Prediction of women's long-term cardiometabolic risks using glycemic indices during pregnancy

Authors


Dr Wing-Hung Tam, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, 1st floor, E Block, The Prince of Wales Hospital, Shatin, New Territories, Hong Kong, China. Email: tamwh@cuhk.edu.hk

Abstract

Aims:  To study the prediction of abnormal glucose tolerance (AGT), diabetes mellitus (DM), hypertension (HT) and metabolic syndrome (MetS) among Chinese women using glycemic indices in the mid-trimester of pregnancy.

Methods:  A cohort of Chinese women who had had either normal glucose tolerance or gestational diabetes mellitus (GDM) during a pregnancy were assessed at a median of 8 and 15 years post-delivery. All women underwent a 50-g glucose challenge test (GCT) and a 75-g oral glucose tolerance test in the mid-trimester of the index pregnancy. A receiver operating characteristic curve was used to assess the prediction of AGT, DM, HT and MetS.

Results:  All glycemic indices were significant predictors of AGT and DM, and the 2-h plasma glucose (PG) and GCT were predictive of HT, at both 8 and 15 years post-delivery. MetS can only be predicted by the fasting plasma glucose (FPG) and was confined to 15 years post-delivery. After adjustment for confounding variables, all glycemic indices were still independent predictors of AGT and DM at both 8 and 15 years post-delivery, except for FPG in predicting DM at 8 years, while only the 2-h PG remains an independent predictor of HT at 15 years. The optimal cut-off values for FPG, 2-h PG and GCT are 4.2 mmol/L, 7.2 mmol/L and 7.7 mmol/L, respectively; all are lower than the current cut-off thresholds for the screening and diagnosis of GDM.

Conclusions:  Women who had a glycemic level below the criteria for a positive screening test and below the diagnostic threshold for GDM still have a significant cardiometabolic risk.

Introduction

Despite a decline in the overall diabetes mellitus (DM)-related deaths in the last 30 years, the mortality rates remain static among women with DM.1 From a woman's health perspective, it would be most cost-effective to identify women at high risk of developing DM and intervene early. It is now well established that a prior history of gestational diabetes mellitus (GDM) conferred a six to nine-fold increased risk of future DM, resulting in a cumulative incidence of 70% at 28 years postpartum.2–4 Furthermore, women with GDM during pregnancy were also at greater cardiovascular risks for conditions such as hypertension (HT), obesity and dyslipidemia, when compared to age-matched controls.5

In a cohort of Chinese women who were diagnosed with GDM using the universal oral glucose tolerance test (OGTT) screening at mid-gestation, their GDM status increased the odds of their future progression to DM and HT by 8.0 and 3.3-fold at 15 years postpartum, respectively.6 In the same cohort, type 2 DM increased from 9% at 8 years to 24.4% at 15 years post-delivery among women with a history of GDM.6,7

There is substantial evidence that the criteria for the diagnosis of GDM are arbitrary. Historically, this was predominantly based on the prediction of DM postpartum in women with relative hyperglycemia during pregnancy.8,9 The very first O'Sullivan criterion was established arbitrarily using glucose levels at two standard deviations above the mean threshold in a pregnant population.8 Neither the criterion modified from O'Sullivan nor the World Health Organization (WHO) criterion was based specifically on perinatal outcomes. It has long been proposed that the relationship between maternal hyperglycemia during pregnancy and an adverse pregnancy outcome may be linear or curvilinear without a threshold value.10–14 These observations were subsequently confirmed by the results of the ‘Hyperglycemia and Adverse Pregnancy Outcome (HAPO)’ study, which demonstrated that a graded association existed between a less severe degree of maternal hyperglycemia below the current standard diagnostic criteria and a number of adverse pregnancy outcomes.15 On the basis of the results of the HAPO study, the International Association of Diabetes and Pregnancy Study Groups Consensus Panel (IADPSG) has finally proposed a new diagnostic criterion in which the threshold of fasting plasma glucose (FPG) was lowered to 5.1 mmol/L.16 Although the glucose challenge test (GCT) was not addressed in the HAPO study, children born to mothers without GDM, but with an abnormal GCT, also had a gradual increase in the rate of macrosomia, large-for-gestational age and cesarean section, in relation to categories of increasing GCT severity.17,18

In a similar analogy, whether the glucose levels of GCT and OGTT have the same linear relationship with the mother's long-term risk or whether there should be an optimal threshold in the prediction of the future risk of DM and other cardiometabolic risks remain to be determined. The objective of the present study is to investigate whether plasma glucose levels at GCT and OGTT can predict a woman's later cardiometabolic risks and whether there is an optimal threshold that is different from that of the current screening and diagnostic criteria.

