Metabolic syndrome and Growth Differentiation Factor 15 in older adults

Abstract

Growth Differentiation Factor 15 (GDF-15) is a cytokine produced in response to tissue injury and inflammatory states that may capture distinct pathways between the risk factors aggregated within metabolic syndrome (MS) and the development of diabetes and cardiovascular disease. This work aims to study the association of MS and its components with GDF-15 among older adults, examining the roles of body fat distribution, glucose metabolism, and inflammation. Data were taken from the Seniors-ENRICA-2 study in Spain, which included 1938 non-institutionalized individuals aged ≥65 years free of diabetes and cardiovascular disease. MS was defined as the presence of ≥3 of the following components: high waist circumference, elevated fasting blood glucose levels, raised blood pressure, increased triglyceride levels, and low serum high-density lipoprotein (HDL) cholesterol. Statistical analyses were performed with linear regression models and adjusted for potential sociodemographic and lifestyle confounders. MS was associated with higher GDF-15 levels (fully adjusted mean increase [95% confidence interval] = 9.34% [5.16,13.7]). The MS components showing the strongest associations were high waist circumference (6.74% [2.97,10.6]), elevated glucose levels (4.91% [0.77,9.23]), and low HDL-cholesterol (8.13% [3.51,13.0]). High waist-to-hip ratio (7.07% [2.63,11.7]), urine albumin (12.1% [2.57,22.5]), and C-reactive protein (10.4% [3.89,17.3]) were also associated with increased GDF-15. In conclusion, MS was associated with higher GDF-15 levels in older adults. Abdominal obesity, hyperglycemia –possibly linked to microvascular disease, as inferred from elevated urine albumin–, low HDL-cholesterol, and inflammation were the main drivers of this association.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11357-021-00370-w.

Background

Metabolic syndrome (MS) is a cluster of five cardiometabolic risk factors comprising abdominal obesity and abnormal blood pressure, blood glucose, triglycerides, and high-density lipoprotein (HDL) cholesterol. MS has been advocated as a simple clinical tool for predicting type 2 diabetes and cardiovascular disease, as it increases the risk for the former fivefold and doubles the risk for the latter [1–3]. Several pathophysiological links between the metabolic risk factors aggregated within MS and the development of diabetes and cardiovascular disease have been proposed [2–4]. First, insulin resistance is the primary cause of hyperglycemia, whose most relevant clinical outcome is microvascular disease, manifested as neuropathy and nephropathy. Insulin resistance may also be linked with hypertension through increased renal reabsorption of sodium leading to an expansion of intravascular volume, further accelerating the development of heart failure and atherosclerosis. Second, high levels of triglycerides and reduced HDL cholesterol, jointly with elevated low-density lipoproteins, seem to be the main cause of atherosclerotic cardiovascular disease. Finally, adipose tissue is associated with: (1) insulin resistance via the supply of ectopic fat to muscle and pancreas; (2) hypertension through activation of the renin–angiotensin–aldosterone and sympathetic nervous systems; and (3) insulin resistance, hypertension, and atherosclerosis via the production of inflammatory adipokines and other bioactive peptides [2–5].

In this regard, Growth Differentiation Factor 15 (GDF-15) is a cytokine produced in response to tissue injury and inflammatory states by cardiomyocytes, adipocytes, macrophages, endothelial cells, and vascular smooth muscle cells, where it plays a tissue-protective role through up- and downregulation of several signaling pathways [6, 7]. Higher serum GDF-15 concentrations may be a clinically relevant biomarker within the context of MS, as they have been associated with several MS components [6, 8, 9] and with the development and progression of cardiovascular and diabetes-related conditions, specifically cardiac hypertrophy, heart failure, atherosclerosis, endothelial dysfunction, insulin resistance, diabetes, and chronic kidney disease, as well as with cardiovascular and all-cause mortality [6, 7]. Nevertheless, little is known about the association of MS with GDF-15 in older adults and its main drivers [8, 10], and it is uncertain if they are related to alterations in glucose metabolism, inflammation, or both. Moreover, these assessments might render different results than in younger populations [8, 11], as both MS prevalence and GDF-15 levels are consistently higher in the elderly than in younger subjects [7, 12–14]. Furthermore, despite being an adipokine, it is unclear whether GDF-15 is associated with abdominal fat –that clustered within MS–, gluteofemoral fat, or overall adiposity [7, 8].

We hence aimed to (1) assess the association of MS and its components with GDF-15 in community-dwelling older adults and (2) delve into these relationships by examining the associations of ancillary adiposity measures (body mass index, hip circumference, and waist-to-hip ratio) and auxiliary metabolic biomarkers (Homeostatic Model Assessment for Insulin Resistance [HOMA-IR], glycated hemoglobin, urine albumin, and high-sensitivity C-reactive protein [hs-CRP]) with GDF-15.

Methods

Study design, setting, and participants

Our data came from the baseline wave of the Seniors-ENRICA-2, a cohort study on cardiovascular health, nutrition, and physical functioning in older adults in Spain (https://clinicaltrials.gov/ct2/show/NCT03541135) [15, 16]. Subjects were recruited between December 2015 and June 2017 by sex- and district-stratified random sampling of the community-dwelling, 65-years and older national healthcare cardholders living in the city of Madrid (Spain) or four surrounding large towns: Getafe, Torrejón, Alcorcón, and Alcalá de Henares.

The study methods were analogous to those of the Seniors-ENRICA-1 cohort, which have been detailed elsewhere [17]. In brief, a comprehensive set of physical examinations, blood, and urine tests were collected during two home visits by trained personnel, whereas data on socio-demographic, lifestyle, and morbidity variables were gathered through a telephone interview [17]. All subjects gave written informed consent, and the Clinical Research Ethics Committee of the “La Paz” University Hospital in Madrid approved the research protocol.

Variables, data sources, and measurements

GDF-15

Fasting blood samples were obtained from every subject at the first home visit in rapid serum tubes with a thrombin-based clot activator and polymer gel (Becton Dickinson). The tubes were centrifuged at 3000 rpm for 10 min and serum was aliquoted and stored at −80°C in the Department of Preventive Medicine and Public Health at Universidad Autónoma de Madrid. Serum GDF-15 was quantified at the Department of Laboratory Medicine of “La Paz” University Hospital by an electrochemiluminescence Elecsys® immunoassay method using a cobas® 6000 analyzer (Roche Diagnostics). The inter-assay coefficient of variation was 5.4% for a mean concentration of 7343 pg/mL and 7.7% for 1428 pg/mL.

Metabolic syndrome

Waist circumference was measured by trained staff with a flexible, inelastic, belt-type tape at the midpoint between the lowermost rib and the iliac crest at the end of a normal expiration [18]. Blood pressure was determined three times under standardized conditions with validated automatic sphygmomanometers (Mobil-O-Graph® 24h PWA, I.E.M., Stolberg, Germany), and the mean of the 2^nd and 3^rd assessments was used in the analyses [19]. 12-h fasting serum glucose, triglycerides, and HDL-cholesterol were measured with colorimetric enzymatic methods using Atellica® solution (Siemens Healthineers).

MS was defined as the presence of ≥3 of the following 5 components: a waist circumference ≥102 cm in men and ≥88 cm in women; fasting blood glucose ≥100 mg/dL or drug treatment of hyperglycemia; a systolic blood pressure ≥130, a diastolic blood pressure ≥85 mm Hg, or being on antihypertensive drug treatment; serum triglycerides ≥150 mg/dL; and serum HDL-cholesterol <40 mg/dL in men or <50 mg/dl in women [1]. Since it has been argued that MS is a pre-morbid condition rather than a clinical diagnosis [2], we excluded individuals with self-reported cardiovascular disease (myocardial infarction, stroke, or heart failure) or established diabetes (blood glucose ≥126 mg/dL, HbA1c ≥6.5%, being treated with antidiabetic drugs, or self-reported diagnosis of diabetes mellitus).

Ancillary adiposity measures

To further investigate the association between body fat and GDF-15, we used data on three additional adiposity measures. First, body mass index (BMI) was calculated as weight (in kg) divided by squared height (in m). Weight and height measurements were conducted under standardized conditions using electronic scales and portable extendable stadiometers, respectively [18]. Normal weight was considered BMI <25, overweight as BMI 25–29.9, and obesity as BMI ≥30 kg/m². Hip circumference was measured by trained staff with a flexible, non-distensible, belt-type tape on the maximum circumference over the femoral trochanters; for analyses, hip circumference was divided into sex-specific tertiles (cutoffs at 98 and 104 cm in men, and 99.6 and 107 cm in women). The waist-to-hip ratio was computed as waist circumference (in cm) divided by hip circumference (in cm). Cutoff values of ≥0.90 in men and ≥0.85 in women were used, for they have been associated with a substantially increased risk of metabolic complications [20].

Ancillary metabolic biomarkers

To explore in more depth the relationship between glucose metabolism and GDF-15 levels, we first measured 12-h fasting serum insulin through a chemiluminescent immunoassay using Atellica® solution (Siemens Healthineers) and calculated the HOMA-IR as blood glucose (in mg/dL) multiplied by serum insulin (in mU/L) and further divided by 405 [21]. We used cutoff values of ≥2.25 in men and ≥2.38 in women, as they have demonstrated a 70% specificity in MS classification in older adults [22]. Second, we determined glycated hemoglobin (HbA_1c) using high-performance liquid chromatography with Arkray Adams™ A1c HA-8180 (Menarini). The threshold level for prediabetes was set at ≥5.7% [23]. We finally measured albumin excretion in spot urine, as an early predictor of progressive renal function loss in prediabetes and diabetes, through the immunoturbidimetry technique using Atellica® solution (Siemens Healthineers). Urinary albumin excretions <10, 10–20, and ≥20 mg/L were considered normal, high normal, and microalbuminuria, respectively [24].

As a marker of inflammation, we also determined hs-CRP levels using the abovementioned immunoturbidimetry technique. Cutoff points were set at 1.0 mg/L and 3.0 mg/dL, according to relative risk categories of cardiovascular disease [25].

Potential confounders

We used data on several self-reported potential confounders of the association between MS and GDF-15, specifically sex, age, educational level (primary or less, secondary, or university), smoking status (never, former, or current), and alcohol consumption (never, former, moderate [≤10 g/day in women and ≤20 g/day in men], or heavy). Physical activity time (min/day) was assessed with an ActiGraph GT9X (ActiGraph Inc) accelerometer. Intensity thresholds were set at <45 miligravitational units (mg) for sedentary behavior, ≥45 and <100 mg for light physical activity, and ≥100 mg for moderate-to-vigorous physical activity [16]. When not available, we used self-reported data on sedentary behavior and physical activity instead [26, 27]. Dietary information, including energy intake (kcal/day), was obtained from a validated diet history [28]. Diet quality was assessed with the Mediterranean Diet Adherence Screener (MEDAS), with higher scores indicating better adherence to the Mediterranean diet [29].

Statistical methods

Study size

From the 3273 individuals recruited in the study (51% of those invited), we excluded 758 with inadequate data (719 subjects had incomplete information on MS or ancillary adiposity measures, 684 on GDF-15, and 488 on potential confounders; note that one individual may lack data in more than one variable). We also excluded 577 individuals with established diabetes or known cardiovascular disease. Hence, the main analytical sample comprised 1938 individuals. For the analyses regarding ancillary metabolic biomarkers, we excluded another 1085 participants without data on HOMA-IR, HbA_1c, urine albumin, or hs-CRP. Thus, this secondary analytical sample included 853 subjects (Online resource 1: Supplementary Fig. 1).

Statistical methods

Differences in characteristics of study participants across the categories of MS and its components were evaluated with Pearson’s chi-squared tests for discrete variables and Wilcoxon rank-sum tests for continuous variables.

Main analyses were conducted with linear regression models where the outcome was log-transformed GDF-15, as a continuous variable, and exposures were MS or its components, as dichotomous variables. We calculated GDF-15 mean percentage differences and their 95% confidence interval (CI) between the subjects with and without MS and each of its components. This was done by taking 1 from the exponentiated β coefficients in the regression models and multiplying the result by 100. Dose–response relationships were assessed by restricted cubic spline regression. To control for potential confounding, two a priori hierarchical models were used: (1) adjusted for sociodemographic characteristics (age, sex, and educational level), and (2) additionally adjusted for lifestyle variables (tobacco smoking, alcohol consumption, diet quality, energy intake, light physical activity, moderate-to-vigorous physical activity, and sedentary behavior).

The associations of ancillary adiposity measures and metabolic biomarkers with GDF-15 were examined alike, except that we used tests for trend to check for dose–response relationships instead, modeling the median values per category as a single continuous variable.

Statistical significance was set at a two-sided p value <0.05. Analyses were performed with Stata® (StataCorp LLC), version 14.

Missing data, interactions, and sensitivity analyses

First, to investigate how incomplete data may have affected our findings, we compared the characteristics between participants who were and were not included in the analyses because of missing values on any variable of interest. Second, we tested whether the association of MS with GDF-15 differed in men and women, or subjects ≤70 and >70 years, as both MS prevalence and GDF-15 levels appear to steadily increase with age [7, 12–14]. To do so, we used Wald tests that compared models with and without interaction terms, defined as the product of sex or the age subgroup by the dichotomous MS variable. Since no statistically significant interactions were found, results are presented for the total sample. Finally, to provide further insights into the dose–response relationship between MS and GDF-15 at higher levels of its components –particularly fasting glucose–, we replicated the analyses without excluding individuals with established diabetes or known cardiovascular disease.

Results

Descriptive data

The prevalence of MS [95% CI] was 26.9% [25.0,28.9]. Subjects with MS were more likely to be women, slightly older, more sedentary, did less light and moderate-to-vigorous physical activity, and had a higher energy intake. Table 1 also shows a detailed distribution of each MS component by sociodemographic and lifestyle variables.

Table 1.

Characteristics of 1938 older adults free of diabetes and cardiovascular disease,^a by metabolic syndrome status and its components

			Components of metabolic syndrome
	Metabolic syndrome		Waist circumference		Fasting glucose		Blood pressure		Triglycerides		HDL-cholesterol
	No	Yesb	Normal	Highc	Normal	Highd	Normal	Highe	Normal	Highf	Normal	Lowg
n	1417	521	1035	903	1480	458	440	1498	1612	326	1557	381
GDF-15 (pg/mL), geometric mean [geometric SD factor]	1088 [1.50]	1227 [1.53]*	1094 [1.53]	1159 [1.50]*	1105 [1.50]	1188 [1.54]*	1082 [1.50]	1136 [1.52]*	1115 [1.51]	1166 [1.51]	1096 [1.50]	1243 [1.55]*
Sex, %
Men	47.6	40.7*	57.8	31.9*	42.8	55.2*	41.1	47.1*	45.4	47.2	47.2	39.6*
Women	52.4	59.3	42.2	68.1	57.2	44.8	58.9	52.9	54.6	52.8	52.8	60.4
Age (years)	71.3 [4.22]	71.9 [4.48]*	71.2 [4.11]	71.7 [4.49]*	71.3 [4.27]	71.8 [4.36]*	70.8 [4.22]	71.6 [4.30]*	71.5 [4.25]	71.2 [4.50]	71.3 [4.25]	71.8 [4.44]
Educational level, %
Primary or less	61.2	64.9	57.2	67.9*	61.6	64.0	60.2	62.8	62.0	62.9	61.8	63.5
Secondary	19.5	18.8	22.2	15.9	19.1	20.1	19.8	19.2	18.9	21.5	19.9	16.8
University	19.3	16.3	20.6	16.2	19.3	15.9	20	18.1	19.1	15.6	18.2	19.7
Tobacco smoking, %
Never	54.3	52.6	52.1	55.9	55.8	47.6*	53.2	54.1	54.5	50.9*	53.8	54.3*
Former	36.6	37.4	37.3	36.2	34.6	43.9	37.3	36.6	37.0	35.6	37.7	33.1
Current	9.1	10.0	10.6	7.9	9.6	8.5	9.6	9.3	8.5	13.5	8.5	12.6
Alcohol consumption, %
Never	17.2	20.9	15.6	21.3*	19.0	15.7*	20	17.7	17.8	20.2	17.1	22.6*
Former	6.35	5.76	6.76	5.54	6.55	5.02	7.95	5.67	6.33	5.52	5.33	9.71
Moderateh	53.0	53.6	53.7	52.5	53.3	52.6	50.7	53.9	53.7	50.3	51.9	58.3
Heavy	23.4	19.8	24.0	20.7	21.1	26.6	21.4	22.8	22.1	23.9	25.6	9.4
Diet quality (MEDAS)	7.15 [1.67]	7.09 [1.78]	7.12 [1.67]	7.15 [1.74]	7.12 [1.68]	7.19 [1.75]	7.07 [1.63]	7.15 [1.72]	7.17 [1.69]	6.98 [1.74]*	7.18 [1.70]	6.94 [1.68]*
Energy intake (kcal/day)	1936 [337]	1985 [366]*	1950 [342]	1947 [350]	1924 [333]	2029 [374]*	1915 [337]	1959 [348]*	1939 [345]	1996 [346]*	1954 [346]	1927 [345]
Light physical activity (min/day)	159 [52.2]	145 [52.8]*	156 [52.1]	153 [53.5]	157 [52.3]	148 [53.7]*	162 [51.5]	153 [53.0]*	157 [52.6]	144 [52.5]*	158 [52.1]	141 [53.2]*
Moderate-to-vigorous PA (min/day)	65.5 [39.4]	50.3 [34.6]*	67.1 [39.5]	54.8 [36.9]*	61.8 [38.1]	59.9 [40.8]	67.1 [41.2]	59.7 [37.9]*	63.2 [39.4]	52.6 [34.4]*	64.8 [39.8]	47.6 [30.5]*
Sedentary behavior (min/day)	769 [138]	822 [158]*	774 [152]	794 [137]*	776 [140]	806 [161]*	772 [147]	787 [145]*	775 [139]	823 [170]*	775 [142]	818 [155]*

MEDAS = Mediterranean Diet Adherence Screener. PA = physical activity. Values are means [standard deviations] unless otherwise indicated. ^*p value <0.05 for differences in means (Wilcoxon rank-sum) or proportions (Pearson’s chi-squared) across the categories of metabolic syndrome and its components

^aParticipants who had a blood glucose level ≥ 126 mg/dL, had HbA_1c levels ≥6.5%, were treated with antidiabetic drugs, or had a diagnosis of diabetes mellitus or cardiovascular disease (myocardial infarction, stroke, or heart failure)

^b≥3 components of metabolic syndrome

^cWaist circumference ≥102 cm in men and ≥88 cm in women

^dFasting glucose levels ≥100 mg/dL

^eSystolic blood pressure ≥130 mm Hg, or diastolic blood pressure ≥85 mm Hg, or treatment with antihypertensive medication

^fTriglyceride levels ≥150 mg/dL

^gSerum HDL-cholesterol <40 mg/dL in men and <50 mg/dL in women

^hModerate drinking: ≤10 g/day in women and ≤20 g/day in men

Compared to participants included in our analyses, those with incomplete data had higher GDF-15 levels (1389 vs 1124 pg/mL). They also were more likely to have MS (41.7 vs 26.9%), older (72.8 vs 71.4 years), less educated (68.1% vs 62.2% had primary or lower studies), more sedentary (1066 vs 783 min/day), did less moderate-to-vigorous physical activity (31.4 vs 61.2 min/day), and had lower diet quality (6.91 vs 7.13 MEDAS scores).

Main results

The geometric GDF-15 means [95% CI] were 1227 [1183,1272] and 1088 pg/mL [1065,1111] for participants with and without MS, respectively (mean percentage difference = 12.7% [8.17,17.5]) (Table 1, Table 2). This association remained when adjusting for sociodemographic (model 1 = 11.4% [7.12,15.8]) and lifestyle variables (model 2 = 9.34% [5.16,13.7]), and it was dose-dependent (Table 2, Fig. 1). The components of MS that contributed the most to this finding were high waist circumference (model 2 mean percentage difference = 6.74% [2.97,10.6]), high glucose levels (4.91% [0.77,9.23]), and low HDL-cholesterol (8.13% [3.51,13.0]). Conversely, the associations of high blood pressure and triglycerides with GDF-15 were modest and did not reach statistical significance (1.49% [−2.55,5.69] and 2.54% [−2.04,7.34], respectively) (Table 2, Fig. 1).

Table 2.

Association of metabolic syndrome and its components with GDF-15 in 1938 older adults free of diabetes and cardiovascular disease^a

		Mean percentage difference in GDF-15 [95% confidence interval]
	n	Crude	Model 1b	Model 2c
Metabolic syndrome
No	1417	0 (reference)	0 (reference)	0 (reference)
Yesd	521	12.7 [8.17,17.5]***	11.4 [7.12,15.8]***	9.34 [5.16,13.7]***
Components of metabolic syndrome
Waist circumference
Normal	1035	0 (reference)	0 (reference)	0 (reference)
Highe	903	5.94 [2.10,9.93]**	7.40 [3.63,11.3]***	6.74 [2.97,10.6]***
Fasting glucose levels
Normal	1480	0 (reference)	0 (reference)	0 (reference)
Highf	458	7.57 [3.00,12.3]***	4.40 [0.21,8.76]*	4.91 [0.77,9.23]*
Blood pressure
Normal	440	0 (reference)	0 (reference)	0 (reference)
Highg	1498	4.99 [0.46,9.72]*	1.64 [−2.48,5.93]	1.49 [−2.55,5.69]
Triglyceride levels
Normal	1612	0 (reference)	0 (reference)	0 (reference)
Highh	326	4.54 [−0.50,9.83]	5.24 [0.50,10.2]*	2.54 [−2.04,7.34]
Serum HDL cholesterol
Normal	1557	0 (reference)	0 (reference)	0 (reference)
Lowi	381	13.4 [8.26,18.7]***	12.6 [7.81,17.5]***	8.13 [3.51,13.0]***

