The Washington Post is reporting about a study that looks for various lifestyle correlates of miscarriage (hat-tip to Jessica). It turns out women who eat fruit and vegetables or take dietary supplements are at a 50% lower risk, women who eat chocolate are at a 15% lower risk, and women who are underweight are at a 72% higher risk. Non-physical factors apply as well: planned pregnancies have a 40% lower risk than unplanned ones.
Usually I read the study in question, but this time I can’t find it anywhere. So instead, I’ll have to rely on one critical analysis that focuses on the lack of a clear causal link (and unfortunately doesn’t link to the study).
The researchers took care to try and ensure they were looking at a sample of women representative of the general population. They gathered data collected on surveys from about 6,700 British women, which asked the women about their reproductive histories, considering each pregnancy in turn. They were also asked to complete a lifestyle survey and a food frequency recall questionnaire surrounding their most recent pregnancy. About 600 first trimester miscarriages were reported; but to increase the information on miscarriage risks, women whose latest pregnancy was not a miscarriage were given another questionnaire about their most recent miscarriage. The researchers applied statistical modeling of the data to look for correlations among factors they thought might be important (such as smoking and caffeine) and determine the odds ratios.
This type of study design is called “case-controlled” and is retrospective because it looks backwards through histories trying to explain an adverse outcome among a certain group. There are several caveats to these types of studies. Researchers look for things they think may be related to a health problem and could miss more important ones. The data relies on people’s memories and is subject to recall bias, where people are more apt to remember things they believe related to their illness. It’s the — “Oh, it must have been those burgers I ate” — guilt phenomenon. And, of course, they are looking for correlations, hoping to find ones that are statistically significant, but which can never prove cause.
The best study design would of course be prospective – that is, it would track pregnancies over nine months. Ideally it would also divide the women into two random groups, but that’s only possible with dietary supplements; with everything else, it would have to look for control variables like income, number of hours worked, etc.
Another limitation of the study is that it, of course, only considers the risk of spontaneous abortion once the pregnancy is confirmed. The bulk of spontaneous abortions occur before the pregnancy is even detectable with today’s pregnancy tests; 75% of all concepti fail, compared with 15-20% of confirmed pregnancies.