Pretreatment cognitive performance predicts survival in patients with leptomeningeal disease
Version of Record online: 5 SEP 2002
Copyright © 2002 American Cancer Society
Volume 95, Issue 6, pages 1311–1316, 15 September 2002
How to Cite
Sherman, A. M., Jaeckle, K. and Meyers, C. A. (2002), Pretreatment cognitive performance predicts survival in patients with leptomeningeal disease. Cancer, 95: 1311–1316. doi: 10.1002/cncr.10816
- Issue online: 5 SEP 2002
- Version of Record online: 5 SEP 2002
- Manuscript Accepted: 16 APR 2002
- Manuscript Revised: 21 MAR 2002
- Manuscript Received: 3 JAN 2002
- leptomeningeal disease;
- cognitive patterns;
Leptomeningeal disease (LMD) involves the spread of malignant cells from solid tumors to the cerebrospinal fluid or to the leptomeninges. LMD has a very poor prognosis and it is difficult to diagnose and follow with traditional diagnostic tools. The purposes of this study were to characterize cognitive functioning of LMD patients before treatment and to determine if measurement of cognitive functioning could be used to predict survival time.
Thirty-seven subjects with LMD were administered the Mattis Dementia Rating Scale (DRS) before they received treatment and statistical analyses were performed.
The Conceptualization subtest of the DRS was the most sensitive to disease course. It was the only individual subtest to predict survival and it was the most impaired subtest across individual subjects. These results support earlier findings of a frontal-subcortical pattern of dysfunction in LMD patients. Cognitive performance at time of LMD diagnosis predicts survival time. Age and clinical status, two factors often correlated with survival in the cancer literature, did not predict survival time for our LMD population. In addition, there were no correlations between survival time and previous medical history or demographic factors.
Neuropsychological assessment appears to be a valuable tool both for tracking disease course over time and for predicting survival of patients. Cancer 2002;95:1311–6. © 2002 American Cancer Society.
Leptomeningeal disease (LMD) involves the spread of malignant cells from solid tumors to the cerebrospinal fluid (CSF) or to the leptomeninges (pia mater and the arachnoid), which are tissues that envelop the brain and spinal cord. It is a relatively rare process, occurring in approximately 5% of cancer patients.1 However, it is believed that its incidence will increase as survival time for patients with systemic cancers increases. This disease has a very poor prognosis, with estimates of only 20% response to treatment and a median survival time of only 4 months.2
Patients with LMD exhibit a wide variation in clinical presentation, as the disease affects all aspects of the central nervous system, including the cerebral hemispheres, the cranial nerves, and the spinal cord. Symptoms may include headache, mental status change, nausea, hemiparesis, papilledema, diplopia, difficulty with vision or hearing, gait instability, and seizures.1
Because many of these symptoms are nonspecific, LMD can be particularly difficult to diagnose and follow clinically. It is diagnosed often with a combination of CSF cytology and radiographic data. However, these methods have proven problematic, as results are often nonspecific and inconclusive. CSF studies may show abnormalities in intracranial pressure, cell count, and glucose and protein levels, all of which are nonspecific findings. Often, the CSF yields positive cytology for LMD, but this is not always the case, and there is controversy concerning appropriate cutoff criteria for positive versus negative classification. Although neuroimaging is often used to support an LMD diagnosis, there is a high rate of false-negative results for both computed tomographic (CT; 58%) and magnetic resonance imaging (MRI; 30%) scans.1
Therefore, it is important to find ways to better pinpoint and classify the clinical status of patients with LMD. This is particularly vital for tracking disease progression during clinical trials for treatment protocols, as the traditional methods of CSF analysis and radiographic information may not be sensitive or specific enough to provide clear data points. One marker of clinical status and disease progression that has been relatively unexplored is cognitive status. Cognitive assessment can be a relatively time-efficient and cost-effective procedure. Several recent studies have described the cognitive effects of various protocols on patients with LMD.3–6 However, there have been very few clear descriptions of cognitive effects for patient populations with LMD before treatment.
The purpose of the current study is to describe and characterize the cognitive functioning of a relatively large group of patients with recently diagnosed LMD before they received treatment. This will establish a baseline from which cognitive changes resulting from treatment can be examined in future research. In addition, this study examines the correlation between cognitive functioning at baseline and survival time. It is predicted that patients exhibiting higher degrees of impairment at baseline neuropsychological testing will also have a shorter time of survival after LMD diagnosis.
MATERIALS AND METHODS
Patients were diagnosed with leptomeningeal carcinoma secondary to melanoma, breast carcinoma, lung carcinoma, or other solid tumors resulting from systemic cancer. LMD diagnosis was defined as a positive CSF cytology for malignant cells or radiographic evidence of CSF disseminated tumor on CT or MRI scans.
All patients received treatment at the M. D. Anderson Cancer Center (Houston, TX) and were included in the subject pool if they were able to undergo a baseline neuropsychological examination before they received any therapy. Baseline neuropsychological assessment occurred within several days of the LMD diagnosis and before the patients received any treatment. For the purposes of this study, the date of Ommaya shunt placement was considered the date of LMD diagnosis. Data collection began in 1990 and continued through 1998. Informed consent was obtained from each patient.
The population for the current study consisted of 37 patients who had been diagnosed with LMD and received baseline neuropsychological assessments. Following baseline assessment, subjects were enrolled in one of two protocols, or they did not receive treatment. Twenty subjects received treatment with intraventricular interleukin-2 (IL-2) and seven subjects received chemotherapy treatment with cytosine arabinoside (ara-C). Ten subjects did not receive any treatment. Of the 37 subjects, 21 had a primary diagnosis of melanoma, 11 of breast carcinoma, 4 of lung adenocarcinoma, and 1 of glioblastoma multiforme. There were 25 females and 12 males. The mean age of all subjects was 43.9 (standard deviation [SD] = 9.7) years and the mean level of education was 14.2 (SD = 2.8) years.
Patients were administered the Dementia Rating Scale (DRS).7 The DRS was selected because it can be administered quickly (usually between 15 minutes and 1 hour) and easily at a patient's bedside, as only a test booklet, pencil, and paper are required. In addition, the DRS is well validated and normed and it assesses five general areas of cognition: attention (Attention subtest), frontal lobe executive functions (Initiation subtest), reasoning (Conceptualization subtest), visual-construction skills (Construction subtest), and short-term memory (Memory subtest). It was designed to track dementia in elderly patients and is, therefore, an easy examination for younger, healthier subjects to perform without error. Normal elderly subjects generally obtain 97% correct or higher (scoring 140–144 points out of a possible 144). Therefore, any scores below 95% are considered below expectation, particularly for younger subjects.
The DRS was designed to provide an overall score (out of 144 points), as well as scores for each of the five general cognitive domains. However, it is difficult to draw specific conclusions about subtest scores in terms of brain functioning due to the fact that each subtest of the DRS contains multiple individual items that assess different cognitive functions. Hoffer et al.8 investigated the DRS by performing a statistical factor analysis on DRS items and found that although the analysis yielded clusters generally similar to the subtest clusters, some items that are grouped together in subtests may not be assessing similar cognitive functions.
Therefore, group performance on individual items was examined. Following the example of several studies that used the DRS to examine patterns of cognitive dysfunction,9–11 the method of examining percentage correct of responses within individual items was utilized to compare performance across items that contained different point values and different numbers of exemplars. Individual items included repetition of digits forward, repetition of digits backward, the ability to follow one and two-step commands, word fluency for a semantic category (supermarket items), word fluency for a semantic category with visual cues (name things examiner is holding or wearing), verbal repetition of nonsense syllables, motor programming, visuospatial construction (copy) of alternating figures, visuospatial construction (copy) of geometric designs, abstract visual reasoning (similarities of shapes), abstract verbal reasoning (similarities between verbally presented items), orientation, visual attention and discrimination of target stimuli, visual recognition memory for words and designs, and recall of sentences that were read or composed recently.
Effects of Previous Treatment and Complications
Because many subjects had experienced long and complex medical problems and treatments before receiving their LMD diagnosis and entering our study, other factors that might affect cognitive functioning were investigated. As the data were distributed parametrically, two-tailed t tests were used to compare group cognitive performance at baseline according to these various factors. A conservative P level was used to correct for multiple comparisons (P < 0.01; Table 1). Previous treatment and medical complications did not significantly influence neuropsychological performance at baseline.
|Subject group comparison||Significance (two-tailed)|
|Previous chemotherapy (n = 26) vs. none (n = 11)||0.32|
|Previous immunotherapy (n = 14) vs. none (n = 23)||0.65|
|Both chemotherapy/immunotherapy (n = 11) vs. others (n = 26)||0.84|
|Neither chemotherapy/immunotherapy (n = 8) vs. others (n = 29)||0.74|
|Brain metastases (n = 20) vs. none (n = 17)||0.77|
|If brain metastases (n = 20):|
|Radiation therapy (n = 13) vs. no radiation therapy (n = 7)||0.81|
|Surgery (n = 12) vs. no surgery (n = 8)||0.18|
|Single metastasis (n = 7) vs. multiple metastases (n = 13)||0.67|
|Right vs. left hemisphere||0.16|
|Melanoma (n = 21) vs. breast carcinoma (n = 11)||0.37|
|Melanoma (n = 21) vs. all other diagnoses (n = 16)||0.59|
To ensure that subjects did not differ significantly according to the treatment protocol they received (IL-2 vs. ara-C), treatment groups were compared. They did not differ significantly on baseline cognitive results (P = 0.25).
Implications for Survival Time
Of the 37 subjects, 32 had died in the time interval between baseline neuropsychological assessment and data analysis (August 1998). The Kaplan–Meier procedure was used to statistically account for and censor survivors from further analysis. Table 2 shows the survival time and Figure 1 shows a survival curve. Baseline assessment occurred within several days of the subjects receiving the LMD diagnosis. For the entire subject group, the mean survival time from baseline assessment to date of death was 36.6 weeks (SD = 9.35). The median survival time from baseline testing was 15.0 weeks (standard error of the mena [SEM] = 3.5).
|Groups||No.||Mean (SD)||Median (SEM)|
|All subjects||37||36.6 (9.35)||15.0 (3.5)|
|Interleukin-2 group||20||42.3 (12.3)||18.0 (4.1)|
|Chemotherapy group||7||20.5 (6.6)||10.0 (4.0)|
The the age of the subjects did not influence their survival time (Spearman correlation coefficient = −0.183, two-tailed significance = 0.28). To determine if general clinical status at baseline was related to survival time, a correlation was performed between survival time and rating on the Karnofsky performance scale (KPS). It should be noted that there is no direct reflection of cognitive functioning in this scale. The correlation between KPS status and survival time was not significant (P = 0.17), likely due to the restricted range of KPS scores (≥ 70).
We investigated whether there was a correlation between cognitive performance at baseline and ultimate survival time, the results of which are reported in Table 3. At baseline, total DRS scores across the subject population correlated significantly with survival time past baseline (Spearman correlation = 0.46, one-tailed significance = 0.0016) and higher scores at baseline correlated positively with a longer survival time. This finding remained completely consistent whether or not the five surviving subjects were statistically censored or included in the correlation. This finding is particularly striking given the fact that survival did not correlate with factors that would be expected to influence survival time, such as age and general clinical status (KPS rating).
|Subtests||Spearman correlation coefficient||Significance (one-tailed)|
|Total DRS score||0.46||0.0016|
To determine whether performance on particular subtests of the DRS correlated with survival time, correlations between each subtest and survival time were computed. Only the Conceptualization subtest score correlated significantly with survival time (Spearman correlation = 0.42, one-tailed significance = 0.008), with higher scores correlating to longer time of survival from baseline. Scores on the Initiation subtest approached a significant correlation with survival time (Spearman correlation = 0.40, one-tailed significance = 0.011). However, other subtest scores (Memory, Construction, Attention) did not correlate significantly with survival time.
In additional analyses, the group receiving IL-2 (n = 20) was compared with the group receiving chemotherapy (n = 7). There was a notable difference in survival time between the groups. The IL-2 group survived longer, with a mean survival time of 42.3 weeks (SD = 12.3) and a median of 18.0 weeks (SEM = 4.1), whereas the group receiving chemotherapy survived for a mean of 20.5 weeks (SD = 6.6) and a median of 10.0 weeks (SEM = 4.0). However, the difference between the two groups was not statistically significant due to the low statistical power resulting from too few subjects in the chemotherapy group.
It is noteworthy that the total group receiving either chemotherapy or IL-2 treatment (n = 27) survived longer (mean = 34.6 weeks, SD = 48.6) than the group receiving no treatment (n = 10; mean = 16.4 weeks, SD = 10.3), although this result was not statistically significant (one-tailed significance = 0.04). However, there were no statistical differences in survival time between each treatment group individually and the no-treatment group, or when all three groups were compared in an analysis of variance analysis, likely due to low statistical power.
Pattern of Cognitive Impairment
Baseline scores on the DRS were examined for the entire subject population. The percentage of items correct for all possible items was calculated for each subject and averaged across subjects (Table 4). The average percentage correct for total DRS score was 91.2%. This finding is below the expected performance level of normal subjects (95% or greater). On subtests, the highest average percentage correct was obtained for Attention (96.6%) followed by Construction (93.7%), Initiation (89.5%), Conceptualization (89.0%), and Memory (89.0%).
|Subtests||Percentage correct||No. subjects impaired|
|Total DRS score||91.2||10/37|
To examine further the range of impairment suggested by these scores, z-scores were calculated for the total score and subtest scores according to the normative data published in the DRS manual, which were based on a sample of patients between the ages of 65 and 81 years (Table 4). Subjects whose scores were more than 1.5 SDs below the mean of the normative sample were considered to be impaired. Using this criterion, 10 of the 37 subjects were within the impaired range for total DRS score. On subtest scores, the fewest subjects fell within the impaired range on Attention (3 of 37) followed by Construction (4 of 37), Memory (5 of 37), Initiation (9 of 37), and Conceptualization (10 of 37). Further data analysis revealed that these 10 impaired subjects did not differ from the rest of the subject population according to variables of demographics, previous medical history, or previoustreatment (P > 0.15 for all two-tailed t tests).
When individual items were examined, again using the percentage correct technique, the lowest scores were obtained on the following items: Digits Backward (78.4%), Sentence Recall (83.4%), Supermarket Fluency (84.3%), Verbal Conceptualization (87.0%), and Orientation (87.1%). Table 5 contains the average percentage correct for each item.
|Item||Percentage correct (mean)|
|Alternating Designs (copy)||93.2|
The current study yielded several interesting findings. The most noteworthy finding is that baseline cognitive performance (at the time of LMD diagnosis and before treatment) correlated positively and significantly with the length of survival time. Surprisingly, baseline cognitive performance was the only variable that predicted survival. Age and clinical status (KPS rating) at baseline did not correlate significantly with survival, although these variables predicted prognosis and survival time among high-grade malignant glioma patients.12 Previous medical history or treatment variables did not influence cognitive performance at baseline assessment. In particular, radiation therapy to the brain can cause a similar pattern of cognitive impairments. However, patients who received whole brain radiation therapy for brain metastases did not perform differently from those who did not receive radiation treatment. No patient received radiation therapy for bulky LMD. Our data suggest that assessment of cognitive function may be helpful in tracking the disease course. This is of particular importance because commonly used surrogate endpoints of LMD progression such as MRI and cytology can be unreliable or variable among patients or between assessments.
Of the DRS subtest scores, only the Conceptualization subtest score (which assesses reasoning ability) correlated significantly with the length of survival time, suggesting that performance on this subtest may be especially sensitive to disease course and may hold possible implications for survival. Usually, conceptualization and other frontal lobe functions are not assessed in brief screening mental status examinations, such as the Mini-Mental Status Examination (MMSE).13 Meyers et al.14 found that the MMSE can be insensitive to cognitive changes in mental functioning in clinical trials of neurotoxic cancer drugs, as it was designed as a screening tool for dementia rather than to track more subtle cognitive changes. Conversely, The DRS assesses a wide range of cognitive domains, including frontal lobe functioning, in a relatively short amount of time.
