Changes Over Time in Occurrence, Severity, and Distress of Common Symptoms During and After Radiation Therapy for Breast Cancer

Transcription

1 98 Journal of Pain and Symptom Management Vol. 45 No. June Original Article Changes Over Time in Occurrence, Severity, and Distress of Common Symptoms During and After Radiation Therapy for Breast Cancer Kristin Hofsø, RN, Tone Rustøen, RN, PhD, Bruce A. Cooper, PhD, Kristin Bjordal, MD, PhD, and Christine Miaskowski, RN, PhD, FAAN Centre for Shared Decision Making and Collaborative Research (K.H., T.R.), Oslo University Hospital, Rikshospitalet, Oslo, Norway; Lovisenberg Diaconal University College (T.R.), Oslo, Norway; Division of Surgery and Cancer Medicine (K.B.), Department of Oncology, Oslo University Hospital, Radiumhospitalet, Oslo, Norway; and Office of Research (B.A.C.), Department of Community Health Systems, and Department of Physiological Nursing (C.M.), School of Nursing, University of California, San Francisco, California, USA Abstract Context. Little is known about changes over time in multiple dimensions of the symptom experience in patients with breast cancer undergoing radiation therapy (RT). Objectives. This study evaluated for changes in and predictors of occurrence, severity, and distress of six common symptoms (lack of energy, worrying, difficulty sleeping, feeling drowsy, sweats, and pain) during RT for breast cancer. Methods. Patients (n ¼ 88) completed the Memorial Symptom Assessment Scale before, during, and after the completion of RT, over a six-month period. Changes in symptom occurrence were evaluated using multilevel logistic regression analysis. Changes in severity and distress scores were evaluated using multilevel proportional odds ordinal logistic regression. The impact of five demographic and clinical characteristics (age, functional status, comorbidities, axillary lymph node dissection, and previous chemotherapy) was evaluated in these analyses. Results. The trajectories for occurrence, severity, and distress for the six symptoms followed similar patterns. For three of the six symptoms (lack of energy, feeling drowsy, and worrying), all three dimensions changed over time. For the other three symptoms (difficulty sleeping, sweats, and pain), no changes over time occurred for any of the symptom dimensions. The overall effect of the five covariates was to increase symptom burden across all three dimensions. Conclusion. Findings from this study provide a more complete picture of the symptom experience of women who undergo RT for breast cancer. These findings can be used to identify patients at higher risk for more severe symptoms before, Address correspondence to: Kristin Hofsø, RN, Centre for Shared Decision Making and Collaborative Research, Oslo University Hospital, Rikshospitalet, Ó U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved. Postboks 495 Nydalen, 44 Oslo, Norway. kristin.hofso@rr-research.no Accepted for publication: June 4, /$ - see front matter

2 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 98 during, and after RT. J Pain Symptom Manage ;45:98e. Ó U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved. Key Words Lack of energy, worrying, difficulty sleeping, feeling drowsy, sweats, pain, occurrence, severity, distress, breast cancer, radiation therapy Introduction Almost years ago, Portenoy et al. developed the Memorial Symptom Assessment Scale (MSAS) to provide multidimensional information about a diverse group of common symptoms. Although valid symptom assessment scales were available,, Portenoy et al. argued that these instruments did not illuminate the multiple dimensions of a patient s symptom experience and were rarely used to investigate the relationships among symptoms and other factors that influence quality of life (QOL)., In addition, the authors emphasized that an evaluation of the occurrence, frequency, severity, and distress of common symptoms has the potential to clarify the impact of these symptoms on patient outcomes that could not be captured with unidimensional measures. The MSAS was used in several studies 4e to describe the symptom experience of patients with a variety of cancer diagnoses. Over 5% of these studies were cross-sectional in their design, 4,5,7, and only four evaluated for changes in one or more symptom dimensions over time.,8e In a study that evaluated for changes in all four dimensions of the MSAS over days in hospitalized patients with advanced cancer (n ¼ ), mean scores for each of symptoms and a total symptom score (i.e., a mean score derived from the frequency, severity, and distress scores) were evaluated. Whereas lack of energy was the most prevalent and frequently occurring symptom, pain was the most severe and distressing symptom. In addition, individual trajectories of symptom prevalence varied greatly within and among patients. In another longitudinal study, 9 four dimensions of the MSAS were evaluated before and one week after chemotherapy (CTX) in adolescents (n ¼ 5). In this study, a modified version of the MSAS was used (MSAS 7e) that assessed only eight symptoms. No differences were found in symptom occurrence rates before and after CTX. Nausea was the only symptom with a significantly higher distress rating after a cycle of CTX. 9 In another study that used the MSAS (MSAS e8) to describe changes in symptom occurrence and severity in children (n ¼ ) during a cycle of CTX, 8 a decreasing linear trend in the occurrence of six symptoms (i.e., fatigue, sadness, irritability, worrying, weight loss, and sweating) and significant quadratic changes in three symptoms (i.e., feeling drowsy, nausea, and vomiting) were found. However, no significant differences in symptom severity scores over time were found for any of the MSAS symptoms. 8 In terms of patients with breast cancer, only four longitudinal studies were found that described changes in multiple dimensions of the symptom experience during radiation therapy (RT). e In a study that used a 8- item version of the Oncology Treatment Toxicity Assessment Tool, both the occurrence of symptoms and their severity increased over the course of RT. In another study that used a 7-item version of the Oncology Treatment Toxicity Assessment Tool, 4 the occurrence and burden of symptoms were evaluated in breast cancer patients for six months after the completion of RT. Fatigue had the highest occurrence rate and the greatest impact on these patients lives. In a prospective study that evaluated the occurrence, severity, and pattern of symptoms experienced by women during and after primary RT, 5 symptom occurrence rates increased over the course of RT and then declined after treatment. Although the severity of symptoms was described as mild to moderate, they were significantly higher at the end of RT. However, this study had several limitations, including a small sample size (n ¼ ), the use of a symptom inventory that included only symptoms, and the inclusion of women with only Stages to II breast cancer. 5 In another study that investigated symptom distress in a sample of

