Linguistic processing speed in glioma patients
The relationship between linguistic and nonlinguistic cognitive abilities
Master’s Thesis
Saskia Mooijman
11317183
July 11
th, 2018
Supervisors
Dr. Laura S. Bos (University of Amsterdam) Dr. Djaina Satoer (Erasmus Medical Center Rotterdam)
Second reader
2 Abstract
This research aimed to disentangle glioma patient’s (N=37) linguistic processing speed abilities from nonlinguistic cognitive abilities and to compare these to a control group (N=35). The sensitivity of the objective measures was assessed by comparing it to reported complaints. Finally, the effects of awake surgical treatment were examined at two moments post-surgery.
The results showed that the patient group did not perform significantly worse on the Sentence Judgment Test assessing receptive linguistic processing speed. However, there was a subgroup of patients with gliomas in the dominant hemisphere showing impaired performance on this test. Importantly, linguistic processing speed appeared to be significantly correlated with nonlinguistic cognitive abilities, while a significant correlation was absent in the control group. The processing speed abilities were positively correlated with the severity of the reported word-finding problems of patients. The results of the postoperative assessments exemplified variation across participants but did not reveal severe postoperative deficits.
Highlights
• Patients with left-hemispheric gliomas show linguistic processing speed impairments • Receptive linguistic processing speed is correlated with nonlinguistic functions • Linguistic processing speed abilities are correlated with anamnestic complaints
Keywords
Glioma, Awake surgery, Processing speed, Anamnestic complaints, Cognitive functioning, Linguistic abilities, Preoperative assessment, Postsurgical outcome.
3
1. Introduction
1.1 Gliomas
Gliomas are the most common type of primary brain tumors. Yearly, there are approximately 1000 diagnoses in the Netherlands, mostly affecting young adults (Houben et al., 2006). Gliomas are categorized into four grades, according to the World Health Organization (WHO). High-grade gliomas (HGG, grades III-IV) are more aggressive and more common than low-grade gliomas (LGG, grades I-II; Sanai & Berger, 2012). DeAngelis (2001) identified the most frequent symptoms of LGGs and HGGs: headaches (40% vs. 50%, respectively), seizures (65-95% vs. 15-25%), hemiparesis (5-15% vs. 30-50%), and mental-status abnormalities (10% vs. 40-60%). LGGs are often located in the vicinity of eloquent areas of the brain (Duffau & Capelle, 2004). In these cases, surgery is aimed at resecting the tumor whilst preserving the cognitive functions (De Witte & Mariën, 2013; Ilmberger et al., 2008; Sanai & Berger, 2008).
The asleep-awake-asleep procedure in tumor surgery is used increasingly often and is currently considered to be the ‘gold standard’ for surgery in glioma patients (De Witt Hamer, Robles, Zwinderman, Duffau, & Berger, 2012). In this procedure, the patient is awakened and the surgeon applies direct electrical stimulation (DES) to the affected areas, which causes a transient (4 second) lesion to that site (Duffau, 2007). The assumption of this procedure is that this transient lesion allows the neurosurgeon to identify the brain regions involved in cognitive processes (Desmurget, Song, Mottolese, & Sirigu, 2013). By doing so, neurological and language functions can be mapped aiming for maximal tumor resection and fewer postoperative neurological deficits. A larger resection, in turn, is indicative of a longer life expectancy for patients with both LGGs and HGGs (Sanai & Berger, 2008). Due to the localization of LGGs in eloquent areas of the brain, patients may experience cognitive impairments pre- and postoperatively. The cognitive deficits of glioma patients are discussed in the following section.
1.2 Cognitive impairments
1.2.1 Cognitive assessment in glioma patients
As gliomas appear to have a preferential location for eloquent areas of the brain (Duffau & Capelle, 2004), patients suffering from a glioma regularly have cognitive deficits. Often-assessed higher cognitive functions are language, memory, attention, and executive functioning (EF). In the present study, EF is defined as consisting of three subcomponents: updating of working memory, shifting between tasks, and inhibition of prepotent responses (Miyake et al., 2000). The nature of the impairments appears to be different from stroke patients; glioma patients
4 may have more global deficits as compared to the very site-specific deficits often seen in stroke patients (Anderson, Damasio, & Tranel, 1990). Even after close matching of lesion locations, Anderson et al. found significant differences between tumor- and stroke patients, with the latter experiencing more severe language deficits. They further note that some tumor patients do not show any deficits on standardized tests. The subtlety of the difficulties of glioma patients pose problems for the assessment of their cognitive functions.
There is a growing body of research investigating the cognitive abilities of glioma patients. The methods of these studies vary, and the findings are not always clear-cut. Variability in findings can be due to differences in operationalization of cognitive functions or different inclusionary criteria of the patient groups. In the following, the pre- and postoperative cognitive deficits of glioma patients are discussed. This discussion is based on systematic reviews by Satoer, Visch-Brink, Dirven, and Vincent (2015); van Kessel, Baumfalk, van Zandvoort, Robe, and Snijders (2017); and van Loon et al. (2015). Van Loon et al. conducted a systematic review of the prevalence of cognitive dysfunction in LGG patients. The authors combined findings on pre- and postoperative impairments, taking together various therapeutic interventions (i.e., surgical resection, chemotherapy, radiotherapy). Satoer et al. reviewed the literature on the cognitive outcome after treatment but restricted their review to surgical treatment. Van Kessel et al. investigated cognitive abilities prior to tumor-treatment. In the present discussion, all studies from the reviews that looked at pre-treatment impairments resulting from a glioma are included. In reviewing the post-treatment findings, only studies investigating the effects of surgical treatment are discussed. As the systematic reviews included studies until November 2016, more recently published studies were added to the overview. The overview is summarized in Appendix A.
1.2.2 Preoperative cognitive outcome
The results for the preoperative findings are divided into analyses on the individual patient-level and group-level comparisons. The findings on the individual patient level are summarized in Figure 1 and the group-level findings in Figure 2. In these figures, only the results of tests assessing more specific cognitive functions are presented. A few studies used short neurological screening tools, such as the Karnosfky Performance Scale (KPS; Karnofsky, 1949) or the Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975). Studies that used the KPS (Bryszewski, Tybor, Ormezowska, Jaskólski, & Majos, 2013; Duffau et al., 2003; Duffau, Gatignol, Mandonnet, Capelle, & Taillandier, 2008; Teixidor et al., 2007) and the MMSE (Sarubbo et al., 2011) mostly find either normal performance or very mild disturbances, from which it can be concluded that these general measures are not sensitive enough for glioma patients. Language comprehension measures too, generally operationalized with the Token Test (TT; De Renzi & Vignolo, 1962), typically do not pose difficulties for patients. Three out of four studies in which a group analysis was conducted found no impairments in comprehension, and the percentage
5 of impaired patients in most studies does not exceed 20%. To detect language problems in glioma patients, more sensitive measures need to be used.
From the overviews in Figure 1 and 2 it becomes clear that the average percentage of patients showing impaired naming abilities lies between 3%-41%. Four studies show a significantly worse naming performance on group-level, whereas two studies failed to find a significant difference between patients and control participants. Fluency (semantic and phonemic) appears to be the most-often impaired linguistic ability, with the average percentage of impaired patients between 10%-55%. All six studies that conducted a group analysis, found glioma patients to perform significantly worse on fluency tasks compared to a healthy control group.
