Emotional Repression, Stress Disclosure Responses, and Epstein Barr Viral Capsid Antigen Titers

Emotional Repression, Stress Disclosure Responses, and Epstein-Barr Viral Capsid Antigen Titers

BRIAN A. ESTERLING, B.A., MICHAEL H. ANTONI, PH.D., MAHENDRAKUMAR, PH.D., NEIL SCHNEIDERMAN, PH.D.

Based on the theory of psychosomatic inhibition, we hypothesized that subjects who abstainedfrom disclosing emotional material on a laboratory task would have poorer control of latentEpstein-Barr virus (as evidenced by high titers for the viral capsid antigen), and similarly, thosesubjects with psychometrically derived repressive interpersonal styles would show the highestEpstein-Barr viral capsid antigen titers (EBV-VCA). Eighty first-year undergraduates completeda personality inventory and were asked to write an essay/letter for 30 minutes about a stressfulevent that had happened in their life. Blood was collected from each subject immediately afterwriting. Essays were scored for degree of emotional disclosure according to the ratio ofemotional-to-total words used. Degree of disclosure was found to be associated with impairedcontrol of latent EBV (high antibody titers to the EBV-VCA) controlling for medication use,recent sleep loss, physical activity, lean body mass, caloric intake, and alcohol and recreationaldrug use. Further, individual differences in interpersonal style (characterized by emotionalsuppression) related to this immunologic marker in a similar fashion, and these two factorsinteracted in determining EBV-VCA titers. That is, Repressers who were either high or lowdisclosers had high levels of antibody titer to EBV-VCA, whereas only those Sensitizers whodid not disclose had high antibody titers to EBV-VCA. In addition to supporting the hypothesisthat emotional repression is associated with some aspects of host-virus interaction, the presentfindings highlight the importance of obtaining behavioral and psychometric assessments inpsychoimmunologic investigations of this abstract affective construct (i.e., repression).

INTRODUCTION

Recently, some attention has been paidto the relationship between cellular im-mune deficits and affective disruption (1-6). It has been proposed that inhibiting theexpression of emotional material relatedto experienced stressful events may leadto immunological consequences (7, 8).

From the Department of Psychology (B.A.E.,M.H.A., N.S), the Department of Psychiatry (M.K.,N.S.), and the Department of Medicine (N.S.), Uni-versity of Miami, Coral Gables, Florida.

Address reprint requests to: Dr. Michael H. An-toni, University of Miami, Department of Psychol-ogy, P.O.Box 248185, Coral Gables, Fl 33124.

Received November 27, 1989; revision receivedMay 2, 1990

Numerous studies have demonstrated in-dividual differences in the tendency toreport experienced emotions (e.g., anxi-ety, tension, anger, depression) across var-ied conditions of challenge (9, 10) andsuch tendencies characterized as repres-sive personality and anger suppressionhave been related to increased risk orworsened clinical course of neoplastic dis-ease (11-16). Since the course of someneoplastic diseases may be mediated byimpairments in cellular immune surveil-lance, these psychosocial factors havebeen postulated to relate to such diseaseprocesses, in part, by way of immunomod-ulation.

In addition to tumor surveillance, cel-lular immune competence has beenthought to be a critical factor in control-

Psychosomatic Medicine 52:397-410 (1990) 397

0033-3174/90/5204-0397S02 00/0Copyright © 1990 by the American Psychosomatic Society

ESTERLING et al.

ling primary herpes virus and Epstein-Barr virus (EBV) infections, as well assubsequent control of latent viral reacti-vation (17). Interestingly, experiencing astressful event (e.g., marital disruption)has been associated with greater mentaldepression (decreased emotional expres-sion and increased rumination) and im-pairments in several indices of cellularimmune function, including a reductionin control of latent EBV (18, 19). Othershave also shown a relationship betweendecreased emotional expression, in-creased rumination, and decreased con-trol of latent EBV (20, 21).

Recently, Pennebaker et al. (8) demon-strated substantial variability in emo-tional disclosure among college studentswho were given directions to revealstress-related emotional material in writ-ten format, and this difference was relatedto immune function (i.e., mitogenic stim-ulation responses). What was not ad-dressed in this study was whether differ-ences in immune competency could bedue to situational factors or subjects' be-havioral predispositions which could beoperating independent of the emotion-re-lated task at hand. Moreover, these inves-tigators did not attempt to relate emo-tional disclosure to immunologic markersof latent EBV control, thereby making itdifficult to integrate these findings withprevious studies of EBV.

The purpose of the current study wasto assess stable interpersonal characteris-tics and differential disclosure responsesindicative of emotional repression to astandardized stimulus among healthy col-lege students, and to relate these differingstyles to differences in immunologicalcontrol of EBV viral capsid antigen (VCA).Specifically, we hypothesized that sub-jects with repressive personalities would

have poorer control of latent EBV andsimilarly, those who abstained from dis-closing emotional material on a laboratorytask would also show the highest EBV-VCA titers. The present findings sup-ported the hypotheses that inhibitingemotional material in the face of discuss-ing a stressful event is associated withimpaired control of latent EBV, that indi-vidual differences in interpersonal style(characterized by emotional suppression)are associated with this immunologicmarker in a similar fashion, and that thesetwo factors interact in determining EBV-VCA titers.

Methods

Exclusion CriteriaIn order to reduce extraneous sources of variance

in immunologic measurements due to subject differ-ences in medication use, physical status, and certainpotentially immunomodulatory behaviors, severalexclusionary criteria were employed in the presentstudy. Subjects were excluded from the study if theyreported use of prescription medications includingantibiotics within the past month, or cortisol or otheranabolic steroids within the last six months. Subjectswho reported heavy cigarette smoking (>1 pack perweek), or excessive alcohol intake (>10 drinks perweek), or who were pregnant or who were currentlyundergoing psychotherapy were also excluded.Other exclusion criteria included psychiatric orother medical illness within the prior three months,or a self-reported history of chronic mental or phys-ical illness or skin reactive diseases (e.g., eczema).Further, subjects were required to be currently freeof symptoms such as fever, drowsiness, sore throat,mouth sores, and nasal congestion that might indi-cate an active upper respiratory or other infectionat the time of the study.

SubjectsNinety-three subjects between the ages of 17 and

25 (M = 18.4, SD = 1 3) were recruited from the first-year undergraduate class of which 80 (30 males, 50

398 Psychosomatic Medicine 52:397-410 (1990)

DISCLOSURE AND EBV

females) satisfied the inclusion criteria. These sub-jects were participating in order to fulfill the require-ments of an introductory psychology course.

ProcedurePrior to the commencement of the study, all sub-

jects completed a personality inventory, as well as abattery of questions comprising physical status itemsand questions pertaining to potential behavioralimmunomodulatory confounds (see "Self-ReportControl Measures"). All subjects were blind withrespect to the instructions and purpose of the study.For the study protocol, all subjects were escorted toa 4 X 5 foot private room which consisted of a table,chair, and a pad of paper where they were asked torelax for a 5-minute period. Following this baselineperiod, the subjects were asked to compose a letterto a close friend about an event which they hadexperienced and which had been highly stressful.All subjects were asked to select an event whichthey had not widely discussed with others.

All subjects were given an unlimited amount ofpaper and were allowed 30 minutes to completetheir essay/letter. Blood for immunologic assessmentwas collected between 1:00 PM and 4:00 PM on thesame day with the time elapsing between writingand blood collection ranging from 45 minutes to 4hours (M = 1.0 hr). In order to determine the stabilityof the writing paradigm, all subjects returned onemonth later and were given the same instructions towrite another essay/letter. All subjects were de-briefed en mass once the entire study was com-pleted.

Behavioral IndexEach essay was read independently by two judges

blind to the other data in the study, as well as toeach others' ratings. In scoring the essays, the num-ber of emotional words was computed in line withprevious research (8) in order to maximize interlabcomparisons in healthy college samples. Words werecomputed without respect to intensity, occurrence,or direction (i.e., positive and negative) of emotion.Emotional words were scored provided they satisfiedtwo criteria: (1) the word that describes a humanfeeling (e.g., she loves her mother) and (2) the worddoes not describe a physical action (e.g., they wentto make love). The difference between the criteria

is that any word describing a human emotion canhave a more intense (or less intense) emotional wordsubstituted in its place. In the example above, "sheloves her mother" can be turned into "she likes hermother" and still make sense. This cannot be donewith the second sentence (i.e., emotional word) be-cause the referent of the emotional word is the actof having sex. Thus the first word would have beenscored and the second would have not. The totalnumber of words in each essay was determined andpercent emotional words served as the index ofemotional repressiveness on this behavioral task.Subjects were divided into the top, middle, andbottom third of the distribution of percent emotionalwords; subjects who reported fewer than 1% (lowdisclosers, N = 26), between 1 % and 2% (middle disc-losers, N = 24) or greater than 2% (high disclosers,N = 30) were used in the statistical analyses. Onlydata from the first writing session (when blood wasdrawn) were used in the statistical analyses. Theseclassification criteria are consistent with previousresearch involving stress disclosure and immunefunction (8).

Personality IndexPrior to the day of essay writing, each subject had

completed the Millon Behavioral Health Inventory(MBHI; 22) which assesses individual differences ininterpersonal coping styles. This inventory is com-posed of 150 true-false forced choice items, is com-puter scored, and yields base rate scores pertainingto eight basic coping styles: Introversive, Inhibited,Cooperative, Sociable, Confident, Forceful, Respect-ful, and Sensitive. These eight basic coping stylesare based on the personality theory of Millon andhave a test-retest reliability over 4.5 months rangingfrom 0.77 to 0.88 with a mean of 0.82 (22, 23).

Subjects were classified into one of three person-ality groups depending on their elevations on eachof the eight coping scales as is consistent with thetheory underlying the instrument. One group wasdesignated as Repressive, a second group as Sensi-tizer, and a third comprised subjects with neitherpersonality style (NP). The groupings used for thisstudy were formed on the basis of high-point codesdetermined by a cut off point of 75 base rate pointsas prescribed by the MBHI test authors (22). Onlythose subjects who met these criteria for Sensitizers(elevations on scales 2, 6, and/or 8; N = 44), Repres-sors (elevations on scales 1,3, and/or 7; N = 8), andneither personality (elevations on scales 4 and/or 5;


ESTERLING et al.

N = 12) were included in the study analyses. Datafor the remaining 16 subjects who could not beclassified into any of the above groupings (i.e., nosignificant elevations) were not analyzed further.

As portrayed by the MBH1 (22), subjects withelevations on Repressive interpersonal styles osten-sibly have an inner need to deny negative feelingsto themselves and others, tend to appear content inthe face of problems, and may attempt to pleaseothers with self-sacrificing behaviors. Individualswith elevations on Sensitizer styles ostensibly comeacross to others as being overbearing, aggressive,rivalrous, and confident, tend to have a low level offrustration tolerance, and are quick to express theirnegative feelings.

