Cleeland, Charles S.

Background In clinical trials and clinical practice, patient-reported outcomes are almost always assessed using multiple patient-reported outcome measures at the same time. This raises concerns about whether patient responses are affected by the order in which the patient-reported outcome measures are administered. Methods This questionnaire-based study of order effects included adult cancer patients from five cancer centers. Patients were randomly assigned to complete questionnaires via paper booklets, interactive voice response system, or tablet web survey. Linear Analogue Self-Assessment, Patient-Reported Outcomes Version of the Common Terminology Criteria for Adverse Events, and Patient-Reported Outcomes Measurement Information System assessment tools were each used to measure general health, physical function, social function, emotional distress/anxiety, emotional distress/depression, fatigue, sleep, and pain. The order in which the three tools, and domains within tools, were presented to patients was randomized. Rates of missing data, scale scores, and Cronbach’s alpha coefficients were compared by the order in which they were assessed. Analyses included Cochran–Armitage trend tests and mixed models adjusted for performance score, age, sex, cancer type, and curative intent. Results A total of 1830 patients provided baseline patient-reported outcome assessments. There were no significant trends in rates of missing values by whether a scale was assessed earlier or later. The largest order effect for scale scores was due to a large mean score at one assessment time point. The largest difference in Cronbach’s alpha between the versions for the Patient-Reported Outcomes Measurement Information System scales was 0.106. Conclusion The well-being of a cancer patient has many different aspects such as pain, fatigue, depression, and anxiety. These are assessed using a variety of surveys often collected at the same time. This study shows that the order in which the different aspects are collected from the patient is not important.

Background: The Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events is an item library designed for eliciting patient-reported adverse events in oncology. For each adverse event, up to three individual items are scored for frequency, severity, and interference with daily activities. To align the Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events with other standardized tools for adverse event assessment including the Common Terminology Criteria for Adverse Events, an algorithm for mapping individual items for any given adverse event to a single composite numerical grade was developed and tested. Methods: A five-step process was used: (1) All 179 possible Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events score combinations were presented to 20 clinical investigators to subjectively map combinations to single numerical grades ranging from 0 to 3. (2) Combinations with <75% agreement were presented to investigator committees at a National Clinical Trials Network cooperative group meeting to gain majority consensus via anonymous voting. (3) The resulting algorithm was refined via graphical and tabular approaches to assure directional consistency. (4) Validity, reliability, and sensitivity were assessed in a national study dataset. (5) Accuracy for delineating adverse events between study arms was measured in two Phase III clinical trials (NCT02066181 and NCT01522443). Results: In Step 1, 12/179 score combinations had <75% initial agreement. In Step 2, majority consensus was reached for all combinations. In Step 3, five grades were adjusted to assure directional consistency. In Steps 4 and 5, composite grades performed well and comparably to individual item scores on validity, reliability, sensitivity, and between-arm delineation. Conclusion: A composite grading algorithm has been developed and yields single numerical grades for adverse events assessed via the Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events, and can be useful in analyses and reporting.

7018 Background: Standard reporting of symptomatic adverse events (AEs) in oncology relies on clinicians to rate patient (pt) experience using CTCAE; each symptom is represented by a single graded item. To capture direct pt experience, NCI developed PRO-CTCAE to supplement CTCAE. In PRO-CTCAE, the pt answers up to 3 questions per AE about a symptom’s frequency, severity and interference with daily activities. To align PRO-CTCAE with CTCAE, we developed an algorithm for mapping sets of questions for an AE to a single composite numerical grade. Methods: We used a 5-step process. (1) All 187 possible PRO-CTCAE score permutations were presented to clinical investigators to subjectively map permutations to single numerical grades (range 0-3). (2) Permutations with < 75% agreement were presented to investigator committees at a National Clinical Trials Network meeting to gain majority consensus via anonymous voting. (3) The resulting algorithm was refined via graphical and tabular approaches to assure directional consistency. (4) Validity, reliability and sensitivity were assessed in a national study dataset. (5) Accuracy for delineating AEs between study arms was measured in 2 phase III clinical trials (Alliance for Clinical Trials in Oncology A091105 and Exelixis COMET-2). Results: (1) 12/187 score permutations had < 75% initial agreement. (2) Majority consensus was reached for all permutations. (3) 5 mappings were adjusted to assure directional consistency. (4) Composite grades for 46/59 (78%) AEs were higher in pts with ECOG performance status 2-4 vs 0-1 (median effect size 0.23 [range -0.49-0.73]; 32/59 effect size ≥0.2; 25/59 p< 0.05), similar to when conducting analysis on individual unmapped items. The test-retest reliability for 24 selected composite grades ranged from 0.57-0.96 (median intraclass correlation coefficient [ICC] 0.77) with 18/24 (75%) grades having ICC ≥0.7. Median (range) standardized response means in pts reporting worsening, no change, and improvement were 0.20 (0.03-0.34), -0.06 (-0.20-0.03) and -0.12 (-0.32-0.06). (5) Pattern, directionality and statistical significance of between-arm differences in both trials were preserved with composite grades as compared to individual unmapped items. Conclusions: A composite grading algorithm for PRO-CTCAE was rigorously developed and validated. PRO-CTCAE composite grades may be useful in analyses to provide a single metric for each pt-reported AE for trial and real-world reporting. Support: UG1CA189823; U01CA233046; HHSN261200800043C; Bayer (A091105); https://acknowledgments.alliancefound.org .

9047 Background: Symptomatic adverse events (AE) in cancer trials are reported by clinicians using the National Cancer Institute's (NCI) Common Terminology Criteria for Adverse Events (CTCAE). To integrate the patient perspective into AE reporting, NCI contracted (HHSN261201000043C) to create a patient-reported outcomes companion tool (PRO-CTCAE). We report the validity and reliability of PRO-CTCAE’s 124 items reflecting 78 symptomatic AEs. Methods: English-speaking subjects (n=869; 44% male; median [mdn] age 59; 32% racial/ethnic minority; 34% high school or less; 17% ECOG Performance Status [PS] 2-4) receiving treatment for a range of cancers at 4 NCI-designated cancer centers and 5 sites in NCI's Community Cancer Centers Program completed a web-based survey in clinic including PRO-CTCAE and EORTC QLQ-C30. Pearson correlations were computed between PRO-CTCAE items and QLQ-C30 scales. Differences in PRO-CTCAE item scores between clinician-reported ECOG PS (0-1 vs 2-4) groups were computed. Test-retest reliability of 48 prespecified items was evaluated in a subset (n=79). Results: Correlations in the expected direction were observed for 116/124 PRO-CTCAE items with the QLQ-C30 global health scale (mdn r=-.21; range .08 to -.57). Stronger correlations were seen for PRO-CTCAE items with conceptually related QLQ-C30 domains: fatigue with physical, role and social functioning (-.54 to -.66); anxiety and depression with emotional functioning (-.54 to -.70); and concentration and memory problems with cognitive functioning (-.62 to -.72) [all p <0.001]. Fatigue, nausea, vomiting, pain, dyspnea, insomnia, appetite loss, constipation and diarrhea items strongly correlated with corresponding QLQ-C30 symptom scales (.69 to .79, all p <0.001). Scores for 98/124 PRO-CTCAE items were higher in the ECOG PS 2-4 vs 0-1 group (mdn effect size .3). Test-retest reliability was observed across all tested items (mdn intraclass correlation coeff .77; range .53 to .96). Conclusions: Results demonstrate favorable validity and reliability of PRO-CTCAE in a large, heterogeneous sample of patients undergoing cancer treatment. Further testing in multicenter treatment trials is underway.

e11051 Background: Limited data exist on patient (pt) experience and work productivity (WP) in MBC. VIRGO is a prospective OCS following >1200 pts with locally advanced or MBC receiving 1st-line hormonal therapy (HT) or chemotherapy (CT) in a real-world setting. We report baseline characteristics of 277 pts from the VIRGO PRO substudy and correlations between health-related quality of life (HRQoL), symptoms, activities of daily living, and WP. Methods: Symptom severity and interference (MD Anderson Symptom Inventory [MDASI]), functional status (Activity Level Scale [ALS] from the Rotterdam Symptom Checklist) and WP (Work Productivity and Activity Impairment Questionnaire) were assessed. Pts rated their HRQoL during the past week on a scale of 0–10. Results: See table. The five most severe symptoms at baseline were fatigue, decreased sexual interest, disturbed sleep, drowsiness, and emotional distress; these were reported with less severity in the HT cohort (24%-37% vs 42%-59%). Overall MDASI severity and interference correlated with WP measures in the CT (R = 0.46 to 0.78) and HT cohorts (0.36 to 0.94). Mean of the five most severe symptoms also significantly correlated with WP indices (R = 0.47 to 0.66). HRQoL correlated (p < 0.05) with all WP measures (R = -0.46 to -0.56) in the CT cohort and with % impairment while working (R = -0.65) and % overall work impairment (R = -0.71) in the HT cohort. In univariate regression analysis, MDASI symptom interference score was the best predictor of reduced WP (R2 = 0.52 while working; R2 = 0.48 for nonwork activities). Conclusions: MBC pts receiving CT and HT report significant work impairment. Results indicate moderate correlations between WP indices, HRQoL and symptom burden. [Table: see text]