3 98 Hofsø et al. Vol. 45 No. June patients with a variety of cancer diagnoses (i.e., head and neck, lung, and breast) who underwent radical RT, the most frequently occurring symptoms did not necessarily cause the most distress. In addition, the only symptom that increased significantly during treatment was tiredness. For the women with breast cancer in this study, difficulty concentrating, pain, and sleep disturbance emerged as significant causes of distress, when duration and intensity of the symptoms were controlled for in the statistical analyses. Of note, patients selected symptoms from a list of 4 symptoms and then rated the distress associated with these symptoms using a linear self-assessment scale. Given the paucity of research on the longitudinal assessment of multiple dimensions of the symptom experience in patients undergoing cancer treatment, additional research is warranted that evaluates not only for occurrence, but also for frequency, severity, and distress of common symptoms that occur during and after cancer treatment. This type of longitudinal information on the multiple dimensions of a patient s symptom experience is critical for clinicians to know when to target symptom management interventions both during and after cancer treatment. In this study, longitudinal data on the occurrence, severity, and distress for the six most prevalent symptoms found in a recent crosssectional analysis performed by our research team are presented. 7 In a sample of women recruited before the initiation of RT for breast cancer (n ¼ 88), the symptoms that occurred in over 5% of the patients were lack of energy, worrying, difficulty sleeping, feeling drowsy, sweats, and pain. These six symptoms were similar to those found in other studies that used the MSAS to evaluate patients with breast cancer., Furthermore, in our cross-sectional pre- RT analysis, women who had received CTX or who had a lower Karnofsky Performance Status (KPS) score or a higher comorbidity score reported a higher number of symptoms. In the present study, in addition to reporting on changes over time in ratings of symptom occurrence, severity, and distress, the influence of selected demographic and clinical characteristics (i.e., age, KPS score, comorbidity score, axillary lymph node dissection [ALND], previous CTX treatment) on patients baseline levels as well as on changes over time in the various dimensions of the symptom experience was evaluated. The rationale for not including frequency in this analysis is based on the original study of Portenoy et al., which stated that the distress scale yields the most information about the relationship between symptoms and QOL. In addition, they noted that the combination of a distress measure with either a frequency or a severity measure provides significantly more information than the distress measure alone. Therefore, to minimize the amount of data presented, the MSAS dimensions of occurrence, severity, and distress are reported in this article. Finally, a more appropriate method of analysis for ordinal data (i.e., multilevel logistic regression analysis) 8e was used to evaluate for changes over time in ratings of symptom occurrence, severity, and distress from before the initiation of RT to four months after the completion of RT (i.e., a period of six months). Materials and Methods Patients and Settings This study is part of a larger longitudinal study of symptoms in oncology outpatients and their family caregivers. Patients (n ¼ 88) were included if they were adults (8 years of age or older); able to read, write, and understand Norwegian; and were scheduled to receive RT for breast cancer or ductal carcinoma in situ. Patients were excluded if they were to receive RT to the brain or had a disease that affected their cognitive ability. Patients were recruited at the RT Department of the Norwegian Radium Hospital, at Oslo University Hospital. The study was approved by the Regional Committee for Medical and Health Research Ethics, the Norwegian Directorate of Health, the privacy ombudsman at the hospital, and the institutional review board at Norwegian Radium Hospital. Study Procedures Atthetimeoftheirfirstappointmentinthe RT Department (approximately eight days before the start of RT), patients were introduced to a member of the research team and provided with information about the study. After obtaining written informed consent, patients completed a number of self-report questionnaires to obtain information on demographic and clinical

4 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 98 characteristics and important life changes, as well as the KPS scale, the Self-Administered Comorbidity Questionnaire (SCQ-9), 4 and the MSAS. To assess for changes in the multiple dimensions of the symptom experience, patients completed the MSAS at one, two, three, and six months after enrollment. Most patients received a total of 5 Gy over a period of five weeks. Instruments Demographic Characteristics. Patients provided information on marital status, living situation, level of education, and employment status. Study nurses completed information on age. Clinical Characteristics. Stage of disease, treatment goal (curative or palliative), and types of previous treatments were obtained from the medical record. Previous cancer treatments were categorized as surgery, CTX, RT, hormonal therapy, and immunotherapy. Based on the patient s tumor (T), node (N), and metastasis (M) classification at the time of diagnosis, stage of disease was reclassified into five stages ( ¼ ductal carcinoma in situ to 4 ¼ advanced stage disease) using the TNM Classification of Malignant Tumors used by the Union for International Cancer Control guidelines. 5, Self-Administered Comorbidity Questionnaire-9. Patients completed the SCQ-9, which evaluated the number of, treatments for, and functional impact of health problems. The SCQ includes common comorbidities and three optional conditions. A higher total score indicates a more severe comorbidity profile. Using the SCQ-9, patients were asked to indicate whether they had the comorbid condition (yes/no); if they had the condition, they were asked if they received treatment for it (yes/no); and finally if it limited their activities (yes/no). The total SCQ-9 score can range from to 57 when the three optional items are used. 4 The SCQ- 9 is a clinical scale, with established validity and reliability for the assessment of comorbidities in patients with chronic medical conditions. 4 The SCQ-9 was translated into Norwegian using a standard forward-backward translation process 7 and was used in two previous studies in Norway. 8,9 KPS Scale. Information about patients level of physical functioning was self-reported using the KPS scale. The KPS scale ranges from (dead) to (the individual is able to carry on normal activities), in -point increments. Because patients in this study were outpatients, the 4 to range of the scale was used. The KPS scale is used extensively and has wellestablished reliability and validity., Memorial Symptom Assessment Scale. The MSAS comprises physical and psychological symptoms that occur as a result of cancer or its treatment. Changes in the occurrence, severity, and distress for the six most common symptoms reported by these patients at the initiation of RT for breast cancer are presented in this study. Using the MSAS, patients were asked to indicate whether or not they had the symptom during the past week (i.e., occurrence). If they had experienced the symptom, they were asked to rate its severity and distress. Symptom severity was rated using a four-point Likert scale ( ¼ slight, ¼ moderate, ¼ severe, and 4 ¼ very severe). Symptom distress was rated using a five-point Likert scale ( ¼ not at all, ¼ mild, ¼ moderate, ¼ severe, and 4 ¼ very severe). When analyzing these data, patients were coded as having the symptom if any one of three boxes were checked (i.e., occurrence, severity, and distress) even if the patient had checked do not have the symptom. All missing items were interpreted and coded to mean that the patient did not have the symptom. The reliability and validity of the MSAS is well established. The MSAS was translated into Norwegian using a standard forward-backward translation process. 7 Formal psychometric testing was not performed for the translated version, as the population in the present study (breast cancer) was part of the population in the original study by Portenoy et al. However, the Norwegian version was pilot tested with cancer patients, and only minor linguistic adjustments were needed based on feedback from patients. Data Analysis Data were analyzed using SPSS v. 8 (SPSS, Inc., Chicago, IL) and Stata v.. (StataCorp LP, College Station, TX). Descriptive statistics were used to characterize the sample. Appropriate parametric and nonparametric tests were performed to evaluate for differences in demographic and clinical characteristics

5 984 Hofsø et al. Vol. 45 No. June between patients who did and did not participate in this study. The six symptoms that occurred in 5% or more of the patients at the initiation of RT (i.e., lack of energy, worrying, difficulty sleeping, feeling drowsy, sweats, and pain) were evaluated in the subsequent longitudinal analyses. Demographic and clinical variables that had a bivariate association (r) of $ with the total number of symptoms at baseline (i.e., age, KPS score, SCQ score, ALND, and previous CTX treatment) were selected for evaluation as predictors in the conditional models. Using multilevel generalized linear models, these six symptoms were evaluated for changes over time in symptom occurrence, severity, and distress (a total of five assessments over six months). Symptom occurrence was coded as a binary variable (yes ¼ and no ¼ ) and examined using multilevel logistic regression. Symptom severity items were coded as ordinal, with ¼ not present and with increasing severity ratings of to 4. Symptom distress was coded as ordinal data, with ¼ not present, ¼ present but causing no distress, and with increasing distress ratings of to 5. Therefore, changes in symptom severity and distress were examined with multilevel proportional odds ordinal logistic regression (also called cumulative logit logistic regression). 8e For both types of models, random intercepts were estimated, with the first assessment being treated as the baseline (or intercept) for the growth trajectory. Unconditional models were examined first to estimate the linear change in the symptom reports. Given the possibility that the growth trajectory might not be only linear, quadratic effects were examined. Furthermore, the treatment period and a lengthy follow-up invited the examination of shifts (also called discontinuities ) in the growth trajectories. Therefore, piecewise models were examined. Two of the symptoms (lack of energy and feeling drowsy) exhibited discontinuous change. The change trajectory for lack of energy had two segments, baseline to two months and two to six months. The change trajectory for feeling drowsy had three segments, baseline to one month, one to three months, and three to six months. The other four symptoms were best fit with linear trajectories. After identifying the best fitting growth trajectory for each symptom, conditional models were fit to examine the associations for each of the covariates (age, KPS score, SCQ score, ALND, and previous CTX treatment) on the reported symptom dimensions at baseline and on the change in symptom dimensions over time (cross-level interaction). Two of the covariates (KPS and age) were reverse coded to make interpretation easier. An odds ratio (OR) greater than one for the KPS indicates that poorer functional status predicts a greater odds of reporting the symptom or a greater odds of reporting greater symptom severity or distress. An OR greater than one for age indicates that younger age predicts the greater odds of reporting the symptom or greater symptom severity or distress. The multilevel logistic regression models were estimated with Stata Release. (program xtlogit) using mean and variance adaptive Gauss-Hermite quadrature with integration points. The multilevel ordinal regression models were fit in Stata with gllamm, a program written for Stata. For these models, estimation was performed with an ologit link and using adaptive Gauss-Hermite quadrature with 5 integration points. For multilevel generalized linear models, estimation with adaptive quadrature was shown to be superior to pseudo-likelihood estimation, as used in other statistical programs. For all analyses, a P-value of <5 was considered statistically significant. Results Demographic and Clinical Characteristics A total of 45 patients were approached, 85% consented to participate, and 9% (n ¼ 88) of these patients completed the questionnaires before RT. Patients who declined participation were significantly older (.5 years, SD ) than those who participated (57.8 years, SD 9., P ¼ 5). As shown in Table, most of the sample was married or living with a partner (7%) and 45% had a secondary level of education. The mean KPS score for the total sample was 88 (SD ), and 77% of women had Stage I or II breast cancer. Approximately half (4%) of the women had received CTX before starting RT.

