Abbreviations 1 INTRODUCTION

Medulloblastomas (MB) are malignant embryonal neuroepithelial brain tumors located in the posterior fossa. They are common in children and have a prevalence of 9.2% in pediatric brain tumors between the ages 0 and 14 years, with the highest rates between 3–4 and 8–10 years of age.1

Treatment for MB consists of surgery, followed by chemotherapy and radiation therapy.2 Different types and doses of radiation are used, depending on risk classification. The overall 5-year event-free survival rate for patients with MB is 70%.2 Five-year event-free survival in average-risk patients (>3 years of age, <1.5 cm2 of tumor present after surgery and no metastatic disease) is up to 85%.3 In contrast, children younger than 3 years have an unacceptable risk of severe neurocognitive impairment after radiotherapy and were previously treated without upfront radiotherapy, resulting in a poor 5-year event-free survival.4-6 Long-term cognitive and academic performance may be impaired, especially for those who are younger at diagnosis and receive higher doses of radiation.6 Reduction of craniospinal radiation dose and efforts to reduce the boost volume could not prevent this effect in young children.6 Surgery aims for maximum tumor resection but is not without risks, as 7%–50% patients may develop postoperative cerebellar mutism syndrome (pCMS).7-9 According to the consensus between experts associated in the Posterior Fossa Society (http://www.posteriorfossa.org/), typical features of pCMS are delayed onset of mutism or severely reduced speech and emotional lability after cerebellar or fourth ventricle tumor surgery in children.9 Additional common features include hypotonia and oropharyngeal dysfunction/dysphagia.9 It is frequently accompanied by cerebellar motor syndrome and brainstem dysfunction, including long tract signs and cranial neuropathies.9 Because the cerebellum also contributes to cognitive functions such as linguistic processing, executive functioning, visual spatial abilities, and affective modulation, symptoms compatible with the cerebellar cognitive affective syndrome (CCAS) may also be present.10, 11 The mutism or severely reduced speech is always transient, but recovery may be prolonged, speech and language may not return to normal, and other deficits of cognitive, affective, and motor functions often persist.9

At relatively short-term (12 months post diagnosis), lower full-scale intelligence quotient (FSIQ) and processing speed index (PSI) were found in 22 pCMS patients compared to a matched control group.12 In a study of 107 patients with pCMS 1 year post diagnosis, intellectual disabilities were found in 59% of patients with severe pCMS (defined as mutism or severely reduced speech lasting >4 weeks) compared to 17% of patients with moderate pCMS (defined as mutism or severely reduced speech between 1 and 4 weeks).13 Compared to children with no pCMS after resection of MB, children with pCMS also exhibited significantly more obsessive-compulsive types of behavior, withdrawal behavior, social problems, and internalizing problems 1–2 years after completion of therapy.14

Long-term intellectual outcome studies show contradictory results. At a mean of 4.5 years after surgery, a nonsignificant lower intelligence of 16 points was present in five MB patients with pCMS compared to 12 MB patients without pCMS.15 In another study, nine children with pCMS had a significantly lower intelligence compared to 67 children without pCMS 2 years after MB diagnosis.16 In a study of 165 children 3–5 years after diagnosis of MB, a significant decline over time of intellectual and academic performances was observed.17 However, the performance of the majority of patients remained within the average range at 5 years after diagnosis.17 Also, serious hearing loss and pCMS independently predicted below-average estimated mean intellectual ability at 5 years post diagnosis, and patients with high-risk MB and young age (<7 years) at diagnosis exhibited the biggest drop.17 In another study, 36 patients with pCMS and 36 age-matched controls were compared at 1, 3, and 5 years after diagnosis of MB.18 The pCMS group exhibited estimated mean scores that were at least one standard deviation (SD) below the mean for FSIQ, PSI, broad attention, and working memory across all time points and for 5 years after MB diagnosis.18 Attention and working memory declined over time.18 In another long-term follow up study of 121 patients at a mean of 6.1 years after MB diagnosis, 34 identified pCMS patients had a lower total intellectual functioning compared to patients without pCMS.2 In 58 adult survivors of childhood MB, FSIQ assessed at a mean of 6.6 years after the end of treatment ranged from 46 to 131.19 Lower socio-economic status (SES) and presence of pCMS, but not a hearing deficit, had significant negative effect on overall IQ.19 Of 13 out of 58 patients with pCMS, only one had an overall IQ >80 (most had IQs in well-below average range).19 The living situations of the patients with pCMS reflected these negative effects on IQ at a mean of 14.9 years after end of treatment. For example, 10 out of 58 were not fully independent for some activities of daily living and had attended specialized schools and 13 were unemployed.19 These findings contrast with those in a recent long-term follow-up study in patients after MB treatment in which neuropsychological outcome of 18 children with pCMS was compared to matched controls 3 years after diagnosis.20 In the pCMS group, expressive vocabulary and fine motor speed were diminished, but no group differences were found in overall intelligence, receptive vocabulary, visuomotor integration, inhibition, emotional control, depression, and anxiety.20