Measurement of Antibody to EBV-VCA and EBV-EABlood samples were collected from all subjects

and the serum fractions were evaluated for IgG an-tibodies against the viral capsid antigen (VCA) andearly antigen (EA) of EBV using reagents supplied askits (Organon Teknika, No. 9100-11) by the indirectimmunofluorescence method (IFA) according toHenle and his colleagues (24). All serological assaysfor antibody determination were performed by 2-fold dilutions of serum samples in phosphate buff-ered saline (PBS), pH 7.4. The diluted samples wereadsorbed for 30 minutes at 37°C on the suppliedslides which contained acetone-fixed lymphocyticcells containing the EBV-VCA or EBV-EA. The cellswere washed with PBS and readsorbed with goatanti-human IgG conjugated to fluorescein lsothy-ocyanate (FITC) in an Evans Blue Counterstain com-bined preparation for 30 minutes at 37°C. The cellswere then washed with PBS, mounted with bufferedglycerine (pH 7.5), and examined immediately witha Nikon inverted DIAPHOT-TMD microscopeequipped with an epifluorescence attachment. Allsera were examined in triplicate, and antibody titersto EBV-VCA and EBV-EA were determined by thehighest dilution of serum able to demonstrate IFA-positive cells. All titer values are reported as themean log-transformed dilution factor. O/rhe original80 subjects, 26 were seronegative and were excludedfrom all immunological analyses. As a result, in allimmunological analyses, the group sizes are slightlyaltered to reflect this change. In the discloser groups,the sample sizes are 19, 20, and 15 for the low,middle and high disclosers, respectively. In the per-

sonality groups, the sample sizes are 13, 9, and 20for the repressors. neither, and sensitizers, respec-tively.

Self-Report Control MeasuresSubjects were asked to describe any medication

(e.g., prescription medication, cold formulas) takenin the previous month prior to participation. Theywere also asked to describe any infectious diseasesymptoms, particularly symptoms associated withrespiratory tract infections experienced within thelast one and six months. Consistent with previousmethodological suggestions in behavioral immuno-logic research (25), we also collected self-report dataon the amount of restful sleep, hard physical activity(sports and work-related), average caloric intake,and recreational drug and alcohol use occurring overthe previous six months. In addition, each subject'slean body mass (LBM) was calculated according tothe Metropolitan Life Insurance actuarial tableswhich take into account body weight and frame sizein determining the subject's deviation from his/herideal weight (i.e., weight which maximizes life ex-pectancy).

Results

Control Measures

Prior to any interpretation of the im-mune data, it was important to inves-tigate self-reported confounding variableswhich might account for the varianceobserved in the antibody outcome data.The descriptive statistics for these vari-ables in our sample are included inTable 1. To test the effect of all of thesepotential confounding variables account-ing (uniquely and in combination) for im-mune data variance, a stepwise multipleregression was performed. It was deter-mined that none of these variables con-tributed to a significant amount of thevariance in EBV-VCA titer variability,F(6,47) = 0.87, N.S. In fact, together, allconfounding variables accounted for 11%


DISCLOSURE AND EBV

of the variance and no single variableaccounted uniquely for more than 7%. Nosignificant effects were found when sub-ject gender was included as an additionalvariable, so data for both sexes were com-bined. In addition to testing for directeffects of the confounding variables (e.g.,sleep, lean body mass) on EBV-VCA titers,a MANOVA procedure was performed foreach psychosocial factor (essay and MBHIgroups) to assess group level differenceson these potentially confounding vari-ables. It was determined that in combi-nation, none of these variables differedamong groups classified by either behav-ioral response (F(5,48) = 0.47, N.S.) orpersonality cluster, F(10,104) = 0.91, N.S.

As a further control, antibody to Ep-stein-Barr virus early antigen (EBV-EA)was assessed to determine if any of thesubjects were recovering from a primaryEBV viral infection. Subjects in such a"recovery" phase might show EBV-VCAlevels that are not representative of latentvirus control and more indicative of aprimary response to initial viral challenge(26). It was found that all subjects hadantibody titers to EBV-EA in the normalrange (i.e., a Iog2 dilution factor of 3.32which corresponds to a dilution factor of1:10) suggesting that the sample was EBV-VCA seropositive yet not undergoing aprimary viral infection (27).

TABLE 1. Means (+SEM) for SampleCharacteristics [N =

Age (years)Physical Activity (hours/day)Drug use (# days in month)Kilocaloric Intake (per day)Sleep (hours/night)Lean Body Mass (% of normal')

= 80)

18.48(0.15)3.90 (0.55)1.36(0.20)1.99(0.07)7.04 (0.13)

97.14 (1.98)1 Normal values are based on 1979 Build Study, Soci-ety of Actuaries, and Association of Life InsuranceMedical Directors of America, 1980.

Repressive Behaviors

The interrater reliability of the essayscoring procedure was assessed by com-puting a Pearson correlation coefficientfor percent emotional word scores as de-termined by the two raters. The highlysignificant interrater correlation (r = 0.90,p < 0.001) and the absence of a differencein mean percent emotional words scored(M, = 1.73, SD = 0.94; M2 = 1.67, SD =1.0) between the raters (t(79) = 1.19, N.S.)suggests that this scoring procedure was areliable one.

