Friday, August 17, 2007

Medical study concluding for dummies

If you read the abstract of a medical study, you'll be confronted with the conclusion of the authors in most cases. My advice: Do not buy it but draw a conclusion of your own. Here is an example where this task looks not too difficult.

The numbers in the table below are odds ratios for preterm birth, drawn from half a million birth records of Missouri. Births from white mothers and white fathers serve as a reference.

white father black father
white mother1.001.13
black mother 2.102.28

What do you conclude from this table?

I guess that you see a big difference between the upper and the lower row and a much smaller difference between the left and the right column. This is not difficult to explain. Obviously, for black mothers, the odds of a preterm birth are double as high as for white mothers. There is also an effect for black vs. white fathers, but it is by an order of magnitude smaller.

So far so simple. Now, what to you think about the causes of this disparity? The following possible reasons may come to your mind, conditions that could be different in black and in white women:
  • health status
  • income
  • education
  • maternal age
  • stress
  • social support
  • lifestyle
  • etc.
What do the study authors conclude from this table?

The study has been published under this title: Paternal race is a risk factor for preterm birth. You may think this is a typo given the fact that the influence of maternal race is nearly ten times as big. But it is no typo. The authors conclude:
«Paternal black race is associated with an increased risk of preterm birth in white mothers, which suggests a paternal contribution to fetal genotype that ultimately influences the risk for preterm delivery.»
Racists would very likely draw such a conclusion, but maybe the authors know more than we do. They may see the big difference in black and white mothers as a given fact that is widely known and covered by research. They may point out one additional finding, not a big one but a new one. But their conclusion is most likely to be used by racists.

Why the %*¿k do the authors leave aside the main finding and comment on the tiny fraction of a tiny fraction? If you translate the odds ratio to a relative risk and then to an absolute risk difference, this will be close to zero.

I am not a specialist in genetics, gynecology and obstetrics but I have my strong doubts if the genetic explanation of the authors is the most plausible one. I suspect that this is a case of undue jumping from correlation to cause. Nowhere in the abstract I can find any mention of controlling for factors or that the odds ratios are adjusted.

We need more evidence

In search of more facts that may solve this case I have found a study on the black/white difference in neonatal mortality, mostly from preterm births and low birthweight and in the same magnitude as found in our table. Can you believe that mortality at birth may have a genetic cause? I can't.

Photo credit:

Comment by James A. Grant
The big difference in preterm births between black and white mothers is widely known and covered by research. This is not a novel conclusion and likely why the authors did not make comment of it. In addition, there is evidence that correcting for socioeconomic status does not fully account for this difference, leading to an alternative hypothesis: there may be genetic bases for the observed difference.

Why should there be a paternal role in the likelihood of preterm birth particularly since it is only necessary that a male contribute his gametes at conception? Does a genetic explanation here truly seem so outlandish, particularly given others’ work on differing genetic polymorphisms among racial groups.

What do the authors’ conclusions have to do with racism and on what basis do you justify associating (if not accusing) the authors with racism? Empirically, the conclusion the authors reached is reasonable: Paternal black race is associated with an increased risk of preterm birth in white mothers. Causality remains undetermined BUT, given a wider body of research examining racial variations in genetic polymorphisms and the role on preterm births, paternal genetics remains a legitimate hypothesis for the reported study observations.

You can’t believe that mortality at birth may have a genetic cause? I guess you might also not believe that disorders such as Turner’s and Down Syndrome don’t have a genetic cause.
Author's reply
The control for socioeconomic status is fine and obviously explains a part of the difference. But the rest is not necessarily genetic factors; there may be other factors that should be controlled before we can make such a statement.

As you clearly state I do not accuse the study authors of racism. My association comes from the fact that I have read many racist statements based upon scientific findings, and this one may be very ready to be misquoted in such a way.

As to a possible genetic cause of mortality at birth: You are right. I have to correct myself and should put it this way: Can you believe that mortality at birth may have a genetic cause other than a genetic disease? While Turner and Down syndromes are such diseases, a black skin certainly is not.
Final comment
Thanks for your reply, Christian. I enjoy your blog and I agree with your assessment that scientific findings are consistently mischaracterized in the media and, particularly in ethnicity related research, have the potential for gross misuse. So I think you are performing a valuable service in providing a better context for understanding of literature. Keep up the good work.

James A. Grant, Duke University

No comments: