Marital quality and survivorship

Slowed recovery for breast cancer patients in distressed relationships

Authors


Abstract

BACKGROUD:

Although marital distress has been implicated in difficulties with adjustment to a breast cancer diagnosis, its long-term effects, especially on physical recovery, are unknown.

METHODS:

Longitudinal data from newly diagnosed breast cancer patients (N = 100) who were married or cohabiting were used. Patients were assessed after diagnosis and surgery (baseline) and then reassessed every 4 or 6 months for the next 5 years. Women in stable, distressed relationships (n = 28) were compared with those in stable, nondistressed relationships (n = 72). Stress, health behavior, and health outcomes were examined using mixed-effects modeling.

RESULTS:

Overall, marital distress was associated with slowed recovery trajectories and poor outcomes. At baseline, both groups had equivalent, high levels of stress, but diverged thereafter. Stress declined more slowly for the Distressed group, and by 5 years it remained significantly higher. Differential reductions in physical activity were also observed. With regard to health, the Distressed group was found to have a slower recovery in performance status and more symptoms/signs of illness and treatment side effects through 3 years. Finally, all the effects were observed above and beyond reductions occurring with depressive symptomatology, which was significantly higher in the Distressed group.

CONCLUSIONS:

Marital distress is not only associated with worse psychologic outcomes for breast cancer survivors, but poorer health and a steeper decline in physical activity. These novel data demonstrate recovery trajectories for breast cancer survivors to be constrained for those also coping with ongoing difficulties in their marriage. Cancer 2009. © 2008 American Cancer Society.

The health-enhancing properties of personal relationships are well known.1 Population-based studies have reliably shown that individuals with partners have lower morbidity and mortality than the unpartnered. This is also the case in studies of individuals with serious illness, such as cancer2 or cardiovascular disease.3 Despite the advantage that the mere presence of a partner appears to confer, other data suggest that this robust finding may be constrained for those in troubled relationships.

For the majority, marriage is 1 of the most important interpersonal relationships, and thus, relationship distress is a major stressor. Psychologic adjustment for those in distressed marriages is generally poor. Mood disorders, typically depression, are a particular problem.4, 5 This may be because of the finding that, unfortunately, marital distress tends to be a chronic stressor.6, 7 Stability in marital satisfaction (or lack thereof) has also been found for those with cancer.8, 9 Moreover, among cancer patients whose marriages fail, it is not because relationships become discontented after cancer diagnosis. Divorce and breakups occur primarily among those reporting that marital difficulties predated their diagnosis.8, 10

In addition to general psychologic distress or depression in particular, marital distress may also contribute to poor health. One route for health effects may be via poor health habits.11 For example, marital satisfaction has been associated with increased compliance with medical regimens,12 whereas marital dissatisfaction has been associated with poorer compliance (eg, problematic weight gain after surgery13). Another route may be through the effects of stress on the body. Both naturalistic and laboratory studies with couples suggest that cardiovascular, endocrine, and immune responses may differ among the maritally distressed compared with the nondistressed.14, 15

However they arise, there is the potential for adverse health consequences after chronic relationship distress, and this impact may be important for those currently or previously ill. Thus, we examined the relationship between marital distress and psychologic, health behavior, and health outcomes among breast cancer patients as they became 5-year cancer survivors. Recovery trajectories of patients coming from (and remaining in) stable, distressed relationships were contrasted with trajectories of those in stable, nondistressed relationships. In the formulation of hypotheses, we considered both the previously described marital literature and also reviewed the few relationship distress studies with cancer patients. The latter primarily provided psychologic outcomes only. Three single-assessment studies showed relationship distress to be correlated with worse psychologic well being,16–18 poorer self-reported physical functioning,16 less positive healthcare orientation, and more illness-related, family difficulties.18 The remaining 3 studies assessed patients on 29, 19 or 3 occasions20 within ≤20 months, and again, marital distress predicted poor emotional adjustment.9, 19, 20

