Although it is widely recognized that exposure to “stress” alters lymphocyte distribution (Herbert and Cohen, 1993; Kiecolt-Glaser et al., 2002) surprisingly few studies have reported immune data from individuals with posttraumatic stress disorder (PTSD). Four controlled studies could be found: [Male combat veterans:Boscarino and Chang, 1999; Laudenslager et al., 1998; Adult female child abuse victims: Wilson et al., 1999; Mixed gender natural disaster victims: Ironson et al., 1997]. Only CD4+ and CD8+ cells were reported in all studies. Two studies (Laudenslager et al., 1998; Wilson et al., 1999) found no PTSD-related group differences, one study found PTSD-related reductions (Ironson et al., 1997), while another reported elevations (Boscarino and Chang, 1999) in CD4+ and CD8+ counts under resting conditions.

The lack of consistency in results across these studies may be due to differences in gender, stressor type, and timing of the assessment in relation to trauma onset. In addition, examination of lymphocyte redistribution following laboratory challenge suggests that sometimes group differences are evident only when subjects are acutely stressed. For example, blunted natural killer (NK) cell cytotoxicity and B-endorphin response, increased epinephrine and NK cell number and slow cortisol recovery were found in response to an acute psychological stressor in men reporting a high degree of “life stressors” as compared to men reporting a low degree of “life stressors” (Pike et al., 1997). In a study of male adults reporting very high or very low daily “hassles”, blunted NK cell number among the high stress group was reported in response to a brief psychological stressor (Benschop et al., 1994). Likewise, college students who scored high on a measure of trait “worry” had a blunted NK response to fear provocation compared to low worriers (Segerstrom et al., 1999).

These laboratory challenge studies found no group effects under resting conditions, but immune differences emerged across low and high stress groups under challenge. Extending a challenge procedure to individuals with PTSD symptoms may help to clarify the inconsistent immune findings of previous resting condition studies. Furthermore, the use of a challenge procedure among those with a history of high life stress with and without current PTSD symptoms may help determine whether lymphocyte distribution in PTSD is distinct from the pattern for high life stress alone.

Mothers of healthy children (controls) were compared to mothers of child cancer survivors with and without posttraumatic stress symptoms (PTSS). All cancer mothers can be characterized as having a high stress history due to frequent acute and prolonged stress associated with their child’s cancer diagnosis and treatment (Stuber et al., 1998). However, only a subset of cancer mothers show clinically significant posttraumatic stress symptoms (PTSS) (Stuber et al., 1996). Based on previous immune studies, we first predicted baseline (resting) group differences in lymphocyte distribution across PTSS versus non-symptomatic women. Due to the equivocal nature of CD4+ and CD8+ resting levels in previous studies of PTSD, we did not predict a direction for these resting group differences. Second, based on the life stress challenge studies, we hypothesized that all cancer mothers would show blunted immune responsivity (i.e., smaller change in NK, CD4+ and CD8+ cells following challenge) relative to control mothers. Finally, because interpersonal stressors appear to have a greater impact on lymphocyte distribution than non-social stressors (Herbert and Cohen, 1993), lymphocyte distribution was tested following a personalized challenge (individualized trauma imagery) and after a generic challenge (mental arithmetic). This allowed us to determine if group differences were limited to interpersonal (high impact) stressors or would emerge even under generic stress.

In the generic mental arithmetic test (MAT), subjects were asked to count backwards by 7’s, beginning with 4500. A metronome was set at a fast pace, to increase psychological pressure for subjects to calculate quickly. In addition, a research assistant stood in front of the seated subject, giving feedback on accuracy and prompting subjects in a stern tone to “go faster please.” The individualized trauma imagery challenge followed procedures from Shalev et al. (1993). Personalized scripts were composed to help subjects invoke memories of their trauma experience. To equate the trauma imagery for control mothers whose children had no serious illness, controls were asked to select and describe a traumatic event. All control subjects reported a serious trauma (e.g., witnessing the death of a family member, automobile accident). Using the scripts, subjects were guided through an imagery session of their “worst moments.” During the final recovery period subjects were left alone and instructed to select an activity that would help them “return to normal.” They were provided with a video, reading material, music, and a game.

