Sunday, September 07, 2008

Are C-section mothers less responsive to baby's cry?

You may have read recently about the widely publicized study by Swain and colleagues claiming that mothers who had C-sections respond to their baby’s cry differently than mothers who had vaginal deliveries (Maternal brain response to own baby-cry is
affected by cesarean section delivery
). The study shows nothing of the kind, and the authors have been utterly irresponsible in the way that they have publicized their findings.

The study itself suffers from so many serious problems that it is hard to know where to begin. At a minimum we can say: The study is too small to draw any conclusions, let alone the conclusions drawn by the authors. We don’t know if there is any validity to the image of brain activity that the authors were looking at. There is no evidence that this image is correlated in any way with maternal care taking.

These problems are easier to under stand in the following thought experiment:

Suppose I asked my children to toss coins. My son tossed a coin 6 times and got 6 heads, and my daughter tossed a coin 6 times and got 6 tails. Then suppose I wrote a study claiming that gender determines whether a coin lands heads or tails. Wait a minute, you’d say, your study doesn’t prove anything, and you’d be right. My study would prove nothing for the exact same reasons that the Swain study proves nothing.

First, my study, like the Swain study, is underpowered. I haven't included enough coin tosses for my son or my daughter. We know from probability theory that if I repeated the experiment with my son tossing the coin 6000 times and my daughter tossing the coin 6000 times, they would get the exact same proportion of heads to tails; they would both get 50:50. My results are not valid because I didn’t include enough coin tosses.

Swain and his co-authors looked at the brain imaging patterns of 6 women who had vaginal deliveries and found that they were different from the 6 women who had C-sections. Just like the heads-tails study, the results could simply reflect the fact that the study is far too small.

Second, drawing the conclusion in my coin experiment that gender determines the result of a coin toss rests on the assumption that correlation equals causation. However, we know from the rules of statistics that correlation does not demonstrate causation. Correlation means that two events appear to be related. Causation means that one event caused the other. In my coin toss experiment, gender and coin toss result appeared to be correlated, but that does not mean that gender caused the coins to land heads or tails. It was just a coincidence.

Similarly, in the Swain study, the two different modes of birth appeared to be correlated with brain image, but that does not mean that the mode of birth cause the specific brain image pattern. It could just have easily been coincidence.

Third, my study claiming that gender determines the result of a coin toss would have rest on another flawed assumption, that there is something fundamentally different, beyond appearance, between a coin landing heads and a coin landing tails. In reality, though, there is no difference; it’s just a matter of chance whether a coin lands heads or tails.

Similarly, Swain and colleagues have made a seriously flawed assumption that the two different brain image patterns they observed reflect a fundamental difference in the actual reaction of the mother, not just a matter of chance. In reality, the authors present no evidence that the brain image pattern has anything to do with the mother’s response to her baby’s cry.

This study shows nothing. It could potentially represent an interesting finding that deserves more investigation, or it could be entirely the result of chance. It is extremely irresponsible for the authors to claim that they showed that mode of delivery determines a mother’s response to an infant. It is no better than a coin toss study claiming that gender determines the result of a coin toss.


0 Old Comments: