The impact of a second breast cancer diagnosis on health related quality of life

Transcription

1 Breast Cancer Research and Treatment (2005) 92: Ó Springer 2005 DOI /s Report The impact of a second breast cancer diagnosis on health related quality of life Andrea A. Thornton 1, Lisa Madlensky 2, Shirley W. Flatt 2, Robert M. Kaplan 3, John P. Pierce 2, for the Women s Healthy Eating and Living (WHEL) Study Group 1 Department of Supportive Care, Pain, and Palliative Medicine, City of Hope National Medical Center, Duarte, California, USA; 2 Cancer Prevention and Control Program, Cancer Center, University of California, San Diego, CA, USA; 3 Department of Health Services, UCLA School of Public Health, University of California, Los Angeles, USA Key words: breast cancer, new primary, psychosocial adjustment, quality of life, recurrence, response bias Summary A growing body of research documents the negative impact of recurrent breast cancer on quality of life (QOL), however few prospective investigations are available that compare QOL outcomes across different types of second cancer events (e.g., local/regional recurrence (LR) versus distant/metastatic recurrence (DR) versus a new primary breast cancer (NP)). In addition, although participant attrition is a major issue in this group of cancer patients, the potential impact of attrition or response bias on QOL outcomes is not typically examined. To address these issues, we prospectively examined QOL data obtained from a sample of women (n=140) participating in the Women s Healthy Eating and Living (WHEL) Study who experienced a second cancer event (LR, DR or NP). Women in our study reported significant worsening in most areas of QOL, except for emotional wellbeing, from pre- to postsecond cancer event. Although the patterns of change to QOL differed slightly within each category (LR versus DR versus NP), participants were more similar than different in their QOL ratings. However, the perceived health status of women with DR was significantly lower than women with LR or NP. We also identify important sources of response bias that should be considered when interpreting findings. Specifically, women diagnosed with DR (a) were more likely to decline to complete the post-recurrence QOL questionnaire (27% refusal rate versus 14% LR and 13% NP), and (b) had higher death rates (28% death rate versus 4% LR and 6% NP) after agreeing to but before completing the post-second diagnosis QOL questionnaire. Introduction Breast cancer is the most commonly diagnosed invasive cancer among women in the U.S., and it is the second leading cause of cancer deaths [1]. Although advances in detection and treatment have improved survival rates, the threat of a second cancer diagnosis is an ongoing and significant issue for breast cancer survivors [2,3]. Recurrence rates for breast cancer vary considerably according to features of the disease, its treatment, and the patient, but may be as high as 40 45% in some subgroups up to 15 years after initial treatment [4 6]. Given that a substantial proportion of women diagnosed with breast cancer will experience disease recurrence or be diagnosed with a new primary, it is important to understand the implications of these diagnoses for health-related quality of life. Several cross-sectional studies suggest that cancer survivors who experience a relapse report worse quality of life compared to those who remain disease-free [7 9]. In addition, three prospective investigations have reported on the changes in quality of life that accompany breast cancer recurrence [10 13]. Both Bull et al. [10] and Hurny et al. [11] reported that most quality of life domains worsened significantly following recurrence. In the former study, improvements also were noted in several areas of quality of life at follow-up (6 months after recurrence); nonetheless, functioning remained impaired in most domains compared to pre-recurrence [10]. Oh and colleagues also found significant declines in physical functioning, general health, and social functioning after recurrence, however no significant declines to other quality of life domains were identified, including those measuring mental health/emotional wellbeing, or depression [12,13]. However, compared to a matched sample of disease-free survivors, women with recurrent disease endorsed poorer functioning across most quality of life domains. Furthermore, the relative impairments in health related quality of life in women with recurrent disease were largely attributable to the poorer functioning of the subgroup of women diagnosed with distant or metastatic breast cancer, with no differences in quality of life being noted between women who remained diseasefree and women treated for a local recurrence. One of the issues that has not been well-elucidated in the literature is the impact of participant attrition and

