Hostility is related to clusters of T-cell cytokines and chemokines in healthy men

+ Models

PNEC-1385; No of Pages 10

Hostility is related to clusters of T-cell cytokinesand chemokines in healthy men

Paula M.C. Mommersteeg a,c, Eric Vermetten b, Annemieke Kavelaars a,Elbert Geuze b, Cobi J. Heijnen a,*

a Laboratory for Psychoneuroimmunology, University Medical Center Utrecht, KC03.068.0, Lundlaan 6,3584EA Utrecht, The NetherlandsbResearch Centre, Military Mental Healthcare, Ministry of Defense, Utrecht, The NetherlandscCoRPS-Center of Research on Psychology in Somatic diseases, Tilburg University, Tilburg, The Netherlands

Received 29 November 2007; received in revised form 15 April 2008; accepted 9 May 2008

Psychoneuroendocrinology (2008) xxx, xxx—xxx

KEYWORDSHostility;Cytokines;Chemokines;Factor analysis;Risk factor;Male

Summary Hostility is a risk factor for adverse health outcomes as diverse as cardiovasculardisease and post-traumatic stress disorder (PTSD). Cytokines have been suggested to mediate thisrelationship. We investigated whether in healthy men a relation existed between hostility and T-cell mitogen-induced cytokines and chemokines. Male Dutch military personnel (n = 304) wereincluded before deployment. Eleven cytokines and chemokines weremeasured in supernatants ofT-cell mitogen-stimulated whole blood cultures by multiplex immunoassay. Factor analysis wasused to identify clusters of cytokines and chemokines. In a regression analysis hostility wasrelated to the cytokine/chemokine clusters, and the potential risk factors age, BMI, smoking,drinking, previous deployment, early life trauma and depression.

Explorative factor analysis showed four functional clusters; a pro-inflammatory factor (IL-2,TNFa, IFNg), an anti-inflammatory factor (IL-4, IL-5, IL-10), IL-6/chemokine factor (IL-6, MCP-1,RANTES, IP-10), and MIF. Hostility was significantly related to decreased IL-6/chemokine secre-tion and increased pro- and anti-inflammatory cytokines. There was an inverse relation betweenage and hostility scores. Early life trauma and depression were positively and independentlyrelated to hostility as well.

This study represents a novel way of investigating the relation between cytokines andpsychological characteristics. Cytokines/chemokines clustered into functional factors, whichwere related to hostility in healthy males. Moreover this relation appeared to be independent ofreported depression and early trauma.# 2008 Elsevier Ltd. All rights reserved.

ava i lab le at www.sc ienced i rect .com

journa l homepage: www.e l sev ie r.com/locate/psyneuen

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitymen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008.

* Corresponding author. Tel.: +31 88 7554351.E-mail address: [email protected] (C.J. Heijnen).

0306-4530/$ — see front matter # 2008 Elsevier Ltd. All rights reservedoi:10.1016/j.psyneuen.2008.05.007

1. Introduction

Hostility has been identified as a risk factor for the devel-opment of adverse health outcomes as diverse as post-trau-matic stress disorder (PTSD; Heinrichs et al., 2005; Ouimette

is related to clusters of T-cell cytokines and chemokines in healthy05.007

d.

+ Models

PNEC-1385; No of Pages 10

2 P.M.C. Mommersteeg et al.

et al., 2004; Orth and Wieland, 2006) and cardiovasculardisease (CVD; McEwen, 2000; Jorgensen et al., 2001; Smithet al., 2004). Hostility can be described as having negativebeliefs about and attitudes toward others, which includescynicism and mistrust. Hostility correlates with other con-structs of negative affect like depression and anxiety, but it isconsidered to be a stable personality trait (Everson-Rose andLewis, 2005). Adverse health behaviors, including smoking,alcohol intake, a high body mass index (BMI), as well as socialeconomic status (SES) have been suggested to mediate therelation between hostility and adverse health outcomes,with hostile persons being more prone to health risk beha-viors. However hostility is valuable as an independent riskfactor as well (Christensen et al., 2004; Bunde and Suls,2006).

From a psychoneuroimmunological point of view it hasbeen suggested that a relation between cytokines and hos-tility could be important for health outcomes. There isevidence that hostility is related to a pro-inflammatorycytokine profile and adverse health outcomes, and may thusserve as a mediating factor (Gidron et al., 2002; Everson-Rose and Lewis, 2005). In normal healthy men and womenserum IL-6 levels were positively related to cynicism, hostileaffect, depression and vital exhaustion (Sjogren et al.,2006). Couples showing high-hostile marital interactionshad larger increases in plasma IL-6 and TNFa after a maritalconflict, and their wound healing was impaired (Kiecolt-Glaser et al., 2005). Hostility and aggression in men werealso associated with increased TNFa secretion by LPS sti-mulated peripheral blood monocytes (Suarez et al., 2002).Moreover, in healthy men high scores on both hostility anddepression were related to increased plasma IL-6 levels(Suarez, 2003), whereas Miller et al. observed that higherplasma IL-6 levels were related to high hostility, but lowdepression (Miller et al., 2003).

The aim of our study was to determine whether a relationbetween hostility and various inflammatory markers could beestablished in a healthy group of males. So far the relationbetween cytokines and hostility has been examined by deter-mining levels in serum or plasma, or the release of cytokinesand chemokines by stimulated monocytes. Thus, the majorfocus has been on the relation between hostility and theinnate immune system. To date no studies have investigatedwhether hostility is related to the activity of the adaptiveimmune system, i.e. the release of cytokines and chemokinesafter stimulation of T-cells. Therefore, we investigatedwhether cytokine and chemokine release induced by stimu-lation of T-cells was related to hostility as measured by theCook—Medley Hostility Score (Barefoot et al., 1989). Sincewe investigated a cross-sectional sample we explored asso-ciations rather than a possible causal relationship.

We analyzed eleven different cytokines and chemokinesin supernatants of whole blood cultures stimulated with a T-cell mitogen. The cytokines measured were the pro-inflam-matory cytokines IL-2, IL-6, TNFa, and interferon gamma(IFNg), and the anti-inflammatory cytokines IL-4, IL-5 and IL-10. In addition we measured the chemokines MonocyteChemotactic Protein-1 (MCP-1, or CCL2), Regulated uponActivation, Normal T-cell Expressed, and Secreted (RANTES,CCL5), as well as Interferon-g-Induced Protein (IP-10,CXCL10). Apart from these chemokines we also determinedMacrophage Migration Inhibitory Factor (MIF), a cytokine

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitmen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008

with a key regulatory role in innate immunity, which pro-motes the pro-inflammatory function of immune cells(Calandra and Roger, 2003).

