Relationships between regional and ectopic adiposity and insulin sensitivity in early and late pregnancy

Abstract

Objective

To determine relationships between measurements of total body, visceral, and ectopic (liver, skeletal muscle) fat with insulin sensitivity in pregnancy.

Research Design and Methods

Pregnant women of varying pre-pregnancy weights were prospectively studied in early (n=59) and late (n=47) gestation. At each visit participants underwent body composition measurements including fat mass (FM), fat-free mass (FFM), abdominal subcutaneous (SAT), visceral (VAT) fat, ectopic lipid amounts in liver (IHL) and calf skeletal muscle (IMCL, EMCL) and hyperinsulinemia-euglycemic clamp to determine insulin sensitivity (Rd), endogenous glucose production (EGP), hepatic insulin sensitivity index (HISI), and free fatty acid (FFA) levels.

Results

In early pregnancy, Rd ((mg/kg FFM/min)/μIU/mL) inversely correlated (P-values < .05) with body mass index (BMI), FM, SAT, VAT, IHL, IMCL, and FFA. HISI inversely correlated (P-values < .05) with BMI, FM, SAT, VAT, IMCL, and FFA, but not IHL. In late pregnancy, however, neither EGP ((mg/kg FFM/min)/μIU/mL) nor Rd correlated with regional or ectopic fat measures, but HISI remained inversely correlated with BMI, FM, SAT, VAT, and IMCL. Early pregnancy IHL levels did not predict late pregnancy insulin sensitivity. Pregnant women with pre-pregnancy obesity were more insulin resistant but gained less gestational weight, VAT, and SAT, and experienced less decline in insulin sensitivity, than normal pre-pregnancy weight women.

Conclusions

Insulin resistance in early pregnancy is strongly associated with total, regional, and ectopic adiposity. However, in late pregnancy, factors other than regional and ectopic adiposity predominately influence insulin sensitivity. Pre-pregnancy weight categories proportionately alter gestational weight gain, adiposity distribution, and glucometabolic responses.

Graphical Abstract

Twitter Summary:

Maternal visceral, hepatic, and skeletal muscle fat accumulation tracks with insulin resistance in early but not late pregnancy when placental factors may become dominant.

Obesity and insulin resistance are significant risk factors for adverse maternal and fetal pregnancy outcomes (1–3). Previous studies of small groups of women have shown that insulin sensitivity declines with pregnancy in women who are normal weight and those with obesity (4–6), both with and without gestational diabetes (GDM) (7–9). Interestingly, the reduction in late-gestation insulin sensitivity was greater in the normal-weight women (5) than those with obesity (6). These studies also demonstrate wide variability in insulin sensitivity among pregnant women with similar body mass index (BMI), the cause of which remains poorly understood.

One explanation may be that BMI is not an accurate measure of regional and ectopic fat accumulation, collectively referred to as adiposity. One of the strongest links between adiposity and insulin resistance among non-pregnant adults is excess accumulation of fat in the abdominal visceral adipose tissue (VAT) depot, more so than abdominal subcutaneous adipose tissue (SAT) depots (10–13). Individuals with increased visceral adiposity also tend to accumulate excess liver fat (intrahepatic lipid or IHL) and are at risk for hepatic steatosis (14–16). Studies have shown that both VAT and IHL accumulation are directly related to insulin resistance in non-pregnant adults (15; 17; 18), with newer data indicating that variations in IHL amounts are more strongly associated with both hepatic and whole body insulin sensitivity than VAT, providing an explanation for insulin resistance variations in persons of similar visceral adiposity (15; 18). While limited evidence suggests that pregnancy contributes to changes in central and visceral fat content (19–21), no studies to date have assessed whether changes in visceral, muscle, and liver adiposity are associated with insulin sensitivity parameters during gestation. Since regional and ectopic adiposity measures can vary greatly in people of the same BMI, this could explain variations in insulin sensitivity in pregnant women of similar BMI noted above.

Therefore, we prospectively quantified abdominal, skeletal muscle, and hepatic lipid content using magnetic resonance imaging (MRI) and spectroscopy (MRS) and performed hyperinsulinemic-euglycemic clamps in early and late pregnancy to test associations of these body adiposity measures and insulin sensitivity in women with pre-pregnancy BMIs ranging from normal to those with obesity. We hypothesized that changes in fat distribution would closely correlate with insulin sensitivity during gestation, and specifically that greater fat deposition in the visceral and ectopic liver depots would be associated with greater insulin resistance in early and late pregnancy. We also hypothesized that of all the adiposity measures, early pregnancy IHL would best predict changes in insulin sensitivity during gestation (15; 18).

