Subjects

Thirty-three individuals participated in this study: twenty-one of them were clinical patients and twelve were controls. The astrocytoma group (CE, without treatments) comprised thirteen individuals (ten girls and three boys), from them five had vermal tumor (11.36 ± 2.78 years) and eight had hemispheric location of the tumor (11.59 ± 3.77 years). The mean group age at evaluation was 11.50 ± 3.3 years old (range 6.66–18). The mean age at surgery was for the global CE group 8.24 ± 3.95 years (range 1.91–14.58), and, specifically, vermis CE: 8.39 ± 3.19 years and hemispheric CE: 8.15 ± 3.43 years. The mean time surgery-evaluation was: CE group: 3.25 ± 2.74 years (range 0.41–10), vermis CE: 2.97 ± 1.03 and hemispheric CE: 3.43 ± 3.49 years.

The medulloblastoma group (CE+, with treatments) comprised eight participants (three girls and five boys), but the performance of one patient was eliminated from the CE+ clinical sample. This patient suffered two surgeries due to tumor recidiva and presented a high resected volume (tumor + margin = 134.8 cc). Therefore the CE+ clinical sample was n = 7. However, it must be taken into account that the neuropsychological profile of the CE+ children, considering all 8 patients, would be worse than the one presented in this report. The mean age at evaluation was 13.59 ± 1.62 years old (range 12–17). Five children had vermis located tumor (14.11 ± 1.64 years) and two children hemispheric (12.29 ± 0.41 years). The mean age at surgery was for the global CE+ group: 7.12 ± 2.06 (range 4–9.83) and vermis CE+ group: 7.21 ± 1.44 years. The mean time surgery-evaluation was: global CE+ group: 6.47 ± 2.77 (range 2.75–10.92) and vermis CE+ group: 6.9 ± 2.69.

Variable resected volume. The resected tumor volume (tumor + margin) was computed following the reduced equation of ellipsoid (a × b × c/2), being a, b and c the diameters of the lesion as measured in MRI post surgery using the Volume 1.0 software (designed by University of Seville). The mean resected volume in CE group was: 49.23 ± 26.97 cc (range 7.5–122.1 cc), vermis CE: 52.97 ± 15.67 cc (range 31.4–67 cc), hemispheric CE: 46.89 ± 33.02 cc (range 7.5–122.1 cc). The mean resected volume in CE+ group was: 35.75 ± 15.73 cc (range 18–48) and vermis CE+ group 29.63 ± 16.44 cc. The data of hemispheric CE+ group (n = 2) are not presented.

Variable dentate-affected. The surgeon (after analysis of pre- and post-surgery MRI) categorized 6 patients as dentate affected and 7 as dentate unaffected in CE group.

In the dentate-affected subgroup there was a secondary lesion (uni or bilateral, without nucleus exeresis) due to the nucleus compression exerted by the tumor and/or the surgical operation. This circumstance appeared often related to medial and high volume tumors. In CE group the dentate-affected children were: 4 with hemisphere and 2 with vermis location.

The control group (C) comprised twelve subjects (seven girls and five boys); the mean age at evaluation was 11.33 ± 1.82 years old (range 9–14). There were no significant group differences with regard to age, sex ratio or scholarization years. Given the broad age range, the direct test scores (the values obtained by each participant) were corrected by age following normative data. Some of the clinical variables have a high variability, but this is characteristic of our patients sample collected in a two years period from the Virgen del Rocio Hospital in Seville (Spain).

The inclusion criteria for the clinical groups were: (i) normal social and emotional behavior before the diagnosis, (ii) normal academic performance before diagnosis and (iii) motor and visual performance not excessively deteriorated after surgery. The latter aspects were assessed and have been previously described in Quintero et al. [67].

The present experimental protocol followed the Helsinki Declaration norms regarding experiments involving human participants and, in addition, was approved by the ethic committees of the Virgen del Rocio Hospital and the University of Seville.

