Prepregnancy Fat Free Mass and Associations to Glucose Metabolism Before and During Pregnancy

Abstract

Objective

Our aim was to quantify the individual contribution of prepregnancy (PP) fat-free mass (FFM), expressed as [PP-FFM index (PP-FFMI) = FFM (kg)/height (m²)], on markers of glucose homeostasis before and during pregnancy.

Methods

Body composition was assessed in 43 women before pregnancy using air-displacement plethysmography. Blood was drawn at PP and gestational weeks ∼8 and 30. Relationships between body composition (independent) variables and glucose homeostasis (dependent) variables were assessed using adjusted correlations and simple and multiple linear regression analyses.

Results

PP-FFMI was the strongest predictor of plasma insulin concentration [squared partial correlation (Pr²) = 17, P = 0.007] and homeostasis model assessment of insulin resistance (HOMA2-IR) (Pr² = 16, P = 0.010). At gestation week 30, PP-FFMI and gestational weight gain (GWG) were the strongest predictors of insulin concentration (PP-FFMI: Pr² = 20, P = 0.010; GWG: Pr² = 12, P = 0.052) and HOMA2-IR (PP-FFMI: Pr² = 19, P = 0.012; GWG: Pr² = 13, P = 0.045). After accounting for PP fat mass index (PP-FMI), PP-FFMI and GWG were independently associated with first-phase insulin response (PP-FFMI: Pr² = 20, P = 0.009; GWG: Pr² = 15, P = 0.025) and second-phase insulin response (PP-FFMI: Pr² = 19, P = 0.011; GWG: Pr² = 17, P = 0.016). PP-FMI was the strongest predictor of an oral glucose tolerance test‒derived estimated metabolic clearance rate of glucose (PP-FMI: Pr² = 14, P = 0.037) and estimated insulin sensitivity index (PP-FMI: Pr² = 13, P = 0.047).

Conclusions

PP-FFMI was a predictor of markers of glucose homeostasis before and during pregnancy. Studies assessing the effect of skeletal muscle quality on metabolic regulation during pregnancy are warranted.

Prepregnancy body composition was measured with displacement plethysmography. Lean and fat compartments were independently associated with markers of glucose homeostasis before and during pregnancy.

Although excessive accumulation of fat mass (FM) is the main characteristic of obesity [body mass index (BMI) ≥30 kg/m²], skeletal muscle (SM) may contribute more to obesity-related metabolic disturbances than earlier acknowledged (1–3).

Greater fat-free mass (FFM), along with its main component SM, is usually associated with better insulin sensitivity (4), but this is not the case in individuals with obesity whose absolute content of FFM may be higher than that of normal-weight individuals (BMI, 18.5 to 24.9 kg/m²) (5). Low-grade chronic inflammation, macrophage infiltration, increased free fatty acid uptake, and increased fat deposition in the SM are some of the obesity-induced changes linked with SM insulin resistance (6, 7). Dysregulation of SM metabolism may be further aggravated with physical inactivity, leading to changes in SM morphology (e.g., fewer type I fibers and less capillary density) and altered secretion of important regulating myokines, such as myostatin, IL-6, and leukemia inhibitory factor (8, 9).

Despite this, very few studies have assessed the association between FFM and markers of glucose homeostasis in woman who are overweight and obese. Krotkiewski et al. (10) were among the first to report that among sedentary women with obesity, elevated abdominal adiposity [waist/hip ratio (WHR) >0.80] and FFM correlated with higher fasting insulin levels. Later, Brochu et al. (3) showed that postmenopausal women with a high visceral fat plus high FFM phenotype were more insulin resistant than women with a high visceral fat plus low FFM phenotype, suggesting that higher FFM is associated with glucose disturbances in this population.

Prospective studies using laboratory-based body-composition techniques to assess the potential contribution of FFM to glucose metabolism in women of reproductive age are largely absent from the literature. However, evidence supports that FFM is actively contributing to cardiovascular health and perinatal outcomes in young women (11–14). Thus, the purpose of the current study was to examine how prepregnancy fat-free mass (PP-FFM) content, measured using air-displacement plethysmography, relates to markers of glucose homeostasis during pregnancy while accounting for PP body fat content, PP body fat distribution, and gestational weight gain (GWG).

Methods

Subjects

Forty-three women recruited before conception for the Growing Life, Optimizing Wellness study (GLOWING; NCT01131117) were included for analysis. The GLOWING study is a longitudinal observational study assessing the effect of maternal programming of offspring metabolism on the development of childhood obesity. Although the GLOWING study included a total of 300 women, only a subset was recruited before conception, and these subjects provided the information reported herein. Fasting blood was sampled within 6 months before conception and at gestational weeks ∼8 (7.6 ± 1.7; range, 4 to 10) and 30. Data reported at the 8-week time point correspond to n = 42 because of lack of plasma for one participant. By gestational week 30, four participants were lost to follow-up and four more were dropped from the study because they did not meet the inclusion criteria (one miscarriage, one BMI <18.5 kg/m², one breastfeeding while pregnant, and one multiple pregnancy).

Participants of the GLOWING study responded to study advertisements, which were distributed in the form of flyers in various locations of Central Arkansas (e.g., physician’s office, health fairs, daycare centers), print ads, and social media, as well as television and radio advertisements. Inclusion criteria for the GLOWING study were PP BMI of 18.5 to 35 kg/m², second parity, singleton pregnancy, ≥21 years old, and conception without assisted fertility treatments. Exclusion criteria included preexisting or ongoing medical conditions, medical complications during pregnancy, medications during pregnancy known to influence fetal growth, active smoking, alcohol consumption in any amount, and being an athlete (defined as being engaged in a professional sports activity). The institutional review board at the University of Arkansas for Medical Sciences approved the study protocol, and all participants gave written informed consent.

Measures

Metabolic variables

Blood was drawn from an antecubital vein (no anticoagulant) following an overnight fast at PP and at ∼8 and 30 weeks of gestation. Also, a 100-g oral glucose tolerance test (OGTT) was performed after an overnight fast at gestational week 30. Blood was sampled at baseline and after 30, 60, 90, and 120 minutes. Plasma glucose and insulin concentrations at the various time points were measured using an RX Daytona^® clinical analyzer (Randox Laboratories–US Limited, Kearneysville, WV).

The updated homeostasis model assessment-2 calculator from the Oxford Centre for Diabetes, Endocrinology and Metabolism was used to estimate homeostasis model assessment of β-cell function (HOMA2-%β) and insulin resistance (HOMA2-IR). Fasting plasma insulin concentration, HOMA2-IR, and HOMA2-%β were used as indirect measurements of insulin sensitivity and β-cell response at the PP, 8-week, and 30-week time points. At 30 weeks of gestation, OGTT-derived measurements of insulin sensitivity and β-cell function were also calculated using the Stumvoll et al. (15) equations and Matsuda and DeFronzo index (16).

The HOMA2-IR and Matsuda index (computed from a 100-g OGTT) have been validated against the euglycemic-hyperinsulinemic clamp in pregnant women (17). HOMA2-%β and the Stumvoll equations have not been validated in pregnant women against the euglycemic clamp. However, they have been used in several pregnancy studies (18–20). The Stumvoll indices were obtained using multiple linear regression analysis to predict direct measurements of insulin sensitivity and β-cell response during a euglycemic clamp.

OGTT-derived measurements of insulin sensitivity and β-cell response

Equation 1. Estimated metabolic clearance rate (MCR) of glucose (mg·kg⁻¹ min⁻¹)

$$18.8-\left(0.271\times \text{BMI}\right)-\left(0.0052\times \text{insulin}\hspace{0.17em}120\text{ min}\right)-\left(0.27\times \text{glucose}\hspace{0.17em}90\text{ min}\right)$$

Equation 2. Estimated insulin sensitivity index [(ISI); µmol·kg⁻¹·min⁻¹·pM⁻¹]

$$0.226-\left(0.0032 \times \text{BMI}\right)-\left(0.0000645\times \text{insulin}\hspace{0.17em}120\text{ min}\right)-(0.00375 \times \text{glucose}\hspace{0.17em}90\text{ min})$$

Equation 3. Matsuda index

$$10000÷\sqrt{(FPG\times FPI)}\times \left(G\times I)\right)$$

where FPG = fasting plasma glucose, FPI = fasting plasma insulin, G = mean of glucose, and I = mean of insulin.

Equation 4. First-phase insulin response (pM)

$$1283+\left(1.829\times \text{insulin}\hspace{0.17em}30\text{ min}\right)-\left(138.7\times \text{glucose}\hspace{0.17em}30\text{ min}\right)+(3.772\times \text{insulin}\hspace{0.17em}0)$$

Equation 5. Second-phase insulin response (pM)

$$287+\left(0.4164\times \text{insulin}\hspace{0.17em}30\text{ min}\right)-\left(26.07\times \text{glucose}\hspace{0.17em}30\text{ min}\right)+(0.9226\times \text{insulin}\hspace{0.17em}0)$$

Body composition

FM (in kilograms), FFM (in kilograms), and total body mass (in kilograms) were assessed using whole-body air-displacement plethysmography (Bod Pod; Cosmed, Concord, CA) for all 43 participants at the PP time point. Because taller individuals have more FFM than shorter subjects (21), the fat-free mass index [(FFMI) = FFM (kg)/height (m²)] was calculated to allow a more adequate comparison among the participants. Fat mass index [(FMI) = FM (kg)/height (m²)] was used as a measure of adiposity.

Anthropometry

Body weight was measured to the nearest 0.1 kg on a calibrated scale (Perspective Enterprises, Portage, MI), and the participant’s height was measured using a standard wall-mounted stadiometer to the nearest 0.1 cm (Tanita Corp., Tokyo, Japan). During the measurements, subjects were asked to wear only a standard hospital gown while in the upright position. Waist circumference (WC) was measured at the midpoint between the lower margin of the last palpable rib and the top of the ileal crest using a stretch-resistance tape with a spring providing constant tension. Prepregnancy WC (PP-WC) was used as the indicator of body fat distribution. All measurements were taken in duplicates and, if not in accordance, in triplicates. In this study, GWG refers to the cumulated weight gain (in kilograms) from PP to each of the studied time points during pregnancy (e.g., gestational weeks 8 and 30).