6 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 985 Table Demographic and Clinical Characteristics of Patients (n ¼ 88) at Enrollment Characteristics Mean SD Range (years) e e Self-Administered Comorbidity.. e8 Questionnaire score Total number of symptoms e Total dose of radiation therapy (Gy) 49.. e n % Education Primary Secondary 8 45 College/university 5 Marital status Married/living with partner 7 7 Unmarried/divorced/widowed 5 7 Employment status Employed 5 8 Not employed 7 9 Stage of disease Stage 7 9 Stage I 7 9 Stage II 8 8 Stage III 5 4 Type of surgery Breast conserving surgery 7 Mastectomy 8 Sentinel lymph node biopsy Previous treatment Hormonal therapy 5 Radiotherapy 5 Immunotherapy Of the patients who enrolled in the study, 85% completed the end of study questionnaires six months after the start of RT. Except for age being significantly higher for patients who did not complete the study, no differences were found in any other demographic or clinical characteristics between patients who did and did not complete the study. Most of the results are presented in tabular format. For symptoms with cross-level interactions, figures are provided to illustrate the effects of categorical covariates on changes in the symptom dimension over time. Lack of Energy Occurrence rates for lack of energy increased significantly over the first two months (P) and then decreased by 5% for each additional month from to (P) (Table, Fig. a). Each of the covariates increased the likelihood of reporting lack of energy at the initiation of RT. Younger patients were more likely to report the occurrence of lack of energy. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. times and.4 times more likely, respectively, to report lack of energy. Patients who had an ALND and those who received CTX were 7.4 times and 7. times more likely, respectively, to report lack of energy. In addition, cross-level interactions were found between both ALND and CTX and changes over time in the odds of reporting lack of energy. As shown in Table and Figs. a and b and a and b, compared with patients who did not have an ALND or CTX, for patients who had an ALND or had received CTX, the odds of reporting lack of energy decreased from baseline to two months and then increased from two to six months. Changes in overall severity ratings for lack of energy were nonlinear, first increasing during the first two months and then decreasing from to (Table, Fig. 4a). Each of the covariates influenced the severity of this symptom at the initiation of RT. Younger patients were more likely to report higher severity ratings for lack of energy. For each - point decrease in KPS score and each one unit increase in SCQ score, patients were. and. times more likely, respectively, to report a higher severity rating for lack of energy. Having had an ALND (Figs. c and d) or having received CTX (Figs. c and d) before RT were the only two predictors that influenced both the severity of this symptom at enrollment and the changes over time in the severity of this symptom. As shown in Table and Figs. c and d, compared with patients who did not have an ALND, for patients who had an ALND, the odds of reporting a higher severity rating for lack of energy were.5 times higher and their severity increased at a lesser rate to two months and then increased. As shown in Table and Figs. c and d, for patients who received CTX, the odds of reporting a higher severity rating for lack of energy were 8. times higher at enrollment and their severity decreased to two months, compared with patients who had not received CTX, whose trajectory increased from baseline to two months. Then the trajectory for patients who had received CTX continued to decrease to six months but at a lesser rate than those who had not received CTX.

7 98 Hofsø et al. Vol. 45 No. June Table Results of the Multilevel Regression Analyses of Occurrence Ratings for Six Symptoms Reported by Patients With Breast Cancer (n ¼ 88) Unconditional Model Conditional Model OR SE CI P-value OR SE CI P-value Lack of energydoccurrence Piecewise model P months.5.e. P months 8 9.e9 <5 5 e e5. <5.4..e. < e9.94 <5 P months e.8 <5 P months 4 7.4e <5 Pdcross-level interaction.. e <5 Pdcross-level interaction.9.5.8e8.4 < e.7 <5 P months e.7 <5 P months 5 7.5e <5 Pdcross-level interaction. 9.7e.54 <5 Pdcross-level interaction e9.9 <5 Worryingdoccurrence Decreasing linear 9 5 e.98 8 e e.5 <5. 9 e e4.4 <5.9.4.e7.57 Difficulty sleepingdoccurrence Decreasing linear.95 5 e5.7 9 e e7.4 8 Feeling drowsydoccurrence Piecewise model (Continued)

8 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 987 Table Continued Unconditional Model Conditional Model OR SE CI P-value OR SE CI P-value P months 7. 5.e8 <5 P months e8 <5 P months.8 7.7e4.48 < e4. <5.4 9.e.59 <5.7.e e.4 <5 P months.8. e7.4 <5 P months e4 <5 P months.7.45e5. Pdcross-level interaction.5. 4e.54 4 Pdcross-level interaction.8..5e Pdcross-level interaction Sweatsdoccurrence Decreasing linear e5 7 e e.78 Paindoccurrence Decreasing linear e9.9.e4. <5 Baseline.55..5e.78 <5.49.e5.5 5 OR ¼ odds ratio; ¼ not significant. Changes in overall distress ratings for lack of energy were nonlinear, first increasing during the first two months and then decreasing from to (Table 4, Fig. 5a). Each of the covariates influenced the distress ratings of this symptom at the initiation of RT. Younger

