Sot.Sci.Med. Vol.33,No. 2.pp. 177-183. 1991 Rimed in Great Britain.
0277-9536191 S3.00+0.00 Copyright 0 1991 Pergamon Press plc
AI1rights reserved
EFFECTS OF SOCIAL RELATIONSHIPS ON SURVIVAL FOR WOMEN WITH BREAST CANCER: A PROSPECTIVE STUDY NANCY
WAXLER-MORRISON’, T. GREGORY H~~LoP~, BRONWEN MEARS’ and LISA KAN*
‘Department of Anthropology/Sociology, University of British Columbia, Vancouver and ‘Division of Epidemiology, Biometry and Occupational Oncology, British Columbia Cancer Agency, Vancouver, Canada Abstract-In this study we examine the relationship between a woman’s social contexts at the time of diagnosis and her chances of having survived breast cancer four years later. A cohort of 133 women were followed prospectively after initial diagnosis and treatment and data were obtained from a questionnaire mailed soon after diagnosis and from hospital charts. Using multivariate methods to examine predictors of survival, two clinical factors, pathologic nodal status and clinical stage of disease, were significantly associated with survival. In addition we found significant and independent effects on survival ofz number of supportive friends, number of supportive persons, whether the woman worked, whether she was unmarried, the extent of contact with friends and the size of her social network. Thus, the woman’s social context, particularly contexts of friendship and work outside the home, are statistically important for survival. Using existing literature, further data analyses and interviews with some survivors, we speculate on the ways in which social contexts may influence survival and suggest research methods suitable to this question. Key words-breast
cancer, social networks, social support, survival.
INTRODUCTION
Many researchers have attempted to link women’s social and psychological characteristics to their ability to adjust to the distress of breast cancer [l-5]. But it was not until Berkman and Syme [6] reported that size of one’s social network was predictive of mortality in a community sample that some began to ask whether the women’s social contexts or psychological traits might tell us something about their chances for survival after breast cancer had been diagnosed. Breast cancer prognostic studies have generally followed several different theoretical perspectives and used a variety of analytic methods. One set uses psychological theory to understand survival, with inconsistent results. Both Greer [A and Derogatis [8] focused mainly on the women’s psychological methods of coping with the initial illness as a predictor of short- or long-term survival. Greer reported that short-term survivors tended to be those who revealed, at diagnosis or during the early stages of treatment, helplessness or stoic acceptance. Derogatis found that women who survived were more anxious and hostile at diagnosis than were short-term survivors; long-term survivors were, from the point of view of health professionals, also less well adjusted to their illnesses and less cooperative with the doctors. Others have used sociological theory, sometimes implicitly, in an attempt to link characteristics of the woman’s social relationships or social network to her survival and have again reported mixed findings. Funch and Marshall [9] showed that a death, an illness or unemployment in the breast cancer patient’s family predicts poorer survival just as did the patient’s perception that family income was in-
adequate. Her membership in formal organizations, however, increased chances of survival. Neale (lo] reported that being married, as opposed to being a widow, is protective. However, Goodwin [I l] has shown that when those women who are currently married at the time of breast cancer diagnosis are compared with all others (single, divorced, separated and widowed) the relative risk of death is the same for each group. Many studies have one or more methodological limitations, particularly lack of multivariate analytic methods, lack of statistical control for clinical variables, and, most important, relatively poor measures of women’s social characteristics. Our earlier report [12] dealt with some of these methodological problems and examined some psychological and social predictors of survival. After controlling for clinical factors, three psychosocial factors were significant for survival: the extent to which the woman, at the time of diagnosis, engaged in expressive activities in the home (e.g. social activities); her level of extroversion; and the extent to which she expressed/felt anger. In that initial report the range of social and psychological characteristics investigated was very narrow. In this paper we have expanded our analysis to reflect a sociological perspective, one suggesting that a woman’s social role relationships may play a part in her response to breast cancer and ultimately in her survival. We have thus examined the effect of the woman’s social network, her relationships with family and friends, and her work experience, on survival. In addition, using interviews of a small group of longterm survivors, we have speculated on the meaning of these factors in women’s lives.
177
NANCY WAXLER-MORRISON et
178 METHODS
The study group comprised women who were referred to the A. Maxwell Evans Clinic (AMEC) in Vancouver, who had a confirmed diagnosis of primary ductal breast cancer between June 1980 and May 1981, and who were under 55 years of age at diagnosis (in order to examine mainly premenopausal disease). One hundred and thirty-three of the 168 eligible patients returned both study questionnaires (u 80% response); the clinical characteristics of responders and nonresponders were similar [12]. AMEC is the only facility where cancer patients can receive radiotherapy on the mainland of British Columbia; 2/3 of all breast cancer patients on the mainland are seen here. The type of initial treatment was dependent upon the clinical stage and site of the tumour [12]. Most patients received modified radical mastectomy. Referrals to AMEC usually occurred shortly after diagnosis. These patients were mailed a self-adrninistered questionnaire to obtain demographic and social network information, followed by a second selfadministered questionnaire to obtain the additional psychosocial information reported in Hislop [12]. Ninety percent of the returned questionnaires were completed within 3 months of diagnosis. Sources and scoring methods for our earlier analysis of psychosocial factors are described elsewhere [12]. Additional sociological data were obtained from the initial questionnaire on educational level; the extent to which she was responsible for others in the home; the extent of her contact with friends and/or relatives; her perception of help/support available from relatives, neighbours, and friends; employment status; marital status; and a measure of social network derived from Berkman and Syme [6] that combines marital status, contact with friends/ relatives, and church membership. Specific questions and scoring methods for these additional social factors are in the Appendix. Breakpoints for ordinal factors (e.g. numbers of supportive friends) which had more than 4 levels were based on frequency distributions in the total study group; breakpoints for the earlier psychosocial factors remained the same as in the original analysis [12]. Clinical factors were abstracted from the patient’s medical record by a trained clerk using a standardized form; the clerk was unaware of the questionnaire responses. The clinical factors included age at diagnosis, clinical stage, pathologic nodal status, histological tumour grade, estrogen receptor (ER) tumour status, type of initial treatment, time to first recurrence of disease (local, regional or metastatic), survival and disease status when last seen, and time to last follow-up or death. Recurrence and survival data was abstracted in January, 1985. Survival was calculated from the time of diagnosis using the product-limit method of Kaplan and Meier [ 131.Observed differences in survival for single factors were examined using the log-rank statistic. The Cox proportional hazards analysis [ 141 was used to determine the maximum likelihood estimates of regression coefficients for each psychosocial factor separately, adjusting for the clinical prognostic factors; the Pvalue indicates the statistical difference between the
al.
levels for a given factor and is not a test for trend. This method of analysis was also used to test for independent effects by examining the significant social and psychological factors together, again adjusting for clinical factors. For analyses examining each social factor separately, patients were excluded with missing information for the factor being analyzed. For analyses examining independent effects, patients were excluded with missing information for any of the factors being compared. Fifteen patients were excluded from this latter analysis because of missing information; hence the final multivariate analysis was restricted to 118 patients. Finally, in order to better understand the importance of social relationships, 18 surviving patients from the original study group who lived in the Greater Vancouver area were interviewed in 1986. Open guided interviews (phenomenological) were used to obtain rich descriptive information from the patients. FINDINGS
Overall survival In the 4 years of follow-up there were 26 deaths (25 attributed to breast cancer) reported for the study group. Over 90% of the patients were followed to at least 2 years, 80% to 3 years and approximately 30% to 4 years. The survival rates were 96%, 89%, 80% and 76% at 1, 2, 3 and 4 years, respectively. Clinical factors All clinical factors were related to survival in the expected direction, however, only two were significantly related to survival: pathological nodal status and clinical stage (Table 1). Social factors The Cox proportional hazards method was used to calculate estimates of the instantaneous relative death rate for each new social factor separately after including the two significant clinical factors (pathological nodal status and clinical stage). Six of the eleven measures of social relationships were significantly associated with survival: marital status; support from Table 1. Percent survival bv clinical factors Survival P-value.
