| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Epidemiology Research Group (E.M., J.B., M.M., L.D.), Department of Social and Preventive Medicine (E.M., J.B., M.M., R.V.), and Department of Surgery (L.D.), Université Laval, Québec, Québec, Canada.
Address reprint requests to: Dr. E. Maunsell, Groupe de recherche en épidémiologie, Centre de recherche, Pavillon St-Sacrement, 1050 chemin Ste-For, Québec, QC, Canada, G1S 4L8. Email: elizabeth. maunsell{at}gre.ulaval.can
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONPATIENTS AND METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
METHODS: This study was based on women with histologically confirmed, newly diagnosed, localized or regional stage breast cancer first treated in 1 of 11 Quebec City (Canada) hospitals from 1982 through 1984. Among 765 eligible patients, 673 (88%) were interviewed 3 to 6 months after diagnosis about the number and perceived impact of stressful events in the 5 years before diagnosis. Three scores were calculated: number of events; number weighted by reported impact; and for almost 80% of events, number weighted by community-derived values reflecting adjustment required by the event. Scores were divided into quartiles to assess possible dose-response relationships. Survival was assessed in 1993. Hazard ratios and 95% confidence intervals (CIs) comparing all-cause and breast cancer–specific mortality were calculated with adjustment for age, presence of invaded axillary nodes, adjuvant radiotherapy, and systemic therapy (ie, chemotherapy and hormone therapy).
RESULTS: When quartiles 2, 3, and 4 were compared with the appropriate lowest quartile, adjusted hazard ratios for all-cause mortality were 0.99 (CI = 0.70–1.38), 0.97 (CI = 0.73–1.31), and 1.04 (CI = 0.78–1.40) for number, number weighted by impact, and number weighted by community-derived values, respectively. Results were essentially similar for the relation between stressful life events limited to those occurring within the 12 months before diagnosis and overall mortality and between stressful life events in the 5 years before diagnosis and breast cancer–specific mortality.
CONCLUSIONS: Stress was conceptualized as life events presumed to be negative, undesirable, or to require adjustment by the person confronting them. We found no evidence indicating that this kind of stress during the 5 years before diagnosis negatively affected survival among women with nonmetastatic breast cancer. Evidence from this study and others on the lack of effect of this type of stress on survival may be reassuring for women living with breast cancer.
Key Words: breast neoplasms • stress • survival • prognosis • quality of life • survivor.
Abbreviations: CI = confidence interval; HR = hazard ratio; LES = Life Events Survey; OR = odds ratio.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONPATIENTS AND METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Few clinical or epidemiologic studies of breast cancer prognosis have information on women’s experience of stress. As well, those few studies that have assessed the relation of stress with survival among women with localized or regional disease at diagnosis have provided contradictory results, in some cases based on quite small numbers of women (4–10). The main objective of this study was to assess the contribution of stressful life events in the 5-year period preceding diagnosis to all-cause mortality over the 7 years after diagnosis among women with newly diagnosed localized or regional stage breast cancer.
|
PATIENTS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONPATIENTS AND METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Study data were obtained through interviews conducted at home by a specially trained nurse interviewer and by review of patients’ medical records. Of the 765 eligible patients during the study period, 673 (88%) were interviewed 3 to 6 months after diagnosis. Principal reasons for exclusion were metastatic disease at diagnosis, absence of surgical treatment providing a date of diagnosis, inability to conduct a home interview for reasons of time or distance, and refusal by the treating physician to allow contact with the patient. Among women interviewed, histologic data were unavailable for review for seven women, and one woman provided no information on stressful life events. Thus, this analysis was based on data from 665 women.
Information on personal and disease characteristics at diagnosis, including age, number of invaded axillary lymph nodes, and stage of disease, was abstracted from patients’ hospital records about 6 weeks after diagnosis. Stage was defined as localized or regional according to the surveillance, epidemiology, and end results classification (18).
Data on adjuvant radiation treatments, chemotherapy, hormone therapy, vital status, nature and timing of any new disease events (local, regional, or distant recurrence), and related treatments were obtained in 1993. To obtain this information, patients’ medical records from either the hospital where initially treated or the facility where they had undergone radiotherapy were consulted. For some women, it was necessary to write to the patients’ family or new treating physician to obtain information on vital and/or disease status.
Stress was conceptualized in terms of "stressors," that is, life events presumed to be negative, undesirable, or to require adjustment by the person confronting them (19–21). The number and perceived impact of stressful life events in the 5 years before diagnosis were ascertained using a modified version of the LES ( Table 1) (22). Most events assessed were relatively objective and represent discrete changes in the woman’s social or personal environment (23). Only items that seemed most appropriate given the age of the majority of women with breast cancer were retained (24, 25). Women were not questioned about items that could be symptoms of psychiatric illness (eg, major changes in eating or sleeping habits or sexual difficulties). For each event the patient was asked whether the event had occurred in the year before diagnosis or during the 4-year period before that, thus covering the 5 years before diagnosis. For a small number of events, some women (N = 124) were unable to decide between these two periods. Therefore in the main analyses, events considered were those occurring in the 5-year period preceding diagnosis without distinction as to whether they occurred in the year before diagnosis or the 4 years before that (26). No personal health events were included among events considered for analysis. This exclusion was made because women were interviewed up to 6 months after diagnosis. This exclusion ensured that health status at the time of the interview, which may have included early signs of disease recurrence for some women, would not influence reports of personal health problems.
|
Overall and disease-free survival was assessed on the basis of the information obtained in 1993. For women who died during the 7-year period, survival time was calculated as days between mastectomy and either death due to any cause or death due to breast cancer. In the analyses of breast cancer–specific mortality, women who died of other causes were treated as censored observations. For women who had not died, follow-up was terminated 7 years after mastectomy.
For the principal analyses presented here, event scores were divided into approximate quartiles. This was done to assess whether there was any evidence of a dose-response relationship. The Cox proportional hazards model was used to calculate HRs and their 95% CIs (28). Using the score for total number of events as an example, the HR represents the death rate over the 7-year period among women reporting 2 to 3, 4 to 5, or 6 to 15 life events divided by the rate in the reference group (those reporting 0 to 1 event). This reference group was chosen because very few women reported no stressful life events in the 5-year period. A HR of 1 indicates that the death rates in the two groups compared were the same, whereas a HR of >1 and <1 means a higher and lower death rate, respectively, compared with women in the reference group. Event scores were also analyzed as a continuous variable and with divisions at the median and in terciles. This was done to ensure that results were unchanged when different divisions were used. The propor-tional hazards assumption, meaning constant HRs over time, was verified for all event scores and adjustment variables before model specification.
A number of medical, sociodemographic, and psychosocial characteristics were considered individually as potentially confounding variables. These characteristics were presence of invaded axillary nodes at diagnosis, tumor size at histologic examination, stage at diagnosis, hormone receptor status, type of surgery, adjuvant therapy, age, years of schooling, weight, and marital status. On the basis of these analyses and to ensure that the HR assessment of the strength of the stress-survival associations was independent of medical and treatment prognostic factors, all models presented were adjusted for age, presence of invaded axillary nodes, adjuvant local therapy (radiotherapy), and systemic therapy (chemotherapy and hormone therapy).
Finally, in complementary analyses we verified whether stress in the preclinical phase negatively affected key intermediate markers of disease progression measured at diagnosis before the start of treatment, including axillary node status, tumor size, disease stage, and histologic grade. Women with one or more invaded nodes, a larger tumor (>2 cm in diameter), regional stage at diagnosis, and a poorly differentiated tumor, respectively, were considered to have a poorer prognosis. We first examined frequency distributions for each of these prognostic indicators according to each stress variable. We also used logistic regression to calculate ORs and their 95% CIs to quantify the association of stress with these factors.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONPATIENTS AND METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Descriptive information on disease, treatment, sociodemographic characteristics, and stressful life events among these patients is presented in Table 2. Most women (86.6%) had infiltrating ductal carcinoma at histologic examination. Women were heterogeneous on other characteristics.
|
Overall survival was 67.3% at 7 years. However life event information was scored, no consistent survival differences disfavoring women with higher numbers of stressful life events or more heavily weighted events in the 5 years preceding diagnosis were observed when these women were compared with women in the respective reference group ( Table 3). Crude and adjusted HRs were slightly higher and slightly lower than the null value of 1, and all 95% CIs contained the null value. No clear gradient demonstrating that more events or more heavily weighted events were increasingly associated with higher death rates was apparent. Adjustment for age, nodal status, and adjuvant treatments generally tended to decrease any differences observed. HRs for breast cancer–specific mortality comparing these stress groups were even closer to the null value than those described in Table 3 (data not shown). When quartiles 2, 3, and 4 were collapsed into one exposed group and compared with the appropriate reference group, adjusted HRs for all-cause mortality were 0.99 (CI = 0.70–1.38), 0.97 (CI = 0.73–1.31), and 1.04 (CI = 0.78–1.40) for number, number weighted by impact, and number weighted by Holmes and Rahe values, respectively. None of these results were significantly changed when personal health events were included in the calculation of the three life event scores (data not shown).
|
3 (N = 107) events were 1.09 (CI = 0.78–1.5) and 0.71 (CI = 0.45–1.12), respectively. These estimates were virtually unchanged when events reported during the 4-year period preceding the year before diagnosis were included among the adjustment variables (data not shown). This analysis was based on 83.2% of 2480 events reported that could be assigned to either the year immediately preceding diagnosis or the 4 years before that.