Methods

Eligible subjects were from a cohort of 1031 Chinese women who had participated in a study that aimed at defining the optimal screening method and diagnostic criteria for GDM in Chinese women.19 In the original study, which took place between 1992 and 1994, all Chinese mothers who had no previous history of DM had blood taken for a 50-gram GCT (between 24 and 28 weeks' gestation), which was then followed by a 75-gram OGTT administered two to four weeks later, regardless of whether the glucose level at GCT was above the cut-off of 7.8 mmol/L or not.19 Among the 942 women who had completed both the GCT and the OGTT, all 134 who were diagnosed with gestational diabetes (either impaired glucose tolerance [IGT][n = 127] or GDM [n = 7] according to the 1999 WHO criteria), as well as 268 who had normal glucose tolerance (NGT) matched for maternal age, were invited for a follow-up study at a median of 8 and 15 years postpartum for the assessment of DM and other metabolic risks such as HT and hyperlipidemia.6,7

Women who consented to the study underwent an OGTT after an overnight fast of ≥8 h at both 8 and 15 years after the index pregnancy, except that a FPG was performed with a measurement of the glycosylated hemoglobin (HbA1c) in case the subject had been diagnosed with DM and had been prescribed a pharmacological treatment. The plasma glucose (PG) was measured using the hexokinase method (DP Modular Analytics; Roche Diagnostics, Indianapolis, IN, USA). The latest American Diabetes Association diagnostic criteria were used to define glycemic status; that is, DM was defined as a FPG level ≥7.0 mmol/L or a 2-h PG level ≥11.1 mmol/L; IGT was defined as a FPG level <7.0 mmol/L and a 2-h PG level ≥7.8 and <11.1 mmol/L; and impaired fasting glucose (IFG) as a FPG level ≥5.6 mmol/L and <7.0 mmol/L. Abnormal glucose tolerance (AGT) was defined as IFG, IGT or DM. Body weight, height, hip and waist circumferences were measured in light clothing. Blood pressure (BP) was measured in the non-dominant arm using an automated vital signs monitor (Model 53000; Welch Allyn, Beaverton, OR, USA) with cuffs of the appropriate bladder size (which covered at least two-thirds of the arm circumference). Measurements were performed thrice, at 1-min intervals, after ≥5 min of rest. The mean BP reading was used for analysis. Women with a known diagnosis of HT and who were taking medication, or with a mean BP reading ≥140/90 mmHg, were defined as having HT. All other significant medical conditions were recorded. All subjects diagnosed with HT or DM at the 8 year follow-up were referred to the outpatient specialist clinic for further management. Women diagnosed with IGT or IFG were provided with health education and advised on diet and lifestyle modification, as well as encouraged to seek regular medical follow-up.

The diagnosis of metabolic syndrome (MetS) was based on the International Diabetes Federation definition,20 modified with the latest American Diabetes Association criteria for IFG21 and the Asian-specific definition of central obesity, defined as three or more of the following risk factors: (i) waist circumference ≥80 cm; (ii) FPG ≥5.6 mmol/L; (iii) systolic BP ≥130 mmHg or diastolic BP ≥85 mmHg; (iv) fasting plasma triglyceride ≥1.7 mmol/L; and (v) high density lipoprotein cholesterol (HDL-C) <1.3 mmol/L. The studies were approved by the Chinese University of Hong Kong Clinical Research Ethics Committee (CRE-2000.354; CRE-2008.265). Written informed consent was obtained from all participants.

