at SciVerse ScienceDirect

Social Science & Medicine 77 (2013) 75e83

Contents lists available

Social Science & Medicine

journal homepage: www.elsevier .com/locate/socscimed

How does race get “under the skin”?: Inflammation, weathering, and metabolicproblems in late life

Aniruddha Das*

Department of Sociology, McGill University, Room 712, Leacock Building, 855 Sherbrooke Street West, Montreal, Quebec H3A 2T7, Canada

a r t i c l e i n f o

Article history:Available online 14 November 2012

Keywords:USARace disparitiesBlackWeatheringBlood sugarDiabetesCardiovascularInflammation

* Tel.: þ1 514 398 4582; fax: þ1 514 398 3403.E-mail address: aniruddha.das@mcgill.ca.

0277-9536/$ e see front matter � 2012 Elsevier Ltd.http://dx.doi.org/10.1016/j.socscimed.2012.11.007

a b s t r a c t

Using nationally representative data from the 2005e2006 U.S. National Social Life, Health, and AgingProject, this study queries the mechanisms underlying worse metabolic outcomesdblood-sugar controland cardiovascular healthdamong black than white men ages 57e85. Results indicate that contrary tomuch of the academic literature as well as media accountsdimplicitly rooted in a “culture of irre-sponsibility” modeldolder black men’s social isolation, poor health behaviors, or obesity may not playa major role in their worse metabolic problems. Instead, these outcomes seem to derive more consis-tently from a factor almost unexamined in the literaturedchronic inflammation, arguably a biological“weathering” mechanism induced by these men’s cumulative and multi-dimensional stress. Thesefindings highlight the necessity of focusing attention not simply on proximal behavioral interventions,but on broader stress-inducing social inequalities, to reduce men’s race disparities in health.

� 2012 Elsevier Ltd. All rights reserved.

Introduction

An established literature documents greater diabetic andcardiovascular problems among black men at any age, andespecially in late life (Heron, 2007). Recent biomedical studiessuggest a syndromal clustering of these two metabolicconditionsdsuggesting common, possibly social, antecedents(Goossens, 2008; Grundy, Brewer, Cleeman, Smith, & Lenfant, 2004;Yudkin, 2003). The sparse sociological literature on underlyingmechanisms generally examines prevalence of diagnosed disease(Dupre, 2008; Morenoff et al., 2007)dproblematic given commonunderdiagnosis of these issues among minority groups (Geiss et al.,2006; Timmermans & Haas, 2008). Few studies directly examineblood-sugar control and cardiovascular health, and comparemultiple mechanisms. Moreover, both the academic literature andnews media frequently posit obesitydconceived as a result of poorsocial control and unhealthy behaviorsdas the “biologicalgateway” through which social and behavioral mechanisms affectthese distal metabolic outcomes. Recent studies on stress-physiology linkages, however, suggest other biosocial pathways(Geronimus, 2001; McDade, 2005). Due to a lack of large-samplesurveys containing requisite social as well as biological indicators,these mechanisms remain underexamined.

All rights reserved.

Using data from the 2005e2006 U.S. National Health and SocialLife Project (NSHAP)da nationally representative probabilitysample of adult Americans aged 57e85dthe present study beginsto fill these gaps. Specifically, it examines and compares a series ofsocial, behavioral, stress-process, and biological mediators of theimpact of race on men’s blood-sugar control and cardiovascularhealth.

Race and metabolic problems

Four potential mechanisms can be extracted from the socio-logical and biomedical literatures, for older black men’s worsemetabolic outcomes: (1) social support and/or control; (2) healthbehaviors; (3) obesity, possibly as a consequence of social andbehavioral factors; and (4) a stress-inflammation chain.

Social support and control

An established literature indicates the health benefits of moreand stronger social tiesdwhether through more control ofunhealthy behaviors, or through emotional and instrumentalsupport, including caregiving (Coleman,1988; Kawachi, Kennedy, &Glass, 1999). These control and support mechanisms can workthrough multiple features of one’s egocentric networkdstartingwith network size, which may simply index a greater number ofsocial alters invested in one’s health. Similarly, the likelihood ofsuch investments arguably increases with the strength of the social

A. Das / Social Science & Medicine 77 (2013) 75e8376

relationship. Multiple studies indicate that especially at older ages,extra-family networks shrink, with retirement and the curtailmentof social activities due to functional or health problems(McPherson, Smith-Lovin, & Brashears, 2006; Schnittker, 2007). Inthis life stage, strong ties are most likely with one’srelativesdespecially those related by blooddwho may becomecrucial sources of health-monitoring and caregiving (Shaw, Krause,Liang, & Bennett, 2007). Hence, networks richer in consanguinealor blood relatives arguably have more control and support poten-tial. Finally, studies have long emphasized the importance of“closure”dthe degree to which network members areinterconnecteddin enhancing social control, by allowing alters tocombine forces to sanction a focal individual’s adverse behaviors(Coleman, 1988).

Sociological literature is more ambiguous about men’s racedifferentials in these health-producing social assets. Some studiesindicate greater social isolation among black mendwith accountsof underlying mechanisms reflecting an implicit moralistic narra-tive of these men’s non-engagement in social and especiallyfamilial roles (Burton & Snyder, 1998; Jarrett, Roy, & Burton, 2002).In contrast, a separate literature suggests that black men may bemore strongly embedded than their white counterparts in denselyinterconnected network structures, centered in church member-ship and the matriarchal kin group (Hill, 1999; Sarkisian & Gerstel,2004). Consistent with the isolation argument, Cornwell, Schumm,and Laumann (2008) combined-gender NSHAP study (additivelyincluding gender and ethnicity) finds smaller networks among allblack participants. However, it also finds few race differentials inother egocentric network attributesdwith existing differentialsgenerally favoring black participants. Finally, if race associations aregender-differentiateddwith black men having worse and blackwomen better social tiesdcombined analyses may possibly maskthe former deficits.

To test these competing arguments about the role of older blackmen’s social deficits in mediating their worse diabetic and cardio-vascular outcomes, Hypothesis 1 (Table 1) is included.

Behaviors, obesity, inflammation: culture or stress?

Next, an established literature suggests that blackmenmay haveworse health behaviors (Williams, 2001; Williams & Collins, 1995).In turn, diabetic and cardiovascular outcomes have been linked toboth actively risky behaviors such as smoking (Ockene & Miller,1997) and alcoholism (Lakka et al., 2002)das well as passive onesincluding too much (Wilson, 2005) or too little sleep (Cappuccioet al., 2007; Miller et al., 2009), and inactivity (Haskell et al.,2007). A final health-related behavior that might affect theseconditions is underutilization of healthcare, which studies suggestis particularly common in black communitiesdarguably due to lowsocial compatibility with and/or trust in medical practitioners (vanRyn, 2002). These conjectures lead to Hypothesis 2 (Table 1).

Moreover, especially in the sociological literature, many of thesebehaviors are assumed to flow through obesity. Whether due tomore cultural acceptance of weight (Bennett & Wolin, 2006),

Table 1Summary of hypotheses.

Relative to their white counterparts, older black men’s worse diabetic andcardiovascular outcomes will be mediated by their:1. worse social deficitsdindicated by smaller networks, with less close ties,

less closure, and smaller proportions of blood relatives.2. worse health behaviorsdboth active (smoking, alcohol consumption) and

passive (inactivity, poor sleep).3. greater obesity.4. worse inflammation, due to greater chronic stress.

neighborhood cultural patterns (Boardman, Saint Onge, Rogers, &Denney, 2005), or fatalism about health (Plowden, 2003), blackindividuals’ obesity-inducing behaviors have been presented asa major cause for their worse health outcomes. Or in other words,obesity has implicitly or explicitly been portrayed as the “biologicalgateway” through which behavioral mechanisms affect metabolicproblems. Indeed, the “obesity epidemic” in recent decades hasbeen depicteddin both the news media and academicliteraturedas key to increases in these chronic conditions (Campos,Saguy, Ernsberger, Oliver, & Gaesser, 2006; Saguy&Almeling, 2008).

Taken together, these conceptions arguably represent an implicit“culture of irresponsibility” model that depicts black men’sunhealthier social and behavioral propensities as distal antecedentsthat, acting partly through obesity, lead to their greater morbidityand mortality (Oliver, 2006; Saguy & Gruys, 2010). Consistent withthis notion, an extensive literature does link worse health behaviors(diet, inactivity) with obesity, and documents the causal role ofobesity in blood-sugar control, diabetes, and heart disease (Grundyet al., 2004; Yudkin, 2003). However, while a range of studies indi-cates greater obesity among black individuals of all ages (Denney,Krueger, Rogers, & Boardman, 2004; Mokdad et al., 2003), recentevidence suggests a gender-differentiated patterndwith black menless obese than their white counterparts, and the reverse holding forwomen (Chang & Lauderdale, 2005). Moreover, some biomedicalliterature also suggests that obesity, in itself, may not be asresponsible for metabolic problems as previously thought (Abbasi,Brown, Lamendola, McLaughlin, & Reaven, 2002; Reaven, 1995).Accordingly, Hypothesis 3 (Table 1) is included to help adjudicatebetween these conflicting accounts of obesity’s mediatory role.

Rather than race-specific cultural patterns, moreover, metabolicpathologies may also derive from older black men’s greater stress-induced “weathering” (Geronimus, 2001; Geronimus, Hicken,Keene, & Bound, 2006). Weathering refers to the cumulativehealth impact of black individuals’ repeated experiences withsocial, economic, or political exclusiondwith the presence of thesestressors in multiple life dimensions, and consequent high-effortcoping, potentially inducing morbidity both directly and throughunhealthy behaviors. Thus, an extensive literature links stress tomultiple behavioral issues, including smoking (Feldner, Babson, &Zvolensky, 2007; Koenen et al., 2005), alcohol use (Davis, Uezato,Newell, & Frazier, 2008; Grant & Harford, 1995), poor sleep(American Psychiatric Association, 2000) as well as physical inac-tivity (Verger, Lions, & Ventelou, 2009)dfactors, as noted above,linked to metabolic problems. More directly, studies indicate racedisparities in “allostatic load”dmulti-systemic physiological wear-and-tear through long-term exposure to stress-induced fluctua-tions or elevations in neuroendocrine response (Geronimus et al.,2006; McEwen, 1998; Singer, Ryff, & Seeman, 2004; Sterling &Eyer, 1981). Finally, recent biodemographic literature suggeststhat low-grade chronic inflammation, triggered by prolongedexposure to stressful environments, may be a key mechanismthrough which weathering works, and produces allostatic load.Multiple studies indicate a pathway from extended psychosocialstress to inflammation (McDade, Hawkley, & Cacioppo, 2006;Melamed, Shirom, Toker, Berliner, & Shapira, 2006; Weinstein,Vaupel, & Wachter, 2007)dpossibly due to norepinephrine-driven gene expression of inflammatory mediators (Bierhauset al., 2003; Kiecolt-Glaser, Gouin, & Hantsoo, 2010). Accordingly,C-reactive protein (CRP), an inflammation indicator, is increasinglybeing used as a biomarker of chronic stressdwhether in the BritishWhitehall II studies (Owen, Poulton, Hay, Mohamed-Ali, & Steptoe,2003), Chicago (McDade et al., 2006), or in the United States (Alleyet al., 2006). Downstream, inflammation has a demonstratedcausative role in cardiovascular problems, as well as poor blood-sugar control due to insulin resistance (Grundy et al., 2004; Yudkin,

Table 2Descriptive statistics for variables used in analyses.

Variable Mean (SE) Percentage Rangea N

Race/ethnicityWhite (non-Hispanic)b 0.81 (0.02) 80.78% 0e1 1451Blackb 0.09 (0.01) 9.17% 0e1 1451Hispanic/otherb 0.10 (0.02) 10.04% 0e1 1451

Control variablesAgec 67.55 (0.28) 57e85 years 1455Less educationd 2.26 (0.05) 1e4 1455Diabetesb 0.22 (0.01) 21.51% 0e1 1455Hypertensionb 0.53 (0.02) 53.43% 0e1 1455No healthcare

accessb0.09 (0.01) 9.08% 0e1 1344

DepressionDepressedb 0.16 (0.01) 15.92% 0e1 1436

Social deficitsFewer altersd 2.74 (0.05) 1e6 1455Less close tiesc 1.88 (0.01) 1e4 1409Less network

closurec0.54 (0.01) 0e1 1410

Smaller kinproportionc

0.64 (0.01) 0e1 1411

Poor health behaviorsPoor sleepb 0.39 (0.02) 38.96% 0e1 1450Inactivityd 1.67 (0.03) 1e5 1453Smokerb 0.21 (0.02) 20.98% 0e1 1185More alcohold 1.37 (0.07) 0e7 1452Binge drinkingd 0.38 (0.02) 0e2 1444Less healthcare

utilizationd4.52 (0.05) 0e7 1346

Biological mediatorsObesityc 40.64 (0.20) 27.00e66.00

inches1412

Inflammationc,e 0.89 (0.03) 0.00e2.25log scale

879

Metabolic outcomesBlood-sugar

controlc1.80 (0.01) 1.44e2.65

log scale839

Systolic BPc 137.13 (0.45) 85.00e212.67mm Hg

1424

Diastolic BPc 80.45 (0.36) 44.50e127.67mm Hg

1424

Heart ratec 70.50 (0.46) 40.00e133.33beats/min

1422

Note: all values are specific to men. Italicization denotes reference category insubsequent analyses. All estimates are weighted to account for differential proba-bilities of selection and differential nonresponse. Design-based standard errors aregiven in parentheses. For dummy variables, both means and percentages areprovided, as baseline information. Subsequent analyses use a standard 0/1 codingfor these measures.

a Range specific to men.b Dummy variable.c Continuous variable.d Ordinal variable.e Observations of CRP greater than 8.6 mg/l excluded from analysis.

A. Das / Social Science & Medicine 77 (2013) 75e83 77

2003). More specifically, recently developed high-sensitivity CRPassays suggest that chronic, low-grade inflammation (indicated bylower CRP levels) is linked to subsequent incidence of cardiovas-cular disease (Danesh et al., 2000), type 2 diabetes (Pradhan,Manson, Rifai, Buring, & Ridker, 2001), and the “metabolicsyndrome” (McDade & Hayward, 2009; Ridker, Buring, Cook, &Rifai, 2003). Moreover, in their combined-gender NSHAP study(additively including gender and ethnicity), McDade, Lindau, andWroblewski (2011) find higher inflammation/CRP among olderblack than white adults. Thus, in addition to obesity, inflammationrepresents a seconddand underexaminedd“biological gateway”through which older black men’s social environments may affecttheir diabetic and cardiovascular outcomes.

