Identification of distinct fatigue trajectories in patients with breast cancer undergoing adjuvant chemotherapy

Transcription

1 Support Care Cancer (2015) 23: DOI /s x ORIGINAL ARTICLE Identification of distinct fatigue trajectories in patients with breast cancer undergoing adjuvant chemotherapy Doerte U. Junghaenel & Jules Cohen & Stefan Schneider & Anu R. Neerukonda & Joan E. Broderick Received: 11 August 2014 /Accepted: 13 January 2015 /Published online: 27 January 2015 # Springer-Verlag Berlin Heidelberg 2015 Abstract Purpose The goal of this study was to characterize changes in daily fatigue in women undergoing chemotherapy for breast cancer. We examined whether there are subgroups of patients with distinct fatigue trajectories and explored potential psychosocial and biomedical predictors of these subgroups. Methods Participants were 77 women with breast cancer receiving adjuvant chemotherapy with AC-T (2-week cycle) and TC or TCH (3-week cycle) regimens. They completed 28 daily ratings online using an adapted version of the Patient-Reported Outcomes Measurement Information System (PROMIS ) fatigue instrument. Results Both regimens followed an Binverted-U-shaped^ fatigue pattern over approximately 2 weeks. Growth mixture modeling identified three patient subgroups with distinct trajectories. Fatigue scores in the Blow fatigue^ group (23 %) increased following the infusion and quickly abated. The Btransient fatigue^ (27 %) group had a very pronounced increase. Patients in the Bhigh fatigue^ (50 %) group reported consistently elevated fatigue with a relatively small increase. Demographic and medical variables were not associated with fatigue trajectory. Patients in the Bhigh fatigue^ group reported significantly poorer physical, emotional, and social functioning, poorer general health, and more depressed mood than patients in the Blow fatigue^ group. The Btransient fatigue^ group reported significantly better physical and social D. U. Junghaenel (*): S. Schneider : J. E. Broderick Dornsife Center for Self-Report Science and Center for Economic and Social Research, University of Southern California, Verna & Peter Dauterive Hall, 635 Downey Way, Los Angeles, CA , USA junghaen@usc.edu J. Cohen: A. R. Neerukonda School of Medicine, Stony Brook University, Stony Brook, NY, USA functioning than the Bhigh fatigue^ group, but emotional distress and depression similar to the Bhigh fatigue^ group. Conclusions The identification of patient subgroups with distinct fatigue trajectories during chemotherapy is an essential step for developing preventative strategies and tailored interventions. Our results suggest that different trajectories are associated with patients psychosocial and general health. Keywords Fatigue. Breast cancer. Trajectory. Daily. Treatment. Chemotherapy Introduction Fatigue is a major symptom in cancer affecting 70 to 100 % of patients [1]. Considerable research efforts have been directed toward examining the detrimental consequences of fatigue on patients well-being and functioning [e.g., 2 4]. Fatigue in cancer is complex and multi-causal and is an important side effect of chemotherapy [5 7]. For example, in women with breast cancer, prevalence rates have been shown to range from 58 to 94 % during adjuvant chemotherapy [8, 9]. Moderate to severe fatigue levels during treatment have been reported in % of women [10, 11]. Previous studies typically report a marked increase in fatigue after an infusion. This Broller coaster effect^ is characterized by a peak in fatigue in the week following an infusion and a subsequent decline prior to the next infusion [8, 12, 13]. Despite the evidence that fatigue is pronounced during treatment, it still remains underrecognized and underdiscussed and we know very little about the daily course of fatigue during chemotherapy; multiple treatment guidelines for managing fatigue exist; however, cancer-related fatigue is not optimally dealt with, particularly during active treatment [15 18].