In this context, we used data from a convenience sample of breast cancer patients who participated in a clinical trial, were assessed at diagnosis/surgery, and then were reassessed through the ensuing 5 years. The objective of this substudy was to determine whether marital distress would covary not only with heightened stress but also with poor health behaviors and slowed recovery. On the basis of previous work,21, 22 we anticipated that at diagnosis, women from distressed relationships would respond in a fashion similar to those in nondistressed relationships. However, as these stable relationships remained distressed or nondistressed during the next 5 years, we anticipated that the trajectories of the 2 groups would diverge. On the basis of prior literature, we expected psychologic recovery to be slower for those in distressed relationships, but we also tested whether health behaviors and physical recovery trajectories would be similarly slowed.

MATERIALS AND METHODS

Description of the Clinical Trial

Patients were consecutive cases at a university-affiliated National Cancer Institute-designated Comprehensive Cancer Center and accrued to a clinical trial testing the efficacy of a psychologic intervention. Patients with newly diagnosed, surgically treated regional breast cancer were eligible. Details regarding informed consent procedures, accrual, and randomization have been published.23 The trial received approval from the local institutional review boards in accord with an assurance filed with and approved by the Department of Health and Human Services. Briefly, patients (N = 227) completed face-to-face interviews and questionnaires assessing cancer stress, depressive symptoms, health behaviors, and other areas at the initial assessment/baseline before beginning adjuvant therapy. A research nurse completed a health status evaluation. Study arms were Intervention and assessment or Assessment only. Study arms were equivalent with regard to sociodemographic, prognostic, and cancer treatment variables (all P values < .23).23 Patients were reassessed at 4 months, 8 months, and 12 months; by the 12-month follow-up, all cancer therapies were completed, and the intervention sessions had ended. For those randomized to the Intervention arm, there were significant reductions in emotional distress, improvements in health behaviors, less variability in chemotherapy dose intensity, and higher immune responses at 4 months23 and similar outcomes and health improvements at 12 months.24 Thereafter, all patients were followed and reassessed in person every 6 months during years 2 to 5.

Participants and Procedures

Data from a subset of patients (N = 100) were used. Criteria for patient selection were that the patients had to be married or cohabitating at the time of the initial assessment, and thereafter, remain a trial participant and remain disease-free up to and including the 60-month assessment. Of the 227 trial participants, 120 were eligible and 107 (47%) were excluded because of the absence of a partner (24%) or disease progression/death (23%). Of the 120 eligible partnered patients, data from 20 (17%) were incomplete because of trial withdrawal before 60 months. Comparison of the substudy sample (N = 100) to the eligible but withdrawn patients (n = 20) found no significant group differences (all P values > .05) with regard to sociodemographics (age, race, partner status, family income, and employment status), disease characteristics (stage, hormone receptor status, and number of positive lymph nodes), or treatments received (psychologic intervention, surgery, radiotherapy, and hormonal therapy), excepting chemotherapy. The substudy sample was more likely to have received chemotherapy (92%) than the eligible patients who withdrew (70%) (P = .005). Table 1 provides descriptive characteristics of the sample.

Table 1. Equivalence of Groups (Distressed Versus Nondistressed) With Regard to Sociodemographic, Prognostic, and Treatment Received Variables
VariableNondistressedDistressedTotal
(n=72), Mean (SD), %(n=28), Mean (SD), %(n=100), Mean (SD), %
  • SD indicates standard deviation; ER/PR, estrogen receptor/progesterone receptor status.

  • *

    Significant group difference: P < .05.