Apart from stress, however, inflammation has also been linkedto obesity (Yudkin, 2003), smoking and heavy alcohol consumption(Chiu et al., 2008; Mukamal, 2006), and lack of sleep (Meier-Ewertet al., 2004). Much of this evidence comes from cross-sectionalstudies, leaving ambiguity over whether inflammation is a conse-quence of these factorsdor, as a long-term stress marker, a cause. Itis argued, therefore, that inflammation represents a stress-inducedbiological weathering mechanism to the extent that the pathwaysthrough which it mediates race differentials in metabolic problemsdo not include obesity and/or health behaviors as causal anteced-ents. To test this potential mediatory role of inflammation,Hypothesis 4 (Table 1) is included.

To summarize, then, the four mechanisms drawn from the liter-ature fall into two categories. A first set of implicitly moralisticapproaches sees men’s blackewhite differentials in diabetic andcardiovascular health as stemming from social deficits due to disen-gagement from social and family roles; from unhealthy behaviors asa function of race-specific cultural patterns; and (consequently) fromobesity. A second emphasizes long-term biological weatheringinduced by older black men’s cumulative and multi-dimensionalstressdwhether in directly causing metabolic pathologies viachronic inflammation; or through unhealthy behaviors.

Methods

Data

Data are from the 2005e2006 U.S. National Social Life, Health,and Aging Project. NSHAP is a nationally representative probabilitysample of 1550 women and 1455 men aged 57e85, with an over-sampling of blacks, Hispanics, men, and those 75e85. In addition toself-reports, data include assessments of physical function, heightand weight, as well as salivary and blood samplesdall collected atthe time of interview by non-medically trained interviewers. Thesurvey has an unweighted response rate of 74.8% and a weightedresponse rate of 75.5% (O’Muircheartaigh & Smith, 2007). Partici-pant consent was obtained prior to interview. Institutional reviewboards at the Division of the Social Sciences and NORC at theUniversity of Chicago approved data collection procedures.

Measures

Table 2 presents summary statistics for all variables used in theanalyses.

Predictors and controlsA participant’s race is indexed by a set of dummy variables for

black and Hispanic/other, with non-Hispanic White as the referencecategory. Seventy percent of men in the Hispanic/other category arenon-black Hispanics, with the remainder comprised of AmericanIndians or Alaskan natives, Asian or Pacific Islanders, and “other.”All analyses linearly control one’s age (in years). Models including

less healthcare utilization also add no healthcare access, a dummyself-report indicating that a participant has no place to go whensick. Finally, Tables 4 and 5 include controls for less educationdaninteger score ranging from to 1 (Bachelors degree ormore) to 4 (lessthan a high school education)dand any lifetime diagnosis of dia-betes or hypertension by a medical doctor.

DepressionA current depressed state is used to validate stress-inflammation

linkages in preliminary analysis. NSHAP depression items are

A. Das / Social Science & Medicine 77 (2013) 75e8378

derived from the Iowa form of the Center for EpidemiologicalStudies Depression Scale (CES-D; Shiovitz-Ezra, Leitsch, Graber, &Karraker, 2009), with each 4-category Likert item indexing expe-riences over the preceding week of one of 11 depressive states.Those with scores above 9.0 on the full additive scale are catego-rized as depressed (Zauszniewski & Graham, 2009).

MediatorsFour attributes of a participant’s egocentric social network tap

deficits in support and monitoring/control. All are based onnominations of alters during the network-roster portion of theface-to-face interview (Cornwell et al., 2008). First, fewer alters isa simple reverse-coded count of one’s network members, and runsfrom 1 (5 ormore) to 6 (no alters). The next three variables all rangecontinuously from 0 to 1, with higher values denoting worse ties.Less close ties indicates a participant’s average (self-reported) lack ofcloseness with his/her alters. Less network closure is proxied by theaverage (participant-reported) frequency of contact among allmembers of one’s networkdincluding the frequency of contactamong nominated alters (in alterealter dyads) as well as betweena participant and each alter (in ego-alter dyads). Finally, smaller kinproportion is indexed by the proportion of one’s alters who areconsanguineal kin (i.e., related by blood).

Measures of poor health behaviors include poor sleepdbased onself-reports of “usually” sleeping less than 7 or more than 9 ha night. Physical inactivity is measured by one’s self-reportedfrequency of participationdon a regular basisdin activities suchas walking, dancing, gardening, physical exercise or sports. Thisvariable ranges from 1 (3 or more times per week) to 5 (never).Next, studies indicate race differences in underreports of smokingin large-sample surveys (Fisher, Taylor, Shelton, & Debanne, 2008).Hence, a participant’s smoking habits are measured through hiscotinine levels (Drum, Shiovitz-Ezra, Gaumer, & Lindau, 2009).Cotinine is a metabolite of nicotine, derived from saliva samplestaken during the biomeasure collection portion of the interview.Based on the Wells-Stewart method (Wells, 1993), men with coti-nine greater than 100 ng/ml are classified as regular smokers. Next,a participant’s alcohol consumption is indexed by two ordinal self-reports. More alcohol indicates average weekly drinking frequency,and runs from 0 (none) to 3 (daily or more). Binge drinking indicates

Table 3Associations of race with social, behavioral, and biological factors among U.S. men aged

Social deficits

Fewer altersa Less close tiesb Less networkclosureb

Smallepropo

Race (ref: non-Hispanic White)Black 0.51** (0.15) �0.02 (0.04) �0.05 (0.04) �0.06Hispanic/other 0.72** (0.19) 0.00 (0.05) L0.14** (0.03) L0.07N 1451 1405 1406 1407

Poor health behaviors (contd.) Biological mediators

Binge drinkinga Less healthcareutilizationa,d

Obesityb Inflam

Race (ref: non-Hispanic White)Black L0.64* (0.29) L0.57** (0.18) �0.88 (0.55) 0.24*Hispanic/other 0.18 (0.18) 0.17 (0.23) L1.33* (0.56) �0.01N 1440 1336 1408 878

Note: all models control a participant’s age. All estimates are weighted to account for differerrors are given in parentheses. Figures in bold represent associations significant at leasþp < .10; *p < .05; **p < .01.

a Ordinal outcome. Results are from ordinal logit regression model.b Continuous outcome. Results are from OLS regression model.c Dummy outcome. Results are from logistic regression model.d Lack of healthcare access controlled.e Observations of CRP greater than 8.6 mg/l excluded from analysis.

the number of days in the preceding three months a participant hashad four or more drinks on one occasion, and runs from 0 (none) to2 (two or more). Finally, less healthcare utilization is indexed byone’s ordinal self-reported frequency of seeing a doctor in thepreceding 12 months, and ranges from 0 (30 or more) to 7 (none).

Of the two posited biological mediators, abdominal obesity isindicated by waist size (in inches). Next, in keeping with previousstudies (Alley et al., 2006; McDade et al., 2006; Owen et al., 2003),low-grade chronic inflammation is indexed by C-reactive protein(CRP)dderived from high-sensitivity assays of dried blood spotscollected through capillary finger-sticks (Williams & McDade,2009). Due to heavy right skew, raw CRP values are log-transformed, with values greater than 8.6 mg/l excluded fromanalysis as more indicative of acute (short-term) infection thanprolonged stress-exposure (McDade, Burhop, & Dohnal, 2004).

Metabolic outcomesDiabetic problems, or issues with blood-sugar control, are indi-

cated by log-transformed hemoglobin A1c (HbA1c)dglycosylatedhemoglobin as a percentage of total hemoglobin. As with CRP,HbA1c is derived from dried blood spots, and measures plasmaglucose concentration, with higher levels indicating worse blood-sugar control. Next, cardiovascular health is indexed by systolic anddiastolic BP (blood pressure, in mm Hg), and heart rate (in beats perminute)deach indicator representing the mean of two successivereadings.