Potential health modifiers that correlated with an immune indicator (Pearson’s r > .30) were included as statistical covariates in all analyses of that indicator. The following variables were examined: age, and for the last 24 h, number of alcohol and caffeinated drinks (e.g., coffee) consumed, minutes of aerobic exercise and a rating of whether the amount of sleep the previous night was adequate (yes or no). F and p values for all variables in each analysis are provided.

CD8+ cells were generally reciprocal of CD4+ cells across number and percentage, thus we report only CD4+:CD8+ ratio statistical analyses. ANCOVA assumptions of normality and homogeneity of variance were met for NK and total lymphocytes, but CD4+:CD8+ ratios for each time period were transformed with a reciprocal transformation before analysis to correct for violations in homogeneity of variance.

As expected, groups were significantly different on the number of PTSD symptoms and symptom severity on the PDS. See Table 1. PTSS mothers had more symptoms [F(2,24) = 31.91, p < .001] and greater symptom severity [F(2,24) = 19.44, p < .001] than the other two groups, who did not differ. The number of years since the “trauma” (for cancer mothers, this was taken as the time since their child’s cancer diagnosis) did not differ across groups and ranged from 2 to 12 years. The number of prior traumas reported by each group was not significantly different and ranged from 1 to 9.

Table 1 Mean (and SD) number of prior traumas, PTSD symptoms and symptom severity as a function of group as assessed by PDS (Foa, 1995)

Subjects ranged from 29 to 55 years of age (M = 39.9 ± 7.4). Cancer mothers were significantly older than controls [F(2,25) = 9.22, p = .001]. The greatest age difference was between No PTSS (M = 47 ± 6.0) and Control (M = 34.44 ± 5.22), with PTSS mothers in between (M = 40.00 ± 6.2). Age was retained as a covariate for all analyses.

Subjects identified themselves as Caucasian (64%), Asian (8%), Latina (8%), African American (8%) or “other” (12%). Groups did not differ on the basis of race/ethnicity, economic status (low-middle-high income), or health behaviors. Subjects reported for the past 24 h a range of 0-60 min of aerobic exercise, consumption of 0-2 alcoholic drinks and 0-5 cups of coffee, 6-12 h of sleep, 0-1 cigarettes and no drug use.

Fig. 1. Mean (and SE) SUD rating across phases. Note: A procedural error resulted in no recorded SUD rating for one control subject.

3.3.1. Resting values Table 2 shows resting lymphocyte distributions (numerative and percentage of total) as a function of group. Repeated measures ANCOVA of total number of lymphocytes over phases indicated no group [F(1,22) = .32, p = .58] or age [F(1,22) < 1, p = .94] effects overall or at any phase. Group and phase patterns were similar for numerative and percentage CD4+, CD8+, NK and total lymphocytes. Thus, we report only percentage data here.

Table 2 Resting mean (and SD) lymphocyte distributions (numerative and percentage of total) as a function of group

3.3.2. CD4+:CD8+ ratio After controlling for age, sleep quality, and alcohol consumption, repeated measures analyses indicated trends for Order 4 effect of phase [F(1,19) = 3.71, p = .07] and group × phase interaction [F(2,19) = 2.69, p = .09]. Also, there was a significant group effect [F(2,19) = 5.88, p = .01]. See Fig. 2. Contrasts indicated the group effect was due to significant differences between PTSS versus controls (p = .04) and PTSS versus No PTSS (p = .006), but not between No PTSS and controls (p = .35). Parameter estimates for group differences at each phase showed PTSS CD4+:CD8+ ratios were marginally lower than controls at baseline, Recovery 1 and Recovery Final [p = .06, .08, .10, respectively] and significantly lower than controls at the MAT and Imagery Stressor phases [p = .03, .02, respectively]. There were no significant differences between No PTSS and controls, overall or at any time point. The statistical trends for phase and group X phase interaction are likely due to a response pattern most consistent with SUD anxiety changes across phases among controls (see Fig. 1), little CD4+CD8+ ratio change for the No PTSS group, and moderate reactivity across phases among PTSS mothers.

Fig. 2. Mean (SE) CD4+:CD8+ Ratio across phases. Covariates: age, alcohol, and sleep.