It is known that there is a high level of interrelationbetween cytokines and chemokines and that there is func-tional redundancy in the cytokine and chemokine system. Ifsingle cytokines/chemokines show a similar relation with thepsychological variable, clustering them into a limited numberof factors could strengthen the relation between these clus-ters and the psychological variable. Therefore, we firstperformed an explorative factor analysis to cluster the ele-ven different cytokines and chemokines into groups. Thefactor scores of these clusters were then used in a regressionanalysis to analyse their association with hostility.

We included healthy male military personnel within twomonths before deployment to Afghanistan. We controlled forthe known health risk factors as age, BMI, smoking, drinking,and depression. Furthermore pre-deployment PTSD-relatedrisk factors like previous deployment, and score on an earlylife trauma inventory were included (Stein et al., 2005; Gahmet al., 2006; Nemeroff et al., 2006).

The goal of this study was first to explore whether multiplecytokines and chemokines would form functional clusters,and second to determine whether an association existsbetween these clusters and hostility scores in healthy males.

2. Methods

2.1. Participants

In The Netherlands a large prospective study has started in2005. Biological and psychological variables are assessed inmilitary personnel before and after deployment to Afghani-stan for peace keeping missions. Data collection in this cross-sectional part of the study occurred between March 2005 andOctober 2006. 304 participants, mean age 30.7 years(S.D. = 10.0, range 18—55 years) were recruited to partici-pate in this project in the training period before deployment.Information about the project was provided and participantssigned an informed consent. In a morning session, the parti-cipants filled out the questionnaires, and blood was collectedin heparinized vacu-tubes between 8.00 and 11.00 a.m. Thisproject was approved by the medical ethical committee ofthe University Medical Center Utrecht.

2.2. Questionnaires

Demographic information on age, BMI, previous deployment(yes/no), marital status, education level, military rank,smoking, alcohol use and medication use were filled out.There was no separate information for income, but as incomescales per rank are fixed, we used military rank as an indi-cator for income.

The total score on the Cook—Medley Hostility Scale 50item true/false version was used to measure hostility (Cron-bachs alpha = 0.78; Barefoot et al., 1989).

General complaints were measured with the symptomchecklist 90 (SCL90; Arrindel and Ettema, 1981), 90 itemson a 5-point scale (Cronbachs alpha = 0.94). The SCL90 sub-scale depression consists of 16 items, with an internal con-sistency of 0.77.

y is related to clusters of T-cell cytokines and chemokines in healthy.05.007

+ Models

PNEC-1385; No of Pages 10

Hostility and stimulated T-cell cytokines 3

The short version of the ‘Early Trauma Inventory’ assessestraumatic experiences before the age of 18 in four domainson a 0—1 scale; general trauma (11 items), physical abuse (5items), emotional abuse (5 items) and sexual abuse (6 items;Bremner et al., 2000). The overall consistency of the totalscore of this questionnaire was good, Cronbachs alpha = 0.76.

In a subgroup of the participants (n = 180), PTSD symptomsof the pasts four weeks were assessed with the Self-ratingInventory for Posttraumatic Stress Disorder list (‘SRIP’, 22items, 1—4 scale, Cronbachs alpha = 0.88; Hovens et al.,1994). The total score (mean = 26.9, S.D. = 5.3) was notdifferent from a group of predominantly male (92%) Dutchfirefighters and policemen (n = 1168, mean = 27.2,S.D. = 6.2; t = 1.03, d.f. = 1375, p = n.s.; Witteveen et al.,2006). Due to its limited sample size the SRIP was notincluded in the regression analysis.

2.3. Leukocyte subpopulations

The distribution of subsets of T-cells, B-cells and NK-cellswithin the lymphocyte population was analyzed in hepari-nized whole blood using dual color fluorescence analysis witha Becton Dickinson Calibur flow cytometer. Whole blood wasstained using monoclonal antibodies (Becton Dickinson, Bel-gium) labeled with either fluorescine isothiocyanate orphyco-erythrin to quantify CD3+ (total T), CD4+ (T-helper),CD8+ (T suppressor/cytotoxic), CD19+ (B), CD16+56+ (NK),and CD3+/CD16+CD56+ (NK-T) cells. Absolute numbers ofcells were calculated from a total leukocyte count.

2.4. Cytokine production

T-cell mitogen-induced cytokine secretion was measured insupernatants of whole blood cultures diluted 1:10 with RPMI-1641 (Gibco, Grand Island, NY), 100 U/ml penicillin, 100 mg/ml streptomycin and 2 mM L-glutamine after stimulation withthe mitogenic anti-CD2/CD28 monoclonal antibodies (CLB,Amsterdam, The Netherlands, final concentration anti-CD2.1/anti-CD2.2 0.33 mg/ml and anti-CD28 1.33 mg/ml)for 72 h at 37 8C/5% CO2 in 96 wells round-bottom plates.Supernatants were stored at �80 8C. Cytokine productionwas measured using multiplex analysis to assess severalcytokines and chemokines in a single sample as describedelsewhere (De Jager et al., 2003). We determined the fol-lowing cytokines and chemokines IL-2, IL-4, IL-5, Il-6, TNFa,IFNg, MIF, MCP-1 (CCL2), IP-10 (CXCL10), and RANTES (CCL5).

2.5. Explorative factor analysis

The 11 cytokines and chemokines were clustered into factorsusing principal axis factoring with oblique (oblimin) rotationin SPSS 12.0.1. In line with the known functional relationsbetween cytokines, bivariate correlational analyses showedthat values for many cytokines and chemokines are corre-lated. Therefore, we chose the oblique rotation, whichpermits correlation between the factors, rather than theoften used varimax rotation, which does not allow correla-tion between the factors. The cytokines and chemokineswere either not (MCP-1), log 10 (IL-2, IL-4, IL-5, IL-6, IL-10, MIF) or square-root (TNFa, IFNg, RANTES, IP-10) trans-formed to obtain a normal distribution. The sample size for

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitymen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008.

factor analysis was sufficiently large (n = 243; Tabachnick andFidell, 2007). The factor loadings, the correlation betweenthe factor and a variable, varied between 0.48 (IL-10) whichis considered fair (20% overlapping variance), and 0.98 (IL-4),which is considered excellent (>50% overlapping variance).The lowest rotated factor loading was 0.30 for IP-10. Com-munalities were high, and the four factors accounted for 65%of the variance of the variables. The Kaiser-Meyer-Olkinmeasure of sampling adequacy was 0.725, which is consid-ered good (Tabachnick and Fidell, 2007). Factors were basedon examination of the screen plot and had initial eigenvalues>1.0. All (transformed) variables were used to calculate theregression factor scores which were then used in the regres-sion analysis.