Analytic plan

The analysis is set up in stages, aimed at progressively elimi-nating posited mechanisms. While the tables present results for“Hispanic or other” men, this is solely for baseline information.Given the theoretical focus of the study, the account below focuseson blackewhite differentials.

The first stage (Table 3) explores men’s age-adjusted racedifferentials in posited social, behavioral, and biological (obesity,inflammation) mediators, as well as metabolic outcomes (blood-sugar control and the three cardiovascular factors). As noted, themodel examining healthcare utilization also adjusts for access tocare. A wider array of controls is deliberately not used in these

57e85: coefficients (standard errors).

Poor health behaviors

r kinrtionb

Poor sleepc Inactivitya Smokerc More alcohola

(0.04) 0.72** (0.13) 0.47* (0.19) 0.05 (0.24) L0.50* (0.21)þ (0.04) �0.04 (0.21) 0.30 (0.21) �0.27 (0.32) L0.39* (0.17)

1446 1449 1181 1448

Metabolic outcomes

mationb,e Blood-sugarcontrolb

Systolic BPb Diastolic BPb Heart rateb

* (0.07) 0.12* (0.05) 3.79* (1.81) 2.06* (0.83) 2.16þ (1.20)(0.06) 0.06þ (0.03) �2.37 (1.59) �0.71 (0.97) 1.50 (1.58)

838 1420 1420 1418

ential probabilities of selection and differential nonresponse. Design-based standardt p < .10.

A. Das / Social Science & Medicine 77 (2013) 75e83 79

analyses, to avoid overcontrolling race differentials. As the lastsection clarifies, these dependent variables include continuous,ordinal, as well as dichotomous indicators. Accordingly, results arefrom OLS, ordinal logit, and logistic regression models. To facilitatepattern-visualization, coefficients rather than Odds Ratios arepresented for categorical outcomes.

Next, associations between each posited mediator witha significant blackewhite differential, and each distal metabolicoutcome, are examined one-at-a-time in Table 4. As noted, theproposed mechanisms are potentially interrelated. Hence, these“gross models” also analyze interdependencies between thesefactors. To adjust for potential endogeneity, additional controls areadded in Table 4 to models examining biological outcomes(inflammation, metabolic states). These include any lifetime diag-nosis of diabetes or (separately) hypertension, which net outfeedback from health conditionsdand more generally, the possi-bility that observed associations are confounded by prior disease.(To the extent that current social and behavioral factorsdandespecially chronic inflammationdindex long-term processesresponsible for race differentials in the prevalence of diagnoseddiabetes and hypertension, these analyses are overcontrolled, andyield conservative estimates of mediator-outcome linkages.) Toadditionally control reverse causation from current spikes in healthstatus, healthcare access and utilization in the preceding year areaddeddon the argument that behavioral and stress responses tosuch issues largely occur through medical notification. Thesefactors are not adjusted in models with social mediators as inde-pendent variables, since it is partly by inducing changes inhealthcare behaviors that social ties may affect biologicaloutcomes. To avoid data lossdand solely as control variables inthese and subsequent path analysesdno healthcare access (111)and less healthcare utilization (109) have missing values replacedwith 0, with the maximum number of substitutions in parentheses.Finally, whether in representing less knowledge of health issues orlower material resources, less education may potentially confoundexamined associations, and is therefore controlled.

In the final analytic stage, a path model is used to jointly testpotential mediators linked tometabolic states in the “grossmodels”(Table 4), for their role inmediatingmen’s race differentials in theseoutcomes. Results for this “net model” are presented in Table 5. (To

Table 4Associations between potential mediators, and between mediators and metabolic outcom

Poor health behaviors Biological mediator Me

Poor sleepa Inactivityb Inflammationc,d,e Blo

Social deficitsFewer alters �0.06 (0.05) 0.06þ (0.04) 0.00 (0.02) 0.0N 1446 1449 878 83

Poor health behaviorsPoor sleep 0.11*f (0.05) L0N 876 83Inactivity 0.08**f (0.02) 0.0N 877 83

Biological mediatorInflammation 0.0N 78

Note: variables entered one-at-a-time into models controlling a participant’s age and educand differential nonresponse. Design-based standard errors are given in parentheses. Figþp < .10; *p < .05; **p < .01.

a Dummy outcome. Results are from logistic regression model.b Ordinal outcome. Results are from ordinal logit regression model.c Continuous outcome. Results are from OLS regression models.d Any lifetime diagnosis of diabetes or (separately) hypertension controlled.e Observations of CRP greater than 8.6 mg/l excluded from analysis.f Lack of healthcare access and less healthcare utilization controlled.

avoid analytic problems due to large variances, some variables (age,systolic and diastolic BP, and heart rate) are first divided by 10,solely in these analyses.) Final inferences about mediation aredrawn from postestimation analyses of specific indirecteffectsdeach subsuming all paths leading from race to a distalmetabolic outcome through a specific mediator (Muthén &Muthén, 1998e2010). The model also examines interdepen-dencies betweenmechanisms, by specifying covariances/correlatedresiduals between the mediators as well as proximal controls fromTable 4dthus avoiding assumptions about causal order. While thisconservative strategy precludes statistical testing of indirect effectsrouted through multiple mechanisms, direct associations betweenmediators do yield tentative inferences about such pathways.

With the exception of the path model, all analyses are con-ducted with the STATA 12.1 statistical package (Stata Corp., 2011).Results are weighted using svy methods, first using populationweights that adjust for the intentional oversampling of blacks andHispanics, and also incorporate a nonresponse adjustment basedon age and urbanicity (O’Muircheartaigh & Smith, 2007). Standarderrors are adjusted for sample stratification (sampling strataindependently) and clustering (sampling individuals within each of100 primary sampling units). Path analyses are conducted with theMplus Version 6 statistical package (Muthén & Muthén, 1998e2010), also adjusting for sample stratification, clustering, andunequal probability of selection.

Results

As noted, an extensive literature now documents inflammatoryresponses to antecedent and prolonged psychosocial stress. Tofurther validate this linkage, preliminary age-adjusted analysisexamines inflammation’s association with a current depressedstate. In the results, inflammation is significantly higher amongdepressed men (Coeff. ¼ 0.16, p < .01).

Race differentials: social ties, behaviors, biology

Table 3 contains age-adjusted OLS models for men’s racedifferences in posited social, behavioral, and biologicalmediatorsdas well as current metabolic states. Consistent with

es, among men: coefficients (standard errors).

tabolic outcomes

od-sugar controlc,d Systolic BPc,d Diastolic BPc,d Heart ratec,d

0 (0.00) 0.37 (0.42) 0.11 (0.21) 0.22 (0.30)8 1420 1420 1418

.02þf (0.01) �0.27f (1.39) 0.42f (0.92) 0.00f (1.05)6 1417 1417 14150f (0.01) L0.85þf (0.43) �0.35f (0.35) 0.75*f (0.31)7 1419 1419 1417

3**f (0.01) 3.11*f (1.27) 1.85*f (0.84) 4.39**f (1.06)4 875 875 874

ation. All estimates are weighted to account for differential probabilities of selectionures in bold represent associations significant at least p < .10.

Table 5Path model: coefficients (standard errors).