*p<0.05. **p<0.01. ***p<0.001

^aParticipants who had a blood glucose level ≥ 126 mg/dL, had HbA_1c levels ≥6.5 %, were treated with antidiabetic drugs, or had a diagnosis of diabetes mellitus or cardiovascular disease (myocardial infarction, stroke, or heart failure)

^bModel 1: Linear regression model adjusted for sex, age, and educational level (primary or less, secondary, or university)

^cModel 2: As Model 1 and further adjusted for smoking status (never, former, or current), alcohol consumption (never, former, moderate [≤10 g/day in women and ≤20 g/day in men], or heavy), diet quality (Mediterranean Diet Adherence Screener), energy intake (kcal/day), light physical activity (min/day), moderate-to-vigorous physical activity (min/day), and sedentary behavior (min/day)

^d≥3 components of metabolic syndrome

^eWaist circumference ≥102 cm in men and ≥88 cm in women

^fFasting glucose levels ≥100 mg/dL

^gSystolic blood pressure ≥130 mm Hg, or diastolic blood pressure ≥85 mm Hg, or treatment with antihypertensive medication

^hTriglyceride levels ≥150 mg/dL

ⁱSerum HDL cholesterol <40 mg/dL in men and <50 mg/dL in women

Fig. 1.

Association of metabolic syndrome and its components with GDF-15 in 1938 older adults free of diabetes and cardiovascular disease [participants who had a blood glucose level ≥ 126 mg/dL, had HbA_1c levels ≥6.5 %, were treated with antidiabetic drugs, or had a diagnosis of diabetes mellitus or cardiovascular disease (myocardial infarction, stroke, or heart failure)]. Plotted values are geometric means (95% confidence intervals) obtained from a linear regression model adjusted as Model 2 in Table 2: sex, age, educational level (primary or less, secondary, or university), smoking status (never, former, or current), alcohol consumption (never, former, moderate [≤10 g/day in women and ≤20 g/day in men], or heavy), diet quality (Mediterranean Diet Adherence Screener), energy intake (kcal/day), light physical activity (min/day), moderate-to-vigorous physical activity (min/day), and sedentary behavior (min/day). The restricted cubic spline knots are located at 1–2–3 components for metabolic syndrome, 81–95–109 cm for waist circumference, 79–91–107 mg/dl for fasting glucose levels, 112–133–158 mm Hg for systolic blood pressure, 67–80–94 mm Hg for diastolic blood pressure, 64–99–169 mg/dl for triglyceride levels, and 40–54–74 mg/dl for serum HDL-cholesterol levels

Consistent with the findings for waist circumference alone, a high waist-to-hip ratio was associated with 7.07% [2.63,11.7] higher levels of GDF-15, even though there was little to no association for hip circumference alone (0.77% [−3.50,5.23]) and BMI ≥30 (2.38% [−2.47,7.48]) (Table 3).

Table 3.

Association of ancillary adiposity measures with GDF-15 in 1938 older adults free of diabetes and cardiovascular disease^a

		Mean percentage difference in GDF-15 [95% confidence interval]
	n	Crude	Model 1b	Model 2c
Body mass index (kg/m2)
Categories
<25	559	0 (reference)	0 (reference)	0 (reference)
25 to 30	932	−2.39 [−6.55,1.94]	−3.08 [−6.95,0.96]	−2.27 [−6.13,1.74]
≥30	447	3.30 [−1.89,8.77]	2.90 [−1.96,8.01]	2.38 [−2.47,7.48]
p for trend	1938	0.220	0.249	0.353
Hip circumference
Tertiles d
1 (lower)	694	0 (reference)	0 (reference)	0 (reference)
2	681	−1.17 [−5.41,3.26]	−1.25 [−5.21,2.88]	−0.56 [−4.49,3.53]
3 (higher)	563	1.77 [−2.82,6.58]	1.55 [−2.74,6.02]	0.77 [−3.50,5.23]
p for trend	1938	0.448	0.477	0.724
Waist-to-hip ratio
Categories
Normal	462	0 (reference)	0 (reference)	0 (reference)
High e	1476	11.9 [7.16,16.8]***	8.18 [3.65,12.9]***	7.07 [2.63,11.7]**
p for trend	1938	<0.001	<0.001	0.002

*p<0.05. **p<0.01. ***p<0.001

^aParticipants who had blood glucose levels ≥ 126 mg/dL, had HbA_1c levels ≥6.5%, were treated with antidiabetic drugs, or had a diagnosis of diabetes mellitus or cardiovascular disease (myocardial infarction, stroke, or heart failure)

^bModel 1: Linear regression model adjusted for sex, age, and educational level (primary or less, secondary, or university)

