SEARCH

SEARCH BY CITATION

Keywords:

  • brain neoplasm;
  • neuropsychologic tests;
  • longitudinal studies;
  • forecasting (projections and predictions)

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

BACKGROUND

Neuropsychologic tests are widely used to predict the course of progressive neurologic diseases, and recent research has demonstrated the specificity of cognitive measures, even in relatively diffuse diseases. However, the cognitive effects of brain tumors of similar histology and location are known to be highly variable. The authors used the specificity of cognitive function principle to compare two models for the early detection of low-grade brain tumor recurrence prior to detection with clinically scheduled neuroimaging.

METHODS

To test the feasibility of these prediction models, 34 patients with supratentorial, low-grade brain tumors prospectively were administered serial comprehensive neuropsychologic examinations; 11 patients developed recurrent tumors during the series and 23 patients did not. A general model based on tests sensitive to malignancy and white matter disease was compared with a tumor-specific model based on indices related to each patient's tumor locus. A Cox proportional hazards model was used to identify the predictor variables that significantly changed immediately prior to recurrence.

RESULTS

Only the tumor-specific model achieved statistical significance (P < 0.02). A tumor-specific index decline of 1 standard deviation indicated a 5–fold increase in the probability of tumor recurrence.

CONCLUSIONS

Although this method needs to be tested with more frequent and regular observations and with a larger sample, these results provide evidence of the feasibility of the subject-specific model as a predictor of recurrence. The evidence of the predictive value of a tumor-specific model is consistent with studies that identify only limited, brain structure-specific cognitive decline from broad neuropsychologic batteries. Cancer 2003;97:649–56. © 2003 American Cancer Society.

DOI 10.1002/cncr.11099

Neuropsychologic testing has demonstrated predictive value in numerous neurologic disorders,1–3 including brain tumors, multiple sclerosis (MS), and dementia-related illnesses. Changes in neurologic disease status can be identified through variations in cognitive performance, arguably correlating with the degree of disease severity. Identifying neuropsychologic changes earlier in disease progression may lead to earlier therapeutic interventions for improved outcome. However, brain tumors are notoriously variable in their presentation and characteristic associated cognitive impairments. In this article, we present a pilot study to test whether neuropsychologic assessment is a viable diagnostic tool with which to detect early changes that are indicative of low-grade brain tumor recurrence.

Tumor recurrence may be a more sensitive predictor of cognitive impairment than tumor histology, location, or treatment effects,4, 5 but the cognitive impairments specific to individual tumor sites are quite variable and can be difficult to identify.6 Currently, to our knowledge only Meyers et al.2 have reported on the use of cognitive variables from a broad neuropsychologic battery to predict tumor recurrence or mortality. They assessed 80 malignant glioma patients (glioblastoma multiforme [GBM] and anaplastic astrocytoma [AA]) at monthly intervals using cognitive, daily performance, and quality of life measures after treatment. The results indicated that of the 11 cognitive indices used, the most predictive were 2 indices of verbal memory (recall and recognition of a word list), which were found to correlate positively with longer survival. There also was a trend for the GBM patients, who had the largest and most aggressive tumors, to achieve poor maintenance of cognitive set, exhibited by their performance on the Trail Making Test. Formal measures of daily performance and quality of life were unrelated to survival in this study.2

Measures of cognitive functioning have been associated with tumor location, histology (i.e., low grade vs. high grade), and treatment modality. The regionally specific effects of tumors on cognitive function may depend in part on the presence of adjuvant treatments such as irradiation and chemotherapy and on tumor size. Anderson et al.6 found that the deficits of patients with brain tumors, compared with those who had a stroke, were mild and highly variable. However, some studies have shown that neuropsychologic measures are associated specifically with lesion location.7, 8 Scheibel et al.8 conducted a cross-sectional study of 245 adult glioma patients who underwent surgery, with the majority receiving radiotherapy and/or chemotherapy. They investigated whether the tumor locus itself or malignancy and its associated neurologic complications accounted for the type of cognitive impairments found. Neuropsychologic tests across several cognitive domains revealed a strong relation between cognitive test and hemispheric tumor location. Regardless of the degree of malignancy, patients with left hemisphere lesions had language, verbal learning, and intellectual impairments and patients with right hemisphere lesions had poor visuoperceptual skills. Thus, current cognitive functioning can indicate tumor loci in surviving patients, substantiating neuropsychologic assessment as a predictor for tumor presence. Consequently, if a patient demonstrates neuropsychologic decline, it may be predictive of a change in tumor status.

