Psychological Well-being and Metabolic Syndrome: Findings from the MIDUS National Sample

Abstract

Objective

Psychological well-being predicts favorable cardiovascular outcomes, but less evidence addresses biological mediators underlying these effects. Therefore, cross-sectional and longitudinal associations among hedonic and eudaimonic well-being and metabolic syndrome (MetSyn) were examined in a national sample.

Methods

Survey of Midlife in the U.S. participants (MIDUS; N = 1,205; 56.8% female; M_age(SD) = 57.5(11.6)) provided survey assessments of hedonic (positive affect, life satisfaction) and eudaimonic well-being (e.g., personal growth, purpose in life) at two waves 9-10 years apart. MetSyn components (waist circumference, systolic and diastolic blood pressure, triglycerides, HDL cholesterol, and fasting glucose) were measured during an overnight clinic visit at time two only. Outcomes included a count of components meeting clinical criteria and a binary outcome reflective of MetSyn status.

Results

The unadjusted prevalence of MetSyn at time two was 36.6%. Life satisfaction [B(SE)= -.12(.04), p=.005], positive affect [B(SE)= -.10(.04), p=.009], and personal growth [B(SE)= -.10(.04), p=.012] predicted fewer MetSyn components and also lower risk of meeting diagnostic criteria in fully adjusted models. Results did not change when adjusting for depressive sypmtoms nor were they moderated by age, gender, race, or socioeconomic status. Life satisfaction [B(SE) = -.11(.05), p = .023] and a eudaimonic well-being composite [B(SE) = -.11(.05), p = .045] also predicted fewer components and lower risk of meeting diagnostic criteria in longitudinal models.

Conclusions

Psychosocial resources, including positive affect, life satisfaction, and personal growth, predicted reduced risk for MetSyn both cross-sectionally and longitudinally. Further work should examine consequences of these linkages for cardiovascular outcomes in intervention contexts.

Growing research addresses the salubrious health effects of positive psychological functioning, which converges with the World Health Organization's view of health as a state of well-being and more than the absence of disease [1]. Evidence supports independent health benefits of psychological well-being, which is more than the absence of negative psychological functioning, such as depression, anxiety, or anger [2]. Indeed, well-being is a multidimensional domain that has been differentiated into related, but distinct, components. Hedonic well-being typically encompasses constructs such as positive affect, happiness, and life satisfaction. Eudaimonic well-being, in contrast, refers to evaluative judgments about people's lives, such as their sense of purpose and meaning and whether they perceive that personal talents and abilities are being realized [3,4]. Major reviews have synthesized the wealth of evidence linking both hedonic and eudaimonic well-being to optimal health, including in the realm of cardiovascular risk, morbidity, and mortality [5,6]. For example, purpose in life, a key component of eudaimonic well-being, has been found to prospectively predict both reduced risk of myocardial infarction [7] as well as reduced risk of stroke [8], independent of traditional cardiovascular risk factors. This work suggests that well-being predicts lower cardiovascular morbidity and mortality in both healthy and patient populations, independent of health behaviors and traditional risk factors.

Identifying links between well-being and reduced cardiovascular morbidity and mortality invites inquiry into the biological processes mediating such associations. Hypothesized mediators include autonomic and neuroendocrine regulation (hypothalamic pituitary adrenal axis and sympathetic nervous system activation), inflammation, and cardiometabolic functioning (e.g., obesity, blood pressure, cholesterol, and glucose regulation). In the current report, we focus on cardiometabolic functioning, assessed via metabolic syndrome, using a large sample of adults across five decades of age. Metabolic syndrome is a constellation of central obesity, hypertension, dysregulated lipids, and insulin resistance or hyperglycemia. Individuals with metabolic syndrome have increased risk for cardiovascular disease, stroke, and type II diabetes [9]. Metabolic syndrome has a high prevalence rate (over 30%) among US adults, with rates increasing in recent years [10]. There are several definitions of metabolic syndrome in the literature, including the National Cholesterol Education Program's Adult Treatment Panel III (ATP III), the World Health Organization, and the International Diabetes Foundation, which vary somewhat in the clinical cut points defining risk [9]. In the current study, our primary outcome was based on ATP III criteria given that this is the definition most commonly used for studies of psychological factors and metabolic syndrome [11].