Abstract Background Inflammatory deregulation may be a major factor in cancer related symptoms (Dantzer Nature Rev Clin Oncol 2012). Improved symptoms in myelofibrosis (MF) patients treated with single arm ruxolitinib (JAK1/JAK2 inhibitor) studies has been correlated with normalization of selected cytokines increased pre-therapy (c-reactive protein, IL-1ra, MIP-1b, TNF-a, and IL-6; Verstovsek NEJM 2010). We sought to assess the impact of JAK1/JAK2 inhibitor therapy on cytokine levels and the relationship between cytokine levels and symptoms prior to and during JAK1/JAK2 inhibitor therapy in the phase III placebo controlled COMFORT-I trial (Verstovsek NEJM 2012). Methods Cytokine levels (89 cytokines measured at baseline and at weeks 4 and 24) and MF symptoms (assessed by MFSAF 2.0 – Mesa JCO 2013) were collected during the blinded phase of COMFORT-I. Patients were randomized to ruxolitinib vs. placebo. Plasma was used for the measurement of cytokines using Rules-Based Medicine, Inc. (Austin, TX) Human MAP panel. Associations between the MFSAF total symptom score (TSS) and log2-transformed cytokine data were investigated at baseline using Spearman correlations and linear regression. Mixed models were used to assess cytokine and TSS changes over time within each arm and overall. Logistic regression was used to assess the relationship between baseline cytokines and TSS response (>/=50% reduction from baseline) at week 24, and between week 4 cytokine changes and TSS response (>/=50% reduction from baseline) at week 24 within each arm and overall. Mixed and logistic regression models combining data across arms also included terms for visit, arm, and visit-by-arm interaction. All models also included age, gender, and body mass index (BMI). Given the large number of cytokines being investigated, p<0.001 was considered statistically significant. Cytokine values below the limit of detection were set at the lowest limit of detection and 25 cytokines were excluded from statistical analysis due to having more than 30% of data missing or below the limit of detection. Results Patients: 309 subjects were randomized in COMFORT-I with median age 68 (range 40-91), 46% female, 50% primary myelofibrosis, and 61% high risk. All 309 subjects had cytokines measured at one or more of the three visits included in this analysis, with 308 having cytokine values paired with a TSS score at the same visit. Cytokines at Baseline: At baseline, the highest Spearman correlations with symptomatic burden (as assessed by the TSS) were observed for APOA1 (rho=-.21) and FERRITIN (rho=-.20), followed by INTLK5, MIP1A, MMP3, INTLK2, MGB, INTLK1A, and INTLK7 with correlations between -.15 and -.17 (all p<0.01). After adjusting for age, sex, and BMI, VCAM1 and APOA1 were significantly associated with TSS at baseline. Cytokine and Symptom changes during trial: Changing levels of 5 cytokines were significantly associated with change of TSS over time beyond the change due to visit, arm, visit-by-arm interaction, age, sex and BMI including VCAM1, LEPTIN, TIMP1, B2MICG, and TNFRII. Within the placebo arm, only 5 cytokines significantly changed over time compared to 46 cytokines in the ruxolitinib arm (visit-by-arm interaction was significant for 43 in the overall models). VCAM1, B2MICG, and TNFRII were among the 5 cytokines which changed in the placebo arm, and VCAM1, LEPTIN, TIMP1, B2MICG, and TNFRII were among the 46 cytokines which changed in the ruxolitinib arm. No baseline cytokines and no changes in cytokines at week 4 univariately predicted week 24 TSS response within either arm at the p<0.001 level. Conclusions In the first serial assessment of MF symptoms and plasma cytokine levels in the conduct of a placebo controlled trial we found 5 cytokines in which improved levels correlated with decreased MF symptom burden in ruxolitinib treated patients. Development of a multivariate model for predicting symptom response, correlations with splenic response and impact on survival benefit of therapy is ongoing. Disclosures: No relevant conflicts of interest to declare.

Abstract BACKGROUND: Most cancer patients with bone metastases report significant pain and substantial impairment in their ability to function. In particular, patients with MM are susceptible to bone pain because of bone damage. Patient-reported outcomes such as a general health outcome measure and a legacy pain instrument may provide valuable information in determining the impact of this condition. The goal of this study was to provide a descriptive profile of health status for cancer patients using the EQ-5D-3L (EQ-5D), a commonly used tool to measure health utility scores. Another goal is to evaluate the predictive information provided by the Brief Pain Inventory (BPI), a self-report pain assessment tool designed to capture pain severity and interference in patients. These information may be useful in estimating the health status of patients where the EQ-5D was not administered. METHODS: Data was pooled across three denosumab registrational trials '20050136', '20050103', and '20050244'. Patients completed the EQ-5D and the BPI at baseline. Eligible patients had histologically confirmed advanced cancers, radiographic evidence of at least one bone metastasis (or lytic bone lesion from multiple myeloma), ECOG of 2 or better and adequate organ function. Using data from patients with MM (N=168), descriptive statistics were used to present summary information on health-related quality of life states and utility scores. Linear regressions were used to examine how BPI severity and interference subscale scores relate with EQ-5D utility scores. Ordinal regressions were used to evaluate how well BPI items predict EQ-5D individual items. Bonferroni adjustments were made when evaluating the relationships of BPI items to individual EQ-5D items. These analyses were repeated using data across several groups of cancer patients to determine if similar results can be obtained (breast cancer, N=2,044, prostate cancer, N=1,819 and others, N=1,509). RESULTS: About 9% of MM patients reported perfect health compared with 11% of all cancer patients. Of the 5 dimensions measured by the EQ-5D, the MM cohort rated pain (80%) the highest with moderate to extreme problems. This was followed by usual activities (68%), mobility (63%), anxiety/depression (49%), and self-care (36%). Similar order was observed across all cancer groups, but with slightly lower percentages for most items: pain (78%), usual activities (58%), mobility (55%), anxiety/depression (57%), and self-care (26%). Although pain interference subscale score explained more of the variability in the EQ-5D utility scores compared with pain severity items in the MM cohort (39% vs 31%) and for all patients (41% vs 34%), model fit was less than ideal (mean squared error=0.23 - 0.25). Hence, BPI items were fitted into individual EQ-5D items. Ordinal regression models fitting BPI items into individual EQ-5D items for the MM cohort showed good predictive power as measured by the concordance index c for mobility (86%), self-care (80%), usual activities (83%), pain (90%) and anxiety/depression (78%). For all patients, similar model fit were obtained for mobility (83%), self-care (78%), usual activities (81%), pain (90%) and anxiety/depression (72%). Further examination of the ordinal regression models across all groups showed that three BPI pain severity items, 'pain at its worst', 'average pain' and 'pain now' in addition to interference with activity and interference with work significantly predicted the EQ-5D pain item. Five interference items, activity, walking ability, work, sleep and enjoyment of life significantly predicted the EQ-5D usual activities item. Finally, 'pain now' and four interference items, relationship with others, enjoyment of life, mood, and activity significantly predicted the EQ-5D anxiety/depression item. CONCLUSIONS: Patients with MM reported several functional limitations on the EQ-5D. Majority of the patients reported moderate to extreme problems in all EQ-5D dimensions except self-care. Model fit of ordinal regression models relating BPI items with EQ-5D items are reasonable underscoring the benefit of using the BPI for planning patient care and health status evaluation. This consistency was also found across cancer groups that primarily include breast and prostate. The resultant model demonstrates the significant contribution of pain severity and interference in determining health status in this patient population. Disclosures Mendoza: Amgen Inc.: Consultancy. Shi:Amgen Inc.: Consultancy. Ma:Amgen Inc.: Employment. Zhang:Amgen Inc.: Employment. Qian:Amgen Inc.: Employment. Cleeland:Amgen Inc.: Consultancy.

Background There is currently a lack of consensus on how health-related quality of life and other patient-reported outcome measures in cancer randomized clinical trials are analyzed and interpreted. This makes it difficult to compare results across randomized controlled trials (RCTs) synthesize scientific research, and use that evidence to inform product labeling, clinical guidelines, and health policy. The Setting International Standards in Analyzing Patient-Reported Outcomes and Quality of Life Endpoints Data for Cancer Clinical Trials (SISAQOL) Consortium aims to develop guidelines and recommendations to standardize analyses of patient-reported outcome data in cancer RCTs. Methods and Results Members from the SISAQOL Consortium met in January 2017 to discuss relevant issues. Data from systematic reviews of the current state of published research in patient-reported outcomes in cancer RCTs indicated a lack of clear reporting of research hypothesis and analytic strategies, and inconsistency in definitions of terms, including “missing data,”“health-related quality of life,” and “patient-reported outcome.” Based on the meeting proceedings, the Consortium will focus on three key priorities in the coming year: developing a taxonomy of research objectives, identifying appropriate statistical methods to analyze patient-reported outcome data, and determining best practices to evaluate and deal with missing data. Conclusion The quality of the Consortium guidelines and recommendations are informed and enhanced by the broad Consortium membership which includes regulators, patients, clinicians, and academics.

AbstractBackgroundWe aimed to provide a simple, descriptive health‐status profile for cancer patients with bone metastases, based on the EuroQol EQ‐5D, a tool commonly used to measure health utility scores, and to evaluate its association with the Brief Pain Inventory (BPI), a legacy pain‐assessment tool. Although pain is one of five health‐status dimensions measured by the EQ‐5D, our understanding of how pain relates to the other EQ‐5D dimensions is limited.MethodsWe derived data from 5500 patients with bone metastases who completed the EQ‐5D and BPI. Regression analyses examined how BPI severity and interference scores correlated with EQ‐5D utility scores and how BPI items associated with EQ‐5D items, for the entire sample and by disease‐type subgroup.ResultsRegardless of cancer site, the percentage of patients reporting moderate/severe problems in each of the five EQ‐5D dimensions were pain/discomfort, 78%; usual activities, 58%; mobility, 55%; anxiety/depression, 57%; and self‐care, 26%. BPI pain interference explained more of the variability in the EQ‐5D utility scores than did pain severity (R2 = 41% vs. 34%). BPI worst pain, average pain, pain now, interference with general activity, and interference with work significantly predicted EQ‐5D pain/discomfort, with odds ratio estimates <1.ConclusionsPain/discomfort was the worst‐rated dimension of the EQ‐5D in this population, but the relationship of this item to BPI pain severity was modest, suggesting that the single pain item of the EQ‐5D may be of limited utility in studies for which pain is an endpoint.SignificanceHealth‐status dimensions include more than pain. We examine the contribution of pain severity and pain‐related functional interference in determining the health status of cancer patients with bone metastases. The pain dimension from a health‐status measure may be an inadequate metric in clinical trials/clinical practice when pain is an important outcome.

Aims: The US National Cancer Institute recently developed the PRO-CTCAE (Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events). PRO-CTCAE is a library of questions for clinical trial participants to self-report symptomatic adverse events (e.g. nausea). The objective of this study is to inform evidence-based selection of a recall period when PRO-CTCAE is included in a trial. We evaluated differences between 1-, 2-, 3-, and 4-week recall periods, using daily reporting as the reference. Methods: English-speaking patients with cancer receiving chemotherapy and/or radiotherapy were enrolled at four US cancer centers and affiliated community clinics. Participants completed 27 PRO-CTCAE items electronically daily for 28 days, and then weekly over 4 weeks, using 1-, 2-, 3-, and 4-week recall periods. For each recall period, mean differences, effect sizes, and intraclass correlation coefficients were calculated to evaluate agreement between the maximum of daily ratings and the corresponding ratings obtained using longer recall periods (e.g. maximum of daily scores over 7 days vs 1-week recall). Analyses were repeated using the average of daily scores within each recall period rather than the maximum of daily scores. Results: A total of 127 subjects completed questionnaires (57% male; median age: 57). The median of the 27 mean differences in scores on the PRO-CTCAE 5-point response scale comparing the maximum daily versus the longer recall period (and corresponding effect size) was −0.20 (−0.20) for 1-week recall, −0.36 (−0.31) for 2-week recall, −0.45 (−0.39) for 3-week recall, and −0.47 (−0.40) for 4-week recall. The median intraclass correlation across 27 items between the maximum of daily ratings and the corresponding longer recall ratings for 1-week recall was 0.70 (range: 0.54–0.82), for 2-week recall was 0.74 (range: 0.58–0.83), for 3-week recall was 0.72 (range: 0.61–0.84), and for 4-week recall was 0.72 (range: 0.64–0.86). Similar results were observed for all analyses using the average of daily scores rather than the maximum of daily scores. Conclusion: A 1-week recall corresponds best to daily reporting. Although intraclass correlations remain stable over time, there are small but progressively larger differences between daily and longer recall periods at 2, 3, and 4 weeks, respectively. The preferred recall period for the PRO-CTCAE is the past 7 days, although investigators may opt for recall periods of 2, 3, or 4 weeks with an understanding that there may be some information loss.