Sociodemographic   
 Age, y49.0 (9.5)47.6 (9.4)48.6 (9.4)
 Race, % white97.292.996.0
 Education, y15.5 (3.0)15.4 (2.4)15.5 (2.8)
 Marital status, % married91.792.992.0
 Relationship duration, y23.6 (12.5)19.4 (12.3)22.4 (12.6)
 Family income, $1000/y99.2 (104.3)73.0 (54.0)92.0 (93.6)
Prognostic   
 Stage, II vs III, % II88.996.491.0
 No. of positive lymph nodes3.0 (5.1)3.7 (4.3)3.2 (4.9)
 ER/PR, % positive*80.650.072.0
Treatment   
 Surgery, % modified radical mastectomy54.260.756.0
 Radiotherapy, % yes59.760.760.0
 Chemotherapy, % yes90.396.492.0
 Psychologic intervention, % yes52.857.154.0

The level of marital distress among the substudy patients was determined with Satisfaction Scale of the Dyadic Adjustment Scale (DAS) data. Patients completed the Satisfaction subscale at baseline and annually thereafter, for a total of 6 assessments. The DAS is the most widely used and psychometrically validated measure of relationship quality, and it is routinely used to discriminate between distressed and nondistressed relationships.25–28 A DAS score of 100 (possible range, 0-151) is used to distinguish distressed and nondistressed individuals.29 The Satisfaction subscale (DASS; 8 item-version; possible range, 0-41) is used as a short form,26, 27 with the corresponding cutoff being 28. A recent meta-analysis of studies using the DAS reported an average Cronbach α of .92 for the full scale and .85 for the Satisfaction subscale.30 Cronbach α for the substudy patients ranged from .87 to .91. One-year interval reliabilities for this sample ranged from .65 to .89, which is comparable to those in the DAS literature for intervals from weeks to months (eg, .69 to .96).9, 31, 32

As the reliability data suggest, the average DASS score for the sample varied little across the 6 assessments, with it being, for example, 31.3 (standard deviation [SD] of 6.1) at baseline and 31.5 (SD of 5.3) at 5 years. To obtain a reliable DASS score for each patient, the mean of a patient's 6 yearly DASS assessments was calculated. By using the cutoff of 28, 2 groups were defined: patients who reported, on average, relationship distress during the 5-year period (Distressed; n = 28; DASS grand mean = 23.3, SD of 3.5; range, 14.5-28.0) and those who did not (Nondistressed; n = 72; DASS grand mean = 33.6, SD of 2.7; range, 28.3-38.0). Plots of the raw DASS scores are provided (see Fig. 1) and show the stability of scores for each group. We found 28% (28 of 100) of the sample to score within the distressed range, similar to the percentage reported by Weihs et al, who also used the DAS to study stage II and III breast cancer survivors (32%; N = 44).9

Figure 1.

Marital distress (Dyadic Adjustment Scale [DAS]-Satisfaction) during the 5 years after the diagnosis of breast cancer is shown for the Distressed and Nondistressed groups. Error bars represent 95% confidence intervals.

Sociodemographic, disease, and treatment characteristics for the sample and the Distressed and Nondistressed groups are provided (see Table 1). The groups were contrasted using analysis of variance or chi-square analyses as appropriate. There were no significant differences noted between groups (all P values > .156) excepting that for hormone receptor status. A significantly (P = .002) larger proportion of the Distressed group was hormone receptor negative (50%) compared with the Nondistressed group (19%).

Measures

Stress

Cancer-specific stress

The Impact of Events Scale (IES) examines intrusive and avoidant thoughts and behaviors related to cancer diagnosis and treatments.33 Scores >19 reflect clinically relevant levels of stress.33 Internal consistency was .87.

Global stress

The Perceived Stress Scale 10-Item Version (PSS-10) measures an individual's appraisal of life as stressful (ie, unpredictable, uncontrollable, and overwhelming).34 Internal consistency was .86.

Health behaviors

Dietary patterns

The Food Habits Questionnaire assesses dietary choices and eating patterns such as avoiding fat, food substitution with lower-fat alternatives, modification of food preparation, replacing high‒fat with low‒fat foods, and fruit and vegetable intake.35 Higher scores indicate healthier eating habits. Internal consistency was .79.

Physical activity

A 7-day report of physical activity, based on the Seven-Day Exercise Recall of the Stanford Heart Disease Prevention Program,36 indexed energy expenditure.

Health

A research nurse completed 2 measures based on patient interview, medical chart review, and physician consultation when needed.

Performance status

The Karnofsky Performance Status Scale (KPS) is a functional status scale ranging from 100 (Normal, no complaints, no evidence of disease) to 0 (Dead) with 10-point intervals.37 Inter-rater reliability ranges from .70 to .97.38, 39

Symptoms, signs, illnesses, and toxicities

Items for evaluating symptoms, signs, and illnesses came from the toxicity measure used by the Southwest Oncology Collaborative Group (1994 version).24 Items are grouped within 22 body categories (eg, gastrointestinal), with 4 to 6 items for each (eg, nausea) rated on a 5-point scale (0 = none to 4 = life threatening). Items were averaged for a total score.

Control variables

Depressive symptoms

Previous research has shown significant associations between marital distress and depressive symptoms.16–18 Therefore, depressive symptoms were included as a control variable to determine the relationship between marital distress and outcomes above and beyond the potential contribution of depressive symptoms reported at the time of the initial assessment and change in depressive symptoms across time. The short form of the Center for Epidemiological Studies-Depression Scale (CES-D) was used.40 A cutoff of ≥10 is recommended for identification of clinically significant symptoms.40 Internal consistency was .81.

Other control variables

The following were also considered: age, education, cancer stage, hormone receptor status, number of positive lymph nodes, surgery type, chemotherapy, radiotherapy, and study arm.

Statistical Analysis

Mixed-effects modeling41 tested for group differences at baseline as well as changes over time. Mixed-effects models are advantageous for analyzing longitudinal data in that the procedure accounts for the correlations among repeated assessments within an individual and allows the number of repeated assessments to vary across individuals. The rate of missing data was 5% for all the outcome variables except for the physical activity measure (14%), which was implemented after the trial began. Both the fixed-effects (group average effects) and random-effects (within-individual variability) were estimated. The form of change, linear versus quadratic, was determined by comparison of relative fit of models using a likelihood ratio test. If the fit of the quadratic model was not significantly better (α = .05) than that of the linear model, the linear model was retained.

Specifically, the Group effect tests differences between the Distressed and Nondistressed at baseline. The Time effect tests whether the outcome changes during the follow-up in the Nondistressed group. The Group × Time effect tests whether the rate of change in the Distressed is significantly different from that of the Nondistressed. To test the effects of marital distress beyond any effects of depression, baseline depressive symptoms (continuous CES-D) and CES-D × Time effects were included in all models. In addition, the following were also considered as controls and included as appropriate: age, education, cancer stage, hormone receptor status, number of positive lymph nodes, surgery type, chemotherapy, radiotherapy, and study arm. All main effects and 2-way interactions with Time were entered into the model. A backward elimination process was used in which nonsignificant terms (P > .05) were eliminated from each model until a final solution was reached.42 The same final models with the intercept coded to be the 60-month assessment rather than the initial assessment were used to test significance of group differences at 5 years. All statistical tests were 2-sided.

RESULTS

Descriptive statistics for the variables at initial and 5-year follow-up are presented in Table 2. There was a significant group difference in baseline depressive symptoms, with the Distressed group having higher depressive symptoms compared with the Nondistressed group (P = .034), as expected. Correlations among the outcome variables ranged from .07 to .57, and the only moderate correlation (.57) was between the 2 stress measures (IES and PSS-10). Despite this relationship, IES and PSS-10 were analyzed separately to examine broader range of stress responses of the patients. Quadratic models provided significantly better fit than linear ones for all outcomes except global stress. Table 3 summarizes results from the mixed-effects models.