The merits of a brief versus lengthy neuropsychologic approach for the prediction of disease exacerbation can be informed by research on MS - a multifocal neurologic disease that often affects the brain's white matter. Brief cognitive screening batteries (e.g., Mini-Mental State Examination) and functionality tests (e.g., Kurtzke Expanded Disability Status Scale) have been reported to be insensitive,9, 10 whereas comprehensive cognitive assessment can detect the nature and severity of deficits.10, 11 However, frequent comprehensive assessment, which often takes ≥ 6 hours, may not be feasible in longitudinal studies. Both comprehensive and brief neuropsychologic batteries have shown value in predicting pharmacologic treatment effects in patients with MS. Fischer et al.12 examined the neuropsychologic outcomes of interferon-β-1a treatment or placebo, giving comprehensive assessments prior to treatment and after a 2-year interval and brief neuropsychologic assessments every 6 months. The comprehensive battery demonstrated improved performance in specific components of cognition in the group treated with interferon-β-1a and the brief battery revealed significant between-group slopes of change over 6-month intervals, also suggesting treatment benefit. We postulated that a brief, tumor-specific method for predicting tumor recurrence would combine the specificity of comprehensive assessments to detect varying tumor effects with the sensitivity of brief cognitive batteries to detect subtle changes over time.

Brain tumor patients often are neurologically symptom-free,5 or have subtle and variable cognitive impairment,6 which may be attributable to the slow growth of the neoplasms and their infiltrative pathophysiology that disturb the function of neural tissue without destroying it.6, 7 For example, certain patients with slowly growing tumors, such as meningiomas, can be asymptomatic for several years and diagnosed incidentally,5 and gliomas can be silently quiescent in the brain for an unknown period prior to the detection of initial clinical symptoms. Therefore, clinical cognitive deterioration from brain tumors may not occur until tumor activity changes.

Our goal was to develop neuropsychologic assessment as a brief, inexpensive, and sensitive method of predictive measurement, yet one that was sufficiently comprehensive to detect anatomically heterogeneous patterns of disease. Predictive batteries of tests that are comprehensive can detect early stages of disease, and brief batteries often are used to predict the slope of change. However, the varying location of patients' brain tumors precludes a uniform brief assessment technique. Some studies have shown that even lengthy batteries for diffuse diseases such as dementia often demonstrate the sensitivity of a small number of cognitive indices.13, 14 The development of an effective predictive model focusing on patients with low-grade tumors has the advantage of developing the model's temporal sensitivity using frequent observations without the confounding effects of histology at baseline and brief survival rates.

We present herein the results of a pilot study that compared a concise, tumor-specific model of cognitive assessment with a general cognitive model, with the intention of increasing the sensitivity of cognitive assessment to the early recurrence of cerebral neoplastic disease, despite varying lesion location.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Patient Selection and Characteristics

Patients were adults with primary low-grade brain tumors who were part of a longitudinal study of the effects of radiotherapy on brain function. All patients agreed by informed consent procedures to be volunteer participants in the current study. Tumor types included supratentorial gliomas (70%), pineal (9%), meningiomas (9%), pituitary (6%), and other (6%), and were unselected in terms of standard risk of tumor recurrence. Patients were identified as recurred by the University of Pennsylvania Brain Tumor Center based on the neuroradiologist's interpretation of magnetic resonance imaging (MRI). Recurred patients were matched to the extent possible with nonrecurred controls from the same population, on the bases of tumor type (73% vs. 70% gliomas, respectively), treatment with radiotherapy (64% vs. 78%), timepoint after surgery, age (38.2 years vs. 42.6 years; P < 0.16), education (15.0 years vs. 14.8 years; P < 0.39), gender (54.6% vs. 50% Male, respectively; P < 0.75), and right hand dominance (90.9% vs. 78.3%, respectively; P < 0.86). Matching based on the extent of surgical resection was not possible (no surgery; 54.6% vs. 17.4%, respectively; and macroscopic total resection: 18.2% vs. 52.2%, respectively). The tumor types/grades (WHO I–IV) and treatments of the individual patients are shown in Table 1.

Table 1. Recurred and Nonrecurred Patient Tumor Type, Grade, and Treatments
Tumor locationTumor typeGrade (WHO I–IV)ResectionRadiotherapy
  1. R: right; macro: macroscopic; L: left.