Figure 1: Preoperative findings on the individual patient-level. Language abilities are in blue (light gray), nonlinguistic cognitive abilities in green (dark gray). The gray dots refer to the reported percentage of impaired patients for one particular study. The total number of patients per cognitive domain is given behind the bars. Note that the verbal fluency tasks are categorized into the linguistic abilities, though some authors classify fluency skills as part of the executive functions.
Studies that include nonlinguistic cognitive functions in the assessment protocol of glioma patients typically incorporate tests for memory, attention, executive functions, and sometimes processing speed. Impairments in the working memory domain are found in 8%-64% of patients, and four studies that conducted a group analysis found it to be significantly worse in patients compared to a control group. One study failed to find a significant difference in working memory abilities between groups. Verbal memory impairments are observed in 0%-63% of patients, and the only study that investigated this on a group-level found it to be significantly worse than in healthy subjects. Attention poses problems for 7%-66% of the assessed patients. Three studies that carried out a group comparison found a significantly worse performance, while one study failed to find such an effect. Divided
6 attention is problematic for 9%-33% of patients. On group-level, two studies report a difference in patients compared to healthy subjects and two studies did not find a significant difference. Regarding executive functioning, 19%-78% of patients show impaired performance. Three studies find a significant difference between patients and healthy participants, whereas one study did not find such an effect. Lastly, 12-28% of patients have impaired processing speed. One study that conducted a group analysis found information processing to be significantly worse in patients, and one study failed to find a significant difference between patients and a control group.
Figure 2: Summary of studies with group-level comparisons of cognitive functions in glioma patients. Red (dark gray) bars indicate the number of studies that found a significantly worse performance compared to a healthy control group, green (light gray) bars indicate the number of studies that failed to find a significant difference between glioma patients and the control group.
In short, the reviewed studies have large variability in their results, though both linguistic and nonlinguistic deficits are frequently observed in glioma patients. Teixidor et al. (2007) conclude that while glioma patients often fail to show a deviant performance on functional scales, they may still experience cognitive deficits. Consequently, the authors stress the importance of extensive neuropsychological assessment. Importantly, there is a significant effect of the number of failed cognitive functions and quality of life (Le Rhun, Delbeuck, Devos, Pasquier, & Dubois, 2009), providing another motivation to find sensitive measurements for these patients. A final reason to implement more sensitive assessments preoperatively is that only those measures will give sufficient information for intraoperative testing. This, in turn, increases the likelihood of positive postoperative outcomes, which is discussed in the following section.
0 1 2 3 4 5 6 7 Naming Comprehension Verbal fluency Verbal memory Working memory Executive functioning Information processing Attention Divided attention
7 1.2.3 Postoperative cognitive outcome
Section 1.2.2 summarized the results for the preoperative assessment of glioma patients. However, the surgery itself can have effects on cognitive functioning. Despite intensive intraoperative monitoring, there appears to be a transient impairment, relative to the preoperative baseline, immediately after surgery. Some studies show a return to baseline in the months that follow the immediate phase after surgery, while other studies show a persisting deterioration compared to baseline. Satoer et al. (2015) conducted a systematic review of the literature on the postsurgical cognitive abilities in glioma patients. They found that from the included studies (N=17) that incorporated an immediate postoperative test moment, 79% found worsening in the language domain, 33% in the memory domain, and 75% in the EF domain. After the acute stage, five studies report an improvement to baseline (i.e., preoperative test moment), whereas six studies report a further deterioration. Satoer et al. conclude that the cognitive functions tend to decline immediately postoperatively, and in some cases return to the preoperative level in the following months. However, Satoer et al. (2014) investigated the longer-term trajectory after surgery and found that recovery can take up to one year. It is therefore important to consider the recovery after the three-month period post-surgery.
1.2.4 Relationship linguistic and nonlinguistic abilities
Most studies discussed in Sections 1.2.2 and 1.2.3 that included language- and nonlinguistic tests assessing glioma patients, find deficits in both modalities. In practice, it appears to be difficult to disentangle nonlinguistic cognitive abilities from linguistic abilities. Successful communication relies on integration of nonverbal and verbal abilities, and linguistic measures always (partially) rely on a nonlinguistic cognitive component. The role of (working) memory for language acquisition, production, and comprehension has long been recognized (e.g., Baddeley & Hitch, 1974), and there is increasing attention being paid to the contribution of EF to language ability. Reading and naming of objects, for example, requires appropriate functioning of the above-mentioned subcomponents of EF, as described by Altani, Protopapas, and Georgiou (2017). A participant needs to sequentially engage and disengage previous and upcoming stimuli (shifting); (s)he needs to suppress previously activated items (inhibition); and the phonological representations of the incoming stimuli in working memory need to be monitored (updating). Consequently, it is relevant to see how these nonverbal cognitive abilities are related to specific linguistic capacities in glioma patients. Santini et al. (2012) investigated the correlation between naming abilities and nonlinguistic cognitive abilities and did not find a significant correlation between these modalities. However, these authors only included word retrieval abilities in their comparison. It has been argued that single-word retrieval may be less dependent on other cognitive abilities as it does not pose large demands on additional computational processes (Rofes et al., 2018). Therefore, other linguistic abilities may be more suitable candidates to investigate the relationship between linguistic and nonlinguistic functions. In Section
8 1.3, (linguistic) processing speed will be discussed in more detail and it will be argued that adding this factor may provide a more sensitive way of assessing glioma patients.
1.3 Processing speed
Section 1.2 discussed the preoperative (linguistic) cognitive abilities of glioma patients and the cognitive outcome after surgery. In this section, one particular cognitive ability will be discussed in more detail. Lageman et al. (2010) nominated processing speed as one of the most important cognitive domains in the assessment of brain tumor patients. In accordance with this, Ek, Almkvist, Kristoffersen Wiberg, Stragliotto, and Smits (2010), found it to be the most-often impaired ability in glioma patients. Other studies have found processing speed in glioma patients to be significantly worse than in a healthy control group (Habets et al., 2014; Wefel, Noll, Rao, & Cahill, 2016). In addition, the tumor grade seems to significantly affect processing speed, as patients with higher-grade gliomas have more difficulties with processing speed (Noll, Sullaway, Ziu, Weinberg, & Wefel, 2014). These studies typically operationalize processing speed with a Symbol Digit Modalities Test or the Trail Making Test Part A (TMT-A, US Army, 1944), but these tests do not place a (high) demand on linguistic abilities.
From the clinical practice, however, there is evidence that glioma patients also have difficulties performing linguistic tasks with a time constraint. The question is whether these problems with processing speed in language tasks are due to an impaired underlying linguistic mechanism, or if they are caused by a more general cognitive impairment. Evidence for a potential discrepancy between language processing speed and accuracy comes from a study by Moritz-Gasser, Herbet, Maldonado, and Duffau (2012). They studied the correlation between lexical access speed and the ability to return to work after surgery in LGG patients and found that naming speed, rather than naming accuracy, significantly predicted return to work. While none of the patients showed deviant accuracy scores on the naming test, the group of patients that was able to return to professional activities was significantly faster in naming. Importantly, none of the patients in their study were classified as ‘aphasic’ according to the results of the Boston Diagnostic Aphasia Examination (BDAE, Goodglass & Kaplan, 1972). The BDAE is a test battery originally designed for stroke patients, and therefore may not be sensitive enough for glioma patients. Naming speed abilities could not be explained by nonlinguistic processing speed measured with the TMT-A. Moritz-Gasser, et al. consequently hypothesize that the increased naming times for these patients is related to the deterioration of the cognitive skills that are involved in lexicosemantic processing, exemplified by co-occurring deficits on measures of verbal working memory and executive functioning. Another recent study has showed that glioma patients are significantly slower on a speeded naming test, compared to healthy participants (Ras, Satoer, & Visch-Brink, 2018). Again, this difference could not be explained by nonlinguistic processing speed measured with the TMT-A or by naming ability measured with the Boston Naming Test (BNT; Kaplan, Goodglass, & Weintraub, 2001).
