The latest issue of The Lancet contained an interesting article describing a positive association between birthweight and the amount of weight a woman gained during pregnancy.

The authors were able to compare several pregnancies by the same mothers, thereby controlling for genetic or hereditary factors.  I liked the within-subjects design and thought it was a novel use of birth certificate data.  In a sample of 500,000 women and 1.1 million term (37-41 weeks) infants, the authors found birthweight increased linearly with maternal weight gain.  On average, each kilogram gained by the mother resulted in an infant born about 7 grams heavier.

What really caught my attention (and prompted this post) was their conclusion that “obesity prevention efforts targeted at women during pregnancy might be beneficial for offspring.”  The reason I balked at this interpretation is that they established statistical significance, but didn’t demonstrate that the association was of practical significance.  My (admittedly basic) understanding is that maternal weight gain is associated with an infant’s birthweight, and infant birthweight is associated with being obese later in life.  But, how much of the variation of birthweight (and obesity later in life) is attributable to maternal weight gain during pregnancy?

I did some exploring using the 2000 NCHS natality files for the United States. (I had the file handy.)  I restricted the sample to their criteria–singleton births between 500 and 7,000g, gestational age 37 to 41 weeks. I forgot to exclude maternal diabetes, and this is simply a cross-sectional design.

That leaves 2.6 million births, and here is a bin plot of maternal weight gain in pounds (x axis) versus birthweight (y axis). Click for larger image:

A couple of things stood out:

• People love to round to the nearest 5 pounds*
• The loess smoothing line (with 95% confidence interval) shows an overall positive association
• There is a lot of variation at any every level of maternal weight gain.
• ggplot2 is seriously awesome, but R used almost all 8GB of my RAM loading the STATA file from NCHS.

In a linear model, an infant gained 14g for each kg gained by the mother (I converted pounds to kilograms to match the paper). This is about double what was found in The Lancet study–probably because they had a superior study design and were able to control for individual/genetic differences. Although the paper didn’t present unadjusted or univariate associations, it would seem that the within-subjects design helped control for confounders such as pre-pregnancy BMI.

The r-squared of my basic linear model was 3.078%.  This means that maternal weight gain only explains 3% of the variation in birthweight, leaving nearly 97% of the variation in birthweight unexplained or due to other factors.  I really wish the paper would have included this result in the article. It would seem that controlling maternal weight gain among term infants would only affect birthweight in a small way.   This could, however, have significant implications when applied to a large population.

As the authors note, they were not able to follow the infants to observe later health outcomes.  It also would not surprise me if maintaining healthy maternal weight before, during, and after pregnancy could help prevent obesity in offspring through pathways other than birthweight. It’s pretty interesting stuff.

*It could be argued that people have an even stronger preference for rounding to the nearest 10 pounds–check out the histogram of maternal weight gain :