Poor health behaviors Biological mediator Metabolic outcomes

Inactivitya Inflammationb,c Blood-sugar controlb Systolic BPb Diastolic BPb Heart rateb

Race (ref: non-Hispanic White)Black Direct effect 0.11* (0.05) 0.09** (0.03) 0.01þ (0.01) 0.14 (0.10) 0.15þ (0.08) 0.02 (0.07)

Total effect 0.14** (0.05) 0.11** (0.03) 0.04** (0.01) 0.19* (0.08) 0.11* (0.04) 0.10þ (0.05)Hispanic/other Direct effect 0.05 (0.07) �0.03 (0.03) 0.02 (0.01) �0.11 (0.08) �0.02 (0.05) 0.05 (0.09)

Total effect 0.09 (0.06) �0.01 (0.03) 0.02þ (0.01) �0.14 (0.09) �0.05 (0.05) 0.07 (0.09)

Poor health behaviorsInactivity Direct ¼ total effect 0.12** (0.02) 0.00 (0.01) L0.18** (0.07) L0.07þ (0.04) 0.04 (0.04)

Biological mediatorInflammation Direct ¼ total effect 0.03** (0.01) 0.30þ (0.16) 0.17þ (0.09) 0.45** (0.09)

Note: numbers are weighted coefficients. Sample size is 1455. The only exogenous control is a participant’s age. Endogenous controls include less education, any lifetimediagnosis of diabetes or (separately) hypertension, no healthcare access, and less healthcare utilization. To condense presentation, only effects for theoretically-relevant pathsare reported. Full results available on request. All estimates are weighted to account for differential probabilities of selection and differential nonresponse. Design-basedstandard errors are given in parentheses. Figures in bold represent associations significant at least p < .10.þp < .10; *p < .05; **p < .01.

a Ordinal variable.b Continuous variable.c Observations of CRP greater than 8.6 mg/l excluded from analysis.

A. Das / Social Science & Medicine 77 (2013) 75e8380

arguments about older black men’s greater social isolation, thesemen report significantly fewer network members. However, theproportion of these alters who are blood relatives is significantlyhigher among black than white men. In other words, the findingssuggest that older black men have smaller networks rich in ties tothose most likely to provide caregiving, emotional, and instru-mental support. As with social factors, behavioral results onlypartly conform to expectations. Specifically, older black men are nomore likely than their white counterparts to smoke regularly, andless likely to consume alcohol or binge drinkdactively unhealthybehaviors that the biomedical literature links to metabolic prob-lems. Underutilization of healthcare is also less common amongblack men. These men are, however, more likely to have passiveunhealthy behaviorsdphysical inactivity and poor sleep habits.

Next, contrary to both media reports and some academic liter-ature, there is no statistically significant blackewhite differential inolder men’s obesity. Indeed, the direction of the effect suggestsa “negative” differential, with black men thinner than their whitecounterparts. (Separate analysis, not shown, suggests similar resultsfor BMI.) In contrast, older blackmen are significantlymore likely tohave chronic inflammation (with the magnitude of this associationincreasing, in supplementary analysis, with obesity controlled).Finally, as expected, all four distal metabolic outcomesdblood-sugar control, systolic and diastolic blood pressure, and heartratedare significantly worse among older black men. (As withinflammation, the magnitude of each of these effects also increasesin separate analysis controlling obesitydfurther indicating a lack ofmediation by this posited “gateway”.)

Associations: mediatoremediator, mediator-outcome

Next, Table 4 presents results for models examining the asso-ciations of men’s diabetic and cardiovascular states with theremaining potential mediators (inflammation, smaller networks,inactivity, and poor sleep habits), added one-at-a-time to controlvariables. As argued, these social, behavioral, and biological factorsare all potentially interdependentdand may play either mediatoryor confounding roles among themselves. Hence, the tables alsoexamine such interdependencies. These models include somemediators as dependent and some as independent variables. Thisordering is solely to simplify analysis and presentation, and doesnot reflect causal assumptions. As noted, subsequent path analysis

more correctly specifies covariances/correlated residuals betweenmediators, entered as a block along with proximal controls.

Results (Table 4) are inconsistent with network size as a poten-tial mediator. While a smaller network has a weak association witholder men’s physical inactivity, it is not linked to any distal meta-bolic outcome. Among behaviors, poor sleep habits and inactivityare both associated with more inflammation. Whether this linkagerepresents a causal effect, or a feedback from prolonged stress tothese behaviors, remains unclear. Moreover, while sleep is notassociated with any diabetic or cardiovascular outcome (and is thusarguably not a mediator), inactivity does have a positive linkagewith a higher heart rate. In contrast, inflammation has consistentpositive associations with all four metabolic outcomes.

Path analysis

As noted, in the final analytic stage, a path model jointlyexamines the role of the remaining potential mechanisms (inac-tivity, inflammation) in mediating men’s race-metabolic linkages(Table 5). The model’s Comparative Fit Index (CFI) and TuckereLewis Index (TLI) are 1.00, while the Root Mean Square Error ofApproximation (RMSEA) is 0.00, suggesting excellent fit.

Findings suggest that much of the race effect passes throughlifetime diabetes or hypertension. However, even net of theseconditions as well as inactivity, inflammation is correlated with allfour current distal metabolic states (Table 5). Postestimation anal-yses of specific indirect effects support conjectures about inflam-mation’s mediatory roledwith significant indirect effects leadingfrom black men, through their inflammation, to blood-sugarcontrol (Coeff. ¼ 0.003, p < .05), systolic (Coeff. ¼ 0.03, p < .10) anddiastolic BP (Coeff. ¼ 0.02, p < .05), and heart rate (Coeff. ¼ 0.05,p < .01). In contrast, inactivity does not mediate any of men’sindirect race effects. However, as in Table 4, inactivity is directlyassociated with inflammationdsuggesting that inflammation iseither a confounder (indexing chronic stress) of the inactivity-heartrate linkage from Table 4, or a mediator through which inactivityhas an indirect association with heart rate. (In separate path anal-ysis excluding only inflammation, inactivity retains its significantdirect association with heart rate (Coeff. ¼ 0.10, p < .05), and alsomediates a significant indirect race effect on this outcome(Coeff. ¼ 0.01, p < .05)dalso suggesting that it is only the inclusionof inflammation that suppresses this linkage.)

A. Das / Social Science & Medicine 77 (2013) 75e83 81

Discussion

This study began with four overlapping mechanisms for blackewhite differentials in older men’s diabetic and cardiovascularoutcomes: (1) deficits in social support and/or control; (2) poorhealth behaviors, potentially due to these social deficits; (3) obesity,possibly as a consequence of social and behavioral factors; and (4)inflammation, potentially induced by prolonged stress-exposure.Results were most uniformly consistent with the inflammationmechanism.

To recall, while the health benefits of social resources are wellestablished, the literature yields competing arguments about men’srace differences in these assets. Some studies suggest greater socialisolation among black mendpotentially leading to less support andcaregiving, less monitoring and control of negative behaviors, and,ultimately, worse health outcomes. Others indicate their greaterembeddedness in densely interconnected networks centered in thematriarchal kin group. Hypothesis 1 (Table 1) was included to testthese conjectures. Results ran counter to the social-deficits argu-ment. Consistent with Cornwell et al.’s (2008) combined-genderfindings, black men had smaller social networks (Table 3).However, this factor did not have any significant linkages withmetabolic problems (Table 4). Moreover, the proportion of bloodkin (those most likely to provide caregiving as well as instrumentaland emotional support) was in fact greater among black men thantheir white counterparts. It is possible, however, that the NSHAPdata omitted important social precursors to metabolic problems,such as the quality (and hence support potential) of one’s networkties.

