Longitudinal study of cognitive function in glioma patients treated with modern radiotherapy techniques and standard chemotherapy

Abstract Introduction: Cognitive function is an important outcome measure in patients with brain tumor, providing information about the patient’s clinical situation, treatment effects and possible progressive disease. The aim of this longitudinal study was to evaluate effects of the currently used radiation and chemotherapy treatment on cognitive function and to investigate associations between cognitive function at baseline and progression as well as overall survival. Methods: 32 patients newly diagnosed with malignant glioma were evaluated at baseline with CNS Vital Signs (CNS-VS), a computerized standardized neuropsychological test battery, prior to arc-based radiotherapy and concomitant chemotherapy with Temozolomide. CNS-VS measures the cognitive functions known to be affected in patients with brain tumor, covering nine cognitive domains. Follow-up cognitive evaluations were performed in 26 patients after 3.5 months and in 13 patients 1 year after treatment start. Results: Overall cognitive scores were lower in the studied patient cohort at baseline compared to standardized domain scores. At 3.5 months follow-up cognitive functioning was slightly decreased, but only in 1/9 cognitive domains – visual memory – where significant changes were found compared to baseline test results. Similarly, at 12 months follow-up no significant changes in cognitive test results were seen compared to baseline examination, except for a decrease in the visual memory domain. In relation to early progression, the most significant cognitive deficits were dysfunctional visual memory and low executive functioning at baseline. Low executive function at baseline correlated most significantly with shorter overall survival. Conclusion: The present study suggests that the currently used arc-based radiotherapy and chemotherapy might affect cognitive function less negatively than previously described during treatment and in the first year after treatment in malignant glioma patients. In general, a high cognitive test score at baseline was associated with longer time to progression and with longer survival.


Introduction
Cognitive function has long been regarded as an important outcome measure in patients with brain tumor, providing information about a patient's clinical situation and results regarding treatment effects [1,2]. Cognitive deterioration can be the first indicator of progressive disease after treatment, as has been found in patients with low-grade gliomas [1,3] and in an early study on recurrent high-grade gliomas [4].
Cognitive function is now also recognized as an independent prognostic factor for the survival of patients with gliomas [5]. Measured by specific neuropsychological tests, cognitive impairment has been independently associated with poor prognosis in patients with newly diagnosed glioblastoma, where executive function, attention and verbal memory are the cognitive domains most closely associated with prognosis [5,6]. In patients treated for recurrent glioma, verbal memory has been independently and strongly related to overall survival after accounting for age, Karnofsky performance status score, histology and time to further tumor recurrence [7]. It has also been observed that treatment with radiotherapy plus concomitant and adjuvant temozolomide per se did not deteriorate cognition in glioblastoma patients with progression-free disease within the first 6 months of treatment [8]. Cognitive functioning may hence be important for treatment planning of individual patients and as a possible prognostic marker for tumor progression [4,9]. Documentation on cognitive functioning in patients with high-grade glioma more than 6 months after diagnosis has, until now, been scarcely found. This may be explained by the severity of the disease, with a short expected overall survival and even shorter time to tumor progression, which strongly limits the possibility of follow-up with cognitive testing even at 12 months from diagnosis.
Recent development of novel techniques such as volumetric-modulated arc therapy (VMAT) as well as helical tomotherapy (HT) have altered the conventional treatment paradigm in radiation treatment of patients with high-grade gliomas. These arc-based radiotherapies may reduce peak radiation doses in critical regions such as the hippocampus without compromising tumor control. Data on these new techniques is limited to planning and feasibility studies [10,11].
In this longitudinal prospective study, we studied the role of cognitive function in patients newly diagnosed with primary high-grade glioma before and up to 12 months after chemoand radiotherapy. Our major aim was to explore the effects of treatment on cognitive function and to describe clinical parameters such as tumor site, radiation dose, age and gender, depressive symptoms, operation method, steroid intake and antiepileptic medication and to investigate if these parameters are associated with cognitive function. The second aim was to explore if there are possible correlations between cognitive function at baseline before adjuvant treatment and time to tumor progression (TTP) and overall survival.
We hypothesize that chemotherapy and modern arcbased rotation radiation therapy per se have less adverse effects on cognitive function than previously described [1,2] in glioblastoma patients within the first year after radiotherapy, and that specific cognitive metrics at baseline before treatment might be a prognostic marker for TTP and survival.