The association of stress with all-cause mortality was examined separately for women living with a spouse and those not and for those with 
7, 8 to 11, and 12 years of education. Adjusted HRs comparing 2 events to 0 to 1 event in the 5-year period were as follows: women living with a spouse, 1.01 (CI = 0.64–1.61); women not living with a spouse, 0.96 (CI = 0.58–1.58); 7 years of education, 1.05 (CI = 0.64–1.72); 8 to 11 years of education, 0.91 (CI = 0.51–1.64); and 12 years of education, 0.83 (CI = 0.39–1.77), respectively.
Finally, no negative effect of stress on tumor size, disease stage, or histologic grade was observed (data not shown). The only observed, although not statistically significant, difference concerned nodal status. The observed likelihood of having invaded nodes was lower among women in quartiles 2, 3, and 4 (combined) than among those in the lowest quartile for number of events weighted by impact and number weighted by Holmes and Rahe values. Observed ORs were 0.77 (0.54–1.10) and 0.74 (0.52–1.04), respectively. Upper limits of the 95% CIs are also consistent with no effect of stress on prognosis. These results were not changed by adjustment for age at diagnosis.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONPATIENTS AND METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Seven previous studies have assessed the relation of stress with recurrence or survival after breast cancer (4–10). Key characteristics of these studies are presented in Table 4. Measures of stress used have differed, as have the periods over which stressors have been considered as exposures. All but one study (10) took into account age and some medical prognostic factors in the analyses presented. In this area of inquiry, the preferred design is one in which stress exposure is measured before disease recurrence is diagnosed (32). This is because it is impossible to rule out recall bias as the explanation for any observed deleterious effects of stress on rate of recurrence if stressors are measured after disease recurrence has occurred. Recall bias could occur if women with recurrence report more events or describe them as more serious as part of their own attempt to explain why illness recurred compared with well survivors. Six studies used the stronger design (4, 5, 7–10), and one assessed stressors after diagnosis of recurrence using a case-control design (6). Findings from five of the seven studies are consistent with our results and show no deleterious effect of stressful life events on survival after breast cancer (4, 5, 8–10). All five studies used the preferred design. Two studies have reported a deleterious effect of stress on prognosis. One of these studies did not use the preferred design (6), and the other included a number of possibly correlated variables in multivariate models (7).
|
|
Because checklist measures of stressful events of the type we used have been criticized, it is important to assess possible limitations or effects on interpretation of our results. We used a modified version of the LES. Two types of modifications were made. First, certain events thought to be less pertinent to the majority of women with breast cancer (based on life stage of the women) were excluded. Tailoring event lists to the population and/or life stage studied has the advantage of increasing sensitivity and relevance (25). The second modification was not asking patients to rate events as either positive or negative before asking them to rate perceived impact. This was done to cut down on interview time so that we could assess a relatively large number of events. As a result, life events are presumed to be stressful, that is, negative, undesirable, or to require adjustment by the person confronting them. Our observations from 205 of these same patients, who also responded to an additional interview section on quality of life, indicates that they were (15). We observed increasing psychological distress in women reporting 2 to 3, 4 to 5, and 6 to 15 events in the 5 years preceding diagnosis compared with those reporting 0 to 1 such event, and this was observed both 3 and 18 months after initial treatment. Finally, despite the absence of positive/negative ratings, the application of Holmes and Rahe weights, which reflect population judgments about the amount of adjustment required, produced virtually identical results.
Analyses of simple event counts have also been criticized because implicitly this means that an event like death of spouse gets the same weight as a child leaving home (25). However, three different scoring conceptualizations (a count and two others that incorporated either the woman’s or population judgments about impact) produced very similar results. We verified that results did not depend on cutoff points for defining high stress groups. We also observed that despite variations in order, the five events most frequently mentioned by women in quartiles 1, 2, 3, and 4 of the number score were fairly similar. This indicates that some events of similar stressfulness were reported by women in all quartiles and suggests that women in the highest quartile did not simply report a larger number of possibly less stressful events.