Next, it was argued that whether due to social deficits, orcultural factors, older black men would have worse health behav-iors, and that these behaviors wouldmediate their worsemetabolicoutcomes (Hypothesis 2). At least for the list of behaviors availablein NSHAP, the results were generally inconsistent with thesearguments. Older black men were no more likely than their whitecounterparts to smoke, and less likely to consume alcohol or bingedrink (Table 3)dactively unhealthy behaviors that the biomedicalliterature strongly links to diabetic and cardiovascular problems.However, other passively unhealthy behaviorsdphysical inactivityand poor sleep habitsdwere more likely among black than whitemen, with both factors linked to chronic inflammation (Table 4).Poor sleep had no associations with either men’s blood-sugarcontrol or their cardiovascular problemsdwhile the “gross” asso-ciation between inactivity and heart rate (Table 4) was suppressedin men’s path results (Table 5) net of inflammation. Overall, then,contrary to both media accounts and academic literature, poorhealth behaviors do not seem to consistently mediate older blackmen’s worse diabetic and cardiovascular health.

A third argument centered in obesity. To recall, both the socio-logical and biomedical literatures frequently portray obesity as the“biological gateway” through which black men’s poor healthbehaviors lead to their worse metabolic outcomes (Hypothesis 3). Itwas argued that these assumptions implicitly represent amoralistic“culture of irresponsibility” model for race disparities in men’shealth. Findings (Table 3) ran completely contrary to thishypothesisdwith older black men less likely than their whitecounterparts to be obese.

Rather than social or behavioral factors, or obesity, it was thefourth and least examined mechanismdinflammation, posited asan alternative “gateway” through which cumulative and multi-dimensional stress may induce biological weatheringdthat mostconsistently accounted for older black men’s greater metabolicproblems (Hypothesis 4). Not only were these men more likely tohave chronic inflammation (Table 3), but in both the “gross”(Table 4) and path (Table 5) results, inflammation was the one

mediator consistently associated with blood-sugar control as wellas all three cardiovascular outcomes. Moreover, analysis of specificindirect effects confirmed that inflammation significantlymediatedmen’s blackewhite differentials in all four distal metabolic states.To recall, it was noted that some cross-sectional studies also linkinflammation to obesity and specific health behaviors, with thecausal direction unclear. It was argued, then, that inflammationrepresents a stress-induced weathering mechanism to the extentthat its role in mediating race differentials in metabolic outcomes isnot confounded or causally influenced by these factors. In theresults, obesity was, as noted, (non-significantly) lower amongolder black than white men, and behaviors other than inactivitywere either no higher among blackmen or had no associationswiththeir diabetic or cardiovascular health. Hence, it is argued that thepathways through which inflammation induces black men’s worsemetabolic states may not involve these factors. The only potentialexception is inactivity. If, as noted above, the association betweenthis factor and men’s higher heart rate was mediated by inflam-mation, then the role of inflammation for this outcomewas (at leastpartly) that of a behaviorally-induced mechanism, and not (only)a stress-metabolic “gateway.” Alternately, inflammation may haveconfounded the linkage between heart rate and this passivelyunhealthy current behavior precisely in marking prolonged stress-exposure. Moreover, this potential behavioral-induction appliedsolely to a single metabolic outcomedwith only inflammationplaying a plausible mediatory role for black men’s worse blood-sugar control, and their higher systolic and diastolic BP. Along withthe depressioneinflammation associations noted above, thisweight of evidence suggests that the inflammation results mayindeed reflect the direct metabolic impact of older black men’sgreater long-term stress.

While this study provides important new information on thebiosocial mechanisms behind a major set of ethnically stratifiedhealth issues in late life, it also has several limitations. First, the listof health behaviors examineddwhile based on current biomedicalliteraturedcould have missed important causal factors, with thesame true of possible network precursors tometabolic issues. In theabsence of direct measures, most also had to be crudely proxiedthrough self-reportsdleading to potential bias from race patternsin reporting. Next, efforts were made to properly conceptualizepotential feedback pathways, and confoundingdand to includerelevant controls. However, the NSHAP data were cross-sectional,making it difficult to examine causal order. In particular, thereremains a possibility of residual feedback from health states toposited mediators. Moreover, the social, behavioral, and biologicalmechanisms outlined above were potentially interrelated. All threesets of factors had possible autonomous linkages with metabolicoutcomes. Additionally, social deficits could lead to obesity andinflammation through lower control of unhealthy behaviors, and toinflammation via less alleviation of chronic stress. Worsebehaviorsdwhether active (smoking, drinking), or passive (inac-tivity, poor sleep)dcould be associated with diabetic and cardio-vascular health, as well as with obesity and inflammation. Finally,obesity had a potential biological linkage with inflammation. Inother words, the role of each factor in accounting for blackewhitedifferentials in metabolic outcomes could either be mediated orconfounded by the other mechanisms. It was thus vital to examineassociations not just between single mechanisms and outcomes,but among mechanisms, as well as potential indirect pathwaysleading from race to metabolic pathologies via multiple mecha-nisms. While these cross-sectional analyses explored such inter-dependencies, and presented tentative inferences, comprehensiveaccounts of confounding and mediation must await longitudinaldata. Perhaps most importantly, the web of interrelationshipsbetween social and metabolic pathologies has only begun to be

A. Das / Social Science & Medicine 77 (2013) 75e8382

examineddsuggesting additional potential for confounding byunexplored pathways. The present study should thus be seen asa broad analysis that establishes baseline linkages and lays thegroundwork for more detailed exploration.

Conclusion

Standard mechanisms in the literature do not consistentlyaccount for men’s blackewhite differentials in diabetic andcardiovascular problems. Specifically, some academic studies aswell as media accounts take an implicit “culture of irresponsibility”approachdimplicating black men’s greater social isolation, worsehealth behaviors, and obesity as the key factors behind their worsehealth outcomes. In contrast, findings from a nationally represen-tative study of older Americans suggest that older black men areless obese than their white counterparts, have social networks richin blood relatives, and their few unhealthier behaviorsmay not playa major role in their current metabolic health. Instead, theseoutcomes seem to derive more consistently from a factor almostunexamined in the literaturedchronic inflammation, arguablya biological weathering mechanism induced by older black men’scumulative and multi-dimensional stress. In other words, theseproblems seem at least in part to be directly rooted in the systemsof social stratification in which these men have lived their lives. Associologists have long argued (Link & Phelan, 1995), reducing racedisparities in health will require refocusing attention to thesefundamental causes of disease.

Acknowledgments

The National Social Life, Health, and Aging Project is supportedby the National Institutes of Health (R01-AG021487), (R37-AG030481), and (R01-AG033903). I thank Edward O. Laumann,Linda Waite, the Editors of Social Science & Medicine, and threeanonymous reviewers for their helpful comments and suggestions.

References

Abbasi, F., Brown, B. W., Jr., Lamendola, C., McLaughlin, T., & Reaven, G. M. (2002).Relationship between obesity, insulin resistance, and coronary heart diseaserisk. Journal of the American College of Cardiology, 40(5), 937e943.

Alley, D. E., Seeman, T. E., Kim, J. K., Karlamangla, A., Hu, P., & Crimmins, E. M.(2006). Socioeconomic status and C-reactive protein levels in the U.S. pop-ulation: NHANES IV. Brain, Behavior, and Immunity, 20(5), 498e504.

American Psychiatric Association. (2000). Diagnostic and statistical manual of mentaldisorders: DSM-IV-TR. Washington, DC: American Psychiatric Association.