Although no prior studies have examined links among hedonic or eudaimonic well-being with diagnostic metabolic syndrome, several have documented associations among well-being and individual components comprising metabolic syndrome. For example, HDL cholesterol has been linked with greater optimism in the present sample [12] as well as with greater positive affect, personal growth, and purpose in life in a sample of older women [13]. In the same sample of older women, positive affect further predicted lower levels of glycated hemoglobin over time [14]. Positive emotions were associated with lower rates of hypertension in a sample of older Mexican Americans [15], and happiness was inversely related to ambulatory blood pressure in the Whitehall psychobiology study [16]. A related positive psychological construct, perceived control, was cross-sectionally associated with higher HDL cholesterol and lower glycated hemoglobin and waist circumference in a national sample of middle aged and older Americans [17]. Finally, life satisfaction was inversely associated with excess weight in a community sample of adolescents and young adults [18]. In a recent review, emerging evidence supported associations among well-being and metabolic function, although limited evidence precluded drawing meaningful distinctions among different types of well-being [6]. Findings were also less consistent regarding associations among eudaimonic well-being and glucose regulation and body composition. Few studies have incorporated both hedonic and eudaimonic well-being measures, which is essential to empirically test their comparative effects [cf., 12,18–22]. Although hedonic and eudaimonic well-being measures are moderately correlated, they have previously demonstrated unique associations with central and peripheral health outcomes [4,6,13,23].

The aim of the current study was thus to examine associations between both hedonic and eudaimonic well-being with metabolic syndrome in a national sample of adults known as MIDUS (Midlife in the U.S.). In line with prior evidence, we hypothesized that both types of well-being, although capturing distinct components of positive psychological experience, would be associated with lower risk of metabolic syndrome. That is, both feeling good and being actively engaged in life, may predict reduced risk for cardiometabolic factors implicated in multiple disease outcomes. Our initial analyses focused on cross-sectional associations, but we augment the analyses with longitudinal associations between a subset of well-being assessments, measured 9-10 years earlier, and currently assessed metabolic syndrome.

Method

Sample

Participants were from the MIDUS survey, which included over 7,000 non-institutionalized adults in the first wave of data collection (1995-1996), recruited via random digit dialing (RDD) from the 48 contiguous states, siblings of the RDD sample, and a large sample of twins [24,25]. MIDUS I data collection went from January 1995- September 1996. Detailed information on the MIDUS I assessments and longitudinal retention are previously reported [22, 23]. The second wave (MIDUS II) began in 2004, with 75% of surviving respondents participating. Biological data were collected from a subset of MIDUS II respondents who agreed to travel to one of three General Clinical Research Centers (GCRC) for an overnight visit. MIDUS II survey data collection ran from January 2004- August 2005; biological data collection occurred between July 2004 and May 2009. There was a 43% response rate, reflective of the demanding protocol and extensive travel required for many participants [26]. The biological subsample was comparable to the full MIDUS II sample on most demographic and health characteristics, though was better educated and less likely to smoke than nonparticipants. Detailed information on the biological sample, protocol, and available measures are previously reported [26]. This study was approved by Institutional Review Boards at Georgetown University, University of California, Los Angeles, and University of Wisconsin, Madison. All participants provided written informed consent. Descriptive statistics by metabolic syndrome status are provided in Table 1.

Table 1. Descriptive statistics for study variables (N = 1,205).

	No Metabolic Syndrome (n = 764)		Metabolic Syndrome (n = 441)
Variable	M(SD) or %	Range	M(SD) or %	Range
Age	57.2(11.8)	35-86	58.0 (11.2)	37-85
Gender (% female)*	61.4		49.4
Race (% Black/African-American)	17.6		20.2
Education*
≤ High School (%)	26.1		31.4
Some College (%)	27.8		33.0
≥ College Degree (%)	46.1		35.7
Marital Status (% married)	63.2		67.7
Positive Affect*	3.7 (0.8)	1-5	3.6 (0.7)	1-5
Life Satisfaction*	7.8 (1.2)	2-10	7.6 (1.4)	2.75-10
M1 Life Satisfaction (n = 982)*	7.9 (1.1)	2.5-10	7.7 (1.2)	3.33-9.75
Autonomy	37.2 (6.7)	17-49	37.6 (6.6)	14-49
Environmental Mastery	38.6 (7.6)	11-49	38.0 (7.5)	12-49
Personal Growth*	40.1 (6.6)	14-49	38.5 (6.8)	18-49
Positive Relations with Others	40.7 (7.3)	7-49	40.4 (7.0)	9-49
Purpose in Life	39.8 (8.0)	15-49	39.0 (7.0)	10-49
Self-Acceptance*	38.9 (8.0)	7-49	37.7 (8.5)	10-49
M1 Well-being Composite (n =981)*	0.06 (0.6)	-2.8-1.3	-0.07 (0.7)	-1.83-1.2
Waist Circumference (cm)*	91.5 (15.1)	60-187	107.9 (12.8)	75-170
Waist circumference criteria (% yes)*	37.0		90.5
Systolic Blood Pressure (mm Hg)*	127.7 (18.1)	83-222	138.4 (16.0)	95-195
Diastolic Blood Pressure (mm Hg)*	74.2 (10.8)	48-125	78.1 (10.1)	51-114
Blood pressure criteria (% yes)*	41.6		77.1
HDL Cholesterol (mg/dL)*	61.7 (17.3)	24-121	44.9 (13.9)	19-103
HDL cholesterol criteria (% yes)*	10.5		61.0
Triglycerides (mg/dL)*	96.6 (43.9)	25-431	180.4 (95.3)	42-765
Triglycerides criteria (% yes)*	8.5		58.5
Glucose (mg/dL)*	96.4 (25.6)	56-418	111.8 (29.9)	67-335
Glucose criteria (% yes)*	19.7		71.8
# Met S symptoms*	1.2 (0.8)	0-2	3.6 (0.7)	3-5
Physical Activity (minutes/week)*	384.9 (643.2)	0-5,040	256.8 (495.0)	0-4,080
Alcohol Consumption (drinks/month)	14.5 (27.2)	0-278	13.4 (30.0)	0-405
Current Smoking (% yes)	14.8		14.7
Cholesterol Medication (% yes)*	23.6		36.7
Blood Pressure Lowering Meds (% yes)*	30.6		47.4
Glucose Lowering Meds (% yes)*	5.8		18.8