Factor
Age
Clinical stage
Pathological nodal status
Histologic grade Estrogen receptor status
Level
n
2 Yr
4 Yr
log rani;
I I people Work: Employed Not emnloycd Social network size Small Medium Large
I
Estimate*
101
1.00
17
0.17
46
1.00
:;
0.61
54 48 14 2
P-value
O.CGO4
0.0001
1.00
I .29 3.14 14.29
0.0005
:: 29
1.00 1.27 1.01 0.65
0.0802
62 56
1s2
40 61 17
1.00 0.86 1.59
28
1.00 0.0003
0.0006
friends; contact with friends; total support (friends, relatives, neighbours); employment status; and social network. Five measures (responsibility for others; educational level; support from relatives; contact with relatives; and total contact with friends and relatives) were not significantly associated with survival. From our earlier analysis three psychosocial factors had been associated with survival after adjustment for clinical factors: frequency of involvement in social activities in the home, degree of extroversion and degree of anger [12]. We then repeated the Cox proportional hazards analysis to identify independent factors, limiting the sample to the 118 with no missing information for any of the factors being compared. The six significant social factors and the three significant psychosocial factors from our earlier analysis were examined after including the clinical factors, pathological nodal status and clinical stage. (The distribution of the three psychosocial factors, and the rate of survival, in this group of 118 women were similar to that in our earlier report.) Table 2 shows the instantaneous relative death rate for the significant independent factors. (Rates above 1.OOindicate decreased survival times, whereas rates below 1.00 indicate increased survival times; the greater the difference from 1.00, the stronger the effect upon survival.) The six social factors remained significant whereas the psychosocial factors were no longer significant. The extent to which a woman feels that she can call upon three or more friends for support or help was most strongly associated with survival. Correlations between our earlier psychosocial factors and new social factors
With the loss of significance for our earlier psychosocial factors, we examined the relationship between 3312-F
these three factors and the significant social factors. The ‘expressive activities at home’ factor was significantly correlated in the expected direction with ‘contact with friends’ (Speannan p = 0.32; P = 0.006), ‘support from friends’ (p = 0.24; P = 0.05). and ‘social network’ (p = 0.32; P = 0.002). Thus ‘expressive activities at home’ seems to reflect in a concrete and limited way the support and contact dimensions of friendship. The ‘extroversion’ factor was significantly correlated with ‘contact with friends’ (p = 0.25; P = 0.01) and ‘support from friends’ (p = 0.30; P c O.OOl), suggesting again that this personality trait reflects several dimensions of the friendship relationship. The third earlier factor, ‘anger’, was not significantly related to any social factors in our model. We examined also for correlation between the significant social factors and other social or demographic factors; only ‘marital status’ showed significant relationships, with both ‘work’ (p = 0.19; P = 0.01) and ‘education’ @ = 0.17; P = 0.03). Interviews of a sample of survivors
*The smaller the estimates (that are less than 1.00) the greater the likelihood of survival, whereas the greater the estimates (that are greater than 1.00) the less the likelihood of survival.
SW
179
In order to better understand the significance of social relationships in long-term survivors, 18 patients were interviewed by the phenomenological method. Our interviews gave some hints as to the dimensions of friendship that may be important. Many survivors mentioned the practical, often timethings that friends offered such as consuming, cooking, child-care and shopping. Having a friend to provide transportation twice a week to the hospital, who waited in the corridor, was the kind of commitment that many women found important. ‘Having a shoulder to cry on’ was mentioned, but seemed secondary to concrete help. There was also the observation that, during the crisis of diagnosis and treatment, ‘you discovered who your real friends are’. Some women lost friends, either temporarily or permanently, apparently because of the friend’s inability to tolerate life-threatening illness. Some women found new friends, often women who had had breast cancer in the past who offered help; sometimes new friendships were established with women who were in the hospital at the same time. Most of the married survivors we interviewed reported very supportive and helpful responses from their spouses in the initial treatment period. The women themselves and their spouses usually recognized that breast loss was a minor issue compared with the threat to life that cancer posed. This threat to life was, in some instances, though, enough to break up a marriage when the woman’s husband could not tolerate a wife with cancer. Married survivors tended to report complex patterns of support and protection between family members which were often experienced by the women as ambivalent and sometimes very stressful. It was the young adult children that women felt to be least supportive, most difficult and unhelpful to them. Many women ‘had to support them by being cheerful’ and ‘carrying on as if things were normal’ since children of this age appeared to be fearful for their mothers’ lives; some children simply stayed away during the most stressful periods.
NANCY WAXLER-MORRWN er al.