Patients and study procedure
In this prospective study, 32 patients, (mean age of 59 years, range 35-68 years), newly diagnosed with malignant glioma were enrolled for cognitive testing from October 2013 to October 2017 at the Departments of Neurology and Oncology at Skåne University Hospital in Lund, Sweden. Patients' inclusion criteria were age >18 years, primary surgery with histopathologically confirmed diagnosis of World Health Organization (WHO) grade III or IV malignant glioma, standard postoperative oncological therapy plan, patient's consent for study participation in imaging monitoring and neurocognitive testing, including all patients admitted at the neuro oncological unit for standard postoperative treatment during the predefined study period who agreed to study participation. Cognitive examination was performed at baseline approximately four weeks after surgery but before starting chemo-radiotherapy according to the Stupp protocol [12], after 3.5 months and, if patients were alive and testing still feasible, after 12 months.
Information regarding survival, TTP and radiological response of the tumor lesion were collected until 1 October 2017, which was the predefined endpoint for this study. Evaluation of quality of life using the EORTC QLQ C-30 [13] questionnaire, for psychiatric symptoms according to the Hospital Anxiety and Depression Scale [14] and for neurological and clinical conditions were carried out on all three test occasions. Patient characteristics and treatment parameters are presented in Table 1.

Assessment of cognitive function, neurological function and quality of life parameters
Cognitive assessment was performed using CNS Vital Signs (CNS-VS), a computerized standardized test battery measuring nine basic cognitive functions, known to be impaired in patents with brain tumor. CNS-VS has been tested and validated in brain tumor patients [15], in traumatic brain injury, dementia, attention deficit/hyperactivity disorders [16,17] and in multiple sclerosis [18]. The CNS-VS test battery comprises seven conventional neuropsychological tests: verbal and visual memory, finger tapping, symbol digit coding, the Stroop Test, a test of shifting attention and a continuous performance test. These seven tests in turn cover nine cognitive  [20] and of clinical performance status according to The Eastern Cooperative Oncology Group (ECOG) Scale of Performance Status [21] were registered. Patients' medication including use of corticosteroids and antiepileptic drugs were noted.

Imaging
The imaging protocol included MRI examinations at baseline and at 3, 6, 12, 24 and 52 weeks after start of postoperative treatment. MRI was performed on a 3 T Siemens MAGNETOM SkyraV R (Erlangen, Germany), with a 20-channel head/neck coil. The MR protocol included the following sequences: axial T2 2 D TSE; axial 2 D FLAIR (fluid attenuated inversion recovery); T1pre-and post-Gadolinium (Gd) contrast administration with 1 mm 3 isotropic resolution and dynamic susceptibility contrast perfusion MR, with the aim to distinguish treatment effects such as pseudo-progression from true tumor progression.
Additionally, conventional clinical radiological monitoring was performed at 36 weeks from baseline. Experienced neuroradiologists performed evaluation of the imaging data continuously as part of clinical routine.

Statistical analyses
Semiparametric Cox regression analyses were conducted with censoring for patients still alive/free of progression on October 1, 2017. The proportional hazards assumption was tested and appeared reasonable for all cognitive domains. In the first step, analyses adjusted for one variable at a time were performed, where p < .05 was considered statistically significant. Significant variables were subsequently included in the final model.
Collinearity diagnostics were carried out to investigate if several variables responded similarly using standard techniques as implemented in SAS PROC REG https://support.sas. com/en/documentation.html. For details, see the online SAS documentation. For the remaining analyses, standard regression techniques based on Gaussian errors were used. Intraindividual changes in cognitive domain scores were analyzed by parametric regression-based t-test. The model assumptions were tested by QQ-plotting and plotting of standardized residuals by predicted values. A two-sided pvalue below .05 was considered statistically significant. All analyses were carried out using SAS 9.4, Cary, NC, USA.