Research Design and Methods

Study population

The methods for this prospective cohort study are published elsewhere (22; 23). Briefly, pregnant women receiving care at Kaiser Permanente Northwest (KPNW) and Oregon Health & Science University (OHSU) were recruited between November 2014 and April 2017. Patients were eligible for the study if they were between 18 and 45 years of age, less than 12 weeks pregnant with a singleton gestation, had a BMI between 18.5 kg/m² and 38 kg/m² at time of enrolment, were fluent English speakers, were willing and able to undergo MR measurements, and did not have any excluded conditions (22), including pre-existing or gestational diabetes. Potential participants underwent an explanatory visit at the OHSU Clinical Translational and Research Center and those who consented to participate in the study returned for study visits in early (between 12-17 weeks of gestation, n=59) and late (between 32-36 weeks of gestation, n=47) pregnancy (Supplemental Figure 1).

Demographic measures

Demographic variables were extracted from the electronic health record including parity and preconception BMI (22) and from the participant survey (race/ethnicity). Categories for normal weight, overweight, and obesity were defined using cut-offs of 18.5-24.9 kg/m², 25-29.9 kg/m², and ≥ 30 kg/m² (24).

Body Composition Measures

Height was measured at the first study visit to allow for the calculation of BMI. At each study visit, weight was measured using the same calibrated scale; fat mass (FM), fat-free mass (FFM), and percent body fat (% BF) were determined using air displacement plethysmography (BOD POD, COSMED USA, Inc., Concord, CA) with adjustments for gestational age using van Raaij equations (25); MRI and MRS (Siemens Magnetom Tim Trio 3 Tesla whole body system, Siemens Medical Solutions, Malvern Pennsylvania and Erlangen, Germany) were used to quantify total abdominal fat, SAT, VAT, liver volume, intrahepatic lipid (IHL), intramyocellular lipid (IMCL), and extramyocellular lipid (EMCL) (at the mid-soleus muscle) as previously described (22).

Insulin Sensitivity

On a separate day and within a week of the body composition measures, participants underwent a hyperinsulinemic-euglycemic clamp to measure insulin sensitivity. For the three days prior to the clamp, participants were instructed to consume a standard diet consisting of 30% of total calories from fat sources, 15% from protein, and 55% from carbohydrates. The night before the study visit, participants were instructed to fast for 11 hours overnight. The morning of the clamp, a primed constant infusion of [6,6-²H₂] glucose (Cambridge Isotope Laboratories, Andover, MA) was started at 0.133 ml/min to achieve an intended enrichment ~1.0 mol percent excess and continued for the next four hours. At the completion of the first two hours of glucose isotope infusion, a primed constant infusion of regular insulin at 40 mU/m²/min was begun along with a variable infusion of D20 glucose (enriched with [6,6-²H₂]) to achieve at target glucose level of 5.0 mmol/L (range 4.7 to 5.3 mmol/L) during the final two hours of the study (Supplemental Figure 2).

Blood samples were collected prior to insulin infusion start and every 5 minutes during the clamp to allow calculation of the rates of glucose disposal (clamp Rd) and endogenous glucose production (EGP) using non–steady-state equations based on plasma [6,6-²H₂] glucose enrichment determined by gas chromatography mass spectrometry (26; 27). Because of changes in body composition, plasma volume, rate of clearance of insulin, and differences in achieved insulin levels during the clamp between early and late pregnancy (Supplemental Figure 2), clamp Rd values were adjusted for fat-free mass and insulin levels. Fat-free mass from early pregnancy was used in the adjustments at both study visits to avoid confounding from the influence of uterine contents and volume expansion on FFM in later pregnancy. Hepatic Insulin Sensitivity Index (HISI) was calculated using the reciprocal of the product of the basal endogenous glucose production rate and the fasting plasma insulin concentration as previously described (16; 28).

Insulin, glucose, and free fatty acid levels

Blood insulin levels were measured by ELISA (Mercodia AB, Uppsala Sweden); glucose levels were measured by hexokinase based colorimetric assay (Stanbio laboratory, Boerne Tx 78006); and free fatty acid (FFA) levels were measured by an enzymatic colorimetric kit (Wako Chemicals USA, Inc., Richmond, VA) in the Laboratory Core of the Oregon Clinical and Translation Research Institute (OCTRI).

Statistical analyses

We calculated means and standard deviations for body composition, insulin sensitivity, insulin, glucose, and FFA measures. Change scores between visit 1 and 2 (visit 2 measurement – visit 1 measurement) and their means and standard deviations were also calculated. Prior to analysis, we transformed variables that showed severe deviation from a normal distribution: for IHL at visit 1 and 2, we performed a negative reciprocal root transformation; for IMCL and EMCL at visit 1 and 2, we performed a natural log transformation. The winsorization method was used to address extreme outliers observed in basal EGP at visit 1 and adjusted clamp Rd at visit 2. Specifically, the most extreme value of each measure was replaced with the next less extreme value in the data; change scores were then calculated using the winsorized values. One-sample t-tests and the Wilcoxon signed-rank test (for IHL) were performed on change scores to determine whether each variable changed significantly from visit 1 to visit 2. To assess relationships between body composition measurements and insulin sensitivity at baseline and during pregnancy, we calculated correlation coefficients between each fat variable (body fat, percent body fat, fat mass, fat free mass, IHL, IMCL , EMCL, VAT, SAT, and free fatty acid levels) and insulin sensitivity (clamp Rd, EGP, and HISI) at first and second visits. Correlation coefficients were also calculated between the change in each fat variable and change in insulin sensitivity between the early and late pregnancy visits. Spearman correlation coefficients are reported for the relationship between change in IHL and changes in insulin sensitivity; Pearson correlation coefficients are reported for the remaining bivariate relationships. One-way analysis of variance (ANOVA) and Kruskal-Wallis tests were performed on baseline body composition and insulin sensitivity measurements and change scores to evaluate differences in mean (and median for IHL) values by preconception BMI groups. Hierarchical multiple regression analysis was performed with HISI as the dependent variable; baseline BMI was entered as a covariate in the first block and baseline IHL, VAT and fat mass as independent variables in the second block to assess change in insulin sensitivity during pregnancy after controlling for preconception BMI. Standardized regression coefficients are reported to evaluate the relative importance of each variable in the model.

Results

Demographics

A total of 59 women underwent both body composition and insulin sensitivity measurements at an early pregnancy visit and most (n=47) returned for repeat measurements at a late pregnancy visit (Supplemental Tables 1 and 2). The average (min-max) age was 30.3 (19.0-39.0) years and preconception BMI was 26.6 (17.1-37.7) kg/m². Participants were predominantly White (84.7%) and non-Hispanic (88.1%), and just over half were nulliparous (57.6%). The average (SD) gestational age was 15.5 (0.9) weeks at the first early-pregnancy visit and 34.1 (1.4) weeks at the second late-pregnancy visit.

Body Composition Changes During Gestation

Body weight, BMI, and fat mass increased 15% from early to late gestation, while percent body fat did not significantly change (+0.2%) (Figure 1, Supplemental Table 2). Other measures of body composition that showed increases during gestation included SAT (+23%) and VAT (+56%). In early pregnancy, mean (SD) IHL levels were low at 1.1 (2.0%) with levels > 5% (a threshold for hepatic steatosis (29)) found in only 2 participants. In late pregnancy, mean (SD) IHL increased slightly to 1.4 (2.4%), with > 5% levels found in 4 participants, one of which fell into this category at both study visits. Neither IMCL nor EMCL levels changed significantly between early and late gestation.

Figure 1:

Graphs of body composition and glucose metabolism by pre-conception BMI grouping in early pregnancy, late pregnancy, and change during pregnancy. SAT: subcutaneous abdominal adipose tissue. VAT: visceral abdominal adipose tissue. IHL: intra-hepatic lipid. IMCL: intra-myocellular lipid. Rd: rate of glucose disposal. HISI: Hepatic Insulin Sensitivity Index. Boxplot elements: Box=25^th to 75^th percentile; Whiskers=minimum and maximum observations below or above 1.5 (interquartile range or IQR) of the 25^th and 75^th percentile, respectively; vertical line within box=median, filled circle=mean; circle, triangle, x symbols indicate outliers beyond 1.5 (IQR) of the 25^th and 75^th percentiles, respectively. Maroon bar and lines: overall group. Light blue bar and lines: pre-pregnancy normal weight group (BMI 18.5-24.9 kg/m²). Medium blue bar and lines: pre-pregnancy overweight group (BMI 25-29.9 kg/m² kg/m²). Dark blue bar and lines: pre-pregnancy obesity weight group (BMI ≥ 30 kg/m²).

Compared to women with normal or overweight pre-pregnancy BMI, women with pre-pregnancy obesity had the highest percent body fat, fat mass, SAT, VAT, and IMCL in early pregnancy, P<.001 (Figure 1, Supplemental Table 3). On the other hand, intra-hepatic lipid content was not significantly different among the three preconception weight categories. Although not always reaching statistical significance, early to late pregnancy changes in percent body fat (P<.05), fat mass, SAT (P<.01), VAT, and IMCL (P<.05) were greatest in women in the normal pre-pregnancy BMI category and least in those with pre-pregnancy obesity (Figure 1, Supplemental Table 3). Notably, in the group of women with pre-pregnancy obesity, the mean gestational changes of several body composition parameters were reductions, including SAT and IMCL.

Glucose Metabolism Changes During Gestation

In the entire cohort from early to late pregnancy, fasting glucose levels decreased (P<.001), insulin levels increased (p<.001), and FFA levels did not significantly change (Table 1). Glucose disposal rates (clamp Rd adjusted for FFM in early pregnancy and insulin levels at each respective pregnancy visit) significantly declined by −16%. Both unadjusted basal and clamp EGP significantly increased by +22% and +144%, but neither were significant after adjusting for FFM-insulin levels (Table 1) and mean percent suppression of EGP during the clamp did not significantly change. On the other hand, the mean HISI decreased by −41%, P<.001.

Table 1.

Measurements of glucose metabolism in the entire cohort.

	N	Early Gestation Mean (SD)	N	Late Gestation Mean (SD)	N	Absolute Change Early to Late Mean (SD)	Percent Change Early to Late	P-value
Fasting Glucose (mmol/L)	59	4.5 (0.3)	47	4.2 (0.3)	47	−0.33 (0.37)	−7%	<.001
Fasting Insulin (pmol/L)	59	29.7 (21.4)	48	41.8 (22.8)	48	17.5 (15.2)	+81%	<.001
Fasting FFA (μM)	59	456 (151)	48	477 (159)	48	44.9 (195)	+21%	.12
Rd (mg/min)	59	440 (118)	47	383 (80)	47	−63.1 (96.0)	−11%	<.001
Rd adjusted for FFM in early pregnancy and insulin (mg/kg/min)/μIU/mL	59	0.17 (0.07)	47	0.14 (0.05)	47	−0.04 (0.06)	−16%	<.001
Basal EGP (mg/min)	59	196 (57)	46	219 (45)	46	28.0 (59.6)	+22%	.003
Basal EGP (mg/kg/min)/μIU/mL	59	0.08 (0.04)	46	0.08 (0.06)	46	0.01 (0.07)	+19%	.41
Clamp EGP (mg/min)	59	79.2 (45.3)	46	94.2 (33.1)	46	20.7 (47.5)	+31%*	.005
Clamp EGP adjusted for FFM in early pregnancy and inulin (mg/kg/min)/μIU/mL	59	0.03 (0.02)	46	0.03 (0.02)	46	0.01 (0.03)	+25%*	.05
% suppression EGP	59	61.1 (18.7)	46	57.4 (10.9)	46	−5.3 (21.9)	--	.11
HISI	59	12.7 (7.3)	46	7.1 (3.8)	46	−7.0 (6.6)	−41%	<.001

Values are reported as mean (SD). FFA: free fatty acid. Rd: rate of glucose disposal. EGP: endogenous glucose production. HISI: Hepatic Insulin Sensitivity Index.

^*Median percent change reported due to presence of extreme outlier.”

While a limited number of participants in the pre-pregnancy obesity category allows only descriptive analysis of many variables, several trends are evident in the data comparing across pregnancy weight groups. For example, examining changes across pre-pregnancy by BMI strata, early pregnancy glucose, insulin, and FFA levels were highest in the group with obesity compared to the normal and overweight groups (Figure 1, Table 2). All groups experienced reductions in fasting glucose levels and increases in fasting insulin levels. Changes in FFA levels amongst the weight groups were mixed and not significantly different between groups. Early-pregnancy clamp Rd was lowest in the group with obesity, and although clamp Rd decreased between early and late pregnancy in all three weight category groups, the group with normal preconception BMI experienced the greatest proportional decline, followed by the overweight group, with the least proportional decline in the group with obesity. Likewise, early pregnancy HISI was lower in participants with pre-pregnancy obesity with all groups experiencing further reductions during pregnancy, the greatest of which occurred in women with normal pre-pregnancy weight and the smallest in the women in the pre-pregnancy obesity category (Figure 1, Table 2).

Table 2.

Measurements of glucose metabolism by pre-pregnancy BMI weight category.

		All Participants	Normal Weighta (< 25 kg/m2)	Overweightb (25.0-29.9 kg/m2)	Obesec (≥ 30.0 kg/m2)	P-value*
		mean (95% CI)
Fasting glucose (mmol/L)	N	47	24	17	6
	Early Pregnancy	4.5 (4.4, 4.6)	4.5 (4.4, 4.7)	4.4 (4.2, 4.5)	4.8 (4.5, 5.1)	.03 b-c
	Absolute change (early to late)	−0.33 (−0.44, −0.23)	−0.34 (−0.49, −0.19)	−0.25 (−0.42, −0.071)	−0.55 (−0.85, −0.25)	.22
Fasting insulin (pmol/L)	N	48	24	17	7
	Early Pregnancy	24 (20, 28)	19 (14, 24)	25 (19, 31)	40 (30, 49)	.001 a-c, b-c
	Absolute change (early to late)	17 (13, 22)	16 (9.5, 22)	17 (9.8, 25)	24 (12, 35)	.48
Fasting FFA (μM)	N	48	24	17	7
	Early Pregnancy	433 (392, 474)	356 (307, 405)	491 (433, 549)	552 (461, 642)	<.001 a-b, a-c
	Absolute change (early to late)	45 (−12, 102)	104 (25, 182)	−11 (−104, 81)	−20 (−164, 125)	.11
Rd (mg/min)	N	47	24	17	6
	Early Pregnancy	446 (412, 479)	479 (434, 525)	419 (364, 473)	389 (298, 480)	.10
	Absolute change (early to late)	−63 (−91, −35)	−100 (−137, −63)	−31 (−75, 13)	−7.4 (−81, 67)	.02
Rd, adjusted for FFM in early pregnancy and insulin level at each pregnancy visit (mg/kg/min)/μIU/mL	N	47	24	17	6
	Early Pregnancy	0.17 (0.15, 0.19)	0.19 (0.17, 0.22)	0.16 (0.13, 0.19)	0.12 (0.075, 0.17)	.01 a-c
	Absolute change (early to late)	−0.035 (−0.053, −0.018)	−0.051 (−0.075, −0.026)	−0.024 (−0.053, 0.005)	−0.008 (−0.057, 0.041)	.19
Basal EGP (mg/min)	N	46	23	17	6
	Early Pregnancy	191 (175, 207)	178 (156, 200)	195 (169, 220)	230 (187, 273)	.10
	Absolute change (early to late)	28 (10, 46)	38 (13, 62)	27 (−2.0, 56)	−5.2 (−54, 44)	.30
Basal EGP, adjusted for FFM in early pregnancy and insulin at each pregnancy visit (mg/kg/min)/μIU/mL	N	46	23	17	6
	Early Pregnancy	0.076 (0.064, 0.087)	0.072 (0.056, 0.088)	0.074 (0.055, 0.092)	0.095 (0.063, 0.13)	.42
	Absolute change (early to late)	0.008 (−0.011, 0.027)	0.008 (−0.020, 0.035)	0.022 (−0.010, 0.054)	−0.030 (−0.084, 0.023)	.25
Clamp EGP (mg/min)	N	46	23	17	6
	Early Pregnancy	73 (62, 85)	67 (51, 84)	80 (61, 100)	78 (45, 110)	.57
	Absolute change (early to late)	21 (6.6, 35)	23 (2.9, 44)	16 (−7.9, 39)	25 (−15, 65)	.87
Clamp EGP, adjusted for FFM in early pregnancy and insulin at each pregnancy visit (mg/kg/min)/μIU/mL	N	46	23	17	6
	Early Pregnancy	0.027 (0.023, 0.032)	0.026 (0.020, 0.033)	0.029 (0.022, 0.037)	0.024 (0.012, 0.036)	.72
	Absolute change (early to late)	0.008 (−0.000, 0.015)	0.007 (−0.004, 0.018)	0.010 (−0.003, 0.023)	0.004 (−0.018, 0.026)	.84
EGP suppression (%)	N	46	23	17	6
	Early Pregnancy	63 (57, 68)	64 (56, 71)	58 (50, 67)	71 (56, 85)	.31
	Absolute change (early to late)	−5.3 (−12, 1.2)	−6.0 (−15, 3.2)	−0.80 (−12, 9.9)	−16 (−34, 2.5)	.37
HISI	N	46	23	17	6
	Early Pregnancy	14 (12, 16)	17 (15, 20)	12 (9.1, 15)	7.1 (1.7, 12)	.002 a-c
	Absolute change (early to late)	−7.0 (−8.9, −5.0)	−9.2 (−12, −6.5)	−5.5 (−8.5, −2.4)	−2.8 (−7.9, 2.4)	.05

^*Results from one-way ANOVA analyses are reported for all measures except absolute change in IHL (Kruskal-Wallis test).

^†Analyses performed on the negative reciprocal root of IHL values. Reported values are the squared reciprocal value of the negative reciprocal root estimates.

^‡Reported values are the median (IQR) of the absolute change in IHL from early to late pregnancy.

^§Analyses were performed on the natural log of IMCL and EMCL at early pregnancy. Reported values are exponentiated value of the log estimates.

FFA: free fatty acid. Rd: rate of glucose disposal. EGP: endogenous glucose production. HISI: Hepatic Insulin Sensitivity Index. Adjustments for FFM use only value from early pregnancy visit (see methods).

Correlations between adiposity measures and insulin sensitivity in early and late pregnancy

In early pregnancy, adjusted clamp Rd correlated inversely with most measures of body composition except EMCL and percent body fat, with the highest correlation with VAT (r = −0.52) (Table 3, Supplemental Figures 3 and 4). Unadjusted basal EGP was positively correlated with BMI, % BF, FM, and SAT, but not with other body composition parameters or FFA levels (Table 3); whereas basal EGP adjusted for FFM and insulin was positively correlated only with % BF and FM. The HISI was also inversely correlated with BMI, %BF, FM, SAT, VAT, and IMCL but not significantly correlated with IHL (Table 3, Supplemental Figures 3 and 4). Both clamp Rd and HISI were inversely correlated with fasting FFA levels (Table 3).

Table 3.

Pearson correlations between insulin sensitivity measures and body composition in early and late pregnancy.

	N	BMI	N	% fat	N	FM	N	SAT	N	VAT	N	IHL†	N	EMCL†	N	IMCL†	N	FFA
Early Pregnancy
Clamp Rd	59	−.22	59	−.14	59	−.16	59	−.19	59	−.36**	58	−.42**	53	.07	53	−.21	59	−.57**
Clamp Rd (adjusted for FFM in early pregnancy and insulin at each pregnancy visit)	59	−.46**	59	−.23	59	−.34**	59	−.38**	59	−.52**	58	−.42**	53	.01	53	−.35**	59	−.49**
Basal EGP	59	.44**	59	.48**	59	.47**	59	.41**	59	.24	58	−.14	53	.26	53	.14	59	.14
Basal EGP (adjusted for FFM in early pregnancy and insulin at each pregnancy visit)	59	.07	59	.27*	59	.15	59	.06	59	−.09	58	−.27*	53	.15	53	−.08	59	.09
Clamp EGP	59	.28*	59	.25	59	.27*	59	.25	59	.10	58	−.06	53	.37**	53	.16	59	−.04
Clamp EGP (adjusted for FFM in early pregnancy and insulin at each pregnancy visit)	59	.08	59	.18	59	.11	59	.07	59	−.09	58	−.18	53	.33*	53	.03	59	−.11
% EGP suppression	59	−.06	59	−.01	59	−.03	59	−.04	59	.01	58	.02	53	−.34*	53	−.13	59	.14
Hepatic Insulin Sensitivity Index (HSI)	59	−.71**	59	−.74**	59	−.71**	59	−.64**	59	−.64**	58	−.16	53	−.26	53	−.44**	59	−.36**
Late Pregnancy
Clamp Rd	47	.21	47	.10	47	.24	43	.22	43	.22	43	−.02	44	.07	44	.06	47	−.31*
Clamp Rd (adjusted for FFM in early pregnancy and insulin at each pregnancy visit)	47	−.22	47	−.17	47	−.21	43	−.17	43	−.18	43	−.16	44	.14	44	−.09	47	−.15
Basal EGP	46	.33*	46	.20	46	.36*	42	.29	42	.31*	42	.03	43	.15	43	.01	46	−.29*
Basal EGP (adjusted for FFM in early pregnancy and insulin at each pregnancy visit)	46	−.11	46	.01	46	−.06	42	−.02	42	−.07	42	−.11	43	.05	43	−.15	46	−.02
Clamp EGP	46	.36*	46	.13	46	.32*	42	.17	42	.21	42	−.06	43	.20	43	−.01	46	−.29*
Clamp EGP (adjusted for FFM in early pregnancy and insulin at each pregnancy visit)	46	−.01	46	−.01	46	−.02	42	−.07	42	−.08	42	−.14	43	.21	43	−.12	46	−.13
% EGP suppression	46	−.21	46	.01	46	−.12	42	.04	42	−.00	42	.10	43	−.18	43	.04	46	.17
HISI	46	−.41**	46	−.34*	46	−.42**	42	−.39*	42	−.40**	42	−0.2	43	−0.07	43	−.30*	46	−0.04

^*P-value<.05;

^**P-value<.01

^†IHL transformed using negative reciprocal root; EMCL and IMCL transformed using natural log.

Clamp Rd: rate of glucose disposal. EGP: endogenous glucose production. HISI: Hepatic Insulin Sensitivity Index. BMI: body mass index. FM: fat mass. SAT: subcutaneous abdominal adipose tissue. VAT: visceral abdominal adipose tissue. IHL: intra-hepatic lipid. IMCL: intra-myocellular lipid. EMCL: extramyocellular lipid. FFA: free acid.

In late pregnancy, adjusted clamp Rd was not significantly associated with any measure of body composition or fasting FFA levels (Table 3, Supplemental Figure 5 and 6). Unadjusted basal EGP in late gestation was positively associated with BMI, FM, VAT, and FFA, but the relationships were not significant after adjusting for FFM and insulin levels. On the other hand, HISI in late pregnancy correlated inversely with BMI, %BF, FM, SAT, VAT, and IMCL, but not IHL, EMCL, or FFA (Table 3, Supplemental Figures 5 and 6).

Correlations between changes in body composition and changes in insulin sensitivity

No significant correlations were found between changes in any measure of body composition from early to late gestation and changes in clamp Rd, EGP, or HISI (Supplemental Table 4). Only a single significant inverse relationship between changes in clamp Rd (adjusted for FFM and insulin) and fasting FFA was detected.

Correlations between early pregnancy adiposity measures and changes in insulin sensitivity

We next tested if early pregnancy IHL correlated with gestational changes in insulin sensitivity and found no significant relationships (Supplemental Table 5). Additional exploratory correlational testing between early pregnancy BMI, FM, and VAT and insulin sensitivity measure changes during pregnancy similarly found no relationships.

Discussion

Previous studies of pregnant women examining relationships between body fat distributions and metabolic outcomes utilized techniques that have either not been validated for use during pregnancy (ultrasound) (30; 31) or that can only be used before and after pregnancy due to radiation risk, such as DEXA or CT scans (32). By using MRI and MRS methodologies that are considered safe during gestation, this is the first study to explore relationships between changes in regional and ectopic lipid stores and insulin sensitivity during gestation in mothers with pre-pregnancy body weights ranging from normal to obesity.

As previously reported (4; 6; 8), we found that pregnant women with obesity are more insulin resistant than normal-weight pregnant women, and women in all pregestational weight categories (normal weight, overweight, obese) become more insulin resistant during pregnancy, both in regards to clamp Rd and HISI. However, these relationships exhibit much inter-individual variation. We hypothesized that known relationships between insulin sensitivity and regional fat distribution, independent of total body weight and including visceral and ectopic fat accumulation in liver and muscle, might account for some of this variability within and between individuals.

We found this to be true in early pregnancy, with clamp Rd and HISI correlating with measures of total, regional, and ectopic adiposity, including BMI, percent body fat, fat mass, SAT, VAT, IHL, and IMCL. We further hypothesized that early pregnancy IHL levels would be the strongest predictor of later insulin sensitivity measures. Contrary to our hypotheses, however, by late pregnancy none of the body composition measures were significantly correlated with glucose production (EGD) or disposal (clamp Rd) and none of the changes in regional and ectopic fat during gestation significantly correlated with changes in clamp Rd, EGP, or HISI, including early pregnancy IHL. Interestingly, HISI remained strongly associated with FM, SAT, and VAT in late pregnancy; and despite the increase in EGP in late pregnancy, suppression during hyperinsulinemic conditions was similar to early pregnancy. This may indicate that factors leading to insulin resistance late in pregnancy discussed below may supplant or supersede influences of regional and ectopic adiposity in muscle but not liver.