At present, two prediction tools have been published to predict the occurrence of pCMS, both of which still have to be validated.8, 21 No prediction tool has yet been published for duration of the mute or severely reduced speech phase of pCMS. The findings of Robertson et al. suggest that mutism or severely reduced speech of >4 weeks duration has a negative effect on intellectual function.13 If severe pCMS or long duration of pCMS are negative prognostic factors for intellectual deficits, then perioperative prognostic factors that can be influenced may become critical for predicting intellectual impairment. In both studies in which a prediction tool for development of pCMS was described, damage to the superior cerebellar peduncles (SCP) during surgery was found to be such a contributing risk factor.8, 21 A 0.5°C higher mean body temperature (BT) in the first 4 postoperative days increased the odds ratio for the development of pCMS almost five-fold.22 The latter risk factor was found to be a major contributor in the Rotterdam prediction tool for pCMS.8 In some studies, children with cerebellar damage are particularly at risk for long-term neuropsychological dysfunction, thus requiring active rehabilitation measures.15, 16

The purpose of the present study is to describe the impact of pCMS on intellectual functioning and identify relevant pre- and postoperative risk factors for lower cognitive scores. To account for late effects of radiotherapy, we assessed FSIQ at a mean of 3 years post diagnosis in MB patients with and without pCMS.6

2 MATERIALS AND METHODS

This is a single-center retrospective cohort study in Erasmus MC/Sophia Children's Hospital of patients with MB who were treated between 1992 and 2017 with surgery, chemo- and adjuvant craniospinal radiotherapy and who completed a neuropsychological assessment at least 2 years post diagnosis. Neuropsychological evaluation followed the protocol of the Childhood Oncology Group in the Netherlands. Children treated for a MB complete neuropsychological evaluation 6 months after completion of chemotherapy, as well as 3 and 5 years after tumor resection. Intelligence is included in this protocol and is assessed with the age-appropriated Dutch form of the Wechsler Intelligence Scales (FSIQ, total verbal intelligence [VIQ], total performance intelligence [PIQ], PSI). Data were acquired from the standardized clinical, radiological, and neuropsychological follow-up of brain tumor patients in our multidisciplinary clinical and outpatient setting as documented in the electronic patient record. In order to make a reliable diagnosis of pCMS, children were not included when age at surgery was <2 years. pCMS assessment included a daily postoperative neurological examination and assessment of speech, language, and behavior in the first 2 weeks after surgery.23, 24 After diagnosis of pCMS, patients were subsequently assessed every second week until the recurrence of speech, language, or normalization of behavior to the premorbid situation as reported by the parents.23, 24 Following Robertson et al., pCMS was considered severe when the phase with mutism or extremely reduced speech lasted 4 weeks or more.13 A pediatric neurologist examined all children at regular intervals to establish residual neurological impairments. Neurologic examination 10 days post surgery was scored systematically and focused on the following domains: (a) swallowing problems: absent or severe (needs tube feeding); (b) pyramidal symptoms: absent, present unilateral, or present bilateral; (c) axial ataxia: absent or mild (can sit/stand without support) versus severe (cannot sit or stand without support); (d) limb ataxia: absent or mild limb ataxia (no functional impairment) versus severe ataxia (through which functional impairment); and (e) oculomotor symptoms: absent, mild (saccadic eye movements and/or first-degree horizontal nystagmus), or severe (second- or third-degree nystagmus). Brain T1, T2, FLAIR magnetic resonance imaging (MRI) images at diagnosis with and without enhancement were reviewed for preoperative size and site of the tumor and brainstem invasion.8 We computed the bicaudate index (BI) as a measure of ventricular dilation.8 Ventricular dilation was distinguished as none (BI < 0.19), mild (BI = 0.19–0.26), or severe (BI > 0.26).8 Follow-up MRI at the time (±3 months) of neuropsychological assessments was used to determine the ventricle width and vermis incision.

