Goer fails to address specific criticisms of the MacDorman studyAs I mentioned several days ago, I was able to engage Henci Goer in a discussion of the MacDorman study that purports to show that C-sections lead to an increased neonatal death rate. After she responded to my initial comments with a critique, I wrote back with substantive, detailed criticsms of the study. In return, she was unable to offer any evidence to rebut the fact that the study suffers from such serious flaws that we must conclude that it tells us essentially nothing about the risk of neonatal death from C-section.
What is most ironic is that the MacDorman study is based on birth certificate data, just like the Pang study. Goer objects to that study because, as she points out, birth certificate data are not always accurate. Furthermore, although Pang et al. show an increased rate of neonatal death at homebirth, they did not show a direct relationship between the neonatal deaths and the place of birth. The EXACT same criticisms apply to the MacDorman study, yet Goer is willing to ignore these problems in this study.
The unwillingness to address these specific criticisms is basically what I expected. It is easy to write for the general public who have neither the knowledge base nor the familiarity with statistics to recognize the serious problems with so many of the arguments. It is much harder, dare I say impossible, to use the same tactics on medical professionals.
I will excerpt highlights of the discussion here, and inside the thread I'll post the whole discussion.
My comments in response to Goer's initial defense of the study:
"Let's start with the most important point. In order to be able to say that C-sections caused the neonatal deaths, you need to be able to show that the C-sections plausibly caused the deaths. Since none of the deaths in the C-section group appears to be caused or even related to C-section, you can't draw the conclusion that the C-sections caused the deaths."Goer's response does not even attempt to address these serious criticisms:
"... birth certificate data is notoriously unreliable for leaving out risk factors and complications. For example: How Well Do Birth Certificates Describe the Pregnancies They Report? The Washington State Experience with Low-Risk Pregnancies, Dobie, Baldwin, et al., Maternal Child Health Journal, 2:3, September 1998.
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, American Journal of Obstetrics and Gynecology, Volume 193, Issue 1, 125-134, July 2005, M. Lydon-Rochelle, V. Holt, et al. 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,...
Look at the results of this study. For virtually every serious pregnancy complication, more than half the cases appeared on the birth certificate as having no risk factors. It seems pretty clear that there is a serious problem with relying on birth certificate data to determine pregnancy complications..."
"Let's see . . . You seriously misrepresent the findings of the MacDorman 2006 study by cherry picking, distorting, and, most problematic of all, leaving out what doesn't suit you.You offer no evidence to support your own position beyond "I believe it therefore it must be so."All I can say is you seem to be suffering from a case of "my mind is made up; don't confuse me with the facts."I wrote back:
"If that were the case, you could demonstrate, point by point (with appropriate citations) exactly where I have gone wrong, so that others who are knowledgeable about the scientific method and statistical analysis could decide who is correct. You have chosen not to do that, and I can only conclude that you cannot defend a study based on birth certificate data that is notoriously grossly unreliable.The complete discussion is posted in the comments section.
I find this especially ironic in light of your vociferous criticism of the Pang (2002) study. I agree with certain of your objections to the Pang study, so it is quite surprising that you are willing to overlook the exact same problems that occur in the MacDorman study. In your essay When Research is Flawed: The Safety of Home Birth, you say: "The study doesn't investigate whether the choice to birth at home was at fault in neonatal deaths: The investigators note that 10 of the 20 babies who died had diagnoses of congenital heart disease or respiratory distress. An additional three babies had major congenital anomalies. They do not say whether the deaths occurred at home or in hospital, nor do they speculate about how the choice of birth setting may have affected neonatal outcome in individual cases. The investigators could have reviewed the medical records on these 20 deaths, which would have provided a more accurate picture of whether home birth care affected outcome, but they did not."
So in the case of the Pang study, you thought it was quite important that the investigators actually document the connection between the cause of death and the place of birth. I could respond that "Not knowing the exact connection is not essential. As epidemiologists observe, you don’t need to know that mosquitoes are the vector in order to know you need to drain the swamp." I wouldn't do that because knowing the connection IS important. It is important in the interpretation of the Pang study and it is equally important in the MacDorman study. I will venture to say that you know as well as I do that C-sections did not cause the neonatal deaths in the MacDorman study.
In discussing birth certificate data you say: "The Washington State birth certificate form, from which the data for the study were collected, provides no means of ascertaining whether the woman intended to birth at home. The way the form is constructed, a woman who gave birth precipitously at home with no attendant and was brought into the hospital by ambulance would be considered a “planned home birth” according to Pang et al." There are reasons to believe that the birth certificate data in the Pang study is not entirely accurate, and the exact same reasons apply to the birth certificate data in the MacDorman study. The issue is hardly theoretical. As the two papers I cited show, birth certificates are notoriously unreliable for determining pregnancy complications.
The MacDorman study suffers from the exact same problems as the Pang study. Both papers fail to demonstrate a connection between the deaths and the place of birth (Pang) or mode of birth (MacDorman). Both studies rely on birth certificate data which is known to be unreliable.
Personally, I think your most compelling statement about the Pang study was this: "The investigators could have reviewed the medical records on these 20 deaths, which would have provided a more accurate picture of whether home birth care affected outcome, but they did not." The same situation applies here. The authors could have reviewed the medical records of the neonatal deaths to provide a more accurate picture of whether mode of delivery affected outcome, but they did not. The MacDorman study tell us NOTHING about the whether C-section cause neonatal deaths.
I suspect that you are not going to be interested in discussing any other studies. If so, I will be disappointed. I'd love to discuss the Johnson and Daviss study about homebirth. People quote it often because it has the fewest methodological errors of any studies done by homebirth advocates. Nonetheless, it does not show homebirth to be as safe as hospital birth since it uses the wrong control group.
I don't suppose I can interest you in discussing that."
Labels: Henci Goer