3.3.3. NK cells After controlling for age, alcohol and coffee consumption, sleep quality, and minutes of aerobic exercise, the repeated measures ANCOVA of NK cells revealed no significant effect of phase [F(1,18) = .64, p = .44], a trend for a group × phase interaction [F(2,19) = 3.24, p = .06] and no overall group effect [F(2,19) = 1.41, p = .27]. See Fig. 3. Contrasts confirmed no significant group differences overall, but parameter estimates showed after the MAT Stressor controls were significantly higher than PTSS (p = .04) but not the No PTSS group (p = .16). Groups were not different at baseline or at any other phase. Visual inspection of Fig. 3 suggests the statistical trend for a group × phase interaction reflects the reduced MAT Stressor reactivity among PTSS mothers relative to controls. Analysis of MAT Stressor change scores confirmed significantly greater reactivity among controls (M = 5.05, SE = 1.0) compared to PTSS mothers (M = 1.93, SE = 0.8) (p = .03) but not the No PTSS group (M = 2.31, SE = 1.2) (p = .14).

Fig. 3. Mean (SE) NK cells across phases. Covariates: age, alcohol, coffee, aerobics, and sleep.

3.3.4. Impact of covariates Age and sleep quality were not significantly associated with CD4+ or CD8+ levels, but alcohol consumption (0, 1 or 2 drinks) was associated with significantly higher CD4+ [F(1,19) = 5.02, p = .04] and lower CD8+ levels [F(1,19) = 5.91, p = .03]. Neither alcohol nor coffee consumption was significantly associated with NK levels [F(1,17) = 2.07, p = .17; F(1,17) = 2.89, p = .11, respectively]. In contrast, older age [F(1,17) = 5.68, p = .03], better sleep quality [F(1,17) = 4.47, p = .05], and less aerobic activity [F(1,17) = 8.49, p = .01] were significantly associated with higher NK levels in this sample.

3.4.1. PTSD symptoms To examine whether group differences in baseline and change score immune indicators were related to PTSS, correlations were conducted. As shown in Table 3, the greater the number and severity of symptoms, the higher the CD4+ and the lower the CD8+ levels at rest. Symptoms were unrelated to NK levels at rest, but were strongly associated with less NK change following the MAT stressor. Symptoms were unrelated to CD4+ and CD8+ MAT stressor changes and to all immune changes in response to the Imagery stressor.

Table 3 Bivariate correlations between PTSD symptoms and immune outcomes at rest and following stressor phases

3.4.2. SUD anxiety To determine the influence of baseline anxiety on immune outcomes, all ANCOVA analyses were repeated with resting SUD scores as a covariate. Immune group and phase effects were unchanged. Multivariate linear regression analyses with backward elimination were conducted to assess the relative contribution of SUD changes in response to the MAT stressor as a predictor of MAT immune changes. The best model for NK cells [F(2,24) = 7.67, p = .003, R² = .41] included aerobic activity (β = -.05, p = .09) and number of PTSS symptoms (β = -.33, p = .002). Neither MAT SUD change nor PTSS group status was predictive of NK change in any model. For CD4+, the best model [F (3,23) = 3.68, p = .03, R² = .36] included age (β = -.10, p = .13), PTSS status (β = 1.21, p = .06) and MAT SUD change (β = -.59, p = -.02). Number of PTSS symptoms was not predictive of CD4+ change in any model. For CD8+, the best model [F(4,23) = 4.99, p = .006, R² = .51] included alcohol (β = 1.61, p = .11), PTSS status (β = 2.10, p = .02) MAT SUD change (β = .91, p < .001) and number of PTSS symptoms (β = .28, p = .08). Thus, subjective distress and PTSS status appear to be independent contributors to CD4+ and CD8+ immune changes in response to a stressor whereas only PTSS status predicts NK changes.

3.4.3. Symptoms of depression BDI scores did not correlate with baseline immune levels or change scores for any immune indicator.

PTSS effects were evident in several ways. PTSS mothers had a significantly smaller change in NK cells following the MAT challenge than controls. Resting CD4+ and CD8+ levels and NK MAT change scores were all significantly correlated with number of PTSS symptoms, providing further evidence of tonic and phasic immune differences associated with PTSS. Multivariate regression results showing the separate predictive value of changes in subjective distress and PTSS status in estimating CD4+ and CD8+, but not NK MAT changes, indicates PTSS immune differences cannot be attributed to subjective distress alone.