2.1 Ethical considerations

According to Dutch law, no approval of a Medical Ethical Committee is needed when patient data are studied that are obtained as part of routine patient care. As is required by Dutch law, parents and patients >12 years old signed a consent form before neuropsychological assessment, in which they allowed the data to be used for research.

2.2 Statistical analysis

The data of the neuropsychological assessment were compared with the normative data for the Dutch population and corrected for age. FSIQ, VIQ, PIQ, and PSI were compared to the mean IQ scores in the normal population using a one-sample t-test. IQ scores were compared between the groups with and without pCMS using an independent sample t-test. Before computing the independent sample t-test, assumptions of normality of the dependent variables and homogeneity of variance had to be met.25 A post hoc paired sample t-test was computed to check differences between VIQ and PIQ within the two groups.

Included risk factors for a lower intelligence in patients with pCMS were brainstem invasion, vermis incision, pre- and postoperative hydrocephalus, tumor size, severity and duration of pCMS, mean BT on days 1–4 post surgery, and age at resection.8 These risk factors were analyzed with a two-sided bivariate Kendall's Tau correlation and before computing this analysis, assumptions of normality of the dependent variables had to be met.25 Missing data were deleted pairwise because data were missing randomly. A multiple regression analysis based on backward elimination procedure was conducted. Before computing this regression, assumptions of linearity, multicollinearity, homoscedasticity, normally distributed errors, and independence of the error were checked.25 To analyze data, Statistical Package of Social Science (SPSS) version 25 was used and results were considered statistically significant at a p-value of <.05.

3 RESULTS

Between 1992 and 2017, a consecutive series of 118 children were diagnosed with MB in Erasmus MC/Sophia Children's Hospital. Eighty-seven patients did not meet the inclusion criteria. Forty-nine children died within 2 years after diagnosis. Thirty-one of them were 3 years or younger at surgery and did not receive upfront radiotherapy. Of the 18 children who were older and received upfront radiotherapy, eight were diagnosed with pCMS. The interval diagnosis assessment was too short in 16 patients. The diagnosis of pCMS was not reliably possible because of age <2 years at surgery in six surviving patients. Five children did not receive adjuvant radiotherapy because of young age. Three children were lost to follow-up. Two patients could not perform the Wechsler intelligence test because of severe visual or auditory deficit. All children had routine follow-up and treatment in a pediatric rehabilitation center and had annual neurologic evaluation at our multidisciplinary neuro-oncological follow-up outpatient clinic. In six children, the neuropsychological evaluation was performed in their rehabilitation center because of a shorter travel distance.

