Reply to Khashan AS & Kenny LC letter. Re: Narchi H, Skinner A. 2010. Overweight and obesity in pregnancy do not adversely affect neonatal outcomes: new evidence. Journal of Obstetrics and Gynaecology 30:679–686

Dear Sir,

We thank Drs Khashan and Kenny for their interest in our paper (Narchi and Skinner 2010). They are surprised that no women in our study were underweight (BMI<18.5 kg/m²) compared to 3% in their own study (Khashan and Kenny 2009). We believe that this is explained by the difference in the ethnicity distribution in the two populations and the association between BMI and ethnicity (whites having lower BMI than others). The absence of underweight women in our study, which had fewer white women, is therefore not surprising.

They are also concerned that only 2.3% of the neonates were classified as small for gestational age (SGA) as compared to their study (6.3% by our calculations). This might be related to the differences in the classification used: while we used the standard gender-specific neonatal growth charts, they used individualised birth weight ratios. The difference is expected, as it has already been shown that customised birth weight centiles diagnose more SGA infants than weight centiles alone (Narchi et al. 2010).

It should not come as a surprise that the crude odds ratios (ORs) and adjusted (aORs) were quite similar for those outcomes where no confounding was present, as compared to other outcomes where we found a difference after adjusting for prematurity, for example (such as retinopathy of prematurity or chronic lung disease). A perfect example of this is that maternal obesity was associated with an increased risk of macrosomic infants and reduced risk of SGA, as confirmed by Khashan’s own study.

We entered the BMI as a categorical variable (with four ordered categories) in a multivariate logistic model. However, we did not calculate the ORs for each BMI category. This is because the four categories had a linear relationship in relation to each outcome, confirmed by the linear increase (or decrease) in the regression coefficient (log odds) per unit increase in the exposure (BMI). The association between maternal BMI and the log odds of each outcome could therefore be easily expressed by a single linear term, rather than by a series of indicator variables representing each BMI category. We are confident that our methodology is sound, as our results corroborate the conclusions of a recent meta-analysis (Heslehurst et al. 2008).

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.