9 988 Hofsø et al. Vol. 45 No. June a Probability of occurrence b Probability of occurrence Lack of energy Worrying Observed Predicted d e Feeling drowsy Sweats c Probability of occurrence Difficulty sleeping f Pain Fig.. Observed (open circles) and predicted (filled circles) trajectories for the probability of occurrence of a) lack of energy, b) worrying, c) difficulty sleeping, d) feeling drowsy, e) sweats, and f) pain across the six months of the study. patients were more likely to report higher distress ratings for lack of energy. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were.8 and.4 times more likely, respectively, to report a higher distress rating for lack of energy. Having had an ALND (Figs. e and f) or having received CTX (Figs. e and f) before RTwere the only two predictors that influenced both the distress of this symptom at enrollment and the changes over time. As shown in Table 4 and Figs. e and f, compared with patients who did not have an ALND, for patients who had an ALND, the odds of reporting a higher distress rating for lack of energy at enrollment were. times higher. Furthermore, their distress decreased to two months, whereas the probability of reporting higher distress for those who had not had an ALND increased. As shown in Table 4 and Figs. e and f, compared with patients who had not received CTX, the odds of reporting a higher distress rating for lack of energy at baseline were 9. times higher for patients who received CTX and decreased to two months. Then, both groups reported a decrease in the probability of reporting a higher distress rate to six months. Worrying Occurrence rates for worrying decreased over time (Table, Fig. b). For each additional month, the odds of reporting worrying

10 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 989 a No ALND Observed Predicted b ALND Probability of occurrence Probability of severity score c None vs slight to very severe (observed) None vs slight to very severe (predicted) None/slight vs moderate to very severe (observed) None/slight vs moderate to very severe (predicted) None to moderate vs severe/very severe (observed) None to moderate vs severe/very severe (predicted) None to severe vs very severe (observed) None to severe vs very severe (predicted) d e Probability of distress score Did not have vs not at all to very severe (observed) Did not have vs not at all to very severe (predicted) Did not have/not at all vs mild to very severe (observed) Did not have/not at all vs mild to very severe (predicted) Did not have to mild vs moderate to very severe (observed) Did not have to mild vs moderate to very severe (predicted) Did not have to moderate vs severe/very severe (observed) Did not have to moderate vs severe/very severe (predicted) Did not have to severe vs very severe (observed) Did not have to severe vs very severe (predicted) f Fig.. Observed (open circles) and predicted (filled circles) trajectories for the probability of lack of energy across the six months for occurrence in patients who a) did not have and b) did have an ALND, for higher vs. lower severity ratings for patients who c) did not have and d) did have an ALND, and for higher vs. lower distress ratings for patients who e) did not have and f) did have an ALND. Severity ratings are plotted as observed (open symbols) and predicted (filled symbols) for none vs. slight to very severe, none/slight vs. moderate to very severe, none to moderate vs. severe/very severe, and none to severe vs. very severe. Distress ratings are plotted as observed (open symbols) and predicted (filled symbols) for did not have vs. not at all to very severe, did not have/not at all vs. mild to very severe, did not have to mild vs. moderate to very severe, did not have to moderate vs. severe/very severe, and did not have to severe vs. very severe. ALND ¼ axillary lymph node dissection. decreased by %. All five covariates were associated with the occurrence of worrying at enrollment. Younger patients were more likely to report worrying. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. and. times more likely, respectively, to report worrying. Patients who had an ALND and those who had received CTX were 5.5 and.9 times more likely, respectively, to report worrying. Severity ratings for worrying decreased over time (Table, Fig. 4b). For each additional month, the odds of reporting a higher severity rating for worrying decreased by %. All five covariates were associated with higher severity ratings for worrying at enrollment. Younger

11 99 Hofsø et al. Vol. 45 No. June a No CTX Observed Predicted b CTX Probability of occurrence c Probability of severity score None vs slight to very severe (observed) None vs slight to very severe (predicted) None/slight vs moderate to very severe (observed) None/slight vs moderate to very severe (predicted) None to moderate vs severe/very severe (observed) None to moderate vs severe/very severe (predicted) None to severe vs very severe (observed) None to severe vs very severe (predicted) d e Probability of distress score Did not have vs not at all to very severe (observed) Did not have vs not at all to very severe (predicted) Did not have/not at all vs mild to very severe (observed) Did not have/not at all vs mild to very severe (predicted) Did not have to mild vs moderate to very severe (observed) Did not have to mild vs moderate to very severe (predicted) Did not have to moderate vs severe/very severe (observed) Did not have to moderate vs severe/very severe (predicted) Did not have to severe vs very severe (observed) Did not have to severe vs very severe (predicted) f

