The role of adiponectin, tumour necrosis factor α (TNFα), leptin and C-reactive protein in the insulin resistance of pregnancy is not clear. We measured their levels in women with gestational diabetes (GDM) and in controls, during and after pregnancy, and related them to insulin secretion and action.
Nineteen women with GDM and 19 BMI-matched healthy pregnant women underwent intravenous glucose tolerance tests in the third trimester of pregnancy and 4 months postpartum to determine insulin sensitivity (SI) and insulin secretion. Adiponectin, TNFα, leptin and high sensitivity CRP (hsCRP) were measured in fasted blood.
Of the circulating factors, only leptin (r = −0.41, p = 0.01) correlated with SI in pregnancy. Leptin and hsCRP levels were elevated in pregnancy compared to postpartum (leptin (mean ± SEM): 27.8 ± 2.4 vs 19.3 ± 2.1 ng/mL, p < 0.001; hsCRP: 5.2 ± 0.7 vs 3.2 ± 0.6 mg/L, p < 0.001). Adiponectin levels did not change from pregnancy to postpartum, despite a marked increase in SI. All four factors correlated with SI postpartum (adiponectin: r = 0.38, p = 0.01; TNFα: r = −0.48, p = 0.002; Leptin: r = −0.61, p = 0.001; hsCRP: r = −0.48, p = 0.002). TNFα correlated inversely with insulin secretion in pregnancy (r = −0.35, p = 0.03) and was significantly higher in the GDM group (2.62 ± 0.3 vs 1.88 ± 0.3 pg/mL, p = 0.01) in pregnancy.
During pregnancy, insulin sensitivity (IS) declines by 50% by the third trimester 1, which may be mediated by increases in hormones such as oestrogen, progesterone, β-human chorionic gonadotrophin, human placental lactogen, growth hormone and cortisol 2. A recent study of 15 subjects suggested a role for tumour necrosis factor α (TNFα) 3 in this pregnancy-induced insulin resistance, and other studies have found associations between TNFα and surrogate measures of insulin sensitivity such as fasting C-peptide 4. The role of adiponectin, which enhances insulin action and exerts opposite effects to TNFα 5, is beginning to be explored during normal and diabetic pregnancy 6, 7. No studies to date, however, have examined adiponectin levels in pregnancy employing the techniques that permit the precise measurement of insulin sensitivity.
Women with gestational diabetes (GDM) demonstrate defects in insulin secretion during pregnancy and defects in insulin sensitivity and secretion postpartum 8, 9. Given that there are possible important roles for the adipocytokines in the early defects of type 2 diabetes, GDM women represent an ideal population to study these inter-relationships further. To date, there is limited cross-sectional information that TNFα may be elevated in GDM pregnancy 4, 10. Other cross-sectional reports have shown that adiponectin levels are reduced in women with GDM 6, 7, 11, 12. Leptin, an adipocytokine, which is also produced by the placenta and is involved with weight regulation and metabolism, has been variously reported as being elevated 13, normal 14 or reduced 15 in GDM pregnancy. Given the uncertainties with the current cross-sectional adipocytokine data, a role for these cytokines in the pathophysiology of GDM has not been clearly established.
The purpose of this study was therefore to determine if these cytokines were (1) linked to the insulin resistance or insulin secretion of late pregnancy, (2) altered in GDM compared to non-diabetic pregnant women and (3) related to persisting insulin secretion and action defects of former GDM women postpartum. Since adipocytokines are secreted from adipose tissue and relate strongly to adiposity, it is important to match for adiposity to remove it as a confounding factor, when investigating the possible role in GDM. To this end, we employed a frequently sampled intravenous glucose tolerance test (FSIVGTT) to measure insulin sensitivity and insulin secretion during late pregnancy and postpartum in a group of women with GDM and age- and BMI (body mass index)-matched healthy pregnant women.