2 2580 Support Care Cancer (2015) 23: Several methodological factors have impeded optimal measurement and clinical management of fatigue during treatment. For example, prior research has not usually been longitudinal examining the daily change in fatigue [15]. This has precluded insight into patients daily symptom patterns as well as the heterogeneity in fatigue response across patients as is evident in wide-ranging estimates for fatigue during and after chemotherapy [5, 19]. Exceptions are studies by de Jong and colleagues [13] studying daily fatigue in patients with breast cancer between two cycles of adjuvant chemotherapy, Badr and colleagues [14] examining intraday changes in fatigue in ovarian cancer survivors, and Jim and colleagues [15] studying daily and intraday variation and associations between fatigue and other symptoms in patients with gynecologic cancer during chemotherapy treatment. All three studies demonstrate significant fluctuations in patients symptoms and highlight the need for determining if there may be subgroups of patients with distinct fatigue trajectories. Another issue involves a lack of commonly accepted diagnostic criteria and assessment tools for daily cancer-related fatigue [20, 21]. While daily diary measures of fatigue exist [22], they have rarely been subjected to rigorous psychometric evaluation [23]. This has hampered comparisons among studies and precluded substantive guidance on which patients to assist regarding their fatigue. Finally, many fatigue instruments are lengthy and burdensome for patients, particularly on a daily basis, as they go through active treatment. This makes them less appealing for use in daily clinical practice. The Patient-Reported Outcomes Measurement Information System (PROMIS ), a NIH initiative, offers instruments to measure fatigue that can alleviate some of the concerns with prior approaches to daily fatigue measurement during treatment. PROMIS developed state-of-the-art self-report instruments for common medical symptoms ( that can be easily incorporated in research and clinical practice. They are brief, reliable, valid, and cover a wide range of symptom severity. The PROMIS fatigue item bank was developed in a comprehensive fashion including reviews of existing fatigue scales, qualitative item review by fatigue experts and fatigued patients [24 26], item response theory (IRT) methods [27], and testing in a large sample of the US population [28, 29]. It shows good correspondence with other established fatigue instruments and performs well across the age range, diverse symptom levels, and clinical populations [30 32]. In the present study, we examined daily changes in fatigue in women receiving chemotherapy for breast cancer using an adapted Bdaily^ version of the PROMIS fatigue measure. We had three specific aims: the first aim was to characterize symptom levels and changes in fatigue over the course of the treatment cycle; the second aim was to identify subgroups of patients with distinct fatigue trajectories during the cycle; and the third aim was to examine demographic, psychosocial, and medical variables predicting the identified subgroups. Some evidence suggests that changes in body weight, menopausal symptoms, social support, biochemical changes, and treatment-related variables may be predictors of patients cancer-related fatigue; however, overall, knowledge about predictors is limited and inconclusive [see, e.g., 5, 10]. Methods Participants and procedure The study was approved by the Stony Brook University Institutional Review Board. Participants were 77 women with breast cancer receiving standard adjuvant chemotherapy with dose-dense AC-T (2-week cycle, n=31) and TC or TCH (3- week cycle, n=46) regimens. A convenience sample of women was recruited from Stony Brook University Cancer Center and a local community oncology practice. Brochures with information about the study were handed out by clinic staff and healthcare providers. Interested women could contact our research office to be telephone-screened for eligibility. General inclusion criteria were age 21 years, English fluency, availability to participate for days, and high-speed home Internet access. Specific inclusion criteria were adjuvant chemotherapy treatment through infusion every 2 or 3 weeks and no chemotherapy infusion within 5 days prior to enrollment to avoid recording fatigue from the previous infusion. All study interactions were conducted via telephone and Internet. Participants completed fatigue ratings over the PROMIS Assessment Center SM ( assessmentcenter.net),afreeonlinedatacollectiontool. Initially, research staff instructed participants how to provide electronic informed consent and how to use the Assessment Center. The first day of fatigue ratings was scheduled 8 days prior to a chemotherapy infusion. Participants completed daily fatigue assessments between 6 PM and midnight for 28 consecutive days. Compliance was monitored daily. Participants were compensated up to $150 for study completion. A lottery for an additional $150 was conducted after every 25th study completer for participants who had completed all daily assessments. Measures Daily fatigue measure Daily fatigue was assessed using a brief scale adapted from PROMIS. The original PROMIS seven-item fatigue short form uses a 7-day reporting period [28]. We used a modified daily diary version in which the reporting period was changed from BIn the past 7 days ^ to BIn the last day ^ [23]. The