2 26 AA Thornton et al. response bias on quality of life outcomes in survivors diagnosed with cancer for a second time (recurrence or a new primary). That is, are these participants representative of the larger population of women who experience a recurrence or new primary, or do they differ in significant ways from nonparticipants? As pointed out by Oh et al. [12], women with distant recurrence may experience worse quality of life outcomes than women with a local breast cancer recurrence. In addition, higher levels of symptom distress and poorer physical functioning may make it less likely for women with recurrent disease to agree to participate in quality of life research by virtue of their physical limitations and/or distress levels. In addition, women with distant recurrence who also progress rapidly to the terminal phase may not have sufficient opportunity to participate in quality of life research before death [10]. Thus, existing quality of life research in the recurrence arena may be biased in favor of women who experience fewer limitations to health related quality of life. To our knowledge, the question of whether response rates or attrition vary according to the type of diagnosis that a patient receives (local recurrence, distant/metastatic recurrence, or new primary) has not been well examined. Given that prospective data regarding the impact of receiving a second cancer diagnosis are difficult to obtain, this gap in the literature is not surprising. However, the nature of our dataset allowed us to also shed light on this issue. We undertook the current investigation to clarify the impact of being diagnosed with breast cancer for a second time on health-related quality of life. The goals of the present study were as follows: (1) to prospectively examine changes in health related quality of life associated with a second diagnosis of breast cancer, (2) to compare health related quality of life across categories of second diagnosis (local/regional recurrence, distant/ metastatic recurrence, and new primary), and (3) to delineate participant attrition and response bias in this sample. Although the diagnosis of a new primary does not necessarily represent recurrence, this group of patients has been virtually ignored in the quality of life literature. Thus, we include patients diagnosed with a local/regional recurrence, a distant/metastatic recurrence, or a new primary in order to empirically evaluate potential differences in health related quality of life among these three categories of second diagnoses. Methods The data presented here were gathered in the context of a large, randomized controlled trial investigating the effect of a plant-based dietary pattern on additional breast cancer events and survival [14]. For the current report we used basic demographic, medical, and quality of life data for the subset of study participants who experienced a medically confirmed second breast cancer event between their enrollment into the study and prior to January 1, 2003, when these data were analyzed. Key features of the WHEL Study are described below. Participants The WHEL Study enrolled 3088 breast cancer survivors (see Pierce et al. [14] for a complete description of the WHEL Study methodology, including participant recruitment, enrollment, and randomization). Eligibility criteria for the WHEL Study included: (1) years of age at initial breast cancer diagnosis; (2) diagnosis of early stage invasive breast cancer categorized as Stage I ( 1 cm), II or IIIA, within the past 4 years [15]; (3) completed treatment consisting of either total mastectomy and axillary dissection, or breast-sparing surgery with radiation; (4) completed chemotherapy regimen, if indicated; (5) no evidence of recurrent disease or new breast cancer since completing initial treatment (confirmed by physician); (6) accessible geographically and by telephone; and (7) ability to communicate dietary data via a 24-hour food recall. Exclusion criteria included the following: (1) objective evidence of recurrent disease; (2) current enrollment in another dietary clinical trial; (3) a comorbid diagnosis requiring a specific diet or medication that contraindicates a high-fiber diet; (4) pregnancy; (5) receiving estrogen replacement therapy; (6) cirrhosis; (7) other primary or recurrent invasive cancer within the last 10 years; (8) unable to commit to the dietary intervention schedule; (9) a life-threatening medical condition or disease other than breast cancer that would interfere with participation in a lengthy dietary study; (10) any prior diagnosis of invasive breast carcinoma. Procedures Enrollment into the WHEL Study took place between March, 1995 and November, 2000 at the following seven clinical sites located in the southern and western United States: (1) University of California, San Diego; (2) University of California, Davis; (3) Kaiser Permanente Medical Group Inc., Oakland; (4) Kaiser Permanente Center for Health Research, Portland; (5) University of Arizona Cancer Center at Tucson and Phoenix; (6) Northern California Cancer Center (incorporating Stanford University and University of California, San Francisco); and (7) University of Texas M.D. Anderson Cancer Center, Houston. Potential participants were identified through tumor registries, community oncologists, and community outreach activities. Participants were randomized to either a dietary intervention group or comparison group by the Coordinating Center. Randomization was stratified by age (with two thirds of women being less than 55 years of age), stage of tumor at diagnosis (Stage I versus Stages II and IIIA), and clinic site. The institutional review board of each participating institution approved the WHEL Study protocol and informed consent was obtained from all participants prior to study entry. Recruitment into the WHEL trial is closed, however data collection is ongoing. Because data collection for the WHEL Study is ongoing, we were unable to remove