Materials and methods
Subjects were recruited from the obstetric clinics of the Mercy Hospital for Women, Box Hill and Werribee Mercy Hospitals. Nineteen women with GDM as diagnosed by the Australasian Diabetes in Pregnancy Society criteria 16 (75 g OGTT: fasting glucose >5.5 mmol/L; 2 h glucose >8.0 mmol/L) in their 28-week OGTT were recruited. All were negative for GAD antibodies, except for one subject who had a normal OGTT postpartum. Seven GDM subjects were on insulin treatment at the time of the pregnancy studies, the remainder were controlled on diet alone. Nineteen control women, pair matched for BMI in pregnancy (±4 kg/m2) and age (±4 years) and with no first-degree relative with diabetes, a history of polycystic ovarian syndrome or previous GDM, were recruited. Subjects were studied in the third trimester of pregnancy (mean 34.0 ± 0.3 weeks' gestation) and at 4.0 ± 0.4 months' postpartum. An additional GDM and control subject were studied postpartum (as the GDM subject declined to be studied in pregnancy, but consented to the postpartum studies), therefore n = 20 for the postpartum studies. All were Australian of European descent. Of the 20 control subjects, 18 were breastfeeding at the time of the postpartum study, compared to 14 of 20 GDM subjects. Six control women were taking the progesterone-only pill, and one the combined contraceptive pill; of the GDM women, four were on progesterone-only preparations and one on the combined pill. These differences were not statistically significant. The studies were approved by the Research and Ethics Committees of the hospitals, and all subjects gave informed written consent.
The women fasted from 10 PM the night before, having consumed three meals of >40-g carbohydrate the previous day. The subjects were weighed, asked to rest on a bed or chair and a vein in the antecubital fossa was cannulated (‘Insyte’ 18 gauge, Becton Dickenson, Sydney, Australia). In the postpartum studies, waist–hip ratio (WHR) and percentage of fat by bioelectrical impedance analysis were also measured. These measures of adiposity were not utilized in the pregnant state because of their unreliability during pregnancy. After resting, fasting blood was taken and 0.3 g/kg of 50% dextrose was diluted half by saline and administered intravenously over 1 min. A maximum dose of 20 g was used during pregnancy and 25 g postpartum (in order to not overestimate the glucose dose based on a late pregnancy weight) and the cannula flushed with 20 mL of normal saline immediately after the IV dextrose. Samples were drawn from the same cannula at 2, 3, 4, 6, 8, 10, 15, 20, 30, 40, 60, 75, 90 and 120 min, with flushing of the cannula with 3 mL of normal saline between each draw 17. Samples were placed into pre-chilled tubes (serum/EDTA for TNFα, Ad, high sensitivity CRP (hsCRP), leptin and lipids; 4% sodium fluoride 25 µL/mL blood for glucose and insulin), centrifuged within 40 min and stored at −80 °C until time of assay. A standard 75-g OGTT was performed at 6–9 weeks postpartum in all subjects.
Plasma glucose was analysed by a glucose oxidase method (YSI 1500 ‘Sidekick’ analyser), coefficient of variation (CV) 2.4%. Insulin was measured by RIA, with <1% cross-reactivity to pro-insulin and CV of 8.7% at insulin levels 6 mU/L, 4.3% at 20 mU/L and 3.5% at 30 mU/L. Plasma insulin antibodies were not present in any samples. Samples were tested in duplicate in Ad, TNFα and Leptin assays, with pregnant and postpartum samples for each GDM subject and her respective control performed in the same assay. Adiponectin was analysed using a RIA (Linco Research, St Charles, USA); inter-assay CV 4.6%. TNFα was measured using a sandwich ELISA (Quantikine, R&D, Minneapolis, Minnesota), with a sensitivity of 0.5 pg/mL. Precision was assessed using Bland–Altman analysis: the limits of agreement of duplicates were 0.001 ± 0.009 pg/mL 5 of 80 samples fell outside these limits. Leptin was measured by RIA (Linco Research, St Charles, USA), inter-assay CV 8.4%. CRP was measured by high-sensitivity nephelometry (Dade Behring BNII, Marburg, Germany); Intra-assay CV 3.7%, inter-assay CV 6.5%. Lipids were measured on a Roche Modular analyser using standard enzymatic, colourimetric methods: cholesterol was measured using cholesterol esterase and cholesterol oxidase, triglyceride using the Roche lipoprotein lipase method and high-density lipoprotein (HDL) using a homogeneous HDL-cholesterol method.