2.6. Linear regression analysis

SPSS 12.0.1 was used for linear regression analysis, withhostility as the dependent variable. To prevent overfittingof the model, the regression method ‘enter’ was used, whichkeeps all variables in the final model (Babyak, 2004). Beforeanalysis the data were screened for the assumptions ofmultivariate analysis. The hostility scale showed a normaldistribution. Exclusion of a single outlier with a total hostilityscore of 44 did not change the outcome of the analysis.Multivariate outliers and multicollinearity diagnostics indi-cated no cause for concern. A p-value of 0.05 was consideredsignificant.

In the regression analysis the presence of a single missingvalue results in listwise exclusion of all data for thatperson. In total 3% (n = 10) of the immune data wereconsistently missing from individuals for whom question-naire scores and demographic variables were available. Thedataset also lacked data due to handling and non-responseof some questionnaire items. These missing variables weremissing completely at random, and accounted for themajority of the missing variables. After listwise deletion227 cases remained in the regression analysis, or 75% of theoriginal total dataset.

We explored whether missing data imputation could pro-vide a tool for a more complete dataset. Multiple Imputation(Schafer and Graham, 2002; Donders et al., 2006) was usedfor imputing missing data with WinMICE (www.multiple-imputation.com). Compared to the incomplete dataset,the dataset with missing data imputation showed similarcorrelations between related variables (e.g. cytokines); how-ever the correlations became less strong. Comparing theregression model of the dataset with missing variables(n = 227) to the dataset with imputed variables (n = 304),the latter model explained less variance and did not fit thedata as well as the smaller dataset. The imputed modelshowed diluted variance, therefore we chose not to reportthe results of the complete, imputed dataset.

3. Results

In Table 1 a description of the group is shown. The BMI showeda normal distribution, with a range from 18.5 to 35.5; 8% hada BMI>30. Military rank had a significant and high correlationwith age (r = 0.732), and we chose to include only age in theregression model. Alcohol use was dichotomized into either

is related to clusters of T-cell cytokines and chemokines in healthy05.007

+ Models

PNEC-1385; No of Pages 10

Table 1 Demographic variables

Mean S.D.

Age 30.67 10.0BMI 25.15 2.90

n %

Marital stateMarried/cohabiting 146 49Longstanding relation 57 19Single/divorced 97 32

EducationLow 117 39Medium 148 49College education or higher 39 12

Military rankOfficers

Senior/field 26 9Junior/company grade 25 8

Non-commissioned officersSergeant 89 32Corporal 50 16

Other enlisted personnelPrivate 105 35

Previously deployed (yes) 133 44Smoking (yes) 135 45Drinking >5 units/week 152 51Medication use 36 12

Of whichAnti-histamine 8 3Beta-blocker 6 2Cholesterol 4 1NSAID 6 2SSRI 2 0.7Angiotensin 1 0.3Glucocorticoid 1 0.3

4 P.M.C. Mommersteeg et al.

less than 5 units/week (0) or greater than 5 units (1). Ninepercent (n = 26) reported no alcohol use at all.

Table 2 summarizes the scores on the questionnaires. Thehostility scale showed a normal distribution, with a mean

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitmen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008

Table 2 Questionnaire scores

Mean Med

Cook—Medley Hostility 16.81 16Cynicism 4.72 5Aggressive responding 4.39 4Hostile affect 1.39 1

SCL90 total score 102.9 98Depression 18.10 17

ETI: Early Trauma Inventory 4.07 4General traumatic experience 2.11 2Physical abuse 1.35 1Emotional abuse 0.50 0Sexual abuse 0.12 0

score of 16.8 (S.D. = 6.7). The Early Trauma Inventory (ETI)and the depression scale of the SCL90 were extremely posi-tively skewed; most men reported no complaints or very lowscores. It was not possible to obtain a normal distribution bytransforming the scores. Therefore, we chose to dichotomizethe scales (MacCallum et al., 2002). After examination of thefrequency curve of both scales, we chose an arbitrary cut-offpoint which separated the peak of the distribution (the lowscores) from the rest of the scores (the higher scores). For theETI total score the cut-off point was set at�7 (n = 58, 19.2%),and for the depression score it was set at�19 (n = 74, 24.4%).The dichotomized score was used in the regression analysis.

The immune parameters are shown in Table 3. Both meanand median scores are shown since the values for cytokineand chemokine production and number of blood cells arepositively skewed. The distribution of leukocytes (WBC), HB,HCT, monocytes, granulocytes and lymphocytes are wellwithin the 95% normal range. There is no known norm-scorefor the range of cytokines and chemokines.

3.1. Cytokine and chemokine clustering

To bundle the effect of several related cytokines and che-mokines, they were clustered by factor analysis. Interest-ingly, this explorative analysis resulted in extraction of fourfactors which represent biologically functional clusters ofmediators. The first factor that emerged was an anti-inflam-matory factor, which included the three anti-inflammatorycytokines determined in our assay: IL-4, IL-5 and IL-10.Second the three pro-inflammatory cytokines IL-2, TNFa

and IFNg clustered into the pro-inflammatory factor. Thethird factor was designated ‘IL-6/chemokine’, and consistedof IL-6, MCP-1, RANTES and to a lesser extent IP-10. Macro-phage Migration Inhibitory Factor, or MIF, did not load well onthe other factors, and we chose to include it in the furtheranalysis as a separate factor. In Table 4 the b-values of thefactor analysis are shown, which represent the relativeproportion of explained variance controlled for the othervariables. In line with its role as a pro-inflammatory cytokine,IL-6 loads on the pro-inflammatory factor as well. As we usedan oblique rotation, correlations between factors were main-tained. We observed a negative correlation between the proand anti-inflammatory factor (r = �0.46), which is consistentin view of the known mutual inhibition of pro- and anti-

y is related to clusters of T-cell cytokines and chemokines in healthy.05.007

ian S.D. Minimum Maximum

6.72 4 442.83 0 131.86 1 91.20 0 5

14.84 90 1873.25 16 42

3.28 0 161.62 0 71.50 0 51.14 0 50.44 0 4

+ Models

PNEC-1385; No of Pages 10

Table 3 Immune variables

Mean Median S.D. 5% 95%

BasicWBC [106/ml] white blood cell count 6.52 6.32 1.68 4.27 9.72HB [mmol/l] hemoglobin 9.50 9.51 0.58 8.56 10.50HCT [l/l] hematocrit 0.460 0.459 0.029 0.419 0.509

