- © 2006 by American Society of Clinical Oncology
Detrimental Effects of Tumor Progression on Cognitive Function of Patients With High-Grade Glioma
- Paul D. Brown,
- Ashley W. Jensen,
- Sara J. Felten,
- Karla V. Ballman,
- Paul L. Schaefer,
- Kurt A. Jaeckle,
- Jane H. Cerhan and
- Jan C. Buckner
- From the Mayo Clinic, Rochester, MN; Toledo Community Hospital Oncology Program, Toledo, OH; and the Mayo Clinic, Jacksonville, FL
- Address reprint requests to Paul D. Brown, MD, Mayo Clinic, 200 First St SW, Rochester, MN 55905; e-mail: brown.paul{at}mayo.edu
Abstract
Purpose There is growing recognition that the primary cause of cognitive deficits in adult patients with primary brain tumors is the tumor itself and more significantly, tumor progression. To assess the cognitive performance of high-grade glioma patients, prospectively collected cognitive performance data were analyzed.
Patients and Methods We studied 1,244 high-grade brain tumor patients entered onto eight consecutive North Central Cancer Treatment Group treatment trials that used radiation and nitrosourea-based chemotherapy. Imaging studies and Folstein Mini-Mental State Examination (MMSE) scores recorded at baseline, 6, 12, 18, and 24 months were analyzed to assess tumor status and cognitive function over time.
Results The proportion of patients without tumor progression who experienced clinically significant cognitive deterioration compared with baseline was stable at 6, 12, 18, and 24 months (18%, 16%, 14%, and 13%, respectively). In patients without radiographic evidence of progression, clinically significant deterioration in MMSE scores was a strong predictor of a more rapid time to tumor progression and death. At evaluations preceding interval radiographic evidence of progression, there was significant deterioration in MMSE scores for patients who were to experience progression, whereas the scores remained stable for the patients who did not have tumor progression.
Conclusion The proportion of high-grade glioma patients with cognitive deterioration over time is stable, most consistent with the constant pressure of tumor progression over time. Although other factors may contribute to cognitive decline, the predominant cause of cognitive decline seems to be subclinical tumor progression that precedes radiographic changes.
INTRODUCTION
Cranial radiotherapy remains the single greatest advance in the treatment of high-grade gliomas to date, with a doubling of median survival in randomized trials compared with patients not receiving radiotherapy.1,2 In addition, the use of combined-modality therapy (eg, temozolomide) has allowed the majority of patients to achieve a survival of greater than 1 year, with more than one fourth of patients on a multicenter study surviving greater than 2 years.3 Consequently, concerns many investigators have had regarding the late cognitive toxicities of cranial radiotherapy for the treatment of brain tumors (such that these anxieties have had a significant influence on the choice of treatment)4-7 are now also somewhat more relevant for patients with high-grade gliomas. The belief that the late effects of treatment account for a significant proportion of cognitive deficits in brain tumor patients (and therefore, that modification of treatment of brain tumors is necessary to decrease this risk) has dominated during the last couple of decades and continues even to this day.
More recently, there has been a paradigm shift,8,9 with several studies finding brain tumors and the progression of these tumors to be the predominant causes of cognitive deterioration in this population.10-12 One of the earliest studies that led this paradigm shift was an initial report by Taylor et al13 analyzing the Folstein Mini-Mental State Examination (MMSE)14 scores obtained prospectively on two consecutive North Central Cancer Treatment Group (NCCTG) high-grade glioma trials. Taylor et al13 found no clear trend to cognitive worsening from treatment; instead, tumor progression seemed to have the strongest impact on cognitive function, although the sample sizes were quite small at longer follow-up periods and larger numbers were needed to substantiate these results. Therefore, in an effort to confirm these results with a larger study population, the current study analyzes MMSE scores prospectively obtained in 1,244 patients on eight consecutive NCCTG treatment trials.
PATIENTS AND METHODS
Patients
Patients enrolled onto eight consecutive NCCTG prospective clinical therapy trials (857251, 860351, 887202, 887203, 887252, 907201, 917201, and 937252) for newly diagnosed primary brain tumors were assessed. Institutional review board–approved consent was obtained from all patients before accrual at all participating sites. All trials enrolled adult patients (age older than 18 years) with newly diagnosed supratentorial, histologically confirmed WHO grades 3 and 4 astrocytomas, gliosarcomas, or oligoastrocytomas. Patients were treated with radiotherapy and chemotherapy (nitrosourea based). The results and methods of these trials have been published elsewhere,15-21 except for 887203 (accelerated radiation therapy and carmustine) and 907201 (radiotherapy with a radiosurgical boost and carmustine).
Evaluations
At study entry (ie, after surgery and histologic confirmation of high-grade glioma), at intervals corresponding to treatment, and at each follow-up evaluation after completion of treatment, the MMSE was administered. The MMSE is a brief, standardized tool to grade cognitive function and is a well-validated screening test for dementia and cognitive impairment.14,22-25 It begins with an assessment of orientation to place and time. A maximum of 10 points may be achieved on this section (the number of correct points provides a subscore). Next, memory is tested by having the patient immediately repeat the names of three objects presented orally. Following this, the patient subtracts 7 serially from 100. The patient is then asked to recall the three items previously repeated. The final section of the MMSE evaluates aphasia and apraxia by testing naming, repetition, compliance with a three-step command, comprehension of written words, writing, and copying a drawing, for a total of 9 points. The maximal score that can be achieved for the entire MMSE is 30 points.
The patient's functional status was also assessed via the Eastern Cooperative Oncology Group (ECOG) performance score (0 = normal, 4 = bedridden). In addition, computed tomography or magnetic resonance imaging studies and neurologic status assessment were performed at each assessment. Patients were removed from the trial if they demonstrated unequivocal clinical or radiologic evidence of tumor progression.
Analysis Methods
The evaluations performed at baseline and approximately 6, 12, 18, and 24 months after study entry were designated key evaluations. Patients were included in the analyses at a specific key evaluation if they were evaluated within 60 days of the specified time. At each key evaluation, participants were classified as having experienced either disease progression (PRs) or no disease progression (NPRs) according to whether they did or did not meet the criteria for disease progression at that interval. Preliminary analyses comparing patients with a grade 4 astrocytoma versus all others with a nonglioblastoma histology (ie, grade 3 astrocytomas, anaplastic oligodendrogliomas, anaplastic oligoastrocytomas) was conducted but the statistical power was quite limited due to the small number of patients in the nonglioblastoma subgroup; hence, the results are not reported in detail in this article.
A loss of greater than 3 points in the MMSE was considered clinically significant deterioration; conversely, a gain of greater than 3 points was considered clinically significant improvement.25 Proportions of the population with a specific characteristic were estimated with binomial point estimators and 95% CIs. Linear correlation between pairs of variables was assessed using Spearman rank correlation coefficients. Equality of distributions of patient characteristics at baseline and at each of the key evaluations was tested with the Wilcoxon rank sum test. Distributions of time to disease progression and time to death were estimated with Kaplan-Meier26 curves and compared with log-rank tests.27 Time to death was measured from time of study entry until death; patients alive at the time of analysis were censored. Time to progression was measured from time of study entry until progression; patients who did not experience progression by the time of analysis were censored and patients who died without documented progression were assumed to have experienced progression at the time of their death.
