Insulin Resistance and Impaired Pancreatic β-Cell Function in Adult Offspring of Women With Diabetes in Pregnancy

Abstract

Context:

Offspring of women with diabetes during pregnancy have an increased risk of glucose intolerance in adulthood, but the underlying mechanisms are unknown.

Objective:

We aimed to investigate the effects of intrauterine hyperglycemia on insulin secretion and action in adult offspring of mothers with diabetes.

Design, Setting, and Participants:

A cohort of 587 Caucasian offspring, without known diabetes, was followed up at the age of 18–27 years. We included 2 groups exposed to maternal diabetes in utero: offspring of women with gestational diabetes mellitus (n = 167) or type 1 diabetes (n = 153). Two reference groups were included: offspring of women with risk factors for gestational diabetes mellitus but normoglycemia during pregnancy (n = 139) and offspring from the background population (n = 128).

Main Outcome Measures:

Indices of insulin sensitivity and insulin release were calculated using insulin and glucose values from a standard oral glucose tolerance test (120 minutes, 75 g glucose). Pancreatic β-cell function taking the prevailing insulin sensitivity into account was estimated by disposition indices.

Results:

Both groups of offspring exposed during pregnancy to either maternal gestational diabetes or type 1 diabetes had reduced insulin sensitivity compared with offspring from the background population (both P < .005). We did not find any significant difference in absolute measures of insulin release. However, the disposition index was significantly reduced in both the diabetes-exposed groups (both P < .005).

Conclusion:

Reduced insulin sensitivity as well as impaired pancreatic β-cell function may contribute to the increased risk of glucose intolerance among adult offspring born to women with diabetes during pregnancy.

There is good evidence from both animal studies and human epidemiological studies that maternal hyperglycemia during pregnancy induces lasting changes in offspring metabolism (1–3). We have previously shown that exposure to intrauterine hyperglycemia was associated with an increased risk of overweight and the metabolic syndrome as well as glucose intolerance and overt type 2 diabetes in young adult Caucasian offspring of women with either gestational diabetes mellitus (GDM) or type 1 diabetes during pregnancy (4, 5). The underlying causes are unclear, but a fuel-mediated mechanism induced by a hyperglycemic intrauterine environment has been hypothesized (6). This is supported by human studies, in which the confounding effect of genetic predisposition to type 2 diabetes was sought eliminated in the study design (7–9). Women with GDM and consequently their offspring have a high genetic predisposition to type 2 diabetes (10), whereas this is not believed to be the case for women with type 1 diabetes and their offspring (8).

The relative contribution of impaired insulin sensitivity and reduced insulin release in the pathogenesis of type 2 diabetes has been intensively debated (11–13). However, there is now general consensus that type 2 diabetes is a disease of multiple organs characterized by both insulin resistance and impaired pancreatic β-cell function. Whether the increased prevalence of prediabetes and diabetes in offspring exposed to intrauterine hyperglycemia is mainly related to reduced insulin sensitivity, impaired insulin production, or both is sparsely described. Studies are conflicting and mostly performed in cohorts of young children (14–19), whereas the topic has not previously been evaluated among adult offspring of women with GDM.

We aimed to evaluate insulin sensitivity and insulin release in adult Caucasian offspring of women with either diet-treated GDM or type 1 diabetes during pregnancy compared with unexposed offspring.

Materials and Methods

Study design (Figure 1)

Figure 1.

Study design: flow chart of subjects participating in the study (n = 587) and subjects lost to follow-up (n = 469) or excluded due to previously known type 1 or type 2 diabetes (n = 10). O-GDM included offspring of mothers with gestational diabetes (GDM); O-Type1 included offspring of mothers with type 1 diabetes; O-NoGDM included offspring of mothers with risk factors for GDM, but normoglycemic in pregnancy; O-BP included offspring of mothers from the background population without risk factors for diabetes.

We did a follow-up study of a cohort including 1066 individuals, born between 1978 and 1985 at Rigshospitalet, Copenhagen, Denmark. Offspring from all singleton pregnancies complicated by either diet-treated GDM or type 1 diabetes were invited, and an equal number of control subjects, randomly selected among healthy women giving birth in the same time period were included. From the mothers' medical records, we extracted maternal baseline data including pregestational weight, blood glucose levels during pregnancy, the mode of antidiabetic treatment, and other relevant data from the pregnancy and delivery. Coupling between the mothers' medical record and the adult offspring was possible through the Danish Civil Registrar System.

The study was in accordance with the Declaration of Helsinki and approved by the regional ethical committee. All participants gave written consent before inclusion in the study.

Participants

Subjects were classified into four groups, depending on exposure to intrauterine hyperglycemia and estimated genetic predisposition to type 2 diabetes and overweight (Figure 1).

We invited two groups of offspring, who had been exposed to intrauterine hyperglycemia but with different genetic predisposition to type 2 diabetes and overweight. These two groups included the following: 1) offspring of women with diet-treated GDM (O-GDM), who were estimated to have a high genetic predisposition to type 2 diabetes and overweight; and 2) offspring of women with type 1 diabetes (O-Type1), in which the genetic predisposition to type 2 diabetes and overweight was assumed to be low and comparable with the background population.

Likewise, two unexposed reference groups with a different genetic predisposition to type 2 diabetes and overweight were also invited. These groups included the following: 1) offspring of women who had undergone an oral glucose tolerance test (OGTT) during pregnancy due to the risk factors for developing GDM but who had presented a normal glucose tolerance (O-NoGDM) (this offspring group was estimated to have a high genetic predisposition to type 2 diabetes and overweight, comparable with the offspring of women who developed GDM); and 2) offspring from the background population not exposed to intrauterine hyperglycemia (O-BP). The genetic predisposition to type 2 diabetes and overweight was assumed to be low because the Danish population in general is a low-risk population).

We hypothesized that a possible effect of intrauterine hyperglycemia could be evaluated through comparing the 2 diabetes-exposed groups (O-GDM and O-Type1) with the offspring of women from the background population (O-BP) as well as comparing O-GDM with O-NoGDM. Finally, a comparison between the 2 unexposed reference groups (O-NoGDM and O-BP) reflects the possible effects of genetic predisposition to type 2 diabetes.

Exclusion criteria in the offspring at follow-up were previously known type 1 diabetes or type 2 diabetes.

Maternal screening procedure and care

In Denmark the prevalence of GDM is 2%–3%, and screening for GDM has traditionally been based on risk factors (20–22). During 1978–1985 the following risk factors were used: pregnancy overweight of 20% or greater, a family history of diabetes, previous GDM, previous delivery of child weighing greater than 4500 g, and glucosuria (20, 23). Women with risk indicators and 2 consecutive fasting capillary blood glucose measurements of 4.1 mmol or greater were offered a 3-hour, 50-g OGTT. The OGTT was defined as abnormal if more than 2 of 7 values during the test exceeded the mean +3 SD for a reference group of normal-weight nonpregnant women without a family history of diabetes (22, 24).

All mothers with GDM in this study were diet treated only. Offspring of mothers with insulin-treated GDM were excluded for participation to minimize the risk of misclassification among the mothers in the GDM group (eg, undiagnosed type 2 diabetes, early stage type 1 diabetes, and maturity-onset diabetes of the young [MODY]).

Mothers with type 1 diabetes fulfilled 3 criteria: 1) onset of diabetes at age 40 years or younger, 2) a classical anamnesis, and 3) insulin treatment initiated 6 months or sooner after diagnosis.

Mothers of O-NoGDM had risk factors qualifying for an OGTT, but the glucose values during the OGTT were below the mean +2 SD of the reference group.

Mothers of O-BP were from the local community routinely referred for antenatal care and delivery at Rigshospitalet (Copenhagen, Denmark).

Examination of offspring at follow-up

After an overnight fast, participants underwent a 120-minute, 75-g OGTT, with venous sampling at 0, 30, and 120 minutes. Weight (kilograms) and height (meters) were measured and a questionnaire including information on occupation, health, medication, smoking habits, diet, and levels of physical activity status was completed.

Biochemical analyses

Blood samples for glucose measurements were drawn in heparin-sodium fluoride vials, kept on ice, centrifuged, plasma separated within 30 minutes, and analyzed on a Cobas Mira analyzer by either the enzymatic UV test, HK/G-6PHD method (ABX Diagnostics Glucose HK 125; Horiba-ABX) or the glucose dehydrogenase catalyzed oxidation method (Gluc-DH method; Merck). The interassay coefficient of variation for both analyses was less than 4%. Serum C-peptide and insulin were measured automatically by a fluoroimmunoassay using monoclonal antibodies (AutoDELFIA C-peptide kit and AutoDELFIA insulin kit; PerkinElmer-Wallac Life and Analytical Sciences). The interassay coefficient of variation for both C-peptide and insulin was less than 6%.

Hemolysed blood samples were excluded in further calculations in insulin or C-peptide levels. Estimation of hemolysis was based on clinical assessment as well as on calculations on the insulin to C-peptide ratio being less than 0.03.

Primary outcome variables (Figure 2)

Figure 2.

Calculations of estimates.

Among several possible indices, we a priori chose to evaluate insulin sensitivity, insulin response, and the disposition index (DI) based on the β-cell function insulin sensitivity glucose tolerance test (BIGTT) method (25) as primary outcomes. BIGTT-derived estimates are based on data from an OGTT, which is easy to perform and less invasive than an iv glucose tolerance test (IVGTT). The BIGTT method was developed by 2 of the authors of this paper (T.H. and O.P.) and has been validated against the Bergman's minimal model and found to produce sound estimates of insulin sensitivity and insulin response with strong correlations to the IVGTT-based estimates (25). Moreover, body mass index (BMI) and sex are taken into account by the BIGTT method. Estimates based on other well-described methods were chosen as secondary outcomes: the Matsuda index (26), homeostatic model assessment of insulin resistance (HOMA-IR) (27) (both validated against the euglycemic clamp), the insulinogenic index (28) (validated against IVGTT), and the corrected insulin response (CIR) (29) as well as the DI based on these indices.

Exposure variables

We used five different estimates of intrauterine hyperglycemia as exposure variables: in analyses based on all participants, the assignment into four groups (O-GDM, O-Type1, O-NoGDM, and O-BP) was used as surrogate measure of different exposures to intrauterine hyperglycemia. In subanalyses of offspring of women who underwent an OGTT during pregnancy, we used blood glucose values at 0 and 120 minutes during this OGTT as exposure variables. In the subanalyses of O-Type1, we used mean maternal blood glucose in the first and third trimesters as exposure variables, as previously described (4).

Covariates

Maternal covariates were: age at delivery (years); pregestational overweight represented by BMI of ≥ 25 kg/m² (yes vs no); parity (≥1 vs 0 previous deliveries); family history of diabetes (unspecified diabetes in a first degree relative, yes vs no); and ethnicity (Nordic Caucasian, if the woman originated from Denmark, Norway, Sweden, or Iceland, yes vs no).

Covariates related to the offspring at time of birth were sex (male vs female), gestational age (days), and birth weight (grams).

Covariates associated with the offspring at follow-up were age (years), level of physical activity (≥30 vs < 30 min daily), and quality of diet, which was divided into a diet rich in vegetables and fruits vs a diet rich in meat and animal fat (healthy vs poor diet). Information on the participant's diet was based on a questionnaire, in which meal frequency and quantity as well as dietary composition were reported. Socioeconomic position was based on the highest occupational status of the parents at present and coded into family social class I-V in accordance with the standards of the Danish National Institute of Social Research, similar to the British Registrar General's Classification I-V. We added a social class VI representing people on transfer income, including sickness benefits and disability pension, and dichotomized the variable into family social class (I-II vs III-VI) (30). Glucose tolerance status was evaluated due to World Health Organization criteria of 1999 (31) and overweight represented by BMI 25 kg/m² or greater (yes vs no).

Statistical analyses

Normally distributed continuous data are presented as mean (SD), whereas nonnormally distributed data are presented as median (25th to 75th percentiles). Nonnormally distributed data were log10 transformed before entering statistically significant analysis and following presented as geometric mean and 95% confidence intervals (CI). Differences between groups were analyzed with 1-way ANOVA, independent t test, or χ² test when appropriate. Post hoc tests were corrected for multiple comparisons using the Bonferroni method: P values were multiplied by 4 because we compared O-BP with the other 3 groups and O-GDM with O-NoGDM according to our hypothesis.

To control for effects of confounding, we did multiple linear regression analyses. We tested associations between the five different exposure variables and the primary (BIGGT sensitivity index [S_I]_{/0–30–120}, BIGTT-acute insulin response (AIR)_0–30–120, DI₁) as well as the secondary outcome variables (Matsuda index, HOMA-IR, insulinogenic index, CIR, DI₂, DI₃). Exposure variables were assigned into the four different exposure groups (O-GDM, O-Type1, O-NoGDM, O-BP), the maternal OGTT data at 0 and 120 minutes, or the mean maternal blood glucose in the first and third trimesters from the women with type 1 diabetes.

Based on previous studies and theoretical considerations, we included 9 potential confounders: maternal age at delivery, maternal pregestational overweight, parity, maternal family history of diabetes, maternal ethnicity, offspring age at follow-up, level of physical activity, quality of diet in the adult offspring, and socioeconomic position. Birth weight and gestational age were considered to be possible mediators of the effect of intrauterine hyperglycemia on offspring outcome and was entered one by one in separate regression analyses. The effect of sex and BMI was tested in the analysis including the secondary outcomes because these covariates are included in the primary BIGTT-indices.

All tests were two tailed, and a significance level of P = .05 was chosen. Data were processed using SPSS version 18 (SPSS Inc).

Results

Baseline data and follow-up characteristics

The overall participation rate was 56% (597 of 1066) with no differences between the 4 groups. Ten subjects met the exclusion criteria of either previously known type 2 diabetes (n = 1) or type 1 diabetes (n = 9) at follow-up, giving data on 587 subjects (Figure 1).

Per definition, women with GDM had higher blood glucose levels than women in the NoGDM-group: fasting glucose 5.2 vs 4.7 mmol/L; 120-minute glucose: 7.9 vs 5.2 mmol/L (P < .0001 for both). The women with type 1 diabetes had a mean blood glucose of 8.9 (2.8) mmol/L in the first trimester and 6.8 (1.8) mmol/L in the third trimester. Table 1 presents baseline data on the mothers and the offspring at the time of delivery and on offspring at the time of follow-up, including the results of OGTT at follow-up [previously published (4)].

Table 1.

Baseline Data and Follow-Up Characteristics

Variable	O-GDM	O-Type1	O-NoGDM	O-BP	P Valuea
n	167	153	139	128
Maternal data
Age at delivery, y	29.5 (5.4)b	26.5 (4.2)	28.2 (5.0)	27.6 (4.3)	.001
Pregestational overweight (BMI ≥ 25 kg/m2)	38 (63/167)b,c	6 (9/145)	23 (27/119)b	11 (14/126)	<.0001
Parity ≥ 1 previous deliveries	58 (97/167)	46 (70/153)	53 (74/139)	45 (57/128)	.059
Family history of diabetes	31 (51/167)b	20 (32/153)	35 (45/139)b	16 (20/128)	<.0001
Nordic Caucasian (ethnicity)	91 (152/167)	99 (151/153)b	93 (129/139)	92 (118/128)	.027
Offspring birth data
Male sex	54 (91/167)	45 (69/153)	45 (62/141)	49 (63/128)	.26
Gestational age, dd	273 (269–276)b,c	260 (253–262)b	281 (275–287)	280 (275–286)	<.0001
Birth weight, g	3408 (531)	3282 (760)b	3493 (498)	3474 (481)	.009
Offspring follow-up data
Age, y	21.5 (1.8)b	22.5 (2.2)	21.1 (2.1)b	22.9 (2.2)	<.0001
Physical activity ≥ 30 min/d	56 (93/167)	46 (71/153)	55 (77/139)	50 (64/128)	.30
Healthy diet with a high content of vegetables and fruits	14 (24/167)b	20 (30/153)	13 (18/139)b	27 (34/128)	.015
Family social class group I or II	37 (62/167)b	29 (46/153)b	38 (52/138)	52 (66/128)	.003
Abnormal glucose tolerance (IFG, IGT, or type 2 diabetes)e	21 (35/167)b	11 (17/153)	12 (17/139)	4 (5/128)	<.0001
Overweight (BMI ≥ 25 kg/m2)	40 (66/167)b	40 (61/153)b	30 (40/139)	24 (31/128)	.008

Abbreviations: IFG, impaired fasting glucose; IGT, impaired glucose tolerance. Table 1 presents data on 587 offspring. Ten participants were excluded due to exclusion criteria of previously known type 2 diabetes or type 1 diabetes. Data are mean (SD) or percentage unless otherwise indicated. Values of P < .05 are in bold.

^aOverall analysis of differences between the 4 groups (means, proportions, and medians were performed by ANOVA, a χ² test, and Kruskal-Wallis, respectively).

^bCompared with the O-BP group, P < .05 (post hoc test: independent samples t test, χ² or Mann-Whitney test; P values are multiplied by 4).

^cCompared with the O-NoGDM group, P < .05 (post hoc test: independent samples t test, χ² or Mann-Whitney; P values are multiplied by 4).

^dData are median (25th to 75th percentiles) because data were not normally distributed.

^eOGTTs were evaluated according to World Health Organization criteria of 1999.

Estimates of insulin sensitivity

O-GDM had a significantly reduced insulin sensitivity compared with O-BP (Table 2). This was the case for the primary outcome: BIGTT-S_{I/0–30–120} (P < .0001) as well as for the secondary indices: Matsuda index (P < .0001) and HOMA-IR (P = .004). Also, offspring exposed to maternal type 1 diabetes showed significantly reduced insulin sensitivity in comparison with O-BP for the primary outcome BIGTT-S_{I/0–30–120} (P < .0001). The same trend was found for the secondary outcomes: Matsuda index and HOMA-IR, although the difference did not reach statistical significance. O-NoGDM had significantly reduced insulin sensitivity compared with O-BP; it was the case for both the primary (BIGTT-S_{I/0–30–120,} P < .0001) and secondary outcomes (Matsuda index, P < .0001; HOMA-IR, P = .004).

Table 2.

Indicies of Insulin Sensitivity and Insulin Secretion

	O-GDM	O-Type1	O-NoGDM	O-BP	P valuea
n	167	153	139	128	(total 587)
Insulin sensitivity
BIGTT-SI/0–30–120	7.97 (3.57)b	8.27 (3.36)b	8.18 (3.08)b	9.79 (3.39)	<.0001
Matsuda indexc	20.80 (18.98–22.80)b	23.28 (21.37–25.36)	21.73 (19.96–23.66)b	27.00 (24.77–29.45)	<.0001
Insulin resistance
HOMA-IRc	10.53 (9.58–11.57)b	9.28 (8.49–10.15)	10.57 (9.65–11.57)b	8.47 (7.71–9.31)	.002
Insulin secretion
BIGTT-AIR0–30–120c	2060 (1871–2265)	2270 (2096–2459)	2239 (2073–2417)	2177 (1991–2381)	.37
Insulinogenic indexc	86.90 (76.58–96.36)	84.70 (76.44–93.86)	97.19 (85.27–110.76)	90.34 (80.13–101.86)	.35
CIRc	765 (688–850)	831 (749–922)	928 (825–1044)	919 (818–1033)	.04
Disposition index
DI1 (BIGITT-SI/0–30–120 × BIGTT-AIR0–30–120)	16 101 (7694)b	18 148 (7482)	17 867 (6701)b	21 454 (13697)	<.0001
DI2 (CIR × Matsuda index)c	15 743 (13 877–17 861)b,d	20 059 (17 334–21 587)b	19 346 (17 750–22 667)b	24 820 (22 197–27 752)	<.0001
DI3 (CIR/HOMA-IR)c	72.53 (63.50–82.81)b	89.15 (78.92–100.69)	87.74 (77.37–99.52)	108.47 (96.14–122.35)	.028

Includes offspring with normal glucose tolerance, impaired fasting glucose, impaired glucose tolerance and screen detected, and treatment-naive type 2 diabetes. Data are mean (SD) unless otherwise indicated.

^aAnalyses between proportions in the 4 groups were performed by 1-way-ANOVA.

^bCompared with O-BP, P < .05 (post hoc test: independent samples t test and Bonferroni).

^cData are given as geometric mean and CIs because of log transformation to obtain normal distribution.

^dCompared with O-NoGDM, P < .05 (post hoc test: independent samples t test and Bonferroni).

Direct estimates of insulin release

The four studied groups of offspring had comparable values of insulin release for primary and secondary outcomes (Table 2).

Disposition index (estimate of insulin release taking insulin sensitivity into account)

O-GDM had a significantly lower DI based on the BIGTT-indices than O-BP (DI₁: P < .0001) (Table 2). The two other estimates of DI confirmed this: both DI₂ and DI₃: P < .0001. Regarding DI₂, a significant difference between O-GDM and O-NoGDM was also found (P = .028). O-Type1 had lower DI than O-BP, but this was statistically significant only with respect to DI₂ (P = .008). O-NoGDM also had significantly lower DI₁ (P = .03) and DI₂ (P = .048) compared with O-BP.

