This simple observational study looks at perinatal mortality in women with gestational diabetes, pre-gestational diabetes and gestational IGT over a period of three years. My main concern relates to the use of ‘pregnancy’ rather than ‘woman’ as the unit of analysis (see point 1 below). The ‘pairwise’ statistical comparisons between the groups (see point 4 below) also needs to be clarified.

Design and Methods:

1. The study takes the total number of pregnancies rather than the number of women to estimate the incidence of the various outcome rates and the relative risks. It is the latter that should be used as the ‘unit of analysis’. Only the first pregnancies occurring during the study period should be included for each woman.
2. No power statement is given. The large 95% confidence intervals for some of the relative risk estimates (Table 2) are of slight concern. However none of the statistically non-significant findings appear to be clinically significant.

 

1. Does the ‘background population’ include women with diabetes or IGT? It would be more appropriate for these data to exclude such women and hence be independent of the three diabetic/IGT groups.
2. Were appropriate multiple comparison tests used to look at ‘pairwise’ comparisons between the three groups following one-way analyses of variance (see p-values quoted in Table 3 and Page 11)?
3. The ‘mean+/-SD’ notation should be changed to ‘mean(SD)’.

Reviewer: Julie Morris