Tumor-related neurocognitive dysfunction in patients with diffuse glioma: a systematic review of neurocognitive functioning prior to anti-tumor treatment

Introduction

Diffuse gliomas are progressive brain tumors that are almost invariably fatal, despite recent advances in treatment. Patients with diffuse glioma (WHO grade II–IV) often have deficits in neurocognitive functioning (NCF). Frequently reported problems concern word-finding, short-term memory and carrying out complex tasks [1–4]. Impairments in NCF negatively influence health-related quality of life (HRQOL) of patients and their partners [5–7]. For this reason, NCF is an important outcome measure of treatment, especially in patients with low-grade glioma (LGG, WHO grade II), who show relatively long periods of disease-free survival compared to patients with high-grade glioma (HGG, WHO grade III-IV).

The neurocognitive deficits in glioma patients may be found in one or more cognitive domains, such as executive functioning, language and memory. The neurocognitive deficits can be subtle and focal neurological deficits (leg or arm paresis, visual disturbances etc.) can overshadow these neurocognitive problems [5–8]. Therefore, to detect neurocognitive deficits in glioma patients, neuropsychological evaluation with sensitive and wide-ranging tests is required. Standard screening tests aimed at cognitive decay, such as the Folstein Mini Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA), cannot always provide relevant information, because they lack sensitivity and domain-specific information.

Factors affecting NCF can be related to the patient (e.g. age, level of education), the tumor (e.g. tumor grade, tumor location, biologic characteristics of the tumor) and the treatment [1, 9, 10]. The tumor can give rise to neurocognitive deficits through many mechanisms, such as local mechanical effects of the tumor mass and hereby ischemic changes, but also cell death by tumor-released excitotoxins and disturbances in synaptic transmission [11, 12]. These different mechanisms give rise to direct neuronal damage in the region of the tumor, as well as more widespread disturbances of functional brain networks [13]. Because NCF in glioma patients has primarily been studied postoperatively, data about the role of the tumor itself in neurocognitive outcome are scarce [3, 8, 14, 15]. After surgery, the effects of the tumor, surgery and medical therapies on NCF are difficult to distinguish.

Since NCF is an important outcome measure of treatment and the information about the occurrence of neurocognitive deficits in glioma patients before treatment is limited, we performed a systematic literature search to identify studies that evaluated NCF in glioma patients before surgery, chemotherapy or radiotherapy [16]. The aim of this systematic review is to give an overview of data about the occurrence of neurocognitive problems in glioma patients before anti-tumor treatment and about the cognitive domains in which these deficits occur, and to identify gaps of knowledge that may be the subject of future research.

Methods

We used the PRISMA-P (Preferred Reporting Items for Systematic review and Meta-Analysis Protocols) checklist to ensure good quality of the review (see appendix 1 in supplementary material for the PRISMA-P checklist). Because data from different articles was very heterogeneous and sometimes relevant data was missing, we had to make some important methodological choices. We summarized these choices in box 1 in supplementary material and explain them below.

Results

Discussion

Cognitive dysfunction in glioma patients has mostly been studied postoperatively. Therefore, the role of the tumor and its direct effects on cognitive functioning is not clear. Through this systematic review of the literature on NCF prior to treatment in diffuse glioma patients, we provide clinicians and researchers in this field a summary of the data on baseline (treatment naive) neurocognitive condition.

We found 23 heterogeneous, mainly small-scale studies with selected groups. NCF was analyzed in 14 studies at the group level, of which 13 studies (92.9%) found significantly decreased mean NCF scores in at least one of the tested domains. The proportion of individuals with a cognitive impairment was reported in 15 studies (covering an n = 709). The median proportion of patients with an impairment in at least one domain was 62.6% (IQR 31.0–79.0). Patients with high-grade glioma more often exhibit cognitive impairments than patients with low-grade glioma (OR 2.50; 95% CI 1.71–3.66).

Based on group analyses cognitive impairment appears to occur in the large majority of glioma patients prior to any anti-tumor treatment, and impairments can be found across all main cognitive domains. These findings support the hypothesis that the tumor and the direct effects of this tumor affects NCF in diffuse glioma. Looking more closely at the level of individual patients, a median proportion of impaired patients of 62.6% indicates that there is a considerable proportion of patients (one in every three) without clear cut cognitive impairments prior to treatment. Speculatively, this variability in the occurrence of cognitive impairment in treatment-naive glioma patients could be explained by tumor-related factors such as tumor location, tumor volume, mass effect of the tumor, WHO grade, histology and molecular markers. Concerning molecular markers, Wefel et al. recently found a complex interrelationship between patients’ NCF, growth velocity and the presence or absence of an IDH-mutation [38]. In addition, Correa et al. found an association between (germline) genetic polymorphisms and NCF in glioma patients, underscoring the role of variability in individual patients’ vulnerability for glioma-related cognitive dysfunction [39]. The available data were insufficient to formally test for the association of these tumor-related factors with NCF across studies.

The problem of neurocognitive dysfunction in glioma is often considered to be most relevant in LGG patients, given their relatively long survival and (often) young age. Analysis for subgroups based on glioma grade showed that HGG patients more often exhibited cognitive impairments than patients with LGG. For interpretation of this finding, it should be taken into account that this analysis was limited to eight studies. No separate data was available for the other included studies. Still, this finding supports the notion that adequate monitoring and treatment of cognitive symptoms is also of great importance in high-grade glioma patients.