Physical activity

Physical activity was assessed with the Actical accelerometer (Philips Respironics Co. Inc., Bend, OR) at the time of enrollment (PP). The Actical is a small (2.8 × 2.7 × 1.0 cm; weight, 17 g) accelerometer measuring omnidirectional gross-motor activity. The Actical was placed on the participant’s ankle on the nondominant side. Each participant wore the Actical for 3 to 5 days. The monitor was programmed to record movement activity beginning at 11:59 pm on a given day. The participants were instructed to undertake usual activities while wearing the monitor. To be included in the analyses, each participant needed to record at least 3 valid days of accelerometer data including 1 weekend day.

Sedentary activity was defined as <100 counts/min, light activity as 100 to <1535 counts/min, moderate activity as 1535 to <3962 counts/min, and vigorous activity as ≥3962 counts/min (22, 23). Moderate to vigorous physical activity (MVPA) was defined as counts/min ≥1535. Total daily minutes spent at each level of physical activity were quantified. We also assessed adherence with the World Health Organization’s recommendation of 150 minutes per week of moderate to vigorous physical activity accumulated in at least 10-minute bouts (24). To count as a bout, at least 8 of 10 consecutive minutes had to be above the cut point for moderate physical activity. Bout-accumulated minutes of moderate to vigorous physical activity per week were estimated using the average of all valid days × 7.

Statistical analyses

Data measures in the interval scale are summarized as mean ± SD or median (Q1, Q3), whereas data measures in the ordinal or nominal scale are summarized as percentages and counts. Simple and stepwise multiple linear regression analyses were performed with HOMA2-%β, HOMA2-IR, fasting insulin levels, estimated MCR of glucose, estimated ISI, Matsuda index, estimated first-phase insulin response, and estimated second-phase insulin response as the dependent variables and PP-FFMI, PP-FMI, PP-WC, and GWG as independent variables. PP-FFMI and PP-FMI were entered in step 1 (model 1) of the stepwise multiple regression analysis, GWG was added at step 2 (model 2), and PP-WC was added at step 3 (model 3). Adjusted Pearson correlations were used to assess the associations between anthropometric measurements and markers of glucose homeostasis. There were no statistically significant interactions between FFMI and other independent variables; thus, interaction statements were not retained in the final regression models. The variance of inflation factor method was used to assess multicollinearity among the independent variables. SAS 9.3 software (SAS Institute Inc., Cary, NC) was used for the statistical analyses.

Results

Participant characteristics are described in Table 1. Participants were sedentary and did not meet the World Health Organization’s recommendation of weekly physical activity. Overall, they did 58 (18, 96) minutes per week of moderate to vigorous physical activity, accumulated in bouts of at least 10 minutes.

Table 1.

Prepregnancy Characteristics of Participants

Variable	n = 43
Age, y	30.6 ± 3.5
Ethnicity, n (%)
Caucasian	40 (93)
Other	3 (7)
Body weight, kg	67.4 ± 10.4
BMI, kg/m2	25.2 ± 4.3
BMI range	18.3–34.4
Height, cm	163.9 ± 6.5
WHR	0.80 ± 0.07
Body composition
FM, kg	24.2 ± 8.6
FM, %	34.6 ± 7.0
FMI, kg/m2	9 ± 3.4
FFM, kg	43.5 ± 4.8
FFMI, kg FFM/m2	16.2 ± 1.60
Fasting insulin level, uIU/mL	10.5 ± 4.3
Physical activity
Total minutes per day
Sedentary	1079.4 ± 86.7
Light	293.2 ± 72.4
Moderate	56.8 ± 20.5
Vigorous	13.0 ± 8.3
MVPA, min/wk (10-min bouts)	58 (18, 96)

Values are means ± SD, % and counts, and median (Q1, Q3).

Abbreviations: FM, %, percentage of fat mass; MVPA, moderate to vigorous physical activity per week accumulated in at least 10-min bouts.

Simple linear regression analyses

Table 2 shows results obtained when markers of glucose metabolism (fasting insulin concentrations, HOMA2-IR, HOMA2-%β, estimated MCR of glucose, estimated ISI, Matsuda index, and first- and second-phase insulin responses) were regressed on participants’ PP-FFMI, PP-FMI, GWG, and WC. At PP, fasting insulin concentrations, HOMA2-IR, and HOMA2-%β were significantly associated with PP-FFMI, PP-FMI, and PP-WC. At gestational week 8, fasting insulin concentrations and HOMA2-IR were significantly related to PP-FFMI and PP-FMI and marginally associated with WC (P = 0.05). HOMA2-%β was significantly related to PP-FFMI only. Finally, at gestational week 30, fasting insulin concentrations, HOMA2-IR, HOMA2-%β, estimated MCR of glucose, estimated ISI, and Matsuda index were significantly related to PP-FFMI, PP-FMI, and PP-WC. First- and second-phase insulin responses were significantly related to PP-FFMI and PP-WC. As shown in Table 2, the coefficients of determination (R²) for PP-FFMI were higher than their corresponding value for PP-FMI for all variables except for estimated MCR of glucose and estimated ISI at gestational week 30.

Table 2.

Linear Regression Analyses Assessing the Relationships Between FFMI, FMI, GWG, and WC With Fasting Insulin Levels, HOMA2-IR, HOMA2-%β, and OGTT-Derived Measurements of Insulin Sensitivity and β-Cell Response Before Pregnancy and at Gestational Weeks 8 and 30

	FFMI			FMI			GWG			WC
	β	R 2	P	β	R 2	P	β	R 2	P	β	R 2	P
Prepregnancya
Fasting insulin	0.14	0.30	<0.001	0.05	0.21	0.002	—	—	—	0.02	0.22	0.001
HOMA2-IR	0.14	0.29	<0.001	0.05	0.19	0.004	—	—	—	0.02	0.21	0.002
HOMA2-%β	0.10	0.20	0.003	0.10	0.20	0.003	—	—	—	0.02	0.18	0.004
8 Weeks
Fasting insulin	1.22	0.18	0.005	0.56	0.14	0.007	0.58	0.04	0.222	0.20	0.09	0.052
HOMA2-IR	0.16	0.17	0.007	0.07	0.14	0.008	0.08	0.04	0.206	0.03	0.09	0.051
HOMA2-%β	9.15	0.29	<0.001	94.9	0.06	0.104	1.32	0.005	0.641	1.07	0.07	0.084
30 Weeks
Fasting insulin	1.71	0.32	<0.001	0.42	0.12	0.040	0.16	0.02	0.482	0.21	0.20	0.007
HOMA2-IR	0.21	0.31	<0.001	0.06	0.12	0.041	0.02	0.02	0.447	0.03	0.19	0.008
HOMA2-%β	16.46	0.39	<0.0001	4.13	0.12	0.040	0.31	0.00	0.870	2.09	0.26	0.002
MCR of glucose	−0.71	0.33	<0.001	−0.37	0.45	<0.0001	0.04	0.01	0.639	−0.13	0.45	<0.0001
ISI	−0.01	0.31	<0.001	−0.004	0.42	<0.0001	0.00	0.00	0.702	−0.002	0.43	<0.0001
Matsuda indexa	−0.12	0.18	0.01	−0.04	0.12	0.046	−0.02	0.04	0.243	−0.019	0.18	0.011
First-phase insulin response	99.97	0.22	0.005	22.37	0.06	0.173	19.94	0.05	0.216	14.63	0.19	0.009
Second-phase insulin response	23.00	0.22	0.005	5.41	0.06	0.153	4.96	0.053	0.182	3.37	0.19	0.009

Bolded variables indicate significant (P < 0.05) associations.

^aDependent variables have been log transformed.

Stepwise multiple linear regression analyses

There was a strong correlation between FMI and WC (r = 0.82; P < 0.0001). Overall, adding WC to the models did not increase total adjusted R-squared (adj-R²)², whereas variance of inflation factor values increased by approximately twofold for FMI and WC (Tables 3 and 4). At the PP time point, after adjustments for all other variables (Table 3, model 2), PP-FFMI was the strongest predictor of insulin levels and HOMA2-IR, explaining 17% (P = 0.007) and 16% (P = 0.010) of the observed variance, respectively.

Table 3.

Results of Stepwise Multiple Regression Analyses Regarding FFMI, FMI, GWG, and WC as Predictors of Insulin Levels, HOMA2-%β, and HOMA2-IR Before Pregnancy and at Gestational Weeks 8 and 30

Variable	FFMI		VIF	FMI		VIF	GWG		VIF	WC		VIF	Adj-R2	P Value
	Pr2	P Value		Pr2	P Value		Pr2	P Value		Pr2	P Value
Prepregnancy
Insulin levels
Model 1	0.207	0.002	1.613	0.004	0.707	1.613							0.295	<0.001
Model 2	0.170	0.007	1.839	0.000	0.942	3.146				0.006	0.627	3.486	0.281	0.001
HOMA2-IR
Model 1	0.195	0.004	1.613	0.001	0.814	1.613							0.265	<0.001
Model 2	0.158	0.010	1.839	0.001	0.855	3.146				0.006	0.617	3.486	0.251	0.003
HOMA2-%β
Model 1	0.043	0.185	1.613	0.041	0.197	1.613							0.147	0.016
Model 2	0.042	0.201	1.839	0.027	0.308	3.146				0.000	0.881	3.486	0.125	0.042
8 Weeks
Insulin levels
Model 1	0.061	0.121	1.583	0.045	0.185	1.583							0.179	0.008
Model 2	0.046	0.185	1.645	0.054	0.150	1.607	0.030	0.288	1.039				0.183	0.014
Model 3	0.027	0.314	1.860	0.008	0.579	3.116	0.029	0.300	1.040	0.013	0.487	3.419	0.172	0.026
HOMA2-IR
Model 1	0.050	0.162	1.583	0.047	0.174	1.583							0.164	0.011
Model 2	0.035	0.244	1.645	0.057	0.138	1.607	0.033	0.257	1.039				0.171	0.017
Model 3	0.019	0.402	1.860	0.008	0.583	3.116	0.033	0.268	1.040	0.015	0.452	3.419	0.162	0.032
HOMA2-%β
Model 1	0.167	0.008	1.583	0.011	0.516	1.583							0.262	0.001
Model 2	0.162	0.010	1.645	0.011	0.523	1.645	0.000	0.989	1.040				0.243	0.004
Model 3	0.146	0.015	1.860	0.008	0.592	3.116	0.000	0.987	1.040	0.000	0.940	3.419	0.222	0.009
30 Weeks
Insulin levels
Model 1	0.227	0.004	1.771	0.002	0.830	1.771							0.280	0.002
Model 2	0.241	0.004	1.774	0.012	0.550	2.070	0.130	0.040	1.260				0.353	0.001
Model 3	0.200	0.010	2.210	0.005	0.711	3.000	0.119	0.053	1.365	0.000	0.940	4.154	0.331	0.003
HOMA2-IR
Model 1	0.214	0.006	1.771	0.001	0.867	1.771							0.266	0.003
Model 2	0.229	0.005	1.774	0.015	0.504	2.069	0.137	0.034	1.260				0.346	0.001
Model 3	0.191	0.012	2.210	0.007	0.647	3.996	0.127	0.045	1.365	0.000	0.982	4.154	0.324	0.003
HOMA2-%β
Model 1	0.313	0.001	1.771	0.011	0.562	1.771							0.360	<0.001
Model 2	0.318	0.001	1.774	0.000	0.981	2.069	0.058	0.177	1.260				0.378	<0.001
Model 3	0.232	0.005	2.210	0.013	0.536	3.996	0.038	0.288	1.365	0.025	0.388	4.154	0.373	0.001

Bolded variables indicate significant (P < 0.05) associations.

Abbreviations: Pr², squared partial correlation; VIF, variance of inflation factor.

Table 4.

Results of Stepwise Regression Analyses Regarding FFMI, FMI, GWG, and WC as Predictors of Estimated MCR of Glucose and First- and Second-Phase Insulin Responses at Gestational Week 30

Variable	FFMI		VIF	FMI		VIF	GWG		VIF	WC		VIF	Adj-R2	P Value
	Pr2	P Value		Pr2	P Value		Pr2	P Value		Pr2	P Value
MCR of glucose
Model 1	0.057	0.174	1.771	0.226	0.004	1.771							0.451	<0.0001
Model 2	0.057	0.182	1.774	0.307	0.001	2.069	0.113	0.056	1.260				0.497	<0.0001
Model 3	0.027	0.368	2.210	0.137	0.037	4.154	0.088	0.098	1.365	0.015	0.506	4.154	0.488	<0.0001
ISI
Model 1	0.049	0.208	1.771	0.207	0.007	1.771							0.418	<0.0001
Model 2	0.048	0.219	1.774	0.288	0.001	2.069	0.112	0.056	1.260				0.466	<0.0001
Model 3	0.022	0.416	2.210	0.126	0.047	4.000	0.090	0.098	1.365				0.456	<0.0001
Matsuda index
Model 1	−0.084	0.097	1.771	0.007	0.636	1.771							0.139	0.035
Model 2	0.090	0.090	1.774	0.067	0.145	2.069	0.186	0.012	1.260				0.277	0.004
Model 3	0.062	0.169	2.210	0.024	0.401	4.000	0.166	0.021	1.365	0.003	0.767		0.255	0.011
First-phase insulin response
Model 1	0.184	0.011	1.771	0.012	0.531	1.771							0.181	0.015
Model 2	0.198	0.009	1.774	0.002	0.790	2.069	0.152	0.025	1.260				0.283	0.004
Model 3	0.111	0.063	2.210	0.016	0.487	3.996	0.111	0.063	1.365	0.053	0.205	4.153	0.299	0.005
Second-phase insulin response
Model 1	0.175	0.014	1.771	0.009	0.603	1.771							0.177	0.017
Model 2	0.191	0.011	1.774	0.006	0.661	2.069	0.174	0.016	1.260				0.298	0.003
Model 3	0.110	0.064	2.210	0.008	0.636	3.996	0.133	0.040	1.365	0.042	0.259	4.154	0.305	0.005

Bolded variables indicate significant (P < 0.05) associations.

Abbreviations: Pr², squared partial correlation; VIF, variance of inflation factor.

At ∼8 weeks of gestation, no significance was found in the variables included in models 1 to 3 for fasting insulin concentrations and HOMA2-IR. PP-FFMI was the strongest predictor of HOMA2-%β, accounting for ∼15% (P = 0.015) of the observed variance after adjustments for PP-FMI, GWG, and WC (model 3). At gestational week 30 (Table 3), after adjustments for all other variables (model 3), PP-FFMI explained 20% (P = 0.010), 19% (P = 0.012), and 23% (P = 0.005) of the variance in fasting insulin levels, HOMA2-IR, and HOMA2-%β, respectively.

Table 4 shows the results of stepwise multiple regression analyses from OGTT-derived insulin sensitivity and insulin secretion indices measured using the Stumvoll and Matsuda equations (15, 16). PP-FMI was the strongest predictor of the estimated MCR of glucose and estimated ISI, even after accounting for WC, and it explained 14% and 13% of the variance, respectively (model 3). GWG accounted for 11% of the variance in the estimated MCR of glucose and estimated ISI (P = 0.056, model 2). Interestingly, GWG was the strongest predictor of the Matsuda index in models 2 and 3 (P < 0.05).

Finally, PP-FFMI and GWG were the strongest predictors of β-cell response and explained 20% (P = 0.009) and 15% (P = 0.025) of the variance in the first-phase insulin response and 19% (P = 0.011) and 17% (P = 0.016) of the variance in the second-phase insulin response when FMI was accounted for (model 2). Adding WC (model 3) did not contribute to the overall adj-R² (model 2 adj-R² = 0.283 vs model 3 adj-R²= 0.299) and led to loss of power. Still, FFMI and GWG each explained ∼11% of the variance in first-phase insulin response (P = 0.063), whereas they explained 11% (P = 0.064) and 13% (P = 0.040), respectively, of the variance of second-phase insulin response.

Adjusted correlations

Table 5 presents adjusted Pearson correlations between PP measurements (FFMI, FMI, and WC) and GWG with markers of glucose homeostasis. At PP, FFMI was positively correlated with fasting insulin levels (adj-r = 0.41; P = 0.007) and HOMA2-IR values (adj-r = 0.40; P = 0.010). At gestational week 8, PP-FFMI was the variable with the strongest correlation with HOMA2-%β (adj-r = 0.39; P = 0.015). Finally, at 30 weeks of gestation, PP-FFMI correlated with fasting insulin level (PP-FFMI: adj-r = 0.45; P = 0.010), whereas GWG was marginally correlated with fasting insulin level (adj-r = 34; P = 0.053). HOMA2-IR correlated with PP-FFMI (adj-r = 0.44; P = 0.012) and GWG (adj-r = 0.35; P = 0.045). When adjusted for GWG, FMI, and WC, PP-FFMI marginally correlated with first- and second-phase insulin responses (adj-r = 0.33; P = 0.063).

Table 5.

Adjusted Correlations Between Anthropometric Measurements and Variables of Glucose Homeostasis

Variable	FFMI	FMI	GWG	WC
Prepregnancy
Fasting insulin	0.42 a	−0.01		0.08
HOMA2-IR	0.40 b	−0.03		−0.02
HOMA2-%β	0.20	0.16		0.08
8 weeks
Fasting insulin	0.17	0.59	0.17	0.11
HOMA2-IR	0.14	0.09	0.18	0.12
HOMA2-%β	0.39 b	0.09	0.00	−0.02
30 weeks
Fasting insulin	0.45 b	0.07	0.34 c	0.01
HOMA2-IR	0.44 b	0.08	0.35 b	0.00
HOMA2-%β	0.48 a	−0.11	0.19	0.16
MCR of glucose	−0.16	−0.37b	−0.30	−0.12
ISI	−0.15	−0.35b	−0.30	−0.12
Matsuda index	−0.23	−0.16	−0.40b	−0.06
First-phase insulin response	0.33 d	−0.13	0.33 d	0.23
Second-phase insulin response	0.33 d	−0.09	0.36 b	0.21

Correlations adjusted among anthropometric variables. Bolded variables indicate significant (P < 0.05) associations.

^a P < 0.01.

^b P < 0.05.

^c P = 0.05.

^d P = 0.06.

Discussion

The aim of the current study was to assess the relationship between PP-FFM, expressed as PP-FFMI, and indirect markers of insulin sensitivity before and during pregnancy while accounting for the effect of PP adiposity (FMI), PP body fat distribution (WC), and GWG. The major finding in this group of sedentary women was that PP-FFMI independently correlated with fasting insulin levels and HOMA2-IR values before and during pregnancy. In addition, at gestational week 30, both GWG and PP-FFMI were independently associated with increasing first- and second-phase insulin responses when FMI was accounted for. Finally, PP-FMI was the strongest predictor of a lower estimated MCR of glucose and estimated ISI, whereas GWG was the strongest variable associated with the Matsuda index.

Multiple indirect methods are available to assess insulin sensitivity/resistance. They vary in complexity of measurement and the physiological components that inform them. For instance, HOMA-IR reflects the effects of insulin on hepatic glucose production during the postabsorptive steady state. However, because SM sensitivity correlates with that of the liver in most cases, it is assumed that HOMA-IR is also a reflection of peripheral insulin sensitivity during fasting conditions (25). OGTT-derived methods incorporate glucose and insulin levels following a glucose load. The Matsuda index, for instance, is used as a marker of whole-body insulin sensitivity. It combines a reduced formula of homeostasis model assessment as an indicator of hepatic insulin sensitivity with the mean of glucose and insulin levels during the OGTT as an indicator of peripheral insulin sensitivity (16). On the other hand, the Stumvoll indices are derived from multiple regression analyses and are used to predict markers of insulin sensitivity and β-cell function as measured during the euglycemic-hyperinsulinemic clamp (15). In this context, interpretation of different indirect measurements of insulin sensitivity must be done, considering their physiological basis, strengths, and limitations.

It is well recognized that changes in FFM are directly proportional to changes in total body weight (26). In agreement with this, our data showed a positive correlation between total body weight and FFM (r = 0.73; P < 0.0001; data not shown). Greater muscle mass in subjects with obesity has been regarded not only as an adaptive response to the stress imposed by excessive weight on bodily structures but also as a necessary response to preserve functional capacity (27). However, some studies report that greater FFM in elderly people with obesity may be a risk factor for insulin resistance, cardiovascular disease, and overall mortality (3, 28–30).

In earlier studies, Krotkiewski et al. (10) reported that in women with obesity, WHR—an indicator of increased abdominal adiposity—and FFM (in kilograms) were independently associated with higher fasting insulin levels. Similarly, our results show a positive association between PP-FFMI and markers of insulin resistance before pregnancy and up to 30 weeks of gestation. In the current study, WC was used as an indicator of body fat distribution, as studies have shown WC to be more sensitive than WHR for identifying abdominal obesity (31). WC did not associate with any marker of insulin sensitivity when multiple regression was conducted. In fact, when added in stepwise analysis, WC did not contribute to the overall model for most of the studied variables. The latter may relate to the collinearity between FMI and WC; most likely, more accurate measurements of visceral adiposity are required to assess the independent effects of visceral vs subcutaneous fat.

As expected, at gestational week ∼8, GWG did not have any effect on the measured outcomes, as mean weight gain at this time point was only 0.90 ± 1.56 kg. FFMI and FMI were not associated with fasting insulin level and HOMA2-IR. However, it is worth noting that early in pregnancy, blood was collected any time between 4 and 10 weeks of gestation, which may have increased the variability of measured outcomes (e.g., fasting insulin level and HOMA2-IR). Despite this, FFMI was positively associated with HOMA2-%β in all three regression models.

Given the collinearity between FM and FFM, research efforts have routinely focused on the effects of excess adiposity on maternal-offspring health, thus disregarding a role for the FFM compartment. A discussion has emerged around how the increasing prevalence of obesity, which also parallels the adoption of sedentary behaviors in modern societies (32), may be linked to changes in the role of SM regarding health outcomes in different populations, including women and their offspring (11–14, 33). For instance, studies show that FFM may be an important contributor to the risk of cardiovascular disease (CVD) in adolescent girls and to perinatal outcomes in women of reproductive age. Gracia-Marco et al. (11) measured the association between clustered CVD risk with body composition indices in ∼15-year-old boys and girls. The clustered CVD risk score incorporated systolic blood pressure, aerobic capacity, HOMA-IR, C-reactive protein, total cholesterol to high-density lipoprotein cholesterol, and triglyceride levels. The authors found independent effects of both adiposity and lean mass on clustered CVD risk in girls but not in boys, for whom adiposity, but not lean mass, was the main contributor to CVD risk. In line with this research, at least three studies have shown greater maternal FFM as an important predictor of offspring birth weight (12–14). The mechanisms involved in these findings are not known, but the results point to the importance of tissue quality. Also, this evidence highlights the limitation of BMI in discerning between the relative contributions of the FM and FFM compartments to specific health outcomes.

After we accounted for PP-FMI, our observation that PP-FFMI contributed to markers of glucose homeostasis not only during pregnancy but also before conception is relevant and leads to questions about the role of SM quality and function in sedentary women who are overweight and obese. In a previous study, Simoneau et al. (34) reported that midthigh muscle areas displaying lower CT attenuation values, a marker of fat deposition, increased in proportion to weight in women who were obese. Citrate synthase activity and peripheral disposal rates of glucose negatively correlated with increasing areas of low attenuation. This suggests that impaired substrate oxidation of the SM may contribute to regional fat deposition and, consequently, insulin resistance. The two-compartment model used to assess body composition in the current study is limited by the assumption of uniformity of the lean tissue. Thus, correlations between fat deposition within the lean compartment and markers of glucose homeostasis could not be assessed.

We also determined insulin sensitivity and insulin secretion at gestational week 30 using OGTT-derived methods, which have been validated against the gold standard method, the clamp technique (15, 16). In postmenopausal women with obesity, greater FFMI has been suggested as an independent correlate of decreased disposal rate of glucose (3). Brochu et al. (3) compared women with similar visceral fat content but different FFMIs. Interestingly, women with a phenotype of high visceral fat combined with high FFMI had a 39% higher concentration of insulin than women with a phenotype of high visceral fat plus low FFMI.

Similarly, our study showed an independent association between PP-FFMI and fasting insulin levels and homeostasis model assessment-2 indices. However, at gestational week 30, PP-FMI was the strongest predictor of the estimated MCR of glucose and estimated ISI (Stumvoll indices), whereas GWG was the strongest predictor of the Matsuda index. We did not find associations with FMI when HOMA2-IR was used as an index of peripheral insulin sensitivity. Catalano and Kirwan (17) compared HOMA-IR with the clamp technique before and during pregnancy and found it was an acceptable predictor of insulin sensitivity in this population. It may be that when the complex relationship between body compartments and glucose metabolism is assessed, a more robust tool to quantify peripheral insulin sensitivity is needed.

After accounting for FMI, both PP-FFMI and GWG were independently associated with higher early and late insulin responses at gestational week 30. This observation may be related to the inherent effect of pregnancy. Normally, as pregnancy progresses, SM becomes more resilient to the effects of insulin, and greater insulin secretion is required to preserve normal transport of glucose into the muscle. This compensatory response may be greater in pregnant women who are overweight and obese and who have an underlying state of insulin resistance (35). However, our finding that greater PP-FFMI in sedentary women is associated with increased early and late insulin responses later in pregnancy may also relate to unique morphological and metabolic characteristics of the SM, such as fiber types, capillary density, oxidative capacity, and lipid content (8, 36–38). Future studies are needed to better understand the role of FFM in insulin responses during an OGTT in pregnancy.

In nonpregnant women with obesity, exercise training without caloric restriction increases capillary density and the proportion of type IIa fibers relative to IIb fibers, whereas it decreases plasma insulin levels (36, 39). More importantly, improvement of SM quality and glucose homeostasis with physical training can be achieved without changes in total body weight or body composition (36, 40, 41). van Poppel et al. (42) reported that women at risk for gestational diabetes who spent ∼30 min/d in moderate to vigorous physical activity early in pregnancy had better insulin sensitivity at gestational week 32, as demonstrated by lower first- and second-phase insulin responses. More recently, the authors suggested that this improvement may relate to higher levels of muscle-derived IL-6, which is an important regulator of insulin-stimulated glucose uptake in the context of increased physical activity (9, 20). Associations between SM quality, myokine profile, and metabolic outcomes during pregnancy have not yet been defined. Of relevance, conditioning of the SM with a structured physical training program starting before or early in pregnancy could potentially improve carbohydrate metabolism, thus making the effects of PP-FFMI on glucose homeostasis a modifiable risk factor. Participants of the current study were sedentary. We did not find an effect of sedentary and moderate to vigorous physical activity on any of the measured outcomes when these variables were incorporated in regression models (data not shown). The authors believe a broader range of physical activity levels among participants may be required to identify an effect.

Our study is strengthened by the use of direct measurements of body composition before conception. Although BMI is widely used as a surrogate of adiposity, it fails to differentiate between the relative contribution of the fat and lean compartments (43). Also, physical activity was objectively measured using accelerometry. However, the authors acknowledge that more days of physical activity recordings, including both weekend days, may be necessary to reliably capture habitual physical activity. Study limitations include the lack of direct measurements of insulin sensitivity. Also, OGTT-derived indices were measured at gestational week 30 only; thus, we were unable to assess the effects of FFMI on these parameters in the pregravid state. Finally, air-displacement plethysmography assumes uniformity of the density of lean tissue (1.1 kg/L) (44) and does not provide information on fat deposition within the lean compartment, which has been shown to correlate with insulin resistance (7, 34). Although overestimation of FFM can occur in states of increased fluid retention, our participants were young women with no comorbidities. Body composition assessment was completed before conception, ruling out the possibility of pregnancy-induced fluid overload.

In recent years, studies have shown that SM is an organ with endocrine functions that allow cross communication with other organs, such as the liver, pancreas, and adipose tissue (9, 45). Because the release of some peptides from the SM is dependent upon contraction (9), physical inactivity, particularly in populations with obesity, could also contribute to metabolic dysregulation of the SM. It is also possible that adipokines released from the adipose tissue may affect SM function (46).

We have shown that PP-FFMI, as measured by air-displacement plethysmography, is an independent predictor of increased markers of insulin resistance in sedentary women of childbearing age. Assessment of SM mass and lipid content within the SM with more accurate methods (e.g., magnetic resonance imaging) is warranted to validate our results. Also, future studies of the mechanisms underlying our findings are needed, including studies of SM tissue samples to determine muscle quality.

Glossary

Abbreviations:

BMI

body mass index

CVD

cardiovascular disease

FFM

fat-free mass

FFMI

fat-free mass index

FM

fat mass

GLOWING

Growing Life, Optimizing Wellness

GWG

gestational weight gain

HOMA2-%β

homeostasis model assessment of β-cell function

HOMA2-IR

homeostasis model assessment of insulin resistance

ISI

insulin sensitivity index

MCR

metabolic clearance rate

OGTT

oral glucose tolerance test

PP

prepregnancy

PP-FFMI

prepregnancy fat-free mass index

PP-FMI

prepregnancy fat mass index

PP-WC

prepregnancy waist circumference

Pr²

squared partial correlation

SM

skeletal muscle

WC

waist circumference

WHR

waist/hip ratio