Nonrecurred tumors    
R frontoparietalOligodendroglioma2Macro totalYes
L frontalOligodendroglioma2Macro totalYes
R parietalAstrocytoma2Macro totalYes
PituitaryAdenomaLow-gradeMacro totalYes
PinealPineocytoma2Macro totalYes
Posterior fossaEpendymoma2PartialYes
L temporalOligodendroglioma2Macro totalYes
L temporoparietalMixed glioma2PartialYes
L frontalAstrocytoma2PartialYes
L frontalPlasmocytoma1Macro totalYes
L frontalEpendymoma2Macro totalYes
R frontalAstrocytoma2Macro totalYes
R frontalAstrocytoma2PartialNo
R cavernous sinusMeningioma1NoneYes
L cavernous sinusMeningioma1PartialYes
PinealCystLow-gradeNoneNo
R frontalHemangiopericytoma2Macro totalYes
L parietalAstrocytoma2PartialNo
L frontalEpendymoma1Macro totalNo
R frontalMixed glioma1Macro totalNo
R frontalAstrocytoma2BiopsyYes
R temporalAstrocytoma2PartialYes
L frontalMixed glioma2BiopsyYes
Recurred tumors    
PinealPineocytomaLow-gradePartialYes
L frontotemporalMixed glioma2BiopsyYes
Corpus callosumAstrocytoma2BiopsyYes
L frontalAstrocytoma2PartialNo
R frontalAstrocytoma2BiopsyYes
L frontalOligodendroglioma2BiopsyYes
PituitaryAdenomaLow-gradeMacro totalNo
L frontoparietalAstrocytoma2PartialNo
L frontalMeningioma1Macro totalNo
L parietalAstrocytoma2BiopsyYes
L temporoparietalAstrocytoma2BiopsyYes

Exclusionary criteria for all patients included history of learning disability, widespread cognitive impairment, or a comorbidity that caused potential neuropsychologic deficits or that would provide further risk to mortality (e.g., stroke, encephalitis, mental retardation, history of moderate or severe head injury, uncontrolled hypertension, kidney disease, immunosuppressive disease, pulmonary disease, or coronary artery disease). A total of 14 recurred patients and 23 nonrecurred patients were identified.

Cases were restricted so that recurrences were analyzed only if they occurred within 6 months of the last neuropsychologic evaluation. Observations of patients in the at-risk group (i.e., nonrecurred glioma patients) also stopped 6 months after the patient's last evaluation. This resulted in 11 recurred patients and 23 nonrecurred patients. Although not a population study, the group sizes were considered sufficient to explore the effectiveness of the tumor specific versus general cognitive model.

Neuropsychologic Battery

A comprehensive, repeatable 4-hour battery of neuropsychologic tests was administered at baseline by a fellow in neuropsychology who was supervised by the senior neuropsychologist. Patients were tested in 10 neuropsychologic domains; tests of mood, fatigue, and functional living are included in the battery but not in these analyses of cognitive predictors of tumor recurrence. The broad battery includes parallel visual/verbal attention, short-term/long-term, automatic/effortful, and perceptual/semantic versions of mnemonic and cognitive tests. We excluded tests that were unsuitable for repeated measures designs, and use alternate forms of some tests that are counterbalanced across testing epochs: Auditory Selective Attention, Rey Auditory Verbal Learning, Revised Visual Retention, Rey-Osterrieth Complex Figure Test [with Taylor revision], Biber Figure Learning, Sentence Repetition, Icon Memory, and Pseudoword Memory. A similar battery16 previously demonstrated sensitivity and reliability. The selection of tests that were tumor-specific was based on theories of the neurologic associates of each test, but also were based empirically on the presence of deficit on a test in an individual patient. The tests that eventually were included in the tumor-specific model for this set of patients are indicated by an asterisk (*) that precedes the test name. The following battery is inclusive of all tests used and is described in detail elsewhere,15, 16 whereas new experimental tests are described below:

1) Attention:Auditory Selective Attention Test,17Continuous Performance Test,18 and Bells Test;192) Information Processing Speed: *Symbol Digit Modalities Test-oral version20 and *Paced Auditory Serial Addition Test;213) Motor Control: *Finger Oscillation Test20 and *Alternating Collaborative Hand Movements; 4) Verbal Short-Term Memory:Digit Span Test20 and *Word Span Test;105) Verbal Associative and Long-Term Memory: *Rey Auditory Verbal Learning Test (RAVLT),22 *Pseudoword Memory Test (follows the same learning and recall format of the RAVLT, using pronounceable pseudowords with little similarity to real words, and was developed by Vitevitch et al.23); and *Word Recognition Test - computer administration of pictures of meaningful words presented for memorization during a study condition, followed by the test condition that required the recognition of the words from others not previously presented24); 6) Visuospatial Short-Term Memory: *Visual Memory Span Test,20Benton Revised Visual Retention,25 and *Rey-Osterrieth Complex Figure Test (immediate free recall20); 7) Visual Associative and Long-Term Memory: *Biber Figure Learning Test,26, 27 *Icon Memory Test (follows the same learning and recall format of the Biber, using internationally understood, simply drawn [to reduce output interference] meaningful object icons [developed/normed for this study]), *Picture Recognition Test (computer administration of pictures of objects presented for memorization during a study condition, followed by the test condition that requires recognition of the seen objects from others not previously presented24), and *Rey-Osterrieth Complex Figure Test (delayed free recall of complex design20); 8) Visuospatial Perceptual Processing:Road Map Test,28, 29 *Rey-Osterrieth Complex Figure Test (copy of a complex design20), and *Visual Pursuits (Employee Aptitude Survey);309) Language:Sentence Repetition Test22, *Controlled Oral Word Association Test,22 *Category Fluency Test22, and Test of Written Expression;2210) Intellectual/Conceptual Processes: *Wisconsin Card Sorting Test (one deck version22); 11) Personality, Mood, and Fatigue:Fatigue Severity Scale,32Minnesota Multiphasic Personality Inventory-2,22Beck Depression Inventory,33Hamilton Rating Scale for Depression,34 and Functional Living Index a self-rating scale of patients' responses to a 20-item questionnaire to measure functional quality of life35).

Each follow-up assessment included an interview that obtained current medical treatments, changes in treatment since last assessment, MRI findings, other neurodiagnostic findings, follow-up schedule, medications, seizure status, mood, cognitive complaints, vocational status, and psychosocial status (including living arrangement). Recurred and nonrecurred patients received the same battery, which was administered at every follow-up timepoint. The schedule of contacts was: 1) baseline (6 weeks after completing surgery if any and immediately prior to starting radiotherapy; 2) 3 months; 3) 3 months later; 4) 6 months later or 1 year after baseline; and 5) on a yearly basis thereafter.

Pedictor Variables

We sought a model of measurement that would be cost-effective (defined as the time needed to administer the tests) and yet highly beneficial (defined as sensitivity to a known tumor). To accomplish this, we limited the elements of each of two models to no more than five and no less than three test indices so that the relative sensitivity of the general and varying tumor-specific models would not be affected by differences in the number of indices used in the model.

The general model applied a uniform set of sensitive tests to the objective of predicting tumor recurrence. It was developed from the predictive variables from the study by Meyers et al.2 and included the RAVLT-5 Trial Total and Recognition Discrimination (equivalent to the 5 Trial total and Recognition measures from the Hopkins Verbal Learning Test used by Meyers et al.2) and Symbol Digit Modalities Test (SDMT) (similar to the Digit Symbol Test used by Meyers et al.2). In addition, the general model included two other measures that are sensitive to white matter integrity: the Semantic Fluency-animal naming (SemFlu) and Paced Auditory Serial Addition Test (PASAT).10, 36 All measures demonstrated a normal distribution in a normal control population. The tumor-specific model was developed individually for each patient based on the results of the baseline comprehensive neuropsychologic evaluation, from which we selected three to five neuropsychologic indices that met the following criteria: 1) the index was distributed normally 2) there was a theoretic structure-function association between the patient's known tumor location and distribution and the cognitive index (e.g., a visual memory impairment in a patient with a right prefrontal, temporal, or parietal tumor locus), and 3) the index had to be relatively impaired at baseline for that patient. Relative impairment was defined as either a score that was at least 1 standard deviation below the mean, or a score that was at least 1 standard deviation below the patient's total performance mean. Tumor-specific scores were selected for patients in both the recurred and nonrecurred groups. Each score was normalized individually for every patient, and the mean of the patient's Z scores was used as the tumor-specific model index.