We chose to concentrate analyses on events during a 5-year period preceding diagnosis, although some believe that a 1-year period is more appropriate (25), partly because of the problem of falloff in reporting with time. Our choice of a 5-year period was based on several considerations. It is thought that many breast cancers have likely been present for several years before clinical diagnosis. Thus, a 5-year period seems relevant for assessing the effect of stressful life events on evolution of breast cancer because during most of this period disease was likely to be present even though it had not been diagnosed. Although falloff in reporting is possible with longer recall periods, it is probably less serious for more objective events such as the ones assessed here (25). In fact, we observed that the cumulative number of events per participant increased from the first period, the year before diagnosis, to the second, which included the 4 years before that. Although it is impossible to tell whether women reported all events, any failure to report events is unlikely to be differential because all these women were confronting the same stressor, breast cancer diagnosis and treatment, when interviewed. Finally, we assessed the relation of stressful events occurring only in the year before diagnosis with all-cause mortality. Results from this analysis were similar to those based on events from the entire 5-year period.
Finally, some authors question the value of checklists compared with a semistandardized interview for assessing events and difficulties (33). Interviews are better able to date event occurrence and duration and to characterize events in terms of undesirability and independence. However, on the basis of results of available studies of stress and survival after breast cancer, we suggest that study design may be more important than event measurement method. The potentially superior interview method yielded diametrically opposed results when used in studies in which events were assessed before (4) and after (6) recurrence had occurred. On the other hand, results from this and most of the other studies reviewed, in which events were ascertained before disease recurrence, were similar, although both interview and checklist methods were used.
One of the main pathways proposed by which stress could affect survival involves immune function. In this view, stress leads to distress, which may lead to immune downregulation directly through changes in immunologically active cells (21, 34, 35), as illustrated in Figure 2. In a subgroup of these women, we were able to verify the relation between A and B, that is, that those with increasing numbers of stressful life events were also more likely to be distressed (15). Specifically, among a consecutive series of 227 of these 673 women first treated in seven of the participating hospitals in 1984 who also responded to a brief interview about different aspects of quality of life (13–17), the odds of having high distress increased linearly and significantly as the number of stressful life events increased and were 2.9 and 3.8 times higher in women reporting 6 to 15 events compared with those reporting 0 to 1 event 3 and 18 months after surgery, respectively. Others have demonstrated a relation between A and C in the proposed model. Negative life events, considered either singly or together, reliably affect a number of immune parameters (21), and lowered cellular immune responses have also recently been reported in women stressed by newly diagnosed breast cancer (36). Among the subgroup of women who responded to quality-of-life questions, we were also able to assess the relation between B and D, that is, whether increasing levels of distress were associated with poorer survival. The HR comparing the mortality experience of women with high distress levels with that of those with lower levels, adjusted for age, nodal status, and localized and systemic therapy, was 0.70 (95% CI = 0.40–1.22). It should also be mentioned that the relation between stressful life events and mortality in this subgroup was the same as that reported here for the entire group. Together these observations are consistent with the idea that if immune changes occurred, their magnitude or duration was insufficient to affect length of life (34).
|
Finally, social support may help reduce the negative effects of stress by acting through either immune, endocrine, or behavioral pathways and thus may counteract any deleterious effects of stress on survival (34). We have observed in two successive studies that a very substantial proportion of women with newly diagnosed disease, 87%, report having a confidant at some time during the year after diagnosis (17, 37). Furthermore, in the subgroup of this cohort also interviewed about their quality of life, the relative mortality of women who had a confidant in the 3 months after surgery was 39% lower 7 years after diagnosis than that of women without such a confidant (17).
In summary, although data from further studies must be awaited, we believe our findings of the effects of this kind of stress on survival provide some reassurance for women living with breast cancer. These women wonder about possible effects of stress on their survival. Our study presents the strongest evidence to date showing no effect of multiple stressors in the preclinical period on either early indicators of prognosis measured at diagnosis or on overall survival, whether early on or later during the course of the disease.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONPATIENTS AND METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Received for publication May 25, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONPATIENTS AND METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  L. R. Schover Myth-Busters: Telling the True Story of Breast Cancer Survivorship J Natl Cancer Inst, December 15, 2004; 96(24): 1800 - 1801. [Full Text] [PDF]
|
|
|
|
 |
|
 Other articles noted Evid. Based Ment. Health, November 1, 2001; 4(4): 104 - 104. [Full Text] [PDF]
|
|
|
|
 |
|
 Other Articles Noted Evid. Based Nurs., October 1, 2001; 4(4): E1 - 11. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