12 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 99 patients were more likely to report higher severity ratings for worrying. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. and. times more likely, respectively, to report a higher severity rating for worrying. Patients who had an ALND and those who had received CTX were 4.7 and. times more likely, respectively, to report a higher severity rating for worrying. Distress ratings for worrying decreased over time (Table 4, Fig. 5b). For each additional month, the odds of reporting a higher distress rating for worrying decreased by %. All five covariates were associated with higher distress ratings for worrying at enrollment. Younger patients were more likely to report higher distress ratings for worrying. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. and. times more likely, respectively, to report a higher distress rating for worrying. Patients who had an ALND and those who had received CTX were 4.9 and. times more likely, respectively, to report a higher distress rating for worrying. Difficulty Sleeping Occurrence rates for difficulty sleeping did not change over time (Table, Fig. c). For each one unit increase in SCQ score at baseline, patients were. times more likely to report difficulty sleeping. Patients who had an ALND were.8 times more likely to report difficulty sleeping at enrollment. Severity ratings for difficulty sleeping did not change over time (Table, Fig. 4c). All five covariates were associated with higher severity ratings at the initiation of RT. Younger patients were more likely to report higher severity ratings for difficulty sleeping. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. and. times more likely, respectively, to report a higher severity rating for difficulty sleeping. Patients who had an ALND and those who had received CTX were.8 and.5 times more likely, respectively, to report a higher severity rating for difficulty sleeping. Distress ratings for difficulty sleeping did not change over time (Table 4, Fig. 5c). Three covariates were associated with higher distress ratings at enrollment. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were.7 and. times more likely, respectively, to report a higher distress rating for difficulty sleeping. Patients who had an ALND were.7 times more likely to report a higher distress rating for difficulty sleeping. Feeling Drowsy The odds of reporting the occurrence of feeling drowsy were predicted to be times greater over the first month (P), then decreased by % from to (P), and then decreased 5.5% per month from to (P) (Table, Fig. d). Four of the covariates influenced the occurrence of this symptom at the initiation of RT. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. and.4 times more likely, respectively, to report feeling drowsy. Patients who had an ALND and those who had received CTX were.7 and 7.5 times more likely, respectively, to report feeling drowsy. In addition, a crosslevel interaction was found between CTX and changes over time in the odds of reporting feeling drowsy. As shown in Table and Figs. a and b, for patients who received CTX, the odds of reporting feeling drowsy were 4. times greater for each month from baseline to Month, compared with.8 times greater for the patients who had not received CTX. The estimated value of 4. was calculated from the coefficients for the logits and can be = Fig.. Observed (open circles) and predicted (filled circles) trajectories for the probability of lack of energy across the six months for occurrence in patients who a) had not and b) had received CTX, for higher vs. lower severity ratings for patients who c) had not and d) had received CTX, and for higher vs. lower distress ratings for patients who e) had not and f) had received CTX. Severity ratings are plotted as observed (open symbols) and predicted (filled symbols) for none vs. slight to very severe, none/slight vs. moderate to very severe, none to moderate vs. severe/very severe, and none to severe vs. very severe. Distress ratings are plotted as observed (open symbols) and predicted (filled symbols) for did not have vs. not at all to very severe, did not have/not at all vs. mild to very severe, did not have to mild vs. moderate to very severe, did not have to moderate vs. severe/very severe, and did not have to severe vs. very severe. CTX ¼ chemotherapy.

13 99 Hofsø et al. Vol. 45 No. June Table Results of the Multilevel Regression Analyses of Severity Ratings for Six Symptoms Reported by Patients With Breast Cancer (n ¼ 88) Unconditional Model Conditional Model OR SE CI P-value OR SE CI P-value Lack of energydseverity Piecewise model P months.4. e.8 P months 9 4e 5 e. 7..e5. < e.75 < e7. <5 P months...4e.85 <5 P months. 9 e.58 <5 Pdcross-level interaction 8..e.77 Pdcross-level interaction.4.7.9e e.7 <5 P months..7.e.9 <5 P months.5 8 e.54 <5 Pdcross-level interaction. 9 e.58 <5 Pdcross-level interaction e5.44 Worryingdseverity Decreasing linear.9 4 e.98 7 e e.8 <5. 8 e e...4.e Difficulty sleepingdseverity Decreasing linear.94 4 e e....e e.4..9.e e. 48 Feeling drowsydseverity Piecewise model (Continued)

14 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 99 Unconditional Model Table Continued Conditional Model OR SE CI P-value OR SE CI P-value P months e.58 <5 P months 5 e. <5 P months e4. <5 e..79.4e5.9 <5 P months e.84 <5 P months 5 e. <5 P months e4. <5 Pdcross-level interaction.4.8e.94 Pdcross-level interaction Pdcross-level interaction.5 9.e.9 <5.8 4e e.7 <5 P months e48. <5 P months e5 <5 P months.4.4e4. Pdcross-level interaction. 4e. <5 Pdcross-level interaction e9.7 4 Pdcross-level interaction Sweatsdseverity Decreasing linear e8 e e. <5.5. 7e.4 4 Paindseverity Decreasing linear 5.97e e5.8 <5 Baseline.44..e.4 <5 e..44.7e7.7 OR ¼ odds ratio; ¼ not significant.

15 994 Hofsø et al. Vol. 45 No. June a Probability of severity score Lack of energy d Feeling drowsy b Probability of severity score Worrying None vs slight to very severe (observed) None vs slight to very severe (predicted) None/slight vs moderate to very severe (observed) None/slight vs moderate to very severe (predicted) None to moderate vs severe/very severe (observed) None to moderate vs severe/very severe (predicted) None to severe vs very severe (observed) None to severe vs very severe (predicted) e Sweats c Probability of severity score Difficulty sleeping f Pain Fig. 4. Observed and predicted trajectories for the probability of a higher vs. a lower severity rating for a) lack of energy, b) worrying, c) difficulty sleeping, d) feeling drowsy, e) sweats, and f) pain across the six months of the study. Severity ratings are plotted as observed (open symbols) and predicted (filled symbols) for none vs. slight to very severe, none/slight vs. moderate to very severe, none to moderate vs. severe/very severe, and none to severe vs. very severe.

