Prevalence and profile of cognitive impairment in adult glioma: a sensitivity analysis Mathieu Boone, Martine Roussel, Bruno Chauffert, Daniel Le Gars & Olivier Godefroy Journal of Neuro-Oncology ISSN 0167-594X J Neurooncol DOI 10.1007/s11060-016-2152-7
1 23
Your article is protected by copyright and all rights are held exclusively by Springer Science +Business Media New York. This e-offprint is for personal use only and shall not be selfarchived in electronic repositories. If you wish to self-archive your article, please use the accepted manuscript version for posting on your own website. You may further deposit the accepted manuscript version in any repository, provided it is only made publicly available 12 months after official publication or later and provided acknowledgement is given to the original source of publication and a link is inserted to the published article on Springer's website. The link must be accompanied by the following text: "The final publication is available at link.springer.com”.
1 23
Author's personal copy J Neurooncol DOI 10.1007/s11060-016-2152-7
CLINIcAL STUDY
Prevalence and profile of cognitive impairment in adult glioma: a sensitivity analysis Mathieu Boone1 · Martine Roussel1 · Bruno Chauffert2 · Daniel Le Gars3 · Olivier Godefroy1
Received: 1 February 2016 / Accepted: 19 May 2016 © Springer Science+Business Media New York 2016
Abstract Cognitive impairment has been reported in 27–83 % of adults with World H ealth Organization (WH O) grade I–III glioma. H owever, the few studies in this field used different methods for cognitive assessment. The objective of the present study was to establish the prevalence of cognitive impairment in patients with WHO grade I–III primary brain tumors and determine the effect sizes of a comprehensive battery of tests. This study used a comprehensive neuropsychological battery to examine 27 patients. To control for false positives, prevalence was estimated from the overall neuropsychological score. Size effects were determined using Cohen’s d. Cognitive impairment was observed in 51.9 % (95 % CI 33−70.7 %) of the patients; the impairment affected action speed (38.5 %), cognitive (33 %) and behavioral (21.7 %) executive functions, oral expression (29.6 %), episodic memory (29.6 %) and visuoconstructive abilities (19.2 %). The largest effect sizes (d ≥ 1.645) were observed for the Digit Symbol Substitution test, global hypoactivity, free recall, Stroop time, the Boston Naming test (BNT), the Trail Making test B (TMTB), verbal fluency and the Rey–Osterrieth Complex Figure Test. Four of these scores (global hypoactivity, the Digit Symbol Substitution test, the TMTB perseveration,
Mathieu Boone
[email protected] 1
Laboratory of Functional Neurosciences & Department of Neurology, Amiens University Hospital, 80054 Amiens Cedex 1, France
2
Department of Oncology, Amiens University Hospital, Amiens, France
3
Department of Neurosurgery, Amiens University Hospital, Amiens, France
and the BNT) were combined to make a shortened battery (AUC 0.872; 95 % CI 0.795–0.949). The overall neuropsychological score was the sole factor associated with the functional outcome. Our results suggest that about half of survivors with a grade I–III primary brain tumor suffer from cognitive impairment. Tests with a large effect size should be included in future large-scale studies. Keywords Glioma · Cognitive disorders · Neuropsychological tests · Executive disorders · Attention
Introduction Neuro-epithelial tumors account for 47 % of primary intracranial tumors [1] with an incidence of 6.61 per 100,000 [2]. Survival has been improved by standardized treatment strategies that combine surgery, radiotherapy and chemotherapy. This improvement has highlighted the need to determine functional outcomes in survivors. Brain diseases may lead to functional impairments in the sensorimotor, psychiatric and cognitive domains. Sensorimotor impairment is determined in standardized clinical examination, whereas the psychiatric and cognitive domains are assessed with a variety of methods. Accordingly, the few neuropsychological studies to have assessed adults with primary brain tumors graded I–III according to the World H ealth Organization (WHO) classification [3] have reported prevalences of cognitive impairment ranging from 27 to 83 % [4–7]. A recent systematic review [8] revealed marked variability in the definition of cognitive dysfunction; this probably accounts for the variability in the reported prevalence and profiles of cognitive impairment. These findings emphasize the need to harmonize cognitive assessments [9]. Accordingly, a
13
Author's personal copy 2
three-test battery has been proposed by the International Cognition and Cancer Task Force (ICCTF) [10]. With a view to determining the optimal strategy for assessment cognitive in glioma patients, it is advisable to analyze the size effect associated with various cognitive tests in a large battery [11, 12]; this determines a test’s sensitivity and thus its ability to detect the presence of cognitive impairment. Surprisingly, our review of the literature failed to identify this type of sensitivity analysis in the field of brain tumors. The primary objective of the present study was therefore to determine the size effect associated with the various tests assessing cognitive and behavioral impairments in patients having been treated for a primary brain tumor (astrocytoma, oligodendroglioma, oligoastrocytoma or ependymoma) graded from I to III according to the WHO classification [3]. To this end, we administered a comprehensive cognitive battery while controlling for sensorimotor impairment. The study’s secondary objectives were to determine to determine the prevalence of cognitive and behavioral impairments, to build a shortened battery and to assess the impairment associated with disability.