3 Second diagnosis of breast cancer and quality of life 27 the study blind and analyze the data according to group assignment. Issues related to condition assignment and the role of diet in quality of life can be examined once data collection is completed. Assessment procedures The health-related quality of life of all participants is assessed via mailed questionnaires that participants are instructed to bring to their scheduled clinic visits at several points over the course of the study beginning at study entry (baseline), and again at 12 months, 24 or 36 (split sample, with 50% at each time point) months, 48 months, and 72 months after randomization. Study participants are contacted every 6 months by the coordinator of their clinic site for cohort maintenance and to monitor the two primary WHEL Study endpoints: (1) additional breast cancer events and (2) death due to any cause. If a participant reports a second breast cancer event (recurrence or new primary), a post-second diagnosis telephone interview is conducted and her medical records reviewed for confirmation of diagnosis. Although the diagnosis of a second breast cancer/recurrence is one of two WHEL Study endpoints, the overwhelming majority of women who experience a second diagnosis request to remain involved in the Study in some capacity. Thus, although the WHEL Study was initially designed to shed light on the relationship between diet and secondary breast cancer events or death, it now also presents a unique opportunity to examine quality of life issues in women diagnosed with a second breast cancer event in a large, prospective, longitudinal sample. Measures Demographic and medical information Demographic and medical information was collected from participants by their self-report at the time of enrollment into the study. Medical information related to a woman s initial cancer diagnosis and treatment, and recurrence or new primary was confirmed by reviewing her medical record. Medical records were accessed with the participant s signed permission, as well as human subjects approval at the WHEL coordinating center and each clinical site. Health-related quality of life General health-related quality of life was assessed with the Rand 36-Item Health Survey, also known as the SF- 36, a widely used self-report quality of life (QOL) instrument [16,17]. This survey is comprised of 36 items, and 35 items are used to quantify QOL in 8 areas (published reliabilities in parentheses): physical functioning (a=0.93), role limitations due to physical health (a=0.84), pain (a=0.78), general health perception (a=0.78), social functioning (a=0.85), role limitations due to emotional problems (a=0.83), emotional wellbeing (a=0.90), and energy/fatigue (a=0.86). Alpha reliabilities in the present sample were comparable for all scales at both time points (Cronbach s a range 0.76 to 0.90). As specified in the scoring protocol provided by Hays et al. responses to individual items were recoded and linearly transformed to a 0 to 100 point scale with higher scores indicating better QOL [16]. Scale scores in each area were then calculated by averaging the items that comprise a given scale. According to Ware et al. [17], five point differences in scale scores represent clinically relevant changes and 10 point differences represent moderate effect sizes. Statistical analyses In order to assess potential response bias, we compared four groups of women: (1) those who agreed to complete the QOL questionnaire but died prior to completion (n=45); (2) those who received a post-second diagnosis telephone call but declined to complete the QOL questionnaire (n=69); (3) those who were alive and agreed to complete the QOL questionnaire, but did not (n=48); and (4) those who completed the QOL questionnaire (n=143). We compared these four groups of women on demographic variables, initial disease characteristics, baseline QOL, and type of second diagnosis (local/regional recurrence (LR), distant/metastatic recurrence (DR), or new primary (NP)) using chi-squared tests for categorical variables and ANOVA for continuous variables (with Tukey s test for any post-hoc comparisons). For the comparisons of pre- to post-second diagnosis QOL scores, paired t-tests were used for the sample overall, as well as within each of the LR, DR and NP groups. To test whether the three groups differed from each other with regard to the mean change in each scale from pre- to post-second diagnosis, three independent t-tests were used (LR versus DR, LR versus NP and DR versus NP), setting p<0.01 for significance given that multiple tests were used. In addition, we evaluated the clinical relevance of the difference scores using the criteria provided by Ware et al. [17]. Results Sample characteristics and participant attrition The numbers of women who were eligible and able to participate is summarized in Figure 1. There were 337 women with a documented recurrence or new primary as of January 1, Of these, 23 had been diagnosed within the past 6 months, and were considered pending contact by the coordinating center and were thus not eligible for inclusion in the analyses. An additional nine women died before they were contacted by the coordinating center, thus they are included in the denominator for determining the completion rate. A total of 143 postsecond diagnosis QOL questionnaires were received, for an overall completion rate of 45.5% (143/314 women thought to be alive).