Statistical analyses

Data are expressed as the mean ± SD or proportion for between-group comparisons. Between-group differences were compared using the Student's t and χ2 tests for continuous and categorical variables, respectively, using the SPSS 16.0 (SPSS, Chicago, IL, USA) program. Receiver operating characteristic (ROC) analysis was used to determine the best predictor of progression to AGT, DM, HT and MetS among different glycemic variables. Areas under the ROC curve (AUC-ROC) calculated from these glycemic variables were compared by the Delong paired test using Analyse-it, version 2.09 (Analyse-it Software, Leeds, UK). A P-value <0.05 (2-tailed) was considered to be significant.

Results

A total of 203 women (136 with NGT and 67 with gestational diabetes) completed the follow-up assessment at 8 years postpartum; the conversion rates to AGT and DM were 25.1% and 4.4%, respectively. At 15 years, 139 women (94 with NGT and 45 with gestational diabetes) completed the assessment with a drop-out rate of 31.5%. Among them, 30.2% progressed to AGT and 13% were diagnosed with type 2 DM. Table 1 summarizes the demographic characteristics and cardiometabolic status between women with NGT and gestational diabetes at the index pregnancy, 8-year and 15-year follow-up assessments. The response rates, the baseline parameters between non-responders and responders, and the maternal metabolic status at both follow-up assessments between women with NGT and gestational diabetes have been discussed in previous publications.6,7 The glycemic indices FPG, 2-h PG and GCT at the mid-gestation of the index pregnancy were normally distributed. Tables 2 and 3 show the AUC-ROC of all the glycemic indices in the prediction of women's risk of AGT, DM, HT and MetS at 8 and 15 years, respectively. All glycemic indices were predictive of AGT, DM and HT, but only the 2-h PG and GCT were predictive of HT at both 8 and 15 years. However, MetS can only be predicted by FPG and was confined to 15 years. The GCT gave a good prediction of the mothers' future risk of DM at both 8 and 15 years (AUC-ROC = 0.81 and = 0.84, respectively). Nevertheless, only the FPG gives a moderate prediction of MetS at 15 years (AUC-ROC 0.65; P = 0.022). There is no significant difference in the performance among the three glycemic indices in the prediction of AGT, DM and HT using the Delong paired test.

Table 1. Demographic characteristics and cardiometabolic status between women with normal glucose tolerance and gestational diabetes at the index pregnancy, at 8-year and 15-year follow-up assessments
 Normal glucose toleranceGestational diabetes P-value
  1. Data was expressed in either mean ± SD, or number (%). BMI, body mass index; DM, diabetes mellitus; GCT, glucose challenge test; IFG, impaired fasting glucose; IGT, impaired glucose tolerance; HDL-C, high-density lipoprotein cholesterol; HOMA-BCF, homeostasis model assessment of β-cell function; HOMA-IR, homeostasis model assessment insulin resistance index; LDL-C, low-density lipoprotein cholesterol; Matsuda-ISI, Matsuda insulin sensitivity index; NGT, normal glucose tolerance; OGTT, oral glucose tolerance test; QUICKI, quantitative insulin sensitivity check index.