Patient characteristics
In the study, 32 right-handed patients (23 males, 9 females) were included, all undergoing treatment with chemo-and radiotherapy. Ten patients (31%) had a tumor in the right hemisphere, 18 (56%) had a left-sided tumor localization and in 4 cases the tumors were bilaterally located. Fourteen patients (44%) had a predominantly frontal tumor localization and 18 patients (56%) had a predominantly non-frontal tumor. Eight patients (25%) had a gross total resection, 16 patients (50%) underwent a partial tumor resection and eight patients (25%) had only a biopsy for diagnosis. Histological analyses revealed glioblastoma WHO grade IV in all patients, except for one case diagnosed with anaplastic oligodendroglioma WHO grade III, (Table 1).
Thirty-one of the 32 patients underwent evaluations of cognitive function with neurocognitive examination at baseline. Follow-up cognitive evaluation was performed in 26 patients at 3.5 months after treatment start and 13 patients were also assessed at 12 months after treatment start. However, patients' clinical deterioration due to disease progression (7) or poor health, including medical side effects of therapy e. g. infections and general fatigue and deterioration in clinical performance status (ECOG Zubrod score) (2) or due to the tumor such as thromboembolism (1), or death (8) or drop out due to patients living long distance from the study center (2) limited the number of participating patients in 12 months , follow-up examinations. On the other hand, 7 patients participated at 12 months , testing despite progressive tumor disease.

Postoperative treatment parameters
All patients were treated with postoperative arc-based rotation radiotherapy by using volumetric modulated arc therapy (VMAT) or helical tomotherapy (HT), all patients receiving concomitant temozolomide (TMZ). Twenty-nine patients received 60 Gray (Gy), two patients received 34 Gy due to massive tumor burden and one patient received 10 Gy þ 56 Gy with interruption due to wound infection. Median equivalent biological radiation doses were: to left hippocampus 11 Gy, right hippocampus 13 Gy, left amygdala 11 Gy, right amygdala 11 Gy. After radiotherapy all patients were treated with adjuvant temozolomide. If tumor progression occurred during the study, second line therapy was offered, including, if feasible, reoperation, otherwise renewed TMZ therapy if recurrence occurred > 6 months after ending adjuvant TMZ, or CCNU (Lomustine) if recurrence occurred earlier. In summary, the treatment was in accordance with routine management and Swedish national guidelines for malignant glioma.
Among the 32 patients, 22 patients were medicated with corticosteroids at baseline and 20 patients received antiepileptic medication (carbamazepine, levetiracetam, valproic acid) at any time point during their participation in the study. A more detailed description of the patient characteristics, treatment and additional medication is presented in Table 1.

Survival and tumor progression (TTP)
For the complete patient cohort, the median survival was 608 days (range 83-1524 days) and the median TTP was 271 days (range 83-930 days) after diagnosis. Out of the 32 included patients, 21 patients died during the time of the study. Among these 21 patients, the median survival was 468 days (range 134-1052 days) with a median TTP of 257 days (range 133-662 days). The 11 patients who were still alive at the end of the study on 1 October 2017, had lived a median of 881 days (range 83-1524 days) after primary surgery and had a median TTP of 388 days (range 83-930 days).
Survival was negatively correlated with age, (p ¼ .0005), but showed no correlation with the sex of the patient in the Cox regression analysis. There was no correlation between TTP and the age or gender of the patients in the Cox regression analysis.