Bennett, G. G., & Wolin, K. Y. (2006). Satisfied or unaware? Racial differences inperceived weight status. International Journal of Behavioral Nutrition and Phys-ical Activity, 3, art. no. 40.

Bierhaus, A., et al. (2003). A mechanism converting psychosocial stress intomononuclear cell activation. Proceedings of the National Academy of Sciences ofthe United States of America, 100(4), 1920e1925.

Boardman, J. D., Saint Onge, J. M., Rogers, R. G., & Denney, J. T. (2005). Race differ-entials in obesity: the impact of place. Journal of Health and Social Behavior,46(3), 229e243.

Burton, L., & Snyder, A. (1998). The invisible man revisited: comments on the lifecourse, history, and men’s roles in American families. In A. Booth (Ed.), Men infamilies: When do they get involved? What difference does it make? (pp. 31e39).Hillsdale, NJ: Lawrence Erlbaum Associates.

Campos, P., Saguy, A., Ernsberger, P., Oliver, E., & Gaesser, G. (2006). The epidemi-ology of overweight and obesity: public health crisis or moral panic? Interna-tional Journal of Epidemiology, 35(1), 55e60.

Cappuccio, F. P., et al. (2007). Gender-specific associations of short sleep durationwith prevalent and incident hypertension: the Whitehall II study. Hypertension,50(5), 693e700.

Chang, V. W., & Lauderdale, D. S. (2005). Income disparities in body mass index andobesity in the United States, 1971e2002. Archives of Internal Medicine, 165(18),2122e2128.

Chiu, H.-M., Lin, J.-T., Chen, T. H.-H., Lee, Y.-C., Chiu, Y.-H., Liang, J.-T., et al. (2008).Elevation of C-reactive protein level is associated with synchronous andadvanced colorectal neoplasm in men. American Journal of Gastroenterology,103(9), 2317e2325.

Coleman, J. S. (1988). Social capital in the creation of human capital. AmericanJournal of Sociology, 94, S95eS120.

Cornwell, B., Schumm, L. P., & Laumann, E. O. (2008). The social connectedness ofolder adults: a national profile. American Sociological Review, 73(2), 185e203.

Danesh, J., et al. (2000). Low grade inflammation and coronary heart disease:prospective study and updated meta-analyses. BMJ, 321(7255), 199e204.

Davis, L., Uezato, A., Newell, J. M., & Frazier, E. (2008). Major depression andcomorbid substance use disorders. Current Opinion in Psychiatry, 21(1), 14e18.

Denney, J. T., Krueger, P. M., Rogers, R. G., & Boardman, J. D. (2004). Race/ethnic andsex differential in body mass among U.S. adults. Ethnicity and Disease, 14(3),389e398.

Drum, M. L., Shiovitz-Ezra, S., Gaumer, E., & Lindau, S. T. (2009). Assessment ofsmoking behaviors and alcohol use in the national social life, health, and agingproject. Journals of Gerontology e Series B Psychological Sciences and SocialSciences, 64(Suppl. 1), i119ei130.

Dupre, M. E. (2008). Educational differences in health risks and illness over the lifecourse: a test of cumulative disadvantage theory. Social Science Research, 37(4),1253e1266.

Feldner, M. T., Babson, K. A., & Zvolensky, M. J. (2007). Smoking, traumatic eventexposure, and post-traumatic stress: a critical review of the empirical literature.Clinical Psychology Review, 27(1), 14e45.

Fisher, M. A., Taylor, G. W., Shelton, B. J., & Debanne, S. (2008). Age and race/ethnicityegender predictors of denying smoking, United States. Journal ofHealth Care for the Poor and Underserved, 19(1), 75e89.

Geiss, L. S., Pan, L., Cadwell, B., Gregg, E. W., Benjamin, S. M., & Engelgau, M. M.(2006). Changes in incidence of diabetes in U.S. adults, 1997e2003. AmericanJournal of Preventive Medicine, 30(5), 371e377.

Geronimus, A. T. (2001). Understanding and eliminating racial inequalities inwomen’s health in the United States: the role of the weathering conceptualframework. JAMA, 56(4), 133e136, 149e150.

Geronimus, A. T., Hicken, M., Keene, D., & Bound, J. (2006). “Weathering” and agepatterns of allostatic load scores among blacks and whites in the United States.American Journal of Public Health, 96(5), 826e833.

Goossens, G. H. (2008). The role of adipose tissue dysfunction in the pathogenesis ofobesity-related insulin resistance. Physiology and Behavior, 94(2), 206e218.

Grant, B. F., & Harford, T. C. (1995). Comorbidity between DSM-IV alcohol usedisorders and major depression: results of a national survey. Drug and AlcoholDependence, 39(3), 197e206.

Grundy, S. M., Brewer, H. B., Jr., Cleeman, J. I., Smith, S. C., Jr., & Lenfant, C. (2004).Definition of metabolic syndrome: report of the National Heart, Lung, and BloodInstitute/American Heart Association conference on scientific issues related todefinition. Arteriosclerosis, Thrombosis, and Vascular Biology, 24(2), e13ee18.

Haskell, W. L., et al. (2007). Physical activity and public health: updated recom-mendation for adults from the American College of Sports Medicine and theAmerican Heart Association. Medicine & Science in Sports & Exercise, 39(8),1423e1434.

Heron, M. P. (2007). Deaths: leading causes for 2004. National Vital Statistics Reports,Vol. 56(5), Hyattsville, MD: National Center for Health Statistics.

Hill, R. B. (1999). The strengths of black families: Twenty-five years later. Lanham, MD:University Press of America.

Jarrett, R., Roy, K., & Burton, L. (2002). Fathers in the hood: qualitative research onlow-income African American men. In C. Tamis-LeMonda, & N. Cabrera (Eds.),Handbook of father involvement: Multidisciplinary perspectives (pp. 211e248).New York: Lawrence Erlbaum Associates.

Kawachi, I., Kennedy, B. P., & Glass, R. (1999). Social capital and self-rated health:a contextual analysis. American Journal of Public Health, 89(8), 1187e1193.

Kiecolt-Glaser, J. K., Gouin, J.-P., & Hantsoo, L. (2010). Close relationships, inflam-mation, and health. Neuroscience and Biobehavioral Reviews, 35(1), 33e38.

Koenen, K. C., Hitsman, B., Lyons, M. J., Niaura, R., McCaffery, J., Goldberg, J., et al.(2005). A twin registry study of the relationship between posttraumatic stressdisorder and nicotine dependence in men. Archives of General Psychiatry, 62(11),1258e1265.

Lakka, H.-M., et al. (2002). The metabolic syndrome and total and cardiovasculardisease mortality in middle-aged men. JAMA, 288(21), 2709e2716.

Link, B. G., & Phelan, J. (1995). Social conditions as fundamental causes of disease.Journal of Health and Social Behavior, (Spec no.), 80e94.

McDade, T. W. (2005). The ecologies of human immune function. Annual Review ofAnthropology, 34, 495e521.

McDade, T. W., Burhop, J., & Dohnal, J. (2004). High-sensitivity enzyme immuno-assay for C-reactive protein in dried blood spots. Clinical Chemistry, 50(3),652e654.

McDade, T. W., Hawkley, L. C., & Cacioppo, J. T. (2006). Psychosocial and behavioralpredictors of inflammation in middle-aged and older adults: the Chicago health,aging, and social relations study. Psychosomatic Medicine, 68(3), 376e381.

McDade, T. W., & Hayward, M. D. (2009). Rationale and methodological options forassessing infectious disease and related measures in social science surveys.Biodemography and Social Biology, 55(2), 159e177.