In order to test for the stability of thewriting task (to what degree it was influ-enced by state fluctuations), all subjectswere asked to return to the lab one monthafter the initial writing and blood draw.Using the entire sample of subjects whocompleted both essays (N = 78), it wasdetermined that the correlation betweenthe emotional words of the first writingand that of the second was 0.34 suggestingthat even though this measure is some-what stable, it may be influenced by statefluctuations over time.

As in previous studies using similar par-adigms, the content of the essays sug-gested that subjects uniformly disclosedhighly personal and upsetting experi-ences. Overall, the primary topics of theessays were as follows: 38% focused oninterpersonal conflicts, 12% loss and lone-liness associated with leaving home, 17%death of a significant other, 16% illness orinjury, 9% on parental problems (e.g., di-vorce), 6% handling academic life, 5%suicidal ideation, 4% humiliation, 3%stressful sport experience that was not anaccident, 1% feelings of depression, and1% sexual abuse by a friend. Even thoughthe subjects were blind with respect towhat they would be writing about, theinstructions pulled for an emotional re-


ESTERLING et al.

sponse. It is interesting that some individ-uals decided to only think about the fac-tual dimension of the stressful act,whereas others focused on the emotionaldimension. As a result, we were able toget a range of responses from no emotionalwords to a few individuals who wrote upto 50 emotional words in a single 400-word essay (12.5% emotional words!). Inline with our first hypothesis, antibodytiter values among the three disclosuregroups were significantly different,(F(2,51) = 17.10, p < 0.001) and a Tukeypost-hoc analysis (p < 0.05) confirmedthat the high disclosers (as measured onthis index) had lower levels of IgG anti-body to EBV-VCA (M = 7.14, SD = 1.14)compared to low disclosers (M = 8.77, SD= 0.70) and middle disclosers (M = 8:21,SD = 0.64) who did not differ from eachother (see Figure 1).

Repressive PersonalityIn order to test the validity of the scaled

scores on the different levels of repres-sion, we compared the base rate eleva-tions for each of the Repressor and Sen-sitizer scales. Consistent with the test au-thors, it was determined that theRepressor scales 1 (Introspective, F(l,50)= 4.26, p < 0.05), 3 (Cooperative, F(l,50)= 39.57, p < 0.001), and 7 (Respectful,F(l,50) = 3.13, p < 0.05) were significantlyhigher for the Repressor personality groupcompared to the Sensitizer personalitygroup. Further, the Sensitizer scales 6(Forceful, F(l,50) = 54.57, p < 0.001), and8 (Sensitive, F(l,50) = 9.63, p < 0.01) weresignificantly higher for the Sensitizer per-sonality groups compared to the Repres-sor personality group. The one anomalyis the Sensitizer scale 2. There was nodifference between the Sensitizer and the

Low Discloser Middle Discloser High DiscloserBehavioral Response

Fig. 1. Mean EBV-VCA antibody titer values (±SEM) for discloser groups based on behav-ioral (essay) index.

Repressor personality groups on this scale(F(l,50) = 1.56, N.S.).

In line with our second hypothesis, an-tibody titer values among the three per-sonality groups were significantly differ-ent (F(2,39) = 5.28, p < 0.01) and a Tukeypost-hoc analysis (p < 0.05) confirmedthat the Repressors had a significantlyhigher level of antibody titer (M = 8.79,SD = 0.54) as compared to both the Sen-sitizer (M = 7.47, SD = 1.09) and NP (M =7.63, SD = 1.39) groups (see Figure 2).

In investigating the association betweenthe behavior and personality classifica-tion systems, it was determined that theemotional words expressed among thethree personality groups were signifi-cantly different (F(2,61) = 5.24, p < 0.01)and a Tukey post-hoc analysis (p < 0.05)confirmed that the Repressors were writ-ing significantly fewer emotional words(M = 1.04, SD = 0.65) as compared to theSensitizers (M = 2.01, SD = 0.97). Theneither personality group had an inter-mediate number of emotional words (M= 1.77, SD = 1.41) which were not signif-


DISCLOSURE AND EBV

Repressor Neither Sensilizer

Personality Group

Fig. 2. Mean EBV-VCA antibody titer valuesSEM) for personality groups.

icantly different from either of the othertwo extremes. Thus, we were able to dem-onstrate criterion validity with our be-havioral assessment procedure using thepersonality instrument as the criterion.

In an attempt to highlight the contri-bution of the extremes on the behavioraltask, and its relationship to the personal-ity data, a 2 x 3 factorial design was uti-lized and a significant interaction be-tween personality groups and behavioralresponse was found, F(2,24) = 3.49, p <0.05. Inspection of cell means suggeststhat within the Repressive personalitygroup, there is no difference in antibodytiter values between levels of the behav-ioral factor (high discloser, M = 8.65, SD= 0.57, N = 4; low discloser, M = 8.85, SD= 0.58, N = 8). Among subjects in theSensitizer personality group, however,high disclosers show lower antibody val-ues (M = 6.68, SD = 0.95, N = 9) than lowdisclosers, M = 8.46, SD = 0.50, N = 3,F(l, 10) = 9.08, p < 0.01. Subjects in theNP group follow a similar pattern to Sen-sitizers with antibody values approaching

significance for the high disclosers (M =6.20, SD = 0.76, N = 3) vs. low disclosers(M = 8.41, SD = 1.53, N = 3), F(l,4) = 5.00,p < 0.10 (see Figure 3).