9 The findings of these studies suggest that there is a discrepancy between naming accuracy, naming speed, and general processing speed. As Moritz-Gasser et al. and Ras et al. both investigated processing speed in the production of language, it leaves the question of receptive linguistic processing speed open for investigation. Moritz-Gasser et al.’s findings highlight the importance of including processing speed in the assessment of glioma patients, as this was correlated with their ability to return to professional activities. Including linguistic processing speed can thus add to the understanding of cognitive difficulties that glioma patients experience by providing a more sensitive measure of their abilities. Another way of gaining more insight into the deficits is by investigating the anamnestic complaints of glioma patients, to be discussed in the following section.
1.4 Subjective experiences
The subjective experience of (language) abilities or impairments of patients is of central importance to quality of life (Cruice, Worrall, & Hickson, 2006). The problems they report are, after all, the issues that are most salient to the patients. Taphoorn and Klein (2004) report that objective cognitive tests may not necessarily reflect the patients’ complaints. This discrepancy can go both ways; patients may overestimate their abilities due to impaired judgment typically caused by a lesion in the frontal lobe (Taphoorn et al., 1992), or, in case of depression, patients may underestimate their abilities (Cull et al., 1996). In most cases, however, it appears that patients report difficulties that are not confirmed with standardized tests, demonstrating the lack of sensitivity of those tests. Satoer et al. (2012) compared scores on the Aphasia Severity Rating Scale (ASRS, Goodglass, Kaplan, & Barresi, 2001) to the self-reported problems by the patients and found that more patients reported issues in daily communication than were shown to have impairments based on the ASRS (57% vs. 39%, respectively). Similar discrepancies between subjective complaints and objective measures are found by Påhlson, Ek, Ahlström, and Smits (2003), Racine, Li, Molinaro, Butowski, and Berger (2015), and Antonsson et al. (2018). Of the self-reported deficits, word retrieval difficulties are among the most common complaints (Racine et al., 2015). Therefore, it is recommended to combine standardized tests and subjective complaints in the assessment of the cognitive abilities of glioma patients (see Taphoorn & Klein, 2004).
1.5 Present study
From this discussion of the literature it has become clear that preoperative language difficulties are frequently observed in glioma patients, although standardized tests are not always sensitive enough to identify impairments in this patient population. In addition, patients may exhibit reduced (linguistic) processing speed, though it remains unclear whether this is due to an underlying linguistic deficit or originates from a more general
10 impairment. Moreover, the reported complaints, reflecting the problems that are most salient to the patient, are not always accompanied by impaired scores on standardized measures. Finally, the effect of (awake) surgery appears to vary between individuals and cognitive domains. These findings are taken as a starting point for this research.
The aim of the present study into the processing speed abilities of glioma patients is threefold. Firstly, the project aims to disentangle glioma patients’ preoperative linguistic processing speed abilities from nonlinguistic cognitive abilities and to compare these to a healthy population. Secondly, to assess the sensitivity of objective measures, the outcomes of these tests are compared to patients’ subjectively reported complaints. Thirdly, the short- and long-term effects of surgery on these measures are examined. In doing so, the project not only adds to the understanding of how linguistic and nonlinguistic abilities are intertwined, but also aims to serve clinical purposes by providing information about prognostic factors and the sensitivity of tests. To meet these research aims, the following research questions are investigated:
1. Are glioma patients’ preoperative impairments of linguistic processing speed caused by an underlying language deficit, or rather by a nonlinguistic cognitive problem?
2. Are the anamnestic language problems reported by glioma patients supported by the results of objective measures? How are they related to the performance on a:
i. Time-pressured sentence judgment task, including semantic, syntactic, and phonological items? ii. Standardized naming task (Boston Naming Test) and a standardized task assessing receptive
language abilities (Token Test)?
iii. Time-pressured nonverbal divided attention test (Trail Making Test)?
3. What are the short- and long-term effects of awake surgery on linguistic and nonlinguistic abilities, as measured objectively and subjectively?
Previous findings in the literature lead to the following hypotheses about these research questions. The mildness of the symptoms experienced by glioma patients can lead to the absence of deviant test scores in oft-used standardized tests, such as the Boston Naming Test or the Token Test; however, preoperative problems with language processing speed (see Moritz-Gasser et al., 2012; Ras et al., 2018) are expected to be evidenced in the assessment of reaction times in a language comprehension task. If the difficulties with such a task are the result of a more global cognitive impairment, the patients will also exhibit a slower reaction time on a nonverbal task. Furthermore, it is hypothesized that there is a discrepancy between the subjective experience and the objectively measured abilities of patients, and that patients will report problems which cannot be confirmed by the results of the standardized tests (see Antonsson et al., 2018; Påhlson et al., 2003; Racine et al., 2015; Satoer et al., 2012). The addition of reaction time measures on the sentence judgment test is expected to lead to a more sensitive measure that can explain the anamnestic complaints. Finally, it is hypothesized that the recovery (up to preoperative baseline) after surgery will take up to one year (see Satoer et al., 2014), with a transient impairment after surgery. The study aims to systematically investigate the relationship between linguistic and nonlinguistic
11 cognitive abilities in glioma patients, taking into account both objective measures and subjective experience. How these research goals are met is explained in the next section.
2. Methods
2.1 ParticipantsThe patient group of the present study consists of glioma patients that underwent awake surgery at the Erasmus Medical Center (MC) in Rotterdam. Patients that underwent the awake procedure after March 2015 generally have been assessed with the required subtests of the Diagnostic Instrument for Mild Aphasia (DIMA, Satoer, De Witte, Vincent, Mariën, & Visch-Brink, 2016; Satoer et al., 2017) as a standard clinical work-up. For this reason, only patients from this time on were included in this study. This resulted in a database of 50 patients in total. Patients with a recurrent tumor that had to undergo surgery for a second or third time were excluded (N=10);1 three patients were excluded because of missing data (i.e., they had performed only one or two tests). One patient had previously been diagnosed with Developmental Dyslexia (DD), but it was decided to include this patient in the dataset. This resulted in 37 participants in the patient group. However, due to limited time before the surgery or to the individual abilities of the patients, not all tasks were administered to every patient. This led to the decision to create subsets of patients for every test moment and for each of the four tests, which will be described in more detail in Section 2.3. The tumor characteristics, including the grade and localization, may also influence test performance and were thus included as factors in the statistical model. The education level of the participants was classified into the seven-point scale proposed by Verhage (1964).
