Gestational diabetes mellitus (GDM), defined as carbohydrate intolerance with first recognition during pregnancy,1 is a common form of prediabetes, affecting up to 14% of pregnancies every year.2 Although most women return to normal glucose tolerance after delivery, GDM carries a great number of complications for both children and mothers.1,3,4 It is now well established that affected women are not only at high risk for recurrent GDM in future pregnancies5 but also for developing type 2 diabetes mellitus in later life.1,4,6 In this context, cumulative incidence rates of 2.6–70% in 6 weeks to 28 years of observation with a markedly increase in the first 5 years after the index pregnancy have been reported.2 Particularly, in a large meta-analysis, Bellamy et al revealed a sevenfold higher risk for diabetes manifestation in participants with a GDM history as compared with women with normoglycemic pregnancy.7
Today, there are few long-term observations available and they are mostly limited to a short period of follow-up or specific ethnic groups.8 Thus, it is not definitely clear which women affected by a recent episode of GDM are under particularly high risk for developing type 2 diabetes mellitus in later life. However, it is of clinical relevance. Once identified as a high-risk group, affected participants could take potential benefit from lifestyle modification9 and risk perception.10
All commonly used definitions of the metabolic syndrome contain comparable parameters of dysglycemia and metabolic alterations strongly associated with incident diabetes.11 Recently, it has been suggested that glucose intolerance during pregnancy is associated with an underlying latent metabolic syndrome. Consequently, a combination of both glycemic and lipid alterations might identify a subgroup at particularly high risk for complications in the future.12 Indeed, a more adverse lipoprotein profile as well as elevated glycemic parameters 3 months after a diabetic pregnancy were shown to be independently associated with early progression to dysglycemia within 1 year postpartum.13 However, threshold values estimating the risk for developing type 2 diabetes mellitus at an early stage after an index pregnancy complicated by GDM are needed.
Thus, the aim of the present study was to assess parameters characterizing the metabolic syndrome (elevated triglycerides, systolic and diastolic blood pressure levels, low high-density lipoprotein [HDL] cholesterol, dysglycemia as well as parameters of body composition) in participants with a recent history of GDM and their association with incident diabetes. A further objective was to estimate a risk factor constellation with the most predictive effect for developing type 2 diabetes mellitus using parameters of the metabolic syndrome up to 10 years of follow-up.
MATERIALS AND METHODS
This is part of the study of the Vienna Post-Gestational Diabetes Project, which aimed to assess parameters associated with diabetes manifestation in women with a recent history of GDM as the primary objective. A detailed description of the study design was reported previously.14 In short, we consecutively included 120 women with a recent history of GDM, attending the diabetes outpatient clinic of the Medical University of Vienna during pregnancy within 3–6 months after delivery. Gestational diabetes mellitus was diagnosed according to the guidelines of the Fourth International Workshop conference on GDM with a 75-g oral glucose tolerance test (GTT).1 We excluded women with overt type 2 diabetes mellitus at the time of recruitment (baseline); thus, we studied 110 recent cases of GDM. At enrollment, all women with recent GDM received general information about their risk of diabetes and potential prevention strategies, including dietary advice and instructions on physical exercise, as recommended. Additionally, 41 women, with normal glucose tolerance during and after pregnancy and without any risk factors for diabetes, any severe chronic diseases, kidney or liver diseases, chronic inflammatory diseases, or chronic medication with drugs known to influence carbohydrate metabolism, served as a control group to compare parameters at baseline. The recruitment of participants with recent GDM and normal glucose tolerance was performed between June 1999 and December 2003 at the Departments of Gynecology and Obstetrics as well as Internal Medicine III (Fig. 1).
All participants received a complete metabolic characterization at baseline (oral GTT as well as examinations of routine laboratory, body composition, and blood pressure). In addition, recent GDM was defined as insulin-resistant (insulin sensitivity index less than 2.8/10–4 min−1/[microunits/mL]) as described previously in detail.14,15 Insulin sensitivity derived from baseline oral GTT was estimated by the quantitative insulin sensitivity check index.16
Until December 2010, women with recent cases of GDM were invited annually for re-examination (oral GTT and biometric parameters) for a maximum of 10 years. A manifestation of overt diabetes was diagnosed if fasting plasma glucose levels (FPG) or 2-hour oral GTT levels exceeded 126 mg/dL or 200 mg/dL, respectively.