Finally, all of the confounding variables(drug use, diet, sleep, activity, LBM, andgender), followed by all of the predictorvariables (disclosure response, personal-ity group, and interaction) were enteredin a hierarchical multiple regressionmodel to predict EBV-VCA antibody titer.The overall model with all variables wassignificant (F(8,27) = 4.78, p < 0.001). Acommonality analysis determined theproportion of variance in antibody titerthat each of the predictor variables con-tributed uniquely (i.e., disclosure re-sponse, personality group and interac-tion). Controlling for the confoundingvariables, it was determined that 24.5%(p < 0.05) of the variance in EBV-VCAcan be accounted for by the disclosureresponse, 21.7% (p < 0.05) by personalitygroup, and an additional 3.5% (N.S.) bytheir interaction. Taken independently,these three predictors accounted for55.6% of the total variance in EBV-VCAand comprise the following predictormodel: Y' =10.82 + 2.14 (Behavioral re-sponse) + 1.78 (Personality cluster) — 1.45(Interaction).

DISCUSSION

This study investigated the relationshipamong concurrent individual differencesin the use of emotional adjectives to de-scribe a recent stressful event, interper-sonal style, and host immune functioningwith respect to latent Epstein-Barr virus.In previous studies, it has been demon-strated that writing about a stressful, emo-tionally intense experience has positiveeffects on the blastogenic response of T-


ESTERLING et al.

10-,

9 -

7 -

6 -

o Low DiscloserHigh Discloser

Repressor Neither Sensitizer

Personality Group

Fig. 3, Mean EBV-VCA antibody values (±SEM) as a function of behavioral response and personalityfactors.

lymphocytes to two mitogens, as well ason autonomic arousal levels (8). In addi-tion, significant changes in antibody toEpstein-Barr virus and EBV-specific T-cell killing have been shown to accom-pany stressful periods involving medicalschool examinations (28). In this study weattempted to integrate these paradigmsand found a linkage between writingabout a stressful event and resting levelsof antibody to EBV-VCA in a populationwhich was similar in age and socioeco-nomic status to those previously studiedcollege undergraduates. In the presentstudy, high disclosers (on a behavioraltask) and those subjects with personalitiescharacterized by emotional expression(Sensitizers) had significantly lower levelsof antibody to EBV-VCA as compared tothose subjects who were low disclosersand/or who had a Repressive personality.

In human psychoimmunologic re-search, the underlying assumption is thatif a behavior change is noted/instituted ina group and there is a subsequent changein immune function, that this immunechange is due, to some extent, to the psy-chological phenomenon which the inves-tigators observed or manipulated. Such anassumption would be unjustified, how-ever, if differences in other confoundingvariables may equally, or more com-pletely, account for the observed immu-nological change(s). For example, in in-vestigating these relationships in a groupof individuals experiencing an emotion-ally stressful event (e.g., death of spouse),there might be changes in nutritional sta-tus, quality of sleep, alcohol intake, ciga-rette smoking frequency, and physical ac-tivity as part of the bereavement process(29) any of which may impact immunity.


DISCLOSURE AND EBV

While there is some limited evidence thatvery gross alterations in these behaviorslead to changes in immune functioning(30, 31), less is known regarding theimmunomodulatory effects of minor fluc-tuations. For all of these reasons, we feltthat it was crucial that potential immuno-modulatory confounding variables be as-sessed to determine if they accounted forany of the observed variance in EBV-VCAantibody levels in the present study. It isimportant that none of the confoundingvariables assessed significantly contrib-uted to EBV-VCA titer variance, and thatthere was no difference between the highand low disclosers, nor between Sensi-tizer or Repressor personality groups.

Regarding the essays written by sub-jects, despite the large variance in theamount of emotional material expressed,there was little variability in the plot andsubject matter. Both low discloser andhigh discloser essays can be found whichaddressed such issues as breaking up witha significant other, contemplating and at-tempting suicide, and witnessing thedeath of a brother, friend, or unknownindividual. This suggests that our assess-ment paradigm was able to evaluate anemotional response category which mayhave been relatively independent of thetype of stressor recalled.

The elevated EBV-VCA antibody titers(M = 8.77, SD = 0.70) noted in our lowdisclosure group suggested that there wassome compromise in the viral antibodyresponse among these subjects. PreviousEBV seroepidemiological studies haveshown characteristic EBV-VCA IgG geo-metric group means of the raw titers(GMTs) ranging from 1:10 to 1:80 (log2 =3.32 and 6.32, respectively) in normaladult populations (24) with extreme casesranging from having no antibody (neverencountering the virus) to greater than

1:640 (log2 = 9.32). The exceptions to thisgeneral serologic pattern have been doc-umented in patients who are immunolog-ically compromised due to immunosup-pressive diseases or therapy, as well asthose with Burkitt's lymphoma and na-sopharyngeal carcinoma-two malignantdiseases with which EBV is associated(32). In these individuals, the antibodytiter may be 1:1280 (log2 = 10.32) andgreater. Elevated antibody titers in indi-viduals not suffering from these condi-tions are thought to occur in response toenhanced expression of latent virus dueto changes in, or dysfunction of, differentT-lymphocyte subpopulations (24, 33, 34).Unfortunately, because all of the abovework has involved in vitro assays, it isdifficult to determine with certainty justhow the viral genome is latently con-trolled, and how it subsequently becomesreactivated. Further investigation in thisarea of emotions and behavior may yieldinformation relevant to the mechanism ofEBV-VCA genome expression.