^cModel 2: As Model 1 and further adjusted for smoking status (never, former, or current), alcohol consumption (never, former, moderate [≤10 g/day in women and ≤20 g/day in men], or heavy), diet quality (Mediterranean Diet Adherence Screener), energy intake (kcal/day), light physical activity (min/day), moderate-to-vigorous physical activity (min/day), and sedentary behavior (min/day)

^dHip circumference tertiles: tertile 1, ≤ 98 cm in men and ≤ 99.6 cm in women; tertile 2, 98 to 104 cm in men and 99.6 to 107 cm in women; tertile 3, >104 cm in men and >107 cm in women

^eWaist-to-hip ratio ≥0.90 in men and ≥0.85 in women

Regarding ancillary metabolic biomarkers, HbA_1c ≥5.7% did not show an association with GDF-15 levels (−0.05% [−4.94,5.09]). Still, some trend was found for high HOMA-IR (3.13% [−1.91,8.43]), while participants with high urine albumin and hs-CRP concentrations did have higher GDF-15 levels (12.1% [2.57,22.5] and 10.4% [3.89,17.3], respectively) (Table 4).

Table 4.

Association of ancillary metabolic biomarkers with GDF-15 in 853 older adults free of diabetes and cardiovascular disease^a

		Mean percentage difference in GDF-15 [95% confidence interval]
	n	Crude	Model 1b	Model 2c
HOMA-IR
Categories
Normal	478	0 (reference)	0 (reference)	0 (reference)
High d	375	4.85 [−0.50,10.5]	3.53 [−1.50,8.82]	3.13 [−1.91,8.43]
p for trend	853	0.076	0.172	0.227
HbA1c (%)
Categories
<5.7	512	0 (reference)	0 (reference)	0 (reference)
5.7 to 6.4	341	2.34 [−2.96,7.93]	0.57 [−4.41,5.80]	−0.05 [−4.94,5.09]
p for trend	853	0.393	0.826	0.984
Urine albumin (mg/L)
Categories
<10	649	0 (reference)	0 (reference)	0 (reference)
10 to 20	132	7.56 [0.086,15.6]*	4.76 [−2.20,12.2]	3.42 [−3.40,10.7]
≥20	72	18.1 [7.56,29.7]***	13.6 [3.87,24.2]**	12.1 [2.57,22.5]*
p for trend	853	<0.001	0.003	0.010
High-sensitivity C-reactive protein (mg/L)
Categories
<1	391	0 (reference)	0 (reference)	0 (reference)
1 to 3	248	−1.02 [−6.91,5.24]	−0.53 [−6.14,5.42]	−0.71 [−6.24,5.15]
>3	214	11.0 [4.09,18.4]**	12.6 [5.96,19.7]***	10.4 [3.89,17.3]**
p for trend	853	<0.001	<0.001	<0.001

*p<0.05. **p<0.01. ***p<0.001. HbA_1c = glycated hemoglobin

^aParticipants who had a blood glucose level ≥ 126 mg/dL, had HbA_1c levels ≥6.5 %, were treated with antidiabetic drugs, or had a diagnosis of diabetes mellitus or cardiovascular disease (myocardial infarction, stroke, or heart failure). We further excluded 1085 participants without data on HOMA-IR, HbA1c, urine albumin, or hs-CRP (Online resource 1: Supplementary Fig. 1)

^bModel 1: Linear regression model adjusted for sex, age, and educational level (primary or less, secondary, or university)

^cModel 2: As Model 1 and further adjusted for smoking status (never, former, or current), alcohol consumption (never, former, moderate [≤10 g/day in women and ≤20 g/day in men], or heavy), diet quality (Mediterranean Diet Adherence Screener), energy intake (kcal/day), light physical activity (min/day), moderate-to-vigorous physical activity (min/day), and sedentary behavior (min/day)

^dHOMA-IR ≥ 2.25 in men and ≥ 2.38 in women

Other analyses

When including in the analyses the 577 participants with established diabetes or known cardiovascular disease, the strength of the association between MS and GDF-15 substantially increased (model 2 mean percentage difference = 24.3% [19.7,29.1]). So was the case for all the MS components, and even the associations with high blood pressure and triglycerides reached statistical significance (Online resource 1: Supplementary Table 1 and Supplementary Fig. 2). Contrary to the analyses restricted to premorbid subjects, BMI ≥30 was significantly associated with increased GDF-15 levels (Online resource 1: Supplementary Table 2), as were both HbA_1c ≥5.7% and high HOMA-IR (Online resource 1: Supplementary Table 3).

Discussion

Key results

In this study of older adults in Spain, MS was consistently associated with higher GDF-15 levels. The MS components that showed the strongest associations were high waist circumference, elevated glucose levels, and low serum HDL-cholesterol. Ancillary adiposity measures as the waist-to-hip ratio and auxiliary metabolic biomarkers as urine albumin and hs-CRP were also associated with increased GDF-15.

Interpretation

Our main results are in line with the few studies that have directly examined the association between MS and GDF-15. Specifically, in two cross-sectional studies (mean ages 80 and 59 years), higher GDF-15 levels were associated with MS (≈2.6 greater odds and p<0.001, respectively) [10, 30], whereas a small case–control study of obese subjects (mean age 34 years) found that patients with MS had ≈120% higher GDF-15 concentrations than otherwise healthy controls [11].

Contrary to other investigations conducted in younger populations [31, 32], we could hardly demonstrate a trend between BMI ≥30 and higher GDF-15 (Table 3), which only became apparent when including those subjects with diabetes or cardiovascular disease (Online resource 1: Supplementary Table 2). It has been argued that the current BMI classification may not be appropriate in older adults, as overweight and even moderate obesity (BMI 24–33) do not seem to be associated with increased mortality in this age subgroup [33, 34]. Anyway, since total body weight is primarily composed of both fat and muscle mass and only the former may be linked with higher GDF-15 concentrations, associations with BMI could be weaker than with other adiposity measures [8, 35]. In this regard, the association between high waist circumference, waist-to-hip ratio, and GDF-15 found by us (Table 3) and others [8, 31] –and the lack of it for hip circumference– may suggest that it is abdominal fat that mediates the corresponding association of obesity with GDF-15. Despite being a cytokine produced in response to adipose-tissue-driven inflammation (note that increased hs-CRP has been linked to both obesity and GDF-15 [Table 4] [30, 36]), abdominal fat leads to a higher release of free fatty acids into circulation than gluteofemoral fat, which may be deposited in other tissues and organs or re-esterified into triglycerides in the liver [3, 37]. On one hand, GDF-15 could then be reflecting the mitochondrial dysfunction associated with ectopic fat accumulation in muscle [12]. On the other hand, both elevated triglyceride levels and reduced serum HDL-cholesterol –as part of atherogenic dyslipidemia– play a role in the initiation and development of atherosclerosis [3], which may, in turn, lead to a rise in GDF-15 levels, as suggested by the strong association of the latter with cardiovascular disease [6, 7]. We and others have indeed found a link between serum HDL-cholesterol, triglycerides, and GDF-15 (Table 2, Fig. 1) [8, 30, 35], though in our study the second was not statistically significant after adjustment for lifestyle variables. As for elevated blood pressure, its contribution to atherogenesis may exert some degree of tissue damage at a systemic level [3], though there is little –if any– evidence linking hypertension with GDF-15 concentrations (Table 2, Fig. 1) [30, 35].

Another pillar of MS is alterations in glucose metabolism. As the primary cause of hyperglycemia in patients with MS, insulin resistance has traditionally been associated with higher levels of GDF-15 [31, 35, 38]. In our study, the magnitude of the estimates for HOMA-IR was rather smaller than for fasting glucose levels –both in main and sensitivity analyses–, waist circumference, and hs-CRP (Table 2, Table 3, Table 4). Since obesity increases insulin requirements, imposes metabolic stress on pancreatic beta cells, and promotes cellular exhaustion via pro-inflammatory signals [5], part of the association between hyperglycemia and GDF-15 might not be mediated by insulin resistance. Moreover, since insulin resistance habitually precedes the rise in blood glucose levels (at the expense of compensatory hyperinsulinemia), it may be of particular relevance during the initial phases of MS [3]. However, in more advanced stages, such as those more likely represented in a prevalence study like ours, blood glucose may already be elevated, and hence the association of glycemia with GDF-15 might predominate over that with insulin resistance. Of note was that the former association seemed to be biphasic (Fig. 1). Nevertheless, the association between lower glycemia and elevated GDF-15 concentrations did not reach statistical significance, contrary to what was observed for higher glucose levels, those clustered within the metabolic syndrome. Specifically, mean percentage differences in GDF-15 [95% CI] for increasing fasting glucose levels quintiles were: 5.39% [−0.23,11.33]; 2.26% [−3.03,7.83]; 0 [Reference]; 3.78% [−1.87,9.75]; 8.37% [2.54,14.53]. Moreover, when including in the analyses the 577 participants with established diabetes or known cardiovascular disease, the association of fasting glucose levels with GDF-15 concentrations did not appear to be J-shaped (Online resource 1: Supplementary Fig. 2). It is then possible that the left tail of the restricted cubic spline in Fig. 1 mimicked a change in trend due to the sparsity of data in the right end of the fasting glucose distribution. Specifically, only 196 subjects were above the third knot of the abovementioned spline (107 mg/dl) vs 597 subjects when including the participants with established diabetes or known cardiovascular disease. Lastly, we have been unable to find any evidence of the association of lower fasting glucose levels with increased GDF-15, as most studies either fit the former as a continuous variable –and hence report a single coefficient for the association– or compare whole groups of patients with and without diabetes or prediabetes [10, 31, 35, 39].

A further discrepancy with other epidemiologic studies is that HbA1c was not associated with higher GDF-15 [35], contrary to fasting glucose levels (Table 2, Table 4). Any explanation must be conjectural, as subjects with altered single time-point glucose measurements likely have altered average glycemia as well [40]. However, HbA1c levels may rise with age beyond the expected elevations in fasting glucose, and the specificity of HbA1c-based diagnostic criteria for prediabetes might decrease with increasing age [41]. Accordingly, in our main analytical sample, the Pearson’s correlation coefficient between fasting glucose and HbA1c was low (r=0.39), yet doubled when including those subjects with established diabetes, in line with the observed substantial increase in the association of HbA1c with GDF-15 (Online resource 1: Supplementary Table 3).

Finally, microalbuminuria, being the primary clinical outcome of hyperglycemia [3, 42], also was associated with elevated GDF-15 levels (Table 4), in line with the robust evidence on the secretion of GDF-15 in response to early endothelial and microvascular damage [8, 43] and on its role as a risk marker of diabetic nephropathy [8, 44].