RESULTS
Patient Characteristics
A total of 1,244 eligible patients were enrolled onto eight consecutive clinical trials. Baseline MMSE scores were available for 1,146 of the patients (92.1%). Patient age ranged from 18 to 84 years (median, 56 years). The vast majority of patients (997; 87%) had a grade 4 glioma, whereas a small subset of patients (149; 13%) had a grade 3 glioma. ECOG performance score was 0 or 1 at baseline for 894 patients (78%) and ECOG 2 or higher for 244 patients (22%). The mean MMSE at baseline was 26.4 (standard deviation, 5.1; median 28). The baseline characteristics of patients with abnormal MMSE scores (0 to 26 points) and those with normal MMSE scores (27 to 30 points) are summarized in Table 1. Results show that patients with abnormal baseline MMSE tended to be older (P < .001), have a worse ECOG performance score (P < .001), were more likely to have undergone a biopsy only (P < .001), less likely to have frontal lobe involvement, and more likely to have parietal lobe involvement (P = .007).
Death, Disease Progression, and MMSE
At the time of this analysis, the median follow-up for all patients was 11.2 months. Median time to progression was 6.9 months, with a total of 1,106 progressions observed. Median survival was 11.2 months, with a total of 1,097 deaths observed. Table 2 lists, for each key evaluation, the number of patients who were alive and the number without disease progression (ie, the NPRs). The number of NPRs with MMSE data both at baseline and at each key evaluation, and with significant decline in their MMSE scores, is also listed. The proportion of patients without tumor progression who experienced clinically significant cognitive deterioration from baseline was stable at 6, 12, 18, and 24 months after study entry (18%, 16%, 14%, and 13%, respectively).
Change in MMSE Scores
Analyses of changes in performance and MMSE scores revealed a weak to moderate negative correlation; these changes at each key evaluation are summarized in Table 3. Preliminary analyses comparing grade 4 astrocytoma patients with all others with a nonglioblastoma histology (ie, grade 3 astrocytomas, anaplastic oligodendrogliomas, anaplastic oligoastrocytomas) found for the nonglioblastoma histologies, in addition to a very slight decrease of the mean MMSE score at 6 months (−0.25), stable scores with a nonsignificant improvement in MMSE scores at 12, 18, and 24 months (+0.32, +0.10, +0.31, respectively).
When we divided the NPRs into three groups (Table 4) of declining MMSE (loss of > 3 points), stable MMSE, and improving MMSE (gain of > 3 points), those patients with cognitive decline tended to have worsening performance scores and this was statistically significant at the 6-month (P < .001) and 18-month evaluations (P = .03). Patients with cognitive decline also tended to be older, and this was statistically significant at the 6-month (P < .001) and 12-month evaluations (P = .03). Notably, the patients with cognitive deterioration tended to have significantly fewer remaining days to progression and death (statistically significant at 6, 12, and 18 months for survival, and at 6 and 12 months for time to progression) than those with stable or improving MMSE scores (Table 4).
Missing Values
To identify possible biases introduced by the large number of NPRs whose MMSE values were not recorded at each evaluation, we compared the characteristics of the NPRs with and without MMSE values recorded at each key evaluation (Table 5). There essentially were no significant differences in baseline characteristics between the groups except NPRs without MMSE values recorded were more likely to have grade 4 tumors (P < .001 at the 6-month evaluation). However, at all evaluations (including 18- and 24-month evaluations; data not shown), NPRs without MMSE values had a shorter time to progression and a shorter time to death. These results suggest health care workers seemed to test patients selectively (consciously or unconsciously) with a better prognosis.
Future PRs Versus NPRs
The baseline characteristics of the participants who did and did not experience progression by the end of the interval defined by ± 60 days of the key evaluations at 6, 12, 18, and 24 months were compared (ie, at the time of comparison, neither future PRs nor NPRs had radiographic evidence of progression; however, the future PRs did later develop radiographic evidence of progression after the evaluation). The future PRs were older, had poorer performance scores, lower MMSE at baseline, and were more likely to have grade 4 tumors (Table 6). The mean change in MMSE from baseline was greater (negatively) for future PRs than for NPRs at all evaluations; the future PRs lost an average of 4.2, 1.7, 6.4, and 1.5 points from their baseline MMSEs at the 6-, 12-, 18-, and 24-month evaluation, respectively, whereas the MMSE scores remained stable for the patients who did not experience tumor progression over the 4-month time interval.