4 28 AA Thornton et al. Figure 1. Participant response and attention n = all recurrence categories (local/distant/new primary). Baseline demographic and medical information, and details regarding the sample are presented for the four response groups in Table 1. Briefly, the women in our sample were predominately White, married, and highly educated. At randomization into the study, when women completed their baseline questionnaires, the 305 women in our eligible sample were a mean of 1.8 years beyond their initial breast cancer diagnosis. Among those who returned a post-second diagnosis QOL questionnaire, the average time between second diagnosis to completion of the questionnaire was 11.8 months (SD = 7.7; range from less than one month to 39.7 months). The women in the four response categories did not differ from each other in age (at initial or second diagnosis), ethnicity, or marital status, however women with higher levels of education were more likely to have completed the post-second diagnosis QOL assessment. There also were no differences in stage, surgical treatment, or radiation therapy at initial diagnosis; however, women who received chemotherapy for their initial diagnosis were more likely to have died prior to the QOL assessment window, or to have agreed to, but not completed the QOL assessment. Women who agreed to, but did not complete the QOL assessment were a mean of 4.5 years between initial and second diagnosis whereas the mean between initial and second diagnosis was 2.9 years for women who died prior to reaching the QOL assessment window. There were no differences between groups on any of the QOL subscales measured at baseline (data not shown). We also compared response rates across the various categories of potential participants. Women diagnosed with a DR had higher death rates (28% versus 4% LR and 6% NP; p < 0.001) after agreeing to, but before completing the post-second diagnosis QOL assessment. In addition, the DR women were also more likely to decline to complete a post-second diagnosis QOL assessment compared to the other categories of women (27% refusal rate versus 14% LR and 13% NP; p=0.023). Quality of life outcomes Five women in this analysis did not have a baseline QOL assessment available; for three of these, their 1 year (pre-second diagnosis) assessment was used. The other two women did not have a pre-second diagnosis QOL assessment, and were excluded from the analyses of preto post-second diagnosis QOL change. One woman completed a post-second diagnosis QOL assessment, but the data were lost due to technical problems. Thus, the sample size for the analyses of pre-post second diagnosis QOL change is 140. Pre- and post-second breast cancer event means and standard deviations are provided in Table 2 along with mailed-survey norms [18]. At baseline, most dimensions of QOL were rated as being high and mean scores appear to be largely comparable to or higher than published norms for U.S. women [17,18]. As shown in Table 2, collapsing across all categories of second diagnosis, 7 of the 8 subscales of the Rand 36 Item Health Survey worsened significantly from pre- to post-second diagnosis including: physical functioning (t[136]=7.16, p<0.001), role limitations due to physical problems (t[136]=4.85, p<0.001), pain (t[136]=6.25, p<0.001), general health perception (t[136]=8.97,

5 Second diagnosis of breast cancer and quality of life 29 Table 1. Sample characteristics by response category Variable Response category Died Pre-QOL Refused QOL Agree, no QOL Completed QOL Mean age at initial diagnosis (SD) 51.2 (8.1) 49.1 (10.6) 47.7 (8.4) 49.4 (10.0) Mean age at second diagnosis (SD) 54.1 (8.7) 52.5 (11.3) 52.2 (8.7) 53.3 (10.4) Mean recurrence-free interval in years (SD) 2.9 (1.5) 3.3 (1.8) 4.5 (1.9) 3.8 (1.8) Characteristics of recurrence Local/Regional recurrence n(%) 2 (4%) 9 (13%) 9 (19%) 46 (32%) Distant recurrence n(%) 41 (91%) 55 (80%) 30 (63%) 75 (52%) New breast primary n(%) 2 (4%) 5 (9%) 9 (19%) 22 (15%) Stage of initial cancer Stage I 11% 17% 17% 23% Stage II 73% 64% 73% 70% Stage IIIA 16% 19% 10% 7% Surgical therapy Lumpectomy 38% 45% 42% 41% Mastectomy 62% 55% 58% 59% Radiation therapy No 33% 33% 33% 42% Yes 67% 67% 67% 58% Chemotherapy a No 9% 28% 11% 23% Yes 91% 73% 89% 77% Ethnicity White 87% 88% 90% 83% Hispanic 9% 9% 4% 4% Black 4% 0% 4% 4% Asian or Pacific Islander 0% 0% 2% 6% Other 0% 3% 0% 0% Marital Status Single 11% 10% 13% 15% Married 71% 75% 72% 77% Separated/Divorced 9% 13% 11% 11% Widowed 9% 1% 4% 3% Education level b Did not attend college 9% 26% 21% 13% Attended some college 53% 32% 35% 22% College graduate 16% 17% 19% 36% Post-college 22% 25% 25% 30% Note. With the exception of variables describing recurrence, sample demographic and medical information was obtained at study enrollment. QOL = quality of life. For response categories: Died Pre-QOL (n=45); Refused QOL (n=69); Agree, no QOL (n=48); Completed QOL (n=143). a p=0.025;. b p < p<0.001), energy (t[136]=3.80, p<0.001), social functioning (t[136]=5.29, p<0.001), and role limitations due to emotional problems (t[136]=2.15, p=0.03). Emotional wellbeing did not worsen appreciably from pre- to post-second diagnosis. Applying Ware et al. s standards, the pre-post second diagnosis change scores represent moderate effect sizes (>10 points) for all significant scales with the exception of emotional limitations and energy, and for all scales registering significant change, the amount of change is likely to be clinically meaningful (i.e., difference score >5 points) [17]. We also compared change scores within each category of second diagnosis (LR, DR, NP) to ascertain whether women in the various second diagnosis categories demonstrated different patterns of change in their QOL scores. The following scales worsened significantly for women with an LR: physical functioning (t[42]=3.13, p=0.003), pain (t[42]=2.78, p=0.008), general health perception (t[42]=2.52, p=0.02), social functioning (t[42]=2.83, p=0.007); change to role limitations due to physical problems, role limitations due to emotional problems, emotional wellbeing, and energy did not attain statistical significance (however the difference score for role limitations due to physical problems is large enough to appear to be clinically meaningful). The difference scores were large enough to be considered clinically significant, and for the most part represent moderate effect sizes.