At the index pregnancy:(n = 136)(n = 67) 
 Age (years)27.4 ± 4.328.6 ± 4.30.07
 BMI (kg/m2)24.4 ± 2.724.8 ± 3.60.20
 Family history of DM (%)18 (13.2)13 (19.4)0.25
 GCT (mmol/L)6.5 ± 1.57.9 ± 1.8<0.001
 75-gram OGTT, fasting glucose (mmol/L)4.0 ± 0.64.5 ± 0.6<0.001
 75-gram OGTT, 2-h glucose (mmol/L)5.8 ± 1.08.8 ± 1.3<0.001
 Neonatal birth weight (g)3272 ± 4293230 ± 4850.26
 Gestational age at delivery (weeks)39.5 ± 1.639.3 ± 2.10.54
 Umbilical cord C-peptide level (nmol/mL)0.56 (0.47)0.70 (0.42)0.12
 Umbilical cord insulin level (pmol/L)48.7 (45.8)62.9 (53.8)0.12
At the 8-year follow-up assessment:(n = 136)(n = 67) 
 Age (years)36.3 ± 4.736.9 ± 4.40.40
 BMI (kg/m2)23.7 ± 3.524.4 ± 4.60.24
 Family history of DM35 (25.7)28 (41.8)0.02
 Family history of hypertension62 (45.6)44 (65.7)0.007
 Exercise levelsSedentary98 (73.1)53 (77.9)0.12
Occasional exercise7 (5.2)7 (10.3)
Regular exercise29 (21.6)8 (11.8)
 Smoker12 (9.0)4 (6.2)0.50
 Hypertension9 (6.6)13 (19.1)0.007
 Glycemic statusNGT112 (82.3)40 (59.7)0.001
IFG and/or IGT21 (15.4)21 (31.3)
DM3 (2.2)6 (9.0)
 Insulin sensitivity indicesHOMA-IR1.93 ± 2.112.82 ± 3.870.08
Matsuda-ISI6.83 ± 4.175.30 ± 3.530.01
QUICKI0.37 ± 0.050.36 ± 0.470.08
 β-cell functionHOMA-BCF140 ± 130207 ± 4430.23
 Plasma HDL-C level (mmol/L)1.64 ± 0.361.43 ± 0.29<0.001
 Plasma LDL-C level (mmol/L)2.74 ± 0.782.75 ± 0.640.94
 Plasma triglyceride level (mmol/L)0.96 ± 0.491.17 ± 1.160.08
 Metabolic syndrome11 (8.1)5 (7.5)0.85
At the 15-year follow-up assessment:(n = 94)(n = 45) 
 Age (years)43.2 ± 4.643.8 ± 4.30.49
 BMI (kg/m2)24.4 ± 3.524.7 ± 4.50.70
 Smoker9 (9.3)2 (4.8)0.53
 Hypertension15 (16.0)16 (35.6)0.01
 Glycemic statusNGT75 (79.8)22 (48.9)<0.001
IFG and/or IGT14 (14.9)12 (26.6)
DM5 (5.3)11 (24.4)
 Plasma HDL-C level (mmol/L)1.59 ± 0.351.49 ± 0.330.12
 Plasma LDL-C level (mmol/L)2.68 ± 0.802.72 ± 0.610.75
 Plasma triglyceride level (mmol/L)1.06 ± 0.571.33 ± 0.800.03
 Metabolic syndrome14 (14.9)10 (22.2)0.41
Table 2. Prediction of women's risk of abnormal glucose tolerance (AGT), diabetes mellitus (DM), hypertension and metabolic syndrome (MetS) 8 years after delivery, using the glycemic indices obtained during pregnancy
Glycemic indicesMean (SD) (mmol/L)AGTDMHypertensionMetS
AUC-ROC curves (95% CI) P-valueAUC-ROC curves (95% CI) P-valueAUC-ROC curves (95% CI) P-valueAUC-ROC curves (95% CI) P-value
  1. 2-h PG, plasma glucose 2-h post-OGTT; AUC, area under curve; CI, confidence interval; FPG, fasting plasma glucose; GCT, 50-g glucose challenge test; ROC, receiver operating characteristic analysis; SD, standard deviation.

FPG4.1 (0.5)0.62 (0.53–0.71)0.010.75 (0.60–0.89)0.010.58 (0.46–0.71)0.190.56 (0.43–0.69)0.43
2-h PG6.8 (1.8)0.69 (0.60–0.77)<0.0010.70 (0.49–0.91)0.040.64 (0.52–0.76)0.030.50 (0.38–0.63)0.97
GCT6.9 (1.7)0.72 (0.63–0.80)<0.0010.81 (0.67–0.95)0.0020.65 (0.52–0.76)0.020.53 (0.41–0.66)0.66
Table 3. Prediction of women's risk on abnormal glucose tolerance (AGT), diabetes mellitus (DM), hypertension and metabolic syndrome (MetS) 15 years after delivery, using the glycemic indices obtained during pregnancy
Glycemic indicesMean (SD) mmol/LAGTDMHypertensionMetS
AUC of ROC curves (95% CI) P-valueAUC of ROC curves (95% CI) P-valueAUC of ROC curves (95% CI) P-valueAUC of ROC curves (95% CI) P-value
  1. 2-h PG, plasma glucose 2-h post-OGTT; AUC, area under curve; CI, confidence interval; FPG, fasting plasma glucose; GCT, 50-g glucose challenge test; ROC, receiver operating characteristic analysis; SD, standard deviation.