For all groups, the direction of lymphocyte distribution in response to stressor challenge (percentage increases in NK and CD8+, decreases in CD4+) was consistent with previous studies (Cacioppo et al., 1998; Mills et al., 1995; Pike et al., 1997; Strauman et al., 1993; Valdimarsdottir et al., 2002). Likewise, PTSS-related immune effects found here provide replication evidence for some previous studies. Elevated resting CD4+ levels in PTSS mothers replicate findings of male combat veterans with chronic partial PTSD (Boscarino and Chang, 1999). Low resting levels of CD8+ are in keeping with some PTSS findings (Ironson et al., 1997), but not others (Boscarino and Chang, 1999; Laudenslager et al., 1998; Wilson et al., 1999). The absence of PTSS baseline differences in NK cells is in contrast to data from a natural disaster sample (Ironson et al., 1997).

Effects for cancer mothers as a chronic stress group are equivocal. Only PTSS mothers showed significantly reduced NK reactivity consistent with previous laboratory challenge studies among individuals high in life stress, daily hassles or tendency to worry (Benschop et al., 1994; Pike et al., 1997; Segerstrom et al., 1999). Although No PTSS mothers appeared to have blunted CD4+:CD8+ ratio reactivity (all phases) and smaller NK MAT Stressor change scores relative to controls, these differences were not statistically significant. Small sample size may be one explanation for the failure to detect significant differences. Alternatively, the absence of No PTSS differences may reflect exceptional resilience among mothers who do not develop long-term PTSS despite exposure to acute and chronic stress associated with their child’s cancer. Past laboratory challenge studies did not assess PTSS, thus it is also possible these symptoms mediated reactivity effects of previous research. Further research will be necessary to clarify whether PTSS confers immune changes separate from those associated with a history of chronic stress.

Group effects found here were not limited to individualized stressors, but emerged under generic challenge. The absence of strong reactivity to the Imagery Stressor in any group was unexpected based on previous data indicating interpersonal stressors have greater impact on lymphocyte distribution than non-social stressors (Herbert and Cohen, 1993). The significant rise in SUD anxiety during the Imagery Stressor as well as the MAT Stressor indicates subjects were as distressed by the Imagery as by the MAT Stressor. Thus, blunted lymphocyte distribution does not appear to be due to failure of the Imagery Stressor to induce distress. Although groups returned to baseline levels after the MAT Stressor, the inability to mount a significant change in lymphocyte distribution after the Imagery Stressor suggests reactivity may be temporarily blocked or reduced by an earlier stressor. Alternatively, circadian rhythm of immune measures may have influenced immune response strengths later in the laboratory session as compared to earlier. However, if that were the case, one would expect a rising pattern in CD4+, CD8+ and NK cell counts during the period coinciding with the lab session (approximately 6-8 p.m.), not stasis or a decline (Kronfol et al., 1997). Replication with generic and individualized stressors counterbalanced for order and separated by days would clarify the relative strength of group effects under each stressor condition.

Additional limitations of this study should be addressed in future research. First, this work must be replicated with a larger sample size to increase confidence in patterns found here. Also, whereas health behaviors such as sleep quality, exercise, use of alcohol, and coffee were statistically controlled here, a larger sample would allow for statistical modeling that could determine a mediating or moderating role for these factors, as has been done for health behaviors and depression (Miller et al., 1999). Second, blunted reductions in NK cell percentages following laboratory stress recently reported in females in the luteal phase of menses (Pehlivanoglu et al., 2001) suggests menses stage may influence outcomes. Finally, recent evidence for dampened NK cytotoxicity in the face of increased cell number (Pike et al., 1997; Strauman et al., 1993) emphasizes the importance of multi-method designs to clarify the mechanisms of immune change.

The present findings are preliminary and should be viewed as a guide for future research. Nonetheless, these data indicate posttraumatic stress symptoms are associated with changes in tonic lymphocyte distribution and immune responsivity to psychological challenge. In addition, these changes may be distinct from patterns associated with subjective distress more generally. Further research will be needed to assess the extent to which differential lymphocyte distribution effects found in the laboratory have direct implications for health in the natural environment.

The first author thanks Michael Irwin for his excellent feedback on an earlier draft, the mothers of child cancer survivors at UCLA for their generous participation; and finally, research assistants Irene Choi, Mark Power, Jennifer Bennett, Marleen Castaneda, and students too numerous to name for their diligent efforts in supporting this work.