Inflammation Among Women With a History of Gestational Diabetes Mellitus and Diagnosed Diabetes in the Third National Health and Nutrition Examination Survey

The sampling strategy and data collection methods for NHANES III have been previously described (5). We excluded women who were currently pregnant, had missing data regarding pregnancy status, or had previous diabetes or hGDM diagnoses, for a total sample of 6,346 women. Women were classified as having diagnosed diabetes if they reported a diagnosis of diabetes outside of pregnancy, as having hGDM if they reported having a diagnosis of diabetes made only during pregnancy, and as unaffected if they did not have hGDM or diagnosed diabetes. Therefore, the categories of unaffected women (n = 6,015), women with hGDM (n = 87), and women with diagnosed diabetes (n = 244) were mutually exclusionary. In NHANES III, undiagnosed diabetes was assessed in a subsample of nondiabetic individuals who were randomly assigned to a morning fasting examination.

Information on demographic and behavioral factors was collected by interview (5). The poverty index was calculated as the poverty-to-income ratio (6). Measurements of height, weight, and waist circumference were performed in a standardized manner. Leg length was calculated by subtracting sitting height from standing height (7). Measurement procedures for inflammatory markers for ferritin (8), leukocytes (9), and C-reactive protein (CRP) (6) in the NHANES III have been previously described.

All analyses accounted for the multistage, stratified, cluster-sampling design of NHANES III by using survey sample weights. We conducted tests for trend across unaffected women, women with hGDM, and women with diagnosed diabetes across exposure variables. We calculated predicted marginal probabilities and 95% CIs in multivariate models. In a sensitivity analysis, we excluded women who had fasting glucose ≥126 mg/dl and who were also classified as having hGDM or as unaffected, but results did not change. We restricted analyses to only parous women, but results did not change. Statistical analyses were conducted using SUDAAN 9.0 (Research Triangle Institute, Research Triangle Park, NC).

Women with diabetes were significantly older than women with hGDM and unaffected women (46 vs. 32 and 36 years, respectively; P < 0.0001) and more often reported non-Hispanic black ethnicity (P < 0.0001), a family history of diabetes (P < 0.0001), the least favorable poverty-to-income ratio (P = 0.0036), the fewest number of alcoholic beverages per day (P < 0.0001), and the greatest waist circumference (P < 0.0001). Waist circumference increased across unaffected women, women with hGDM, and women with diabetes from 85.9 to 92.4 to 103.2 cm, respectively (P < 0.0001). Few women were nulliparous; all of the women in the hGDM and diabetic groups and >91% of unaffected women had at least one delivery. Women with diabetes had the highest ferritin measurements compared with women with hGDM and unaffected women (135.1 vs. 59.4 and 62.7 ng/ml, respectively; P < 0.0001), the highest leukocyte counts (8.4 vs. 8.0 and 7.2 cells/ml, respectively; P < 0.0001), and the highest CRP levels (0.88 vs. 0.51 and 0.40 mg/l, respectively; P = 0.0012).

When we adjusted for patient covariates, the most striking differences were seen between women with diabetes and unaffected women. After adjustment for demographic and behavioral factors (Table 1, model 1), women with diabetes had greater ferritin, leukocyte, and CRP measurements than unaffected women. After further adjustment for waist circumference (Table 1, model 2), women with diabetes had greater inflammatory marker levels than unaffected women, but the differences in CRP were no longer significant. As in unadjusted analyses, we observed no significant differences in inflammatory markers between unaffected women and women with hGDM. Women with hGDM had lower inflammatory markers than women with diabetes, but the differences were statistically significant only for ferritin, not for CRP or leukocyte counts.

This population-based cross-sectional study suggests that differences in inflammatory markers are greatest between women with diagnosed diabetes and unaffected women. Ferritin levels distinguished women with diabetes from unaffected women and women with hGDM before and after adjustment for other risk factors, but ferritin did not differ significantly between unaffected women and women with hGDM.

Our findings are consistent with results from previous studies of women with hGDM versus unaffected women that have not shown robust associations after adjustment of BMI or other markers of adiposity (1–4). Our findings are also consistent with studies that compared unaffected women and women with diabetes that found more robust associations (10).

This study has several limitations. NHANES III was cross-sectional and may not reflect disease progression in individuals. It is possible that the relationship between inflammation and hGDM would have been stronger had we been able to distinguish which women would go on to develop diabetes and if we were able to adjust for time since delivery. It is also possible that examination of other markers with greater discrimination would have shown an association. There may have been unmeasured confounding with other inflammatory conditions not included in our analysis. Finally, the number of women with hGDM was greater than in most other reports, though still small, leading us to report group differences only when the 95% CIs were mutually exclusive.

In conclusion, the differences in inflammatory markers between women with hGDM and unaffected women were much smaller than those between women with diabetes and unaffected women. Prospective longitudinal studies using inflammatory markers with greater discrimination may elucidate how inflammation progresses in women who develop GDM and then diabetes after delivery.

C.K. was supported by grant K23DK071552 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK).