More appendicular lean mass relative to body mass index is associated with lower incident diabetes in middle-aged adults in the CARDIA study

Abstract

Background and Aims

Although lower lean mass is associated with greater diabetes prevalence in cross-sectional studies, prospective data specifically in middle-aged Black and White adults are lacking. Relative appendicular lean mass (ALM), such as ALM adjusted for body mass index BMI, is important to consider since muscle mass is associated with overall body size. We investigated whether ALM/BMI is associated with incident type 2 diabetes in the Coronary Artery Risk Development in Young Adults study.

Methods and Results

1,893 middle-aged adults (55% women) were included. ALM was measured by DXA in 2005–06. Incident type 2 diabetes was defined in 2010–11 or 2015–16 as fasting glucose ≥7 mmol/L(126 mg/dL), 2-hour glucose on OGTT ≥11.1 mmol/L(200 mg/dL)(2010–11 only), HbA_1C ≥48 mmol/mol(6.5%)(2010–11 only), or glucose-lowering medications. Cox regression models with sex stratification were performed.

In men and women, ALM/BMI was 1.07 ± 0.14(mean±SD) and 0.73±0.12, respectively. Seventy men(8.2%) and 71 women(6.8%) developed type 2 diabetes. Per sex-specific SD higher ALM/BMI, unadjusted diabetes risk was lower by 21% in men[HR 0.79 (0.62–0.99), p=0.04] and 29% in women[HR 0.71 (0.55–0.91), p=0.008]. After adjusting for age, race, smoking, education, physical activity, and waist circumference, the association was no longer significant. Adjustment for waist circumference accounted for the attenuation in men.

Conclusion

Although more appendicular lean mass relative to BMI is associated with lower incident type 2 diabetes in middle-aged men and women over 10 years, its effect may be through other metabolic risk factors such as waist circumference, which is a correlate of abdominal fat mass.

Introduction

Type 2 diabetes mellitus is a growing public health problem associated with significant morbidity, mortality, and health care expenditures [1, 2]. BMI has been the preferred anthropometric variable to identify individuals at increased risk for type 2 diabetes. Although there has been intense focus on the deleterious effects of excess adiposity, less is known about the possible protective effects of greater muscle mass, on type 2 diabetes risk, especially in middle-aged adults. Since skeletal muscle is metabolically active and responsible for the majority of the body’s postprandial glucose disposal [3], relatively more muscle mass in middle age may protect against the development of type 2 diabetes.

Lean mass, which includes muscle, skin, and connective tissue, can be measured by dual energy x-ray absorptiometry (DXA) and bioelectrical impedance analysis (BIA). Appendicular lean mass, i.e. the sum of the lean mass of the 4 extremities, by DXA is highly correlated with both magnetic resonance imaging (MRI) and computed tomography (CT) measures of skeletal muscle volume [4, 5]. More lean mass by DXA or BIA, as well as higher muscle cross-sectional area by CT, relative to total body weight, height, or BMI has been associated with less insulin resistance and prevalent type 2 diabetes in cross-sectional studies of adults across the age spectrum [6–12]. Prospective studies have demonstrated that more lean mass and higher muscle cross-sectional area are associated with lower risk of incident type 2 diabetes in Asian populations of middle-aged and older adults [13–15]. However, prospective studies in other race and ethnic groups have demonstrated discordant results [16–18]; and prospective data in middle-aged (37– 54 years old) adults are lacking. In addition, none of these studies adjusted appendicular lean mass for body mass index (ALM/BMI). This is important to consider since appendicular lean mass is positively associated with overall body mass, and ALM/BMI is the measure of relative muscle mass recommended by the Foundation for the National Institutes of Health Sarcopenia Project (FNIHSP) in other populations, such as older adults [19].

Understanding whether relatively less appendicular lean mass is associated with incident type 2 diabetes in middle-aged adults could inform screening, prevention, and/or treatment strategies to reduce the public health burden of type 2 diabetes. Given known sociodemographic differences in the risk of type 2 diabetes and in body composition, it is important to explore these relationships in diverse populations [20–23].

The objective of this study was to investigate whether ALM/BMI is associated with incident type 2 diabetes in Black and White middle-aged adults in the Coronary Artery Risk Development in Young Adults (CARDIA) cohort, and to investigate whether this association varies by sex and race. Our primary hypothesis is that higher ALM/BMI, which reflects higher relative appendicular lean mass to overall body weight for height, is associated with lower incident type 2 diabetes. Further, we studied whether the association persisted following adjustment for health behaviors and cardiometabolic health.

Methods

Study Population

The CARDIA study is an ongoing multicenter longitudinal cohort study. In 1985–86, Black and White men and women aged 18–30 years were enrolled from four field centers: Birmingham, AL; Oakland, CA; Chicago, IL; and Minneapolis, MN. Participants were examined 2, 5, 7, 10, 15, 20, 25, and 30 years after enrollment, with retention rates of 71% of living participants at the 30-year follow-up. The details of the study design have been previously published [24]. Participants underwent whole body DXA for assessment of body composition at year 20 in 2005–06, baseline for this analysis. All data were collected according to standardized protocols across all exams. Participants provided written consent and the institutional review boards at each center approved the study.

From the 5,115 CARDIA participants, 2,704 had completed DXA scans. Among these, we excluded 809 participants with the following conditions at 2005–06: known diabetes (as defined below), breastfeeding, heart failure, HIV, cancer, cirrhosis, renal failure, history of bariatric surgery, or no follow-up data after 2005–06. These medical conditions or chronic diseases were excluded due to their potential effect on appendicular lean mass. This yielded a final sample of 851 men and 1,042 women.

For anthropometric assessment, participants wore light clothes and no shoes. Body weight was measured by a balance-beam scale, and height was measured using a vertically mounted metal centimeter ruler and carpenter’s square. BMI was calculated as weight (kg) divided by height in square meters (m²). Waist circumference was measured laterally midway between the iliac crest and the lowest portion of the rib cage and anteriorly midway between the xyphoid process of the sternum and the umbilicus.

ALM was defined as the total non-fat, non-bone mass of both arms and legs as measured by DXA (Hologic QDR 4500W, Delphi 11.2, Discovery XP 12.1, Discovery XP 2002; Hologic, Bedford, MA). Quality control and calibration processes have been previously published [25]. The primary exposure was ALM adjusted for BMI (ALM/BMI), which is the measurement of relative muscle mass recommended by the FNIHSP in other populations, such as older adults [19].

For laboratory assays, participants were asked to fast overnight for at least 8 hours. Incident type 2 diabetes, the primary endpoint, occurred if any of the following were met in 2010–11 or 2015–16 among persons without diabetes in 2005–06: fasting glucose ≥7 mmol/L(126 mg/dL), 2-hour glucose on OGTT ≥11.1 mmol/L(200 mg/dL)(2010–11 only), HbA_1C ≥48 mmol/mol(6.5%)(2010–11 only), or use of glucose-lowering medications. Glucose levels at a visit where participants were pregnant or fasted <8 hours before examination were designated as missing.

Self-reported age, sex, race, and education were collected using standardized questionnaires. Smoking was defined as never using any tobacco products or smoking <5 cigarettes per week for <3 months in the past (never smokers), smoking ≥5 cigarettes a week for ≥3 months in the past (former smokers), or smoking ≥5 cigarettes a week ≥3 months currently (current smokers) using the 2005–06 CARDIA study exam tobacco forms. Physical activity data were collected using the CARDIA Physical Activity History, which queries the frequency, duration, and intensity of 13 different activities during the previous year. A continuous total physical activity intensity score was derived as previously published [26]. Physical activity data were used because it is easy to collect in the clinical setting and therefore more clinically relevant.

All analyses were performed using SAS version 9.4. Study baseline was defined as 2005–06, which are the years in which participants in the ancillary CARDIA Fitness Study underwent DXA for assessment of body composition. Baseline characteristics are reported as mean ± SD for continuous variables and count (percent) for categorical variables. Follow up duration was calculated between the baseline and the earliest time of the following events: incident type 2 diabetes diagnosis, death, and the last known visit.

The proportional hazards assumption was tested and verified by visually inspecting log negative log (log-log) curves to ensure that they do not cross using ALM/BMI as a binary predictor with and without stratifying by sex. We proceeded to a series of Cox regression models: 1) an unadjusted model, 2) a model adjusted for age, race, smoking status, education, physical activity score, and waist circumference (adjusted model), and 3) a model adjusted for age, race, smoking status, education, and physical activity score (adjusted model without waist circumference). These analyses were stratified by sex a priori given marked differences in ALM between men and women. Secondary analyses of the three Cox regression models above were performed with the sexes combined; exposures of ALM/BMI, ALM/weight, ALM, and ALM/height²; and interaction terms for sex and race. Collinearity Diagnostics was performed on the sex-combined, adjusted models with the exposures of ALM/BMI, ALM/weight, ALM, and ALM/height². Hazard ratios per sex-specific SD higher ALM/BMI are reported with a 95% confidence interval and an associated p-value for each model. A two-sided p-value <0.05 was considered statistically significant for the fully adjusted Cox regression model. A two-sided p-value <0.05 was considered statistically significant for the interaction terms.

Results

Characteristics of the 851 men and 1,042 women are presented in Table 1. In men, mean age was 45.0 ± 3.5 (SD) years (range 37–54 years), mean BMI was 28.0 ± 4.3 kg/m² (range 17.2–52.5 kg/m²), and mean ALM/BMI was 1.07 ± 0.14. In women, mean age was 45.2 ± 3.6 years (range 37–52 years), mean BMI was 28.4 ± 6.4 kg/m² (range 17–52 kg/m²), and mean ALM/BMI was 0.73 ± 0.12. Mean follow-up time was 9.3 ± 1.8 years in men and 9.4 ± 1.7 years in women.

Table 1.

Baseline clinical characteristics (mean ± SD)

	Men (n=851)	Women (n=1042)
Age, y	45.0 ± 3.5	45.2 ± 3.6
Black, n (%)	333 (39)	468 (45)
Body mass index (BMI), kg/m2	28.0 ± 4.3	28.4 ± 6.4
Appendicular lean mass (ALM), kg	29.8 ± 4.6	20.4 ± 3.9
ALM/BMI	1.07 ± 0.14	0.73 ± 0.12
Waist circumference, cm	94.3 ± 10.9	85.6 ± 13.4
Physical activity score	428.3 ± 296.7	294.9 ± 241.2
Education, greater than high school, n (%)	839 (81)	655 (77)

Incident type 2 diabetes was diagnosed in 70 men (8.2%) and 71 women (6.8%) overall; 31 White men (6.0%), 39 Black men (11.7%), 23 White women (4.0%) and 48 Black women (10.3%). In the adjusted model with the sexes combined, there was no evidence of interaction of ALM/BMI by sex (p=0.71) nor race (p=0.43), but ALM/BMI and sex were collinear. We proceeded with our a priori decision to stratify by sex. In men, diabetes risk was lower by 21% per sex-specific SD higher ALM/BMI [HR 0.79 (0.62–0.99), p=0.04] (Table 2). After adjusting for age, race, smoking status, education, physical activity score, and waist circumference, the association between ALM/BMI and type 2 diabetes was no longer significant [HR 1.07 (0.82–1.41), p=0.60]. Upon removing waist circumference from the adjusted model, the association between ALM/BMI and type 2 diabetes remained significant—diabetes risk was lower by 27% per SD higher ALM/BMI [HR 0.73 (0.57–0.94), p=0.01].

Table 2.