We also investigated the pattern of neuropsychological performance of newly diagnosed LMD patients before they received treatment. In the only previous study that investigated and described specific cognitive effects of LMD before treatment,15 Siegel et al. examined 26 patients with recently diagnosed LMD using neuropsychological assessment, CT scan, and measurement of regional cerebral blood flow. Only 15 patients were able to undergo the neuropsychological testing. Of these, the investigators found that 93% of their patient population exhibited low performance on neuropsychological tests compared with a control group. Specifically, impairments were noted in the domains of attention, recent memory, and abstraction.
As predicted, our subject group demonstrated a similar pattern of cognitive impairment. Among the five subtests of the DRS, our group performed worse on the Conceptualization, Memory, and Initiation subtests and better on the Attention and Construction subtests. This pattern was the same for the entire population as well as for the subgroup of subjects falling within the impaired range (i.e., ≤ −1.5 SDs below the mean of a geriatric population). Furthermore, performance was consistently lowest on the Conceptualization subtest.
When individual items from the DRS were examined across the entire group, the group scored lowest on Digits Backward, followed by Sentence Recall, Supermarket Fluency, Verbal Conceptualization, and Orientation. With the exception of Digits Backward, these items belong to the three subtests in which our subject population performed the worst (Conceptualization, Memory, and Initiation). Digits Backward is grouped within the Attention subtest and is the only item requiring complex attention within the subtest. In addition to these items, our subject population fell within the impaired range on Supermarket Fluency and Orientation. When patient impairments on these items are compared with the impairments found by Siegel et al.,15 they reflect the same general areas of dysfunction specified in that study: attention (Digits Backward), recent memory (Sentence Recall), and abstraction (Verbal Conceptualization). In general, this pattern of impairment suggests frontal-subcortical dysfunction, as the frontal lobe is highly involved in abstract reasoning, executive functioning, attention, and working memory.16
Although the pattern of impairment was similar to the previous study, Siegel et al.15 also found that 93% of their subjects fell within the impaired range of performance. In the current study, only 27% of the subject population met our criteria for impaired overall neuropsychological performance. However, the difficulty level between the tests used in the previous study (which were not fully specified) and the DRS may be significantly different. The DRS is a relatively easy test that is used often as a cognitive tool for screening older patients with dementia. This finding suggests that a critical level of assessment breadth and difficulty is necessary to capture subtle changes associated with LMD. Although a more thorough battery of tests (such as that used by Siegel et al.15) would perhaps be more sensitive than a screening measure such as the DRS, the DRS may be more time-efficient and sufficient for capturing and tracking vital changes. This is an area for further exploration.
To conclude, cognitive performance at the time of LMD diagnosis, as measured by total score on the DRS, predicted survival, whereas age and clinical status (KPS rating) did not. In particular, the Conceptualization subtest was the most sensitive to disease course. It was the only individual subtest to predict survival and it was the best measure of impairment across subjects. Conceptualization or reasoning ability is a cognitive process that is strongly dependent on frontal lobe functioning. These results support earlier findings of a frontal-subcortical pattern of dysfunction among LMD patients.15
These conclusions hold implications for future studies, which might investigate the pattern of cognitive dysfunction throughout the course of LMD, both during and after treatment. In particular, the issue of whether cognitive performance, particularly reasoning ability, continues to predict disease course should be explored. If so, cognitive assessment may prove to be a valuable tool in making clinical and treatment decisions regarding these patients.
- 1Leptomeningeal metastases. In: LevinVA, editor. Cancer in the nervous system. New York: Churchill Livingstone, 1996: 281–301., .
- 2Central nervous system emergencies. In: WeissGR, editor. Clinical oncology. Norwalk, CT: Appleton & Lange, 1993: 372–379..
- 3Cytokine responses to intraventricular injection of interleukin-2 into patients with leptomeningeal carcinomatosis: rapid induction of tumor necrosis factor alpha, interleukin-1-beta, interleukin-6, interferon, and soluble interleukin-2 receptor. Cancer Res. 1992; 52: 1123–1128., , , , , .
- 7Dementia Rating Scale. Odessa, FL: Psychological Resources, Inc., 1988..
- 16Neuropsychological assessment, 3rd ed. New York: Oxford University Press, 1995..