Cognitive function at baseline and follow-up cognitive examination
The results of the CNS-VS test are summarized in Table 2.
The mean standard score for the psycho motor speed domain was within the normal range, for visual memory domain within lower average, all other scores in the different tested cognitive domains were lower at baseline compared to standardized domain scores. At follow up after 3.5 months, intra-individual scores in all nine cognitive domains had decreased. However, this was statistically significant only in the domain of visual memory (p ¼ .015) analyzed by t-test.
None of the other eight tested domains showed statistically significant differences between baseline examinations and 3.5 months. Nor were any statistically significant differences present in the cognitive domain scores for the tested domains, apart from visual memory (p ¼ .045) at 12 months' follow-up examination, compared to baseline (Table 3).

Association between cognitive function at baseline and location of the primary tumor
Patients with frontal tumors, regardless of hemisphere, had significantly lower test scores for complex attention (p ¼ .03), executive functioning (p ¼ .03), visual memory (p ¼ .02) and verbal memory (p ¼ .02) compared to patients with nonfrontal tumors, assessed by t-test. Patients with tumors in the right hemisphere performed better in the cognitive domains of executive functioning (p ¼ .04) and both visual and verbal memory (p ¼ .05), compared to those with tumors in the left hemisphere.

Cognitive function at baseline related to type of surgery, steroid and antiepileptic medication
There were no significant correlations at baseline between cognitive function and the type of surgical operation method used, tested by linear regression analysis. However, a  Association between neurological performance status, quality of life and psychiatric symptoms at baseline and cognitive function Statistically significant correlations in the linear regression analysis were seen at baseline, between executive function and neurological (NIHSS) performance status (p ¼ .03) and between verbal memory and NIHSS status (p ¼ .01), where patients with good neurological function had higher scores in the executive function and verbal memory domains compared to patients with impaired neurological performance status. Analysis of cognitive functioning at baseline and change at 3.5 and 12 months versus numerical self-assessment sum scores of health and quality of life (EORTC QLQ C-30) and measurements of anxiety and depression (HADS) at baseline did not show any statistically significant correlations. In summery the patients self-reported their health and quality of life in EORTC QLQ-C30 as average to good and their psychiatric health as average according to the HADS self-assessment test results and this did not change over time.

Association between cognitive function at baseline and TTP
The better the cognitive test results, indicated by higher scores in the visual memory domain, complex attention domain and executive functioning domain at baseline, the lower the risk for tumor progression during the time of the study according to the Cox regression analysis. When taking into account all significant domains in the model, the most significant cognitive domains associated with progression were visual memory (Hazard Ratio (HR) ¼ 0.965; p ¼ .002) and executive functioning (HR ¼ 0.951; p < .0001). As revealed by collinearity diagnostics, the complex attention domain was highly intercorrelated with the executive functioning domain (Table 4).

Association between cognitive function at baseline and survival
High scores in the domains of visual memory, complex attention and executive functioning at baseline prior to radiochemotherapy treatment lowered the risk of death, analyzed at the end of this study by Cox regression analysis, although several of the domains were intercorrelated. The Cox regression analyses included adjustment for age, since a clear association between age and survival was present. Our results suggest that high age (HR¼ 1.134; p ¼ .0005) and low executive functioning (HR ¼ 0.926; p ¼ .001) may be associated with shorter survival (Table 5).

Association between change in cognitive function and tumor progression
Additionally, visual boxplot estimation of change of cognitive functioning in standard scores from baseline in relation to tumor progression or stable disease at 12 months were performed for visual and verbal memory domains, complex attention and executive function domains. This showed divergent median results with more deterioration at 3.5 months in standard scores among patients with progressive disease (n ¼ 14/25) at 12 months in visual and verbal memory and complex attention, but a median improvement in executive function standard scores compared to patients with stable disease at 12 months (n ¼ 11/25). Among patients who were able to participate at 12 months testing (n ¼ 13), median test results had less deterioration or were unchanged among patients with tumor progression (n ¼ 6/13) compared to patients with stable disease (n ¼ 7/13) for all four tested domains. In summary no clear association was seen between cognitive deterioration in CNS-VS and tumor progression in this cohort by visual estimation of boxplot.