Neuropsychological instruments

We assessed some of the components referred as part of executive functions, more specifically, planning, abstraction and conceptual capacity, mental flexibility, phonologic and semantic associative verbal fluency, working memory, selective attention, response conflict, response inhibition, visual-spatial organization and construction. To fulfill this objective different highly sensitive tests were applied [55]. All tests were applied following standardized criteria. Direct scores corrected by age were transformed in T scores (normal mean = 50 ± 10) for the WCST [72], Stroop [73] and Raven-IQ [74]. Direct scores of Rey-copy test [75] were transformed in Z-scores corrected by age following Taylor scoring criteria [76]: (each subject score – mean normative data)/SD normative data. Direct scores of Digit span (WISC-R verbal scale subtests [77]) were transformed in corrected by age scaled scores (normal mean = 10 ± 3). Direct scores of Verbal Fluency test (initial letter and animal categories) were transformed in corrected by age scaled scores following the Neuropsi battery scoring criteria (Spanish-speakers children) [78].

The normal/abnormal deficit criteria considered for T and Z scores was 1.5 SD below the normative mean (severe deficit ≤ 1.5 SD, moderate to slight impairment between -1.4 SD and -1 SD). The deficit criteria considered in scaled scores (SS) was 1 SD below the normative mean (10 ± 3) because this cut-off score (7) maximize sensitivity and specificity in discriminating between normal and brain-damaged participants [72] (severe deficit ≤ 1 SD, moderate to slight impairment between -1 SD and -0.66 SD).

Statistical analysis

We have checked that normality assumption (Kolmogorov-Smirnof with Lilliefors significance correction) was not observed in all the variables of the tests and/or different groups, so the results were analyzed using non-parametric Mann-Whitney U tests (p < 0.05) in order to compare the control group's mean performance with clinical groups. An independent comparison was performed for each clinical group and assessed measurements due to differences in the kind (cellular bases) of tumor and post-surgical treatments.

Possible influences related to the tumor's location (hemisphere or vermis) and to the affectation (or not) of the dentate nucleus on the neuropsychological performance were assessed by means of statistical analysis on the astrocytoma group (surgery without additional treatments). Mann-Whitney U test were computed to compare the performance of vermis CE group (n = 5) vs. C and hemispheric CE group (n = 8) vs. C. These comparisons would allow defining if the vermis or hemispheric located tumor was more disruptive for executive functioning. The low sample in medulloblastoma only allowed to analyze the vermis CE+ group (n = 5) but not the hemispheric CE+ group (n = 2). Additional non-parametric tests based on the Mann-Whitney U test were computed to compare dentate-affected CE group (n = 6) vs. C, dentate-unaffected CE group (n = 7) vs. C and affected vs. unaffected.

Finally, non-parametric Spearman's correlations (p < 0.05) were computed for each clinical group in order to assess the statistical dependence among the evaluated variables and the children's age at surgery, time between surgery-evaluation and resected volume.

Stroop test

Attention and inhibition are important for reducing the interference of irrelevant information on the working memory, but also for suppressing previously relevant information that becomes irrelevant during the task. The impairment on inhibition also can be particularly important for decision-making in social and emotional behaviors indicating the lack of self-regulation. Frontal patients and other groups with frontal-subcortical pathways impairments present a poor execution in the Stroop tasks. Patients with right superior medial-frontal lesions (8B, 9, and superior portion of 32 BA) are the most impaired (errors and slowness) for the incongruent Stroop condition [68].

In our sample, the CE+ group compared with norms presented a clear deficit in accomplishing the Stroop tasks indicating a selective attentional deficit [19, 30], poor mental flexibility and a lack of inhibitory control on interfering stimuli [83].

The CE group (vermis and hemispheric patients) compared with norms showed scores about -1 SD below normal mean for the word and colour conditions but not for the incongruent condition (-0.6 SD), suggesting that the impairment was caused by speed processing slowness. In a posterior fossa tumors patients study [84] moderate impairments in the three Stroop conditions (W = -1.25, C = -1.37 and I = -1.19) were obtained in patients evaluated at 7.5 years from surgery (surgery at 8.25 years old), but they analyzed together diverse tumor patients (most of them astrocytoma). If we collapsed the CE and CE+ groups the results would be similar to those previously described [84]. Our results are also coherent with another study [85] where it is compared the performance of young adults (23 years old) resected of astrocytoma (T ≤ 40) with a group of medulloblastoma patients with chemo- and radiotherapy (T ≤ 33).

Several functional neuroimaging studies have demonstrated the activation of the superior medial prefrontal cortex [86, 87], the anterior cingulate cortex (ACC) [41, 83, 88, 89] and the cerebellum [39, 90, 91] during the performance of the Stroop incongruent condition. A dysfunction in adults with cerebellar damage in the interference was shown using PET neuroimage techniques by Fiez et al. [39]. The most characteristic medial prefrontal dysfunction is the impairment of inhibitory cognitive control (mainly superior medial PFC), but other behavioral disturbances as lack of self-control and affective deficits (behavioral and emotional inhibitory control) are also present (mainly inferior medial PFC). Anterior cingulate cortex (ACC) patients present attentional deficit, distractibility, lack of initiative, apathy and flattened affect. The ACC cortex is a wide region classically related to emotions that has been cytoarchitectonically divided in different regions with sub-regions [92]. The cerebellum connectivity with these frontal regions via thalamus (parafascicular, centromedian intralaminar, dorsomedial, anterior ventral and lateral ventral nuclei) [93] would explain the Stroop deficit in cerebellar patients. In accordance with the triple attentional system described by Posner [94], the ACC is active when a conflict is present between simultaneous processes [95] as occurs in the Stroop incongruent condition, but this is not its unique function, the ACC also coordinates multiple attentional subsystems and plays a crucial role in other cognitively demanding tasks [89].

Verbal fluency

Phonologic and semantic fluency are measurements of associative verbal fluency in restricted situations. Both verbal fluencies require the generation of words based in categories (phonological or semantical) and both require the language executive function in order to establish a search strategy, search of information, retrieval from stored information in long-term memory and active maintenance in the working memory, as well as self-regulation for an efficient performance. The frontal cortex participates in the executive control of processes as initiation, search and active evocation of phonologic and semantic representations. These last are more distributed and therefore should be more affected than phonologic fluency if executive functions are disturbed. The frontal cortex must also monitor intrusions and perseverations. Subjects with prefrontal lesions have a preserved language but show an important disconnection between knowledge and execution, and are impaired in both phonologic and semantic fluency. Lesions studies have shown that patients with left dorsolateral and striatal lesions were the most impaired in phonologic fluency performance (right dorsolateral or inferior medial frontal lesions patients were not significantly affected). Bilateral dorsolateral and (superior and inferior) medial frontal cortices are needed for an adequate performance on semantic fluency task (Stuss et al. [69] for review). Another unilateral prefrontal lesions study [96] also showed that phonological fluency is mediated by the left dorsolateral PFC. In contrast, the semantic fluency involves bilateral dorsolateral PFC and the right ventromedial frontal areas. Several functional neuroimage studies [38, 90, 97–99] have found cerebellar activation in verbal fluency tasks. In addition, it has been obtained an association between the cognitive demand in the semantic task and bilateral cerebellum activation [100]. Ravnkilde et al. [90], by means of PET in normal subjects, showed that during phonological fluency task left dorsolateral prefrontal cortex, left and right inferior frontal cortex, left supplementary motor cortex, left and middle anterior cingulate gyrus, left orbitofrontal cortex, right posterior temporal lobe and cerebellum (mainly in the right hemisphere) were activated.

Our cerebellar patients showed differential verbal fluency impairment. The medulloblastoma group scaled scores (SS) compared with norms (deficit -1 SD, cut-off 7) presented severe deficit (-1.81 SD) in semantic fluency and moderate to slight deficit in phonologic fluency (-0.66 SD, SS = 8). These scores are more severe in vermis CE+ patients (SF: -1.87 SD and PF: -1.13 SD). The astrocytoma group showed different profile in vermal and hemispheric patients. Vermis CE patients showed deficit in semantic fluency (-1.2 SD) and moderate impairment (-0.67 SD) in phonologic fluency. The hemispheric CE dentate-affected patients showed deficit in semantic fluency (-1 SD) but not the unaffected patients, and both had normal scores in phonologic fluency. The absence of this deficit in dentate-unaffected hemispheric CE patients, rules out the possibility that the observed deficit in vermal CE patients would be due to motor-speech impairment. Given that CE patients did not receive chemo and/or radiotherapy, the semantic fluency deficit could be due to the dysfunctional crossed cerebellum-cortical diaschisis and suggests a greater impairment in vermal CE patients and in all dentate-affected CE patients. Verbal fluency deficits have also been reported in other cerebellar patients [25, 46, 47, 51]. For the CE+ group it could be argued that the deficit in verbal fluency is due to the radiation and/or chemotherapy. However, Riva and Giorgi [101] assessed a resected hemisphere-astrocytoma group and a vermal medulloblastoma group before receiving any additional treatment (5–6 weeks after surgery) and they reported deficit in verbal fluency in both groups.

Digits span (WISC-R verbal scale subtests)

The first Digit span task allows testing the number of items maintained in the short-term-memory (STM). Therefore, it allows giving an estimation of the limited retention capacity in the STM and its temporal dimension (digit forward: "say the same numbers as rapid as you can"). The second task (working memory) implies an increased selective attention and executive control for the planning, allocation of attentional resources and the necessary attentional maintenance in order to interfere with the previous retention memory process when coordinates/synchronizes a verbal information (digit backward: "say the same numbers in inverse order as rapid as you can"). This subtle working memory task requires executive control, initial phonological encoding, sub-vocal rehearsal of items that are actively remembered, but also attentional focus and shifting between previous items and the on-line task (mainly parietal posterior contribution).

The multiple-component-model [102] considers two separate systems: the phonological loop for verbal information and the visuospatial sketchpad for non-verbal information, and a separate attentional controller or central executive. The verbal system is subdivided in a passive phonological store that retains some time (decay) verbal information through phonological representations, and an active rehearsal articulatory process that maintains by reactivation these representations.

Some functional neuroimage studies using specific neuropsychological tasks to test the Baddeley's model and other models showed differences between recognition (parietal BA 7 and temporal-parietal, supported by converging psychophysiological evidence) and recall (frontal). [103]. Lesional studies showed that posterior lesions, especially perisylvian (left ventral supramarginal gyrus) brain-damage patients are the most impaired in verbal working memory, while Broca's area patients showed moderate deficit in verbal span and severe deficit in overt or covert rehearsal working memory process, linked to speech deficit by concurrent articulation effect [103]. Probably, the language comprehension, symbolic abstraction and conceptualization capacities are necessary to more demanding tasks (i.e. "in inverse order").

Functional neuroimage studies [24, 39, 43, 98, 104, 105] provide anatomical working memory process support in bilateral dorsolateral PFC (BA 9, 46), ventrolateral PFC (Broca: 44/45), medial frontal pre-SMA (BA 6) and ACC (BA 32) and superior posterior parietal cortices (BA 7), all related to updating information and maintaining sequential order, and bilateral inferior parietal (supramarginal gyrus BA 40) related to specific representations. Paulesu et al. [105] located the phonological store in the left supramarginal gyrus and the articulatory control process in Broca's area. Lateral cerebellar areas are also active during verbal working memory tasks [39, 98, 99]. Desmond and Fiez [98], by means of f-MRI, during a verbal working memory task showed that HVI and superior HVIIA cerebellar activations represent input from the articulatory control system of working memory from the frontal lobes, and that HVIIB activation is derived from the phonological store in temporal and parietal regions.

In our report, both clinical groups showed deficit in language working memory as measured by the WISC-R digit span. Medulloblastoma group showed deficit (-1.53 SD), most increased in vermal CE+ patients (-1.6 SD). Astrocytoma group children did not present evident dysartrhia problems (most related to overt or covert articulatory rehearsal). Vermal CE patients showed the most severe deficit (-1.86 SD below normal mean: 10 ± 3) in verbal working memory. These patients also showed the already described deficit in semantic fluency and moderate to slight problems in phonologic fluency, reading, naming, conceptual responses and non-verbal Raven-IQ, but the interference was less compromised (-0.6 SD), suggesting that this severe deficit in verbal working memory capacity would interfere with the correct performance in other language functions.

