|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2000 American Society for Clinical Oncology Cognitive Function as a Predictor of Survival in Patients With Recurrent Malignant GliomaFrom the Departments of Neuro-Oncology and Biomathematics, The University of Texas M.D. Anderson Cancer Center, Houston, TX. Address reprint requests to Christina A. Meyers, PhD, Department of Neuro-Oncology, Box 100, The University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Blvd, Houston, TX 77030; email cameyers{at}mdanderson.org
PURPOSE: To determine the contribution of cognitive function in predicting the survival of patients with recurrent malignant brain tumors. PATIENTS AND METHODS: A total of 80 patients with recurrent glioblastoma multiforme or anaplastic astrocytoma were seen for baseline evaluations before beginning a phase I or phase II clinical trial. Each patient received a battery of nine brief tests measuring cognitive function, ability to perform activities of daily living (ADLs), and quality of life (QOL). Tests were given monthly after treatment was begun. RESULTS: Performance on a test of verbal memory was independently and strongly related to survival after accounting for age, Karnofsky performance status score, histology, and time since diagnosis. Models incorporating three of nine and all nine tests in the battery accounted for significantly more variance in survival than did the clinical variables alone. Measures of QOL and ADLs (bathing, feeding, and so on) were not independently related to survival, although they provide clinical information that is important for patient care. CONCLUSION: These results indicate that a multifaceted assessment of cognition, QOL, and patient function is practical for brain tumor patients in clinical trials and can provide information regarding the relative risks versus benefits of new treatment regimens that supplements the information from the usual clinical variables.
PRIMARY MALIGNANT brain cancer is a devastating illness, characterized by a low cure rate, short survival time, and significant morbidity as the disease progresses. Survival rates have improved over the past few years for patients with anaplastic astrocytoma (AA) (3-year median survival), although little progress has been made with glioblastoma multiforme (GBM) (1-year median survival).1 The major prognostic factors that predict survival and time to tumor progression are age of the patient, performance status (usually measured with the Karnofsky performance status [KPS] scale), and tumor grade.2 However, none of these assessments reflect the specific brain dysfunction caused by the tumor or its treatment. Brain dysfunction is manifested by neurologic and cognitive impairments. Impairments caused by the tumor itself will vary from patient to patient with the site of the lesion.3 In contrast, the impairments caused by treatment tend to be related to frontal-subcortical white matter dysfunction, including deficits of information-processing speed, frontal lobe executive functions, memory, sustained attention, and bilateral motor coordination.3-5 Assessments of brain dysfunction in brain tumor clinical trials, when performed at all, have been limited to brief screenings of global cognitive function, such as the Mini-Mental State Exam (MMSE).6 However, the MMSE was developed as a screening tool for dementia and is insensitive to mild cognitive impairments or focal lesions.7 The MMSE does not have well-established sensitivity or specificity,8 it does not measure many of the functions known to be impaired in brain tumor patients, and it does not have validated alternative forms for repeated testing. The MMSE has been found to be insensitive to changes in mental functioning (such as memory loss) in clinical trials of neurotoxic cancer drugs.9 In fact, a recent article describing a study that involved the MMSE reported that radiation therapy in patients with high-grade gliomas was not associated with cognitive decline10 despite the extensive literature contradicting this assertion. The use of such an insensitive tool might mean that patients with true disability resulting from cognitive impairments would not be identified and offered appropriate interventions. Furthermore, insensitive measures of cognition invariably correlate strongly with measures of performance status, providing no additional information for clinical trials beyond the usual prognostic indicators.6,10 To be practical, assessment of brain function in clinical trials must be brief, inexpensive, sensitive to change, and simple enough to be completed by most patients, even those with significant neurologic compromise. On the other hand, the assessment must be sufficiently comprehensive to be sensitive to the focal effect of tumors in various locations. In this study, we used a model proposed by the World Health Organization that includes objective measures of neurocognitive function (impairment), ability to conduct activities of daily living (ADLs) (disability), and quality of life (QOL) (handicap) in brain tumor patients undergoing phase I and phase II clinical trials for recurrent tumor.11 In this article we report the prognostic value of these assessments in predicting survival.
Patients A total of 80 patients with recurrent GBM or AA were seen for baseline evaluations before beginning postsurgical treatment. The demographic characteristics of the patients are listed in Table 1. The number of prior surgeries, extent of surgery, and the number of tumor recurrences before enrolling onto these studies are also reported. Tumor volume was not available for this group. However, our previous experiences suggest that tumor size is not significantly related to performance on cognitive tests.12 Evaluations of cognitive function, QOL, and ADLs were conducted as a part of nine different treatment protocols. All of the protocols had similar eligibility and exclusion criteria (eg, KPS score  60). All patients had prior radiation and chemotherapy before developing tumor recurrence and were tested after fully recovering from surgery, if one had been performed. The patients were accrued and tested between May 1995 and November 1997. Of the total group, 58 patients (73%) had at least one follow-up assessment.