Healthy native speakers of Dutch constituted the control group in the study. In total, 35 healthy control participants took part in the study. 2 These were matched with the patients for age and education level, and gender when possible (gender has generally not been shown to influence test results on these tests, e.g., De Witte et al., 2015; Snitz et al., 2009). The demographic information of the patients and control participants per subset is given in Table 1. None of the participants were financially compensated for their participation. The study was approved by the Ethical Committee of Erasmus MC and all participants gave their informed consent.
1 Patients with a recurrent tumor are excluded because it is impossible to attribute their preoperative impairments to the presence of the tumor alone. Instead, their impairments may also be the result of the previous surgery.
2 Exclusionary criteria: no (history of) cardiovascular, neurological, psychiatric, or developmental language disorders; no toxic substance abuse; normal vision and hearing; no sleep medication, psychotropic or neuroleptic drugs.
12
Demographic characteristics for patients and control participants
Group Patients
(N=37) Control participants (N=35)
Gender Female 12 20
Male 25 15
Mean age (range) 44.56 (18-73) 42.75(19-61)
Mean education (range) 5.35 (3-7) 5.53 (3-7)
Handedness Right 32 N/A
Left 5 N/A
Tumor characteristics for 37 patients
Variable Count (%)
Hemispheric lateralization Left hemisphere 25 (68) Right hemisphere 12 (32)
Tumor localization: lobe Frontal 21 (56)
Temporal 6 (16) Insular 1 (3) Parietal 3 (8) Frontoparietal 2 (5) Parietotemporal 1 (3) Temporoparietal 1 (3) Frontotemporal 2 (5)
Tumor histological type Astrocytoma 15 (41) Oligodendroglioma 13 (35)
Glioblastoma 10 (30)
Xanthoastrocytoma 1 (3)
Tumor grade (WHO classification) Grade I 1 (3)
Grade II 21 (56)
Grade II 5 (14)
Grade IV 10 (30)
Table 1: Demographic and tumor characteristics. Education level based on Verhage (1964): Dutch classification system including 7 categories. 1: did not finish primary school, 2: finished primary school, 3: did not finish secondary school, 4: finished secondary school, low level, 5: finished secondary school, medium level, 6: finished secondary school, highest level, and/or college degree, 7: university degree).
2.2 Materials
2.2.1 Sentence Judgment Test
To test comprehension and linguistic processing speed in the semantic, syntactic, and phonological domain, the Sentence Judgment Test (SJT) from the DIMA is used. This test was originally designed for the Dutch Linguistic Intraoperative Protocol (DuLIP, De Witte et al., 2013; De Witte et al., 2015) and includes items specifically designed for this protocol, in addition to items taken from the WEZT, FIKs, and BOX part 6 (Bastiaanse, Maas, & Rispens, 2000; Rijn, Booy, & Visch-Brink, 2000; Visch-Brink & Bajema, 2000, respectively). For the DIMA, the test was shortened and the registration of reaction times (RT) was added to the accuracy measures. This test consists of 30 sentences, half of which contain errors in three different linguistic levels. The phonological items
13 aim to assess phonological awareness by including pseudo-words (Example 1). The syntactic items contain errors regarding verb inflection (tense and agreement), word order, and pronouns (Example 2). Finally, the semantic items include sentences with semantic anomalies (Example 3).
1. De zanper koopt een blando.
The zanper buy-AGR a blando
‘The zanper buys a blando.’
2. Lies zingt gisteren een lied.
Lies sing-AGR.PRES yesterday a song
‘Lies sings a song yesterday.’
3. De loodgieter repareert de regenboog.
The plumber repair-AGR the rainbow
‘The plumber repairs the rainbow.’
The patient group performed the SJT in the E-Prime software (Psychology Software Tools, 2012), whereas the control participants were assessed in Praat (Boersma & Weenink, 2018).3 The participants read the sentences on a computer screen and rated their correctness by pressing the keys ‘F’ for fout ‘wrong’ and ‘J’ for juist ‘right’ on the middle of the keyboard. The items were presented in randomized order, and the test contained four practice items to familiarize the participant with the procedure.
2.2.2 Trail Making Test
In addition to the SJT, nonverbal cognitive abilities of the participants are assessed using the Trail Making Test A (TMT-A) and B (TMT-B). In the TMT-A, the participant needs to connect numbers (1-25) in an ascending order on a paper. The TMT-B requires the participant to connect alternating numbers and letters (i.e., 1-A-2-B-3 etc.). The score on both tasks consists of the time in seconds it takes to finish. Sánchez-Cubillo et al. (2009) investigated construct validity of the TMT and found that visual search speed and perceptual speed are good candidates to underlie the A score, rather than the often-assumed motor speed (cf. Lezak, 1995). The TMT-B relies on working memory and task-switching abilities, whereas the TMT-TMT-BA, operationalized as the difference
3 Besides the assigned software, the design of the experiment is the same. A previous study showed that there was no overall
14 score B-A4, provides a good measure of executive functioning. The TMT has been shown to be a sensitive tool to assess the above-mentioned nonverbal abilities in populations with neurological impairments (Lezak, 1995). Reitan (1958) also showed that patients with brain damage scored significantly worse on the TMT-A, TMT-B, and TMT-BA than control participants.
2.2.3 Short Token Test
The Token Test (TT) was developed by De Renzi and Vignolo (1962) and has been shown to be a sensitive measure to detect the presence and extent of aphasia. Therefore, the TT is one of the most widely used tests for assessing aphasia severity. In this test, the participants are asked to move and point to geometric forms on verbal commands increasing in difficulty. The patients in the present study have been tested with the short form of the TT (De Renzi & Faglioni, 1978), which has 36 items. De Renzi and Faglioni established the cut-off score for a deviant (i.e., aphasic) performance at 29/36.
2.2.4 Boston Naming Test
The Boston Naming Test (BNT, Kaplan et al., 2001) consists of 60 black-and-white drawings of objects and animals ordered based on difficulty and frequency. The participant is asked to name the objects. The BNT has been adapted to Dutch by van Loon–Vervoorn, Stumpel, and de Vries (1995) and van Loon-Vervoorn (2005).
2.2.5 Anamnestic complaints
The subjective experience of the patients was estimated using anamnestic information gathered through a standard set of questions preceding the test-protocol. This set includes questions on encountered language-, memory-, attention-, and executive functioning problems. For the present study, there was a focus on word-finding problems, as they are among the most reported issues in glioma patients (Racine et al., 2015). After collecting the detailed information from the anamnesis, the complaints were categorized using three levels: 0: no complaints, 1: mild complaints, 2: clear complaints. If the patients only reported difficulties after more targeted questions, if they indicate that they ‘sometimes’ experience problems, or if their partner reported word-finding difficulties, their complaints are labeled ‘1’. If the patient presented their word-finding complaints centrally in the anamnesis,
4 Other oft-used calculations of the TMT-BA are ratio score B:A and the logarithmic transformation of the ratio score. In the
present study, difference score B-A is used as this has been found to be the best TMT indicator of executive functioning (Sanchez-Cubillo et al., 2009).
15 or with modifiers such as ‘often’, or ‘severe’, their complaints are labeled ‘2’. Below is an example of a mild complaint (Example 4) and a clear complaint (Example 5) presented during the anamnesis.
4. Patiënt rapporteert geen problemen op cognitief gebied. Na doorvragen meldt hij lichte woordvindproblemen. Handschrift ook iets slordiger.
‘Patient does not report cognitive problems. After additional questions, he reports subtle word-finding difficulties. Handwriting is also messier.’
5. Patiënt rapporteert woordvindproblemen, resulteert in het vermijden met mensen te praten. Woord is wel in het hoofd, maar kan niet worden uitgesproken. Schrijven en typen lukt ook niet. Verder klankverwisselingen, vergeten lidwoorden en functiewoorden.
‘Patient reports word-finding difficulties, that results in avoiding talking to people. Word is in mind but cannot be pronounced. Patient also does not succeed in writing and typing. In addition, there are phonemic paraphasias, and articles and function words that are forgotten.’
2.3 Procedure
2.3.1 Data collection patients
The data of the patients have been collected by the clinical staff at the Erasmus MC between March 2015 and November 2017. All tests were administered as part of a standard test protocol. Following conventional clinical procedure, test moments were scheduled twice or trice: preoperatively (T1), and twice postoperatively (T2 and T3), as recovery after surgery can take up to one year (Satoer et al., 2014). If patients were only tested twice, the assessment took place at T1 and T2. The first postoperative assessment is done at approximately three months after surgery. The average time between T1 and T2 is 14,5 weeks (range: 9 – 21 weeks). T3 takes place at approximately one year after surgery. The number of patients for each of the tests and test moments is presented in Figure 3 below. A database of patients that underwent surgery in this period was created by digitalizing their information and test scores from their files.
The results of the preoperative assessment were compared to the performance of the control participants. In order to gain insight into the short-term effects of surgery, the results of the patients were compared between T1 and T2 (if available). Because there were only four patients that were assessed with the SJT at three test moments, an extensive group comparison for this test was not possible. Instead, a more elaborate description in the form of a case series design of those four patients was carried out.
16 Figure 3: Number of patients per test and test moment. SJT: Sentence Judgment Task, TMT: Trail Making Test, BNT: Boston Naming Test, TT: Token Test. T1: preoperative assessment; T2: approximately three months postoperative; T3: approximately one year postoperative.
2.3.2 Data collection control group
A portion of the data of the healthy participants was already collected (Mooijman, 2018). These participants were tested again to account for test-retest variability of the SJT, as the same test was administered repeatedly in the patient group. The average time between the two test moments in the control group was 6.5 months. For the comparison between the control participants and the patients, the results of the second assessment are used. In addition, new control participants were recruited, resulting in a total control group of 35 participants. All participants were asked to sign the informed consent form. After being informed about the procedure of the tests, they were asked to fill in a short questionnaire including questions about their language-, education-, and medical background. The BNT, SJT, and TMT were administered in a random order. Every test was explained verbally to the participants. The entire procedure lasted approximately fifteen minutes.
2.4 Data analysis
All statistical analyses were carried out in R (R Core Team, 2016). The results on the SJT, TMT, TT, and BNT constitute the dependent variables. The data were analyzed using regression models in the R package lme4 (version 1.1-14, Bates, Mächler, Bolker, & Walker, 2015) and lmerTest (version 2.0-33, Kuznetsova, Brockhoff, & Christensen, 2015) to retrieve p-values. The accuracy scores on the SJT, consisting of a binary outcome variable, were analyzed with a Generalized Mixed-Effects Linear Regression (GLMER) model with random
50 patients DIMA Included patients N=37 BNT T1 N=36 BNT T2 N=23 BNT T3 N=11 TT T1 N=36 TT T2 N=22 TT T3 N=12 SJT T1 N=21 SJT T2 N=15 SJT T3 N=4 TMT T1 N=36 TMT T2 N=22 TMT T3 N=13 Recurrent tumor N=10 Missing data N=3 Excluded
17 slopes for participants and items. The RTs of the SJT were analyzed with a Linear Mixed-Effects Regression (LMER) model, again with random slopes for participant and items. The outcomes on the TMT, TT, and BNT were analyzed using a Linear regression Model (LM). Regression models rely on the assumption that the residuals are normally distributed. This assumption was evaluated after converging the models by visually inspecting the residuals plotted in quantile-quantile-plots. An α-level of 0.05 was adhered to in the present study.
The main predictor in each model was group (patients vs. control participants), and age and education level were included in all models. Gender was included as a counterbalancing predictor. Within the patient group, the effect of tumor grade (LGG vs. HGG) and hemisphere (left vs. right) was estimated in separate models. The categorical predictors were contrast coded so that each level of the predictors is compared to a fixed reference level. Contrast coding was used to ensure that the intercept is the grand mean of all levels of the predictors. All numerical predictors were centered around the mean to get a meaningful intercept in the models.
In the analysis of the SJT results, the linguistic domain (semantics, syntax, phonology), the target answer of the item, and the trial-by-trial sequence (i.e., the position of each item in the test) were included as additional within-participant predictors. Before the group analysis of the RTs of the SJT, the outliers were removed. Items with an RT below 500 milliseconds were removed as it is assumed that participants need at least 500 milliseconds to assess an item. In addition, items with an RT above 10 seconds were removed, as the E-Prime experiment included a time-limit and any responses longer than 10 seconds were classified as null responses. This led to the exclusion of 14 trials (0.8%). Thereafter, outliers per participant were calculated and removed from the dataset using the trimr package (Grange, 2015) in R. An outlier was defined as an RT value of 2 SD above or below the mean for each participant. This led to the exclusion of 86 trials (5%). The remainder of the RTs were log-transformed to normalize the data and meet the model criteria. The log-log-transformed RTs provided a good fit for the raw data (ρ=0.96, p <0.0001).
First, the preoperative abilities were analyzed in group-level analyses between the patients and the control group. Second, the performance on the SJT and the TT was analyzed on the individual patient level by comparing it to normative data. For the TT, the Italian norms by De Renzi and Faglioni (1978) were used. For the SJT, normative data stratified by age and education level in a group of 140 healthy native speakers of Dutch were used (Satoer et al., 2017). A score of 1,5 standard deviations below the mean of the appropriate subgroup indicates mild difficulties, 2 standard deviations from the mean points to severe difficulties.
To estimate the relationship between the anamnestic complaints and the scores on the objective measures, the correlation between these complaints and the scores was calculated using Pearson's product moment correlation coefficient. In the analysis of the effects of awake surgery, test moment (T1, T2, T3) was included as a predictor. To assess the recovery within three months after surgery, the outcomes on the measures at T2 are compared to T1 and to the control group. In order to evaluate the long-term recovery, the scores at T3 are compared to T2, T1, and to the control group. There were four patients that performed the SJT at T3, rendering a
18 group comparison for this test moment impossible. Therefore, a group comparison of only T1 and T2 was carried out. The recovery trajectory of these four patients is discussed in more detail in the results of the recovery after surgery (Section 3.3). In the Appendix C, an elaborate description of each of the cases is given. The results of the data analysis are presented in the following section, and a detailed summary is given in the tables in Appendix B.
3. Results
3.1 Preoperative group analyses 3.1.1 Sentence Judgment Test
The test-retest variability of the SJT in the control group was assessed in a preparatory analysis. At the first assessment, control participants had an average RT of 3632.03ms per item. On the second assessment, the average RT was 3222.18ms. This is a non-significant difference between test moments (ß=-180.7, SE= 707.64, p=0.78). The accuracy scores also did not significantly deviate between test moments (ß=1.01 odds ratio, SE=1.53, p=0.97).