Acute insulin release to glucose (AIRg) and SI were calculated using Minmod 18. Disposition index (AIRg × SI) was calculated as an assessment of SI adjusted AIRg. Kg (glucose disappearance) is equal to the slope (×100) of the natural logarithm of glucose values from 10 to 40 min after glucose infusion. Our FSIVGTTs continued for only 120 min because of the women's comfort and social needs, so the basal values for glucose and insulin were used to create a 180-min time point, which allows accurate estimation of SI by Minmod. This methodology has been demonstrated to be valid in insulin-sensitive individuals 19. We also undertook further validation studies in insulin-resistant individuals (10 relatives of diabetic subjects before and after severe insulin resistance was induced by serial dexamethasone doses 20). The 120-min method correlated well with the standard 180-min calculations both in the pre (r = 0.96, p < 0.001) and post (r = 0.98, p < 0.001) dexamethasone studies; the mean difference was −0.14 ± 0.17[mU/L]−1 × 10−4, t-value −0.84 (pre) and −0.10 ± 0.14[mU/L]−1 × 10−4, t-value −0.76 (post) and there was no error bias in its calculation.
Data for GDM versus control subjects and pregnancy versus postpartum for each subject were analysed using paired t−tests, with transformation of the data where appropriate to normalize their distribution. The Bonferonni correction was applied to these analyses, so p < 0.025 was accepted as the level of significance for the paired t-tests. For Pearson correlations and linear multiple regression, data was transformed where appropriate to normalize the distribution. For non-parametric data (analysis of % change of SI, Triglycerides and HDL), Spearman correlations were used. Four subjects (two control and two GDM) who had extremely low TNFα levels postpartum—giving very high values for the Ad/TNFα ratio above five standard deviations above the mean—were excluded from the analyses involving correlations with Ad/TNFα. Statistical analysis was performed using Minitab and SPSS for windows. Statistical significance was taken at p < 0.05.
During pregnancy, in both groups, SI was significantly reduced and AIRg elevated compared to postpartum. GDM subjects demonstrated significantly reduced AIRg, but equivalent SI compared to their matched control subjects (Table 1). In contrast, postpartum GDM subjects showed defects in both disposition index and SI.
Table 1. Pregnancy and postpartum intravenous glucose tolerance test (IVGTT) results and subject characteristics
Pregnancy (n = 19)
Postpartum (n = 20)
Pregnancy (n = 19)
Postpartum (n = 20)
Data is expressed as mean ± standard error
p < 0.001.
p ≤ 0.005 for postpartum vs pregnant state within each group (excluding the extra subject added to each group postpartum).
Adipocytokines and CRP in pregnancy and postpartum: GDM versus control subjects (Figure 1)
Serum TNFα was significantly higher in the pregnant subjects with GDM compared to control subjects (2.6 ± 0.3 vs 1.9 ± 0.3 pg/mL, p = 0.01). TNFα was not higher in the GDM subjects postpartum (2.0 ± 0.3 vs 1.4 ± 0.2 pg/mL, p = 0.2). In both groups combined, there was a trend towards higher levels during pregnancy (2.3 ± 0.2 vs 1.7 ± 0.2 pg/mL, p = 0.08).