The predictor variables entered into the model were demographic data (two variables: age and gender), the general model (one variable: individually normalized mean Z score for the five general predictors), the individual components of the general model (five variables: RAVLT -5 Trial Total, RAVLT-Recognition Discrimination, SDMT-Oral Version, SemFlu, and PASAT-mean of 2.4 seconds, 2.0 seconds, and 1.6-second trials), treatment condition (two variables: radiotherapy or no radiotherapy), and the tumor-specific model (one variable: a normalized mean z score for the three to five-subject specific predictor test scores).

Statistical Analyses

All neuropsychologic predictor variables were renormalized for each patient based on his/her own time-varying scores. The measure for a given time point was z(t) = [x(t)-m(t)]/s(t) in which m(t) is the mean of all prior observations of x and s(t) is the corresponding standard deviation.

A Cox proportional hazards model37 was used to fit the data to time to recurrence, with time varying covariates as predictors. The hazard rate at any given time (t) is the probability of a tumor recurrence in the next instant given that the subject was free of recurrence up to t. The algorithm computes log-hazard-ratio coefficient estimates for the increase in risk due to a difference of one unit on the predictor's scale, adjusted for all other predictors being tested. We used a backward selection process, removing the least predictive variable at each step, with a threshold for removal (maximum P value = 0.15) that retained more variables than the standard level of significance (P = 0.05). Those variables remaining in the analysis at the end of the process were considered predictive of recurrence.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

At baseline and prior to recurrence, the mean z score for the patients' general model was −1.10 for the recurred group and −0.80 for the nonrecurred group (t(32) = 0.93; P < 0.36). The mean z score for the patient's tumor-specific model at baseline was −1.67 for the recurred group and −2.21 for the nonrecurred group (t(32) = −1.56; P < 0.13). Thus, the groups were comparable with regard to their degree of cognitive impairment as measured by the two models.

The backward stepwise elimination process that began with all predictor variables ended with only four variables remaining in the analysis: radiation therapy (hazards ratio [HR] = 0.315; P = 0.13), tumor-specific model (HR = 5.211; P = 0.02), RAVLT Recognition Discrimination (HR = 1.97; P = 0.08), and SemFlu (HR = 0.337; P = 0.10). Only the tumor-specific score, which added a 400% increased chance of recurrence when it dropped by 1 standard deviation, reached a significance level of P < 0.05 (Table 2). A decline in RAVLT Recognition Discrimination also was associated with an increased chance of recurrence although the magnitude and significance level were not as strong. A lower SemFlu score was found to be associated with a reduced chance of recurrence, but it also was less significant. Radiation therapy was found to be associated with a reduced likelihood of recurrence of relatively low significance. Table 2 shows the Cox proportional hazards regression results for all the predictive variables.

Table 2. Cox Proportional Hazards Regression Results for Cognitive Indices
IndexCoefficient (β)SEP valueHazards ratio
  1. SE: Standard error; y/n: year/number; RAVLT: Rey Auditory Verbal Learning Test; SemFlu: Semantic Fluency Test (Animal Naming); SDMT: Symbol Digit modality Test; PASAT: Paced Auditory Serial Addition Test.

Age−0.020.030.470.98
Gender0.220.800.781.25
Radiotherapy (y/n)−1.150.770.130.32
Tumor-specific model1.650.700.025.21
General model0.111.000.911.12
 RAVLT - 5-Trial Total−0.020.710.980.99
 RAVLT - Recognition Discrimination0.680.380.081.97
 SemFlu−1.090.660.100.34
 SDMT0.450.590.451.56
 PASAT0.160.560.771.18

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

The examination of whether cognitive function can predict tumor recurrence was extended in the current study through the inclusion of patients with low-grade brain tumors, strong functional abilities, and long life expectancies. This reduces the necessity of controlling for histology, tumor growth rates, number of recurrences, and other individual random effects because each patient is studied for one recurrence and neuropsychologic data are compared within individuals across timepoints. The use of a control group of patients without recurrences further controlled for the clinical factors. Our study design was intended to focus on specific cognitive indicators and the comparison of two methods of cognitive prediction. The magnitude of the hazards ratio for the tumor-specific index was found to be strong and more predictive than any other variable, suggesting that a tumor-specific model for predicting tumor recurrence may be feasible. The remaining three variables in the hazards ratio analysis appeared to have a greatly reduced predictive power considering the small sample size and other limitations of the study.