180
Many surviving women downplayed the financial aspects of work, stating that they could have managed financially without work but continued working either as volunteers or for pay, in part, because ‘work took their mind off illness’. However, it was clear from these women that jobs were a source of support, were symbolic of ‘getting back to living’, and more important, provided useful information. DISCUSSION
While there is in the research literature a large body of work linking women’s social and psychological characteristics to their adjustment to breast cancer over the short-term there are relatively few studies attempting to relate these factors to their survival [7,11,15]. The current study combines a number of methodological features that makes it relatively strong. The design is prospective, the sample size is relatively large, social factors were measured directly from the patient rather than from registries or hospital charts, and multivariate analytic methods were used. There are, however, methodological limitations of the study. First, women over age 55 (postmenopausal) were not included in the sample, thus omitting women in the higher incidence age cohorts; this prevented us from understanding how some social factors, particularly widowhood, may relate to survival. As well, the mailed questionnaire method necessitated obtaining limited information and thus leaves us with many unanswered questions. Measures of social network and social support were those available from the literature in 1980; both methodological and theoretical advances have been made since then [16-181. All social factors were measured shortly after diagnosis and surgical treatment (usually mastectomy), hence these factors are likely to be reflections of the women’s social situation during the crisis period. We do not know whether social relationships during this crisis period are in some ways consistent with the women’s earlier social relationships, although we would assume that they are. Finally, we have no information on social relationships during the intervening follow-up years and thus cannot document how these may effect survival. Our earlier study (121 was based on a limited number of social and psychological variables which pointed to three significant correlates of survival after clinical factors had been controlled: expressive (emotional, relational) activities in the home, extroversion and anger. With the addition of social support and network factors in the current analysis, these initial factors were no longer significant, no doubt because they are correlated with the new social factors in our model. What we have found in the current analysis is that several characteristics of the women’s friendship and social network, her marital status and employment status are significantly and independently related to survival. There is little evidence in our findings that relationships within the woman’s family and with her relatives are important for survival. Instead, the social contexts that appear to be significant are outside the woman’s home. Friendship
and work are the themes that seem to underlie the findings. Previous research on breast cancer survival is not very helpful in understanding our findings. Both Greer [I and Derogatis [8] focused mainly on the woman’s psychological methods of coping with the initial illness as predictors of survival and reported shorter survival for those who revealed at diagnosis or during the early stages of treatment, low levels of hostility, high levels of positive mood, stoic acceptance or hopelessness. However, Derogatis’ findings that the treating physician rated long-term survivors as significantly less adjusted to their illness, possessing more negative attitudes to the illness and its treatment, and that the interviewer judged the longterm survivors as having “significantly poorer attitudes toward their physicians*’ [p. 15061 may be important for hypothesis construction. Greer [A does show that an undescribed measure of “interpersonal relationships” was not related to survival, but that being unmarried or reporting poor marital relationships at the time of diagnosis was related to a less favourable outcome. Funch and Marshall [9] found that a measure of “number of relatives and friends” was not a significantly related to survival. However, they report that a factor called “organizational involvement” (unfortunately not further defined) was significant for survival and independent of the significant effects of other stresses (e.g. death in the family). Neale et al. [lo] found that married rather than widowed women survive, even when age is controlled; unfortunately, no divorced, single or separated women were included in the sample, thus preventing any direct comparison with the data reported here. Other investigators have found no significant re.lationship between marital status and survival in breast cancer patients [9, Ill. These studies have not measured directly and comprehensively characteristics of the women’s social network, her social support or her work experience and thus it is difficult to use this research to examine on our own findings. The numerous studies examining the effects of social support, social networks, etc. on udjusttnenr (as indicated by coping methods, psychological distress, etc.) to breast cancer [l-5] might be thought of as a rich source of hypotheses linking social factors to survival, but we are wary of making that link since at least one researcher [8] has shown that the woman who is thought by her physician to be less adjusted is more likely to survive. With the exception of Derogatis’ work, the strength and direction of the relationship between adjustment to having breast cancer and survival has, as far as we know, not been documented. It is premature, as well, to use our very limited set of findings (especially before they have been replicated in other studies) to develop hypotheses linking social relationships in the first few months after diagnosis to long-term survival. We need to know much more about the nature, quality and meaning of these social factors, and how they change during the intervening period, before attempting to understand how they may be linked to survival. Several authors have speculated about mechanisms that link social relationships to survival but none of the proposed mechanisms has been fully and adequately tested and
Breast cancer survival
the sparse body of data collected in this study prevent complete tests as well [16,19-211. However, to better understand the meaning of our findings and to provide more specific and detailed measures of social relationships for future survival studies, we have examined two sets of information. First, we looked at the correlations between the social factors that appear in the model and other measured social/demographic characteristics of the sampled women in order to clarify the meaning of significant factors. Second, we interviewed a sample of women who had participated in the original study [22]. Naturally we could see only those who survived, who were well, and who resided in the Greater Vancouver area. This sample can in no way be considered a representative sample; however, discussions with them using a very open guided-interview method provided rich descriptive information useful in understanding what our findings may actually reflect in women’s lives. From the interviews one aspect of friendship was highlighted: the choice of friends. It is possible to encourage friends who provide something useful and to discourage friends who do not; such choice is usually not possible among family or relatives. Further, friendship relationships are usually less encumbered with past and future obligations than are family ties and one sensed that women perceived the help of friends to be generally unambivalent. This resonates with Nealing and Winefield’s findings [4] that “friends provide less support of every kind than do families yet the reported satisfaction with support given by friends is greater than that from family members” [p. 3901. The fact, too, that women in our sample who had many contacts with, and more support from, friends were also significantly more likely to live in households in which they were the only adult, or one of two adults, suggests that women who are linked into a large social network of choice rather than of family obligation may be able to be more selective about, or in control of, the kind of relationships they have. Therefore, one dimension of social relationships that may be important to pursue in an effort to understand survival is the degree to which the woman has control over who helps, supports or provides information. Large friendship networks may, as well, allow for broader coverage of useful information, especially if these networks are less dense ones [23] in which members of the network are unacquainted with each other and can thus provide great variety of information and support. It is commonly reported that marriage has a positive relationship to good health [24]. We found, however, that the single, divorced, and widowed women were more likely to survive. Yet our findings on marital status echo observations by professionals who work with breast cancer patients who often describe complex patterns of ambivalent, and sometimes negative, relationships with spouse or family [25]. For example, in order to protect her immediate family, the woman suffering from breast cancer may not reveal to them the true extent of her fear or anxiety; consequently her spouse and other kin may not offer help or support when it would be useful or
181
may icsfiond to her in ways inappropriate to her real concerns. Many of the relationships married women reported are similar to those described by Dunkel-Schetter and Wortman [26]. Marriage, for women with breast cancer, may be important because it reflects the extent to which the woman is enmeshed in social relationships that, because of both past events and future expectations and obligations, may be ambivalent and over which she has relatively less choice or control than she has with friendships. Having a job, another predictor of survival, may reflect some aspects of choice and control as well. In our sample, women who work are significantly more likely to have a large family income which, in turn, may provide opportunity for the kind of choices (of treatment persons, of use of time, of help in the home) that lower income or unemployed women may not have. Women tend to work with other women, mainly in offices, hospitals and stores, and working women reported that they could seldom keep their illnesses a private matter; when one woman suffers from breast cancer her co-workers immediately find out and provide information (“I had it two years ago; do you want to ask any questions?‘) and access to others (“My neighbour had it; do you want her phone number?“). Employed women seemed to have much more offered to them while those not employed were left to seek out help from others. Help included information about good physicians, where to buy clothing suitable after mastectomy, alternative treatments, etc.. . . Among those who were not employed some chose to protect their own privacy and thus close off support while others did not actively seek help beyond their immediate family, relatives or neighbours. Working women, on the other hand, often became the centre of a large network of information and support. This ‘colleague network’ (perhaps not defined as ‘friends’ in the woman’s mind) may, as we suggested friends might be, provide less ambivalent, more controllable, and concretely useful relationships. The colleague network working women have may be one example of the “extended network” described by Hammer [21], a large reservoir of distant relationships from which new friends and helpers can be recruited as the smaller core network changes. Most of the literature linking social relationships to illness seems to assume that sick people are, and must be, the recipients of the giving, in other words, that they are dependent and in need of contact and support. Our interviews of survivors suggest a very different hypothesis. Many survivors talked of offering help and support to others; for example, while in the hospital after surgery one woman described walking down the hall to talk with another who had also had a mastectomy because the nurses said she was upset. Many more described seeking out other newly diagnosed women who might need information or advice. It appears from our small set of interviews that friendship networks and support may be important not only because they bring something, particularly concrete help and information, to the women with breast cancer but also because they indicate her active involvement in and initiation of relationships
NANCY WULER-MORRISON
182
Both Pearlin [27] and Jung [28] have observed that social support relationships are by definition interaction ones and that a full understanding of supportive interaction requires looking both at giver and receiver. Many survivors whom we interviewed seemed to be active givers. The findings reported here require replication, work that is currently underway in a prospective study of survival of men and women newly diagnosed with lung, breast, colorectal or prostate cancer. However, given the fact that theory linking social relationships to survival is in its infancy there is certainly need for data from studies that go beyond the usual prospective design using highly structured survey instruments before such a theory can be formulated. It will not be enough to rely on survey-type questionnaires administered at diagnosis and analysed for survival effects some years later. In the intervening period, now largely a ‘black box’, there are undoubtedly complex and changing patterns of social relationships in which the woman is actively involved. To develop a valid survival model in which social relationships may play a part, it will be important to examine that intervening period in detail using panel designs and qualitative methods, that is, listening to what women tell us about their lives. in that network.
Acknowledgements-Funding for this project was provided by the National Health and Research Development Program, Health and Welfare Canada, Grant No. 6610-1215-44 and by British Columbia Health Care Research Foundation. Grant Nos. 20 (79-2l and 16 (80-2l. We thank MS D. ~Skippen and Mrs J. Moody for data collection, and Mrs B. Craig for preparing the manuscript.
REFERENCES
1. Bloom J. Social support, accommodation to stress and adjustment to breast cancer. Sot. Sci. Med. 16, 1329-1338, 1982. 2. Bloom J. and Spiegel D. The relationshin of two dimensions of social support to the psychological well-being and social functioning of women with advanced breast cancer. Sot. Sci. Med. 19, 831-837, 1984. 3. Funch D. and Mettlin C. The role of support in relation to recovery from breast surgery. Sot. Sci. Med. 16, 91-92, 1982. 4. Nealing S. and Winefield H. Social support and recovery after surgery - . for breast cancer: Frequency and correlates of supportive behaviours by- fami ilv, friends and suraeon. Sot. Sci. Med. 27. 385-392. 1988. 5. Penman D. et al. The impact of mastectomy on selfconcept and social function: A combined cross sectional and longitudinal study with comparison groups. Women Hlth 11, 101-130, 1987. 6. Berkman L. and Syme S. L. Social networks, host resistance and mortality: A nine year follow-up study of Alameda County residents. Am. J. Epidemiol. 109, i 186-204, 1979. 7. Greer S., Morris T. and Pettingale K. W. Psychological response to breast cancer: effect on outcome. Z.uncet 13 Oct., 785-787, 1979. 8. Derogatis L., Abeloff M. and Melisaratos N. Psychological coping mechanisms and survival time in
et al.
metastatic cancer. /. Am. Med. Assoc. 242, 1504-1508, 1979. 9. Funch D. and Marshall J. The role of stress, social support and age in survival from breast cancer. J.-Psychosomatic Res. 27, 77-83. 1983. 10. Neale A. V.. Tillev B. and Vernon S. Marital status delay in seeking treatment and survival from breast cancer. Sot. Sci. Med. 23, 305-313, 1986. 11. Goodwin J. S., Hunt W. C., Key C. R. and Samet J. M. The effect of marital status on stage, treatment and survival of cancer patients. J. Am. Med. Assoc. 258, 312&3125. 1987. 12. Hislop T. G., Waxler N., Coldman A., Elwood J. M. and Kan L. The prognostic significance. of psychosocial factors in women with breast cancer. J. Chronic Dis. 40, 729-735, 1987. 13. Kaplan E. L. and Mcier Non-parametric estimation from incomplete observations. J. Am. Stork Assoc. 53, 45741, 1958. 14. Cox D. R. Regression models and life tables (with discussion). J. R. Stat. Sot. E 34, 187-220, 1972. IS. Pettingale K. W., Morris T., Greer S. and Haybittle J. L. Mental attitudes to cancer: An additional prognostic factor. Lancer 3 March, 750. 1985. 16. House J., Umberson D. and Landis K. R. Structures and prccesses of social support. Ann. Rev. Social. 14, 293-318, 1988. 17. House and Kahn R.L. Measures and J. concepts of social support. In Social Support and Health (Edited by Cohen S. and Syme S. L.), pp. 83-108. Academic Press, Orlando, 1985. 18. Orth-Gomer K. and Johnson J. Social network interaction and mortality. A six-year follow-up study of a random sample of the Swedish population. J. Chronic Dis. 49, 949-957, 1987. 19. Bloom J. Social support systems and cancer: A conceptual view. In Psy&oso&l Aspects of Cancer (Edited bv Cohen J.). DD. 129-149. Raven Press. New York. 1982. .. -20. Broadhead W. E. et al. The epidemiological evidence for a relationship between social support and health. Am. J. Epidemiol. 117, 521-537, 1983. 21. Hammer M. Core and extended social networks in relation to health and wellness. Sot. Sci. Med. 17, 405-411, 1983. 22. Mears B. Unpublished data. University of British Columbia, 1989. 23. Hall A. and Wellman B. Social networks and social support. In Social Support and Health (Cohen S. and Syme S. L.), pp. 23-41. Academic Press, Orlando, 1985. 24. Ortmeyer C. F. Variations in mortality, morbidity and health care by marital status. In Mortulity and Morbidity in the United States (Edited by Erhardt L. L. and Berlin J. E.), pp. 159-184. Harvard University Press, Harvard, 1974. 25. Revenson T. A., Wollman C. A. and Felton B. J. Social supports and stress. Pschyosomatic Med. 45, 321-331, 1983. 26. Dunkel-Schetter C. and Wortman C. The interpersonal dynamics of cancer. Problems in social relationships and their impact on cancer. In Interpersonal Issues in Health Cure (Edited by Friedman H. S. and DiMatteo M. R.), pp. 69-100. Academic Press, New York, 1982. L. Social structure and processes of 27. Pearlin social support. In Social Support and Health (Edited bv Cohen S. and Svme S. L.). DD. 43-60. Academic P&as, Orlando, 1985. ‘. - 28. Jung J. Social support and its relation to health: A critical evaluation. Basic appl. Sot. Psycho!. 5, 143-169. 1984.
183
Breast cancer survival APPENDIX Wording of Questions and Coding Categories for Demographic and Social Factors Education of patient: “How much schooling have you had?’
Code: 1 Patient lives 2 With others, 3 With others, 4 With others,
Number of supportive relatives: you
presently
alone all adult adults and children children onIy
Marital status: “What is your present marital status?” Code: 1 Married 2 Single, widowed, separated/divorced Work: “What is your present occupation?” Code: 1 Currently working 2 Not working Contact with friends: “How often do you visit or talk on the phone with your friends?’ Code: 1 2 3 4
Code: Number of friends: 0 = none 9-9 or more
Code: 1 O-11years 2 12 years 3 more than I2 Degree of responsibility for others: “Who do live with?”
Number of supportive friends: “In the event of domestic or emotional problems or other stressful stituations do you have any friends on whom you could call for support or help?’
At least once a day At least once a week At least once a month Less often
Contact with relatives: Same as contact with friends Total contact with friends and relatives: Combine frequency for friends and relatives
Same as supportive friends Total number of support friends, relatives and neighbours: Combine number of persons for friends, relatives and neighbours Social network: Uses data from Marital Status and Total Contact, plus “To which church or faith do you belong?” Code: Marital status: 1 Married 0 Other Total contact with friends and relatives:
1 Above sample median 0 Below sample median Membership in faith: “To which church or faith do you belong?’ 1 Mentions name of faith 0 No name mentioned Score for social network is obtained by summing the above three factors to obtain: 0 Low social network score l-2 Intermediate social network score 3 High social network score.
0277-9536191 S3.00+0.00 Copyright 0 1991 Pergamon Press plc
AI1rights reserved
EFFECTS OF SOCIAL RELATIONSHIPS ON SURVIVAL FOR WOMEN WITH BREAST CANCER: A PROSPECTIVE STUDY NANCY
WAXLER-MORRISON’, T. GREGORY H~~LoP~, BRONWEN MEARS’ and LISA KAN*
‘Department of Anthropology/Sociology, University of British Columbia, Vancouver and ‘Division of Epidemiology, Biometry and Occupational Oncology, British Columbia Cancer Agency, Vancouver, Canada Abstract-In this study we examine the relationship between a woman’s social contexts at the time of diagnosis and her chances of having survived breast cancer four years later. A cohort of 133 women were followed prospectively after initial diagnosis and treatment and data were obtained from a questionnaire mailed soon after diagnosis and from hospital charts. Using multivariate methods to examine predictors of survival, two clinical factors, pathologic nodal status and clinical stage of disease, were significantly associated with survival. In addition we found significant and independent effects on survival ofz number of supportive friends, number of supportive persons, whether the woman worked, whether she was unmarried, the extent of contact with friends and the size of her social network. Thus, the woman’s social context, particularly contexts of friendship and work outside the home, are statistically important for survival. Using existing literature, further data analyses and interviews with some survivors, we speculate on the ways in which social contexts may influence survival and suggest research methods suitable to this question. Key words-breast
cancer, social networks, social support, survival.
INTRODUCTION
Many researchers have attempted to link women’s social and psychological characteristics to their ability to adjust to the distress of breast cancer [l-5]. But it was not until Berkman and Syme [6] reported that size of one’s social network was predictive of mortality in a community sample that some began to ask whether the women’s social contexts or psychological traits might tell us something about their chances for survival after breast cancer had been diagnosed. Breast cancer prognostic studies have generally followed several different theoretical perspectives and used a variety of analytic methods. One set uses psychological theory to understand survival, with inconsistent results. Both Greer [A and Derogatis [8] focused mainly on the women’s psychological methods of coping with the initial illness as a predictor of short- or long-term survival. Greer reported that short-term survivors tended to be those who revealed, at diagnosis or during the early stages of treatment, helplessness or stoic acceptance. Derogatis found that women who survived were more anxious and hostile at diagnosis than were short-term survivors; long-term survivors were, from the point of view of health professionals, also less well adjusted to their illnesses and less cooperative with the doctors. Others have used sociological theory, sometimes implicitly, in an attempt to link characteristics of the woman’s social relationships or social network to her survival and have again reported mixed findings. Funch and Marshall [9] showed that a death, an illness or unemployment in the breast cancer patient’s family predicts poorer survival just as did the patient’s perception that family income was in-
adequate. Her membership in formal organizations, however, increased chances of survival. Neale (lo] reported that being married, as opposed to being a widow, is protective. However, Goodwin [I l] has shown that when those women who are currently married at the time of breast cancer diagnosis are compared with all others (single, divorced, separated and widowed) the relative risk of death is the same for each group. Many studies have one or more methodological limitations, particularly lack of multivariate analytic methods, lack of statistical control for clinical variables, and, most important, relatively poor measures of women’s social characteristics. Our earlier report [12] dealt with some of these methodological problems and examined some psychological and social predictors of survival. After controlling for clinical factors, three psychosocial factors were significant for survival: the extent to which the woman, at the time of diagnosis, engaged in expressive activities in the home (e.g. social activities); her level of extroversion; and the extent to which she expressed/felt anger. In that initial report the range of social and psychological characteristics investigated was very narrow. In this paper we have expanded our analysis to reflect a sociological perspective, one suggesting that a woman’s social role relationships may play a part in her response to breast cancer and ultimately in her survival. We have thus examined the effect of the woman’s social network, her relationships with family and friends, and her work experience, on survival. In addition, using interviews of a small group of longterm survivors, we have speculated on the meaning of these factors in women’s lives.
177
NANCY WAXLER-MORRISON et
178 METHODS
The study group comprised women who were referred to the A. Maxwell Evans Clinic (AMEC) in Vancouver, who had a confirmed diagnosis of primary ductal breast cancer between June 1980 and May 1981, and who were under 55 years of age at diagnosis (in order to examine mainly premenopausal disease). One hundred and thirty-three of the 168 eligible patients returned both study questionnaires (u 80% response); the clinical characteristics of responders and nonresponders were similar [12]. AMEC is the only facility where cancer patients can receive radiotherapy on the mainland of British Columbia; 2/3 of all breast cancer patients on the mainland are seen here. The type of initial treatment was dependent upon the clinical stage and site of the tumour [12]. Most patients received modified radical mastectomy. Referrals to AMEC usually occurred shortly after diagnosis. These patients were mailed a self-adrninistered questionnaire to obtain demographic and social network information, followed by a second selfadministered questionnaire to obtain the additional psychosocial information reported in Hislop [12]. Ninety percent of the returned questionnaires were completed within 3 months of diagnosis. Sources and scoring methods for our earlier analysis of psychosocial factors are described elsewhere [12]. Additional sociological data were obtained from the initial questionnaire on educational level; the extent to which she was responsible for others in the home; the extent of her contact with friends and/or relatives; her perception of help/support available from relatives, neighbours, and friends; employment status; marital status; and a measure of social network derived from Berkman and Syme [6] that combines marital status, contact with friends/ relatives, and church membership. Specific questions and scoring methods for these additional social factors are in the Appendix. Breakpoints for ordinal factors (e.g. numbers of supportive friends) which had more than 4 levels were based on frequency distributions in the total study group; breakpoints for the earlier psychosocial factors remained the same as in the original analysis [12]. Clinical factors were abstracted from the patient’s medical record by a trained clerk using a standardized form; the clerk was unaware of the questionnaire responses. The clinical factors included age at diagnosis, clinical stage, pathologic nodal status, histological tumour grade, estrogen receptor (ER) tumour status, type of initial treatment, time to first recurrence of disease (local, regional or metastatic), survival and disease status when last seen, and time to last follow-up or death. Recurrence and survival data was abstracted in January, 1985. Survival was calculated from the time of diagnosis using the product-limit method of Kaplan and Meier [ 131.Observed differences in survival for single factors were examined using the log-rank statistic. The Cox proportional hazards analysis [ 141 was used to determine the maximum likelihood estimates of regression coefficients for each psychosocial factor separately, adjusting for the clinical prognostic factors; the Pvalue indicates the statistical difference between the
al.
levels for a given factor and is not a test for trend. This method of analysis was also used to test for independent effects by examining the significant social and psychological factors together, again adjusting for clinical factors. For analyses examining each social factor separately, patients were excluded with missing information for the factor being analyzed. For analyses examining independent effects, patients were excluded with missing information for any of the factors being compared. Fifteen patients were excluded from this latter analysis because of missing information; hence the final multivariate analysis was restricted to 118 patients. Finally, in order to better understand the importance of social relationships, 18 surviving patients from the original study group who lived in the Greater Vancouver area were interviewed in 1986. Open guided interviews (phenomenological) were used to obtain rich descriptive information from the patients. FINDINGS
Overall survival In the 4 years of follow-up there were 26 deaths (25 attributed to breast cancer) reported for the study group. Over 90% of the patients were followed to at least 2 years, 80% to 3 years and approximately 30% to 4 years. The survival rates were 96%, 89%, 80% and 76% at 1, 2, 3 and 4 years, respectively. Clinical factors All clinical factors were related to survival in the expected direction, however, only two were significantly related to survival: pathological nodal status and clinical stage (Table 1). Social factors The Cox proportional hazards method was used to calculate estimates of the instantaneous relative death rate for each new social factor separately after including the two significant clinical factors (pathological nodal status and clinical stage). Six of the eleven measures of social relationships were significantly associated with survival: marital status; support from Table 1. Percent survival bv clinical factors Survival P-value.