In reviewing the literature, it is clearthat our results are consistent with earlierwork (8, 28). Pennebaker et al. (8) usedthis writing task and assigned subjects tofactual, emotional, or factual-plus-emo-tional conditions to examine the effects ofdisclosure manipulation on mitogen stim-ulation responses. What was interestingabout this study was that the researcherstook into consideration within group dif-ferences in emotional words written bysubjects in the "factual-plus-emotional"condition. They identified a group of highand a group of low disclosers based onemotional word output and found that thehigh disclosers had a larger lymphocyteresponse to ConA stimulation than thelow disclosers. These findings suggest thatthe immune changes noted were inde-pendent of the experimental manipula-


ESTERLING et al.

tion and due rather to individual differ-ences in response proclivities just as ourdata have indicated.

What potentially might be most impor-tant in our findings is the interaction be-tween the personality (interpersonal)style and the behavioral (disclosure) re-sponse with respect to the level of EBV-VCA titer. In the present design, we firstwanted to address the differences amonggroups for main and interaction effects.Next we determined the proportion ofunique variance accounted for by the in-teraction term. The significant interactioneffect suggested one or more significantsimple main effects of discloser levelwithin (any or all of) the personalitygroups. When entered into a regressionequation, however, the interaction termwas not significant. This data is inter-preted to suggest that there are significantdifferences between the groups that makeup the interaction (i.e., six cells). Further,this suggests that there is considerableoverlap of shared variance with respectto the main effects (i.e., personality styleand behavioral response). Because of thislarge overlap, there is little unique vari-ance in antibody titers left over for theinteraction to predict in the regressionequation. Thus, the proportion of uniquevariance that the interaction accounts foris small because much of its contributionis shared with the other factors, resultingin a nonsignificant unique contribution.

One explanation of this interaction isthe following. With respect to personalitystyle, it was determined that subjects withelevations on Repressive styles had higherlevels of antibody to the EBV-VCA thanthose with Sensitizer elevations. Simi-larly, those who refrained from emotionalexpression (low disclosers) in response toa standardized stimulus had higher anti-body titers as compared to high disclosers.

What is more interesting is that withinthe Repressive personality group, thelevel of disclosure evidenced on the lab-oratory task was not associated with dif-ferences in antibody titers whereas withinthe Sensitizer personality group thosewho engaged in a high degree of disclo-sure on the task showed evidence of thegreatest degree of viral control (i.e., lowestEBV titers). However, the low-disclosingSensitizers had antibody levels as high asthose who were classified as Repressors!If individuals are aware of but choose notto disclose the emotional intensity of thestressful events because of embarrass-ment, fear of punishment, or other rea-sons, the consequence may be to holdback or inhibit their thoughts, feelings,and behaviors from others. According toour data, this pattern of inhibition may bemost salient for individuals who are es-pecially sensitive (i.e., Sensitizers) to en-vironmental stressors.

Although highly speculative, it is pos-sible that a rumination component is im-portant for the Sensitizer personalitytypes because these individuals both ex-perience and actively repress the emo-tional material with respect to the event.Evidence for this hypothesis comes fromthe nonsignificant difference between theSensitizers and the Repressors with re-spect to scale 2 (inhibited) of the MBHI.This scale is clinically related to the ex-perience of emotion (i.e., grouped intoSensitizers), but is confounded by the factthat these individuals tend to be interper-sonally inhibited (22) and may thereforenot publicly reveal the emotions that areexperienced. It is possible that this subsetof the Sensitizers is repressing experi-enced emotion which is reflected in thepoorer control over the latent EBV that isobserved in the Sensitizer-low disclosersubgroup. These data suggest that both


DISCLOSURE AND EBV

dispositional personality style and behav-ioral responses to standardized stimulineed to be incorporated in assessments ofthe repression construct and its relation-ship to host immune status. Even thoughthe sample size of this interaction com-ponent is small and one needs to be cau-tious in making interpretations, we be-lieve that these results are worthy of fu-ture study.

It has been hypothesized that to inhibitongoing behavior, thoughts, or feelings,"physiological work" must be performed(35, 36). Immediately following a stressfulevent, inhibition is accompanied by in-creases in autonomic nervous activity (35)and by increased sympathetic firing in theseptal and hippocampal regions of thebrain (36). It is plausible to propose that ifthis behavior pattern continues, the workof inhibition may serve as a low levelchronic stressor which has the potentialof long-term cumulative damage. If thishypothesis is correct, it may explain theimmunologic differences noted in thoseindividuals who typically express emo-tion in response to a stressful event versusthose individuals who typically expressfactual information following a similarevent.

In searching for a mechanism to linkrepressive personality, disclosure behav-ior and the immune system, a logicalcourse would be to investigate neuroen-docrine pathways. It has been shown thatfor those individuals who demonstrate af-fective disruption and low defensive re-serve, there is a high correlation betweenrepressive behavior, personality andplasma cortisol secretion rate following astressful event (37). This finding may berelevant as it has been shown that cortisolimpairs several components of cell-me-diated immunity (38, 39). Vickers (37) pro-vided cumulative evidence from five

studies (40-44) that over 16% of the vari-ance in cortisol levels could be predictedfrom emotionality scores. These studiesprovide suggestive evidence for a neu-roendocrine mechanism which could ex-plain our findings.