Because at the beginning of the study there were no larger prospective studies on the conversion rate to overt diabetes in women with a history of GDM in Austria available, the present observational study is more of a descriptive character. However, based on the results of previous studies in white women with GDM during pregnancy and of previous experience in this study population, we estimated a conversion rate of 20% during follow-up for 5 years2 and estimated a dropout rate of 10% during the first 5 years.
This study was approved by the local ethics committee (Ethics Committee of the Medical University of Vienna) and performed in accordance with the Declaration of Helsinki. All participants gave written informed consent to participate in the study.
For this study, we used metabolic cutoff values for metabolic disturbances, which are recommended by the International Diabetes Foundation 2005 (FPG 100 mg/dL or greater, waist circumference 80 cm or greater, systolic and diastolic blood pressure levels 130/85 mmHg or greater, triglycerides 150 mg/dL or greater, HDL cholesterol less than 50 mg/dL).17 Because 2-hour oral GTT glucose levels have been reported to be a better screening tool for glucose alterations in women as compared with FPG levels alone,18 we additionally included 2-hour oral GTT serum glucose levels 140 mg/dL or greater as a predictive variable and further body mass index greater than 30 kg/m2, which may be used as a risk marker of body composition by another definition of the metabolic syndrome.18,19 Further exploratory variables were treatment with insulin during pregnancy, a positive family history of type 2 diabetes mellitus, non-European origin, and age older than 35 years.
Continuous variables were summarized by means, standard deviations, 95% confidence intervals (CIs), categorical variables by counts, and percentages and were compared with Student's t test and Fisher's exact test, respectively. Comparisons of metric-scaled variables in three groups were based on Fisher's protected least significant difference tests. The time to an event was summarized by median values as well as interquartile ranges and ranges from minimum to maximum.
Cox regression models, Kaplan-Meier estimators, and log-rank tests were used to assess the association between exploratory variables at baseline and the development of diabetes up to 10 years of follow-up.
For optimized variable selection, we used the Least Absolute Shrinkage and Selection Operator method for Cox models. By this method, the regression coefficients are penalized toward zero in relation to the maximum likelihood estimates by a tuning parameter (λ) to reduce overfitting as a result of collinearity of independent variables.20 Lambda was fitted by 50-fold crossvalidation. Thereafter, sequential analysis of deviance was performed in the unpenalized model (covariables were added sequentially from high to low Least Absolute Shrinkage and Selection Operator fitted regression coefficients) to identify the most predictive independent parameters for diabetes onset in later life.
Statistical analysis was performed with R 2.11.1. A two-sided P≤.05 was considered statistically significant. There was no adjustment for multiple statistical testing.
In 10 years of follow-up, 23 recent GDM participants (21.3%) progressed to overt diabetes (median time to diagnosis 3.0 years, interquartile range 2.0–4.5 years, range 9 years). Patients with no reported diabetes manifestation had a median follow-up of 5.0 years (interquartile range 4.0–7.5 years, range 10 years). Comparisons of baseline characteristics between women with recent cases of GDM who progressed to overt diabetes, normal glucose tolerances, and those with no progression or censored during follow-up are shown in Table 1.
Univariable longitudinal analyses are given in Table 2 and revealed that most diagnostic markers for the metabolic syndrome were significantly associated with the risk for overt diabetes up to 10 years of follow-up. However, no significant effect on diabetes manifestation was observed for a positive family history (hazard ratio 2.22, CI 0.87–5.69, P=.087) and non-European origin (hazard ratio 1.29, CI 0.30–5.57, P=.730). Because waist circumference 80 cm or greater appeared to be not associated with later diabetes incidence (hazard ratio 2.29, CI 0.53–9.83, P=.253), waist circumference 88 cm or greater was used as a cutoff according to another definition of the metabolic syndrome.19
Among significant predictors observed in univariable analysis, 2-hour oral GTT 140 mg/dL or greater, age older than 35 years, and HDL cholesterol less than 50 mg/dL were identified as the best predictors (Table 2). Kaplan-Meier plots for these variables are given in Figure 2A–C. As shown in Figure 2D, the effect of these variables was additive. In women with recent cases of GDM with no selected risk factors, only one participant (2.5%) progressed to diabetes. Diabetes incidence was even more pronounced in those participants with two or three selected risk factors: 12 women (54.5%) progressed to overt diabetes as compared with those who showed only one risk factor; 10 women (20.8%) progressed to overt diabetes (hazard ratio 3.2, CI 1.4–7.7, P=.008). The selected variables remained also significant in a full adjusted multivariable model even when age at baseline was included as a covariate.