Materials and methods Population All consecutive adult survivors operated on for a primary brain tumor (astrocytoma, oligodendroglioma, oligoastrocytoma and ependymoma) graded I–II or 1p/19q-codeleted anaplastic gliomas [13] on the WHO classification [3] in the neurosurgery department at the Amiens University Medical Center between January 1, 2005, and December 31, 2009, were eligible for inclusion. The main exclusion criteria were as follows: a mother tongue other than French, illiteracy, severe motor or sensory impairments, lack of informed consent, a comorbidity interfering with cognitive status [stroke, severe head injury, psychiatric illness with hospitalization for 3 days or more in a psychiatric unit, or alcohol abuse (daily intake: more than three standard drinks)], severe cardiac, renal, hepatic or respiratory insufficiency and malignant disease. All patients gave written informed consents with the approval of our institutional review board. The 87 patients fulfilling these criteria were invited to attend a follow-up consultation including clinical, neuropsychological and morphological evaluations. The following demographic data were recorded: age, gender, educational level, handedness, and prior and current occupations. The following tumor characteristics were determined: date of diagnosis, presence of epilepsy and intracranial hypertension, tumor site, histology and grade (according to the WHO classification [3]), and treatment, including surgery (total or partial
13
J Neurooncol
resection, or biopsy only), radiation (total dose), chemotherapy (type and duration), and the administration of antiepileptic drugs, steroids and other drugs. During the visit, neurological impairment was scored on to the National Institutes of Health Stroke Scale (NIHSS) [14] and disability was scored according to the Karnofsky index (KI) [15]. Of the 87 eligible patients, 60 were not assessable (Fig. 1). The remaining 27 patients were included in the study. The study population’s demographic characteristics are summarized in Table 1. Eleven participants (41 %) had an astrocytoma (1 with grade I and 10 with grade II according to the WHO classification [3]), 11 (41 %) had an oligoastrocytoma (7 with grade II and 4 with grade III), 3 (11 %) had an oligodendroglioma (2 with grade II and 1 with grade III) and 2 (7 %) had an ependymoma (1 with grade I and 1 with grade II). Surgery had been performed in 25 (93 %) cases, with total removal in ten of these (37 %). Nineteen (70.4 %) cases had received radiotherapy (median total dose 59.4 Gy; interquartile range 50.4–60) and six (22 %) had received chemotherapy [Temozolomide: n = 4, PCV (Procabazine, CCNU and Vincristine) therapy: n = 2 or Fotemustine: n = 1]. Tumor recurrence with anaplastic transformation was observed in 8 (30 %) patients. Epilepsy was present in 16 (59 %) patients, of whom 7 (26 %) had reported two or more epileptic
Pa ent with glioma tumor (grade I to III) with surgery between 2005 to 2009
n = 87
Death: n = 27 Lost to follow-up: n = 18 Unable or refused to come to hospital: n = 15
n = 27
Fig. 1 Flow chart
Author's personal copy J Neurooncol
3
Table 1 Characteristics of the population
Table 2 Neuropsychological battery Population
Domain
Tests
Agea Sex male (%) Education level: primary/secondary/high (%) Right handedness (%) Delay post-surgery (month)a
43.7 (36–51) 17 (63 %) 3/17/7 (11/63/26 %) 22 (81 %)
Global cognitive efficiency
Delay after additional treatment (month)a
18 (12–34)
Visuo-spatial abilities
Tumor location Right/left/bilateral Frontal/insula Parietal/temporal/occipital Corpus callosum/posterior fossa Hippocampal atrophy NIHSSa
11/12/4 12/2 8/4/1 5/7 2 (2–4)
Memory
Mini Mental State Examination [20–21] Progressive matrice 47 [22] Golberg scale [24] Token test (shortened version) [25] Boston Naming test (shortened version) [26] Albert cancellation test [27] Rey figure copy [28] Digit and spatial span [29] FCSRT [30] Door test [31] Digit Symbol Substitution subtest [29] Stroop [18] Trail Making test [18] Verbal fluency test [18] Modified Card Sorting test [18] Six Elements test [18] Strategic index of FCSRT [18] Behavioral Dysexecutive Syndrome Inventory [18]
a
MMSE Karnofsky indexa
Language
30 (20–44)
Action speed and executive functions
0 (0–1) 27 (26–28) 90 (70–100)
MMSE minimental state examination, NIHSS National Institutes of Health Stroke Scale a
Expressed as median and interquartile range
seizures. At the time of evaluation, 15 patients were taking antiepileptic drugs, and all were seizure-free. Two patients were taking antidepressants and/or anxiolytics. Brain imaging MRI was performed at the time of cognitive assessment in axial plane using T1, diffusion and T2 weighted imaging, T2*, FLAIR and gadolinium-enhanced T1-weighted imaging. The MRI data were analyzed by an examiner blinded to the data on cognitive performance. The tumor site was determined according to a previously validated method [16] that rated the presence of tumor tissue in the posterior fossa, lobar regions (frontal, insula, parietal, temporal and occipital) and the corpus callosum. The tumor’s sidedness was rated as left-side, right-side or bilateral (with the latter including midline tumors). We examined hippocampal atrophy (on Scheltens et al.’s scale) [17], hydrocephalus and any other lesions, including post-radiotherapy encephalopathy. Neuropsychological evaluation The comprehensive battery of tests (duration 2 h) was chosen in order to assess all the cognitive domains (general intellectual efficiency, language, visuospatial abilities, episodic memory, and executive functions) that may be impaired in this population (Table 2). Dysexecutive behavioral changes were assessed using the Behavioral Dysexecutive Syndrome Inventory [18]—a structured interview of an informant that assesses changes relative to previous behavior in