Generalizability

On one hand, the prevalence of MS found in our study might be lower than that in other settings and countries. For instance, 42.3% of the Spanish population ≥65 years [13] and 50.4% of adults ≥60 years in the USA [14] have MS, compared to 36.3% of our participants—note that all figures comprise individuals with diabetes and cardiovascular disease. Nevertheless, (1) subjects included in our analyses met 0 to 5 MS criteria, while their ranges of waist circumference, fasting glucose, blood pressure, triglyceride, and serum HDL-cholesterol levels were also broad (Fig. 1); (2) the association of MS with GDF-15 was stronger when including in the analyses those subjects with diabetes and cardiovascular disease (Online resource 1: Supplementary Table 1 and Supplementary Fig. 2), who more frequently complied with ≥4 MS criteria and generally had higher levels of its components; and (3) the association between MS and GDF-15 showed a clear dose–response relationship with the number of MS criteria. So was the case for waist circumference, though some saturation seemed to arise at higher triglyceride and HDL-cholesterol levels (Fig. 1).

On the other hand, GDF-15 levels are reported to steadily increase with age [12, 30, 45]. Specifically, we found a 3.31% [2.89,3.74] increment for every additional year of age (Online resource 1: Supplementary Fig. 3). While the mean serum GDF-15 concentration found in subjects without MS (1088 pg/mL, Table 1) was consistent with that of other studies of healthy community-dwelling older adults [45, 46], it was ≈60% higher than that observed in subjects 17–71 years [8, 47]. Also, the strength of the association between MS and GDF-15 in our study appeared to increase somewhat with increasing age (model 2 mean percentage difference = 11.8% [5.73,18.3] for >70 years vs 7.38% [1.46,13.6] for ≤70 years; p for interaction=0.23), in line with the stronger association of obesity with GDF-15 in middle-aged subjects compared to children [32, 48], and perhaps reflecting a cumulative exposure to tissue injury and inflammatory states [8, 45]. It is nevertheless reassuring that our results are consistent with the evidence stemming from younger populations suggesting that both MS [11, 30] and its components [6, 8, 9] are associated with increased GDF-15 levels.

Limitations

Some study limitations should be acknowledged. First, because we used cross-sectional data, we cannot assure that MS always preceded the rise in GDF-15 levels. We cannot rule out that GDF-15 also plays a role in the regulation of food intake and in carbohydrate and lipid metabolism [8, 9], which might be in turn associated with MS itself. Second, anthropometric techniques can be helpful for the assessment of body fat distribution, but they might not be used to make inferences on subcutaneous and visceral fat –note that the former may exceed the latter by twofold or threefold even in subjects with abdominal obesity– [3, 49].

Moreover, despite MS and GDF-15 being assessed with standardized procedures and analytical techniques, some measurement error is unavoidable, though this would usually bias study results toward the null [50]. Also, the self-reported nature of some covariates may not allow to rule out residual confounding, even after adjusting the models for several lifestyle and sociodemographic variables—notably age, as GDF-15 has been reported an aging biomarker in multiple studies [12, 30, 35]. In this regard, since our investigation comprised people ≥65 years, our results may not necessarily apply to younger populations. Finally, some imprecision may have arisen due to the limited sample size, especially in the analyses regarding ancillary metabolic biomarkers (n=853, Table 4).

Conclusions

MS was associated with higher GDF-15 levels in older adults, highlighting its potential role as a biomarker for the development of diabetes and cardiovascular disease. The main drivers of this association were abdominal obesity, hyperglycemia –possibly linked to microvascular disease–, low HDL-cholesterol, and inflammation. Nonetheless, these findings should be confirmed by longitudinal studies covering a wider age range and using, wherever possible, imaging techniques for the assessment of body fat.

Abbreviations

BMI

Body mass index

CI

Confidence interval

HbA1c

Glycated hemoglobin

GDF-15

Growth Differentiation Factor 15

HDL

High-density lipoprotein

hs-CRP

High-sensitivity C-reactive protein

HOMA-IR

Homeostatic Model Assessment for Insulin Resistance

MEDAS

Mediterranean Diet Adherence Screener

MS

Metabolic syndrome

Author contributions

FRA conceived the study. ACC and RO performed statistical analyses. All authors contributed to results interpretation. ACC, FRA, and RO drafted the manuscript. All authors reviewed the manuscript for important intellectual content, read, and approved the final manuscript.

All authors have agreed both to be personally accountable for the author’s own contributions and to ensure that questions related to the accuracy or integrity of any part of the work, even ones in which the author was not personally involved, are appropriately investigated, resolved, and the resolution documented in the literature.

Funding

This work was supported by the Instituto de Salud Carlos III, State Secretary of R+D+I and FEDER/FSE (FIS grants 18/287, 19/319); grant 2020/017 from the National Plan on Drug Addiction (Ministry of Health), and the MITOFUN project grant from the Fundación Francisco Soria Melguizo. Adrián Carballo-Casla has an FPI contract from the Universidad Autónoma de Madrid. Reagents for measuring Growth Differentiation Factor 15 have been provided by Roche Diagnostics International through a Research Agreement with the FUAM (Fundación de la Universidad Autónoma de Madrid). The funding agencies had no role in study design, data collection, and analysis, interpretation of results, manuscript preparation, or the decision to submit this manuscript for publication.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding authors on reasonable request.

Code availability

The custom code used for the analyses is available from the corresponding authors on reasonable request.

Declarations

Ethics approval and consent to participate

Study participants provided written informed consent, and the Clinical Research Ethics Committee of “La Paz” University Hospital in Madrid approved the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.