Adiponectin levels were not significantly different between GDM and matched control women either in pregnancy (8.0 ± 0.8 vs 9.4 ± 0.8 µg/mL, p = 0.28) or postpartum (7.6 ± 0.9 vs 8.6 ± 0.8 µg/mL, NS). There was also no difference in either group, separately or combined, between the pregnancy levels and the postpartum levels (7.8 ± 1.10 vs 7.2 ± 1.1 µg/mL, NS). The ratio of Ad/TNFα tended to be lower, but was not significantly different between GDM and control subjects either in pregnancy (3.78 ± 0.58 vs 5.21 ± 0.79, p = 0.054) or postpartum (5.11 ± 1.21 vs 8.09 ± 1.23, p = 0.07).
Plasma leptin was significantly higher in pregnancy than postpartum in both groups (27.8 ± 2.4 vs 19.3 ± 2.1, p < 0.001). Plasma leptin was higher in control subjects compared to GDM in pregnancy, although this was of borderline significance (32.2 ± 3.9 vs 23.0 ± 2.3 ng/mL, p = 0.05). Postpartum, this trend remained, but was not significant (21.7 ± 3.7 vs 16.9 ± 2.2 ng/mL, p = 0.14).
HsCRP levels were significantly higher for GDM and control women during pregnancy (5.1 ± 0.7 vs 3.2 ± 0.6 mg/L, p < 0.001), but there was no significant difference between groups either during pregnancy or postpartum.
Adipocytokines, CRP and insulin sensitivity in pregnancy (Table 2)
Table 2. Univariate correlations for adiponectin, TNFα, adiponectin/TNFα, leptin and hsCRP with metabolic parameters in all subjects
TNFα levels in pregnancy did not correlate with SI, fasting insulin(r = −0.22, p = 0.19), BMI or lipid parameters. Adiponectin levels also did not correlate with SI, BMI or lipid parameters, nor was the calculated adiponectin/TNFα ratio associated with SI in pregnancy. During pregnancy, leptin correlated with SI and BMI. HsCRP was associated with BMI, but not with SI or the lipid parameters.
Adipocytokines, CRP and insulin sensitivity postpartum (Table 2)
In contrast to the situation in pregnancy, postpartum TNFα correlated negatively with SI and HDL and positively with fasting insulin, BMI, % fat and WHR.
Postpartum adiponectin levels correlated significantly with SI and fasting insulin (r = −0.34, p = 0.03); there was a trend for an association with HDL and TG, but none for BMI, WHR or % fat. The adiponectin/TNFα ratio was significantly associated with SI and its related parameters.
Postpartum leptin was associated with SI, fasting insulin (r = 0.65, p < 0.001), BMI, % fat, HDL and TG. HsCRP was associated with SI, fasting insulin (r = 0.45, p = 0.004), BMI, % fat, WHR and HDL.
Although all of the cytokines examined were associated with postpartum SI, the only independent predictors of SI were TG, HDL, leptin and GDM status, with an r2 of 68% for this linear regression model. In contrast, in pregnancy, only leptin retained a significant association with SI (leptin and SI: r = −0.41, p = 0.01).
There was an association for the pregnancy-induced changes (expressed as % change) between TG and HDL with SI for each individual—such that the greater the drop in SI during pregnancy, the greater the elevation in TG and the lesser the elevation HDL during pregnancy (% change SI and % change TG: r = −0.60, p < 0.001; % change SI and % change HDL: r = 0.39, p = 0.02). There was no association between % change SI and % changes to TNFα, adiponectin, hsCRP or leptin
There was a negative correlation between TNFα and AIRg, and between TNFα and the AIRg corrected for SI (disposition index) for all subjects during pregnancy. Postpartum, these associations were not present. In GDM subjects alone, however, TNFα correlated with the disposition index both in pregnancy (r = −0.46, p < 0.05) and postpartum (r = −0.50, p = 0.02).
This study examined whether four factors (adiponectin, TNFα, leptin and hsCRP) with associations to insulin sensitivity were altered in pregnancy and whether they could be related to the changes of insulin resistance or insulin secretion of pregnancy, particularly in women with GDM.