Table 2. Descriptive Statistics for the Measurement of Depressive Symptoms, Stress, Health Behavior, and Health Outcomes at Baseline and 5 Years for the Nondistressed and Distressed Groups
VariableNondistressed (n=72)Distressed (n=28)
Baseline Mean (SD)5 Years Mean (SD)Baseline Mean (SD)5 Years Mean (SD)
  • SD indicates standard deviation; CES-D, Center for Epidemiological Studies-Depression Scale; IES, Impact of Events Scale; PSS-10, Perceived Stress Scale 10-item version; KPS, Karnofsky performance status.

  • *

    Significant group difference: P < .05.

Depressive symptoms    
 CES-D4.79 (3.28)*3.45 (3.62)6.46 (3.00)*5.25 (3.99)
 ≥10 cutoff10%10%21%15%
Stress    
 IES24.96 (12.89)8.63 (10.27)27.43 (15.59)13.20 (13.95)
 >19 cutoff64%10%71%20%
 PSS-1016.89 (6.30)13.82 (5.54)19.32 (8.24)16.15 (7.20)
Health behavior    
 Dietary patterns2.37 (0.49)2.45 (0.45)2.26 (0.52)2.28 (0.47)
 Physical activity15.72 (22.84)14.06 (18.64)31.99 (51.39)10.91 (11.55)
Health    
 KPS85.56 (7.10)92.09 (7.69)86.43 (7.31)89.00 (9.68)
 Symptoms/signs0.18 (0.11)0.20 (0.09)0.21 (0.11)0.24 (0.10)
Table 3. Mixed-Effects Models Comparing Trajectories of Breast Cancer Patients in Nondistressed (n=72) or Distressed (n=28) Relationships Across 5 Years
OutcomeEffectsEstimateSEtP
  1. SE indicates standard error; IES, Impact of Events Scale; PSS-10, Perceived Stress Scale 10-item version; KPS, Karnofsky performance status.

Stress     
 IESGroup−1.0372.411−0.430.668
Time−0.5130.048−10.706<.001
Quadratic term0.0050.00087.009<.001
Group×Time0.1010.0333.067.002
 PSS-10Group0.4491.0810.415.679
Time−0.0300.008−3.608<.001
Group×Time0.0410.0162.504.012
Health behavior     
 Dietary patternsGroup−0.1420.083−1.705.091
Time0.0080.0025.176<.001
Quadratic term−0.00010.00003−5.212<.001
Group×Time−0.0010.001−0.845.399
 Physical activityGroup9.4654.7242.004.046
Time0.3910.2031.922.055
Quadratic term−0.0090.003−2.671.008
Group×Time−0.2460.121−2.032.042
Health     
 KPSGroup0.1800.9810.184.854
Time0.2680.0357.597<.001
Quadratic term−0.0030.0006−5.366<.001
Group×Time−0.0530.021−2.460.014
 Symptoms/signsGroup0.0500.0173.029.003
Time−0.00020.0005−0.481.631
Quadratic term−0.000020.000007−2.125.034
Group×Time−0.00050.0003−1.953.051

Stress

There was no significant difference noted between the Distressed and Nondistressed groups with regard to baseline cancer-specific stress (IES; P = .668) (Fig. 2a). As expected, both groups' baseline estimates indicated comparable, clinically significant33, 43 mean levels of cancer stress (mean = 19.7 and 20.7 for the Distressed and Nondistressed groups, respectively). With regard to change, the Nondistressed group showed a significant quadratic change in IES (P < .001), such that the average individual's stress rapidly decreased, then stabilized. As hypothesized, the rate of decline in the IES for the Distressed group was significantly slower than that for the Nondistressed group (P = .002). For outcomes at 60 months, the group difference approached statistical significance (P = .053), with a higher level of cancer-specific stress for the Distressed group.

Figure 2.

Stress trajectories are shown. (a) A significantly slower decline in cancer-specific stress was noted for the Distressed group compared with the Nondistressed group. IES indicates Impact of Events Scale. (b) A significant reduction in global stress was noted for the Nondistressed group, but no reduction was noted in the Distressed group. PSS-10 indicates Perceived Stress Scale 10-item version.