6 30 AA Thornton et al. Table 2. Pre- and post-recurrence comparisons on the rand 36item health survey by diagnostic category and overall [Mean (SD)] Recurrence category PF PR P GHP Pre Post Pre Post Pre Post Pre Post LR 87.6 (13.8) a 78.2 (20.0) b 63.6 (36.4) 55.7 (40.7) 69.3 (24.2) a 59.7 (25.6) b 66.6 (18.4) a 60.2 (17.3) b DR 83.3 (18.9) a 69.5 (24.3) b 70.0 (37.7) a 45.7 (42.8) b 74.0 (23.4) a 58.8 (23.1) b 70.3 (19.3) a 50.5 (20.0) b NP 89.8 (13.8) a 80.0 (22.1) b 77.4 (33.5) 52.4 (45.3) 75.6 (21.8) a 54.2 (22.8) b 76.7 (18.8) a 62.5 (18.2) b All 85.6 (16.8) a 73.8 (23.0) b 69.1 (36.7) a 49.8 (42.5) b 72.8 (23.4) a 58.4 (23.8) b 70.1 (19.1) a 55.3 (19.5) b Norms A E SF EWB ER Pre Post Pre Post Pre Post Pre Post LR 55.6 (20.2) 53.5 (19.6) 88.1 (18.5) a 78.1 (23.8) b 75.5 (12.4) 76.2 (13.8) 79.5 (28.1) 76.5 (32.6) DR 58.7 (20.2) a 50.9 (20.4) b 84.3 (22.9) a 74.5 (25.0) b 77.5 (14.9) 76.5 (13.2) 80.9 (34.7) 73.8 (38.9) NP 64.0 (21.9) a 52.2 (23.8) b 90.5 (14.7) a 73.2 (28.9) b 81.1 (9.6) 78.5 (14.0) 95.2 (15.9) a 87.3 (26.8) b All 58.5 (20.5) a 51.9 (20.6) b 86.4 (20.5) a 75.4 (25.1) b 77.5 (13.5) 76.7 (13.4) 82.6 (30.8) a 76.7 (35.5) b Norms A Note. Abbreviations: PF = physical functioning; PR = physical role limitations; P = pain; GHP = general health perception; E = energy/ fatigue; SF = social functioning; EWB = emotional wellbeing; ER = emotional role limitations; Pre = pre-second diagnosis; Post = postsecond diagnosis; LR = local/regional recurrence; DR = distant/metastatic recurrence; NP = new primary breast cancer; All = all recurrence categories and new primaries. For all scales, higher scores indicate better quality of life. For LR, n=44; for DR, n=75; for NP, n=21; for All, N=140. Mean values with different subscripts differ at p<0.05. A From Comparisons of the costs and quality of norms for the SF-36 Health Survey collected by mail versus telephone interview: Results from a national survey. by Colleen A. McHorney, Mark Kosinski, & John E. Ware, Jr., 1994, Medical Care, 32, For women with a DR, change to physical functioning (t[74]=5.75, p < 0.001), role limitations due to physical problems (t[74]=4.26, p < 0.001), pain (t[74]=4.49, p < 0.001), general health perception (t[74]=8.61, p < 0.001), energy (t[74]=3.03, p= 0.003), and social functioning (t[74]=3.54, p=0.001) worsened significantly from pre- to post-second diagnosis; change to role limitations due to emotional problems and emotional wellbeing were not statistically significant (although the difference score for the scale measuring role limitations due to emotional problems is large enough to suggest it would be considered clinically significant). Again, difference scores were sufficiently large to suggest that they represent clinically meaningful change and most represent at least moderate effect sizes. The following scales worsened significantly following the diagnosis of an NP: physical functioning (t[18]=3.27, p=0.004), pain (t[18]=3.77, p=0.001), general health perception (t[18]=4.31, p=0.001), energy (t[18]=2.35, p=0.03), social functioning (t[18]=2.79, p=0.01), and role limitations due to emotional problems (t[18]=2.45, p=0.03). Emotional well being approached statistical significance (t[18]=2.09, p=0.05), but was not large enough to be considered clinically significant. Although the scale measuring role limitations due to physical problems was not statistically significant, the change score was sufficiently large to suggest that change in this domain was in fact clinically significant. The statistically significant changes represent at least moderate effect sizes that are likely to be of clinical significance. To summarize, across each category, the scales measuring physical functioning, pain, general health perception, and social functioning worsened significantly following second diagnosis. Change to the scales assessing role limitations due to physical problems, energy, and role limitations due to emotional problems showed more variability, reaching statistical significance for some, but not all, categories of patients. However, in many cases the worsening in these domains was sufficiently large to be considered clinically meaningful. Change to emotional wellbeing was most pronounced for women diagnosed with a NP, but was not statistically or clinically significant for any of the other groups of women. We conducted subsidiary analyses to determine whether the amount of change in quality of life scores differed significantly by diagnostic category (DR, LR, or NP). Women who were diagnosed with