Factor
Age
Clinical stage
Pathological nodal status
Histologic grade Estrogen receptor status
Level
n
2 Yr
4 Yr
log rani;
I I people Work: Employed Not emnloycd Social network size Small Medium Large
I
Estimate*
101
1.00
17
0.17
46
1.00
:;
0.61
54 48 14 2
P-value
O.CGO4
0.0001
1.00
I .29 3.14 14.29
0.0005
:: 29
1.00 1.27 1.01 0.65
0.0802
62 56
1s2
40 61 17
1.00 0.86 1.59
28
1.00 0.0003
0.0006
friends; contact with friends; total support (friends, relatives, neighbours); employment status; and social network. Five measures (responsibility for others; educational level; support from relatives; contact with relatives; and total contact with friends and relatives) were not significantly associated with survival. From our earlier analysis three psychosocial factors had been associated with survival after adjustment for clinical factors: frequency of involvement in social activities in the home, degree of extroversion and degree of anger [12]. We then repeated the Cox proportional hazards analysis to identify independent factors, limiting the sample to the 118 with no missing information for any of the factors being compared. The six significant social factors and the three significant psychosocial factors from our earlier analysis were examined after including the clinical factors, pathological nodal status and clinical stage. (The distribution of the three psychosocial factors, and the rate of survival, in this group of 118 women were similar to that in our earlier report.) Table 2 shows the instantaneous relative death rate for the significant independent factors. (Rates above 1.OOindicate decreased survival times, whereas rates below 1.00 indicate increased survival times; the greater the difference from 1.00, the stronger the effect upon survival.) The six social factors remained significant whereas the psychosocial factors were no longer significant. The extent to which a woman feels that she can call upon three or more friends for support or help was most strongly associated with survival. Correlations between our earlier psychosocial factors and new social factors
With the loss of significance for our earlier psychosocial factors, we examined the relationship between 3312-F
these three factors and the significant social factors. The ‘expressive activities at home’ factor was significantly correlated in the expected direction with ‘contact with friends’ (Speannan p = 0.32; P = 0.006), ‘support from friends’ (p = 0.24; P = 0.05). and ‘social network’ (p = 0.32; P = 0.002). Thus ‘expressive activities at home’ seems to reflect in a concrete and limited way the support and contact dimensions of friendship. The ‘extroversion’ factor was significantly correlated with ‘contact with friends’ (p = 0.25; P = 0.01) and ‘support from friends’ (p = 0.30; P c O.OOl), suggesting again that this personality trait reflects several dimensions of the friendship relationship. The third earlier factor, ‘anger’, was not significantly related to any social factors in our model. We examined also for correlation between the significant social factors and other social or demographic factors; only ‘marital status’ showed significant relationships, with both ‘work’ (p = 0.19; P = 0.01) and ‘education’ @ = 0.17; P = 0.03). Interviews of a sample of survivors
*The smaller the estimates (that are less than 1.00) the greater the likelihood of survival, whereas the greater the estimates (that are greater than 1.00) the less the likelihood of survival.
SW
179
In order to better understand the significance of social relationships in long-term survivors, 18 patients were interviewed by the phenomenological method. Our interviews gave some hints as to the dimensions of friendship that may be important. Many survivors mentioned the practical, often timethings that friends offered such as consuming, cooking, child-care and shopping. Having a friend to provide transportation twice a week to the hospital, who waited in the corridor, was the kind of commitment that many women found important. ‘Having a shoulder to cry on’ was mentioned, but seemed secondary to concrete help. There was also the observation that, during the crisis of diagnosis and treatment, ‘you discovered who your real friends are’. Some women lost friends, either temporarily or permanently, apparently because of the friend’s inability to tolerate life-threatening illness. Some women found new friends, often women who had had breast cancer in the past who offered help; sometimes new friendships were established with women who were in the hospital at the same time. Most of the married survivors we interviewed reported very supportive and helpful responses from their spouses in the initial treatment period. The women themselves and their spouses usually recognized that breast loss was a minor issue compared with the threat to life that cancer posed. This threat to life was, in some instances, though, enough to break up a marriage when the woman’s husband could not tolerate a wife with cancer. Married survivors tended to report complex patterns of support and protection between family members which were often experienced by the women as ambivalent and sometimes very stressful. It was the young adult children that women felt to be least supportive, most difficult and unhelpful to them. Many women ‘had to support them by being cheerful’ and ‘carrying on as if things were normal’ since children of this age appeared to be fearful for their mothers’ lives; some children simply stayed away during the most stressful periods.
NANCY WAXLER-MORRWN er al.