With regard to global stress (PSS-10), there was no baseline group difference (P = .679) (Fig. 2b). The baseline estimates for both groups (mean = 15.8 and 15.3 for the Distressed and Nondistressed groups, respectively) were 0.5 SD higher than that from normative data (mean = 13.0, SD of 6.4).44 With regard to the trajectory of the Nondistressed group, there was a significant decrease (P < .001) in global stress across time, eventually declining to that of the national average (13.5). Notably, the trajectory of the Distressed group was significantly different (P = .012), with the stress scores actually increasing. At 60 months, global stress was significantly higher (P = .011) for the Distressed group (16.4).

Health behaviors

There was no significant group difference in baseline dietary patterns (P = .091; Fig. 3a). With regard to the trajectory, the Nondistressed group demonstrated a significant quadratic change in healthy dietary habits (P < .001), such that the average patient showed improvements during the first 30 months, then a gradual decrease. The groups did not differ with regard to the pattern of change (P = .399). However, at 60 months the Nondistressed group was found to have significantly better dietary habits than the Distressed group (P = .032).

Figure 3.

Health behavior trajectories are shown. (a) Generally healthier dietary habits were noted for the Nondistressed group compared with the Distressed group. FHQ indicates Food Habits Questionnaire. (b) A significantly higher baseline physical activity followed by a steeper decline was noted for the Distressed group compared with a more stable activity level in the Nondistressed group.

For physical activity, the Nondistressed group demonstrated significantly lower levels than the Distressed group at baseline (P = .046; Fig. 3b). With regard to the trajectory, a significant quadratic change was observed for the Nondistressed group (P = .008), and the pattern of change differed by group (P = .042). Whereas the Nondistressed group showed an increase in physical activity during the first 2 years and then a gradual decrease, the Distressed group showed a stable level during the first 18 months and a rapid decrease thereafter. By 60 months, the groups did not significantly differ (P = .282) in their overall, lowered levels of physical activity.

Health

There was no significant group difference in baseline performance status (KPS; P = .854), with both groups in the 80 to 90 range (ie, “normal activity with effort; some signs/symptoms of disease” to “able to carry on normal activity; minor signs/symptoms”; Fig. 4a). With regard to their trajectories, a significant quadratic change was observed for the Nondistressed group (P < .001), and the pattern of change differed by group (P = .014). The Distressed group demonstrated a curve of improvement during the first 3 years and decrease thereafter. In contrast, the Nondistressed group showed more rapid improvement during the first 3 years and a plateau without any decline in performance status during the next 2 years. Thus by 60 months, the Nondistressed group was evaluated as having a significantly better (P = .005) performance status than the Distressed group.

Figure 4.

Health trajectories are shown. (a) A significantly slower recovery in Karnofsky performance status (KPS) was noted for the Distressed group compared with the Nondistressed group. (b) Significantly higher baseline symptoms/signs followed by a steeper decline was noted for the Distressed group, compared with lower, more stable symptoms/signs for the Nondistressed group.

With regard to symptoms/signs, the Nondistressed group had significantly fewer and/or less severe symptoms/signs at baseline than the Distressed group (P = .003) (Fig. 4b). The trajectory of the Nondistressed group showed a significant quadratic change in symptoms/signs (P = .034), such that the average individual showed a slight increase, peaking at 18 months, and then a decreasing trajectory thereafter. The group difference in the rate of change approached statistical significance (P = .051), and the groups did not differ at 60 months (P = .276). However, follow-up analyses showed that the significantly higher level of symptoms/signs of the Distress group found at baseline remained for 36 months (P = .040).