Adipocytokine and CRP levels in pregnancy and postpartum
In the pregnant state, TNFα levels were elevated, in accordance with previous literature 3, 4, 21. TNFα was further elevated in pregnant GDM compared to their BMI-matched pregnant control subjects. This is also consistent with previous findings 3, 4, although, to our knowledge, this is the first study to document higher TNFα levels in GDM pregnancies compared to a closely BMI-matched group of healthy control women.
No significant change in adiponectin values from the third trimester to four months postpartum emerged, consistent with another recent study 22. Our GDM subjects did not have significantly different adiponectin levels to the controls, although there was a slight trend for the healthy control women to have higher levels postpartum.
Our finding of elevated leptin in pregnancy may be due to increased secretion from adipocytes in the presence of elevated oestrogen 23 and placental production 24, and is consistent with previous reports 15, 25. We observed reduced leptin levels in GDM compared to that in control subjects. Reports of leptin levels in GDM pregnancy are conflicting, with investigators reporting similar 14, reduced 15 and elevated 13 leptin levels in GDM women compared to healthy pregnancy controls.
HsCRP is an acute phase reactant, elevated levels of which are associated with features of the metabolic syndrome 26, and have also been documented to be elevated in pregnancy 27. Previous reports have shown that where BMI and adiposity are taken into account, hsCRP is not significantly associated with GDM 28, 29. Our finding in closely BMI-matched subjects confirms that hsCRP is linked to adiposity, but not glucose tolerance, in GDM women. There was no association with SI suggesting that the elevation of hsCRP observed is not an important cause or consequence of the reduction of SI in pregnancy.
Adipocytokines and CRP: relationship to insulin sensitivity
In the postpartum state, all of the adipocytokines studied demonstrated their expected association with SI, which confirms the robust nature of our methodology with pregnant and postpartum samples for each subject measured in the same assay.
In pregnancy, only leptin retained a relationship with SI, an association that has been noted previously 3, although it remains unknown if leptin has a direct effect on SI in pregnancy.
In contrast to previous investigators, we did not observe an association between TNFα and SI during pregnancy in either GDM or control women, though this relationship was apparent postpartum. Kirwan et al. noted a correlation between TNFα and SI in a group of 15 women in the third trimester, divided equally into lean normal glucose tolerant (NGT), obese NGT and GDM groups 3. The explanation for the difference in our findings and those of Kirwan et al. may lie in the BMI range of subjects tested. Our GDM and control subjects with an average postpartum BMI of 28 kg/m2 correspond to the five GDM subjects (BMI 30.8 kg/m2) and five obese NGT subjects (BMI 27.3 kg/m2) respectively in the study by Kirwan et al. where the serum TNFα was not significantly different (2.84 ± 0.17 vs 2.80 ± 0.72 pg/mL), despite a difference in SI (4.9 ± 0.8 vs 9.5 ± 1.510−2 mg/kg FFM/min/µU/mL). It appears that the association between TNFα and SI in pregnancy is less evident in this overweight group of subjects. A cross-sectional study by Winkler et al4, who found an association between fasting C-peptide and TNFα also tested a leaner group of women, and this relationship was not apparent on multivariate analysis once BMI was accounted for, which is in keeping with our findings.
In our subjects there, was no association between adiponectin and SI in pregnancy, which is consistent with Ranheim et al. who reported no correlation between Adiponectin and fasting insulin during pregnancy in 51 pregnant women 6. In contrast, Cseh et al7 found reduced adiponectin levels in the second and third trimester compared to the first trimester in a cross-sectional study of healthy pregnant non-diabetic subjects. They postulated that reduced adiponectin has a role in the insulin resistance of pregnancy and especially of GDM where they found further reductions. Our prospective study does not suggest such a role for adiponectin.
Our observation of reduced SI in former GDM women postpartum is in keeping with the findings reported by previous investigators 9, 30, and, given the association of adiponectin with SI, one would expect that, in a larger sample, GDM women would have statistically significantly reduced adiponectin levels. Indeed, a recent study of 180 women found reduced adiponectin levels in GDM 12. However, we would speculate that this was due to their underlying lower insulin sensitivity in the non-pregnant state rather than changes in pregnancy, as suggested by Winzer et al's study of 89 former GDM subjects 31. Our paired pregnant and postpartum data highlights that the association of adiponectin with SI in pregnancy is weaker than that observed postpartum.
Adiponectin and TNFα produce opposite effects on insulin signalling—with TNFα inhibiting 32 and adiponectin increasing 33 tyrosine phosphorylation of the insulin receptor. TNFα may inhibit the synthesis of adiponectin 5. The ratio of these cytokines may therefore be an important determinant of insulin sensitivity. In our subjects, postpartum adiponectin/TNFα correlated with SI and its related parameters of TG, HDL, % fat and BMI, and had a higher correlation with SI than either adiponectin or TNFα alone.
Lactation and the oral contraceptive pill use (by a minority of subjects) may have had an impact on adiponectin levels and SI in our subjects. Combs et al. noted an inhibitory effect of prolactin and oestrogen on adiponectin levels in mice 34, which may be important in our study given that 32/40 subjects were lactating at the time of the postpartum study. However, the adiponectin levels were consistent with previous studies for women 35, 36 and we found no evidence of altered adiponectin levels in the non-breastfeeding subjects (non-lactating: mean adiponectin 6.9 ± 1.7 vs lactating 8.4 ± 0.6µg/mL, p = NS).
There are profound alterations in lipid levels during normal pregnancy. Under the influence of elevated oestradiol, TG, HDL and cholesterol synthesis are stimulated and levels increase by 200–310%, 15–40% and 30–65% respectively 37, 38. The normal relationships between plasma TG, HDL and SI are to some extent modulated by these hormonal influences. However, our finding of an association between the changes in HDL and TG with the changes in SI indicates that there is still an influence of SI on HDL and TG in pregnancy. The question then arises as to whether this relationship may exist for the adipocytokines. There was no association, however, between % change TNFα, adiponectin or leptin with change in SI for each individual, which suggests that in our study population neither TNFα nor adiponectin are responding to or driving the changes in SI during pregnancy.
Adipocytokines and CRP: relationship to insulin secretion
In GDM subjects, the physiological decrease in SI in pregnancy may unmask an abnormality in glucose metabolism, which is not apparent in the non-pregnant state. Thus, it is the defect in insulin secretion, which differentiates those women who develop GDM from those who maintain normal glucose tolerance 39. A novel finding of this study was the association between serum TNFα and acute insulin secretion—AIRg. In pregnancy, for all subjects, TNFα correlated with AIRg, although this only remained significant for the GDM group postpartum. Furthermore, for GDM subjects, there was a significant association between TNFα and disposition index in pregnancy. This raises the possibility that in pregnancy TNFα could be having a deleterious impact on insulin secretion in GDM subjects. This concept is supported by in vitro evidence of a TNFα inhibiting beta cell function 40, 41.
We also observed an association between serum leptin levels and AIRg for control subjects in pregnancy and postpartum. This is most likely due to leptin's close association with BMI and SI—which, because of the inverse relationship between AIRg and SI in normal subjects, will produce an association between leptin and AIRg. This was not seen in GDM subjects, where their insulin secretory deficit has uncoupled the association between AIRg and SI.
In conclusion, our findings suggest that adiponectin, TNFα and hsCRP do not appear to contribute greatly to pregnancy-induced insulin resistance in GDM or in women with a healthy pregnancy, while leptin retains its association with SI during pregnancy. TNFα is elevated in GDM pregnancy, and may play a role in impairing insulin secretion in these subjects.
This work was supported by the National Health and Medical Research Council of Australia, the Medical Research Foundation for Women and Babies, Cardiovascular Lipid research grant and a Novo Nordisk Regional Diabetes Support Scheme Grant. The authors also thank our research nurse Noella Johnson, Dr Kevin Rowley for statistical advice and the women who participated.