Several studies also have used disease-specific neurocognitive measures to predict progressive neurologic disease, and these studies have shown that the only cognitive predictors of disease progression were those that were found to be associated closely with the known neurologic disease. Neuropsychologic testing at yearly intervals using a global cognitive impairment factor38 was not found to be sensitive to the early stages of Alzheimer disease; however, specific disease-sensitive cognitive measures (i.e., free recall) did appear to discriminate elderly patients with mild cognitive impairment from those who developed dementia.14 In a study of patients with idiopathic Parkinson disease without dementia, only one measure (Picture Completion test from the Wechsler Adult Intelligence Scale-Revised) from an extensive neuropsychologic battery was found to independently identify those patients who developed dementia over the course of 3.5 years.39 Research to predict cognitive impairment associated with human immunodeficiency virus (HIV) also has been reported to demonstrate the specificity of cognitive measurement. One study20 used both a broad battery of cognitive indices and specific measures sensitive to HIV infection, as well as MRI, to follow asymptomatic and symptomatic patients with the acquired immunodeficiency syndrome (AIDS) every 6 months for 2.5 years. Only 5 of 47 neuropsychologic indices were found to correlate with increased caudate atrophy or change over time. In another study13 only 2 of 11 neuropsychologic tests were found to be associated significantly with caudate volume in patients with advanced HIV disease. A study41 of neuropsychologic changes in intravenous drug users at 6-month intervals, including those who were HIV-positive and HIV-negative, also found selective cognitive measures sensitive to HIV-positive symptoms over time. Together these studies support the validity of using specific cognitive measures associated with varying brain tumor loci to predict the course of disease.

The significant cognitive indices in the study by Meyers et al.2 of recurrent high-grade gliomas that were found to be predictive of tumor growth2 were word list recall (P = 0.013) and a discrimination index from a word recognition task (P = 0.0019), although the Digit Symbol measure also was found to have some degree of prediction (P = 0.10). Comparisons cannot be made with the current study because hazards ratios were not reported. Similar measures were analyzed in the current study both as components of the general model and as individual indices. In the current study, only the discrimination index demonstrated a trend toward prediction of tumor recurrence (P = 0.075), but its power (a decline of 1 standard deviation predicted a 197% increased chance of recurrence) was less than that for the tumor-specific model that predicted a 521% increased chance of recurrence. The multitrial word list learning measure (RAVLT-5 Trial Total) was not found to be significant in the current study although it had been somewhat predictive in the study by Meyers et al.2 Although there were several design differences between the study by Meyers et al.2 and the current study, such as our inclusion of patients with nonmalignant, slowly growing tumors and longer periods to recurrence, the tumor-specific model appears to proffer more sensitive early detection than the significant cognitive measures found in the study of high-grade glioma recurrence. This is notable because the prediction of change in the low-grade tumors required sensitivity to slower and smaller changes in tumor size over more varying time periods than those that occur in highly malignant tumors. Our lack of support for a general model or common set of cognitive predictors may be due in part to our sampling of low-grade tumors. As highly malignant gliomas increase in size, they disrupt parenchyma and cognitive systems to a greater degree than that caused by low-grade gliomas.

We cannot make inferences regarding the sensitivity of either neurocognitive model to extraaxial or neuroendocrine tumors because only two nonparenchymal tumors (pineocytoma and pituitary adenoma) were included in the group of recurred patients. Examination of the change in test scores just prior to recurrence demonstrated no decline in tumor-specific scores, and future studies of these techniques may demonstrate little efficacy of the neurocognitive model for nonparenchymal tumors. However, post hoc individual examination also demonstrated that measures of memory (visual and verbal) declined in the case of the pineocytoma, and visual memory was found to decline in the case of the pituitary adenoma. Refinement of test selection using empiric findings of cognitive change in relation to tumor locus may lead to more sensitive tumor-specific models.

We currently are reviewing this model to improve our methods for selecting tumor-specific predictive measures. We also believe that we can increase the sensitivity and robustness of the tumor-specific model by collecting more frequent observations, especially soon after diagnosis to improve subject-specific computations. Furthermore, the method of analysis in the current pilot study equally weighted the most recent observed scores, irrespective of time prior to recurrence. This was the case because the parent study examining the effects of radiotherapy on brain functions used annual follow-up after the first year after treatment. In a study with more frequent testing, lag times will not be so varied, but an analysis that includes the estimation of some time lag-dependent weight function may be useful for determining how often tests must be administered. Finally, although the results appear promising, this sample size still is very limited. Minor changes, such as adjusting censoring times or adding another subject, can have major effects on parameter estimates.