Neuropsychologic Test Battery Patients received a pretreatment baseline evaluation and on-treatment follow-up before the next course of therapy (ie, monthly) until they left the study because of progressive disease. The tests selected are widely used, standardized psychometric instruments for assessing cognitive functions known to be affected by brain tumors and treatment. All have published normative data that take into account age and, where appropriate, education and sex. The tests were selected to minimize the effect of repeated administration. The memory test has six alternative forms, and the other tests measure motor and information-processing speed and are relatively resistant to the effects of practice. The tests were: (1) Digit Span13 to measure attention span; (2) Digit [  ]13[ ] for graphomotor speed; (3) Hopkins Verbal Learning Test14 for memory; (4) Controlled Oral Word Association15 for verbal fluency; (5) Trail Making Test Part A7 for visual-motor scanning speed; (6) Trail Making Test Part B7for executive function; (7) Grooved Pegboard7 for motor speed and dexterity; (8) Functional Assessment of Cancer Therapy with brain tumor specific module (FACT-BR)16-17 for QOL; and (9) Functional Independence Measure (FIM)18 for ADLs. Those tests that are scored in terms of seconds to complete (Trails A & B and Pegboard right and left) were always discontinued at 300 seconds to reduce the time and burden to the patient. Thus, no test took more than 5 minutes to administer, and the entire battery generally took 40 minutes or less to complete, demonstrating practicality in terms of cost, repeatability, and burden to patient.
Statistical Analysis
The overall median survival time for the group was 35 weeks (95% confidence interval [CI], 30 to 53 weeks). The 26-week survival rate was 66% (95% CI, 56% to 78%); the 52-week survival rate was 39% (95% CI, 29% to 51%); and the 78-week survival rate was 22% (95% CI, 14% to 34%). The baseline scores on the neuropsychologic test battery are displayed in Table 2. The limited number of subjects precluded having sufficient statistical power to correlate the results of the nine tests with the clinical variables of histology, site of tumor, age, and KPS score. However, a relation seemed to exist between the Trails B score and histology (median value for AA patients was 96 seconds v 212 seconds for GBM patients). This suggests that patients with the most malignant tumors also had poorer executive function. In addition, age is known to effect performance on most neuropsychologic tests and is accounted for in the published normative data used to interpret a given individual’s score.
Table 3 displays the results of the univariate Cox analyses for the nine cognitive tests, in which each test was treated as a continuous covariate (with a linear effect on the log hazard ratio). This analysis revealed that baseline performance on seven of the nine variables was statistically related to survival. Table 3 also shows that after adjustment for the four traditional prognostic variables (age, KPS score, histology, and diagnosis-to-test interval), performance on the memory test continued to be highly related to survival. Table 4 displays results from the multivariate Cox model. Of the clinical variables, only histology and number of recurrences was related to survival. Extent of resection, age, KPS, time since diagnosis, and number of previous surgeries were not statistically related to survival, whereas a number of the cognitive tests were.
When the models with the lowest AIC are analyzed, the four clinical prognostic variables accounted for 34% of the variance in survival time. The addition of the nine cognitive test scores to the four clinical variables accounted for 53% of the variance, a significantly greater proportion. The three tests in this analysis that were most strongly related to survival after accounting for the clinical variables were the Hopkins Verbal Learning Test (memory), Digit Span (attention span), and Digit Symbol (graphomotor speed). A model using the clinical variables and only those three tests accounted for 49% of the variance in survival, which is also significantly better than the clinical variables alone. The ability to perform basic ADLs (as assessed by the FIM) was not related to survival time when the score was adjusted for the clinical covariates. Similarly, QOL as assessed by the FACT-BR was unrelated to survival.