Subsets absolute [106/ml]Monocytes 0.461 0.441 0.152 0.246 0.739Granulocytes 4.00 3.87 1.33 2.23 6.66

Lymphocytes 2.07 1.97 0.59 1.26 3.24B-cells 0.236 0.216 0.118 0.087 0.478NK-cells 0.306 0.273 0.169 0.101 0.634

T-cells CD3 1.42 1.33 0.46 0.83 2.28T-helper CD4 0.830 0.785 0.296 0.423 1.407T suppressor CD8 0.433 0.391 0.219 0.191 0.840T-helper/suppressor ratio [CD4/CD8] 2.19 1.99 1.00 0.94 4.14

Cytokines [pg/ml]IL-2 1240 729 1423 69 4185IL-4 80 54 86 10 230IL-5 711 381 1014 40 2372IL-10 342 267 278 66 899IL-6 1649 1086 1965 141 4762TNFa 577 489 459 57 1585IFNg 14,840 10,084 15,045 375 48,440MIF 9759 8789 3953 5304 16,709

Chemokines [pg/ml]MCP-1 23,288 20,976 12,084 7451 47,140RANTES 13,189 11,185 8842 3309 31,042IP-10 6382 5472 3823 1777 13,968

Hostility and stimulated T-cell cytokines 5

inflammatory cytokine production. The pro-inflammatorycytokine factor had a positive correlation with the IL-6/chemokine factor (r = 0.34), and negative with MIF(r = �0.16). Furthermore the IL-6/chemokines factor corre-lated negative with MIF (r = �0.24). The anti-inflammatoryfactor correlated weakly with IL-6/chemokines and MIF(r < 0.08). The factor scores for each factor were used inthe regression analysis.

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitymen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008.

Table 4 Factor analysis pattern matrix

Factor

Anti-inflammatory Pro-inflam

IL-4 S0.986 �0.032IL-5 S0.913 �0.093IL-10 S0.487 0.170

IL-2 �0.222 0.586TNFa 0.020 0.774IFNg 0.004 0.925

IL-6 0.059 0.390MCP-1 0.072 �0.222RANTES �0.038 0.300IP-10 �0.160 0.120

MIF �0.021 0.043

3.2. Regression analysis relation cytokinefactors with hostility

Before performing a regression analysis with hostility, weexplored the correlation between the hostility score and thecytokine/chemokine factors. The anti-inflammatory factorand the IL-6/chemokine factor showed significant andopposite correlations with the hostility score (n = 238,

is related to clusters of T-cell cytokines and chemokines in healthy05.007

matory Il-6/chemokines MIF

�0.138 �0.0720.123 �0.062�0.031 0.120

�0.018 �0.0110.053 �0.2700.019 0.075

0.708 0.1230.831 �0.0520.489 �0.2700.309 0.001

�0.009 S0.794

+ Models

PNEC-1385; No of Pages 10

Table 5 Regression analysis of hostility

Hostility

Step 1 Step 2 Step 3 Step 4 Step 5

Adj. R 2 0.030 0.047 0.086 0.164 0.278DR2% variance

explained3% 1.7% 3.9% 7.8% 11.4%

F value 8.05 ** 6.61 *** 6.35 *** 6.53 *** 8.89 ***

DF 5.02 * 5.80 ** 6.13 *** 12.44 ***

Steps 1—3: factors b-value t-value b-value t-value b-value t-value b-value t-value b-value t-value

IL-6 and chemokines a �0186 �2.838 ** �0.181 �2.782 ** �0.297 �3.973 *** �0.258 �3.506 *** �0.292 �4.223 ***

Anti-inflammatoryb 0.146 2.240 * 0.275 3.706 *** 0.173 2.194 * 0.196 2.652 **

Pro-inflammatoryc 0.279 3.402 *** 0.221 2.761 ** 0.235 3.149 **

MIF �0.006 �0.087 �0.027 �0.418 �0.029 �0.463

Step 4: demographicAge �0.288 �4.004 *** �0.259 �3.556 ***

BMI 0.011 0.161 �0.007 �0.113Smoking Y/N 0.076 1.163 0.088 1.435Drinking Y/N 0.064 1.027 0.070 1.195

Step 5: covariatesPrevious deployment �0.026 �0.410ETI high (�7)d 0.254 4.351 ***

Depression high (�19)e 0.226 3.941 ***

a IL-6, MCP-1, RANTES, IP-10.b IL-4, IL-5, IL-10.c IL-2, TNFa, IFNg.d TI cut-off �7, n = 58, 19.2%.e ep cut-off �19, n = 74, 24.4%.* p < .05.** p < .01.*** p < .001.

6 P.M.C. Mommersteeg et al.

anti-inflammatory factor: r = 0.154, p = .017 and IL-6/che-mokine factor: r = �0.169, p = .009). The pro-inflammatoryfactor and MIF did not show a significant correlation with thehostility score (r = 0.017, p = .799, and r = 0.070, p = .285respectively).

Linear regression analysis was used to determine whetherthe variance in hostility score could be explained by cytokineand chemokine production, while controlling for age, BMI,smoking, drinking, previous deployment, early trauma anddepression. To determine the relative contribution of eachfactor to hostility, the factors were introduced step-wise.The results are shown in Table 5. In the first step, the IL-6/chemokine factor was introduced. A lower IL-6/chemokinelevel is related to increased hostility, which explained 3.0% ofthe variance in hostility score. Subsequent introduction ofthe anti-inflammatory factor revealed that an increased anti-inflammatory cytokine level was related to a high hostilityscore. The relative contribution of each variable in themodelwas determined by comparing the standardized coefficient b

(b-values). When comparing the b-values of the IL-6.chemo-kine between step 1 and step 2, the value did not changesignificantly, showing that both factors have an independentrelation with hostility.

In step 3 the pro-inflammatory factor and MIF were intro-duced. Interestingly, the pro-inflammatory factor showed a

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitmen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008

significant positive relation with hostility as well. A higherhostility score was related to a higher pro-inflammatorycytokine level.