180
Many surviving women downplayed the financial aspects of work, stating that they could have managed financially without work but continued working either as volunteers or for pay, in part, because ‘work took their mind off illness’. However, it was clear from these women that jobs were a source of support, were symbolic of ‘getting back to living’, and more important, provided useful information. DISCUSSION
While there is in the research literature a large body of work linking women’s social and psychological characteristics to their adjustment to breast cancer over the short-term there are relatively few studies attempting to relate these factors to their survival [7,11,15]. The current study combines a number of methodological features that makes it relatively strong. The design is prospective, the sample size is relatively large, social factors were measured directly from the patient rather than from registries or hospital charts, and multivariate analytic methods were used. There are, however, methodological limitations of the study. First, women over age 55 (postmenopausal) were not included in the sample, thus omitting women in the higher incidence age cohorts; this prevented us from understanding how some social factors, particularly widowhood, may relate to survival. As well, the mailed questionnaire method necessitated obtaining limited information and thus leaves us with many unanswered questions. Measures of social network and social support were those available from the literature in 1980; both methodological and theoretical advances have been made since then [16-181. All social factors were measured shortly after diagnosis and surgical treatment (usually mastectomy), hence these factors are likely to be reflections of the women’s social situation during the crisis period. We do not know whether social relationships during this crisis period are in some ways consistent with the women’s earlier social relationships, although we would assume that they are. Finally, we have no information on social relationships during the intervening follow-up years and thus cannot document how these may effect survival. Our earlier study (121 was based on a limited number of social and psychological variables which pointed to three significant correlates of survival after clinical factors had been controlled: expressive (emotional, relational) activities in the home, extroversion and anger. With the addition of social support and network factors in the current analysis, these initial factors were no longer significant, no doubt because they are correlated with the new social factors in our model. What we have found in the current analysis is that several characteristics of the women’s friendship and social network, her marital status and employment status are significantly and independently related to survival. There is little evidence in our findings that relationships within the woman’s family and with her relatives are important for survival. Instead, the social contexts that appear to be significant are outside the woman’s home. Friendship
and work are the themes that seem to underlie the findings. Previous research on breast cancer survival is not very helpful in understanding our findings. Both Greer [I and Derogatis [8] focused mainly on the woman’s psychological methods of coping with the initial illness as predictors of survival and reported shorter survival for those who revealed at diagnosis or during the early stages of treatment, low levels of hostility, high levels of positive mood, stoic acceptance or hopelessness. However, Derogatis’ findings that the treating physician rated long-term survivors as significantly less adjusted to their illness, possessing more negative attitudes to the illness and its treatment, and that the interviewer judged the longterm survivors as having “significantly poorer attitudes toward their physicians*’ [p. 15061 may be important for hypothesis construction. Greer [A does show that an undescribed measure of “interpersonal relationships” was not related to survival, but that being unmarried or reporting poor marital relationships at the time of diagnosis was related to a less favourable outcome. Funch and Marshall [9] found that a measure of “number of relatives and friends” was not a significantly related to survival. However, they report that a factor called “organizational involvement” (unfortunately not further defined) was significant for survival and independent of the significant effects of other stresses (e.g. death in the family). Neale et al. [lo] found that married rather than widowed women survive, even when age is controlled; unfortunately, no divorced, single or separated women were included in the sample, thus preventing any direct comparison with the data reported here. Other investigators have found no significant re.lationship between marital status and survival in breast cancer patients [9, Ill. These studies have not measured directly and comprehensively characteristics of the women’s social network, her social support or her work experience and thus it is difficult to use this research to examine on our own findings. The numerous studies examining the effects of social support, social networks, etc. on udjusttnenr (as indicated by coping methods, psychological distress, etc.) to breast cancer [l-5] might be thought of as a rich source of hypotheses linking social factors to survival, but we are wary of making that link since at least one researcher [8] has shown that the woman who is thought by her physician to be less adjusted is more likely to survive. With the exception of Derogatis’ work, the strength and direction of the relationship between adjustment to having breast cancer and survival has, as far as we know, not been documented. It is premature, as well, to use our very limited set of findings (especially before they have been replicated in other studies) to develop hypotheses linking social relationships in the first few months after diagnosis to long-term survival. We need to know much more about the nature, quality and meaning of these social factors, and how they change during the intervening period, before attempting to understand how they may be linked to survival. Several authors have speculated about mechanisms that link social relationships to survival but none of the proposed mechanisms has been fully and adequately tested and
Breast cancer survival
the sparse body of data collected in this study prevent complete tests as well [16,19-211. However, to better understand the meaning of our findings and to provide more specific and detailed measures of social relationships for future survival studies, we have examined two sets of information. First, we looked at the correlations between the social factors that appear in the model and other measured social/demographic characteristics of the sampled women in order to clarify the meaning of significant factors. Second, we interviewed a sample of women who had participated in the original study [22]. Naturally we could see only those who survived, who were well, and who resided in the Greater Vancouver area. This sample can in no way be considered a representative sample; however, discussions with them using a very open guided-interview method provided rich descriptive information useful in understanding what our findings may actually reflect in women’s lives. From the interviews one aspect of friendship was highlighted: the choice of friends. It is possible to encourage friends who provide something useful and to discourage friends who do not; such choice is usually not possible among family or relatives. Further, friendship relationships are usually less encumbered with past and future obligations than are family ties and one sensed that women perceived the help of friends to be generally unambivalent. This resonates with Nealing and Winefield’s findings [4] that “friends provide less support of every kind than do families yet the reported satisfaction with support given by friends is greater than that from family members” [p. 3901. The fact, too, that women in our sample who had many contacts with, and more support from, friends were also significantly more likely to live in households in which they were the only adult, or one of two adults, suggests that women who are linked into a large social network of choice rather than of family obligation may be able to be more selective about, or in control of, the kind of relationships they have. Therefore, one dimension of social relationships that may be important to pursue in an effort to understand survival is the degree to which the woman has control over who helps, supports or provides information. Large friendship networks may, as well, allow for broader coverage of useful information, especially if these networks are less dense ones [23] in which members of the network are unacquainted with each other and can thus provide great variety of information and support. It is commonly reported that marriage has a positive relationship to good health [24]. We found, however, that the single, divorced, and widowed women were more likely to survive. Yet our findings on marital status echo observations by professionals who work with breast cancer patients who often describe complex patterns of ambivalent, and sometimes negative, relationships with spouse or family [25]. For example, in order to protect her immediate family, the woman suffering from breast cancer may not reveal to them the true extent of her fear or anxiety; consequently her spouse and other kin may not offer help or support when it would be useful or
181
may icsfiond to her in ways inappropriate to her real concerns. Many of the relationships married women reported are similar to those described by Dunkel-Schetter and Wortman [26]. Marriage, for women with breast cancer, may be important because it reflects the extent to which the woman is enmeshed in social relationships that, because of both past events and future expectations and obligations, may be ambivalent and over which she has relatively less choice or control than she has with friendships. Having a job, another predictor of survival, may reflect some aspects of choice and control as well. In our sample, women who work are significantly more likely to have a large family income which, in turn, may provide opportunity for the kind of choices (of treatment persons, of use of time, of help in the home) that lower income or unemployed women may not have. Women tend to work with other women, mainly in offices, hospitals and stores, and working women reported that they could seldom keep their illnesses a private matter; when one woman suffers from breast cancer her co-workers immediately find out and provide information (“I had it two years ago; do you want to ask any questions?‘) and access to others (“My neighbour had it; do you want her phone number?“). Employed women seemed to have much more offered to them while those not employed were left to seek out help from others. Help included information about good physicians, where to buy clothing suitable after mastectomy, alternative treatments, etc.. . . Among those who were not employed some chose to protect their own privacy and thus close off support while others did not actively seek help beyond their immediate family, relatives or neighbours. Working women, on the other hand, often became the centre of a large network of information and support. This ‘colleague network’ (perhaps not defined as ‘friends’ in the woman’s mind) may, as we suggested friends might be, provide less ambivalent, more controllable, and concretely useful relationships. The colleague network working women have may be one example of the “extended network” described by Hammer [21], a large reservoir of distant relationships from which new friends and helpers can be recruited as the smaller core network changes. Most of the literature linking social relationships to illness seems to assume that sick people are, and must be, the recipients of the giving, in other words, that they are dependent and in need of contact and support. Our interviews of survivors suggest a very different hypothesis. Many survivors talked of offering help and support to others; for example, while in the hospital after surgery one woman described walking down the hall to talk with another who had also had a mastectomy because the nurses said she was upset. Many more described seeking out other newly diagnosed women who might need information or advice. It appears from our small set of interviews that friendship networks and support may be important not only because they bring something, particularly concrete help and information, to the women with breast cancer but also because they indicate her active involvement in and initiation of relationships
NANCY WULER-MORRISON
182
Both Pearlin [27] and Jung [28] have observed that social support relationships are by definition interaction ones and that a full understanding of supportive interaction requires looking both at giver and receiver. Many survivors whom we interviewed seemed to be active givers. The findings reported here require replication, work that is currently underway in a prospective study of survival of men and women newly diagnosed with lung, breast, colorectal or prostate cancer. However, given the fact that theory linking social relationships to survival is in its infancy there is certainly need for data from studies that go beyond the usual prospective design using highly structured survey instruments before such a theory can be formulated. It will not be enough to rely on survey-type questionnaires administered at diagnosis and analysed for survival effects some years later. In the intervening period, now largely a ‘black box’, there are undoubtedly complex and changing patterns of social relationships in which the woman is actively involved. To develop a valid survival model in which social relationships may play a part, it will be important to examine that intervening period in detail using panel designs and qualitative methods, that is, listening to what women tell us about their lives. in that network.