Earlier prediction of tumor growth or recurrence can be important in preventing or mitigating neurologic deterioration (e.g., visual loss) or in improving a patient's life span by delivering more diverse clinical treatments earlier in the development of malignancy. However, to our knowledge, this assumption has not been well tested. Currently, there are no noninvasive, brief, and inexpensive methods with which to mark tumor recurrence. Although it is unlikely that neuropsychologic testing will ever replace diagnostic methods such as MRI, these results suggest that neuropsychologic methods may be improved to provide a model that is sensitive to disease progression, and thus help to better calibrate the timing for clinical MRI scans. Because the majority of the nonrecurred patients had undergone macroscopic total resections and the majority of the recurred patients had undergone biopsies, this method of early detection may be most useful for patients for whom only biopsy or partial resections were possible. Other potential applications include the utility of neuropsychologic assessment in combination with MRI for earlier indication of MRI scheduling over the late periods after treatment in patients whose tumors remain stable, and the development of a low-cost/high-benefit method that could be used to test treatment effects more vigorously. A future study is planned to use volumetric MRI measures of tumor growth against which neurocognitive indices of tumor activity would be compared.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES
  • 1
    Jelic V, Johansson S, Almkvist O, et al. Quantitative electroencephalography in mild cognitive impairment: longitudinal changes and possible prediction of Alzheimer's disease. Neurobiol Aging. 2000; 21: 533540.
  • 2
    Meyers C, Hess K, Yung W, Levin V. Cognitive function as a predictor of survival in patients with recurrent malignant glioma. J Clin Oncol. 2000; 18: 646650.
  • 3
    Peyser J, Rao S, LaRocca N, Kaplan E. Guidelines for neuropsychological research in multiple sclerosis. Arch Neurol. 1990; 47: 9497.
  • 4
    Archibald Y, Lunn D, Ruttan L, et al. Cognitive functioning in long-term survivors of high-grade glioma. J Neurosurg. 1994; 80: 247253.
  • 5
    Giovagnoli A, Boiardi A. Cognitive impairment and quality of life in long-term survivors of malignant brain tumors. Ital J Neurol Sci. 1994; 15: 481488.
  • 6
    Anderson S, Damasio H, Tranel D. Neuropsychological impairments associated with lesions caused by tumor or stroke. Arch Neurol. 1990; 47: 397405.
  • 7
    Hom J, Reitan R. Neuropsychological correlates of rapidly vs. slowly growing intrinsic cerebral neoplasms. J Clin Neuropsychol. 1984; 6: 309324.
  • 8
    Scheibel R, Meyers C, Levin V. Cognitive dysfunction following surgery for intracerebral glioma: influence of histopathology, lesion location and treatment. J Neurooncol. 1996; 30: 6169.
  • 9
    Beatty W, Goodkin D, Hertsgaard D, Monson N. Clinical and demographic predictors of cognitive performance in multiple sclerosis. Arch Neurol. 1990; 47: 305308.
  • 10
    Grossman M, Armstrong C, Onishi K, et al. Patterns of cognitive impairment in relapsing-remitting and chronic progressive multiple sclerosis. Neuropsychiatry Neuropsychol Behav Neurol. 1994; 7: 194210.
  • 11
    Wishart H, Sharpe D. Neuropsychological aspects of multiple sclerosis: a quantitative review. J Clin Exp Neuropsychol. 1996; 19: 810824.
  • 12
    Fischer J, Priore R, Jacobs L, et al. Neuropsychological effects of interferon beta-1a in relapsing multiple sclerosis. Ann Neurol. 2000; 48: 885892.
  • 13
    Kieburtz K, Ketonen L, Cox C, et al. Cognitive performance and regional brain volume in human immunodeficiency virus type I infection. Arch Neurol. 1996; 53: 155158.
  • 14
    Reischies F, Hellweg R. Prediction of deterioration in mild cognitive disorder in old age - neuropsychological and neurochemical parameters of dementia diseases. Compr Psychiatry. 2000; 41: 6675.
  • 15