16 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 995 Table 4 Results of the Multilevel Regression Analyses of Distress Ratings for Six Symptoms Reported by Patients With Breast Cancer (n ¼ 88) Unconditional Model Conditional Model OR SE CI P-value OR SE CI P-value Lack of energyddistress Piecewise model P months e.55 P months. 9e9 7 e e5.55 <5 P months P months.7..5e.94 9 Pdcross-level interaction 8 7e Pdcross-level interaction.8.9 e e.57 < e5.8 <5 P months.8..e. <5 P months. e5 <5 Pdcross-level interaction 9..e.77 Pdcross-level interaction..7.7e e5.7 <5 P months.7..5e.8 <5 P months.7 8 5e.5 <5 Pdcross-level interaction.5 8 e.55 <5 Pdcross-level interaction..79e5.9 <5 Worryingddistress Decreasing linear 7 4 e.95 8 e e.5 <5. 8 e e e8.4 Difficulty sleepingddistress Decreasing linear e e e e.8 (Continued)

17 99 Hofsø et al. Vol. 45 No. June Unconditional Model Table 4 Continued Conditional Model OR SE CI P-value OR SE CI P-value Feeling drowsyddistress Piecewise model P months e8. <5 P months. 5e <5 P months.9..e.78 <5..7.8e5.4 <5 P months e7.7 <5 P months. 4 e <5 P months.9.7.e.8 Pdcross-level interaction.4 e.9 Pdcross-level interaction.87.5 e.7 9 Pdcross-level interaction.48..7e.7 <5 P months e7.9 <5 P months. 4 5e <5 P months.9.7.e.8 Pdcross-level interaction Pdcross-level interaction.. e Pdcross-level interaction.8..7e7.59 P months.. 5.e. <5 P months e.4 <5 P months Pdcross-level interaction 5..e 5 Pdcross-level interaction Pdcross-level interaction Sweatsddistress Decreasing linear 4.9e8 5 e. 5..e e.7 Painddistress Decreasing linear e4. <5 (Continued)

18 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 997 Table 4 Continued Unconditional Model Conditional Model OR SE CI P-value OR SE CI P-value Baseline. 8.e.48 <5 4 e..4.49e e OR ¼ odds ratio; ¼ not significant. calculated by back transforming the ORs that are shown. Changes in overall severity ratings for feeling drowsy were nonlinear, first increasing during the first month, then decreasing from to, and then increasing from to (Table, Fig. 4d). Each of the covariates influenced the severity of this symptom at the initiation of RT. Younger patients were more likely to report higher severity ratings for feeling drowsy. For each one unit increase in SCQ score, the odds of reporting a higher severity rating for feeling drowsy were.5 times greater. For patients who had an ALND, the odds of reporting a higher severity rating for feeling drowsy were.4 times greater. For each -point decrease in KPS score, the odds of reporting a higher severity rating for feeling drowsy were.8 times higher at enrollment and the odds of reporting a higher rating were 4% greater from enrollment to one month into RT. Having received CTX before RT influenced both the severity of this symptom at enrollment and the changes over time in the severity of this symptom. As shown in Table and Figs. c and d, for patients who had received CTX, the odds of reporting a higher severity rating for feeling drowsy were.5 times higher at enrollment. For patients with no prior CTX, the odds of reporting a higher severity rating were times greater from enrollment to one month, decreased steeply by % from one to three months, and then decreased less steeply by % from three to six months. Patients with prior CTX had a significantly less dramatic increase in the odds (.4) of reporting a higher severity rating from enrollment to one month and then a % decrease in the odds of reporting a higher severity rating from one to three months. The odds of reporting a higher severity rating for feeling drowsy did not differ between these two groups of patients from three and six months. Changes in overall distress ratings for feeling drowsy were nonlinear, following the same pattern as the severity ratings (Table 4, Fig. 5d). For each one unit increase in SCQ score, the odds of reporting a higher distress rating for feeling drowsy were.4 times greater. For each -point decrease in KPS score, the odds of reporting a higher distress rating for feeling drowsy were. times greater at the initiation of RT. Because KPS score demonstrated a cross-level interaction, the odds of reporting a higher distress rating at Month compared with enrollment was 9.7 times greater for each -point decrease in KPS score. In addition, for each -point decrease in KPS score, the odds of reporting a higher distress rating at Month were 5.8 times greater than at Month. No change associated with KPS scores was found in the odds of reporting a higher distress rating from three to six months. Having received CTX before RT influenced both the distress of this symptom at enrollment and the changes over time. As shown in Table 4 and Figs. e and f, for patients with no prior CTX, the odds of reporting a higher distress rating from enrollment to one month were. times greater and then decreased by 4% from one to three months. For patients who had received CTX, the odds of reporting a higher distress rating for feeling drowsy were. times greater from enrollment to one month, a change that differed significantly from those who had not received CTX. The change in the odds of reporting a higher

19 998 Hofsø et al. Vol. 45 No. June a Probability of distress score b Probability of distress score c Probability of distress score Lack of energy Worrying Difficulty sleeping Did not have vs not at all to very severe (observed) Did not have vs not at all to very severe (predicted) Did not have/not at all vs mild to very severe (observed) Did not have/not at all vs mild to very severe (predicted) Did not have to mild vs moderate to very severe (observed) Did not have to mild vs moderate to very severe (predicted) Did not have to moderate vs severe/very severe (observed) Did not have to moderate vs severe/very severe (predicted) Did not have to severe vs very severe (observed) Did not have to severe vs very severe (predicted) d e f Feeling drowsy Sweats Pain Fig. 5. Observed and predicted trajectories for the probability of a higher vs. a lower distress rating for a) lack of energy, b) worrying, c) difficulty sleeping, d) feeling drowsy, e) sweats, and f) pain across the six months of the study. Distress ratings are plotted as observed (open symbols) and predicted (filled symbols) for did not have vs. not at all to very severe, did not have/not at all vs. mild to very severe, did not have to mild vs. moderate to very severe, did not have to moderate vs. severe/very severe, and did not have to severe vs. very severe.

