There is a dearth of empirical tests, as the majority of randomized clinical trials (RCT) focus on emotional distress or quality of life outcomes rather than health (Andersen, 1992, 2002). Also, the findings are mixed regardless of the definition of ‘health’ used. It is infrequent to rate or measure health directly, but instead patients might provide ratings of the disruptive effects of symptoms on their daily lives. Measures such as the physical health component score of the SF-36 (Ware, Snow & Kosinski, 2000) or similar scales [European Organization for Research and Treatment of Cancer QLQ-30 (EORTC; Aaronson et al., 1993); Functional Assessment of Cancer Therapy Scale (FACT; Cella et al., 1993)] have been utilized. With such measures, a number of interventions have resulted in fewer disruptions in physical health-related quality of life (QoL; Edgar, Rosberger, & Collet, 2001; Helgeson et al., 1999) whereas others have not (Allen et al., 2002; de Wit et al., 1997; Edmonds et al. 1999; Giesler et al., 2005; Stanton et al., 2002). A more direct strategy has been patient ratings of a few specific symptoms, usually symptoms related to treatments received (e.g., a visual analogue scale used for rating nausea/vomiting with chemotherapy). In these studies, results have been generally favorable, with intervention patients self reporting significantly lower chemotherapy related symptoms (Roscoe, et al., 2000), fatigue (Yates et al., 2005; Quesnel et al., 2003), pain (de Wit et al., 1997; Doorenbos et al., 2006; Maguire et al., 1983; Spiegel & Bloom, 1983; Yates et al., 1981), and lymphedema (Maguire et al., 1983), as well as improvements across multiple symptoms (Forester et al., 1985; McCorkle et al., 1989). With the symptoms-focused studies, there have been fewer null findings [no differential improvement of urinary symptoms for prostate cancer patients (Giesler et al., 2005; Johnson et al., 1988); no improvement in lymphedema among breast cancer patients (Dodd, 1988; Stanton et al., 2002)].

Rather than self-evaluations, we were interested in testing the effects of an intervention on objective measures of health status typically used for oncology clinical trials. One global indicator used is the Karnofsky Performance Status Rating (Karnofsky & Burchenal, 1949), a broad evaluation of the patient’s mobility, symptomatology, and the resulting effects on current activity level and self care. As cancer treatment trials monitor patients closely to check for disease progression (rather than control), the tolerance for the next round of therapy, and the possible emergence of treatment toxicities or adverse events, there are multiple specific indicators used, such as findings from physical exams (e.g., blood pressure, neuromotor weakness), laboratory tests (e.g., liver functioning), or radiographs (e.g., chest films). Collaborative groups [(e.g., Gynecologic Oncology Group (GOG), the NSABP (National Surgical Adjuvant Breast and Bowel Project)] have combined these indicators into similar, standardized listings of symptoms, signs, illnesses, lab values, etc. used for their trials. In the listings, each item has a graded severity scale which is rated by nurses or physicians at frequent, protocol dependent intervals. In combination these two measures--functional status ratings and listings of symptoms, signs, etc.--document the health status of cancer treatment trial patients (e.g., Shapiro and Recht, 2001; Mouridsen et al., 2004; Pelegri et al., 2005) and are tested as correlates of disease endpoints (Mouridsen et al., 2003).

Our ongoing clinical trial tests whether a psychological intervention can alter the incidence of and/or time to recurrence for women with breast cancer. Although data are not yet available for an effect on disease endpoints, results are available for intermediate outcomes. Thus far, the intervention has shown robust biobehavioral effects (Andersen et al., 2004). Patients randomized to weekly, group intervention sessions for 4-months showed significant reductions in distress as well as significant health behavior changes, with patients high in initial cancer-related stress experiencing the greatest reduction in distress. Immunity was conceptualized as a potential mechanism linking lowered distress to health outcomes (Andersen et al., 1994). Analyses showed that T-cell proliferation in response to Phytohemagglutinin (PHA) and Concanavalin A (ConA) remained stable or increased for the Intervention group but declined for the Assessment only group (ps < .01). Because the intervention produced significant biobehavioral effects, we reasoned that improved health should, at some point, be evident at a later time point.

To increase the durability of the intervention effects, the 4-month intensive phase was followed by a maintenance phase of monthly sessions for an additional 8 months. At 12-months we reevaluated patients’ emotional distress and immunity, but relevant to the question and discussion above, we also assessed health status (Shapiro and Recht, 2001; Maltoni et al., 2005). We predicted that the intervention would have a direct, positive effect on health, operationalized with the indicators used in oncology trials.