Acknowledgements-Funding for this project was provided by the National Health and Research Development Program, Health and Welfare Canada, Grant No. 6610-1215-44 and by British Columbia Health Care Research Foundation. Grant Nos. 20 (79-2l and 16 (80-2l. We thank MS D. ~Skippen and Mrs J. Moody for data collection, and Mrs B. Craig for preparing the manuscript.
REFERENCES
1. Bloom J. Social support, accommodation to stress and adjustment to breast cancer. Sot. Sci. Med. 16, 1329-1338, 1982. 2. Bloom J. and Spiegel D. The relationshin of two dimensions of social support to the psychological well-being and social functioning of women with advanced breast cancer. Sot. Sci. Med. 19, 831-837, 1984. 3. Funch D. and Mettlin C. The role of support in relation to recovery from breast surgery. Sot. Sci. Med. 16, 91-92, 1982. 4. Nealing S. and Winefield H. Social support and recovery after surgery - . for breast cancer: Frequency and correlates of supportive behaviours by- fami ilv, friends and suraeon. Sot. Sci. Med. 27. 385-392. 1988. 5. Penman D. et al. The impact of mastectomy on selfconcept and social function: A combined cross sectional and longitudinal study with comparison groups. Women Hlth 11, 101-130, 1987. 6. Berkman L. and Syme S. L. Social networks, host resistance and mortality: A nine year follow-up study of Alameda County residents. Am. J. Epidemiol. 109, i 186-204, 1979. 7. Greer S., Morris T. and Pettingale K. W. Psychological response to breast cancer: effect on outcome. Z.uncet 13 Oct., 785-787, 1979. 8. Derogatis L., Abeloff M. and Melisaratos N. Psychological coping mechanisms and survival time in
et al.
metastatic cancer. /. Am. Med. Assoc. 242, 1504-1508, 1979. 9. Funch D. and Marshall J. The role of stress, social support and age in survival from breast cancer. J.-Psychosomatic Res. 27, 77-83. 1983. 10. Neale A. V.. Tillev B. and Vernon S. Marital status delay in seeking treatment and survival from breast cancer. Sot. Sci. Med. 23, 305-313, 1986. 11. Goodwin J. S., Hunt W. C., Key C. R. and Samet J. M. The effect of marital status on stage, treatment and survival of cancer patients. J. Am. Med. Assoc. 258, 312&3125. 1987. 12. Hislop T. G., Waxler N., Coldman A., Elwood J. M. and Kan L. The prognostic significance. of psychosocial factors in women with breast cancer. J. Chronic Dis. 40, 729-735, 1987. 13. Kaplan E. L. and Mcier Non-parametric estimation from incomplete observations. J. Am. Stork Assoc. 53, 45741, 1958. 14. Cox D. R. Regression models and life tables (with discussion). J. R. Stat. Sot. E 34, 187-220, 1972. IS. Pettingale K. W., Morris T., Greer S. and Haybittle J. L. Mental attitudes to cancer: An additional prognostic factor. Lancer 3 March, 750. 1985. 16. House J., Umberson D. and Landis K. R. Structures and prccesses of social support. Ann. Rev. Social. 14, 293-318, 1988. 17. House and Kahn R.L. Measures and J. concepts of social support. In Social Support and Health (Edited by Cohen S. and Syme S. L.), pp. 83-108. Academic Press, Orlando, 1985. 18. Orth-Gomer K. and Johnson J. Social network interaction and mortality. A six-year follow-up study of a random sample of the Swedish population. J. Chronic Dis. 49, 949-957, 1987. 19. Bloom J. Social support systems and cancer: A conceptual view. In Psy&oso&l Aspects of Cancer (Edited bv Cohen J.). DD. 129-149. Raven Press. New York. 1982. .. -20. Broadhead W. E. et al. The epidemiological evidence for a relationship between social support and health. Am. J. Epidemiol. 117, 521-537, 1983. 21. Hammer M. Core and extended social networks in relation to health and wellness. Sot. Sci. Med. 17, 405-411, 1983. 22. Mears B. Unpublished data. University of British Columbia, 1989. 23. Hall A. and Wellman B. Social networks and social support. In Social Support and Health (Cohen S. and Syme S. L.), pp. 23-41. Academic Press, Orlando, 1985. 24. Ortmeyer C. F. Variations in mortality, morbidity and health care by marital status. In Mortulity and Morbidity in the United States (Edited by Erhardt L. L. and Berlin J. E.), pp. 159-184. Harvard University Press, Harvard, 1974. 25. Revenson T. A., Wollman C. A. and Felton B. J. Social supports and stress. Pschyosomatic Med. 45, 321-331, 1983. 26. Dunkel-Schetter C. and Wortman C. The interpersonal dynamics of cancer. Problems in social relationships and their impact on cancer. In Interpersonal Issues in Health Cure (Edited by Friedman H. S. and DiMatteo M. R.), pp. 69-100. Academic Press, New York, 1982. L. Social structure and processes of 27. Pearlin social support. In Social Support and Health (Edited bv Cohen S. and Svme S. L.). DD. 43-60. Academic P&as, Orlando, 1985. ‘. - 28. Jung J. Social support and its relation to health: A critical evaluation. Basic appl. Sot. Psycho!. 5, 143-169. 1984.
183
Breast cancer survival APPENDIX Wording of Questions and Coding Categories for Demographic and Social Factors Education of patient: “How much schooling have you had?’
Code: 1 Patient lives 2 With others, 3 With others, 4 With others,
Number of supportive relatives: you
presently
alone all adult adults and children children onIy
Marital status: “What is your present marital status?” Code: 1 Married 2 Single, widowed, separated/divorced Work: “What is your present occupation?” Code: 1 Currently working 2 Not working Contact with friends: “How often do you visit or talk on the phone with your friends?’ Code: 1 2 3 4
Code: Number of friends: 0 = none 9-9 or more
Code: 1 O-11years 2 12 years 3 more than I2 Degree of responsibility for others: “Who do live with?”
Number of supportive friends: “In the event of domestic or emotional problems or other stressful stituations do you have any friends on whom you could call for support or help?’
At least once a day At least once a week At least once a month Less often
Contact with relatives: Same as contact with friends Total contact with friends and relatives: Combine frequency for friends and relatives
Same as supportive friends Total number of support friends, relatives and neighbours: Combine number of persons for friends, relatives and neighbours Social network: Uses data from Marital Status and Total Contact, plus “To which church or faith do you belong?” Code: Marital status: 1 Married 0 Other Total contact with friends and relatives:
1 Above sample median 0 Below sample median Membership in faith: “To which church or faith do you belong?’ 1 Mentions name of faith 0 No name mentioned Score for social network is obtained by summing the above three factors to obtain: 0 Low social network score l-2 Intermediate social network score 3 High social network score.