3.1 Patient characteristics

Demographic characteristics of the 31 patients are listed in Table 1. Fourteen patients developed pCMS with a duration of 4–212 days (mean [M] = 45, SD = 45 days). Six children had a duration of pCMS symptoms of >4 weeks (duration of pCMS 32, 45, 35, 70, 124, and 212 days), and at least one other symptom classified as severe by Robertson et al.13 One of these children classified as pCMS because of a severely reduced speech and severe irritability. Three children were classified as mild pCMS (duration of pCMS 4, 5, and 9 days) and five children as moderate pCMS (duration of pCMS 12, 14, 16, 20, and 21 days) (Table 2, 3). None of the children classified as mild or moderate pCMS experienced reduced speech. The pCMS and no-pCMS groups differed in mean BT within the first 4 days after surgery, but not in age, tumor size, tumor location, hydrocephalus, chemotherapy protocol, radiation dose, or premorbid disorder.

TABLE 1. Clinical characteristics pCMSMean (SD) No pCMSMean (SD) p-Value Patients (number) 14 17 Sex (male) 64% 65% .72 Deceased 21% 5% .24 Age at resection (years, months) 8,4 (3,7) 8,3 (3,4) .93

Age at assessment (years, months)

Interval resection assessment (years, months)

11,9 (3,7)

3,9 (2,1)

11,7 (3,9)

3,9 (3,0)

.85

.28

Tumor size diameter in cm 4.9 (0.7) 4.6 (0.7) .20 Tumor location Brainstem 71% 60% .51 Vermis 93% 85% .50 Hydrocephalus Preoperative 86% 75% .46 Postoperative 57% 45% .59 Mean temp 1–4 days after resection in °C 37.8 (0.5) 37.3 (0.5) .005 Treatment protocol .11 Standard risk 71% 45% High risk 7% 10% Other 21% 45% Radiation dose posterior fossa in Gy (mean, SD)

48.5 (13.9)

Range 24–68

53.2 (7.2)

Range 31–68

.28 Radiation dose craniospinal in Gy (mean, SD)

30.7 (6.4)

Range 23–36

27.4 (6.5)

Range 18–40

.25 Disorder presurgery 0% 15% .89 Duration mutism or severely reduced speech in pCMS days (mean) 4–212 (45) Neurologic examination on day 10 after surgery Axial ataxia severe 50% 0% Limb ataxia severe 50% 29% Swallowing problems severe 57% 0% Pyramidal symptoms 21% 35% Oculomotor symptoms severe 43% 12% Abbreviations: °C, degrees Celsius; cm, centimeter; Gy, Gray; pCMS, postoperative cerebellar mutism syndrome; SD, standard deviation. TABLE 2. Characteristics of postoperative cerebellar mutism syndrome (pCMS) symptoms Patient diagnosed with pCMS Gender (M/F) Age at surgery (years,months) Days after surgery before pCMS onset pCMS characteristics Duration pCMS (days) Ataxia (days) Hypotonia (days) Irritability (days) Severity score following Robertson et al.13 1 M 6,11 1 Mutism 5 4 2 3 Mild 2 F 7,11 0 Mutism 20 25 20 15 Moderate 3 M 8,7 0 Mutism 4 10 0 3 Mild 4 F 4,8 0 Mutism 21 27 20 18 Moderate 5 F 8,8 0 Mutism 32 60 45 25 Severe 6 M 17,7 1 Mutism 70 120 120 60 Severe 7 M 6,6 0 Mutism 124 35 25 10 Severe 8 M 12 3 Severely reduced speech 212 100 84 154 Severe 9 M 4,8 2 Mutism 14 16 0 0 Moderate 10 M 11,4 1 Mutism 16 14 14 5 Moderate 11 M 5,3 0 Mutism 45 70 50 40 Severe 12 M 4,8 0 Mutism 35 60 50 30 Severe 13 M 11,7 1 Mutism 12 20 18 0 Moderate 14 F 7,9 0 Mutism 9 5 4 5 Mild TABLE 3. Significant independently related predictors for lower FSIQ, VIQ, and PSI scores in total group Intelligence R2 p-Value Predictors Standardized β coefficients FSIQ .35 .01 Age at resection .35 Mean body temperature −.31 Size of tumor −.22 VIQ .31 .04 Age at resection .34 Size of tumor −.25 Mean body temperature −.17 PSI .31 .04 Preoperative hydrocephalus .37 Mean body temperature −.22 Size of tumor −.21 Abbreviations: FSIQ, full-scale intelligence; PSI, processing speed index; VIQ, total verbal intelligence. 3.2 Neuropsychological data