3 Support Care Cancer (2015) 23: daily version of the PROMIS fatigue scale has demonstrated high reliability, consistent psychometric properties across different patient populations, and sensitivity to change [23, 33]. The scale is normed on a Z-score metric relative to a US general population sample, where 0 is the mean and 1 is one standard deviation (SD) above the mean fatigue level in the general population [23]. Predictor variables Sociodemographic characteristics including age, marital status, education, and employment were assessed at the initial telephone screening. Patient medical characteristics including chemotherapy regimen, infusion cycle during study participation, cancer stage, menopausal status, and hemoglobin were abstracted from the clinic charts. Physical, emotional, and social functioning were assessed on the day before the 28-day fatigue assessment period using seven PROMIS Bglobal health^ items [34]. These items were designed to provide a brief measure of several health domains and have been found to correspond with health status measured with the EQ-5D index [35]. We formed summary scores by averaging two items pertaining to physical (Cronbach s α= 0.71 in this sample), emotional (α=0.81), and social functioning (α=0.62), respectively. One additional item asked patients about their health in general. All items were scored on a 1 to 5 scale; higher scores represent poorer functioning. Depression was assessed weekly using the PROMIS depression measure administered via computerized adaptive testing (CAT) [28]. PROMIS CAT dynamically selects items based upon responses to the previous items for each individual patient to optimize measurement precision and was programmed to administer items until >0.90 score reliability was achieved (between 4 and 12 items were administered). Scores are reported on a T-score metric normed to have a mean of 50 and a SD of 10 in the US general population. The four weekly scores were averaged into an overall depression score characterizing each patient. Analysis strategy Data analysis proceeded in three steps. First, the average course of daily fatigue across all patients during the study was graphically examined and plotted separately to describe fatigue patterns during a 3-week (TC/THC) and 2-week (AC-T) cycles. Second, growth mixture modeling (GMM) was used to identify subgroups of patients with different fatigue trajectories over the treatment cycle. GMM is an extension of latent growth curve analysis. It seeks to determine types or Blatent classes^ of individuals with distinctly different patterns of change [36 38]. The GMM models were built gradually: we first estimated alternative latent growth models (linear, curvilinear, piecewise linear, and sinusoidal). We then used the best fitting latent growth model as the foundation for GMM models with an increasing number of patient subgroups. The optimal number of subgroups was determined with the bootstrapped likelihood ratio test (BLRT), which examines whether extracting an additional GMM class leads to significant improvement in model fit [39]. The GMM analyses were conducted simultaneously for patients in both regimens, andthedatawerelimitedtothe2weeksofthefirsttreatment cycle (from 1 day before to 12 days after the infusion). This was done to avoid the influence of the second cycle for patients with AC-T regimen in the formation of GMM classes, and to discriminate types of fatigue trajectories over a single treatment cycle. Missing fatigue scores were accommodated using full information maximum likelihood estimation [40]. In the final step, we examined differences among the identified fatigue trajectories on patients demographic, medical, and psychosocial variables. The method of Bpseudo-class^ draws was used [41]. Using this method, an individual s class membership was determined based on multiple imputations from the posterior probability distribution obtained in the GMM. Regression analyses relating the imputed class membership to the background variables were then conducted for each imputation and combined across imputations using Rubin s rules [42]. Analyses were performed using Mplus Version 7.11 [43]. Results Compliance with the daily assessments was high. Women (n= 77) completed on average 26.3 (SD=2.01, median=27) out of 28 daily assessments; a total of 129 out of 2156 assessment days (6 %) were missed. The average age was 