a DR experienced a significantly greater decline in perceived general health compared to women with an LR (mean decline of 19.9 in DR versus mean decline of 6.4 in LR, p=0.001), likely reflecting the more serious nature of their diagnosis. However, no other comparisons of score changes between diagnostic categories were significant. Thus, although the precise pattern of change in QOL domains differed slightly across each category, comparisons between categories were for the most part not significant. Discussion We found that being diagnosed with breast cancer for a second time had a significant impact on most dimensions of health related quality of life including: physical health, role performance due to physical problems, pain,

7 Second diagnosis of breast cancer and quality of life 31 general health perception, social functioning, energy/ fatigue, and role limitations due to emotional problems. The statistically significant changes in quality of life that we observed were large enough to be considered clinically meaningful and for the most part represent difference scores that would be considered at least moderate in size. Emotional wellbeing did not worsen significantly from pre- to post-second diagnosis, however; nor was change in this domain large enough to be considered clinically meaningful. Consistent with other reports, changes to physical components of health related quality of life appeared to be more pronounced than changes to emotional aspects of health related quality of life in our sample of women. However, as will be discussed further below, we think it is important to remain cautious in forming strong conclusions or generalizing on the basis of these data given the substantial evidence of response bias that we encountered in our sample. Overall, attrition in our sample was high, with preand post-second diagnosis quality of life data being available for only 45.5% of participants who experienced a recurrence or new breast primary and were not pending post-second diagnosis contact by the study coordinating center (42% of all women diagnosed with recurrence or new primary). Although attrition analyses revealed few differences between women who completed the post-second diagnosis quality of life questionnaire and the various categories of non-completers on medical, demographic, and quality of life variables measured at baseline; more highly educated women were more likely to complete the post-second diagnosis questionnaire. In addition, women who received chemotherapy at initial diagnosis were more likely to die before receiving the post-second diagnosis quality of life questionnaire or having the opportunity to complete it. Thus, our conclusions may not apply equally as well to women who were less educated or initially had more systemic disease involvement. In addition, our sample was comprised largely of highly educated, married, White women. The extent to which quality of life outcomes in these women are representative of quality of life outcomes in more diverse groups of women is unknown. And, because these data were obtained from a consecutive series of women who were diagnosed with a recurrence or a new breast primary within the context of a larger, on-going study that included an intensive dietary intervention we also cannot be sure that they are representative of other women with breast cancer. Finally, women in our sample completed their postsecond diagnosis quality of life questionnaires, on average, nearly one year after learning of their change in disease status. Prior research suggests that the point in the disease trajectory when data are gathered may have an impact on quality of life outcomes [10]. Thus, combining data acquired at varying lengths of time from recurrence or second diagnosis may obscure important differences in or patterns of adjustment to this diagnosis. For instance, participants may have initially experienced an acute worsening in physical or emotional components of adjustment that had largely resolved by the time that they completed their post-second diagnosis quality of life questionnaire. Longitudinal research with multiple time points is needed to better clarify the relationship between time since second diagnosis and quality of life outcomes. It is also possible that factors related to the specific treatment that a woman received for her recurrence or new primary may impact quality of life outcomes. Unfortunately, these data were not available for the current report. Future research should consider the impact of medical as well as demographic and psychosocial factors on quality of life outcomes in this population. We also uncovered significant sources of response bias in analysis of our completion rates that limit the generalizability of our findings. Namely, women diagnosed with distant recurrence were less likely to complete a post-second diagnosis questionnaire either because of (a) declining to respond to it, or (b) dying prior to having the opportunity to respond. Women diagnosed with metastatic recurrence likely experience higher symptom distress and may be limited in their willingness and/or ability to participate in quality of life research. Given that women with distant disease were less likely to complete a post-second diagnosis questionnaire, the data that we report may underestimate the problems in this group. Women who were sicker may have experienced more profound declines in mental as well as physical components of quality of life, which are not reflected in the current sample. Furthermore, individual reactions to the diagnosis of recurrence a new primary may vary considerably, and some women may be more psychologically devastated than others by the change in their health status. There has been a movement to reduce the amount of missing QOL data in cancer clinical trials [19], however much of this literature focuses on statistical methods for imputing missing data [20,21], or administrative and coordination improvements in the collection of such data [22]. Moinpour et al. [23] demonstrated in their study of advanced colorectal cancer patients that completion of longitudinal QOL assessments was not random, and strongly related to deterioration of health status. Our findings support their conclusions that longitudinal analyses of QOL data in the advanced cancer setting need to be considered in light of patient attrition rates, particularly when attrition is associated with worsening health. In summary, our data suggest that being diagnosed with breast cancer recurrence or a new primary is associated with significant worsening in virtually all domains of health related quality of life. In addition, given the sources of potential response bias that we uncovered in our sample, we suspect that our data may underestimate the nature of the decrements in quality of life that accompany recurrence, particularly for women diagnosed with more extensive or

8 32 AA Thornton et al. metastatic recurrence. On the other hand, inspection of the mean quality of life scores reported here also suggests that there may be a subgroup of women who endorse relatively good quality of life, particularly in domains related to emotional and/or social functioning. Future research aimed at uncovering medical, demographic, and psychosocial predictors of quality of life and adjustment may improve our ability to identify patients at highest risk for poor outcomes and develop interventions to meet their needs. Acknowledgements The WHEL Study is supported by NCI grant CA69375 and the Walton Family Foundation. The Pete Lopiccola Cancer Research Foundation and the University of California, San Diego Cancer Center supported the first and second authors in the preparation of this manuscript. Special thanks to Cheryl L. Rock, Wayne A. Bardwell, and Cynthia Thomson for their comments on an earlier draft, and to the participants who made this study possible. The Women s Healthy Eating and Living Study Group: WHEL Study Coordinating Center: University of California, San Diego, Cancer Prevention and Control Program, San Diego, CA: John P. Pierce (Principal Investigator); Cheryl L. Rock, Susan Faerber, Vicky Newman, Shirley W. Flatt, Sheila Kealey, Loki Natarajan, Barbara Parker. WHEL Study Clinical Sites: Center for Health Research, Portland, OR: Njeri Karanja, Mark Rarick; Kaiser Permanente Northern California, Oakland, CA: Bette J. Caan, Lou Ferenbacher; Northern California Cancer Center, Union City, CA: Marcia L. Stefanick, Robert Carlson; University of Arizona, Tucson & Phoenix, AZ: Cynthia Thomson; James Warnecke; University of California, Davis, Davis, CA: Ellen B. Gold, Sidney Scudder; University of California, San Diego Cancer Center, San Diego, CA: Vicky E. Jones, Kathryn A. Hollenbach; University of Texas M.D. Anderson Cancer Center, Houston, TX: Lovell A. Jones, Richard Theriault. References 1. American Cancer Society: Cancer Facts and Figures Atlanta, Author, Gotay CC: The experience of cancer during early and advanced stages: the views of patients and their mates. Soc Sci Med 18: , Spencer SM, Lehman JM, Wynings