Mean age at assessment was 11 years and 11 months and mean follow-up period was 3 years and 9 months in the pCMS group and 11 years and 7 months and 4 years and 11 months, respectively, in the non-pCMS group. There were no significant differences between the pCMS and non-pCMS groups for age at resection, age at assessment, time interval resection and assessment, gender, and number of patients deceased at the moment of writing.

In comparison with the normal population, the whole MB group showed weaker performances on FSIQ, VIQ, PIQ, and PSI (FSIQ t = −6.742, p < .001; VIQ t = −4.433, p < .0001; PIQ t = −8.459, p < .0001; PSI t = −10.055, p < .0001). No significant differences were found for FSIQ (pCMS group mean 77 ± 16; non-pCMS group mean 83 ± 16), VIQ (pCMS group mean 80 ± 1; non-pCMS group mean 89 ± 17), PIQ (pCMS group mean 72 ± 15; non-pCMS group mean 76 ± 16), and PSI (pCMS group mean 71 ± 15; non-pCMS group mean 74 ± 12) between the pCMS and non-pCMS groups. A relevant difference was found on VIQ (89 in non-pCMS group vs. 80 in pCMS group). In the Netherlands, this difference in IQ score is of clinical relevance, at a cutoff score of <85 for getting extra help at school. This 1 SD below the standardized mean score may also be used in other countries/states/provinces as a clinically significant cutoff. In the non-pCMS group, only 58% of patients had a VIQ between the range of 85 and 115 (1 SD below and above 100) versus 27% of pCMS (chi-square p = .105). All intelligence scores are shown in Figure 1.

IQ scores pCMS versus non-pCMS groups. FSIQ, full-scale intelligence; IQ, intelligence; pCMS, postoperative cerebellar mutism syndrome; PSI, processing speed index; TPIQ, total performance intelligence; TVIQ, total verbal intelligence

3.3 Risk factors for intelligence in pCMS

In the whole group, lower FSIQ correlated with lower age at resection (r = .39, p = .03), lower age at assessment (r = .42, p = .02), and mean raised postoperative temperature (r = −.41, p = .03). A correlation was found between poorer VIQ and larger tumor size (r = −.45, p < .03). Poorer PSI was also correlated with mean raised BT (r = −.42, p = .04) and preoperative hydrocephalus (r = .48, p = .02).

In the pCMS group, we found correlations between lower FSIQ and greater tumor size (r = −.59, p = .04) and between lower VIQ and greater tumor size (r = −.77, p < .01) and higher temperature (r = −.67, p = .02). We did not find a correlation between duration of pCMS and intelligence scores. However, a mutism duration of more than 15 days in children with pCMS significantly correlated with the presence of severe swallowing problems (r = .41, p < .01) and severe axial ataxia (r = .56, p < .01) 10 days after surgery. Swallowing problems and bilateral pyramidal symptoms were also found to be risk factors for long-term VIQ deficits.

Multiple regression analysis (Table 3) showed significant independently related predictors for lower FSIQ, VIQ, and PSI scores: younger age at resection, higher mean BT 1–4 days after surgery, larger tumor, and presence of preoperative hydrocephalus. There were no significant predictors for PIQ.

4 DISCUSSION

In the present long-term outcome study, MB survivors had significantly lower performance compared to their healthy peers on measures of IQ and processing speed, as can be expected after craniospinal radiation.6 In agreement with the studies of Wells et al.15 and Grieco et al.,20 we found that FSIQ scores in MB patients with pCMS did not significantly differ from those without pCMS. However, a clinically relevant difference was found on VIQ.

There is a general agreement that interruption of the dentato-thalamo-cortical (DTC) pathway is