51 years, most (91 %) women were White, 74 % were married, and 43 % were postmenopausal. About one third of patients (29 %) received their first chemotherapy infusion during the study. Almost half of the patients (40 %) received the AC-T regimen. Participants cancer staging was I (29 %), II (45 %), III (21 %), and IV (4 %). The majority of patients had undergone either mastectomy (51 %) or breast conserving surgery (37 %) (Table 1). Average fatigue patterns Across all patients and days, the mean fatigue level was Z-scores (SD=0.95), indicating generally elevated fatigue relative to the general population (p<0.001). Figure 1 shows the changes in average fatigue scores across the 28 days for both regimens. Patients on TC/TCH regimens received only one

4 2582 Support Care Cancer (2015) 23: Table 1 (n=77) Demographic and medical characteristics of study participants Mean±SD or percent levels did not significantly differ between the two regimens except for study days (i.e., starting 2 days after the second infusion for the AC-T regimen, ps<0.001; differences for days 8 and 7 were nonsignificant). Demographic characteristics Age 50.9±10.0 White 90.9 % Married 73.7 % Education High school 22.1 % Some college 35.1 % College graduate 28.6 % Advanced degree 14.3 % Family income a Less than $35, % $35,000 49, % $50,000 74, % $75,000 and higher 46.8 % Employed 55.8 % Medical and treatment characteristics Postmenopausal 42.9 % Weight (kg) 75.4±18.1 AC-T regimen (versus TC or TCH) 40.3 % First chemotherapy cycle 28.8 % Cancer stage I 29.3 % II 45.3 % III 21.3 % IV 4.0 % Tumor size (cm) 2.46±1.73 Surgery type Mastectomy 51.3 % Breast conserving surgery 36.8 % Neoadjuvant therapy 6.6 % Other 5.3 % Hemoglobin (HB) 12.22±1.23 Stage not available for two participants. Surgery type not available for one participant. Tumor size not available for nine participants. Weight not available for six participants. HB not available for five participants a Income was not reported by four participants Identification of patient subgroups Prior to examining evidence for multiple subgroups in GMM, it is necessary to specify a single-group latent growth model that fits the data. Several linear and nonlinear growth models were compared to capture the observed inverted-u pattern of fatigue. A sinusoidal change model (with a single sine and cosine function and a period of 14 days) provided a reasonable fit to the data (RMSEA=0.085, CFI=0.92, TLI=0.93). The sinusoidal model describes change in fatigue in the shape of a sine-wave. 1 This model was then used in GMM analyses comparing models with 1 to 4 latent growth classes. The BLRT indicated significant improvements in model fit for two-class (p<0.0001) and three-class (p=0.03) models (Table 2). Moving from a three-class to a four-class model did not yield significantly better fit (p=0.17). Thus, the model with three latent classes was retained. The observed mean fatigue scores and estimated growth curves of the three patient subgroups are shown in Fig. 2. The groups were labeled Blow fatigue^ (23.4 %), Btransient fatigue^ (27.1 %), and Bhigh fatigue^ (49.5 %). Fatigue scores in the Blow fatigue^ group were lower than the general population average prior to the infusion (Z-score of about 0.4), increased by about 0.4 to 0.5 Z-scores over 2 3 days, and then quickly returned to pre-infusion levels. The Btransient fatigue^ group showed somewhat low fatigue levels prior to the infusion (Z-score of about 0.2) but had a pronounced increase in fatigue of 1.5 Z-scores, with fatigue levels returning to preinfusion values after about 10 days. Finally, patients in the Bhigh fatigue^ group evidenced consistently elevated fatigue, with Z-scores of +0.8 on the day before the infusion and a further increase of about 0.3 Z-scores during days 2 8 of the cycle. Predictors of fatigue subgroups Demographic characteristics infusion (day 0 in Fig. 1); patients on AC-T regimen received a second infusion 14 days after the first infusion. Both regimens followed an Binverted-U-shaped^ pattern of fatigue over approximately 2 weeks. Mean fatigue levels were near normal (Z-scores of about 0.1 to 0.2) prior to the infusion, increased by about 0.8 to 0.9 Z-scores (a large effect size as per Cohen s conventions) over the following 2 5 days, and returned to near normal by days For the AC-T regimen, this pattern was repeated in the next cycle. The mean daily fatigue The subgroups did not significantly differ on demographic characteristics (ps>0.10, Table 2). There was a trend for patients in the Bhigh fatigue^ group to have lower income than patients in the Blow fatigue^ group (p=0.09). 1 Theequationusedisasfollows:y=β 0 +β 1 sin (2πt/14)+β 2 cos (2πt/14), where t represents the day of the cycle, β 0 is an intercept (representing the Boverall^ fatigue level), and β 1 and β 2 express amplitude (magnitude of increase and subsequent decrease) and phase shift (the Btiming^) ofsinusoidal change.