The model with all four cytokine and chemokine factorsexplained 8.6% of the variance in hostility score. Introductionof thepro-inflammatorycytokine factor increasedtheb-valuesof both the IL-6/chemokine and the anti-inflammatory factor.While the pro-inflammatory factor is not significantly corre-lated with hostility by itself, it does show a significant relationwith hostility in the model. This effect is known as ‘suppres-sion’ (Tabachnick and Fidell, 2007). The relation between thepro-inflammatory factorwith theanti-inflammatoryfactorandthe IL-6/chemokine factor ‘suppresses’ unexplained variance,which increases the relative relation of the variables withhostility. Vice versa both the IL-6/chemokine and the anti-inflammatory factoreachsuppressunexplainedvarianceof therelation between the pro-inflammatory factor and hostility. Itis likely that the non-significant correlation between the pro-inflammatory factorswith hostility is obscured by the oppositecorrelation of the IL-6/chemokine factor and the anti-inflam-matory factor with the pro-inflammatory factor.

After introducing the demographic (age, BMI) and healthrisk (smoking, drinking) variables, the model explained 15.4%of the variation in the hostility score. Age was significant andnegatively related to hostility.

y is related to clusters of T-cell cytokines and chemokines in healthy.05.007

+ Models

PNEC-1385; No of Pages 10

Hostility and stimulated T-cell cytokines 7

In the final step of the regression analysis the binaryvariables previous deployment, early life trauma and depres-sion were introduced (Table 5). Early life trauma and depres-sion had a significant and positive relation with hostility.Being previously deployed was not related to hostility inthe model. Persons reporting more early life trauma anddepression were more hostile.

In the final model the three cytokine/chemokine factorsshowed significant relation with the hostility score. A lowerIL-6/chemokine level was related to a higher hostility score.On the other hand the anti-inflammatory cytokine factor andthe pro-inflammatory cytokine factor had a positive relationwith the hostility score. An increased level of pro- and anti-inflammatory cytokines is related with a higher hostilityscore. This effect was stronger for the pro-inflammatorycytokines compared to the anti-inflammatory cytokines.The fourth factor, MIF had no relation with hostility.

3.3. Regression controlling for number of T-cells

Since we determined cytokine and chemokine productionafter stimulation with a T-cell mitogen, differences in num-ber of T-cells could have affected the results. The regressionmodel with hostility as dependent variable was controlled forthe number of T-cells (log 10 #CD3+). The number of T-cellshad no significant relation with hostility and controlling forthis variable did not affect outcome of the regression model(data not shown).

4. Discussion

This is the first report describing a significant relationbetween stimulated T-cell cytokines/chemokines and hosti-lity scores in a large group of healthy males. Explorativefactor analysis revealed functional clusters of anti-inflam-matory, pro-inflammatory, IL-6/chemokine and the cytokinefactor MIF. Regression analysis showed that hostility wassignificantly and negatively related to IL-6/chemokine, andpositively to both pro- and anti-inflammatory cytokinerelease. In other words; an increased hostility score is relatedto an overall decreased in IL-6/chemokines release and anincreased release of pro- as well as anti-inflammatory cyto-kines. The relation remained intact after controlling for thesignificant confounders age, early life trauma, and depres-sion.

Studies on the relation between hostility and cytokineshave predominantly focused on the innate immune system.These studies observed increased pro-inflammatory cytokinelevels (IL-6, TNFa) in relation to higher hostility scores(Suarez et al., 2002; Miller et al., 2003; Suarez, 2003;Kiecolt-Glaser et al., 2005; Sjogren et al., 2006). Our findingsadd to this field, and show that the pro- and anti-inflamma-tory cytokine profile after stimulation of the adaptiveimmune system was positively related to hostility. On theother hand, the IL-6/chemokine factor shows a negativerelation with hostility. At a first glance, the observation thatboth pro- and anti-inflammatory cytokines are positivelyrelated to hostility is remarkable. It should be noted, how-ever, that in a simple correlational analysis hostility waspositively related to the anti-inflammatory factor and nega-tively to the IL-6/chemokine factor, but not to the pro-

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitymen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008.

inflammatory factor. Moreover, the pro- and anti-inflamma-tory cytokine factors have a negative correlation ofr = �0.46, and IL-6 loads on the pro-inflammatory factor aswell. Such negative correlation between pro- and anti-inflammatory cytokines can be well understood in the phy-siological context, as it is known that production of pro-inflammatory cytokines can inhibit anti-inflammatory cyto-kines and vice versa. In addition, IL-6 has pro-inflammatoryproperties. In our complete model, we observed a positivecorrelation of both the pro- and anti-inflammatory cytokinefactor with hostility, but this relation only became apparentafter all factors were included in the model. We have inter-preted these data as being due to ‘suppression’, the correla-tions among the factors ‘suppresses’ unexplained variancewith hostility.

There are few studies which have investigated the rela-tion between hostility and the adaptive immune system.Kemeny and colleagues found that, averaged over a six-month period, high hostility was related to a lower proportionof suppressor/cytotoxic T-cells (CD8+), but not T-helper cells(CD4+; Kemeny et al., 1989). Acute hostile marital interac-tions were related to increases in the number of total T-cells(CD3+) and T-helper cells (CD4+), and a decrease in T-cellfunction (Kiecolt-Glaser et al., 1993). However, hostility wasnot related to an acute stress induced increase in T lympho-cytes (Mills et al., 1996). These results are difficult toextrapolate to the results observed in our study, especiallysince we have used a different read out, i.e. a broad spec-trum of cytokines and chemokines released after T-cellstimulation rather than circulating cell numbers or T-cellproliferation.

We studied healthy male military personnel beforedeployment. These individuals form a homogenous groupat the time of sampling to the extent that they were exposedto similar environmental factors such as training schedule,food, and vaccination. Moreover they were relatively homo-genous with respect to age, BMI, education, SES, and health.A relatively large proportion (27.7%) of the total variance inhostility was explained, of which 8.6% was accounted for bythe cytokine factors. This is promising for confirming thesefindings in a smaller sample size, if participants are chosen onthe extreme ends of the hostility scale.