DISCUSSION
Whether the predominant cause of cognitive decline in high-grade glioma patients is the treatment or the tumor itself remains an important question. Older studies have emphasized the late neurotoxicity of treatment, especially radiotherapy.6,28 However, these studies have suffered from many deficiencies29 in addition to the inherent weaknesses of retrospective studies, including outdated radiotherapy techniques (ie, whole-brain radiotherapy as opposed to conformal radiotherapy),28,30,31 unknown number of patients treated but not studied (ie, unidentified denominator),32 extremely heterogenous patient population,33 and most importantly, the lack of baseline neurocognitive testing, given that the brain tumor itself is often the primary cause of cognitive difficulties.6,12,28,29,33,34 In contrast, more recent studies (most of them prospective, with baseline cognitive testing) have found tumor to be the dominant cause of cognitive decline in brain tumor patients,10,12 including high-grade glioma patients.35,36 Our study seems to support the concept that the tumor is the predominant cause of cognitive decline in high-grade glioma patients, given that the proportion of patients with cognitive decline over time was stable (18% at 6 months, 16% at 12 months, 14% at 18 months, and 13% at 24 months), most consistent with the constant pressure of tumor progression over time. In contrast, in cognitive neurotoxicity due to radiotherapy, the incidence of late radiation effects increases over time, peaking at approximately 2 years after radiotherapy.28,29,37 In addition, one would expect a more favorable subset of high-grade glioma patients with a longer progression-free survival to better reflect the impact treatment has on cognitive function because there would be less pressure on this more favorable population from tumor progression.38 The current study did find mean MMSE scores stable overtime for the nonglioblastoma patients, again bolstering the premise that tumor progression is the predominant cause of cognitive deterioration.
Cognitive deterioration has long been recognized as a harbinger of tumor progression, and clinical evaluation of cognitive function has been a component of determining brain tumor response of patients on trials for decades, but has only recently been formally analyzed and confirmed.39 In this study, clinically significant deterioration in MMSE scores at a follow-up evaluation seemed to indicate subclinical tumor progression; as a matter of comparison, patients with stable or improved MMSE scores had a significantly better time to tumor progression and death, and at some evaluations time to tumor progression and death was approximately two to three times as long as for those with deterioration in MMSE scores (Table 4). Furthermore, when we compared the distributions of various factors of patients who experienced progression within the 120-day interval after evaluation and those who did not (none had radiographic evidence of progression at the time point of comparison), the patients who were to experience progression in the next 4 months had significantly lower MMSEs than did the NPRs. The average reduction in MMSE scores for future PRs was 4.2, 1.7, 6.4, and 1.5 at 6, 12, 18, and 24 months, respectively, whereas the scores remained stable for the patients who did not experience tumor progression during the 4-month time interval (Table 6). These findings demonstrate that those patients who are experiencing disease progression are more likely to display signs of cognitive decline than those patients who are NPRs, and provide additional evidence that cognitive decline in brain tumor patients after therapy frequently may reflect subclinical tumor progression rather than neurotoxicity due to treatment.