5 Support Care Cancer (2015) 23: Fig. 1 Mean fatigue levels for TC/TCH and AC-T regimens over the course of the 28-day study period. Day 0 is the day of chemotherapy infusion. Day 14 is the day of the second infusion for the AC-T regimen Fa gue z-score AC-T (n = 31) TC/TCH (n = 46) Day of chemotherapy cycle Baseline physical and psychosocial functioning Significant differences between the subgroups were evident on pre-infusion health and psychosocial status. The Bhigh^ and Blow fatigue^ groups differed consistently on all domains of functioning (see Fig. 3): compared with the Blow fatigue^ group, patients in the Bhigh fatigue^ group reported significantly poorer general health (p<0.05), poorer physical, emotional, and social functioning (ps< 0.01), as well as higher depression levels (p<0.001). The Btransient fatigue^ group displayed a more complex pattern. For physical and social functioning, this group scored significantly more favorably than the Bhigh fatigue^ group (ps<0.05) and comparable (ps>0.25) to the Blow fatigue^ group. For emotional problems and depression, the Btransient fatigue^ group did not significantly differ from the Bhigh fatigue^ group (ps>0.14), but showed marginally greater emotional problems (p=0.08) and significantly higher depression (p=0.02) than the Blow fatigue^ group. Medical characteristics No significant fatigue group differences were found for chemotherapy cycle number (first versus later cycle), type of regimen, cancer stage, tumor size, type of surgery, weight, menopausal status, and hemoglobin levels (Table 2). However, given that AC-T patients received a second infusion during the study, we also inspected the fatigue trajectories of each subgroup for days after the first observed chemotherapy infusion (i.e., the days that were not included in the GMM). As expected, for TC/TCH patients, fatigue levels in each subgroup remained relatively constant during the remaining study days. For AC-T patients, the group-specific changes in fatigue were replicated after the second infusion. Discussion We examined daily changes in fatigue in women receiving adjuvant chemotherapy for breast cancer. In line with prior research [8, 13, 14], we found an inverted-u-shaped fatigue pattern over approximately 2 weeks with an increase in levels over the next 2 5 days following an infusion and a return to near pre-infusion levels by days of the study. This trajectory was evident across two different chemotherapy regimens. We then sought to identify patient subgroups with distinct fatigue trajectories during a treatment cycle. We found three classes of patients with Blow fatigue^ (23 %), Btransient fatigue^ (27 %), and Bhigh fatigue^ (50 %). Fatigue scores in the Blow fatigue^ group demonstrated an expected increase in fatigue following the infusion, which then quickly subsided. Their fatigue scores prior to the infusion were lower than the general population average and also returned to below normal levels. This finding adds to evidence that cancer is not necessarily associated with higher symptomatology, such as fatigue, in all patients compared to the general population [31]. The Btransient fatigue^ group had a very pronounced, about 1.5 Z-scores on general population norms, increase in response to the infusion. The transitory nature of their trajectory suggests that they would likely not be identified in fatigue screening before treatment. Patients in the Bhigh fatigue^ group reported consistently elevated fatigue scores with a relatively small increase after the infusion. Their steady elevation of fatigue in this group cannot be attributed to a ceiling effect; the daily fatigue measure adapted from PROMIS has been shown to reliably capture pronounced fatigue at levels exceeding that observed in this study [23]. Determining predictors of fatigue trajectories during chemotherapy is an essential step for developing preventative strategies, identifying patients at risk, and for providing interventions to improve quality of life. With the

6 2584 Support Care Cancer (2015) 23: Table 2 Means (standard errors) or percent by fatigue subgroup on external variables Means (SE) or percent for each group p values for group differences High fatigue (H) (n=38) Transient fatigue (T) (n=21) Low fatigue (L) (n=18) HvsT HvsL TvsL Demographic characteristics Age 50.8 (1.64) 49.1 (2.05) 53.4 (3.03) Married/living together 72.7 % 69.6 % 80.5 % Lower education a 58.2 % 53.0 % 60.0 % Lower income b 41.8 % 24.7 % 17.8 % Employed 54.6 % 44.7 % 71.5 % Health domains General health problems 2.63 (0.16) 2.22 (0.25) 2.15 (0.13) Physical problems 2.54 (0.16) 1.90 (0.16) 2.24 (0.10) < Social problems 2.78 (0.16) 2.24 (0.19) 2.14 (0.15) Emotional problems 2.78 (0.14) 2.52 (0.17) 2.09 (0.17) Depression T-scores (1.30) (1.49) (1.66) 0.14 < Treatment characteristics AC-T regimen 55.5 % 72.1 % 54.0 % Cycle number (first cycle) 26.9 % 24.5 % 38.4 % Cancer stage (early late) 32.4 % 21.6 % 14.3 % Tumor size 2.60 (0.34) 2.56 (0.50) 2.10 (0.29) Surgery type (mastectomy) 58.6 % 67.0 % 48.5 % Weight (3.43) (4.39) (3.69) Postmenopausal status 44.7 % 29.0 % 55.2 % Hemoglobin (0.21) (0.30) (0.32) a Education was dichotomized into Bless than college graduate^ versus Bcollege graduate or more^ b Income was dichotomized into Bless than $50,000^ versus B$50,000 or more^ exception of income, none of the demographic and medical variables predicted group membership. There was a trend for patients in the Bhigh fatigue^ grouptohave lower income than patients in the Blow fatigue^ group. This is consistent with prior research [44] showing that highly fatigued breast cancer survivors had a lower annual income than less fatigued patients. These patients may have less resources available to manage their fatigue and may particularly need support. Our nonsignificant findings regarding patients medical and treatment-related characteristics join in with prior literature that has not identified reliable biomedical predictors of fatigue [5, 19, 21]. As our patients were recruited from two different facilities, the assessment of some variables, such as weight and hemoglobin, was not perfectly aligned across sites. Future research should standardize the timing of biomedical assessments and also examine the predictive role of change on these variables as possible fatigue predictors during treatment. Our data show that about half of the patients selfreported global health was not worse than the average of the general population; patients in the Bhigh fatigue^ group reported much worse global health. BHigh fatigue^ patients also reported significantly poorer physical, emotional, and social functioning, and more depressed mood than Blow fatigue^ patients. Thus, they report both physical and mental health distress. The Blow fatigue^ group was remarkable in its report of better physical and mental health than the average of the US population, indicating a particularly healthy and resilient group. The Btransient fatigue^ group was closer to US population ratings of health. A consistent finding in prior research is a moderate correlation between cancer-related fatigue and psychological distress, such as depression [5, 19, 45]. However, several studies have documented that an increase in fatigue was not accompanied by an increase in anxiety or depression [5, 21]. The present study adds to this by showing that higher depression levels do not necessarily correspond with consistently higher fatigue but may potentially be a marker of a susceptibility for fatigue as evident in the Btransient fatigue^ group. Pending replication, our results suggest that interventions to

7 Support Care Cancer (2015) 23: Fig. 2 Observed means and estimated growth curves of three-class growth mixture model. The horizontal line (at a score of zero) indicates the average fatigue level in the general population mitigate persistent fatigue are warranted for patients in the Bhigh fatigue^ group. The Btransient fatigue^ group would likely benefit from information to manage expectations regarding the brief elevation in fatigue after infusion and methods to cope with it. Several study limitations should be noted. First, we only assessed women s fatigue response to chemotherapy over one month. It is unclear whether our results generalize to an entire course of chemotherapy treatment. In addition, our sample size was limited, particularly for our subgroup analyses of patients receiving 2-week AC-T regimens. Further studies are needed to replicate patients trajectories and confirm their stability over a longer period of time. Nevertheless, it appeared that the distinct trajectories in the subset of our patients receiving 2-week AC-T regimens hold up over more cycles. Second, our sample was predominantly Caucasian and limited to one geographic location; further studies including larger and more diverse samples are needed. Finally, our available data regarding patients medical characteristics were limited. For example, we did not obtain information on patients thyroid function. In terms of strengths, several aspects of our study address limitations and concerns observed in prior research. First, our study is one of the few to examine changes in fatigue on a day-to-day basis during adjuvant chemotherapy. Most prior literature has relied on crosssectional or more periodic assessments, often averaging fatigue experiences over a span of time. Moreover, progress in understanding fatigue during chemotherapy has arguably been hampered by the use of measures with variable reliability and validity. The use of a measure adapted from PROMIS allowed us to capture daily fatigue with high measurement precision [23]. We identified three subsets of patients experiencing distinct fatigue trajectories and differing on psychosocial and general health. Pending replication, our results suggest that Fig. 3 Mean general, physical, social, and emotional health scores and depression levels in each group. Higher scores represent poorer functioning for each domain.the general population average is represented by a score of 0 (Z-scores) and 50 (T-scores), respectively