FPG4.1 (0.6)0.70 (0.61–0.81)<0.0010.74 (0.61–0.86)0.0020.61 (0.49–0.73)0.0530.65 (0.54–0.76)0.02
2-h PG6.8 (1.9)0.72 (0.63–0.81)<0.0010.76 (0.66–0.87)0.0010.70 (0.61–0.80)0.0010.62 (0.51–0.73)0.06
GCT6.8 (1.8)0.75 (0.66–0.84)<0.0010.84 (0.76–0.92)<0.0010.67 (0.58–0.77)0.0030.63 (0.51–0.74)0.05

Optimal cut-offs of individual glycemic indices were selected by using the highest Youden indices (i.e. sensitivity + specificity − 1) for the prediction of DM at 15 years, as shown in Figure 1. Table 4 shows the relative risks and absolute risks of AGT, DM and HT at each of the glycemic thresholds. Mothers who had a GCT ≥7.7 mmol/L had a relative risk of 15 and 14 in the conversion to type 2 DM at 8 and 15 years after pregnancy respectively. Likewise, a 2-h PG result in the OGTT ≥7.2 mmol/L gave relative risks of 5.3 and 4.9 at 8 and 15 years, respectively. Both thresholds are lower than the current criteria for a positive test. After adjustment for maternal age, the body mass index at the time of booking, a family history of DM, gestational HT, pre-eclampsia during the index pregnancy and the number of subsequent deliveries, most of the glycemic indices were not predictive of HT, except 2-h PG for the prediction of HT at 15 years (Table 4).

Figure 1.

ROC curve for the prediction of diabetes mellitus (DM) at 15 years after the index pregnancy; ○, glucose challenge test (GCT); □, fasting glucose (FG); ▵, 2-h plasma glucose (PG).

Table 4. Relative and absolute risks of abnormal glucose tolerance (AGT), diabetes mellitus (DM), hypertension at 8 and 15 years post-delivery using various glycemic indices at the mid-gestation of the pregnancy
Cut-off of glycemic indicesAGTDMHypertension
At 8 yearsAt 15 yearsAt 8 yearsAt 15 yearsAt 8 yearsAt 15 years
  1. †Adjustment for maternal age, BMI at booking, family history of DM during the index pregnancy and the number of subsequent term pregnancies. ‡Adjustment for maternal age, BMI at booking, family history of DM, gestational hypertension, pre-eclampsia during the index pregnancy and the number of subsequent term pregnancies. 2-h PG, second hour OGTT plasma glucose; CI, confidence interval; FPG, fasting plasma glucose; GCT, 50-g glucose challenge test; RR, relative risk.

FPG ≥4.2 mmol/LUnadjusted RR (95% CI)1.7 (1.1–2.7)2.0 (1.2–3.3)4.5 (0.95–21)4.1 (1.4–12)1.3 (0.5–2.8)1.9 (1.0–3.5)
Adjusted RR (95% CI)2.0 (1.0–4.0)3.2 (1.4–7.1)5.0 (0.97–25)5.5 (1.6–20)1.3 (0.5–3.5)2.2 (0.9–5.6)
Absolute risk (%)32.642.47.920.312.230.5
GCT ≥7.7 mmol/LUnadjusted RR (95% CI)3.2 (2.0–5.2)3.1 (1.8–5.2)15 (1.9–118)14 (3.3–58)2.7 (1.2–6.0)1.9 (1.0–3.4)
Adjusted RR (95% CI)4.2 (2.0–8.6)5.4 (2.3–13)16 (1.9–138)16 (3.2–82)1.5 (0.5–4.3)1.4 (0.6–3.5)
Absolute risk (%)46.454.311.630.418.632.6
2-h PG ≥7.2 mmol/LUnadjusted RR (95% CI)2.2 (1.3–3.5)2.5 (1.5–4.0)5.3 (1.1–25)4.9 (1.7–14)2.6 (1.1–5.9)3.0 (1.5–5.7)
Adjusted RR (95% CI)2.6 (1.3–5.2)5.6 (2.4–13)6.1 (1.2–32)5.9 (1.7–21)2.0 (0.7–5.5)4.2 (1.6–11)
Absolute risk (%)37.050.08.622.617.137.7