Multiple linear regression analyses including maternal glucose values during pregnancy

Table 3 gives data from multiple linear regression analyses on the primary outcomes: insulin sensitivity (BIGTT-S_{I/0–30–120}), insulin release (logBIGTT-AIR_0–30–120), and disposition index (DI₁) in O-GDM, O-NoGDM, and O-Type1 using O-BP as the reference. When adjusted for potential confounders O-GDM, O-NoGDM, and O-Type1 had a significantly reduced insulin sensitivity (BIGGT-S_{I-0–30–120}) (all P ≤ .001) and reduced disposition index (DI₁) (O-GDM: P < .0001; O-Type1: P = .004; O-NoGDM: P = .002) compared with O-BP, whereas the absolute insulin response (BIGGT-AIR_0–30–120) did not differ significantly between the four groups.

Table 3.

Multiple Linear Regression Analyses on the Primary Outcomes

Dependent Variable	BIGTT-SI			logBIGTT-AIRa			DI1b
	β	95% CI	P Value	β	95% CI	P Value	β	95% CI	P Value
Exposure variable
O-GDM	−1.56	− 2.38 to − 0.74	<.0001	−0.046	−0.104 to 0.013	.127	−5573.42	− 7943,85 to − 3202.99	<.0001
O-Type1	−1.35	− 2.15 to − 0.55	.001	−0.001	−0.063 to 0.062	.980	−3433.50	− 5738.74 to − 1128.26	.004
O-NoGDM	−1.65	− 2.53 to − 0.78	<.0001	0.007	−0.05 to 0.064	.809	−4098.63	− 6625.06 to − 1572.19	.002

Data given as regression coefficient (β), 95% CIs, and P values. Values of P < 0.05 are bold. Offspring gender and BMI are included in the estimates of the BIGTT indices. The following confounders were included in the models: maternal age at delivery, pregestational overweight, parity, family history of diabetes, Nordic Caucasian ethnicity, offspring age at follow-up, physical activity in the adult offspring, healthy diet, and social-economic position.

^aLog transformed to obtain normal distribution.

^bDI₁: BIGITT-S_{I/0–30–120} × BIGTT-AIR_0–30–120.

Direct measures of the maternal glucose level during pregnancy (blood glucose values at 0 and 120 min from the OGTT during pregnancy in women with GDM or mean blood glucose in first or third trimester in women with maternal type 1 diabetes) was not associated with insulin sensitivity, insulin release, or disposition index in the offspring (data not shown).

The parameter estimates for all three primary outcomes (BIGTT-S_{I/0–30–120}, BIGGT-AIR_0–30–120, and DI₁) did not change when offspring birth weight or gestational age was included as mediators in any of the models (data not shown).

Regarding secondary outcomes of insulin sensitivity (Matsuda and HOMA-IR), insulin release (CIR and insulinogenic index), and disposition index (DI₂ and DI₃), multiple linear regression analyses showed similar results. When offspring sex and overweight were entered in the models regarding the secondary outcomes, the association between the exposure group (O-GDM, O-Type1, and O-NoGDM) and the outcome variable was not changed (data not shown).

The multivariate analyses were essentially unchanged if only offspring with normal glycemic status was included (data not shown).

Discussion

Principal findings

Offspring exposed to hyperglycemia in utero due to maternal GDM or type 1 diabetes were characterized by impaired insulin sensitivity as well as a relatively reduced insulin release in comparison with offspring from the background population. The absolute level of insulin release did not significantly differ between exposed and nonexposed offspring.

Other studies

This is the first study evaluating insulin sensitivity and insulin release in adult offspring of women with GDM. Only one other study has previously examined adult offspring of women with type 1 diabetes during pregnancy (8). The study included offspring aged 22–24 years. Fifteen offspring of mothers with type 1 diabetes were compared to 16 offspring of fathers with type 1 diabetes. None had a family history of type 2 diabetes and the human leukocyte antigen (HLA) DR and DR4 distribution was similar in the groups, whereas no information regarding ethnicity was given. Intrauterine exposure to hyperglycemia was associated with a defect in insulin secretion but not with adiposity or insulin resistance in the offspring. In our study, we found both increased body weight and impaired insulin sensitivity in diabetes exposed offspring, but in accordance with Sobngwi et al (8), we also found diminished insulin secretion when insulin action was taken into account.