C, Arena P, Carver CS, Antoni MH, Derhagopian RP, Ironson G, Love N: Concerns about breast cancer and relations to psychosocial well-being in a multiethnic sample of early-stage patients. Health Psychol 18: , Santiago RJ, Wu L, Harris E, Fox K, Schultz D, Glick J, Solin LJ: Fifteen-year results of breast-conserving surgery and definitive irradiation for Stage I and II breast carcinoma: the University of Pennsylvania experience. Int J Radiat Oncol Biol Phys 1: , Saphner T, Tormey DC, Gray R: Annual hazard rates of recurrence for breast cancer after primary therapy. J Clin Oncol 14: , Voogd AC, Nielsen M, Peterse JL, Blichert-Toft M, Bartelink H, Overgaard M, Tienhoven Gvan, Andersen KW, Sylvester RJ, Dongen JAvan: Danish Breast Cancer Cooperative Group. Breast Cancer Cooperative Group of the European Organisation for Research and Treatment of Cancer: differences in risk factors for local and distant recurrence after breast conserving therapy or masectomy for Stage I and II breast cancer: pooled results of two large European randomized trials. J Clin Oncol 19: , Camilleri-Brennan J, Steele RJC: The impact of recurrent rectal cancer on quality of life. Eur J Cancer 27: , Frost MH, Suman VJ, Rummans TA, Dose AM, Taylor M, Novotny P, Johnson R, Evans RE: Physical, psychological and social well-being of women with breast cancer: the influence of disease phase. Psycho-Oncology 9: , Winer EP, Lindley C, Hardee M, Sawyer WT, Brunatti C, Borstelmann NA, Peters W: Quality of life in patients surviving at least 12 months following high dose chemotherapy with autologous bone marrow support. Psycho-Oncology 8: , Bull AA, Meyerowitz BE, Hart S, Mosconi P, Apolone G, Liberati A: Quality of life in women with recurrent breast cancer. Breast Cancer Res Treat 54: 47 57, Hurny C, Bernhard J, Coates AS, Castiglione-Gertsch M, Peterson HF, Gelber RD, Forbes JF, Rudenstam CM, Simoncini E, Crivellari D, Goldhirsch A, Senn HJ for the International Breast Cancer Study Group: Impact of adjuvant therapy on quality of life in women with node-positive breast cancer. Lancet 347: , Oh S, Heflin L, Meyerowitz BE, Desmond KA, Rowland JH, Ganz PA: Quality of life of breast cancer survivors after a recurrence: a follow-up study. Breast Cancer Res Treat 87: 45 57, Ganz PA, Desmond KA, Leedham B, Rowland JH, Meyerowitz BE, Belin TR: Quality of life in long-term, disease-free survivors of breast cancer: a follow-up study. J Natl Cancer Inst 94: 39 49, Pierce JP, Faerber S, Wright FA, Rock CL, Newman V, Flatt SW, Kealey S, Jones VE, Caan BJ, Gold EB, Haan M, Hollenbach KA, Jones L, Marshall JR, Ritenbaugh C, Stefanick ML, Thomson C, Wasserman L, Natarajan L, Thomas RG, Gilpin EA: Women s Healthy Eating and Living (WHEL) study group: a randomized trial of the effect of a plant-based dietary pattern on additional breast cancer events and survival: the Women s Healthy Eating and Living (WHEL) Study. Control Clin Trials 23: , American Joint Committee on Cancer: Manual for Staging of Cancer, 4th ed. J.B. Lippincott Company, Philadelphia, PA, Hays R, Sherbourne C, Mazel R: The Rand 36-Item Health Survey 1.0. Health Economics 2: , Ware JE, Snow KK, Kosinski M, Gandek B: SF-36 Health Survey: Manual and Interpretation Guide. Lincoln, RI, QualityMetric Incorporated, McHorney CA, Kosinksi M, Ware JE: Comparisons of the costs and quality of norms for the SF-36 Health Survey collected by mail versus telephone interview: results from a national survey. Med Care 32: , Bernhard J, Cella DF, Coates AS, Fallowfield L, Ganz PA, Moinpour CM, Mosconi P, Osoba D, Simes J, Hurny C: Missing quality of life data in cancer clinical trials: serious problems and challenges. Stat Med 17: , Sales AE, Plomondon ME, Magid DJ, Spertus JA, Rumsfeld JS: Assessing response bias from missing quality of life data: the Heckman method. Health Qual Life Outcome 2: 49, Curran D, Molenberghs G, Fayers PM, Machin D: Incomplete quality of life data in randomized trials: missing forms. Stat Med 17: , Simes RJ, Greatorex V, Gebski VJ: Practical approaches to minimize problems with missing quality of life data. Stat Med 17: , 1998

9 Second diagnosis of breast cancer and quality of life Moinpour CM, Sawyers Triplett J, McKnight B: Challenges posed by non-random missing quality of life data in an advanced-stage colorectal cancer clinical trial. Psycho-oncology 9: , 2000 Address for offprints and correspondence: John P. Pierce, Cancer Prevention & Control Program, Cancer Center, University of California, San Diego, 9500 Gilman Drive, 0645, La Jolla, CA ; Tel.: ; Fax: ;