One of the limitations in this study is the insensitivity of the measurement tool. Although the Folstein MMSE is a well-validated measure of cognitive decline when used to assess dementia,14,22-25 it is expected that a more discriminating neurocognitive assessment would have identified cognitive decline earlier and more frequently. For example, some of the stability in the MMSE scores could be accounted for by practice effects, and therefore potentially could mask mild to moderate cognitive deterioration. However, even with the relative insensitivity of the MMSE, the results in this study correspond quite well with other studies with more extensive neurocognitive testing.39
The vast majority of patients in these trials also received corticosteroids and anticonvulsants, which may influence cognitive function. The dosage and length of anticonvulsant and corticosteroid therapy were not studied specifically nor controlled, and this could be a confounding variable.11,35 An additional limitation of this study is the inability to analyze long-term (ie, 5 years) cognitive function due to diminishing sample sizes.40
When NPRs with and without MMSE values recorded were analyzed, there were essentially no differences in baseline characteristics, yet at all evaluations patients without MMSE had a shorter time to progression and shorter remaining time to death (Table 5). The reasons for the significant number of missing values is not clear, but possibly reflects reluctance of medical and nursing staff to perform the MMSE on patients who appear unwell (although without documented progression at the time of evaluation). Another possibility is the patients with a worse prognosis refused to be tested. It is expected that if these patients had been tested, the proportion of patients with significant cognitive deterioration would have increased; however, it is important to note the proportion of NPRs tested with MMSE remained stable over time (and therefore the proportion of untested patients was also stable over time). In addition, with a larger number of patients with poorer prognosis, worse MMSE scores, and significantly shorter time to progression and death, it is likely that the association between cognitive decline and preclinical tumor progression would have been even stronger.
MMSE values were associated independently with other variables, including age; older patients had a higher incidence of cognitive deterioration. Patients with cognitive decline tended to have worsening performance scores, and there was a weak to moderate correlation with MMSE and performance score. Therefore, the age and overall physical condition of the patient, along with subclinical tumor progression, plays a role in initial cognitive performance and the cognitive decline noted with time in some patients.
In summary, these findings suggest that cognitive deterioration often preceded radiographically evident tumor progression and that subclinical tumor progression had a detrimental effect on the cognitive capacity of high-grade glioma patients. In addition, patients with radiographic evidence of tumor progression have significant cognitive declines. Therefore, as long as tumor progression remains the predominant cause of cognitive decline in high-grade gliomas, investigators should continue to focus their scientific endeavors on the ultimate goal of durable tumor control; this will result in the additional benefit of maintaining cognitive function in high-grade glioma patients.
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following authors or their immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO’s conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
| Authors | Employment | Leadership | Consultant | Stock | Honoraria | Research Funds | Testimony | Other |
|---|---|---|---|---|---|---|---|---|
| Paul L. Schaefer | Eli Lilly & Co | Amspan; Casemark; Edward Liferman; Novostes |
Author Contributions
Conception and design: Paul D. Brown, Jan C. Buckner
Financial support: Jan C. Buckner
Provision of study materials or patients: Paul L. Schaefer, Kurt A. Jaeckle, Jan C. Buckner
Collection and assembly of data: Paul D. Brown, Ashley W. Jensen, Karla V. Ballman
Data analysis and interpretation: Paul D. Brown, Sara J. Felten, Karla V. Ballman, Jan C. Buckner
Manuscript writing: Paul D. Brown, Karla V. Ballman, Jane H. Cerhan, Jan C. Buckner
Final approval of manuscript: Paul D. Brown, Karla V. Ballman, Paul L. Schaefer, Kurt A. Jaeckle, Jane H. Cerhan, Jan C. Buckner
Comparison of Baseline Characteristics of Patients With Abnormal (0-26) and Normal (27-30) MMSE Scores at Baseline
Number and Percentage of Patients at Baseline and Key Evaluation
Mean Changes in PS and MMSE at Key Evaluations Compared With Baseline, With Number of Participants, Spearman Correlation Coefficient Between Δ PS and Δ MMSE, and Wilcoxon P for Testing ρ = 0
Average Age, Survival, TTP, and Change in Performance Score for Subgroups Defined by Magnitude of Change in MMSE at Key Evaluations Compared With Baseline, With Number of NPRs and P Values for Testing Equality
Characteristics of NPRs With and Without MMSE Values Recorded at the 6- and 12-Month Evaluations
Characteristics of PRs and NPRs at Key Evaluations
Footnotes
-
Supported in part by Public Health Service Grants No. CA-25224, CA-37404, CA-15083, CA-35415, and the Linse Bock Foundation, and conducted as a collaborative trial of the North Central Cancer Treatment Group and Mayo Clinic.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
- Received August 2, 2006.
- Accepted September 26, 2006.