Discussion

Our study results have confirmed that a threshold of GCT and OGTT glucose levels lower than the current screening and diagnostic criteria, respectively, were predictive of a woman's progression to AGT, DM and HT at both 8 and 15 years post-delivery. Among the various glycemic indices studied during the index pregnancy, the GCT appears to be a good predictor of a woman's progression to DM with a sensitivity and specificity of 87.5% and 73.6%, respectively, at its optimal level. Moreover, a woman's long-term risk of DM was increased 16-fold at both 8 and 15 years when the GCT was ≥7.7 mmol/L during pregnancy. Similar to our findings, other research has also demonstrated a graded increase in a woman's future risk of DM across all GCT quartiles in those without GDM – an adjusted hazard ratio was 3.6 if the mother's GCT was at the highest quartile (≥7.4 mmol/L) compared to the lowest quartile (≤5.3 mmol/L) at a median follow up of 8.8 years.22 Moreover, the optimal threshold of the 2-h PG for the prediction of DM was also lower than many of the current criteria for the diagnosis of GDM. Studies have shown that 1-h glucose levels of the OGTT were a predictor of immediate postpartum insulin resistance and type 2 DM at 5 years after pregnancy;23,24 however, the 1-h PG was not available at the OGTT in our cohort for such analysis.

Despite the fact that various antenatal factors, such as the pre-pregnant body mass index, gestational weeks at diagnosis, persistent fasting and postprandial hyperglycemia requiring insulin therapy, had been found to be significant predictors of future DM,24–26 no previous studies have addressed the optimal threshold of GCT and OGTT in the prediction of future DM and other cardiometabolic risk. Since large epidemiological studies have demonstrated a high prevalence of DM in young Chinese people, as well as an increase in undiagnosed AGT and diabetes,27,28 the current results will provide some useful information on how the various glycemic indices at mid-gestation in a cohort of relatively young women (a median age of 28 years old at the time of pregnancy) may be of help in the long-term prediction of both pre-diabetes and type 2 DM.7

Similarly, where our earlier results failed to show an increased long-term risk of MetS in women with a history of GDM during pregnancy,6,7 the glycemic indices during pregnancy were not found to be predictive of MetS at 8 and 15 years in the present study. This is contrary to several follow-up studies of women with a prior history of GDM who had demonstrated an increased risk of MetS, which in turn is associated with cardiovascular risk;29–34 however, the negative prediction in the present study may be related to the small sample size.

Finally, the main limitation of the current study is the small sample size, as mentioned, and the modest attrition rate at the second follow-up session. Moreover, our results may not be generalizable to other ethnic groups. Last, but not the least, the current study did not address the prediction with adjustment of various confounding factors, such as family history, dietary habit and exercise levels from pregnancy to the follow-up assessment. Our findings could only suggest that the current diagnostic criteria for GDM might not be discriminative in predicting women at risk of pre-diabetes and DM in future. Therefore, further large, multi-ethnic studies would be of interest to confirm this observation and to explore the optimal thresholds. While the result of the HAPO study has lead to a consensus on the diagnostic criteria of GDM on the basis of adverse pregnancy outcomes,16 long-term follow-up of the HAPO cohort would provide useful information for establishing a threshold to predict mothers' future cardiometabolic risk.

Acknowledgments

We would like to thank all the mothers and children who participated in the study, as well as our research staff (Connie Yuen, Lindy Chan, Cherry Chiu and Josephine Chan) for their dedication and professionalism in conducting the survey.

Disclosure

No relevant conflicts of interest need to be disclosed for any of the authors.

Ancillary