There is some controversy in the liter-ature over which psychological variablescontribute significantly to immune func-tioning. In view of this, we believe thatthe present study is a significant contri-bution to resolving this literature and thatfuture work should also take into consid-eration personality style as well as stressdisclosure-related behavioral responseswhen conducting human psychoimmu-nological investigations of the repressionconstruct. However, even though we de-fined our variables as behavioral (use ofemotional adjectives) and psychometric(interpersonal style), we believe that fu-ture studies should include a controlgroup which initially writes about a triv-ial event and then writes again at a latertime about a stressful event. Such a designcould determine with certainty that thebehavioral response is in fact a state meas-ure and is not a function of personalitystyle. That is, it is quite possible that thehigh discloser group could have displayeda large proportion of emotional words indescribing less emotional or trivial events.It is also quite possible that our behavioralmeasure was measuring emotional reac-tivity in contrast to emotional disclosure.This, of course, would have to be deter-mined empirically, but could have impor-tant implications for the interpretation ofthe results if this were to be the case. Inaddition, in that it has been shown thatother psychosocial phenomena have sim-ilar effects on latent EBV titer (19, 28, 45,46), possible sources of the unexplainedvariance in the present study might in-clude transient intrasample differences in


ESTERLING et al.

psychosocial stressors (e.g., examinations,divorce) and physiological status (e.g.,menstrual cycle) not measured in thisstudy. In that we did not have any controlfor the stress experienced, we cannot ruleout the possibility that the repressors/lowdisclosers experienced nonrandom stress-ful experiences. Perhaps having all sub-jects watch a film of a brutal murder andto then talk about it could be used to testthis possibility. The only assurances thatwe have with respect to severity of expe-rience is that our population, as a whole,were freshman who had just left home/

friends/lovers and had a common expe-rience to tap if they wished to do so.Unfortunately, we did not ask the subjectto rate the stressfulness of the experience,but only to choose a very stressful one.Thus, there are no assurances that theirrecent experiences were comparableprior to the study, only that they were allperceived as stressful. An experimentaldesign should be used to test the validityof these results unconfounded by otherpsychosocial and physiological phenom-

REFERENCES

1. Cosyns P. Maes M, Vandewoude M. Stevens WJ, De Clerck LS, Schotte C: Impaired mitogen-inducedlymphocyte responses and the hypothalamic-pituitary-adrenal axis in depressive disorders. J AffectiveDisord 16(l):41-48, 1989

2. Darko DF. Lucas AH, Gillin JC, Risch SC. Goldshan S, Hamburger RN, Silverman MB, Janowsky DS:Age, cellular immunity and the HP axis in major depression. Prog Neuropsychopharmacol Biol Psy-chiatry 12(5).713-720, 1988

3. Evans DL, Pedersen CA, Folds JD: Major depression and immunity: Preliminary evidence of decreasednatural killer cell populations. Prog Neuropsychopharmacol Biol Psychiatry 12(5):739-748,1988

4. Leonard BE: The immune system and depression. Hum Psychopharmacol Clin Exp 3(3):157-158,19885. Schleifer SJ, Keller SE, Bond R, Cohen J. Stein M: Major depressive disorder and immunity: Role of age,

sex. severity and hospitalization. Arch Gen Psychiatry 46(l):81-87, 19896. Levy SM, Herberman RB, Simons A, Whiteside T, Lee J, McDonald R. Beadle M: Persistently low natural

killer cell activity in normal adults: Immunological, hormonal and mood correlates. Nat Immun CellGrowth Regul 8:173-186,1989

7. Pennebaker JW, Beall SK: Confronting a traumatic event: Toward an understanding of inhibition anddisease. J Abnorm Psychol 95(3):274-281, 1986

8. Pennebaker JW. Kiecolt-Glaser J, Glaser R: Disclosure of traumas and immune function: Healthimplications for psychotherapy. J Consult Clin Psychol 56:239-245, 1988

9. Asendorpf JB, Scherer KR- The discrepant repressor: Differentiation between low anxiety, high anxiety,and repression of anxiety by autonomic-facial-verbal patterns of behavior. J Pers Soc Psychol 45:1334-1346, 1983

10. Temoshok L: Biopsychosocial studies on cutaneous malignant melanoma: Psychosocial factors associatedwith prognostic indicators, progression, psychophysiology and tumor-host response. Soc Sci Med 20:833-840, 1985

11 Dattore PJ, Shontz FC. Coyne L: Premorbid personality differentiation of cancer and noncancer groups:A list of the hypotheses of cancer proneness. ) Consult Clin Psychol 48:388-394, 1980

12. Jensen MR: Psychobiological factors in the prognosis and treatment of neoplastic disorders Unpublisheddoctoral dissertation, Yale University, 1984

13. Jensen MR: Psychobiological factors predicting the course of breast cancer. J Pers 55:317-342, 198714 Temoshok L: Psychoimmunology and AIDS. Clin Immunol Newslet 9:113-116, 198715 Levy SM, Herberman R, Lippman M, d'Angelo T: Correlation of stress factors with sustained depression


DISCLOSURE AND EBV

of natural killer cell activity and predicted prognosis in patients with breast cancer. J Clin Oncol5(3):348-353, 1987

16. Levy SM, Herberman RB, Maluish AM, Schlein B, Lippman M: Prognostic risk assessment in primarybreast cancer by behavioral and immunological parameters. Health Psychol 4(2):99-113,1985

17. Glaser R, Gotlieb-Stematsky T: Human herpesvirus Infections: Clinical Aspects. New York, MarcelDekker, 1982

18. Kiecolt-Glaser JK, Fisher L, Ogrocki P, Stout JC, Speicher CE, Glaser R: Marital quality, maritaldisruption, and immune function. Psychosom Med 49:13-34,1987

19 Kiecolt-Glaser JK, Kennedy S, Malkoff S, Fisher L, Speicher CE, Glaser R: Marital discord and immunityin males. Psychosom Med 50:213-229,1988