This is the first study to our knowledge that uses factoranalysis on such a broad range of cytokines and chemokines toaggregate them into a smaller number of factors. It is impor-tant to note that the factors that emerged after factoranalysis represent functional clusters of cytokines. Wholeblood cultures were stimulated with a T-cell mitogen, whichincreases the release of several cytokines and chemokines.These, in turn, can affect the release of cytokines andchemokines by other cell types and also regulate the produc-tion of these factors by the T-cells themselves. Therefore, itis quite remarkable that functional clusters emerged. Onecluster that emerged from the factor analysis consisted of theTh1-type pro-inflammatory cytokines IL-2, TNFa, and IFNg.On the other side the Th2-type anti-inflammatory cytokinesIL-4, IL-5 and IL-10 formed a cluster. The anti-inflammatoryfactor showed negative correlation with the pro-inflamma-tory, which is in accordance with the notion that pro- inflam-matory cytokines or Th1 type cytokines inhibit production ofTh2 or anti-inflammatory cytokines (Elenkov et al., 2005).The third factor was IL-6 with the chemokines MCP-1, RANTES

is related to clusters of T-cell cytokines and chemokines in healthy05.007

+ Models

PNEC-1385; No of Pages 10

8 P.M.C. Mommersteeg et al.

and IP-10. IL-6 formed a separate factor with the chemo-kines, though IL-6 had a factor loading on the pro-inflamma-tory cytokine factor as well. It is known that while IL-6 haspro-inflammatory properties, it is also has a broad range ofother activities, which may explain why it did load on twodifferent factors (for review see Pedersen, 2007). Otherstudies have found similar results; Chan et al. (2002) andKoukkunen et al. (2001) have used factor analysis and wereunable to group IL-6 and TNFa in the same factor. Macro-phage Migration Inhibitory Factor formed a factor on its own.MIF is a regulator of both the innate and the adaptive immuneresponse, and is critically involved in inflammation. Func-tionally it is not unexpected that MIF is loading as a separatefactor, since MIF can function not only as a cytokine, but alsohas additional biological effects, e.g. it can act as a counter-regulator of glucocorticoid action (Calandra and Roger,2003).

A limitation of this study is the fact that the factor analysiswas explorative, and the results need to be replicated in anew sample to confirm the present findings. We had no apriori assumptions to what the outcome would be, and theclustering was based on observation of the factor analysisresults. However, factor loadings were high for the anti- andpro-inflammatory variables, and IL-6/chemokine and MIFformed distinct factors as well. Moreover, the factors thatemerged reflected biologically functional groups, which addsto our confidence that the clusters that were formed repre-sent true meaningful groups, instead of chance findings. MIFdid not fit well with any of the other factors and we chose tofit it in a separate factor rather than force it in with anotherfactor or leave it out.

Both factor analysis andmultiple cytokine determination ina single sample via multiplex analysis have been used pre-dominantly in cardiovascular research. Martins et al. (2006)performed a multiplex cytokine assay of 8 plasma cytokines.Coronary artery disease was positively related to IL-2, IL-4, IL-6, IL-12 and IL-18, however factor analysis was not used toconfirm the a priori group assignment of the cytokines (Martinset al., 2006). Tziakas et al. (2007) used factor analysis in 320patients with acute coronary syndrome. Three clusters wereidentified; a ‘‘systemic inflammation’’ cluster with CRP andfibrinogen, a ‘‘local inflammation-endothelial dysfunction’’cluster with IL-18 and ICAM-1 and an ‘‘anti-inflammation’’cluster comprising IL-10 and HDL cholesterol. The anti-inflam-matory cluster was a significant predictor of adverse cardiacevents (Tziakas et al., 2007). In the study of Chan et al. (2002)factor analysis in a group of 107 persons revealed 6 factorsrelated to components of the metabolic syndrome, a riskfactor for CVD. The cytokines IL-6 and TNFa were indepen-dently and significantly related to severalmetabolic syndromerelated factors. These studies further underline that a combi-nation of multiple cytokine determinations and factor analy-sis, as used in our study, can be a valid tool to improve ourunderstanding of the relation between psychological factors,cytokine/chemokines and health.

As hostility is a personality factor, a stable trait, it can besuggested that the observed relations with cytokine/chemo-kine release represent an intrinsic relation. This could renderpersons with a high hostile profile more vulnerable to adversehealth related outcomes as their adaptive immune systemresponds in a different manner than in persons with a lowhostile profile. Indeed it has been found that hostility is a risk

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitmen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008

factor for adverse health outcomes like PTSD (Ouimetteet al., 2004; Heinrichs et al., 2005; Orth and Wieland,2006), atherosclerosis and CVD (Siegman et al., 2000; Kop,2003; Everson-Rose and Lewis, 2005; Bunde and Suls, 2006),and disturbed immune functions have been described foradverse health outcomes as diverse as autoimmune diseases,infectious diseases, cardiovascular disease, and psychoso-matic conditions like depression and PTSD, presumablymediated by stress (Everson-Rose and Lewis, 2005; Kemenyand Schedlowski, 2007; Kendall-Tackett, 2007).

We have analyzed a cross-sectional dataset, which doesnot give us any information on the causality of our findings.The studied group will be followed after deployment, whichprovides a chance to determine whether the relationbetween hostility and cytokines before deployment as wedescribe it in this study is related to adverse health out-comes.

The relation between cytokine factors and hostility wereport here remained intact after controlling for age, BMI,smoking, drinking, and previous deployment, early lifetrauma and depression. It has been suggested before thathostility is related to CVD via risk factors as BMI, smoking anddrinking (Bunde and Suls, 2006). Hostility has been related toPTSD risk factors as early trauma and depression (Brewinet al., 2000; Stein et al., 2005; Gahm et al., 2006), which isconfirmed in our study. Depression has been related to pro-inflammatory cytokines like IL-6 and TNFa (Elenkov et al.,2005; Sjogren et al., 2006). When controlling for depressionand early life trauma the relation between cytokines andhostility increased slightly (the standardized coefficient b

increased), but the larger part of the relation remainedunaffected. This suggests first of all an independent under-lying mechanism in the relation between cytokines andhostility on the one hand and hostility and depression orearly trauma on the other hand. Our depression and earlytrauma scores were arbitrarily divided in two, and do notrepresent clinical scores. Nevertheless, it can be hypothe-sized that an increased depression or early trauma score mayenhance the relationship between hostility and cytokines/chemokines factors. This is in line with the study of Suarezand colleagues, who observed an additional positive effect ofthe combination of high hostility and high depression scoresin relation to plasma IL-6 levels (Suarez, 2003). Miller et al.(2003) on the other hand rather observed a relation betweenplasma IL-6 and high hostility, but low depression. We can,however, not directly compare our data to these studies sincewe used ex vivo cytokine release rather than plasma cyto-kines and since scores on hostility and depression were lowerin our sample.