In this study, cognitive function was a unique prognostic factor in predicting survival in patients with recurrent malignant glioma. Specifically, performance on a test of verbal memory was independently and strongly related to survival after accounting for age, KPS score, histology, extent of resection, number of recurrences, and time since diagnosis. Models incorporating three of nine and all nine tests in the battery accounted for significantly more variance in survival than the clinical variables alone. Measures of QOL and ability to perform basic ADLs (bathing, feeding, and so on) were not independently related to survival, although they provide clinical information that is important for patient care. These findings bring up several interesting questions for further study. The assessment of subjective QOL in patients with neurologic deterioration is difficult. Many patients have a diminished appreciation for their current circumstances and lack insight into their deficits. Thus, the patient may experience a good QOL even though observers might infer that QOL should be poor. Twelve of our patients were unable to complete the FACT-BR at their baseline assessment. There was no difference in survival or KPS score between patients able or not able to complete the QOL assessment, but those patients who could not complete the FACT-BR had worse cognitive scores than did those who could complete it. In essence, it takes grossly intact cognitive functioning for a patient to adequately assess QOL issues, so it is not surprising that QOL scores are not strongly related to outcome in this population. The ability to perform basic ADLs (as assessed by the FIM) was modestly correlated with survival but not after adjustment for the other clinical factors. This suggests that ability to perform basic self-care activities was fairly well preserved in this population, and that marked reduction in these abilities occurred late in the disease. Thus, ability to perform basic self-care activities alone may not be a good indicator of change during brain tumor clinical trials. The relation between cognitive functioning and survival suggests that cognitive tests are relatively sensitive measures of the functioning of the brain. Brain function can be altered before anatomic evidence of change. The other clinical variables either indirectly assess brain function (eg, the KPS score), address aspects of the host (eg, age), or they assess aspects of the tumor and not the brain (eg, histology strongly reflects the tumor’s growth potential). Thus, a combination of tumor prognostic variables and brain function assessments seem to predict survival better than tumor variables alone. Multiple end points in brain cancer clinical trials may provide more information than the traditional end points of survival and progression-free survival. We are currently studying the relationship between change in performance on cognitive tests over time and evidence from magnetic resonance imaging of tumor progression. Our previous preliminary report found that such cognitive decline occurred nearly a month in advance of magnetic resonance imaging evidence of tumor recurrence, whereas patient QOL did not decline until well after tumor progression, and patient ability to perform basic self-care activities was even less affected.21 These results indicate that a multifaceted assessment that includes cognition, QOL, and ADLs of brain tumor patients in clinical trials can be practical and can provide additional information regarding the relative risks versus benefits of new treatment regimens, that will be more useful than tracking only the usual clinical variables. Use of these assessments will also allow for a more complete characterization of individual patient function that will help the treating physicians optimize total care and offer appropriate behavioral and pharmacologic interventions.
1. Levin VA, Leibel S, Gutin PH: Neoplasms of the central nervous system, in DeVita VT Jr, Hellman S, Rosenberg SA (eds): Cancer: Principles and Practice of Oncology (ed 5). Philadelphia PA,Lippincott-Raven, 1997, pp 2022-2082 2. Medical Research Council Brain Tumour Working Party: Prognostic factors for high-grade malignant glioma: Development of a prognostic index. J Neurooncol 9:47-55, 1990[Medline] 3. Scheibel RS, Meyers CA, Levin VA: Cognitive dysfunction following surgery for intracerebral glioma: influence of histopathology, lesion location, and treatment. J Neurooncol 30:61-69, 1996[Medline] 4. Archibald YM, Lunn D, Ruttan LA, et al: Cognitive functioning in long-term survivors of high-grade glioma. J Neurosurg 80:247-253, 1994[Medline] 5. Grant R, Slattery J, Gregor A, et al: Recording neurological impairment in clinical trials of glioma. J Neurooncol 19:37-49, 1994[Medline] 6. Choucair AK, Scott C, Urtasun R, et al: Quality of life and neuropsychological evaluation for patients with malignant astrocytomas: RTOG 91-14. J Radiat Oncol Biol Phys 38:9-20, 1997 7. Lezak MD : Neuropsychological Assessment (ed 3). New York NY,Academic Press, 1995 8. Wade DT: Measurement in Neurological Rehabilitation. New York NY,Oxford University Press, 1992 9. Meyers CA, Kudelka AP, Conrad CA, et al: Neurotoxicity of CI-980, a novel mitotic inhibitor. Clin Cancer Res 3:419-422, 1997[Abstract] 10. Taylor BV, Buckner JC, Cascino TL, et al: Effects of radiation and chemotherapy on cognitive function in patients with high-grade glioma. J Clin Oncol 16:2195-2201, 1998[Abstract] 11. World Health Organization: International Classification of Impairments, Disabilities, and Handicaps. Geneva, Switzerland, WHO, 1980 12. DeWinter A, Meyers C, Hannay HJ, et al: The effects of brain tumor growth rate on neuropsychological test performance. Neuropsychol Soc 2:65, 1996 (abstr) 13. Wechsler D: Wechsler Adult Intelligence Scale (Revised). San Antonio TX,The Psychological Corp, 1981 14. Benedict RHB, Schretlen D, Groninger L, et al: Hopkins Verbal Learning Test (revised): Normative data and analysis of inter-form and test-retest reliability. Clin Neuropsychol 10:43-55, 1998 15. Benton AL, Hamsher KdeS: Multilingual Aphasia Examination. Iowa City IA,AJA Associates, 1989
16.
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 11:570-577, 1993 17. Weitzner MA, Meyers CA, Gelke CK, et al: The functional assessment of cancer therapy (FACT) scale: Development of a brain subscale and revalidation of the FACT-G in the brain tumor population. Cancer 75:1151-1161, 1995[Medline] 18. Linacre JM, Heinemann AW, Wright BD, et al: The structure and stability of the Functional Independence Measure. Rehabil 75:127-132, 1994
19.
Atkinson AC: A note on the generalized information criterion for choice of a model. Biometrika 67:413-418, 1980
20.
Nagelkerke NJD: A note on a general definition of the coefficient of determination. Biometrika 78:691-692, 1991 21. Meyers CA, Grous JJ, Ford KM, et al: Multifaceted models for assessing quality of life in brain cancer therapy trials. Inform J 30:856-857, 1996 (abstr) accepted September 16, 1999.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2000 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