Post Hoc Analyses

As noted in the Statistical Analysis above, the primary analyses controlled for the effects of depressive symptoms. However, post hoc analyses were conducted to explore the relationship between marital distress and depression, and to further test for depressive symptoms as a moderator for outcomes. Thus, all analyses for the models were repeated but with Group × CES-D and Group × CES-D × Time interactions added. Depressive symptoms were not found to be a significant moderator of martial distress in the models with stress, physical activity, and performance status as an outcome (all P values < .267). However, depressive symptoms were found to be a significant moderator between marital distress and symptoms/signs (P values < .011). Whereas the patients with less depressive symptoms demonstrated similar trajectories as those presented in Figure 4b, the depressed women with no marital distress maintained high symptoms across the follow-up, and the depressed and martially distressed showed an increase in symptoms. A trend toward depressive symptoms as a moderator between marital distress and dietary patterns was also observed (P = .052). The depressed and maritally distressed women showed a constant decline in dietary habits without any improvements, whereas the other 3 groups generally demonstrated a gradual increase for the first 30 months and a gradual decrease thereafter, similar to the trajectories presented in Figure 3a.

These analyses suggest that the combination of depression and marital distress results in selected worse outcomes than those found with either characteristic alone. This is the case for symptoms/signs and dietary habits, which worsen with time. For the stress, physical activity, and performance status outcomes, there appears to be no further worsening of outcome than that found with marital distress.

DISCUSSION

Novel findings emerged from a study of breast cancer patients remaining disease-free for 5 years, with some couples simultaneously coping with long-term marital relationship strain. Overall, results suggest that chronic marital distress may impair recovery in the postdiagnosis years. Specifically, individuals in a distressed relationship had continuously heightened levels of global stress, deteriorating health behaviors, slower improvements in cancer-specific stress and performance status, and eventually more impaired functioning in comparison with those whose relationships were nondistressed. These findings are particularly notable because analyses controlled for baseline depressive symptoms, which were significantly higher for maritally distressed survivors. Thus, we found marital distress to be associated with many poorer outcomes beyond those occurring when individuals have persistent depressive symptoms. Other variables—sociodemographic, disease characteristics, and treatments received—were also controlled.

Our previous studies21, 22 suggested that heightened levels of cancer-specific and global stress would be found at diagnosis and the beginning of adjuvant treatment for all patients. Theorists have proposed that cognitive processing of stressful or traumatic events, such as conversation with a supportive partner, may relieve traumatic stress symptoms.45, 46 Thus, the recovery noted in the Nondistressed group in both cancer-specific and global stress may have, at least in part, occurred because of the partner's facilitation of cognitive processing of the breast cancer experience.

Chronic stress can contribute to physical and psychologic consequences,47 which could be particularly deleterious for individuals diagnosed with cancer. Thus, the health findings are interesting to note. At baseline, all patients were recovering from surgery and anticipating the start of adjuvant chemotherapy and radiotherapy. Subsequently, though both groups showed steady improvement in performance status, that of the Distressed group was notably slower and, at 5 years, remained at a lower level than that of the Nondistressed group. For the symptoms/signs data, significant differences were observed at baseline between the groups. As the groups were statistically equivalent with regard to variables apt to produce symptoms/signs (eg, type of surgery received), it is reasonable to hypothesize that some of the additional symptomatology may have been stress-related. Manne et al have suggested that avoidant or intrusive thoughts related to one's cancer experience are associated with deteriorations in physical performance status, which may account for the similarities between the cancer-specific stress and performance status trajectories observed here.48 Furthermore, post hoc analyses suggest that marital distress in combination with depressive symptoms are related to the poorest recovery.

The current study provides new data regarding health behaviors of breast cancer survivors. Overall, the data suggest that women may make modest dietary improvements and be more active during the first 2 to 3 years after diagnosis, with the exception of maritally distressed and depressed women, who show a constant decline in healthy dietary habits. However, there appears to be a precipitous decline thereafter among all individuals. The analyses demonstrate this trend to be even more striking among the Distressed group, providing empiric support for the notion that health behaviors of the maritally distressed are indeed different from the nondistressed.11 As data suggest that activity level may be positively correlated with breast cancer survival,49 this may be an important finding.