The higher functional status and lower level of symptomatology, toxicities, etc. for the intervention group is clinically important. Characteristics of the present trial—disease endpoints as primary outcomes and tests of mechanisms for an effect—made it important to have longitudinal documentation of the patients’ general health. Search measures in several literatures were unproductive or unsatisfactory. Thus, we turned to the cancer treatment clinical trial studies for solutions. Because of the predominance of ongoing SWOG clinical trials at our institution at the time, their listing was chosen, as many of our patients would be participating in trials for which this listing was designed. Indeed, after disease markers, these data become the secondary endpoints in Phase III trials. When treatments are compared and found to be equally efficacious, these data become, for many patients and physicians alike, the basis for choosing among treatments (see Potosky et al., 1999 for data for prostate cancer; see discussion in Goodwin, et al., 2003, for breast cancer). In addition, these health data provide guidance for dosage modification in Phase II trials (Pelegri et al., 2005).

The health measures in this study also have importance as they covary with behavioral and psychological variables. The magnitude of treatment toxicities has been used to determine risk for poor adherence to treatment (Ayres et al., 1994; Levin et al., 1999; Demissie et al., 2001). Knowing the extent of a slowed recovery and/or disease and treatment morbidities is important for caring for survivors. Poor health status has a negative effect on mood (Ganz et al., 2004) and quality of life (Ganz et al., 2004; Levin et al., 1999), and health worries are the source of greatest concern among survivors (Spencer et al., 1999), as poor health negatively impacts personal relationships (Wimberly et al., 2005), daily activity levels, and return to normal routines such as employment (Bradley et al., 2005). In addition, poorer health impacts the meaning patients derive from life following cancer (Jim & Andersen, in press).

While some psychological trials have used the KPS (Dodd, 1988; Northouse et al., 2005; Scheier et al., 2005) the use of an extensive symptomatology and toxicity listing is novel. These measures are costly, as a high level of rater expertise is required (e.g., a nurse specialist), and they are time consuming with the patient interview, evaluating lab results and chart data, etc. Yet, they provide significant breadth, relevance, and with symptom-specific graded scales, objectivity. These measures are different in content and method from self-reported perceptions of health or symptoms. Correlations between patients’ reports of emotional distress and evaluations of their physical health may be inflated from shared method variance (i.e., both self-report measures). In addition to the objectivity of the items used here, nurses blinded to the patients’ study arm assignment were the raters. These factors greatly reduce measurement error related to the patient (e.g., reporting bias).

A second aim of this report was to identify how this psychological intervention could achieve positive health effects. While there are many possible routes, two likely mechanisms were tested. The path model accounted for the contribution of distress reduction, per se, on health status. In addition, the direct effect of the intervention suggested that additional mechanisms also played a positive role. The intervention included multiple components, and we can only speculate about which factors may be influential. Progressive muscle relaxation (PMR) provided patients with an active, and portable, means to reduce daily stress and cope with and control symptoms (e.g., nausea, disturbed sleep, fatigue). In other studies with cancer patients, PMR has yielded reductions in anxiety (Lyles et al., 1982) and elevations in immune measures (Lekander et al., 1997). Other intervention components also may have had general health benefits also; possibilities include recommendations for regular exercise [which has had symptom reducing effects in other studies with breast cancer patients (e.g., MacVicar et al., 1989)] and teaching patients to use direct, assertive communications with their physicians and nurses to get their medical needs met. Future studies will explore these and other possibilities.

There is great research interest in the potential for interrelationships among stress, immunity, and health (Kiecolt-Glaser et al., 2002). It is suggested that a robust immune response might maintain health by lowering one’s risk for infection or illness (Janeway, 2004), for example. While there is correlational evidence for the covariation of heightened stress, lowered immunity, and selected health outcomes (Kiecolt-Glaser et al., 2002), there are few experimental data in which stress is ‘manipulated’ (via randomization to an intervention) and the consequent effect on immunity and health are tested. To our knowledge, only Fawzy et al. (1993) provided relevant data, though not a test of relationships among the three variables. They reported that lower levels of baseline distress predicted recurrence and death, and higher natural killer cell activity predicted a lower risk for recurrence, though not survival.

In the present study, the intervention-produced changes in blastogenesis levels were statistically significant, but did not relate to health measures. [We also note that the effect size for the PHA observed at 12 months was ‘medium’ to ‘large’ by accepted standards (Cohen, 1988) and compares favorably with other estimates. In a meta analysis of psychological interventions and immunity, effect sizes for blastogenesis measures were small (rs of −.05 for ConA and .13 for PHA (Miller and Cohen, 2001)]. However, our health outcome assessed many bodily systems, a few of which likely had more direct relevance to immunity (‘infection’) than several of the others. Studying a health outcome particularly relevant for T-cell function and selection of assays with a direct link to a specific health outcome (Lutgendorf et al., 2002), may provide additional information regarding this hypothesized mechanism.

01