20 Vol. 45 No. June Changes in Symptom Dimensions During Radiation Therapy 999 a No CTX Observed Predicted b CTX Probability of occurrence c Probability of severity score None vs slight to very severe (observed) None vs slight to very severe (predicted) None/slight vs moderate to very severe (observed) None/slight vs moderate to very severe (predicted) None to moderate vs severe/very severe (observed) None to moderate vs severe/very severe (predicted) None to severe vs very severe (observed) None to severe vs very severe (predicted) d e Probability of distress score Did not have vs not at all to very severe (observed) Did not have vs not at all to very severe (predicted) Did not have/not at all vs mild to very severe (observed) Did not have/not at all vs mild to very severe (predicted) Did not have to mild vs moderate to very severe (observed) Did not have to mild vs moderate to very severe (predicted) Did not have to moderate vs severe/very severe (observed) Did not have to moderate vs severe/very severe (predicted) Did not have to severe vs very severe (observed) Did not have to severe vs very severe (predicted) f Fig.. Observed (open circles) and predicted (filled circles) trajectories for the probability of feeling drowsy across the six months for occurrence in patients who a) had not and b) had received CTX, for higher vs. lower severity

21 Hofsø et al. Vol. 45 No. June distress rating from one to three months and from three to six months did not differ as a function of CTX. Sweats Occurrence rates for sweats did not change over time (Table, Fig. e). Younger patients were. times more likely to report the occurrence of sweats at enrollment. For each -point decrease in KPS score, patients were.7 times more likely to report sweats at the initiation of RT. Severity ratings for sweats did not change over time (Table, Fig. 4e). Younger patients were more likely to report higher severity ratings for sweats at enrollment. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. and. times, respectively, more likely to report a higher severity rating for sweats at the initiation of RT. Distress ratings for sweats did not change over time (Table 4, Fig. 5e). Younger patients were more likely to report higher distress ratings for sweats at enrollment. For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. and. times more likely, respectively, to report a higher distress rating for sweats at the initiation of RT. Pain Occurrence rates for pain did not change over time (Table, Fig. f). For each -point decrease in KPS score and each one unit increase in SCQ score, patients were. and. times more likely, respectively, to report pain at enrollment. Patients who had an ALND were.5 times more likely to report pain at enrollment. Severity ratings for pain did not change over time (Table, Fig. 4f) except conditioned on SCQ scores. For each one unit increase in SCQ score, patients were.4 times more likely to report a higher severity rating for pain at enrollment and time more likely to report a higher severity rating for each additional month. For each -point decrease in KPS score, patients were.7 times more likely to report a higher severity rating for pain at enrollment. Patients who had an ALND were. times more likely to report a higher severity rating for pain at the initiation of RT. Distress ratings for pain did not change over time (Table 4, Fig. 5f), except conditioned on SCQ scores, having an ALND, and previous CTX treatment (Fig. 7). For each one unit increase in SCQ scores, patients were. times more likely to report a higher distress rating for pain at enrollment and time more likely to report a higher distress rating for each additional month. For each -point decrease in KPS score, patients were. times more likely to report a higher distress rating for pain at enrollment. As shown in Fig. 7, patients who had an ALND were. times more likely to report a higher distress rating for pain at enrollment. Patients who had received CTX were 5% less likely to report a higher distress rating for pain for each additional month from the initiation of RT. Discussion Symptom Dimension Trajectories This study is the first to examine changes in multiple dimensions of the symptom experience (occurrence, severity, and distress) for six of the most common symptoms reported by patients before, during, and after RT for breast cancer. Although fatigue-related symptoms (lack of energy, difficulty sleeping, and feeling drowsy) were reported by more than 5% of these patients at the initiation of RT, worrying (59%), sweats (5%), and pain (5%) emerged as significant clinical problems that are not well characterized from a longitudinal perspective. The occurrence, severity, and distress trajectories for these six symptoms could be grouped into three distinct patterns. For lack of energy and feeling drowsy, significant increases in all three symptom dimensions occurred during the first (feeling drowsy) or second (lack of energy) months of RT and were followed by = ratings for patients who c) had not and d) had received CTX, and for higher vs. lower distress ratings for patients who e) had not and f) had received CTX. Severity ratings are plotted as observed (open symbols) and predicted (filled symbols) for none vs. slight to very severe, none/slight vs. moderate to very severe, none to moderate vs. severe/very severe, and none to severe vs. very severe. Distress ratings are plotted as observed (open symbols) and predicted (filled symbols) for did not have vs. not at all to very severe, did not have/not at all vs. mild to very severe, did not have to mild vs. moderate to very severe, did not have to moderate vs. severe/very severe, and did not have to severe vs. very severe. CTX ¼ chemotherapy.