With regard to activity level, the finding that the Nondistressed group had a significantly lower baseline than the Distressed group was surprising and not readily explained. As some individuals with heightened symptom/signs increase their level of physical activity to reduce discomfort, perhaps the women in the Distressed group engaged in more activity to reduce, for example, symptoms of fatigue. Or, it might be the case that these particular individuals used higher levels of activity to cope with stress more so than the Nondistressed group, in that their higher activity levels continued for 18 months after diagnosis. As these are new findings for the literature, future research is needed to clarify this differential pattern of activity if it is found to be reliable.

The routes by which marital distress is linked with poorer health may be through heightened distress or poorer health behaviors, as noted above. We found a 5-year trajectory of sustained (even increasing) global stress simultaneous with slowed recovery. Perhaps if maritally distressed women had maintained a higher level of physical activity, their recovery might have been quicker. Biologic variables, such as endocrine or immune responses,14 may have also covaried with marital distress to impact the health outcomes. In any case, further investigation is warranted to determine additional temporal relationships among distress/health behavior mechanisms.

Limitations regarding generalizability of these findings should be noted. First, data came from those of a larger clinical trial for which only stage II or III breast cancer patients were eligible. The sample was also homogeneous in other respects: primarily white, middle-class, and middle-aged. However, these characteristics may be advantageous in relation to marital distress41; these data may underestimate any impact of marital distress when it occurs among couples who also have fewer socioeconomic resources, for example. The data may not generalize to other racial or ethnic groups, men, gay/lesbian relationships, or those who divorce after cancer. Second, this is an observational study. We did not, of course, randomly assign individuals to Distressed or Nondistressed groups, nor did we obtain marital satisfaction ratings before the women's diagnoses. The prior literature suggesting that marital distress is usually a chronic phenomena and the stability of patients' satisfaction reports during the ensuing 5 years, do, taken together, suggest that our classification of patients, albeit imperfect, was likely accurate for the majority. It would be valuable for future research to examine similar outcomes for subsets of women, particularly those with both depressive symptoms and low marital satisfaction and those few with various patterns of change in marital satisfaction.

In conclusion, we believe the current study provides a first look at the long-term relationship between marital distress and physical, health behavior, and mental health outcomes for breast cancer survivors. The data implicate marital distress as a risk factor for numerous poorer outcomes, some observed immediately after diagnosis and others persisting for years. As such, and given the sizable proportion of patients likely experiencing marital distress (25%-35%), screening patients for marital distress might be considered, although it would warrant a lower priority than screening for psychologic distress. Also, some patients might be puzzled by it or regard it as intrusive. Alternatively, our data demonstrated that approximately 25% of those in the Distressed group reported clinically significant depressive symptoms at baseline (CES-D ≥10) (Table 2), suggesting that screening for depressive symptoms would identify many of those with marital distress as well. Subsequently, appropriate referrals could be made (eg, see couples intervention studied by Manne et al50). Such assistance may offset poor survivorship outcomes for patients doubly burdened by psychologic distress and relationship problems.

Acknowledgements

We thank the participants in the Stress and Immunity Breast Cancer Project and the professional and research staff. Special thanks to Dr. Barbara Andersen, Dr. Lisa Thornton, and Brittany Brothers for comments.

Conflict of Interest Disclosures

Supported by the American Cancer Society (PBR-89 and RSGPB-03-248-01-PBP), Longaberger Company-American Cancer Society Grant for Breast Cancer Research (PBR-89A), US Army Medical Research Acquisition Activity Grants (DAMD17-94-J-4165, DAMD17-96-1-6294, and DAMD17-97-1-7062), National Institutes of Mental Health (1 R01 MH51487), the National Cancer Institute (K05 CA098133 and R01 CA92704), the General Clinical Research Center (M01-RR0034), and the Ohio State University Comprehensive Cancer Center (P30 CA16058).

Ancillary