^*p < 05 when comparing individuals with and without metabolic syndrome by independent samples t-test or Chi-square tests.

The biological sample included 1,255 individuals. In order to examine race as a covariate, a small number of respondents were excluded (n = 50) who identified as a race other than White or Black or African-American; small cell sizes precluded investigating other racial or ethnic groups. Of the remaining respondents (n= 1,205), 379 were twins (51.7% monozygotic), and 6 were siblings. The sample size for longitudinal analyses was reduced, given that the MIDUS II sample had been expanded to include a city-specific sample of African Americans [n = 201; 24]. No prior well-being assessments were available for these African American respondents. Further, 28 respondents did not provide well-being data at MIDUS I. Therefore, the sample size for longitudinal analyses was 981, including 368 twins (51.4% monozygotic) and 6 siblings.

Measures

Well-being

All self-reported well-being scales were completed as part of the MIDUS I and II survey assessments. Eudaimonic well-being was based on Ryff's theoretical framework and included six scales: Autonomy, Environmental Mastery, Personal Growth, Positive Relations with Others, Purpose in Life, and Self-Acceptance [27,28]. The original scales each had 20 items, and other versions with 14 items per scale have been published [28–30]. At MIDUS II, each scale had seven items, and internal consistency ranged from .66 to .84. Well-being was also measured at MIDUS I, but with limited scales (3 items per scale), which had low internal consistency coefficients (.36 to .59). Thus, for tests of longitudinal associations among well-being and metabolic syndrome, we thus utilized a composite measure of well-being from MIDUS I by summing all individual items (18 in total). Assessed this way, internal consistency was .80 for the total eudaimonic well-being measure from MIDUS I.

Hedonic well-being was assessed with positive affect and life satisfaction. Positive affect was assessed by an average rating of how much of the time respondents felt, “enthusiastic,” “attentive,” “proud,” and “active” in the last 30 days on a 4-point scale (α = .85). These adjectives were derived from the Positive and Negative Affect Schedule [31]. Assessed this way, positive affect was only measured at MIDUS II; the same measure was not available at MIDUS I. To assess life satisfaction, respondents were asked to rate five dimensions of their lives, including their life overall, work, health, relationship with their spouse/partner, and relationships with their children, on a scale from 0 (worst possible) to 10 (best possible). The scores for relationship with spouse/partner and relationship with children were averaged to create one “item.” Our measure was calculated as the mean of this new item with the other three items, with higher scores reflecting greater overall life satisfaction [32]. Life satisfaction was assessed identically at MIDUS I and MIDUS II, and internal consistency was .67 at both time points.

Metabolic Syndrome

Metabolic syndrome was assessed at MIDUS II only. Metabolic syndrome was defined by the National Cholesterol Education Program: Adult Treatment Panel III definition [33]. Accordingly, participants were classified as meeting metabolic syndrome criteria when they had at least three out of the following risk factors: central obesity (defined as waist circumference > 102 cm for men or > 88 cm for women), triglycerides ≥ 150 mg/dL, HDL cholesterol < 40 mg/dL in men or < 50 mg/dL in women, blood pressure ≥ 130 mm Hg systolic or ≥ 85 mm Hg diastolic, and fasting plasma glucose ≥ 100 mg/dL. Waist was measured at the narrowest point between the ribs and iliac crest by GCRC staff. Blood pressure was assessed in a seated position three times consecutively with a 30 second interval between each measurement, and the two most similar readings were averaged. Participants rested for five minutes prior to the first blood pressure assessment. The lipid panel and glucose were assessed from a fasting blood sample taken on the morning of the second day of the GCRC visit (Roche Diagnostics, Indianapolis, IN).

We utilized two outcome variables for metabolic syndrome. The first was a count of components described above of which participants met criteria, ranging from 0 to 5. The second outcome variable was dichotomous, reflective of whether participants met the definition of metabolic syndrome [34,35].