8 2586 Support Care Cancer (2015) 23: tailored interventions to mitigate or manage fatigue are warranted for patients with consistently Bhigh^ or Btransient fatigue^ over the course of treatment. Acknowledgments This research was supported by a grant from the National Institutes of Health (NIH 1-U01AR ). We thank our participants and Gim Yen Toh, Laura Wolff, Lauren Cody, and Linda Mahler for their assistance. PROMIS was funded with cooperative agreements from the NIH Common Fund Initiative (U54AR057951, U01AR052177, U54AR057943, U54AR057926, U01AR057948, U01AR052170, U01AR057954, U01AR052171, U01AR052181, U01AR057956, U01AR052158, U01AR057929, U01AR057936, U01AR052155, U01AR057971, U01AR057940, U01AR057967, U01AR052186). The contents of this article use data developed under PROMIS. These contents do not necessarily represent an endorsement by the US Federal Government or PROMIS (see for additional information). References 1. Mock V, Atkinson A, Barsevick A, Cella D, Cimprich B, Cleeland C et al (2000) NCCN Practice Guidelines for Cancer-Related Fatigue. Oncology 14: Curt GA, Breitbart W, Cella D, Groopman JE, Horning SJ, Itri LM et al (2000) Impact of cancer-related fatigue on the lives of patients: new findings from the Fatigue Coalition. Oncologist 5: Tiedtke C, de Rijk A, Dierckx de Casterle B, Christiaens MR, Donceel P (2010) Experiences and concerns about returning to work for women breast cancer survivors: a literature review. Psychooncology 19: Gupta D, Lis CG, Grutsch JF (2007) The relationship between cancer-related fatigue and patient satisfaction with quality of life in cancer. J Pain Symptom Manag 34: Servaes P, Verhagen C, Bleijenberg G (2002) Fatigue in cancer patients during and after treatment: prevalence, correlates and interventions. Eur J Cancer 38: Tiesinga LJ, Dassen TW, Halfens RJ (1996) Fatigue: a summary of the definitions, dimensions, and indicators. Nurs Diagn 7: Magnusson K, Moller A, Ekman T, Wallgren A (1999) A qualitative study to explore the experience of fatigue in cancer patients. Eur J Cancer Care 8: Jacobsen PB, Hann DM, Azzarello LM, Horton J, Balducci L, Lyman GH (1999) Fatigue in women receiving adjuvant chemotherapy for breast cancer: characteristics, course, and correlates. J Pain Symptom Manag 18: Greene D, Nail LM, Fieler VK, Dudgeon D, Jones LS (1994) A comparison of patient-reported side effects among three chemotherapy regimens for breast cancer. Cancer Pract 2: Andrykowski MA, Schmidt JE, Salsman JM, Beacham AO, Jacobsen PB (2005) Use of a case definition approach to identify cancer-related fatigue in women undergoing adjuvant therapy for breast cancer. J Clin Oncol 23: Lawrence DP, Kupelnick B, Miller K, Devine D, Lau J (2004) Evidence report on the occurrence, assessment, and treatment of fatigue in cancer patients. J Nat Cancer Inst Monog Berger AM, Higginbotham P (2000) Correlates of fatigue during and following adjuvant breast cancer chemotherapy: a pilot study. Oncol Nurs Forum 27: de Jong N, Kester ADM, Schouten HC, Abu-Saad HH, Courtens AM (2006) Course of fatigue between two cycles of adjuvant chemotherapy in breast cancer patients. Cancer Nurs 29: Badr H, Basen-Engquist K, Carmack Taylor CL, De Moor C (2006) Mood states associated with transitory physical symptoms among breast and ovarian cancer survivors. J Behav Med 29: Jim HSL, Small B, Faul LA, Franzen J, Apte S, Jacobsen PB (2011) Fatigue, depression, sleep, and activity during chemotherapy: daily and intraday variation and relationships among symptom changes. Ann Behav Med 42: Passik SD, Kirsh KL, Donaghy K, Holtsclaw E, Theobald D, Cella D et al (2002) Patient-related barriers to fatigue communication: initial validation of the fatigue management barriers questionnaire. J Pain Symptom Manag 24: Hilarius DL, Kloeg PH, van der Wall E, Komen M, Gundy CM, Aaronson NK (2011) Cancer-related fatigue: clinical practice versus practice guidelines. Support Care Cancer 19: Koornstra RH, Peters M, Donofrio S, van den Borne B, de Jong FA (2014) Management of fatigue in patients with cancer a practical overview. Cancer Treat Rev 40: de Jong N, Courtens AM, Abu-Saad HH, Schouten HC (2002) Fatigue in patients with breast cancer receiving adjuvant chemotherapy: a review of the literature. Cancer Nurs 25: Jacobsen PB (2004) Assessment of fatigue in cancer patients. J Natl Cancer Inst Monogr 32: Prue G, Rankin J, Allen J, Gracey J, Cramp F (2006) Cancer-related fatigue: a critical appraisal. Eur J Cancer 42: Watson T, Mock V (2004) Exercise as an intervention for cancerrelated fatigue. Phys Ther 84: Christodoulou C, Schneider S, Junghaenel DU, Broderick JE, Stone AA (2014) Measuring daily fatigue using a brief scale adapted from the Patient-Reported Outcomes Measurement Information System (PROMIS(R)). Qual Life Res 23: Christodoulou C, Junghaenel DU, DeWalt DA, Rothrock N, Stone AA (2008) Cognitive interviewing in the evaluation of fatigue items: results from the patient-reported outcomes measurement information system (PROMIS). Qual Life Res 17: Riley WT, Rothrock N, Bruce B, Christodoulou C, Cook K, Hahn EA et al (2010) Patient-reported outcomes measurement information system (PROMIS) domain names and definitions revisions: further evaluation of content validity in IRT-derived item banks. Qual Life Res 19: DeWalt DA, Rothrock N, Yount S, Stone AA (2007) Evaluation of item candidates: the PROMIS qualitative item review. Med Care 45: S12 S Reeve BB, Hays RD, Bjorner JB, Cook KF, Crane PK, Teresi JA et al (2007) Psychometric evaluation and calibration of health-related quality of life item banks: plans for the Patient-Reported Outcomes Measurement Information System (PROMIS). Med Care 45:S22 S Lai JS, Cella D, Choi S, Junghaenel DU, Christodoulou C, Gershon R et al (2011) How item banks and their application can influence measurement practice in rehabilitation medicine: a PROMIS fatigue item bank example. Arch Phys Med Rehabil 92:S20 S Cella D, Riley W, Stone A, Rothrock N, Reeve B, Yount S et al (2010) The Patient-Reported Outcomes Measurement Information System (PROMIS) developed and tested its first wave of adult selfreported health outcome item banks: J Clin Epidemiol 63: Cook KF, Bamer AM, Amtmann D, Molton IR, Jensen MP (2012) Six patient-reported outcome measurement information system short form measures have negligible age- or diagnosis-related differential item functioning in individuals with disabilities. Arch Phys Med Rehabil 93: Rothrock NE, Hays RD, Spritzer K, Yount SE, Riley W, Cella D (2010) Relative to the general US population, chronic diseases are associated with poorer health-related quality of life as measured by the Patient-Reported Outcomes Measurement Information System (PROMIS). J Clin Epidemiol 63:

9 Support Care Cancer (2015) 23: Junghaenel DU, Christodoulou C, Lai JS, Stone AA (2011) Demographic correlates of fatigue in the US general population: results from the Patient-Reported Outcomes Measurement Information System (PROMISA) initiative. J Psychosom Res 71: Schneider S, Choi SW, Junghaenel DU, Schwartz JE, Stone AA (2013) Psychometric characteristics of daily diaries for the Patient- Reported Outcomes Measurement Information System (PROMIS(R)): a preliminary investigation. Qual Life Res 22: Hays RD, Bjorner JB, Revicki DA, Spritzer KL, Cella D (2009) Development of physical and mental health summary scores from the patient-reported outcomes measurement information system (PROMIS) global items. Qual Life Res 18: Revicki DA, Kawata AK, Harnam N, Chen WH, Hays RD, Cella D (2009) Predicting EuroQol (EQ-5D) scores from the patient-reported outcomes measurement information system (PROMIS) global items and domain item banks in a United States sample. Qual Life Res 18: Jung T, Wickrama KAS (2008) An introduction to latent class growth analysis and growth mixture modeling. Soc Personal Compass 2: Muthén B, Muthén LK (2000) Integrating person-centered and variable-centered analyses: growth mixture modeling with latent trajectory classes. Alcohol Clin Exp Res 24: Reinecke J, Seddig D (2011) Growth mixture models in longitudinal research. Asta-Adv Stat Anal 95: McLachlan G, Peel D (2000) Finite mixture models. Wiley, New York 40. Schafer JL, Graham JW (2002) Missing data: our view of the state of the art. Psychol Methods 7: Wang CP, Brown CH, Bandeen-Roche K (2005) Residual diagnostics for growth mixture models: Examining the impact of a preventive intervention on multiple trajectories of aggressive behavior. J Am Stat Assoc 100: Rubin DB (1987) Multiple imputation for nonresponse in surveys. Wiley, New York 43. Muthén LK, Muthén BO ( ) Mplus user s guide, 7th ed. Los Angeles, CA: Muthén & Muthén 44. Bower JE, Ganz PA, Desmond KA, Rowland JH, Meyerowitz BE, Belin TR (2000) Fatigue in breast cancer survivors: Occurrence, correlates, and impact on quality of life. J Clin Oncol 18: Liu LQ, Fiorentino L, Natarajan L, Parker BA, Mills PJ, Sadler GR et al (2009) Pretreatment symptom cluster in breast cancer patients is associated with worse sleep, fatigue and depression during chemotherapy. Psychooncology 18: