Individual trajectories in stress covary with immunity during recovery from cancer diagnosis and treatments

Abstract

Research connects stressful events with altered immune regulation, but the role of subjective stress is uncertain. Using a longitudinal design, we provide a statistically powerful test of the relationship between subjective stress (perceived stress, emotional distress) and immunity (T cell blastogenesis, natural killer cell cytotoxicity, [NKCC]) as individuals adjust to a severe stressor, a cancer diagnosis and its treatments. Women with regional breast cancer (N = 113) were assessed at diagnosis/surgery and reassessed 4, 8, 12, and 18 months later. Latent growth curve analysis tested two hypotheses: (1) initial levels of subjective stress will correlate inversely with initial levels of immunity, and (2) rate of change in subjective stress will correlate inversely with rate of change in immunity. As predicted by Hypothesis 1, participants with high initial subjective stress showed poor initial blastogenesis. As predicted by Hypothesis 2, participants exhibiting an early, rapid decline in subjective stress also showed rapid improvement in NKCC. Follow-up analyses revealed perceived stress to be strongly related to immune function, while emotional distress was not. This is the first study to investigate trajectories in stress and immunity during recovery from a major stressor. Results imply that NK and T cells are sensitive to different aspects of the stress response. While T cell blastogenesis correlated with initial (peak) subjective stress, NKCC correlated with change (improvement) in subjective stress. These data highlight the importance of subjective stress, particularly stress appraisals, in the immune response to a major stressor.

Introduction

More than two decades of research have established an association between psychological stressors and changes in immune function (Herbert and Cohen, 1993, Segerstrom and Miller, 2004). This research body has overwhelmingly defined stress as the presence of a stressful event or circumstance, but a shift is now occurring. Rather than treating stressors as monotonic, there is a greater emphasis on studying individual variability in the stress response (see, for example, the 2003 special issue in Brain, Behavior and Immunity; e.g., Segerstrom, 2003). Investigating individual variability has far-reaching implications, from elucidating the mechanisms by which psychological factors could affect immune function to understanding why stress reducing interventions may (or may not) affect immunity. The present study examines individual variability in subjective stress and immunity during adjustment to a severe stressor: a cancer diagnosis and treatment.

Stressors are thought to produce immune change via cognitive appraisals of stress and increased emotional distress. Specifically, when an event is appraised as stressful, this appraisal elicits emotional distress, and these cognitive and emotional changes initiate a series of central nervous system and endocrine events which ultimately impact the immune system (see Kiecolt-Glaser et al., 2002). One implication of this theory is that the strength of the psychological response (perceived stress, distress) should correlate with the physiological response (e.g., immune function), yet the data supporting this relationship are weak. To investigate this issue in a naturalistic setting, a sample of individuals responding to a similar stressor is needed, and the reports must include correlations between subjective stress and immune function. In a recent meta-analysis, nine studies provided relevant data (Segerstrom and Miller, 2004), but the overlap in immune measures permitted the meta-analysis of only three immune parameters: numbers of circulating T-helper cells and T-cytotoxic cells and natural killer (NK) cell cytotoxicity. Only natural killer cell cytotoxicity (NKCC) showed a consistent correlation with subjective stress (r = −.15, p = .02). Authors of the meta-analysis cite a small number of studies and methodological issues as a potential explanation for the null results.

Meta-analytic results lead to the puzzling conclusion that stressful events are consistently related to immune function (Herbert and Cohen, 1993, Segerstrom and Miller, 2004), while the hypothetical causal factor, perceived stress, is not (see Thornton and Andersen, 2006) for a discussion. Two features of research in subjective stress might explain this discrepancy. First, studies designed to examine stressful events enjoy greater statistical power than studies of subjective stress by virtue of their study designs. Studies of stressful events maximize the potential for finding effects by specifically comparing participants in a “stress” condition (when subjective stress is presumably high) to those in a “no stress” condition (when subjective stress is presumably low). In contrast, studies examining subjective stress typically sample participants from the general population (e.g., a survey of working adults, similar to the “non-stress” condition; Theorell et al., 1990) or participants who are experiencing a common stressor (e.g., hurricane victims, similar to the “stress” condition; Ironson et al., 1997). Within these naturally occurring groups, there is less inter-individual variability in subjective stress than is seen in designs contrasting “stress” and “non-stress” conditions. Consequently, statistical power is reduced. Second, the relationship between subjective stress and immunity may be complex. Self-reports of subjective stress could be influenced not only by changes in response to the stressor, but also by persistent individual differences in the ways people interpret and report stress (see Segerstrom, 2003). Individual differences could obscure relationships between subjective stress and immunity in cross-sectional analysis.

A longitudinal study offers the possibility of controlling for design limitations. Because each subject serves as his or her own control, differences between subjects can be differentiated from change within subjects over time. Further, when longitudinal studies follow participants through high and low stress periods, the range of subjective stress observed is increased, enhancing statistical power. Only three studies have offered such an analysis. The first examined perceived stress. Maes and colleagues measured students’ perceived stress and immune function at two time-points, mid-semester and during an examination period (Maes et al., 1999). The difference in perceived stress across the two time-points correlated with the difference in immune cell counts and in the T-helper/suppressor ratio. Students whose perceived stress increased the most also showed the greatest change in immunity. The second and third studies examined mood over time. Stone and colleagues asked students to rate their mood on 25 occasions over $8{\displaystyle \frac{1}{2}}$ weeks. They discovered that students had lower antibody titers to a harmless protein on days when their negative mood was high relative to days with low negative mood (Stone et al., 1987). Shimamiya and colleagues asked subjects to provide daily mood ratings during a 10-day confinement period. Participants who showed the greatest mood change showed greater immune change across all measures (percentages of granulocytes, natural killer cells, and CD69 positive cells; Shimamiya et al., 2005). Data from these three studies suggest that within-subject change in subjective stress is an important determinant of immune function. Longitudinal studies offer the possibility of studying both within and between subject differences, yielding greater power and flexibility.

It is extensively reported that individuals experience stress upon learning of a cancer diagnosis (e.g., Compas and Luecken, 2002). It is also known that patients vary in their initial psychological response and recovery. While most patients report declines from their initial high distress (Edgar et al., 1992), some continue to report distress years later (Ganz et al., 1996). One four-year longitudinal study of quality of life after breast cancer diagnosis found that, while most patients’ mental functioning either remained steady (43%) or improved (45%), a substantial minority reported declines in quality of life related to mental functioning (Helgeson et al., 2004). Thus, data suggest considerable individual variation in psychological adjustment following a cancer diagnosis.

The present study examines subjective stress and immune trajectories in the first 18 months following a cancer diagnosis. Recruited an average of 37 days after surgical treatment, a primary stressor for these individuals is the diagnosis and early treatment of breast cancer. We have previously reported that those with the highest stress at diagnosis/surgery have the poorest immune function (Andersen et al., 1998) and show reductions in emotional distress from a psychological intervention (Andersen et al., 2004). Here we extend those findings with longitudinal observations. We examine the interim steps between the stressor and immune change: stress appraisals and emotional distress. Although we have differentiated these two constructs in other contexts (Andersen et al., 2004), in this naturalistic study, we expect that stress and distress (a) both arise from a stressful experience, (b) arise in the same individuals (i.e., they correlate), and (c) should hypothetically show similar relationships with immune function. Thus, we examine stress and distress in parallel, as indicators of the subjective response to stress.

Design features of this study offer advantages over prior investigations of subjective stress and immunity. The longitudinal design allows the examination of within-subject change in stress and its relationship with change in immune function. In addition, participants are recruited into the study during a period of very high stress (after diagnosis and surgical treatment) and they are followed as stress declines over an 18-month follow-up period.

We use an analysis suited to the examination of within and between-subject differences over time. Latent growth curve analysis (Meredith and Tisak, 1990) treats change over time as an underlying latent process that is reflected in repeated measures data. In the present study, data are summarized in terms of three latent variables: intercept, linear slope, and quadratic change. Fig. 1 illustrates these concepts and our hypotheses. The upper panel of Fig. 1 illustrates the subjective stress trajectories of four hypothetical individuals. First, consider the starting points (intercepts) for each of the four individuals. Persons A and B enter the study with high subjective stress, Person C enters the study with moderate stress, and D reports low subjective stress. The lower panel illustrates the immunological trajectories of the same individuals over time. We hypothesize that greater subjective stress at study entry will correspond to poorer concurrent immune function (a between-subjects effect). Thus, the figure denotes Persons A and B with the poorest initial immunity, C is slightly higher, and D is depicted with the highest initial immunity. In latent growth curve analysis, Intercept latent variables utilize data from all assessment points to estimate initial levels of stress and immunity. This allows us to test the concurrent relationship between stress and immunity with greater statistical power than cross sectional studies. We expect that the Intercepts of stress will correlate inversely with the Intercepts of immunity (Hypothesis 1). Second, consider the different patterns of change over time. Person A shows no improvement over time, B and C show linear decreases in stress over time, and D shows curvilinear change over time (i.e., denoting the involvement of a quadratic change variable). We hypothesize that rate of change in subjective stress will correspond to rate of change in immunity. Thus, the figure illustrates patterns of change in immunity that mirror those of subjective stress. In the latent curve models, we expect Slope of subjective stress to correlate inversely with Slope of immunity (Hypothesis 2).