Overall, the mean accuracy was higher for correct target items than for incorrect target items (ß=2.2, SE=1.37, p=0.02). The mean accuracy scores for the incorrect target items per group and per linguistic domain are presented in Figure 4.5 For the incorrect target items, there was a significant difference between the control participants and the patients. The control participants were 4.86 times more likely to obtain a better accuracy score on the SJT than the patients (SE=2.16, p=0.04). The accuracy scores were not significantly influenced by tumor grade or -location.
Across all participants (healthy participants and patients), the incorrect target items are responded to faster than the correct target items (ß=-2.71, SE=0.06, p <0.001). The mean RTs for incorrect target items for each linguistic domain in the SJT are presented in a bar plot in Figure 5 below. In the group-level comparison, it appears that the patients had slightly longer RTs for each linguistic domain compared to the control group, but these differences did not reach significance (ß=0.013, SE=0.11, p=0.91). Within the patient-group, tumor grade did not significantly affect RTs, but patients with a tumor in the left hemisphere were slower in performing the task (ß=-1.03, SE=0.26, p<0.001).6 Education level played a beneficial role in all participants; higher education
5 To assess response biases in both groups, A’ as measure of yes-biases were calculated (Grier, 1971). The patient group and
control group both showed values for A’ that reflect good discriminability (A’ value near 1).
6 As mentioned in Section 2.1 Participants, there was one patient in the dataset who had previously been diagnosed with
Developmental Dyslexia (DD). As the SJT is a reading task, and reading performance is typically hampered in people with DD, this may have caused an additional disadvantage for this patient. However, an additional analysis excluding this patient showed that the results of the group comparison were virtually the same.
19 levels were accompanied by shorter RTs (ß=-0.16, SE=0.071, p=0.03). There was an interaction between age and education, indicating that the facilitating effect of higher education level was larger in older participants (ß=0.021, SE=0.006, p=0.01). Both the patient- and the control group became faster at responding to items as the task progressed, exemplified by the trial-by-trial sequential effects (ß=-0.012, SE=0.003, p<0.001). Finally, the RTs for the items with syntactic errors were significantly longer than the RTs of the phonological items (ß=0.39, SE=0.058, p <0.001), as were the RTs for the semantically anomalous items (ß=0.19, SE=0.056, p =0.004).
At the individual patient level, it appears that there is a subgroup of patients with deviant RT scores on the SJT compared to normative data (Satoer et al., 2017). Three out of 21 (14%) patients had mild difficulties (≥1.5SD from population mean) in at least one of the three linguistic domain, and three patients (14%) had severe difficulties (≥2SD from population mean) in at least one domain. All patients with deviant scores on the SJT had a glioma in the dominant hemisphere, and all but one in the frontal lobe. Four of the six patients with deviant scores had a high-grade glioma. Out of the patients with normal RT scores, on the other hand, nine (60%) had a tumor in the nondominant hemisphere. Ten patients (66%) had a low-grade glioma, while five (33%) were diagnosed with a high-grade glioma.
3.1.2 Trail Making Test
The results for the TMT are presented in Figure 6. It took healthy participants on average 30.6s to complete the TMT-A. The mean score for the patients was 29.9s. The difference in scores on the TMT-A between control participants and patients was non-significant (ß=1.95, SE=3.0, p=0.51). Education level influenced test performance, as participants with a higher education level needed less time to complete the task (ß=-4.31, SE=1.83, p=0.02). Patients with a high-grade glioma took on average 8.21s longer to finish (SE=3.77, p=0.04).
To complete the TMT-B, the control participants and patients needed on average 58.14s and 80.22s, respectively. There was no significant main effect of group on task performance, though there was a three-way interaction between group, age, and education level (ß=-5.29, SE=1.39, p < 0.001). This indicates that the main effects of age (older participants need more time) and education level (participants with lower education level need more time) were larger in the patient group compared to the control group. Similar to the results for the TMT-A, patients with a high-grade glioma needed more time for the test (ß=49.69, SE=19.13, p=0.01) than patients with low-grade gliomas.
Lastly, the score on the BA, operationalized as the difference score between the A and TMT-B, was 27.51s for the control participants, compared to 50.29s for patients. The higher the education level on the Verhage (1964) scale, the smaller the difference score of TMT-BA (ß=-13.84, SE=6.0, p=0.03), though this effect of education level was larger in older participants as illustrated by the interaction between age and education (ß=2.61, SE=0.61, p < 0.001). Similar to the results of the TMT-B, there was a three-way interaction between
20 participant group, age, and education level (ß=-4.85, SE=1.23, p < 0.001), which indicates that the interaction between education and age was larger in the patient group than in the control group. The difference between TMT-A and -B was larger for patients with a high-grade glioma compared to low-grade glioma patients (ß=-41.47, SE=17.27, p=0.02).
3.1.3 Token Test
The mean scores on the TT are presented in Figure 7. Since the control participants did not perform the TT, there are no comparisons between groups for this test. The mean score of the patients was 33.72 points. Adhering to the cut-off score of 29 (De Renzi & Faglioni, 1978), only two patients showed deviant scores on the TT. Patients with a low-grade glioma scored on average 5.71 points higher on the TT than patients with a high-grade glioma (SE=1.95, p=0.01). Age and level of education did not significantly influence performance on the TT.
3.1.4 Boston Naming Test7
The mean scores on the BNT are given in a bar plot in Figure 8. The control participants had a mean score of 53 (88%) on the BNT, while the mean score of the patients was 49 (82%). The results of the linear model show that this difference between groups was not significant (ß=2.54, SE=1.79, p=0.16). Although there was no overall effect of age, there was a cross-over interaction between group and age (ß=0.45, SE=0.13, p < 0.001) which indicates that the effect of age had opposite effects in the two groups. Within the patient group, older participants had lower scores while the opposite was true for the control group. There was a significant effect of education level; participants with a higher education level performed better on the naming task (ß=4.19, SE=1.1, p < 0.001). This effect was larger for younger people compared to older people, as illustrated by a significant interaction between age and education (ß=-0.29, SE=0.11, p=0.01). However, these effects could be qualified by a significant three-way interaction between group, education level and age (ß=0.59, SE=0.22, p=0.01), which indicates that the interaction effect of age and education was larger for patients than for control participants. A separate analysis showed that within the patient group, there was no significant effect of grade or location of the tumor.
3.1.5 Relationship between tasks
In the control group, there was no correlation between the RTs on the SJT and the TMT-A (ρ=0.20, p=0.26), the TMT-B (ρ=0.10, p=0.59), and the TMT-BA (ρ=-0.01, p=0.97). There were no significant correlations between
7 Eight patients performed the Object Naming task of the DuLIP, consisting of 100 items, instead of the BNT. Their scores
21 the accuracy scores on the SJT and the TMT. Within the patient group, however, the RTs on the SJT appear to be moderately correlated with the performance speed on the TMT-A (ρ=0.617, p=0.004), TMT-B (ρ=0.637, p=0.003), and the difference score TMT-BA (ρ=0.587, p=0.006), indicating that longer RTs on the SJT were accompanied by longer completion time on the TMT. Similar to the control group, this is in absence of significant correlations between the accuracy scores of the SJT and the RTs on the TMT. The RTs and accuracy scores of the SJT were also not significantly correlated with the results of the TT.