FCSRT free and cued selective reminding test
12 domains: (1) hypoactivity with apathy-abulia; (2) difficulties in anticipation, planning and initiation of activities; (3) disinterest and indifference to his/her own concerns and others; (4) hyperactivity–distractibility–psychomotor instability; (5) irritability–impulsivity–aggressiveness; (6) euphoria, emotional lability and moria; (7) stereotyped and perseverative behavior; (8) environmental dependency; (9) anosognosia–anosodiaphoria; (10) spontaneous confabulations; (11) social behavior disorders; (12) disorders of sexual, eating and urinary behavior. Generalized anxiety disorder and depression were diagnosed according to the DSM IV criteria [19]. Statistical analysis Neuropsychological data were analyzed using normative data obtained in 480 controls (36.4 % male) from the on-going GRECOG-VASC study [32]. The controls had a mean ± SD age 60.7 ± 11.1 years, a mean schooling of 11.5 years, and a mean Mini Mental State Examination (MMSE) score of 28.7 ± 1.4. Frequency of impairment The frequency of impairment was estimated from the overall neuropsychological score, according to a previously validated method [33]. Briefly, the transformed scores of ten cognitive tests were fed into a linear regression analysis with age, educational level, gender and an interaction
13
Author's personal copy 4
term and significant factors in controls were retained. Regression coefficients computed for controls were used to calculate standardized residuals, i.e. z scores (with poor performance corresponding to a negative z score). z scores were calculated for five cognitive domains: oral expression (using the Boston Naming test score), visuoconstructive abilities (using the Rey–Osterrieth Complex Figure Test score), verbal episodic memory (using third and delayed free and total recalls) [34], cognitive executive function (using the error interference index in the Stroop test; perseveration in the Trail Making Test part B and the Modified Card Sorting test; the completion times in the Trail Making Test part A and the naming and reading subtests in the Stroop test; the rank in the Six Elements task; the category achieved in the Modified Card Sorting test, and the two verbal fluency tests) and behavioral executive function (using an average of the 12 behavioral z scores) [17]. The overall neuropsychological score corresponded to the average of the z scores for the five cognitive domains. The dichotomization of performance (normal vs. impaired) was based on fifth percentile cut-offs determined from z scores in controls. The size effect was analyzed using Cohen’s d statistics (z score) and the odds ratio (OR) and its 95 % confidence interval (CI) for dichotomized performance [35]. A Cohen’s d ≥1.5 was considered to be a large effect size. Selection of tests for a shortened battery A shortened battery for the diagnosis of neuropsychological disorders in patients with grade I–III primary brain tumors was determined in a stepwise logistic regression analysis performed on all dichotomized z scores from patients and controls. The diagnostic accuracy of z scores selected by the regression analyses was judged according to the scores’ ability to discriminate between patients and controls via calculation of the area under the receiver operating characteristic (ROC) curves (AUC) and the corresponding 95 % CI. A false-positive rate ≤5 % (i.e. specificity ≥0.95) was used as a rule of thumb. The model that combined optimal diagnostic accuracy with the smallest number of tests was selected. The AUC value provided by the selected tests was compared to that provided by the overall neuropsychological score using Delong et al.’s method [36]. In addition we compared the AUC values of the overall neuropsychological score and the scores corresponding to those selected by the ICCTF (Trail Making test B time and perseveration, third and delayed free recalls of Free and Cued Selective Reminding test and the two verbal fluency tests).
13
J Neurooncol
Relationship between cognitive performance with tumor characteristics and treatment The relationship between the cognitive performance (indexed by overall neuropsychological score) and characteristics of tumor and its treatment (WHO grade; delay of assessment following surgery and following complementary treatment; seizure, NIH SS motor subscore; surgery, chemotherapy, radiotherapy) was examined using Pearson correlation analysis (WHO grade; delay of assessment following surgery and following complementary treatment, NIHSS motor subscore) and t test (surgery, chemotherapy, radiotherapy; seizure). Relationship with the functional outcome The relationship with the functional outcome was examined using the KI [15]. A poor outcome was defined as a KI
