|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 36 (December 20), 2007: pp. 5731-5737 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2007.11.1476 Prognostic Value of Health-Related Quality-of-Life Data in Predicting Survival in Patients With Anaplastic Oligodendrogliomas, From a Phase III EORTC Brain Cancer Group Study
From the European Organisation for Research and Treatment of Cancer, Quality of Life Unit, Data Center, and the Data Center Brain Group, Brussels, Belgium; Medical Center Haaglanden/Westeinde Hospital, Den Haag; Erasmus University Medical Center, Rotterdam; Canisius Wilhelmina Ziekenhuis, Nijmegen; St Elisabeth Ziekenhuis, Tilburg, the Netherlands; Centre Hospitalier Universitaire Pitié-Salpêtrière, Paris; Centre Antoine Lacassagne, Nice, France; and Azienda Sanitaria Locale Ospedale Bellaria-Maggiore, Bologna, Italy Address reprint requests to Murielle Mauer, PhD, European Organisation for Research and Treatment of Cancer Data Center, Quality of Life Unit, Ave Mounier 83/11, Brussels, Belgium 1200; e-mail: murielle.mauer{at}eortc.be
Purpose This is one of a few studies that have explored the value of baseline symptoms and health-related quality of life (HRQOL) in predicting survival in patients with brain cancer. Patients and Methods Baseline HRQOL scores (from the European Organisation for Research and Treatment of Cancer [EORTC] Quality of Life Questionnaire C30 and the EORTC Brain Cancer Module) were examined in 247 patients with anaplastic oligodendrogliomas to determine the relationship with overall survival by using Cox proportional hazards regression models. Refined techniques as the bootstrap resampling procedure and the computation of C indexes and R2 coefficients were used to explore the stability of the models as well as better assess the potential benefit of using HRQOL to predict survival in clinical practice and research. Results Classical analysis controlled for major clinical prognostic factors selected emotional functioning (P = .0016), communication deficit (P = .0261), future uncertainty (P = .0481), and weakness of legs (P = .0001) as statistically significant prognostic factors of survival. However, several issues question the validity of these findings and no single model was found to be preferable over all others. C indexes, which estimate the probability of a model to correctly predict which patient among a randomly chosen pair of patients will survive longer, and R2 coefficients, which measure the proportion of variability explained by the model, did not exhibit major improvement when adding selected or all HRQOL scores to clinical factors. Conclusion While classical techniques lead to positive results, more refined analyses suggest that baseline HRQOL scores add relatively little to clinical factors to predict survival. These results may have implications for future use of HRQOL as a prognostic factor for patients with cancer.
It has become increasingly accepted that, in addition to the traditional assessment of clinical outcomes, health-related quality-of-life (HRQOL) information can play a key role in cancer research and help individual patient care.1,2 Patients' self assessment of HRQOL is now an established end point for treatment comparisons in many cancer disease sites3 particularly in advanced disease.4,5 The use of HRQOL outcomes could have various applications, including supporting clinical decision making by providing the patient's perspective or providing prognostic information. Recent studies have shown that HRQOL parameters can be independent prognostic factors for survival in cancer patients.6 If HRQOL parameters are independent predictors of survival, they could be used in daily practice to identify patients who will benefit from a specific intervention; further, it could prevent overtreatment of patients who will gain no benefit from often toxic and aggressive therapies or to set up more tailored psychosocial intervention programs aimed at improving patients' HRQOL. Furthermore, they could be used to better stratify patients in future cancer clinical trials, hence better interpreting study outcomes, or to identify critical areas that could help in the selection of key end points for future clinical trials. HRQOL prognostic factor analyses have been carried out on several different cancer populations including, among others, lung,7-10 esophageal,11,12 advanced breast,4,13 metastatic malignant melanoma,14 metastatic bladder cancer,15 and head and neck16 cancers, highlighting the importance HRQOL scores may have in predicting survival. Only three studies have examined HRQOL and/or cognitive functioning as a prognostic factor in brain cancer.17-19 This study evaluates the prognostic value of HRQOL data collected from a prospective, large scale international randomized controlled trial, by using various statistical techniques in an attempt to provide robust conclusions on the prognostic value of HRQOL in patients with anaplastic oligodendrogliomas.
Treatment In this international, multicenter study (European Organisation for Research and Treatment of Cancer [EORTC] trial 26951), adult patients with newly diagnosed and histologically proven anaplastic oligodendroglioma or oligoastrocytoma were randomly assigned to treatment with radiation therapy (RT) only or radiation therapy followed by six cycles of procarbazine, lomustine, and vincristine (PCV) chemotherapy. Patients were stratified for institution, performance status (WHO performance status 0 or 1 v 2), age (> 40 or  40), the extent of the resection at surgery (biopsy only v debulking surgery/resection), and possible prior surgery for a low-grade oligodendroglioma (yes or no). The details on trial conduct and clinical outcome have already been reported.20 A central pathologist (J.M. Kros) reviewed all tumor samples and scored the presence or absence of tumor necrosis. The presence or absence of loss of the short arm of chromosome 1 (1p loss) was assessed in a subset of patients with fluorescence in situ hybridization. The trial was approved by the EORTC protocol review committee and the institutional review board of each participating center. All patients provided written informed consent.
Procedures The EORTC QLQ-C30 measure comprises five functioning scales: physical, role, emotional, cognitive, and social; three symptom scales: fatigue, nausea/vomiting, and pain; six single item scales: dyspnoea, insomnia, appetite loss, constipation, diarrhea, and financial impact; and the overall health/global QOL scale. The EORTC QLQ-BN20, designed for use with patients undergoing chemotherapy or RT, includes 20 items assessing visual disorders, motor dysfunction, communication deficit, various disease symptoms (eg, headaches and seizures), treatment toxicities (eg, hair loss), and future uncertainty. The items on both measures were scaled and scored using the recommended EORTC procedures.24 Raw scores were transformed to a linear scale ranging from 0 to 100, with a higher score representing a higher level of functioning or higher level of symptoms. Provided at least one half of the items in the scale were completed, the scale score was calculated using only those items for which values existed. Patients were randomly assigned after surgery and before the start of the RT. RT started within 4 to 6 weeks after surgery. Valid HRQOL assessments were performed at baseline, not more than 6 weeks before or after random assignment, more than 7 days after surgery, and before the start of RT. Follow-up assessments were performed at regular intervals (ie, at the end of RT, then every 3 months for the first year after RT, and then at 6-monthly intervals until recurrence of the disease). HRQOL was a mandatory aspect of this clinical trial protocol. The protocol stipulated that a responsible nurse, clinician, or data manager administered the questionnaire, requesting completion and its return to the EORTC Data Center. EORTC guidelines for administering questionnaires were provided, ensuring a standard approach and optimal compliance of HRQOL data collection by all personnel.
Statistical Analysis
The Cox proportional hazards regression model25 with overall survival measured from time of random assignment as dependent outcome was used for both univariate and multivariate analyses. A Collett's Model selection approach26 was used with a level of significance of 0.15 for the univariate screening and stay and entry criterions of 0.05. The HRQOL scales were included as continuous factors. The model was controlled for the major established prognostic baseline clinical factors—age (< 45; 45 to 54; and
Validation of the final model was undertaken by use of several refined statistical techniques. The stability of the final model was investigated using a bootstrap resampling procedure as proposed by Sauerbrei and colleagues,27 applied in the context of HRQOL.28 This technique generates a number of samples (each the same size as the original data set), by randomly selecting patients and replacing them before selecting the next patient (ie, bootstrap resampling). The frequency of inclusion of the HRQOL scores in the Cox proportional hazards regression models, including all the selected clinical factors and treatment, fitted to each of these data sets using automatic forward stepwise selection (entry level of Deviance residuals from the model with clinical factors were plotted versus the HRQOL scores to explore the relationship between each HRQOL score and the remaining part of the hazard not already explained by clinical factors. Finally, discrimination C indexes and Nagelkerke's R2 coefficients were computed to quantify the predictive accuracy of a model. All data analyses were performed using SAS, version 9 (SAS Institute, Cary, NC).
Between August 1996, and March 2002, 368 patients from 40 institutions in 10 countries were randomly assigned to receive either RT alone (n = 183) or RT plus chemotherapy (n = 185; Fig 1). Of these patients, 288 had baseline HRQOL measures completed (150 in RT and 138 in RT plus adjuvant chemotherapy). Genetic status (loss of the short arm of chromosome 1 (1p) and the long arm of chromosome 19 (19q)) was assessed in 311 patients (156 in RT and 155 in RT plus chemotherapy). Only the 247 patients with both baseline HRQOL measures and with genetic status assessed were included in the analysis (128 in RT and 119 in RT plus chemotherapy), which represents 67.1% of the original sample size.
Clinical Results The clinical results have recently been published.20 In brief, the study demonstrated that immediate postradiotherapy treatment with PCV leads to an increase in progression-free survival compared with RT alone (median progression-free survival of 23 v 13.2 months; P = .002 by the log-rank test). The median survival was 40.3 months in the RT plus PCV group compared with 30.6 months in the RT group (P = .23). Eighty-two percent of patients in the RT arm received chemotherapy at the time of progression, mostly PCV. Combined loss of 1p/19q identified a prognostic favorable subgroup of oligodendroglial tumors, regardless of treatment. No genetic subgroup could be identified that benefited with respect to overall survival from adjuvant PCV. Baseline clinical characteristics for patients with a valid baseline HRQOL questionnaire and with 1p/19q loss assessed are depicted in Table 1.
The distribution of the baseline clinical characteristics was very similar between patients with a valid baseline HRQOL questionnaire and those without and between patients with 1p/19q loss assessed and those without.
Prognostic Factor Analysis Results
Multivariate analysis. The Cox multivariate model selected by Collett's model selection approach contained emotional functioning, communication deficit, future uncertainty, and weakness of legs, in addition to the selected clinical factors (Table 3). However, the signs of the coefficients related to emotional functioning and communication deficit were opposite to what was expected (ie, worse emotional functioning and more communication deficit were related to better survival).
Bootstrap Resampling Procedure Table 4 presents the results of the bootstrap resampling procedure. The inclusion frequencies higher than 50% were related to emotional functioning (58.1%) and weakness of legs (89%). The frequencies of selection of each possible set of HRQOL scores were very low. The most frequently selected model (emotional functioning, future uncertainty, and weakness of legs) was selected only 3.4% of the time. This indicates a high degree of model instability with no single model to be uniformly preferable over all others.
Plots of the Residuals to Explore Trends The graphical exploration of the residuals suggested several issues. First, due to the very low number of patients with extreme values for some of the HRQOL scores, the results can be influenced by a few outliers (eg, for emotional functioning, the positive trend could be driven by the only four patients with a score of 0 but who actually survived considerably longer than most patients). It also suggested that future uncertainty was entered in the model as a correction for emotional functioning. Finally, the plots indicate a high variability in terms of survival between patients with fixed levels of the HRQOL scores.
Discrimination Index C and Nagelkerke's R2 coefficient
Anaplastic oligodendroglioma is an incurable disease that, like any other malignant brain tumor, has considerable impact on the patient's HRQOL. In this study, we aimed to explore if HRQOL data could provide reliable and useful prognostic information. Our study examined 247 patients with anaplastic oligodendrogliomas, controlled for major prognostic clinical factors, and has attempted to overcome limitations of past studies by using a reasonable sample size and sophisticated statistical methodology. Previous studies in patients with brain tumor showed that HRQOL factors and/or cognitive functioning were statistically significant factors in prediction models for groups of patients, comparable with our classical analysis. For example, Sehlen et al17 examined HRQOL in 153 patients with either malignant astrocytoma or brain metastases. Using the Functional Assessment of Cancer Therapy General (FACT-G) HRQOL measure,29 they found two variables (ie, living with a spouse and the FACT-G total score) to predict survival. Two other studies demonstrated objective cognitive functioning to have prognostic significance, both in newly diagnosed and in recurrent high-grade glioma.18,19 Meyers et al18 examined HRQOL and cognitive functioning in 80 patients with recurrent malignant glioma or anaplastic astrocytoma, at baseline, before treatment in phase I and II trials. HRQOL was undertaken with the FACT-Brain module, along with other neuropsychological tests. HRQOL scores did not predict survival, but it was found that cognitive functioning was a significant predictor of survival. It is difficult to compare these findings with our results, given the different measures that were used, along with their sample being relatively small. In addition, the phase I/II setting of Meyers et al is likely to be considerably different (higher expectations and discounting toxicities) to that of a large phase III trial.30 Klein et al19 explored cognitive functioning along with activities of daily living in 68 newly diagnosed high-grade glioma patients. Cognitive functioning was found to have prognostic value, but only in a subsample of older patients. However, it is unclear to what extent studies on such small samples can be relied on for providing definitive conclusions. It is also difficult to make comparisons between our trial and Klein et al due to the different HRQOL measures employed. However, several issues question the validity and the reliability of the results obtained by classical techniques. Some of them are well known—the large number of HRQOL scales and the intercorrelation of these HRQOL scales. It makes the selection of a particular set of HRQOL scores quite difficult as various sets of HRQOL scores may predict equally survival when added to clinical factors. It may also lead to models difficult to interpret (with worst HRQOL status associated with longer survival) as some HRQOL factors may enter the model just as corrections for others. In addition, as HRQOL scores are analyzed as continuous factors, the results could be influenced by a few outliers (ie, patients with some very bad HRQOL scores but who actually survived long or vice versa). Furthermore, the residuals plots also suggested a high variability in duration of survival among patients who have a same level of the HRQOL scores. C indexes and R2 coefficients are thought to better assess the potential benefit of using baseline HRQOL scores in addition to the clinical factors to predict survival in clinical practice and research. These coefficients did not exhibit major improvement when adding selected or all HRQOL scores to clinical factors, suggesting that baseline HRQOL scores in the end add relatively little to clinical factors to predict survival and cannot replace them. Knowing that a prognostic model with many factors becomes too complex, is not easy to use and interpret, we have also to take into account in the comparison of different prognostic models the number of factors in the model. The objective is to find a model which predicts the best overall survival with as few parameters as possible. In our study, the C index of the model with only clinical baseline characteristics (five factors plus treatment) is C = 0.743 and the Nagelkerke's R2-coefficient is R2 = 0.348. For the model without clinical factors but with all preselected HRQOL scores (17 factors), C = 0.653 and R2 = 0.191. For the model with all preselected HRQOL scores and treatment, C = 0.654 and R2 = 0.195. Clinical factors alone better predict survival than HRQOL factors with less factors. When adding clinical factors to all HRQOL scores, C = 0.769 and R2 = 0.436. The increase in C index and especially in R2 coefficient is larger when adding clinical factors to all HRQOL scores (0.653 v 0.769 and 0.191 v 0.436) than when adding HRQOL scores to clinical factors (0.743 v 0.769 and 0.348 v 0.436). Care needs to be taken when interpreting the results of our study, given our study had limitations, particularly as this was an exploratory analysis. Also, while 247 patients represent a considerable sample, other data sets are required to validate these findings. In summary, while traditional methods of analysis suggest HRQOL data are prognostic, more detailed analysis revealed these findings may not be as reliable as expected. Further research should investigate the use of HRQOL with more sophisticated techniques, to obtain reliable results. Further research could also investigate the prognostic value of changes from baseline in HRQOL rather than baseline values and should investigate why HRQOL parameters might be of value in one setting but not another.
The author(s) indicated no potential conflicts of interest.
Conception and design: Martin J.B. Taphoorn, Andrew Bottomley, Corneel Coens, Fabio Efficace, Alba A. Brandes, Martin J. van den Bent Administrative support: Andrew Bottomley, Corneel Coens, Denis Lacombe Provision of study materials or patients: Martin J.B. Taphoorn, Marc Sanson, Alba A. Brandes, Carin C.D. van der Rijt, Hans J.J.A. Bernsen, Marc Frénay, Cees C. Tijssen, Martin J. van den Bent Collection and assembly of data: Marc Sanson, Carin C.D. van der Rijt, Marc Frénay, Cees C. Tijssen, Denis Lacombe, Martin J. van den Bent Data analysis and interpretation: Murielle E.L. Mauer, Martin J.B. Taphoorn, Andrew Bottomley, Corneel Coens, Fabio Efficace, Alba A. Brandes, Denis Lacombe, Martin J. van den Bent Manuscript writing: Murielle E.L. Mauer, Martin J.B. Taphoorn, Andrew Bottomley, Fabio Efficace, Alba A. Brandes, Carin C.D. van der Rijt, Cees C. Tijssen, Martin J. van den Bent Final approval of manuscript: Martin J.B. Taphoorn, Andrew Bottomley, Corneel Coens, Fabio Efficace, Marc Sanson, Alba A. Brandes, Carin C.D. van der Rijt, Hans J.J.A. Bernsen, Marc Frénay, Cees C. Tijssen, Denis Lacombe, Martin J. van den Bent
We thank all the patients and investigators for their involvement, and Laurence Collette, PhD, European Organisation for Research and Treatment of Cancer, for helping and applying the refined statistical techniques to our data.
Supported in part by Grant Nos. 5U10CA11488-30 through 5U10CA11488-34 from the National Cancer Institute and by the European Organisation for Research and Treatment of Cancer (EORTC) Brain Cancer Group; also supported by the EORTC Charitable Trust. Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
1. Velikova G, Booth L, Smith AB, et al: Measuring quality of life in routine oncology practice improves communication and patient well-being: A randomized controlled trial. J Clin Oncol 22:714-724, 2004 2. Detmar SB, Muller MJ, Schornagel JH, et al: Health-related quality-of-life assessments and patient-physician communication: A randomized controlled trial. JAMA 288:3027-3034, 2002 3. Bottomley A, Vanvoorden V, Flechtner H, et al: The challenge and achievements of implementation of quality of life research in cancer clinical trials. Eur J Cancer 39:275-285, 2003[CrossRef][Medline] 4. Efficace F, Biganzoli L, Piccart M, et al: Baseline health-related quality-of-life data as prognostic factors in a phase III multicentre study of women with metastatic breast cancer. Eur J Cancer 40:1021-1030, 2004[CrossRef][Medline] 5. Bottomley A, Flechtner H, Efficace F, et al: Health related quality of life outcomes in cancer clinical trials. Eur J Cancer 41:1697-1709, 2005[CrossRef][Medline] 6. Kramer JA, Curran D, Piccart M, et al: Identification and interpretation of clinical and quality of life prognostic factors for survival and response to treatment in first-line chemotherapy in advanced breast cancer. Eur J Cancer 36:1498-1506, 2000[CrossRef][Medline] 7. Montazeri A, Milroy R, Hole D, et al: Quality of life in lung cancer patients: As an important prognostic factor. Lung Cancer 31:233-240, 2001[CrossRef][Medline] 8. Kaasa S, Mastekaasa A, Lund E: Prognostic factors for patients with inoperable non-small cell lung cancer, limited disease: The importance of patients' subjective experience of disease and psychosocial well-being. Radiother Oncol 15:235-242, 1989[CrossRef][Medline] 9. Herndon JE, Fleishman S, Kornblith AB, et al: Is quality of life predictive of the survival of patients with advanced non-small cell lung carcinoma? Cancer 85:333-340, 1999[CrossRef][Medline] 10. Langendijk H, Aaronson NK, de Jong JM, et al: The prognostic impact of quality of life assessed with the EORTC QLQ-C30 in inoperable non-small cell lung carcinoma treated with radiotherapy. Radiother Oncol 55:19-25, 2000[CrossRef][Medline] 11. Fang FM, Tsai WL, Chiu HC, et al: Quality of life as a survival predictor for esophageal squamous cell carcinoma treated with radiotherapy. Int J Radiat Oncol Biol Phys 58:1394-1404, 2004[CrossRef][Medline] 12. Blazeby JM, Brookes ST, Alderson D: The prognostic value of quality of life scores during treatment for oesophageal cancer. Gut 49:227-230, 2001 13. Luoma ML, Hakamies-Blomqvist L, Sjöström J, et al: Prognostic value of quality of life scores for time to progression (TTP) and overall survival time (OS) in advanced breast cancer. Eur J Cancer 39:1370-1376, 2003[CrossRef][Medline] 14. Coates A, Thomson D, McLeod GR, et al: Prognostic value of quality of life scores in a trial of chemotherapy with or without interferon in patients with metastatic malignant melanoma. Eur J Cancer 29A:1731-1734, 1993[CrossRef] 15. Roychowdhury F, Hayden A, Liepa A: Health-related quality-of-life parameters as independent prognostic factors in advanced or metastatic bladder cancer. J Clin Oncol 21:673-678, 2003 16. de Graeff A, de Leeuw JR, Ros WJ, et al: Sociodemographic factors and quality of life as prognostic indicators in head and neck cancer. Eur J Cancer 37:332-339, 2001[CrossRef][Medline] 17. Sehlen S, Lenk M, Hollenhorst H, et al: Quality of life (QoL) as predictive mediator variable for survival in patients with intracerebral neoplasma during radiotherapy. Onkologie 26:38-43, 2003[Medline] 18. Meyers CA, Hess KR, Yung WK, et al: Cognitive function as a predictor of survival in patients with recurrent malignant glioma. J Clin Oncol 18:646-650, 2000 19. Klein M, Postma TJ, Taphoorn MJ, et al: The prognostic value of cognitive functioning in the survival of patients with high-grade glioma. Neurology 61:1796-1798, 2003 20. van den Bent MJ, Carpentier AF, Brandes AA, et al: Adjuvant procarbazine, lomustine, and vincristine improves progression-free survival but not overall survival in newly diagnosed anaplastic oligodendrogliomas and oligoastrocytomas: A randomized European Organisation for Research and Treatment of Cancer phase III trial. J Clin Oncol 24:2715-2722, 2006 21. Aaronson NK, Ahmedzai S, Bergman B, et al: The European Organisation for Research and Treatment of Cancer QLQ-C30: A quality-of-life instrument for use in international clinical trials in oncology. J Natl Cancer Inst 85:365-376, 1993 22. Osoba D, Aaronson NK, Muller M, et al: The development and psychometric validation of a brain cancer quality-of-life questionnaire for use in combination with general cancer-specific questionnaires. Qual Life Res 5:139-150, 1996[CrossRef][Medline] 23. Cull A, Sprangers M, Bjordal K, et al: EORTC Quality of Life Group Translation Procedure. Brussels, EORTC Publications, ISBN:2-930064-28-5, 2002 24. Fayers P, Aaronson N, Bjordal K, et al: EORTC QLQ-C30 Scoring Manual. Brussels, EORTC Publications, ISBN:2-930064-22-6, 2001 25. Cox DR: Regression models and life tables. J Royal Stat Soc 34:187-220, 1972 26. Collett D: Modeling Survival Data in Medical Research. London, United Kingdom, Chapman & Hall, 1994 27. Sauerbrei W, Schumacher M: A bootstrap resampling procedure for model building: Application to the Cox regression model. Stat Med 11:2093-2109, 1992[Medline] 28. Van Steen K, Curran D, Kramer J, et al: Multicollinearity in prognostic factor analyses using the EORTC QLQ-C30: Identification and impact on model. Stat Med 21:3865-3884, 2002[CrossRef][Medline] 29. Cella DF, Tulsky DS, Gray G, et al: The Functional Assessment of Cancer Therapy scale: Development and validation of the general measure. J Clin Oncol 11:570-579, 1993 30. Cheng JD, Hitt J, Koczwara B, et al: Impact of quality of life on patient expectations regarding phase I clinical trials. J Clin Oncol 18:421-428, 2000 Submitted February 16, 2007; accepted August 13, 2007.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2007 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
55 years), tumor location (frontal v other), performance status (0 v 1 v 2), tumor necrosis (yes v no), and 1p loss (yes v no) — as well as for treatment type by forcing their inclusion in all models. 
= .05), can be considered to be indicative for the importance of the HRQOL factors. This technique was applied to 1,000 bootstrap samples. 