Figure 4: Mean accuracy scores on the SJT for incorrect target items per linguistic domain and overall correct target items.
Figure 5: Mean RTs (ms) on the SJT for incorrect target items per linguistic domain and overall correct target items.
Figure 6: Mean scores TMT, operationalized as time to complete the task (s), and the difference score between B and A (s).
Figure 7: Mean scores on the Token Test for patients (cut-off score for aphasic classification: 29/36). 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Semantics Syntax Phonology Correct Targets Control participants Patients - T1 Patients - T2
0 500 1000 1500 2000 2500 3000 3500 4000
Semantics Syntax Phonology Correct targets Control participants Patients - T1 Patients - T2
0 20 40 60 80 100 120 TMT-A TMT-B TMT-BA
Control participants Patients - T1 Patients - T2 Patients - T3 0 5 10 15 20 25 30 35
22 3.2 Relationship subjective and objective measures
Preoperatively, 16 out of 37 patients (43%) did not report any word-finding difficulties. Twenty-one patients reported word-finding problems of which eleven patients (30%) reported mild word-finding problems, and 10 patients (27%) reported clear word-finding problems. The experienced word-finding problems were not reflected by the outcomes of the BNT; only eight out of 21 patients (38%) with reported word-finding problems also showed deviant scores on the BNT. In addition, the two outcomes (reported word-finding difficulties and BNT scores) were not significantly correlated (ρ=-0.11, p=0.51). The accuracy scores of the SJT and the scores on the TT also did not correlate with the reported complaints. There is, however, a strong uphill linear relationship between the RTs on the SJT and the anamnestic complaints (ρ=0.688, p <0.001), indicating that more severe complaints are accompanied by longer RTs on the SJT. Additionally, the scores on the TMT-A, -B, and -BA are weakly correlated with the reported complaints (ρ=0.351, p=0.04 for TMT-A, ρ=0.415, p=0.01 for TMT-B, ρ=0.393, p=0.02 for TMT-BA).
3.3 Recovery after awake surgery 3.3.1 Group analyses
The mean scores for the measures on the different test moments are presented in Figures 4-9. Overall, the patients did not show significantly different RTs or accuracy scores on the SJT before (T1) and three months after awake surgery (T2) compared to the control group. However, at T2 patients started to show longer RTs for the semantic items, compared to healthy participants (ß=370.06, SE=171.29, p=0.03). The scores on the TMT-A remained Figure 8: Mean scores on the BNT for the control group and the patients
for each test moment.
Figure 9: Reported word-finding difficulties during the anamnesis at the three test moments.
0 10 20 30 40 50 60 Control
participants Patients - T1 Patients - T2 Patients - T3
0% 20% 40% 60% 80% 100% T1 T2 T3
23 stable at three months and one year after surgery and did not significantly deviate from the control group or T1. The scores on the TMT-B and TMT-BA, however, showed a significant increase in time to complete for the patients at T2 compared to the control group (ß=26.42, SE=10.81, p=0.02 and ß=28.86, SE=9.37, p=0.003, respectively). This difference became smaller and no longer deviated significantly from the healthy participants at T3. A similar pattern was found for the results of the BNT; the patient group had a significantly lower score at T2 compared to the control group (ß=-6.63, SE=1.97, p=0.002), a difference that was no longer significant at T3. As the control group did not perform the TT, only a comparison between test moments of the patients was carried out. The differences in scores on the TT did not significantly change at three months or one year after surgery, compared to the preoperative baseline.
Regarding the anamnestic complaints (Figure 9), the distribution of complaints appears similar across the three test moments. Three months after surgery, 8 out of 23 patients (35%) reported no problems finding words, whereas 7 (30%) and 8 (35%) reported mild or clear word-finding issues, respectively. One year after surgery, 4 out of 12 patients (33%) reported no problems finding words, 6 out of 12 patients reported mild word-finding difficulties, and 2 (17%) had clear complaints.
3.3.2. Case descriptions
The SJT was only administered at three test moments in four patients. Therefore, a group analysis of these results was not possible. In this section, the four cases that performed the SJT at three test moments are described in more detail. The average RTs on the SJT for each case are presented in Figure 10. Scores that significantly deviate from the normative data (Tombaugh, 2004) are indicated with an asterisk. The four cases show different recovery patterns after awake surgery, the complete case descriptions can be found in Appendix C. The first case is C1, a 21-year-old male (Verhage 5). He demonstrated preoperative deficits with regard to the BNT, TMT, and accuracy on the syntactic items of the SJT (i.e., >2SD from population mean). At three months after surgery, his RTs on the SJT worsened and deviated from normal for the syntactic items. The accuracy scores improved and were within the normal range. His RTs recovered at one-year post-surgery, as his RTs at this test moment were even slightly shorter than the preoperative baseline. The second case, C2, is a 35-year-old male (Verhage 7). He had a glioma in the nondominant hemisphere and did not show impaired performance on the tests at any test moment besides a deviant accuracy score on the semantic items preoperatively, which was resolved postoperatively. The third case, C3, is a 36-year-old male (Verhage 5) and showed yet another recovery pattern. He did not show impaired performance on the BNT, TT, and TMT, but had deviant RTs on the semantic and syntactic items of the SJT preoperatively. This improved at three months after surgery, though remained impaired for the syntactic items. His RTs on the semantic and syntactic items were similar at one year after surgery, but his RTs on the phonological items increased. C4, a 36-year-old male (Verhage 6), is the final example. He showed
24 unimpaired accuracy scores on the language tests both preoperatively and postoperatively. His linguistic and nonlinguistic processing speed decreased slightly after surgery but did not deviate significantly from the norm, except for the syntactic items on the SJT.
Figure 10: Mean RTs (ms) on the SJT for four cases.
4. Discussion
4.1 Linguistic processing speed in the SJT
Linguistic processing speed of glioma patients measured with a new test, the Sentence Judgment Test (SJT), was the topic of investigation in the present study. The study aimed to examine the underlying mechanisms of linguistic processing speed and to see how this was related to the subjective experience of the deficits. The main research question was whether the performance of glioma patients in a time-constrained language test is influenced by a primary linguistic mechanism or by a more general cognitive mechanism. Previously, language
0 1000 2000 3000 4000 5000 6000 7000 T1 T2 T3 T1 T2 T3 T1 T2 T3 Semantics Syntax Phonology
C1 0 1000 2000 3000 4000 5000 6000 7000 T1 T2 T3 T1 T2 T3 T1 T2 T3 Semantics Syntax Phonology
C2 0 1000 2000 3000 4000 5000 6000 7000 T1 T2 T3 T1 T2 T3 T1 T2 T3 Semantics Syntax Phonology
C3 0 1000 2000 3000 4000 5000 6000 7000 T1 T2 T3 T1 T2 T3 T1 T2 T3 Semantics Syntax Phonology
C4 *
*
* *
25 tasks with a time constraint have been reported to be difficult for glioma patients (Moritz-Gasser et al., 2012; Ras et al., 2018). The results of the current study, however, did not find evidence for a deviant performance on the RT measures of a SJT testing receptive language abilities. The individual patients with deviant RT scores on the SJT were identified in a subsequent analysis. The patients in this group all had a glioma in the left (dominant) hemisphere, while the majority of patients that did not show a deviant RT score had a glioma in the nondominant hemisphere. The occurrence of a high-grade glioma was also more frequent in the group of patients with deviant RT scores. These findings are in line with Hahn et al. (2003) and Habets et al. (2014), who also found that patients with a tumor in the dominant hemisphere experience more cognitive problems.