Association between appendicular lean mass relative to body mass index (ALM/BMI) and incident diabetes

	Unadjusted model		Adjusted model		Adjusted model without waist circumference
	HR (95% CI) per SD increment in exposure	P-value	HR (95% CI) per SD increment in exposure	P-value	HR (95% CI) per SD increment in exposure	P-value
Men
ALM/BMI	0.79 (0.62–0.99)	0.04	1.07 (0.82–1.41)	0.60	0.73 (0.57–0.94)	0.01
Age (years)			0.95 (0.74–1.21)	0.66	0.89 (0.71–1.12)	0.36
Race (Black vs White)			1.81 (1.08–3.03)	0.02	2.15 (1.28–3.62)	0.004
Smoking (current vs never)			0.87 (0.47–1.62)	0.68	0.71 (0.52–1.34)	0.29
Smoking (former vs never)			0.98 (0.49–1.96)	0.95	0.85 (0.43–1.69)	0.64
Education (>high school vs ≤ high school)			0.65 (0.38–1.10)	0.11	0.68 (0.40–1.17)	0.17
Physical activity score			0.89 (0.69–1.15)	0.37	0.87 (0.67–1.12)	0.28
Waist circumference (cm)			2.26 (1.77–2.89)	<0.0001
Women
ALM/BMI	0.71 (0.55–0.91)	0.008	1.10 (0.83–1.46)	0.52	0.79 (0.61–1.03)	0.08
Age (years)			1.21 (0.95–1.53)	0.12	1.21 (0.95–1.55)	0.12
Race (Black vs White)			1.62 (0.94–2.78)	0.08	2.37 (1.40–4.03)	0.001
Smoking (current vs never)			2.32 (1.32–4.07)	0.003	2.26 (1.29–3.93)	0.004
Smoking (former vs never)			1.20 (0.65–2.22)	0.56	1.21 (0.65–2.25)	0.54
Education (>high school vs ≤ high school)			1.04 (0.60–1.79)	0.90	1.02 (0.59–1.75)	0.95
Physical activity score			0.69 (0.50–0.97)	0.03	0.65 (0.46–0.91)	0.01
Waist circumference (cm)			2.17 (1.71–2.76)	<0.0001

In women, diabetes risk was lower by 29% per sex-specific SD higher ALM/BMI [HR 0.71 (0.55–0.91), p=0.008] (Table 2). After adjusting for age, race, smoking status, education, physical activity score, and waist circumference, the association between ALM/BMI and type 2 diabetes was no longer significant (HR 1.10 (0.83–1.46), p=0.52). Upon removing waist circumference from the adjusted model, the association between ALM/BMI and type 2 diabetes remained not significant (HR 0.79 (0.61–1.03), p=0.08).

Secondary analyses with the sexes combined were performed using other traditional measures of relative appendicular lean mass as the exposure. With ALM/BMI as the exposure, the association between ALM/BMI and type 2 diabetes was not significant in the unadjusted model (Supplemental Table 1). When sex was added to the model, diabetes risk was lower by 35% per SD higher ALM/BMI [HR 0.65 (0.49–0.85), p=0.002] (Supplemental Table 1). The association between ALM/BMI and type 2 diabetes was not significant in the adjusted model, but upon removing waist circumference from the model, diabetes risk was lower by 38% per SD higher ALM/BMI [HR 0.62 (0.40–0.98), p=0.04] (Supplemental Table 1). With ALM/weight as the exposure, diabetes risk was lower by 19% per SD higher ALM/weight [HR 0.81 (0.69–0.96), p=0.02) (Supplemental Table 2). The association was not significant in the adjusted model, but upon removing waist circumference from the model, diabetes risk was lower by 48% per SD higher ALM/weight [HR 0.52 (0.34–0.79), p=0.002] (Supplemental Table 2). With ALM as the exposure, diabetes risk was higher by 55% per SD higher ALM (kg) [HR 1.55 (1.32–1.81), p<0.0001]; the association was not significant in the adjusted model, but upon removing waist circumference from the model, diabetes risk was 2.4 times higher SD higher ALM (kg) [HR 2.38 (1.62–3.50), p<0.0001] (Supplemental Table 3). In a model including ALM, 1/BMI, and an interaction term ALM*(1/BMI), there was no evidence of interaction of ALM by 1/BMI (p=0.91). With ALM/height² as the exposure, diabetes risk was higher by 69% per SD higher ALM/height² (kg/m²) [HR 1.69 (1.45–1.98), p<0.0001] (Supplemental Table 4). The association was not significant in the adjusted model, but upon removing waist circumference from the model, diabetes risk was 2.4 times higher per SD higher ALM/height² (kg/m²) [HR 2.38 (1.62–3.50), p<0.0001] (Supplemental Table 4).

Discussion

In this prospective observational study, middle-aged men and women with higher appendicular lean mass relative to their BMI were at a lower risk of developing type 2 diabetes over 10 years. For instance, a man with a BMI of 28.0 kg/m² (mean BMI in this cohort) and ALM of 33.7kg (ALM/BMI of 1.06), has a 21% lower risk of diabetes over 10 years than a man the same BMI but ALM of 28.8 kg (mean ALM in this cohort) (ALM/BMI 1.20). Among men, the association between relative appendicular lean mass and incident type 2 diabetes is largely explained by waist circumference, a more readily measured diabetes risk factor. However, among women, the presence of other diabetes risk factors accounted for the observed associations. In sex-combined models of other traditional measures of relative appendicular lean mass, a similar pattern emerged that relative appendicular lean mass not associated with incident type 2 diabetes after controlling for other diabetes risk factors, which was largely explained by waist circumference.

Previous cross-sectional studies in adults across the age spectrum have demonstrated that more lean mass (measured by BIA or DXA) or muscle cross-sectional area (by CT scan) relative to body weight is associated with lower insulin resistance and prevalent type 2 diabetes [6, 9, 27]. However, prospective studies in middle-aged and older Black and White adults have had discordant results, and prospective data specifically in middle-aged Black and White adults were lacking. Two studies reported that a higher percentage of lean mass measured by DXA was associated with a lower risk of incident type 2 diabetes in White men age 35–80 years [16], and White and Black men age 20–98 years [18], after a median follow-up of 5 – 7 years. The former study found that the association was no longer significant after controlling for other diabetes risk factors, such as age, waist circumference, physical activity, hypertension, and triglyceride levels. The latter study found that relatively higher lean mass over time was associated with a lower risk of type 2 diabetes in men, including after controlling for race, but not women. In contrast, Larsen et al. reported that higher lean mass measured by DXA was associated with a higher risk of type 2 diabetes in both men and women (age 70–79 years, approximately 60% White) after a median follow-up of 11 years, but was not significant after controlling for age, race, physical activity, smoking, lipids, hypertension, BMI, visceral fat, and total body fat [17].

The primary analysis from our prospective study is consistent with prior literature that relative appendicular lean mass is not associated with incident type 2 diabetes after controlling for other diabetes risk factors. We add to the literature in three salient ways. First, we investigated the association between relative appendicular lean mass and incident type 2 diabetes using a specific body composition variable (ALM/BMI) recommended by the FNIHSP to identify low lean mass relative to body mass in other populations, such as older adults [19]. This is an important consideration because higher body weight is both a strong risk factor for type 2 diabetes and associated with higher muscle mass [28], and is consistent with a prior study showing that higher ALM/BMI was associated with lower risk of incident type 2 diabetes in Korean women with a history of gestational diabetes mellitus [29]. Second, we specifically investigated adults during middle-age, which is a critical window as accelerated muscle loss is yet to occur, and type 2 diabetes prevention in middle age reduces lifelong morbidity and mortality. Third, we investigated whether the association between relative appendicular lean mass and incident type 2 diabetes varied by sociodemographic and biological characteristics, race and sex. We add to the literature by demonstrating that the relationship between appendicular lean mass relative to BMI (ALM/BMI) and incident type 2 diabetes is largely explained by waist circumference. We did not investigate whether the rate of change in lean mass, or qualities of muscle other than its mass, such as its level of insulin sensitivity or amount of intermuscular or intramuscular fat, are associated with increased type 2 diabetes risk. This could be an area of further investigation.

The reason for the sexual dimorphism in our primary analysis is unknown. At a given BMI, men have more lean mass and visceral adipose tissue, and a higher risk of type 2 diabetes, than women [20]. Excess visceral adiposity, an established risk factor in the causal pathway of type 2 diabetes pathogenesis, has been associated with loss of lean muscle mass over time [30]. Therefore, in men, relatively less appendicular lean mass, i.e., lower ALM/BMI, may simply be a marker of higher visceral adiposity indirectly measured by waist circumference. In addition, by magnitude of risk per SD, waist circumference was a much stronger risk factor for incident diabetes than ALM/BMI, and waist circumference can be more readily measured by trained staff in a physician’s office or in the field, whereas measurement of appendicular lean mass requires a DXA scanner. In women, ALM/BMI is not independently associated with incident type 2 diabetes regardless of whether waist circumference is in the model. This suggests that relatively less appendicular lean mass may be a marker of different type 2 diabetes risk factors in women vs men. Since skeletal muscle is responsible for the majority of the body’s postprandial glucose disposal [3], further investigation is needed to understand skeletal muscle’s metabolic activity as an endocrine organ rather than simply its mass.

Our secondary analyses advance the literature by demonstrating the association between different measures of relative appendicular lean mass and incident type 2 diabetes in the same population. The idea that different measures of relative lean mass relate differently to insulin resistance is not new. Cross-sectional studies have shown that ALM/BMI and ALM/weight are negatively, while ALM and ALM/height² are positively, associated with insulin resistance and prevalent type 2 diabetes [8, 11, 31, 32]. A previous prospective study that included older (>50 years old) White adults demonstrated that lower ALM/weight, but higher ALM, are associated with higher risk of incident type 2 diabetes, neither of which were significant after controlling for multiple diabetes risk factors [33]. Our data in middle-aged White and Black adults are similar. A recent paper reported that older White adults who had greater lean mass (defined as lean mass/height²) and fat mass (defined as fat mass/height²) were at higher risk of developing type 2 diabetes, but adjustment for other diabetes risk factors was not performed. Our data suggest that—similar to ALM/BMI, ALM/weight, and ALM—ALM/height² is not associated with incident diabetes after adjusting for other diabetes risk factors. In general, a measure of relative appendicular lean mass that does not adjust for adiposity appears to be associated with increased type 2 diabetes risk since adults with excess adiposity also have higher lean mass to carry a greater weight. When added to our models, waist circumference was a strong risk factor for incident diabetes, and relative appendicular lean mass was no longer significant. This suggests that the detrimental effects of abdominal fat mass are greater than any potential effects, positive or negative, of lean mass on type 2 diabetes risk.

The strengths of this study are its large size, prospective cohort, and use of multiple different measures of relative appendicular lean mass in the same population. Our findings should be considered in light of some limitations. There was a low incidence of type 2 diabetes in this population due to the relatively young age, which constrains the power to detect an association in models with covariates. In addition, OGTT and HbA1c data were excluded from 2015–16 due to their availability in only a subset of participants; however, these data were available in all participants in 2010–11 and were included in the analysis. Some covariates were measured by self-report, such as physical activity, and subject to recall bias. We did not stratify results in women by menopausal status as menopause was not defined using biochemical data and thus may not be accurately captured using menstrual cycle data alone, such as in women who are status post hysterectomy. The results of this study may not be generalizable to other age, race or ethnic groups.

In conclusion, more appendicular lean mass relative to BMI is associated with a lower risk of developing type 2 diabetes in middle-aged men and women over 10 years, but not after controlling for other type 2 diabetes risk factors. Our results were consistent across other measures of relative appendicular lean mass. This suggests that the effect of relative muscle mass, as approximated by appendicular lean mass, on type 2 diabetes risk is through other metabolic risk factors.

Highlights.

• Appendicular lean mass relative to BMI is inversely associated with diabetes risk

• The association may be explained by other diabetes risk factors

• Other diabetes risk factors, e.g. waist circumference, are more easily measured

Acronyms:

DXA

dual energy x-ray absorptiometry

BIA

bioelectrical impedance analysis

MRI

magnetic resonance imaging

CT

computed tomography

ALM/BMI

appendicular lean mass for body mass index

FNIHSP

Foundation for the National Institutes of Health Sarcopenia Project

CARDIA

Coronary Artery Risk Development in Young Adults

Data Availability

The datasets analyzed during the current study are avail-able from the CARDIA study (www.cardia.dopm.uab.edu).