It is currently thought that placentally-derived factors play major roles in expression of gestational insulin resistance, especially in later pregnancy (33). These include animal studies and the observation of rapid return of insulin sensitivity shortly after delivery of the placenta following birth (34–36), with hypothesized mediators including gestational increases in free cortisol, maternal cytokines, placental miRNAs, placental chorionic gonadotrophin, placental growth hormone, and placental lactogen (37; 38). Our data support that this is not only likely but also that, in late pregnancy, these factors are more potent than expression of normal gestational adiposity changes in mediating this effect, especially for glucose disposal. Even FFA levels, which have experimental support to directly impair insulin sensitivity during pregnancy (39), demonstrated inverse relationships with clamp Rd and HISI in early pregnancy, but not in late gestation; although of all the tested factors correlating change during pregnancy, only change in FFA levels from early to late pregnancy correlated with the change in insulin sensitivity (clamp Rd). Further understanding of these factors and how they align (or not) with current models of insulin resistance involving altered adiponectin signaling or abnormal fatty acid metabolism in the liver and muscle merit further investigation.

In our study, women with pre-pregnancy obesity had the greatest amounts of total and regional fat measures and the highest FFA levels in early pregnancy. Predictably, they also had the lowest insulin sensitivity in early pregnancy. Interestingly, women in the normal pre-pregnancy BMI category experienced the greatest gestational weight gain while the women with pre-pregnancy obesity experienced the least. In fact, many in the pre-pregnancy obesity group experienced gestational losses of fat mass, SAT, VAT, IMCL, and EMCL as well as declines in FFA levels. As a group, they also experienced the least proportional decline in insulin sensitivity measures during pregnancy compared to the other pre-pregnancy weight groups. The preservation of these expected proportional relationships between body composition (and FFA) and insulin sensitivity measures observed between the BMI weight categories suggests that adiposity changes may still play a role, albeit small, in influencing late-pregnancy glucose metabolism. Regardless, the greater gestational weight gain and decline in insulin sensitivity meant that women with pre-pregnancy normal weight “caught up” in late pregnancy compared to women with pre-pregnancy obesity and had similar measures of both clamp Rd and HISI. It is possible that, compared to pre-gestationally normal weight women, women with obesity may lack to capacity to respond (or have reached a state of maximal responsiveness) to the physiologic forces governing energy balance, gestational weight gain, and insulin resistance in late pregnancy.

A major limitation to our study was the smaller numbers of pregnant women in the overweight and obesity categories compared to normal weight category, making our findings in the women with obesity more exploratory than conclusive. Nevertheless, the consistency in proportional and directional changes make these observations compelling. Another limitation is that independent contributions to insulin sensitivity from the mother and fetus were not quantified. This is especially relevant in late pregnancy when the enlarging fetoplacental unit significantly contributes to (non-insulin mediated) glucose disposal (40), potentially confounding maternal glucose metabolism measurements at this timepoint. Strengths of our study were the use of advanced imaging techniques for quantifying regional and ectopic fat distribution paired with insulin clamp measurements of insulin sensitivity.

In summary, early pregnancy insulin sensitivity measures demonstrate the same relationships to total, regional, ectopic fat measures, and free fatty acid levels as reported in non-pregnant women. In late pregnancy, however, although regional and ectopic fat stores have increased during gestation, they are no longer as potently associated with insulin sensitivity measures, including VAT or IHL levels in early or late pregnancy. The insulin resistance that occurs during later stages of pregnancy appears to be driven primarily by factors other than parameters of regional and ectopic adiposity accumulation, may differentially affect muscle and liver insulin responsiveness, and are relatively less impactful to women with pre-pregnancy obesity compared to mothers who are normal-weight.

Article Highlights.

Why did we undertake this study?

• To understand relationships between total, regional, and ectopic adiposity and insulin resistance (IR) in pregnancy.

What is the specific question(s) we wanted to answer?

• Do abdominal visceral/hepatic fat levels correlate with IR?

• Do these relationships differ by pre-pregnancy BMI weight category?

What did we find?

• Visceral, hepatic, and skeletal muscle fat stores correlate with IR in early, but not late, pregnancy.

• Gestational changes in body composition and IR were less in women with pre-pregnancy obesity than normal weight.

What are the implications of our findings?

• Like non-pregnant women, regional/ectopic fat stores correlate with IR in early pregnancy.

• This is no longer true in late pregnancy, when placental factors likely predominant.

• Better understanding of the glucometabolic physiological differences between pre-pregnancy BMI weight categories is needed.