All hemispheric CE patients showed deficit (-1.29 SD), but it is more severe in dentate affected patients (-1.58 SD) – who also presented semantic fluency deficit – than in dentate unaffected patients (-1 SD). Digit span deficit in cerebellar tumors patients also has been reported [84, 106].

The astrocytoma group impairment suggests that preservation of reciprocal cerebellar-dorsolateral PFC pathways is necessary for correct verbal executive functioning, although the specific systemic participation of the cerebellum in these high level functions is not well known.

Rey complex figure

The obtained results for the Rey-Osterrieth complex figure copy showed that only the medulloblastoma group presented deficit with respect to norms for the exactness of the copy (Z = -1.62). This result implies difficulties in the executive planning and organization of the copy and visual-spatial perception and construction. The copy task is very difficult for prefrontal patients due to the impairment in temporal sequential planning [26–29, 107], and CE+ group showed other frontal-like deficit. Other studies in children with posterior fossa tumors also found this deficit [84] showing a similar Z-score = -1.9 and suggest that the impairment is more associated to a deficit in executive functioning than to visuo-spatial problems. Patients in representative case studies [106] showed deficit too. Our results are also in agreement with those of [85], who compare the performance of medulloblastoma vs astrocytoma groups in the Rey figure copy and showed that medulloblastoma group presented a statistically significant impairment with respect to the astrocytoma group.

Wisconsin card sorting test

The medulloblastoma group compared with norms showed deficit in perseverative responses (-1.56 SD below normal mean), perseverative errors (-1.5 SD) and moderate impairment in errors and conceptual responses about -1 SD below normal mean. Vermal CE+ patients presented a more severe deficit (perseverations: -1.78 SD, perseverative errors: -1.76 SD, errors: -1.18 SD and conceptual responses: -1.16 SD below normal mean). These results indicate difficulties with planning, cognitive flexibility, set shifting and integration of cognitive responses which are assessed by the WCST. Since frontal patients show impairment of these capacities [52, 70, 71], our results in CE+ group indicate a frontal-like deficit [80]. The increased number of perseverative errors and perseverations indicate deficit in the shifting strategy and the increased errors a moderate impairment in maintaining the strategy too.

In vermal-medulloblastoma subgroups at 5–6 weeks after surgery, without treatments, Riva and Giorgi [101] showed the following Z-scores for perseverations-WCST (speech-problem group = 0.5, language-problem group = 2.8, behavioral alteration group = 1.3 and one patient autistic-like = 2.4). When averaging as a global medulloblastoma group they obtained Z = 1.75, which is the same deficit level (-1.78 SD below T-normal mean) we obtained at 6.47 years from surgery and treatments.

In the astrocytoma group (at 3.25 years from surgery), dentate-affected compared with unaffected patients showed an impairment about -1 SD in perseverative errors, perseverations and errors-WCST (table 4).

The WCST is particularly sensitive to different dysfunctions in dorsolateral (left and right) and superior medial frontal cortex patients in lesion studies [70], and f-MRI studies showed the dorsolateral PFC particularly involved while solving the WCST [108, 109]. The CE+ group deficit in this task must be due to cerebellar-dorsolateral connections damage caused by surgery (as in Riva and Giorgi [101]) and increased by post-surgery treatments, given that we also obtained deficit in perseverative error. Dentate-affected CE group showed moderate executive deficit too although they did not received treatments.

Raven test

The obtained mean-score in the CE+ group was low-normal (85.85 ± 14, -1 SD below normal mean). With respect to the lack of impairment in the global CE group it must be remarked that the non verbal Raven-IQ was -1 SD in vermal CE patients, while the hemispheric group was normal. Radiation dosimetry and hydrocephalia also could be some of the factors predicting a low IQ [110–112]. However our vermis CE group neither receive radiation nor presented hydrocephalia.

General discussion

The neuropsychological evaluation showed a differential prefrontal-like deficit: vermal medulloblastoma patients showed the most severe executive deficit and global medulloblastoma group showed severe deficit of all tested executive functions. All astrocytoma patients showed executive deficit in verbal working memory (more increased in vermal patients) and slowing of speed processing. The vermal astrocytoma patients showed semantic fluency deficit, moderate to slight impairment in phonologic fluency, conceptualization and non-verbal Raven IQ. Hemispheric astrocytoma patients showed the global group deficits, but only dentate-affected patients showed semantic deficit. All dentate-affected (vermal and hemispheric) astrocytoma patients showed semantic fluency deficit and moderate impairment (-1 SD) in perseverative errors, perseverations and errors WCST responses.

