What would you say if I told you that a study was just published that showed that visiting Weight Watchers causes obesity?
The authors looked at more than 150,000 people across multiple states and found that those who visited Weight Watchers were 25% more likely to be obese. This relationship was true for all subgroups: male, female, white, African-American, and across all ages.
I can imagine you rolling your eyes, but wait! How about if I added the the authors had provided all manner of charts, including funnel plots, and corrected for publication bias?
I bet I know what you would say (after you picked yourself up from rolling on the floor laughing). You’d point out that Weight Watchers doesn’t cause obesity. You’d probably point out that the people who visit Weight Watchers differ substantially from the people who never visit Weight Watchers. At a minimum, the people who visit Weight Watchers are more likely to be obese to begin with.
You’d be right. In other words, the study is meaningless if the authors did not correct for confounding variables. The investigators might have identified a correlation between visiting Weight Watchers and obesity, but that doesn’t mean that visiting Weight Watchers causes obesity. In fact the opposite is true, being obese makes it more likely that you will visit Weight Watchers.
There is no study that shows that visiting Weight Watchers causes obesity, but if you understand that example, you will understand why the new study published in PLOS ONE, Mode of Delivery and Offspring Body Mass Index, Overweight and Obesity in Adult Life: A Systematic Review and Meta-Analysis, by Darmasseelane, Hyde, Santhakumaran, Gale, and Modi is equally meaningless.
What did the authors do?
15 studies with a combined population of 163,753 were suitable for inclusion in the meta-analysis. Comparing all CS to VD in pooled-gender unadjusted analyses, mean BMI difference was 0·44 kg·m-2 (0·17, 0·72; p = 0·002), OR for incidence of overweight was 1·26 (1·16, 1·38; p<0·00001) and OR for incidence of obesity was 1·22 (1·05, 1·42; p = 0·01). Heterogeneity was low in all primary analyses. Similar results were found in gender-specific subgroup analyses. Subgroup analyses comparing type of CS to VD showed no significant impact on any outcome.
In other words, they found that for people born by C-section, the risk of obesity in later life was increased by 22%. The association was robust, meaning that it was found in most subgroups, but it is actually rather small.
But it is well known that women who undergo C-section differ substantially from those who do not. They are more likely to be obese themselves, more likely to be carrying large babies and more likely to have gestational diabetes, which impacts metabolism. The authors are aware of this, and even mention it, but they do not correct for these confounders, making their conclusions rather suspect.
In media interviews, the authors have indulged themselves in flights of fancy:
Study co-author Dr. Matthew Hyde suggested possible ways that C-sections might influence later body weight.
“The types of healthy bacteria in the gut differ in babies born by cesarean and vaginal delivery, which can have broad effects on health,” Hyde said in the news release. “Also, the compression of the baby during vaginal birth appears to influence which genes are switched on, and this could have a long-term effect on metabolism.”
No, Dr. Hyde, we have as yet no evidence that the differences in gut flora in the first few days of life have any impact on health, let alone broad effects. Furthermore, I am unaware of any studies that show that “compression” of the baby influences which genes are switched on, let alone whether that affects metabolism.
Speaking to The New York Times, Dr. Hyde said:
… [A] woman cannot make a reasoned choice [about a C-section] unless she’s fully informed not only about the short-term outcomes, but also the long-term outcomes. We want to give you the data you need to make an informed decision.”
But you haven’t given us that data, Dr. Hyde. You haven’t given us any evidence of any causal relationship.
In the paper, the authors are a bit more circumspect:
… [S]everal factors associated with increased risk of CS are also associated with increased BMI in offspring, including high maternal BMI , gestational diabetes , and lower socioeconomic status . Of these, maternal obesity is probably the most significant confounding factor in the relationship between CS and offspring BMI. This needs elucidation in datasets which can be properly controlled for maternal BMI. (my emphasis)
Although the authors analyzed data from more than 150,000 individuals, although the relationship between C-section and obesity was found in most (but not all) subgroups, although the authors provide all sorts of elaborate tables and charts, they never corrected for confounding variables, meaning that they are not entitled to conclude anything about a causal relationship between C-sections and obesity.