Statistics from Altmetric.com
Editor,—In a recent paper Narchi and Kulaylat1 analysed the prevalence (erroneously termed incidence2) of trisomy 21 in children of mothers with diabetes (7/1870) and in non-diabetic mothers (28/20 430) and found it to be significantly higher in the former. All seven cases in the diabetes group occurred in mothers with gestational diabetes.
The authors concluded, that: (1) Maternal gestational diabetes is an independent risk factor for Down’s syndrome, irrespective of maternal age, as in an analysis of their data stratified by age, all five age groups showed a higher relative risk for Down’s syndrome in diabetic mothers; (2) Down’s syndrome should be added to the list of congenital anomalies known to occur more frequently in infants of diabetic mothers.
Table 2 in their paper does not fully support their hypothesis. As can be seen from their data, age is a confounding factor for gestational diabetes as fewer than 5% of pregnant women develop this condition in the age group below 30 years but more than 23% in those over 44 (assuming that females with pre-gestational diabetes are distributed randomly over the age groups—the authors combine both types of diabetes in this table). An analysis of their age stratified groups using the Mantel-Haenszel method reveals an odds ratio of 2.33 with a 95% confidence interval (CI) of 0.99 to 5.48 for the whole study population, as opposed to the authors’ unstratified analysis of the whole group (relative risk 2.75; 95% CI, 1.2 to 6.29).
Much more important for the discussion of the results is a violation of the rules of causal inference. The authors laudably made a sharp distinction between mothers with gestational diabetes (n = 1748) and pre-gestational diabetes (n = 122) and found cases (seven in all) with trisomy 21 only in the former group. Thus their analysis of diabetes as a risk factor for Down’s syndrome is only valid for gestational diabetes. Although the use of the term risk factor in the literature is rather loosely defined, in aetiological research for an exposure in the broadest sense (gestational diabetes in this study) to become a genuine risk factor it must precede the occurrence of the outcome (here, trisomy 21). As non-dysjunction leading to trisomy 21 occurs before or shortly after fertilisation, gestational diabetes with onset during pregnancy can hardly be a risk factor for trisomy 21. The study at hand does not justify adding trisomy 21 to the known congenital anomalies associated with pre-gestational diabetes in the mother.
What Narchi and Kulaylat may have shown is that a woman with a trisomy 21 conceptus is more likely to develop gestational diabetes, although the attributable risk may be small, and age, as can be seen from their data, seems to be much more important.
Drs Narchi and Kulaylat comment:
We read with interest Pelz and Kunze’s comments on our study and thank them for the opportunity to clarify our findings.
We agree that the term prevalence is more appropriate than incidence, but our data do not support maternal age as a confounding factor. Although maternal diabetes (mainly gestational) was more common with advancing age, when the prevalence of Down’s syndrome was broken down by maternal age, it remained 1.6 to 3.01 times more common in infants born to mothers with gestational diabetes within each age group. If maternal age was a confounding factor, the prevalence of Down’s syndrome, although increasing with advancing maternal age, would not be expected to be different within the same age group regardless of the presence of gestational diabetes. Even using the Mantel-Haenszel method for age stratified groups as suggested by Pelz and Kunze, the relative risk for a diabetic mother to have a baby with Down’s syndrome was 2.34 (95% CI, 1.02 to 5.33), not very different from the crude data relative risk of 2.35 (95% CI, 1.2 to 6.2) in our initial analysis.
We also disagree that the rules of causal inference were violated: we implied an association rather than a causal relation, as we could prove none. We made it very clear that as mothers with gestational diabetes were euglycaemic at conception, hyperglycaemia was ruled out as a potential mechanism for non-dysjunction; and we suggested the need for further studies of advanced biological aging, autoimmunity, and the role of biochemical factors such as apolipoprotein E. The lack of a direct causal relation should not lead to complacency in ignoring the increased prevalence of Down’s syndrome in infants of mothers with gestational diabetes.