Covariates

Covariates were measured as part of the MIDUS II survey and biological assessments. Demographic variables included age, gender, educational attainment (12-response category variable ranging from no education to professional degree; used continuously), race (coded to reflect White or Black/African-American only), and marital status (married v. all other). Health behavior variables, collected at the GCRC visit, included current smoking status, alcohol consumption over the previous month, physical activity (self-reported minutes per week of moderate and vigorous activity), and medication usage, including blood pressure lowering, cholesterol, or glucose-lowering medications.

Statistical Analyses

Hierarchical linear regression models were employed to test cross-sectional associations among well-being and metabolic syndrome components, and associations between well-being and diagnostic metabolic syndrome status were examined in hierarchical logistic regression models. Model 1 included demographic variables, including age, gender, education, race, and marital status, entered in the first step. The well-being measures were added in the second step of the regression, with each scale entered in respective models (i.e., 8 regressions total for the 8 well-being indicators). Model 2 included demographic and health covariates on the first step, and well-being was added in the second step in separate regression models for each well-being scale.

Preliminary analyses revealed that the linkages between well-being and both metabolic syndrome outcomes were not moderated by age, gender, educational attainment or race (p's > .10). All continuous variables were standardized as z-scores. Thus, coefficients reflect the change in metabolic syndrome risk for a one standard deviation increase in well-being. The alpha level was set to .05. Degrees of freedom varied slightly to reflect different degrees of missing data. No more than five individuals were missing data on any given variable other than race, and the sample size with complete data on all variables was 1,193 for cross-sectional analyses. Because the MIDUS sample includes siblings and twins, assumptions regarding independent observations are violated. Thus, we conducted supplemental analyses using GEE to adjust for biological dependencies in the data.

Identical models and covariates were used to test longitudinal associations among well-being (measured 9-10 years earlier at MIDUS I) and metabolic syndrome (measured at MIDUS II only). Appropriate measures of well-being from MIDUS II were included in the model. The sample size for the longitudinal analyses was reduced (n = 981), given missing data and that the MIDUS II sample had been expanded to include a city-specific sample of African Americans [n = 201; 26]. No prior well-being assessments were available for these African American respondents.

Results

Biological data to assess metabolic syndrome status were only available at time two. Respondents met criteria for two components of metabolic syndrome, on average, and metabolic syndrome prevalence was 36.6%. Descriptive information for individuals with and without metabolic syndrome is presented in Table 1. Table 2 presents bivariate correlations for study variables. Lower educational attainment, male gender, less physical activity, less alcohol consumption, and usage of blood pressure, cholesterol, or glucose lowering medication were associated with greater risk for metabolic syndrome in bivariate models. As would be expected, all individual components of the metabolic syndrome were correlated with metabolic syndrome status in the expected directions. MIDUS II positive affect and life satisfaction were moderately correlated with each other (r = .48) and with the eudaimonic well-being scales (r's: .20 - .54). Correlations among MIDUS II eudaimonic well-being scales ranged from .36 to .77. Of the metabolic syndrome components, well-being measures were consistently correlated with waist circumference, HDL cholesterol, and triglycerides, and less so with blood pressure and glucose.

Table 2. Bivariate correlations among study variables.

	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.
1. MetSyn Status	--
2. # MetSyn symptoms	.84*	--
3. Positive Affect	-.09*	-.08*	--
4. Life Satisfaction	-.11*	-.09*	.48*	--
5. M1 Life Satisfaction	-.09*	-.10*	.35*	.49*	--
6. Autonomy	.04	.02	.28*	.20*	.18*	--
7. Environmental Mastery	-.04	-.04	.51*	.53*	.40*	.49*	--
8. Personal Growth	-.13*	-.11*	.41*	.34*	.21*	.44*	.57*	--
9. Positive Relations with Others	-.02	-.02	.41*	.44*	.35*	.36*	.63*	.59*	--
10. Purpose in Life	-.09*	-.06*	.47*	.38*	.29*	.40*	.63*	.67*	.60*	--
11. Self-Acceptance	-.07*	-.07*	.54*	.53*	.35*	.49*	.77*	.63*	.67*	.69*	--
12. M1 Well-being Composite	-.08*	-.10*	.41*	.39*	.52*	.39*	.51*	.48*	.48*	.51*	.56*	--
13. Age	.09*	.04	.17*	.26*	.22*	.14*	.24*	.07*	.21*	.06*	.18*	.10*
14. Female Gender	-.14*	-.12*	-.01	-.00	-.01	-.09*	-.05	.11*	.13*	.05	-.01	-.01
15. Black/AA Race	.05	.03	.07*	-.21*	.00	.01	-.15*	-.09*	-.17*	-.04	-.11*	-.04
16. Education	-.14*	-.09*	.04	.12*	-.04	.07*	.15*	.22*	.11*	.14*	.20*	.12*
17. Married	.01	.05	.01	.22*	.17*	-.00	.14*	.07*	.18*	.14*	.16*	.17*
18. Waist Circumference	.61*	.48*	-.11*	-.13*	-.11*	.08*	-.04	-.15*	-.10*	-.10*	-.08*	-.07*
19. Systolic blood pressure	.42*	.29*	.05	.02	.04	.03	.07*	-.04	.07*	-.01	.01	.04
20. Diastolic blood pressure	.27*	.17*	-.03	-.10*	-.04	-.03	-.04	-.07*	-.05	-.03	-.06*	-.00
21. HDL Cholesterol	-.52*	-.45*	.10*	.06*	.08*	-.06*	.07*	.11*	.07*	.10*	.07*	.10*
22. Triglycerides	.65*	.55*	-.12*	-.12*	-.11*	.04	-.09*	-.11*	-.05	-.12*	-.10*	-.10*
23. Glucose	.44*	.36*	-.00	-.08*	-.02	.04	.01	-.02	-.01	.01	.00	-.01
24. Physical Activity	-.17*	-.15*	.09*	.13*	.09*	.06*	.13*	.14*	.08*	.13*	.12*	.11*
25. Alcohol Consumption	-.08*	-.06*	.02	-.03	.02	.01	-.01	.01	-.07*	-.00	-.01	.02
26. Current Smoking	.04	-.00	-.11*	-.22*	-.17*	-.01	-.16*	-.12*	-.17*	-.16*	-.19*	-.14*
27. Cholesterol Medication	.16*	.14*	.07*	.05	.03	.04	.05	-.03	.04	.02	.04	-.01
28. BP Lowering Meds	.22*	.17*	.00	-.02	.01	.03	.02	-.03	.06*	-.00	-.01	-.01
29. Glucose Lowering Meds	.23*	.21*	-.02	-.07*	-.03	.01	.02	-.04	-.02	-.02	.01	-.01

Note. Triglycerides, glucose, physical activity, and alcohol consumption were all log-transformed prior to calculating correlations in order to achieve normal distributions.

^*p ≤ .05

Metabolic Syndrome Components

Hierarchical linear regression models were used to examine cross-sectional associations between well-being and metabolic syndrome components (Table 3). In Model 1, the first step of the regression included demographic variables only, including age (B(SE) = .12(.04), t(1193) = 3.04, p = .002), female gender (B(SE) = -.42(.08), t(1193) = 5.17, p < .001), Black or African American race (B(SE) = .16(.11), t(1193) = 1.42, p = .16), educational attainment (B(SE) = -.19(.04), t(1193) = 4.75, p < .001), and being married (B(SE) = .02(.09), t(1193) = 0.20, p = .84). Together, demographic variables accounted for 5.1% of the variance in metabolic syndrome components. Well-being measures were added in the next step, with each scale entered in respective models. Adjusting for demographic variables, higher levels of both dimensions of hedonic well-being (i.e., life satisfaction and positive affect) and three dimensions of eudaimonic well-being (i.e., personal growth, purpose in life, and self-acceptance) significantly predicted fewer metabolic syndrome components (Table 3, Model 1). In Model 2, demographic factors, health behaviors and medication usage were added as covariates in the first step of regression models and well-being was added in the next step, with each scale entered in respective models. In fully-adjusted models, greater life satisfaction (t(1186)=2.83, p = .005), positive affect (t(1183)=2.62, p = .009), and personal growth (t(1181)=2.52, p = .012) remained significant predictors of fewer metabolic syndrome components, while purpose in life (t(1181)=1.86, p = .063) and self-acceptance (t(1181)= 1.82, p = .069) were attenuated (Table 3, Model 2).

Table 3. Linear regression models with well-being predicting number of metabolic syndrome components.

Variable	Model 1 Demographic				Model 2 Demographic + Health Covariates
	B	SE	p	ΔR2	B	SE	p	ΔR2
M2 Hedonic Well-being
Life Satisfaction	-.17	.04	<.001	.012	-.12	.04	.005	.006
Positive Affect	-.14	.04	<.001	.010	-.10	.04	.009	.005
M2 Eudaimonic Well-being
Autonomy	.03	.04	.44	.000	.05	.04	.24	.001
Environmental Mastery	-.06	.04	.12	.002	-.04	.04	.27	.001
Personal Growth	-.13	.04	.002	.008	-.10	.04	.012	.005
Positive Relations	.004	.04	.92	.000	.01	.04	.83	.000
Purpose in Life	-.10	.04	.016	.005	-.07	.04	.063	.003
Self-Acceptance	-.09	.04	.024	.004	-.07	.04	.069	.002

Note. All continuous variables were standardized as z-scores, and coefficients reflect a change in metabolic syndrome risk for a one standard deviation increase in well-being. Model 1 included age, race, gender, marital status, and education. Model 2 included Model 1 covariates plus smoking status, physical activity, alcohol consumption, and usage of cholesterol, blood pressure, and glucose lowering medications. The ΔR² values reflect the amount of additional variance in metabolic syndrome components accounted for by well-being above and beyond the demographic factors (Model 1) and the demographic factors and health covariates together (Model 2), respectively.

Metabolic Syndrome Status

Hierarchical logistic regression models were used to examine cross-sectional associations between well-being and diagnostic metabolic syndrome (Table 4). In Model 1, the first step of the regression included demographic variables only, including age (B(SE) = .07(.06), Wald = 1.21, p = .27), female gender (B(SE) = .49(.13), Wald = 15.39, p < .001), Black or African American race (B(SE) = -.20(.17), Wald = 1.28, p = .26), educational attainment (B(SE) = -.20(.06), Wald = 10.15, p = .001), and being married (B(SE) = -.19(.14), Wald = 1.80, p = .18). Together, demographic variables accounted for 3.6% of the variance in metabolic syndrome status. Well-being measures were added in the next step, with each scale entered in respective models. In models adjusting for demographic factors, life satisfaction, positive affect, personal growth, and self-acceptance were significant predictors of lower risk of meeting metabolic syndrome criteria (Table 4, Model 1). In Model 2, demographic factors, health behaviors and medication usage were added as covariates in the first step of regression models and well-being was added in the next step, with each scale entered in respective models. In fully-adjusted models, the association between self-acceptance and metabolic syndrome was attenuated (Wald = 3.69, p = .055), while the associations between life satisfaction (Wald = 5.53, p = .019), positive affect (Wald = 5.33, p = .021), and personal growth (Wald = 5.38, p = .020) remained significant (Table 4, Model 2).

Table 4. Logistic regression models with well-being predicting metabolic syndrome diagnosis.

	Model 1 Demographic					Model 2 Demographic + Health Covariates
Variable	B	SE	OR	95% CI	ΔR2	B	SE	OR	95% CI	ΔR2
M2 Hedonic Well-being
Life Satisfaction	-.21**	.07	0.81	[0.71, 0.92]	.012	-.16*	.07	0.85	[0.75, 0.97]	.006
Positive Affect	-.19**	.06	0.83	[0.73, 0.93]	.011	-.15*	.07	0.86	[0.76, 0.98]	.006
M2 Eudaimonic Well-being
Autonomy	.04	.06	1.04	[0.92, 1.17]	.001	.07	.07	1.07	[0.94, 1.21]	.001
Environmental Mastery	-.09	.06	0.91	[0.81, 1.04]	.002	-.07	.07	0.93	[0.82, 1.07]	.001
Personal Growth	-.18**	.06	0.83	[0.74, 0.94]	.009	-.15*	.07	0.86	[0.76, 0.98]	.006
Positive Relations	-.002	.07	1.00	[0.88, 1.13]	.000	.01	.07	1.01	[0.88, 1.15]	.000
Purpose in Life	-.09	.06	0.91	[0.81, 1.03]	.003	-.07	.07	0.93	[0.82, 1.06]	.001
Self-Acceptance	-.14*	.06	0.87	[0.77, 0.98]	.006	-.13†	.07	0.88	[0.77, 1.00]	.004

Note. All continuous variables were standardized as z-scores, and coefficients reflect a change in metabolic syndrome risk for a one standard deviation increase in well-being. Model 1 included age, race, gender, marital status, and education. Model 2 included Model 1 covariates plus smoking status, physical activity, alcohol consumption, and usage of cholesterol, blood pressure, and glucose lowering medications. The ΔR² values reflect the change in Nagelkerke R² values between regression blocks with covariates only and blocks with covariates and well-being. This approximates the additional variance in metabolic syndrome status accounted for by well-being above and beyond the demographic factors (Model 1) and the demographic factors and health covariates together (Model 2), respectively.

^**p < .01,

^*p < .05,

^†p < .10.

Role of Depressive Symptoms

To assess whether ill-being was affecting aforementioned results, depressive symptoms (assessed with the CES-D [36]) were added to models with well-being factors that were significant in fully adjusted models. CES-D scores in this sample ranged from 0-54, M(SD)=8.6(8.2). Bivariate correlations between well-being and depressive symptoms ranged from 0.24-0.53, p's < .001. Life satisfaction and positive affect remained significant predictors of both outcomes with depressive symptoms included in fully adjusted models (components: positive affect: t(1175) = 2.42, p = .016, life satisfaction: t(1178) = 2.50, p = .012; status: positive affect: Wald = 4.19, p = 0.041, life satisfaction: Wald = 3.81, p = .051). Personal growth also remained a significant predictor of both outcomes (components: t(1173) = 2.28, p = .023; status: Wald = 3.84, p = .050).

Independence among Well-being Measures

To assess the relative independence among hedonic and eudaimonic well-being, additional models were run that controlled for the other variety of well-being. When a eudaimonic well-being composite was included in models with positive affect and life satisfaction, respectively, these hedonic measures remained significant predictors of metabolic syndrome. In fully adjusted models, positive affect significantly predicted metabolic syndrome components (B(SE) = -0.12(0.05), p =.012) and status (OR(95% C.I.) = 0.84 (0.72, 0.98), p = .023) as did life satisfaction (components: B(SE) = -0.13(0.05), p =.005); status (OR(95% C.I.) = 0.84 (0.72, 0.98), p = .024). Associations among personal growth and metabolic syndrome were attenuated when positive affect was included as an additional control (components: (B(SE) = -0.07(0.04), p =.13; status: OR(95% C.I.) = 0.90 (0.78, 1.04), p = .16).

Longitudinal Analyses

Table 5 presents longitudinal associations among well-being and both metabolic syndrome outcomes. Life satisfaction was assessed identically at MIDUS I and MIDUS II. Life satisfaction at MIDUS I correlated with life satisfaction at MIDUS II at r = 0.50, p < .001. In line with the cross-sectional analyses, life satisfaction at MIDUS I significantly predicted both metabolic syndrome outcomes 9-10 years later controlling for MIDUS II life satisfaction, demographic, and health covariates. The eudaimonic well-being composite at time one correlated with eudaimonic well-being composite at time two at r = 0.60, p < .001. The eudaimonic well-being composite from MIDUS I also significantly predicted number of metabolic syndrome components and metabolic syndrome status in fully-adjusted models (controlling for MIDUS II eudaimonic well-being, demographic factors, and health covariates).¹

Table 5. Longitudinal models with well-being predicting metabolic syndrome outcomes.

	DV: Metabolic Syndrome Components						DV: Metabolic Syndrome Status
	Model 1 Demographic			Model 2 Demographic + Health Covariates			Model 1 Demographic				Model 2 Demographic + Health Covariates
Variable	B	SE	p	B	SE	p	B	SE	OR	95% C.I.	B	SE	OR	95% C.I.
M1 Life Satisfaction	-.13	.05	.015	-.11	.05	.023	-.20*	.08	0.82	[0.70, 0.96]	-.20*	.08	0.82	[0.69, 0.97]
M2 Life Satisfaction	-.11	.06	.058	-.04	.06	.52	-.14	.09	0.87	[0.73, 1.04]	-.05	.09	0.96	[0.80. 1.15]
M1 Eudaimonic Composite	-.12	.06	.025	-.11	.05	.045	-.26**	.09	0.77	[0.65, 0.92]	-.26**	.09	0.77	[0.65, 0.93]
M2 Eudaimonic composite	.01	.06	.92	.02	.06	.71	.06	.09	1.06	[0.88, 1.27]	.08	.10	1.09	[0.90, 1.32]

Note. All continuous variables were standardized as z-scores, and coefficients reflect a change in metabolic syndrome risk for a one standard deviation increase in well-being. Model 1 included well-being at MIDUS II, age, race, gender, marital status, and educational attainment. Model 2 added current smoking status, physical activity, alcohol consumption and usage of cholesterol, blood pressure, or glucose lowering medications.

^**p < .01,

^*p < .05.

Data Dependencies

Because the MIDUS sample includes a considerable number of siblings of the RDD sample and twins (37%), assumptions of independent observations are violated. To address these data dependencies, supplemental analyses employed generalized estimating equations (GEE) models with random intercepts for family clusters. The within-cluster covariance structure was specified as exchangeable. All conclusions drawn from reported results remained identical to those presented above, supporting that biological dependencies in the data did not bias results.

Discussion

This was the first study to examine associations between both hedonic and eudaimonic well-being and metabolic syndrome in a national sample of adults. Previous research identified well-being as prospectively predictive of lower cardiovascular morbidity and mortality, but evidence linking well-being to intermediate biological processes has been limited in number and scope [6]. Metabolic syndrome represents a potential biological mediator, and this study provided an important test of associations among multiple varieties of well-being with metabolic syndrome. Results from the current study demonstrated that several dimensions of well-being predicted lower risk of metabolic syndrome in cross-sectional and longitudinal models.

Specifically, after adjustments for sociodemographic factors, hedonic indicators of life satisfaction and positive affect, as well as eudaimonic indicators of purpose in life, personal growth, and self-acceptance were all significant predictors of lower metabolic syndrome risk in cross-sectional models. Importantly, none of these sociodemographic factors functioned to moderate the associations among well-being and metabolic syndrome. Additional adjustments for health covariates, including health behaviors and medication usage, attenuated associations between self-acceptance and purpose in life, but all other aforementioned associations remained significant in fully-adjusted models. Life satisfaction and the eudaimonic well-being composite also predicted lower risk of metabolic syndrome status in fully-adjusted, longitudinal models. Of the metabolic syndrome components, associations were strongest among well-being and waist circumference, HDL cholesterol, and triglycerides. Thus, body composition and diet, as opposed to glucose metabolism directly, may be the most relevant targets of well-being interventions.

In order to examine the relative independence of these associations, additional models included depressive symptoms as well as both varieties of well-being included together. All associations with well-being and metabolic syndrome remained significant with depressive symptoms included in the model. These results coincide with considerable evidence supporting well-being and distress as separate dimensions, and not simply two ends of the same continuum [4,37]. Further, both hedonic well-being measures remained significant predictors of metabolic syndrome with a eudaimonic composite in the model, supporting the distinctiveness among hedonic and eudaimonic well-being [3]. However, associations among personal growth and metabolic syndrome were attenuated when positive affect was included as an additional control, suggesting that positive affect is implicated in the salubrious associations seen with personal growth. Personal growth reflects a sense of self-improvement, continued development, and realization of one's potential, which could lead to feelings of high positive affect. This observation calls for greater research on how various aspects of hedonic and eudaimonic well-being work together to contribute to better health outcomes. So doing will require studies that incorporate both types of assessment in the same investigation so as to investigate their individual and joint effects. This is one of a few studies that incorporate both hedonic and eudaimonic dimensions in the same manuscript, which is critical to examine their relative contributions to health markers [cf., 12,18–22].

With regard to clinical implications, several promising interventions exist to improve well-being. Specifically “well-being therapy” has been successful at reducing recurrence of major depression [38] and generalized anxiety disorder [39,40]. Other interventions that increase hedonic well-being further show reductions in visits to student health facilities [41]. Early randomized controlled trials have been effective at increasing dimensions of eudaimonic well-being, specifically purpose in life, among cancer patients [42,43]. Hedonic and eudaimonic well-being have been linked to healthier brain functioning, including prefrontal activation asymmetries [23], sustained activity in reward circuitry following positive stimuli [44], and faster recovery from negative emotional stimuli [45] as well as cortisol regulation [13,44], which likely constitute additional mechanisms underlying the health-promoting effects of well-being. Individuals with high well-being generally report lower rates of smoking, less abuse of alcohol, healthier diet, and more leisure time physical activity [6]. These health behaviors are implicated in the pathogenesis of metabolic syndrome [46]. In sum, well-being is modifiable, and such interventions may yield important physical health benefits.

Several study limitations warrant mention. Of primary concern is the lack of biological assessments at MIDUS I, precluding the testing of truly longitudinal relationships. Therefore, causality cannot be determined due to a lack of time-ordering among the predictor and outcome variables. We do, however, note stability in life satisfaction and the eudaimonic well-being composite over the 9-10 year interval. Further, when self-reported chronic conditions at baseline were included as an additional control variable, both longitudinal well-being measures remained significant predictors of metabolic syndrome (data not shown), which attenuates, but does not eliminate, concerns that healthier individuals reported higher well-being at baseline, explaining the observed reduced risk of metabolic syndrome at follow up. Secondly, there was limited representation of individuals from racial and ethnic minority groups, with the exception of city-specific sample of African Americans from Milwaukee, WI in cross-sectional models. Thus, it is unknown whether these results generalize to a more representative sample of African Americans or to other racial and ethnic groups. Another limitation involved dissimilar assessments of well-being at MIDUS I and MIDUS II. A composite of eudaimonic well-being and life satisfaction were the only well-being measures with identical assessments at both time points, though we note that associations were largely similar in cross-sectional and longitudinal models. Internal consistency of the life satisfaction measure at MIDUS I and MIDUS II was relatively low (alpha = .67), likely reflective of the multiple domains assessed with our measure (i.e., health, relationships, work, life overall). Further, the assessment of positive affect only tapped into high activation states, and thus it is unknown whether the same associations with metabolic syndrome would emerge if low or medium activation states were assessed. Prospective analyses that can replicate and extend the current results represent an important avenue for future work. Finally, effect sizes were relatively small, with well-being accounting for 1-2% of the variance in metabolic syndrome outcomes. However, the magnitude of these associations is similar to that seen with age and educational attainment in this sample, which are both recognized as important risk factors for metabolic syndrome [9].

Despite these limitations, this research incorporated a comprehensive formulation of well-being, including its distinct hedonic and eudaimonic varieties. We also incorporated an objectively assessed outcome with important public health implications, namely metabolic syndrome, and for the first time demonstrated that both hedonic and eudaimonic well-being contribute to this index of cardiometabolic risk. Finally, the study questions were investigated in a large sample of sociodemographically heterogeneous participants, including participants' ages spanning five decades. Findings supported modest protective effects of well-being for metabolic syndrome in this sample, providing support for metabolic syndrome as a biological mediator of the links between well-being and cardiovascular morbidity and mortality.

Abbreviations

AA

African American

ATP III

Adult Treatment Panel III

BMI

body mass index

HDL

High density lipoprotein

MetSyn

Metabolic Syndrome

M1

MIDUS I

M2

MIDUS II

MIDUS

Midlife in the United States