Those patients who developed diabetes showed a mean increase in body weight of 3.4 kg from baseline to diagnosis. However, this increase was not significant using Student's t test for the paired sample. Furthermore, we found no indication that participants with diabetes manifestation changed categories of biometric parameters (from waist circumference 88 cm or less to waist circumference greater than 88 cm or body mass index 30 kg/m2 or less to body mass index greater than 30 kg/m2) during follow-up by using the McNemar test (data not shown).
The annually dropout rate in the first 4 years was approximately 3.0% but increased to 6.6% after including the next period. However, we are able to exclude a bias resulting from loss in follow-up, because we found no indication that variables of interest were related to censoring status. Only patients of non-European origin had a significantly increased chance to be lost during follow-up (hazard ratio 2.61, CI 1.33–5.13, P=.004).
In this prospective long-term observation, we assessed parameters of metabolic alterations (dysglycemia, dyslipidemia as well as adverse blood pressure profile, body composition) in women after pregnancy with GDM and the incidence of diabetes. Especially, if parameters exceeded the recommended threshold values of the metabolic syndrome, we found a predictive effect on the incidence of overt type 2 diabetes mellitus up to 10 years of follow-up. Multivariable analysis identified 2-hour oral GTT levels 140 mg/dL or greater as well as HDL cholesterol less than 50 mg/dL and age older than 35 years as the best predictors and revealed an additive effect of these variables, because those women with two or more risk factors were under a higher risk as when only one risk factor was present.
Based on these findings, we agree with Retnakaran et al, who found a tight association of glucose intolerance during pregnancy with the prevalence of the metabolic syndrome 3 months after index pregnancy.12 Our data support the suggestion of an “underlying latent metabolic syndrome”12 in women with GDM. In addition, our study showed that assessing the presence of such latent metabolic alterations early after GDM pregnancy might help to identify those women at the highest risk of developing the disease in their future. Furthermore, identification of simple parameters predicting diabetes at follow-up is of importance for future prevention strategies in women at risk. At present a 2-hour oral GTT after GDM pregnancy is recommended by the American Diabetes Association21; however, this examination alone fails to estimate the risk for developing more severe glycemic disorders in some cases.13
A recent 12-year follow-up study has investigated the association of clinical variables with incident type 2 diabetes mellitus in 72 Hispanic women after pregnancy complicated by GDM and identified an association between impaired insulin sensitivity as well as β-cell compensation with risk for diabetes manifestation.8 In this context, another follow-up study showed a more adverse lipid profile (higher triglycerides and low-density lipoprotein cholesterol, lower HDL cholesterol) as well as higher body mass index, C-reactive protein, leptin, and lower adiponectin levels in women with an early progression to overt diabetes.13 We agree with most of these important findings; however, regarding the use of adipokines for risk assessment, it has to be considered that measurement is costly and threshold values are lacking at present.
We tried to solve the problem by using recommended cutoff values for markers, characterizing the metabolic syndrome, and revealed a significant association with diabetes incidence in most of the observed parameters. Interestingly, waist circumference greater than 80 cm was the only International Diabetes Foundation marker, in which univariable analysis failed to show a significant association. However, waist circumference became significant when 88 cm was used as a cutoff value, recommended by the National Cholesterol Education Program Adult Treatment Panel III. In addition to the commonly used International Diabetes Foundation parameters, we included 2-hour oral GTT glucose levels 140 mg/dL or greater as a further predictor, because impaired glucose tolerance is a better marker for dysglycemia, especially in women18 and particularly in participants with a history of GDM.22 Indeed, this variable was shown to be of higher predictive value as compared with FPG in addition to higher age and more adverse HDL cholesterol profiles by multivariable models. Taken together, the high-risk parameters identified in this study are not fully comparable to those defined by the International Diabetes Foundation to specify the metabolic syndrome. As an explanation, it has to be mentioned that the metabolic syndrome itself was not designed to be an optimal predictor and therefore might not represent the best predictive risk factor constellation.23 However, a modification of the metabolic syndrome as actually recommended, including 2-hour glucose (oral GTT) from World Health Organization criteria and waist circumference greater than 88 cm from National Cholesterol Education Program Adult Treatment Panel III criteria19 as well as higher age (older than 35 years), could be a valuable but simple GDM-specific risk score. Particularly, impaired glucose tolerance (IGT) was shown to be the best predictive marker in multivariable analysis. This is also corroborated by the Diabetes Prevention Program, in which women with IGT with a history of GDM showed a 71% higher incidence of diabetes as compared with IGT women without former GDM.22 In addition, Ratner et al suggest a markedly greater effect of treatment with metformin or lifestyle modification in women with recent cases of GDM with IGT compared with other groups at risk. Furthermore, a protective effect of glitazones has been assumed.24,25 Nevertheless, the use of postpartum screening for alterations in glucose metabolism is limited as a result of some barriers for screening, eg, uncertainty regarding screening guidelines.26 However, our data clearly recommend the use of postpartum oral GTTs and indicate that especially dyslipidemia and age are of further prognostic relevance in addition to and also independent from glucose values. Insulin treatment during pregnancy might be a further predictor, consistent with previous publications2 but also confirmed by univariable analysis of our data.