This review demonstrates that the available literature is very heterogeneous with regard to the neuropsychological tasks used in the different studies. This limited the possibilities for synthesis of neurocognitive data. Our finding of heterogeneity in neuropsychological tests is supported by a recent systematic review of methods of cognitive assessment in glioma research [40]. To allow comparison across studies, we had to make use of a predetermined classification based on international standard and experience within our group (appendix 3 in supplementary material) to align classifications [17]. Given that each neuropsychological test taps more than one cognitive domain, differences can be found between classification in the original paper and our analyses. Therefore interpretation with respect to what cognitive domain is the most affected needs to be done with caution. However, the finding that impairments can be found in every cognitive domain is an important conclusion that is not affected by a possible bias as a result of our predetermined classification.

Another problem we faced was that most studies that reported on the proportion of patients with an impairment for each cognitive test, did not report on the number of individual patients that had below-threshold scores for impairment in at least one test within a certain cognitive domain. Therefore, we were not able to determine the exact frequency of patients with impairment within a certain domain. Our method of estimation of this frequency probably resulted in an underestimation of this frequency. Despite the limitations of these methodological choices, the consistent use of our methodology results in systematically obtained estimations of NCF at the group-level and individual patient-level. In determining the frequency of group—and individual-level of cognitive deficits, we decided to consider a domain to be affected on group level, if the mean-score of the group was statistically significant lower compared to norm/control data on any of the tests in that domain and on individual level if the patient performed below threshold on any of the available tests from this domain. By doing this we are able to take all cognitive tasks into account that have been reported, in addition we use rather stringent criteria of -2SD to indicate impairments. Although this approach carries the risk of overestimation of the severity of abnormalities, we feel that it is the best method to value test performance. Each individual test represents a specific function within a cognitive domain; therefore, we considered that one abnormal test—even with other tests results within normal-range—already represents dysfunction in that give domain. For all other methodological choices, we chose options that lead to possible underestimation of the actual cognitive deficits rather than overestimation. Altogether, our summary measures can be viewed as a conservative estimation of neurocognitive abnormalities in treatment-naïve patients with a diffuse glioma.

Furthermore, the risk of bias in the selected studies is high, because these studies were small observational studies that mostly selected patients based on certain glioma locations (selection on location in 11 studies, 47.8%). This selecting on location can lead to selection bias. For example, patients with a tumor located in the left temporal lobe have an increased chance of disturbances in the domain of language, leading to an overestimation of language dysfunction in glioma patients. Further, selection based on location led to more focus on specific cognitive domains such as language and visuospatial functioning and to a lesser extent on more widely distributed functions, such as executive functioning and memory. Unfortunately, the small sample sizes prohibited subgroup analysis of NCF for different glioma locations.

Another limitation is that another form of selection bias may have played a role; In five studies neuropsychological evaluation is done in patients who are undergoing awake surgery [2, 18, 20, 32, 37]. These are often patients with language dysfunction or a tumor in another eloquent area; hence, such patients are probably over represented in the available studies. Moreover, patients who are affected too severely will not undergo awake surgery, and restrain from comprehensive neuropsychological testing. So, preferably, future studies will include all patients with a diffuse glioma prospectively. If this is not possible, it should at least be described how the sample relates to the overall glioma population.

Finally, another possible confounding factor is the effect of medication. Several drugs may disrupt NCF in glioma patients, especially anti-epileptic drugs. At diagnosis of a glioma, drugs aimed at symptom relief will usually be administered at the shortest possible notice. It is therefore unlikely that someone will perform a study of NCF in patients with a glioma prior to first administration of such medications.

Overall we can conclude that most of the above mentioned limitations lead to an underestimation of cognitive functioning on the individual patient-level and group-level per domain and therefore underlines the importance of acknowledgment of the pre-treatment level of NCF. Only the selection of patients by tumor location may lead to an overestimation of cognitive problems in specific domains (for example language).

Conclusion

The literature about NCF in diffuse glioma patients prior to anti-tumor treatment is heterogeneous with regard to patient characteristics (glioma location and grade) and neuropsychological tests used in the different studies. Risk of bias is high, because of the small sample sizes and certain patient selection criteria. Based on the individual patient data, cognitive impairment occurs in more than half of all glioma patients prior to any anti-tumor treatment, across all main cognitive domains. Based on group analyses cognitive impairment appears to occur in the large majority of glioma patients prior to any anti-tumor treatment. These findings support the hypothesis that the tumor itself contributes significantly to NCF dysfunction in diffuse glioma. The variability in the occurrence of cognitive impairment in treatment-naive glioma patients remains largely unexplained. Research on the influence of tumor-related factors (e.g. localization, volume, grade, histology and molecular profile) is necessary in order to explain this variability. Deeper knowledge of the degree and origins of tumor-related cognitive dysfunction will likely facilitate the development of new strategies for treatment and rehabilitation. Furthermore, it is important to take the existence and extent of pre-treatment cognitive dysfunction into account in the research focusing on cognitive sequelae of treatments. In current clinical practice, adequate detection of cognitive dysfunction, especially in executive functioning, may guide the counseling and symptomatic treatment of patients.

Electronic supplementary material

Compliance with ethical standards