The implications of our findings may become apparent inour follow-up studies when all the analyses will be repeatedafter deployment of the military personnel in Afghanistan.Our results imply an intrinsic relation between cytokines andhostility, which may be part of a preexisting vulnerability ofhostile persons. Therefore, based on our findings and pre-vious research we predict that higher levels of hostility and T-cell cytokine release before deployment are related to thedevelopment of adverse health outcomes.

In conclusion T-cell mitogen-induced cytokines and che-mokines formed functional clusters as identified by factoranalysis. These factors were significantly related to hostilityscores in healthy men.

y is related to clusters of T-cell cytokines and chemokines in healthy.05.007

+ Models

PNEC-1385; No of Pages 10

Hostility and stimulated T-cell cytokines 9

Role of funding source

This study was subsidized by a grant from the Ministry ofDefense of The Netherlands.

The Ministry of Defense had no role in study design, dataanalysis and interpretation of data, and in writing and sub-mitting the manuscript.

Acknowledgements

This study was subsidized by a grant from the Ministry ofDefense of The Netherlands. The authors are greatlyindebted to Col MD C. IJzerman and the commanders andtroops for their time and effort. We thank Kim Kroezen fororganizing inclusion of the participants. We also thank KarimaAmarouchi, Linda Schild, Hanneke Willemen, Marijke Ter-steeg-Kamperman, Esther Rudolph, Mirjam Maas, and JitskeZijlstra for excellent technical assistance in the framework ofthe PRISMO project.

Conflict of interest

The authors Eric Vermetten and Elbert Geuze are employedby the Ministry of Defense of The Netherlands. All otherauthors are employed by the University Medical CenterUtrecht. All authors declare that they have no conflicts ofinterest.

References

Arrindel, W.A., Ettema, J.H.M., 1981. Dimensionele structuur,betrouwbaarheid en validiteit van de Nederlandse bewerkingvan de Symptom Checklist (SCL-90) [Dimensional structure, relia-bility and validity of the Dutch translation of the SymptomChecklist (SCL-90)]. Nederlands Tijdschrift voor de Psychologie.Lisse, Swets & Zeitlinger, 77—108.

Babyak, M.A., 2004. What you see may not be what you get: a brief,nontechnical introduction to overfitting in regression-type mod-els. Psychosom. Med. 66, 411—421.

Barefoot, J.C., Dodge, K.A., Peterson, B.L., Dahlstrom, W.G., Wil-liams Jr., R.B., 1989. The Cook—Medley hostility scale: itemcontent and ability to predict survival. Psychosom. Med. 51,46—57.

Bremner, J.D., Vermetten, E., Mazure, C.M., 2000. Development andpreliminary psychometric properties of an instrument for themeasurement of childhood trauma: the early trauma inventory.Depress. Anxiety 12, 1—12.

Brewin, C.R., Andrews, B., Valentine, J.D., 2000. Meta-Analysis ofrisk factors for posttraumatic stress disorder in trauma-exposedadults. J. Consult. Clin. Psychol. 68, 748.

Bunde, J., Suls, J., 2006. A quantitative analysis of the relationshipbetween the Cook—Medley Hostility Scale and traditional coron-ary artery disease risk factors. Health Psychol. 25, 493—500.

Calandra, T., Roger, T., 2003. Macrophage Migration Inhibitory Fac-tor: a regulator of innate immunity. Nat. Rev. Immunol. 3, 791—800.

Chan, J.C., Cheung, J.C., Stehouwer, C.D., Emeis, J.J., Tong, P.C.,Ko, G.T., Yudkin, J.S., 2002. The central roles of obesity-asso-ciated dyslipidaemia, endothelial activation and cytokines in themetabolic syndrome–—an analysis by structural equation model-ling. Int. J. Obes. Relat. Metab. Disord. 26, 994—1008.

Christensen, U., Lund, R., Damsgaard, M.T., Holstein, B.E., Ditle-vsen, S., Diderichsen, F., Due, P., Iversen, L., Lynch, J., 2004.

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitymen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008.

Cynical hostility, socioeconomic position, health behaviors, andsymptom load: a cross-sectional analysis in a Danish population-based study. Psychosom. Med. 66, 572—577.

De Jager, W., Te Velthuis, H., Prakken, B.J., Kuis, W., Rijkers, G.T.,2003. Simultaneous detection of 15 human cytokines in a singlesample of stimulated peripheral blood mononuclear cells. Clin.Diagn. Lab. Immunol. 10, 133—139.

Donders, A.R., van der Heijden, G.J., Stijnen, T., Moons, K.G., 2006.Review: a gentle introduction to imputation of missing values. J.Clin. Epidemiol. 59, 1087—1091.

Elenkov, I.J., Iezzoni, D.G., Daly, A., Harris, A.G., Chrousos, G.P.,2005. Cytokine dysregulation, inflammation and well-being. Neu-roimmunomodulation 12, 255—269.

Everson-Rose, S.A., Lewis, T.T., 2005. Psychosocial factors and car-diovascular diseases. Annu. Rev. Public Health 26, 469—500.

Gahm, G.A., Lucenko, B.A., Retzlaff, P., Fukuda, S., 2006. Relativeimpact of adverse events and screened symptoms of posttrau-matic stress disorder and depression among active duty soldiersseeking mental health care. J. Clin. Psychol. 63, 199—211.

Gidron, Y., Gilutz, H., Berger, R., Huleihel, M., 2002. Molecular andcellular interface between behavior and acute coronary syn-dromes. Cardiovasc. Res. 56, 15—21.

Heinrichs, M., Wagner, D., Schoch, W., Soravia, L.M., Hellhammer,D.H., Ehlert, U., 2005. Predicting posttraumatic stress symptomsfrom pretraumatic risk factors: a 2-year prospective follow-upstudy in firefighters. Am. J. Psychiatry 162, 2276—2286.

Hovens, J.E., van der Ploeg, H.M., Bramsen, I., Klaarenbeek, M.T.,Schreuder, J.N., Rivero, V.V., 1994. The development of the self-rating inventory for posttraumatic stress disorder. Acta Psychiatr.Scand. 90, 172—183.

Jorgensen, R.S., Frankowski, J.J., Lantinga, L.J., Phadke, K., Spraf-kin, R.P., Abdul-Karim, K.W., 2001. Defensive hostility and cor-onary heart disease: a preliminary investigation of male veterans.Psychosom. Med. 63, 463—469.

Kemeny, M.E., Cohen, F., Zegans, L.S., Conant, M.A., 1989. Psycho-logical and immunological predictors of genital herpes recur-rence. Psychosom. Med. 51, 195—208.