4.2 Underlying mechanisms of linguistic processing speed
The potential underlying mechanisms of linguistic processing speed were investigated by analyzing the correlations of performance on the different tasks. The results showed that the performance speed on the linguistic task (SJT) was correlated with the performance on the nonlinguistic tasks (TMT), indicating that lower performance speed on the TMT co-occurred with longer reaction times on the SJT. The cognitive abilities said to underlie performance speed on the TMT are perceptual speed (TMT-A), working memory and task-switching (TMT-B), and executive functioning (TMT-BA). Remarkably, a significant correlation only existed in the patient group and was absent in the control group. Moreover, the patients’ scores on the TT were not correlated with the RTs or the accuracy measures of the SJT. This would suggest that glioma patients perhaps compensate for their deficits by relying more heavily on a range of different cognitive skills when performing a task.
These results indicate that the receptive linguistic processing speed measured with the SJT partially relies on more general cognitive speed. This finding is in contrast with Ras et al.’s (2018) and Moritz-Gasser et al.’s (2012) results, which failed to find a significant correlation between the RTs on a rapid naming test and overall processing speed measured with the TMT-A. One potential explanation for this discrepancy in findings lies in the difference between modalities of the used language tests. In the present study, receptive reading abilities were measured, whereas Ras et al. and Moritz-Gasser et al. investigated the results of a speeded naming test assessing language production. As Sánchez-Cubillo et al. (2009) noted, the TMT-A mainly relies on visual search and perceptual speed. Therefore, a comparison between a reading task (perceptual and visual) and the TMT-A may result in stronger relationships than with a naming task. Moreover, multiple linguistic levels are combined in the SJT. The participants have to assess correct sentences and sentences that contain a semantic, syntactic or phonological error. These correct and incorrect items are presented in an alternating manner. It could thus be argued that there is constant task-switching within the SJT, placing a higher demand on executive functioning and attention (Rubinstein, Meyer, & Evans, 2001). Combining various tests and presenting them in a rapidly alternating way has previously been shown to be a good way to assess brain tumor patients (De Witte, Satoer,
26 Colle et al., 2015). A final, related explanation is that (speeded) naming tests may measure linguistic abilities in a more isolated manner, making it harder for participants to rely on other abilities, while the SJT requires the participant to integrate various processes in order to perform the task.
It is important to note, however, that there also is a dissociation between the results of the TMT, operationalized as the time to complete the task, and the reaction times on the SJT. While there was a tendency of lower RT scores on the SJT to co-occur with lower scores on the TMT, on the individual patient-level the opposite pattern was also observed. The existence of such a double dissociation is evidence that linguistic processing speed is not entirely dependent on more general cognitive abilities.
4.3 Anamnestic complaints and processing speed
Section 3.2 discussed the preoperative anamnestic complaints of word-finding difficulties of the glioma patients. The preoperative results showed that 57% reported word-finding deficits, of which 30% mild deficits, and 27% clear word-finding problems. This illustrates that experienced word-finding problems are not exceptional in glioma patients. However, their complaints were not always supported by deviant scores on the BNT. This is a problem, as the BNT is a widely used test for assessing word retrieval deficits. The discrepancy between reported complaints and scores on objective measures has been found before (Antonsson et al., 2018; Påhlson et al., 2003; Racine et al., 2015; Satoer et al., 2012). In addition, there is evidence that impaired linguistic variables found in spontaneous speech of glioma patients do not correlate with standardized language tests (Satoer, Vincent, Smits, Dirven, & Visch-Brink, 2013; Satoer et al., 2018). If the scores on the BNT do not reflect the subjective experience, either this test is not sensitive enough or the reported word-finding difficulties originate from another deficit.
When the RT scores on the SJT were compared to the anamnestic complaints, the two turned out to be significantly correlated. More severe reported word-finding complaints appeared to be accompanied by longer RTs on the SJT. In addition, the reported complaints were correlated with the TMT, measuring nonlinguistic functions, although this correlation was much weaker. The relationship between processing speed and the presence of reported word-finding difficulties seems to suggest that the reported complaints may be caused by an impaired network that is involved with both language and general cognitive processes. Everyday communication requires the conversational partners to process information quickly and respond to it in an appropriate manner. Cognitive functions, attention and EF in particular, have been shown to play a crucial role in the successful everyday communication of aphasic speakers (e.g., Ramsberger, 2005). It may therefore be unsurprising that reported complaints are accompanied by longer RTs on linguistic- and nonlinguistic tasks. It could also be the case that word-finding difficulties are more salient problems for patients. This idea is supported by findings of Racine et al. (2015), who showed that word-finding problems are one of the most often reported complaints.
27 Patients experience these issues in every conversation they engage in, whereas it may be easier to avoid usage of- or compensate for other impaired functions. Thus far, the preoperative findings have been discussed. The postoperative performance on the tests is addressed in the next section.
4.4 Transient impairments after awake surgery
Patients were tested at three months and one-year post-surgery. The results showed that the transient impairment reported in the literature (Bello et al., 2007; Chainay et al., 2009; Moritz-Gasser et al., 2013; Sarubbo et al., 2011; Teixidor et al., 2007; Yoshii et al., 2008) should perhaps be extended from the immediate postoperative stage to a longer period after surgery. This is in accordance with Satoer et al. (2014), who found that recovery after surgery can take up to a year; much longer than the typical follow-up in the literature. RTs on the semantic items in the SJT deviated from the control group at the test moment three months post-surgery but improved and no longer deviated at one year after surgery. The same pattern was found for the scores on the BNT, TMT-B and TMT-BA. However, it must be noted that the patients showed few impairments at the preoperative baseline compared to the control group.
The variability in the recovery trajectory of glioma patients was also illustrated with four cases, each of whom performed the SJT at all three test moments. The cases exemplified different trajectories: from unimpaired performance at all test moments (C2), to improvement after surgery (C1 and C3), to subtle worsening postoperatively (C4). The variation in recovery after surgery makes it difficult to draw firm conclusions. However, one important finding is that the results of both the group study and the case studies do not reveal severe postoperative deficits, thereby corroborating Satoer et al.’s (2014) findings. This can serve as additional evidence for the positive postsurgical outcomes of the awake procedure (De Witt Hamer et al., 2012).
4.5 Influence of participant characteristics
Besides the main factors discussed in the previous sections, the demographic characteristics of all participants were included in the analysis. The demographic factors age and education level contributed significantly to the outcomes of the RT measures of the SJT, the BNT, and the TMT. The effect of age and education level on these tests has been found in earlier studies (De Witte et al., 2015; Snitz et al., 2009; Tombaugh, 2004). The way in which these factors influenced the results differed slightly per test. For the SJT, TMT-B and TMT-BA, there is a significant interaction between age and education, indicating that the effect of education is larger in older people than in younger people for these tests. Interestingly, there were three-way interactions between age, education, and group in the results of the TMT-B, TMT-BA, and the BNT. Closer inspection of the results revealed that the