Discussion
This longitudinal, descriptive study suggests that modern chemo-and radiotherapy might not negatively affect cognitive functioning in malignant glioma patients during and up to 12 months after initiated postoperative treatment. Despite a minor decrease in all cognitive domains at first follow-up at 3.5 months after initiated radiotherapy, the more refined radiation techniques currently used, with arc-based rotation radiotherapy, seem to have less neurotoxic impact on the patients, compared to previously used techniques [1,2].
Chemotherapy may have short-and long-term impacts on cognitive function of cancer patients, but in patients with a primary brain tumor this impact is assumed to be low [22].
The higher the cognitive domain scores patients had in general at baseline, the longer their TTP. Higher cognitive scores also were associated with longer overall survival. However, many of the cognitive domains were intercorrelated. The domain that significantly was associated independently with survival was the executive functioning domain. The executive functions are ascribed to the prefrontal networks, also called the executive system. Executive dysfunction may arise with damage to or pathological conditions affecting the prefrontal networks [23]. Antiepileptic drugs and corticosteroids may influence cognitive functioning in brain tumor patients. Previous studies have demonstrated that the cognitive domains that may be sensitive to treatment-induced dysfunction include attention and executive functions, memory and psychomotor speed [24,25]. We found no influence of antiepileptic medication on cognitive functioning. We did find an association between medication of corticosteroids and impaired executive functioning. This effect may be interpreted as a result of aggressive tumor growth requiring steroid use, rather than the steroid treatment per se, but direct disease-driving effects of corticosteroids have also been discussed [26].
It has recently been hypothesized that properties of the tumor itself and not the anti-tumor treatment may well underlie the cognitive deficits. One study of high-grade glioma patients examined before, during and after standard chemo-and radiotherapy demonstrated no clear association between cognitive functioning and white matter hyperintensities (WMH) or cerebral atrophy (CA) [27]. Furthermore, the same study showed that the majority of patients who declined in cognitive functioning at follow-up had tumor progression within 4 months.
Standardized neuropsychological assessment should be considered an essential part of the management and care of brain tumor patients, and possibly a tool toward providing more individualized treatments. In order to ensure reproducibility and for assessment in clinical trials, it is of importance to use standardized, comparable and comprehensive testing of all patients, i.e. testing should assess the cognitive domains sensitive to tumor and treatment effects [7,28]. CNS-VS fulfills all of these criteria, which is why computerized testing in these cases may be a better option than conventional neuropsychological testing or a complement to the latter [29,30]. Since most previous studies have utilized conventional test methods, CNS-VS is suitable for use as a screening instrument, or as a serial assessment measure, but should not be regarded as replacing standard neuropsychological assessments.
This study has limitations such as the modest number of patients and that less than half of the patients could be examined at 12 months, due to death or to being too impaired, or declined to participate. With the small number of patients, the amount of statistical analyses performed may have been over-fitted. In future studies larger patient cohorts would be beneficial for more reliable observations.

Conclusion and implications for future research
This longitudinal study has demonstrated that good cognitive function in patients with primary glioblastoma, at the time of diagnosis, is one of the positive prognostic factors for prolonged TTP and survival. No significant deterioration in cognitive function was seen within 12 months following the start of therapy, possibly due to modern, less neurotoxic radiotherapy methods. Neurocognitive examination should be regarded as an essential part of the pretreatment planning, to ensure that patients with high-grade glioma receive the best possible treatment combined with reasonable quality of life. With improved neuro-oncological treatment options, longer survival in high-grade glioma patients is expected. Further studies to evaluate the effect on neurocognitive function as a tool for therapy monitoring, including long-term follow-up in larger cohorts of patients with highgrade gliomas, are warranted. Ability to recognize and process information, motor speed, fine motor coordination, visual-perceptual ability. Important when, for example driving a car, or in occupational issues Executive function Shifting attention, see above Ability to sequence and manage multiple tasks simultaneously, tracking and responding to a set of instructions Summarized description of test domains and clinical relevance. Summarized and reproduced from: Gualtieri and Johson [16], and Meskal et al. [15].