20 DeLisi LE, Nurnberger JI, Goldin LR, Simmons-Ailing S, Gershon ES: Epstein-Barr virus and depression.Arch Gen Psychiatry 43(8):815-817, 1986

21. Miller AH, Silberstein C, Asnis GM, Munk G, Rubinson E, Spigland I, Norin A: Epstein-Barr virusinfection and depression. J Clin Psychiatry 47 529-530,1986

22. Millon T. Green CJ, Meagher RB: Millon Behavioral Health Inventory. Minneapolis, MN, InterpretiveScoring Systems, 1982

23. Millon T: Modern Psychopathology: A biosocial approach to maladaptive learning and functioning.Philadelphia, Saunders, 1969

24. Henle W, Henle G: Epstein-Barr virus and infectious mononucleosis. In Glaser R, Gottleib-Stematsky T(ods), Human Herpesvirus Infections: Clinical Aspects. New York, Marcel Dekker, 1982, 151-165

25. Kiecolt-Glaser JK, Glaser R: Methodological issues in behavioral immunology research with humans.Brain Behav Immun 2:67-78, 1988

26. Sumaya CV: Epstein-Barr virus serologic testing: Diagnostic indications and interpretations. PediatrInfect Dis 5:337-342, 1986

27. Roberts GH. The many faces of Epstein-Barr virus. Diagn Clin Testing 27'16-21, 198928. Glaser R, Rice J, Sheridan J, Fertal R, Stout J, Speicher C, Pinshy D, Kotur M, Post A, Beck M, et al.:

Stress-related immune suppression: Health implications. Brain Behav Immun 1:7-20, 198729. Gregory MD, Smeltzer MA: Psychiatry: Essentials of clinical practice. Toronto, Little, Brown and

Company, 198330. Gatchel RJ, Baum A, Krantz DS: Introduction to health psychology. New York, Random House, 198931. LaPorte RE, Montoye HJ, Casperscn CJ: Assessment of physical activity in epidemiologic research:

Problems and prospects. Public Health Rep 100:131-146, 198532. Tuckwiller LS. Glaser R: Epstein-Barr virus and nasopharyngeal carcinoma. In Reznick-Schuller (ed),

Comparative Respiratory Tract Carcinogenesis. Cleveland, Ohio, CRC Press, 1983, 171-18533. Klein E, Ernberg I, Masucci MG, Szigeti R, Wu YT, Masucci G, Svedmyr R: T-cell response to B-cells

and Epslein-Barr virus antigens in infectious mononucleosis. Cancer Res 41:4210-4215,198134. Rickinson AB, Moss DJ, Wallace LE, Rowe M, Misko IS, Epstein MA, Pope JH: Long-term T-cell-mediated

immunity to Epstein-Barr virus. Cancer Res 41:4216-4221,198135. Fowles DC: The three arousal model: Implications of Gray's two-factor theory for heart rate, electroder-

mal activity and psychopathy. Psychophysiology 17:87-104, 198036. Gray JA: Elements of a Two-Process Theory of Learning. New York, Academic Press, 197537. Vickers RR: Effectiveness of defenses: A significant predictor of cortisol excretion under stress. )

Psychosom Res 32(l):21-29, 198838. Ader R: Psychoneuroimmunology. New York, Academic Press, 198139. Cupps T, Fauci A: Corticosteroid-mediated immunoregulation in man. Immunol Rev 65:133-155. 198240. Curtis G, Fogel M, McEvoy D, Zarate C: Urine and plasma corticosteroids. psychological tests, and

effectiveness of psychological defenses. J Psychiatr Res 7:237-247, 197041. Katz JL, Weiner H, Gallagher TF, Hellman L: Stress, distress, and ego defenses. Psychoendocrine

response to impending breast tumor biopsy. Arch Gen Psychiatry 23:131-142, 197042. Knight RB, Atkins A, Eagle CJ, Evans N, Finkelstein FW, Fukushima D, Katz J. Weiner H: Psychological

stress, ego defenses and cortisol production in children hospitalized for elective surgery. PsychosomMed 41:323-328, 1979

43. Rose RM, Poe RO, Mason JW: Psychological state and body size as determinants of 17-OHCS excretion.Arch Intern Med 121:406-413, 1968

Psychosomatic Medicine 52:397-410(1990) 409

ESTERLING et al.

44. Wolff CT, Friedman SB, Hofer MA, Mason JW: Relationship between psychological defenses and meanurinary 17-hydroxycorticosteroid excretion rates: I. A predictive study of parents of fatally ill children.Psychosom Med 26:576-591, 1964

45. Glaser R, Kiecolt-Glaser JK, Speicher CE, Holliday JE: Stress, loneliness, and changes in herpesviruslatency. J Behav Med 8(3):249-260, 1985

46. Kiecolt-Glaser JK, Glaser R: Psychosocial influences on herpesvirus latency. In E. Kurstak ZJL (ed),Viruses, Immunity, and Mental Disorders. New York, Plenum, 1987, 403-412

ERRATUM

The first two words of the article by Myra Hunter, Ph.D., appearingon page 357 in the May/June issue of Psychosomatic Medicine, Vol. 52,No. 3, were inadvertently left off. The title should read, "Psychologicaland Somatic Experience of the Menopause: A Prospective Study." Weregret the error.


Emotional Repression, Stress Disclosure Responses, and Epstein Barr Viral Capsid Antigen Titers

Documents

Transcript of Emotional Repression, Stress Disclosure Responses, and Epstein Barr Viral Capsid Antigen Titers