind that vital statistics records are meant for surveillance, and the conclusions taken from the analysis of such databases are intended for hypothesis generation rather than hypothesis proving. The study by MacDorman et al will stimulate the discussion further; however, we encourage a critical evaluation of the data evaluating the risks and benefits of cesarean delivery versus vaginal delivery so that we can, in the end, counsel our patients to make autonomous and informed choices.

References

  1. Top of page
  2. References