Kemeny, M.E., Schedlowski, M., 2007. Understanding the interactionbetween psychosocial stress and immune-related diseases: astepwise progression. Brain Behav. Immun. 21, 1009—1018.

Kendall-Tackett, K.A., 2007. Inflammation, cardiovascular disease,and metabolic syndrome as sequelae of violence against women:the role of depression, hostility, and sleep disturbance. TraumaViolence Abuse 8, 117—126.

Kiecolt-Glaser, J.K., Loving, T.J., Stowell, J.R., Malarkey, W.B.,Lemeshow, S., Dickinson, S.L., Glaser, R., 2005. Hostile maritalinteractions, proinflammatory cytokine production, and woundhealing. Arch. Gen. Psychiatry 62, 1377—1384.

Kiecolt-Glaser, J.K., Malarkey,W.B., Chee, M., Newton, T., Cacioppo,J.T., Mao, H.Y., Glaser, R., 1993. Negative behavior duringmaritalconflict is associated with immunological down-regulation. Psy-chosom. Med. 55, 395—409.

Kop, W.J., 2003. The integration of cardiovascular behavioral med-icine and psychoneuroimmunology: new developments based onconverging research fields. Brain Behav. Immun. 17, 233—237.

Koukkunen, H., Penttila, K., Kemppainen, A., Halinen, M., Penttila,I., Rantanen, T., Pyorala, K., 2001. C-reactive protein, fibrino-gen, interleukin-6 and tumour necrosis factor-alpha in theprognostic classification of unstable angina pectoris. Ann.Med. 33, 37—47.

MacCallum, R.C., Zhang, S., Preacher, K.J., Rucker, D.D., 2002. Onthe practice of dichotomization of quantitative variables. Psy-chol. Methods 7, 19—40.

Martins, T.B., Anderson, J.L., Muhlestein, J.B., Horne, B.D., Carl-quist, J.F., Roberts, W.L., Carlquist, J.F., 2006. Risk factor ana-lysis of plasma cytokines in patients with coronary artery diseaseby a multiplexed fluorescent immunoassay. Am. J. Clin. Pathol.125, 906—913.

is related to clusters of T-cell cytokines and chemokines in healthy05.007

+ Models

PNEC-1385; No of Pages 10

10 P.M.C. Mommersteeg et al.

McEwen, B.-S., 2000. Allostasis and allostatic load: implications forneuropsychopharmacology. Neuropsychopharmacology 22, 108—124.

Miller, G.E., Freedland, K.E., Carney, R.M., Stetler, C.A., Banks,W.A., 2003. Cynical hostility, depressive symptoms, and theexpression of inflammatory risk markers for coronary heart dis-ease. J. Behav. Med. 26, 501—515.

Mills, P.J., Dimsdale, J.E., Nelesen, R.A., Dillon, E., 1996. Psycho-logic characteristics associated with acute stressor-inducedleukocyte subset redistribution. J. Psychosom. Res. 40,417—423.

Nemeroff, C.B., Bremner, J.D., Foa, E.B., Mayberg, H.S., North, C.S.,Stein, M.B., 2006. Posttraumatic stress disorder: a state-of-the-science review. J. Psychiatr. Res. 40, 1—21.

Orth, U., Wieland, E., 2006. Anger, hostility, and posttraumatic stressdisorder in trauma-exposed adults: a meta-analysis. J. Consult.Clin. Psychol. 74, 698—706.

Ouimette, P., Cronkite, R., Prins, A., Moos, R.H., 2004. Posttraumaticstress disorder, anger and hostility, and physical health status. J.Nerv. Ment. Dis. 192, 563—566.

Pedersen, B.K., 2007. IL-6 signalling in exercise and disease. Bio-chem. Soc. Trans. 35, 1295—1297.

Schafer, J.L., Graham, J.W., 2002. Missing data: our view of the stateof the art. Psychol. Methods 7, 147—177.

Siegman, A.W., Townsend, S.T., Civelek, A.C., Blumenthal, R.S.,2000. Antagonistic behavior, dominance, hostility, and coronaryheart disease. Psychosom. Med. 62, 248—257.

Sjogren, E., Leanderson, P., Kristenson, M., Ernerudh, J., 2006.Interleukin-6 levels in relation to psychosocial factors: studies

Please cite this article in press as: Mommersteeg, P.M.C., et al., Hostilitmen, Psychoneuroendocrinology (2008), doi:10.1016/j.psyneuen.2008

on serum, saliva, and in vitro production by blood mononuclearcells. Brain Behav. Immun. 20, 270—278.

Smith, T.W., Glazer, K., Ruiz, J.M., Gallo, L.C., 2004. Hostility, anger,aggressiveness, and coronary heart disease: an interpersonalperspective on personality, emotion, and health. J. Pers. 72,1217—1270.

Stein, A.L., Tran, G.Q., Lund, L.M., Haji, U., Dashevsky, B.A., Baker,D.G., 2005. Correlates for posttraumatic stress disorder in GulfWar veterans: a retrospective study of main and moderatingeffects. J. Anxiety Disord. 19, 861—876.

Suarez, E.C., 2003. Joint effect of hostility and severity of depressivesymptoms on plasma interleukin-6 concentration. Psychosom.Med. 65, 523—527.

Suarez, E.C., Lewis, J.G., Kuhn, C., 2002. The relation of aggression,hostility, and anger to lipopolysaccharide-stimulated tumornecrosis factor (TNF)-[alpha] by blood monocytes from normalmen. Brain Behav. Immun. 16, 675—684.

Tabachnick, B.G., Fidell, L.S., 2007. Using Multivariate Statistics.Pearson Education Inc., Boston.

Tziakas, D.N., Chalikias, G.K., Kaski, J.C., Kekes, A., Hatzinikolaou,E.I., Stakos, D.A., Tentes, I.K., Kortsaris, A.X., Hatseras, D.I.,2007. Inflammatory and anti-inflammatory variable clusters andrisk prediction in acute coronary syndrome patients: a factoranalysis approach. Atherosclerosis 193, 196—203.

Witteveen, A.B., Van der Ploeg, E., Bramsen, I., Huizink, A.C.,Slottje, P., Smid, T., Van der Ploeg, H.M., 2006. Dimensionalityof the posttraumatic stress response among police officers andfire fighters: an evaluation of two self-report scales. PsychiatryRes. 141, 213—228.

y is related to clusters of T-cell cytokines and chemokines in healthy.05.007