McDade, T. W., Lindau, S. T., & Wroblewski, K. (2011). Predictors of C-reactiveprotein in the national social life, health, and aging project. Journals of Geron-tology e Series B Psychological Sciences and Social Sciences, 66(1), 129e136.

McEwen, B. S. (1998). Protective and damaging effects of stress mediators. The NewEngland Journal of Medicine, 388(3), 171e179.

McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2006). Social isolation inAmerica: Changes in core discussion networks over two decades. AmericanSociological Review, 71(3), 353e375.

A. Das / Social Science & Medicine 77 (2013) 75e83 83

Meier-Ewert, H. K., Ridker, P. M., Rifai, N., Regan, M. M., Price, N. J., Dinges, D. F., et al.(2004). Effect of sleep loss on C-reactive protein, an inflammatory marker ofcardiovascular risk. Journal of the American College of Cardiology, 43(4), 678e683.

Melamed, S., Shirom, A., Toker, S., Berliner, S., & Shapira, I. (2006). Burnout and riskof cardiovascular disease: evidence, possible causal paths, and promisingresearch directions. Psychological Bulletin, 132(3), 327e353.

Miller, M. A., et al. (2009). Gender differences in the cross-sectional relationshipsbetween sleep duration and markers of inflammation: Whitehall II study. Sleep,32(7), 857e864.

Mokdad, A. H., et al. (2003). Prevalence of obesity, diabetes, and obesity-relatedhealth risk factors, 2001. JAMA, 289(1), 76e79.

Morenoff, J. D., House, J. S., Hansen, B. B., Williams, D. R., Kaplan, G. A., & Hunte, H. E.(2007). Understanding social disparities in hypertension prevalence, awareness,treatment, and control: the role of neighborhood context. Social Science &Medicine, 65(9), 1853e1866.

Mukamal, K. J. (2006). The effects of smoking and drinking on cardiovasculardisease and risk factors. Alcohol Research and Health, 29(3), 199e202.

Muthén, L. K., & Muthén, B. O. (1998). Mplus user’s guide (6th ed.). Los Angeles, CA:Muthén & Muthén.

Ockene, I. S., & Miller, N. H. (1997). Cigarette smoking, cardiovascular disease, andstroke: a statement for healthcare professionals from the American HeartAssociation. Circulation, 96(9), 3243e3247.

Oliver, J. E. (2006). The politics of pathology: how obesity became an epidemicdisease. Perspectives in Biology and Medicine, 49(4), 611e627.

O’Muircheartaigh, C., & Smith, S. (2007). NSHAP (national social life, health, and agingproject) wave 1 methodology report. Chicago, Ill.: National Opinion ResearchCenter (NORC).

Owen, N., Poulton, T., Hay, F. C., Mohamed-Ali, V., & Steptoe, A. (2003). Socioeco-nomic status, C-reactive protein, immune factors, and responses to acutemental stress: special issue on psychological risk factors and immune systeminvolvement in cardiovascular disease. Brain, Behavior, and Immunity, 17(4),286e295.

Plowden, K. O. (2003). A theoretical approach to understanding black men’s health-seeking behavior. Journal of Theory Construction and Testing, 7(1), 27e31.

Pradhan, A. D., Manson, J. E., Rifai, N., Buring, J. E., & Ridker, P. M. (2001). C-reactiveprotein, interleukin 6, and risk of developing type 2 diabetes mellitus. JAMA,286(3), 327e334.

Reaven, G. M. (1995). Characteristics of metabolic syndrome X. Endocrinology andMetabolism, 2(Suppl. B), 37e42.

Ridker, P. M., Buring, J. E., Cook, N. R., & Rifai, N. (2003). C-reactive protein, themetabolic syndrome, and risk of incident cardiovascular events: an 8-yearfollow-up of 14 719 initially healthy American women. Circulation, 107(3),391e397.

van Ryn, M. (2002). Research on the provider contribution to race/ethnicitydisparities in medical care. Medical Care, 40(1 Suppl.), I140eI151.

Saguy, A. C., & Almeling, R. (2008). Fat in the fire? Science, the news media, and the“obesity epidemic”. Sociological Forum, 23(1), 53e83.

Saguy, A. C., & Gruys, K. (2010). Morality and health: news media constructions ofoverweight and eating disorders. Social Problems, 57(2), 231e250.

Sarkisian, N., & Gerstel, N. (2004). Kin support among blacks and whites: race andfamily organization. American Sociological Review, 69(6), 812e837.

Schnittker, J. (2007). Look (closely) at all the lonely people: Age and the socialpsychology of social support. Journal of Aging and Health, 19(4), 659e682.

Shaw, B. A., Krause, N., Liang, J., & Bennett, J. (2007). Tracking changes in socialrelations throughout late life. Journals of Gerontology e Series B PsychologicalSciences and Social Sciences, 62(2), 90e99.

Shiovitz-Ezra, S., Leitsch, S., Graber, J., & Karraker, A. (2009). Quality of life andpsychological health indicators in the national social life, health, and agingproject. Journals of Gerontology e Series B Psychological Sciences and SocialSciences, 64(Suppl. 1), i30ei37.

Singer, B., Ryff, C., & Seeman, T. E. (2004). Allostasis, homeostasis and the cost ofphysiological adaptation. In J. Schulkin (Ed.), Allostatic load: Operationalizingallostatic load (pp. 113e149). Cambridge, UK: Cambridge University Press.

Stata Corp.. (2011). STATA release 12.0. College Station, TX: Stata Press.Sterling, P., & Eyer, J. (1981). Biological basis of stress-related mortality. Social

Science & Medicine e Part E Medical Psychology, 15(1), 3e42.Timmermans, S., & Haas, S. (2008). Towards a sociology of disease. Sociology of

Health and Illness, 30(5), 659e676.Verger, P., Lions, C., & Ventelou, B. (2009). Is depression associated with health risk-

related behaviour clusters in adults? European Journal of Public Health, 19(6),618e624.

Weinstein, M., Vaupel, J. W., & Wachter, K. W. (2007). Biosocial surveys. Washington,D.C.: National Academies Press.

Wells, A. J. (1993). Appendix B: method for correcting relative risk for smokermisclassification. In Respiratory health effects of passive smoking: Lung cancer andother disorders: The report of the Environmental Protection Agency. NationalCancer Institute monograph 4, smoking and tobacco control, . Washington, DC:U.S. Dept of Health and Human Services.

Williams, D. R. (2001). Racial variations in adult health status: patterns, paradoxesand prospects. In N. Smelser, W. J. Wilson, & F. Mitchell (Eds.), Americabecoming: Racial trends and their consequences (pp. 371e410). Washington, D.C.:National Academy of Sciences Press.

Williams, D. R., & Collins, C. (1995). U.S. socioeconomic and racial differences inhealth: patterns and explanations. Annual Review of Sociology, 21, 349e386.

Williams, S. R., & McDade, T. W. (2009). The use of dried blood spot sampling in thenational social life, health, and aging project. Journal of Gerontology: SocialSciences, 64B(Suppl. 1), i131ei136.

Wilson, J. F. (2005). Is sleep the new vital sign? Annals of Internal Medicine, 142(10),877e880.

Yudkin, J. S. (2003). Adipose tissue, insulin action and vascular disease: inflam-matory signals. International Journal of Obesity, 27(Suppl. 3), S25eS28.

Zauszniewski, J. A., & Graham, G. C. (2009). Comparison of short scales to measuredepressive symptoms in elders with diabetes. Western Journal of NursingResearch, 31(2), 219e234.