The deficit level is clearly related with the kind of tumor (cellular bases), vermal or hemispheric cerebellar tumor's location and dentate nucleus' affectation by compression without exeresis nucleus. In the few existing studies the clinical samples were small. Those reports which considered independently the kind of tumors (cellular bases) showed a greater deficit in medulloblastoma group than in astrocytoma group. Another contribution of present report is to include by first time, according to our knowledge, the dentate affectation variable as a source of possible executive and cognitive impairment.

In the astrocytoma group, higher resected-volume is significantly correlated with dentate-affectation and worse scores in verbal fluency (phonologic and semantic) and Rey-copy exactness (table 6). The older children at surgery (age-at-surgery statistically significant correlations) showed better Rey-copy type and required lower number of intents to complete the WCST than younger children. Longer time elapsed between surgery and evaluation (time surgery-evaluation statistically significant correlations) is related to more correct responses-WCST, suggesting that it is possible some improvement in conceptualization capacity as time increases after surgery. Luria [113] termed as "Frontal Secundary Syndrome" the observed deficits in adult cerebellar-tumor patients and indicated the possibility of recovery in some patients without severe sequelae.

The only significant correlation in medulloblastoma group was related to the age at surgery. Older children at surgery showed better performance in semantic fluency than younger children. Semantic maturation occurs lately than the phonologic. Therefore, younger children (range at surgery: 4–10 years old) were most immature for semantic fluency before surgery, and consequently presented difficulties in organizing this function. Some developmental MRI-studies showed that decreased amounts of normal-appearing white matter correlated with verbal thinking, non-verbal thinking, attention and IQ [114, 115] in medulloblastoma children (with treatments) compared with age and gender-matched normally developing children. Neuroectodermal tumors treatments involve conformal radiotherapy and increased radiation-dose including craniospinal irradiation (without minimizing the dose to any brain structures, including the temporal lobe). Liu et al. [116], using a more sensitive technique with high-resolution MRI and an automated cortical reconstruction technique, have found abnormal changes in the grey-matter development of these patients. They compared the normal differential developmental pattern of cortical thicker (frontal and temporal lobes) and thinner (posterior cortex) in children about 10 years. Results showed significant differences in both (cortical thickness and thinner areas) and reflect significant cortical thinning in the treated medulloblastoma children compared with normal developed children. They concluded that cortical areas undergoing development are more sensitive to the radiation-dose effects.

The present report assessed functions indicate a frontal-like syndrome with differential deficits level: very severe in vermal medulloblastoma patients, severe in global medulloblastoma group and moderate to slight in vermal, hemisphere dentate-affected and hemisphere dentate-unaffected astrocytoma patients respectively. These executive deficits cannot be attributed to the chemo- or radiotherapy exclusively, given that also appeared in children who only received surgery, although it is clear that these treatments can generate and increase the reported impairments. Therefore, the obtained executive impairments support the concept of Cerebellar Cognitive Affective Syndrome (CCAS) proposed by Schmahman and Sherman [31] as a framework to characterize the cerebellar patients, but since we only have tested some executive functions (and learning capacities [67]), we cannot add more evidence about affective functions. The results obtained in our study are coherent with the role of prediction and preparation proposed for the cerebellum [117], given that after cerebellar lesion some executive functions remains, although clearly deteriorated. Paraphrasing the notion of thought dysmetria [31, 32, 66], this sub-optimal functioning could be defined as an executive dysmetria.

Limitations of the study

As in similar studies, the principal limitation is the small sample. In studies with clinical participants the presence of non-controlled variables is always a concern. In this sense it has been shown that oncologic patients can suffer depression, anxiety [118] or post-traumatic stress disorder [119]. Therefore, it is very difficult to control all the possible intervening variables and they can be considered as one of the data variability sources.