Others studies in prepubertal diabetes-exposed Caucasian offspring suggest that the primary abnormality is impaired insulin sensitivity (32, 33). A recently published study found that maternal gestational glucose concentration was negatively associated with offspring insulin sensitivity and positively associated with β-cell responsivity (34). The finding of decreased insulin sensitivity is in concordance with our finding, although found in a small (n = 21) and younger (5–10 y) cohort of mixed ethnicity (34). In both our groups of adult diabetes-exposed offspring, we found a significantly decreased insulin secretion in terms of DI, taking the prevailing insulin sensitivity into account. A low DI is an early marker of β-cell depletion and has been found to predict future development of diabetes (35).

In support of our data, a study of 22- to 27-year-old Pima Indians found a lower insulin secretion in the offspring of parents with an early onset of diabetes, and this effect was even more prominent in subjects exposed to an intrauterine diabetic environment (36).

Regarding O-NoGDM, we found that these offspring were insulin resistant and had a decreased insulin response (lower DI₁ and DI₂) compared with O-BP. Mothers of O-NoGDM were heterogeneous with respect to risk factors for GDM (eg, overweight and family disposition to diabetes). It could therefore be speculated whether these findings were induced by mild hyperglycemia within the nondiabetic range or were the results of genetic predisposition. The latter hypothesis is supported by a previous study finding of a lower insulin sensitivity and DI in white youth with a family history of type 2 diabetes, but the study did not evaluate potential effects of exposure to intrauterine hyperglycemia (37).

Strengths and weaknesses of the study

The present study is based on a large sample size and long-term follow-up with detailed information on potential confounding factors. Furthermore, detailed objective information on pregnancy and delivery were available from maternal hospital files. The data in our study that most strongly support the hypothesis, that hyperglycemia during pregnancy contributes to increased risk of glucose intolerance, come from the comparison between the offspring of women with type 1 diabetes and offspring from the background population: in the adjusted analyses, we found a significantly decreased insulin sensitivity and disposition index (Table 3). However, the fact that offspring of women with GDM differed significantly only from the background population and not statistically significantly from the genetic high-risk reference group (NoGDM), except from the univariate analysis comparing the disposition index based on CIR and the Matsuda Index, may to some extent question the effect of intrauterine hyperglycemia, at least in a population of offspring with a high genetic predisposition to type 2 diabetes. The very similar phenotype found in the O-GDM and O-NoGDM groups indicates strong effects of genes and that numerous other factors may have potential for fetal programming (eg, maternal overweight and lipid profile), which in a genetic high-risk population may overrule the effects of intrauterine hyperglycemia. Furthermore, mothers in the NoGDM group presented slightly higher glucose levels within the normal range than the mothers from the background population (4), which in accordance with the Hyperglycemia and Adverse Pregnancy Outcome (HAPO) data is likely to influence fetal growth and developmental programming mechanisms (38). In turn, this may weaken the status of the O-NoGDM as an unexposed reference group because mild exposure to hyperglycemia cannot be excluded.

Numerous indices based on data from an OGTT have been developed and compared with the gold standard, ie, the euglycemic-hyperinsulinemic clamp or with results from IVGTTs to find surrogate measures of insulin sensitivity and insulin secretion (39, 40). In this large-scale human study, we chose an OGTT to be the diagnostic tool of the metabolic status to obtain data within a reasonable time. Although the BIGTT-S_{I-0–30–120} and Matsuda index primarily describe the insulin resistance in muscle and adipose tissue, the HOMA-IR index gives an estimate of insulin resistance in the nonstimulated fasting state.

The exposure variables in O-GDM (maternal OGTT data at 0 and 120 min) and O-Type1 (mean maternal blood glucose in first and third trimesters) may be considered to be fairly rough. However, these are the only data available because self-monitored glucose measurements and glycosylated hemoglobin (HbA1C) determinations were not introduced in clinical practice between 1978 and 1985. Information on potential confounding lifestyle factors in the adult offspring is based on self-estimated information in a questionnaire with the inherent theoretical risk of recall bias.

Conclusion

Exposure to intrauterine hyperglycemia is associated with impairments of insulin secretion as well as insulin action in young adulthood. Altogether this is contributing to an increased risk of type 2 diabetes in offspring of women with diabetes in pregnancy. The quantitative contribution of these findings to the overall risk of developing type 2 diabetes is unknown and needs to be established.