    Armstrong C, Ruffer J, Corn B, DeVries K, Mollman J. Biphasic patterns of memory deficits following moderate dose/partial brain irradiation: neuropsychologic outcome and proposed mechanisms. J Clin Oncol. 1995; 13: 22632271.
  • 16
    Armstrong C, Corn B, Ruffer J, Pruitt A, Mollman J, Phillips P. Radiotherapeutic effects on brain function: double dissociation of memory systems. Neuropsychiatry Neuropsychol Behav Neurol. 2000; 13: 101111.
  • 17
    Armstrong C. Selective versus sustained attention: a continuous performance test revisited. Clin Neuropsychol. 1997; 11: 1833.
  • 18
    Loong J. The continuous performance test. San Luis Obispo, CA: Wang Neuropsychological Laboratory; 1991.
  • 19
    Vanier M, Gauthier L, Lambert J, et al. Evaluation of left visuospatial neglect: norms and discrimination power of two tests. Neuropsychology. 1990; 4: 8796.
  • 20
    Lezak M. Neuropsychological assessment, 3rd ed. New York: Oxford University Press; 1995.
  • 21
    Gronwall D. Paced auditory serial-addition task: a measure of recovery from concussion. Percept Mot Skills. 1977; 44: 367373.
  • 22
    Spreen O, Strauss E. A compendium of neuropsychological tests. New York: Oxford University Press; 1991.
  • 23
    Vitevitch MS, Luce PA, Charles-Luce J, Kemmerer D. Phonotactics and syllable stress: implications for the processing of spoken nonsense words. Lang Speech. 1997; 40: 4762.
  • 24
    Goldstein B, Armstrong CL, Modestino E, Ledakis GE, John C, Hunter JV. Picture and word recognition memory in adult intracranial tumor patients. Brain Cogn. In press.
  • 25
    Benton A. Revised visual retention test: clinical and experimental application, 4th ed. San Antonio, TX: The Psychological Corporation, 1974.
  • 26
    Drachman D, Glosser G, Fleming P, Longenecker G. Memory decline in the aged: treatment with lecithin and physostigmine. Neurology. 1982; 32: 944950.
  • 27
    Glosser G, Goodglass H, Biber C. Assessing visual memory disorders. Psychol Assess Jour Consul Clin Psychol. 1989; 1: 8289.
  • 28
    Armstrong C, Cloud B. The emergence of spatial rotation deficits in dementia and normal aging. Neuropsychology. 1998; 12: 208217.
  • 29
    Money J, Alexander D, Walker HJ. A standardized Road Map Test of direction sense. Baltimore: Johns Hopkins Press; 1965.
  • 30
    Grimsley G, Ruch FL, Warren ND, Ford JS. Employee aptitude survey. Los Angeles: Psychological Services, Inc.; 1956.
  • 31
    Armstrong C, Seebauer E. A test of written expression. Arch Clin Neuropsychol. 1990; 5: 154a.
  • 32
    Krupp L, LaRocca N, Muir-Nash J, Steinberg A. The fatigue severity scale. Arch Neurol. 1989; 46: 11211123.
  • 33
    Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. An inventory for measuring depression. Arch Gen Psych. 1961; 4: 5363.
  • 34
    Klerman GL, Weissman MM, Rousanville BJ, Chevron ES. Interpersonal psychotherapy of depression. New York: Basic Books; 1984: 223244.
  • 35
    Schipper H, Clinch J, McMurray A, Levitt M. Measuring the quality of life of cancer patients: the Functional Living Index—Cancer: development and validation. J Clin Onc. 1984; 2: 472483.
  • 36
    van Buchem M, Grossman R, Armstrong C, et al. Correlation of volumetric magnetization transfer imaging with clinical data in MS. Neurology. 1998; 50: 16091617.
  • 37
    Cox D. Regression models and life tables (with discussion). J R Stat Soc [Ser B]. 1972; 34: 187220.
  • 38
    Goldman W, Price J, Storandt M, et al. Absence of cognitive impairment or decline in preclinical Alzheimer's disease. Neurology. 2001; 56: 361367.
  • 39
    Machieux F, Fenelon G, Flahault A, Manifacier M, Michelet D, Boller F. Neuropsychological prediction of dementia in Parkinson's disease. J Neurol Neurosurg Psychiat. 1998; 64: 178183.
  • 40
    Hall R, Whaley R, Robertson K, Hamby SJW, Hall C. The correlation between neuropsychological and neuroanatomic changes over time in asymptomatic and symptomatic HIV-1-infected individuals. Neurology. 1996; 46: 16971702.
  • 41
    Stern Y, Liu X, Marder K, et al. Neuropsychological changes in a prospectively followed cohort of intravenous drug users with and without HIV. Neuropsychiatry Neuropsychol Behav Neurol. 1996; 9: 8390.