Although this was not a main outcome of the present study, weight gain is a commonly accepted risk factor for diabetes and was shown to increase the incidence of the disease also in Hispanic women after GDM.8 The increase in body weight from baseline to diagnosis was only 3.4 kg in our cohort of participants developing diabetes during follow-up. This might be attributable to the prospective study design and to a higher risk perception in our cohort, because all women with recent GDM were advised on their risk and thus encouraged to maintain normal body weight.
The advantages and disadvantages of this study may be summarized as follows. Our sample characterizes a special cohort of prediabetic women by complete metabolic examinations early after index pregnancy and by a very long duration of follow-up with annual visits. However, this is an observational trial and, therefore, we suggest that further interventional studies (especially lifestyle modification or pharmacotherapeutic approaches) are needed to assess the best treatment options for women with recent GDM identified to be at highest risk. Further studies should also focus on the offspring, because there is strong evidence that lipid profiles of diabetic mothers are associated with abnormal fetal growth and might also play a key role in programming the susceptibility for atherosclerosis in the newborns' later life.27,28 The absolute numbers of events, dichotomizing of continuous variables as well as overfitting are limiting factors for variable selection procedures. Because our sample size is comparable to another recent study in Hispanic women,8 bias resulting from overfitting was reduced additionally by performing a variable shrinkage procedure in this study. We think that the use of dichotomized variables resulting from a priori clinically established and recommended threshold values is warrantable.
In summary, this study gives evidence for a strong association between type 2 diabetes mellitus incidence and risk factors, characterizing the metabolic syndrome in women with a recent history of GDM. Particularly, 2-hour oral GTT levels 140 mg/dL or higher, HDL cholesterol less than 50 mg/dL, and additionally age older than 35 years were shown to be the best independent predictors in multivariable analysis and might have additive effects. Therefore, we recommend defining a very high risk for diabetes manifestation after GDM if at least two of these factors are present after delivery. Further intervention studies are necessary to focus on therapeutic strategies in women with recent GDM at particularly high risk.
1. Metzger BE, Coustan DR. Summary and recommendations of the Fourth International Workshop-Conference on Gestational Diabetes Mellitus. The Organizing Committee. Diabetes Care 1998;21(suppl 2):B161–7.
2. Kim C, Newton KM, Knopp RH. Gestational diabetes and the incidence of type 2 diabetes: a systematic review. Diabetes Care 2002;25:1862–8.
3. Retnakaran R. Glucose tolerance status in pregnancy: a window to the future risk of diabetes and cardiovascular disease in young women. Curr Diabetes Rev 2009;5:239–44.
4. Kjos SL, Buchanan TA. Gestational diabetes mellitus. N Engl J Med 1999;341:1749–56.
5. Kim C, Berger DK, Chamany S. Recurrence of gestational diabetes mellitus: a systematic review. Diabetes Care 2007;30:1314–9.
6. Buchanan TA, Xiang AH. Gestational diabetes mellitus. J Clin Invest 2005;115:485–91.
7. Bellamy L, Casas JP, Hingorani AD, Williams D. Type 2 diabetes mellitus after gestational diabetes: a systematic review and meta-analysis. Lancet 2009;373:1773–9.
8. Xiang AH, Kjos SL, Takayanagi M, Trigo E, Buchanan TA. Detailed physiological characterization of the development of type 2 diabetes in Hispanic women with prior gestational diabetes mellitus. Diabetes 2010;59:2625–30.
9. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lackin JM, Walker EA, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002;346:393–403.
10. Morrison MK, Lowe JM, Collins CE. Perceived risk of type 2 diabetes in Australian women with a recent history of gestational diabetes mellitus. Diabet Med 2010;27:882–6.
11. Ford ES, Li C, Sattar N. Metabolic syndrome and incident diabetes: current state of the evidence. Diabetes Care 2008;31:1898–904.
12. Retnakaran R, Qi Y, Connelly PW, Sermer M, Zinman B, Hanley AJ. Glucose intolerance in pregnancy and postpartum risk of metabolic syndrome in young women. J Clin Endocrinol Metab 2010;95:670–7.
13. Retnakaran R, Qi Y, Connelly PW, Sermer M, Hanley AJ, Zinman B. Risk of early progression to prediabetes or diabetes in women with recent gestational dysglycemia but normal glucose tolerance at 3-month postpartum. Clin Endocrinol (Oxf) 2010;73:476–83.
14. Winzer C, Wagner O, Festa A, Schneider B, Roden M, Bancher-Todesca D, et al. Plasma adiponectin, insulin sensitivity, and subclinical inflammation in women with prior gestational diabetes mellitus. Diabetes Care 2004;27:1721–7.
15. Kautzky-Willer A, Krssak M, Winzer C, Pacini G, Tura A, Farhan S, et al. Increased intramyocellular lipid concentration identifies impaired glucose metabolism in women with previous gestational diabetes. Diabetes 2003;52:244–51.
16. Katz A, Nambi SS, Mather K, Baron AD, Follmann DA, Sullivan G, et al. Quantitative insulin sensitivity check index: a simple, accurate method for assessing insulin sensitivity in humans. J Clin Endocrinol Metab 2000;85:2402–10.
17. International Diabetes Federation. The IDF consensus worldwide definition of the metabolic syndrome. 2005. Available at: www.idf.org
. Retrieved October 17, 2010.
18. Kautzky-Willer A, Handisurya A. Metabolic diseases and associated complications: sex and gender matter! Eur J Clin Invest 2009;39:631–48.
19. Grundy SM, Brewer HB, Cleeman JI, Smith SC, Lefant C. Definition of metabolic syndrome: Report of the National Heart, Lung, and Blood Institute/American Heart Association conference on scientific issues related to definition. Circulation 2004;109:433–8.
20. Goeman J. L1 and L2 penalized regression models. 2010. Available at: http://cran.r-project.org/web/packages/penalized/vignettes/penalized.pdf
. Retrieved March 28, 2010.
21. American Diabetes Association. Standards of medical care in diabetes—2010. Diabetes Care 2010;33(suppl 1):S11–61.
22. Ratner RE, Christophi CA, Metzger BE, Dabelea D, Bennett PH, Pi-Sunyer X, et al. Prevention of diabetes in women with a history of gestational diabetes: effects of metformin and lifestyle interventions. J Clin Endocrinol Metab 2008;93:4774–9.
23. Cameron AJ, Zimmet PZ, Shaw JE, Alberti KG. The metabolic syndrome: in need of a global mission statement. Diabet Med 2009;26:306–9.
24. Buchanan TA, Xiang AH, Peters RK, Kjos SL, Marroquin A, Goico J, et al. Preservation of pancreatic beta-cell function and prevention of type 2 diabetes by pharmacological treatment of insulin resistance in high-risk Hispanic women. Diabetes 2002;51:2796–803.
25. Xiang AH, Peters PK, Kjos SL, Marroquin A, Goico J, Ochoa C, et al. Effect of pioglitazone on pancreatic beta-cell function and diabetes risk in Hispanic women with prior gestational diabetes. Diabetes 2006;55:517–22.
26. Bentley-Lewis R, Levkoff S, Stuebe A, Seely EW. Gestational diabetes mellitus: postpartum opportunities for the diagnosis and prevention of type 2 diabetes mellitus. Nat Clin Pract Endocrinol Metab 2008;4:552–8.
27. Schaefer-Graf UM, Graf K, Kulbacka I, Kjos SL, Dudenhausen J, Vetter K, et al. Maternal lipids as strong determinants of fetal environment and growth in pregnancies with gestational diabetes mellitus. Diabetes Care 2008;31:1858–63.
28. Göbl CS, Handisurya A, Klein K, Bozkurt L, Luger A, Bancher-Todesca D, et al. Changes in serum lipid levels during pregnancy in type 1 and type 2 diabetic participants. Diabetes Care 2010;33:2071–3.