Long-term neurocognitive and psychological outcomes in meningioma survivors: Individual changes over time and radiation dosimetry

Abstract

Background

This study investigates long-term changes in neurocognitive performance and psychological symptoms in meningioma survivors and associations with radiation dose to circumscribed brain regions.

Methods

We undertook a retrospective study of meningioma survivors who underwent longitudinal clinical neurocognitive assessments. Change in neurocognitive performance or psychological symptoms was assessed using reliable change indices. Radiation dosimetry, if prescribed, was evaluated based on treatment-planning computerized tomography co-registered with contrast-enhanced 3D T1-weighted magnetic resonance imaging. Mixed effects analyses were used to explore whether incidental radiation to brain regions outside the tumor influences neurocognitive and psychological outcomes.

Results

Most (range = 41%–93%) survivors demonstrated stable—albeit often below average—neurocognitive and psychological trajectories, although some also exhibited improvements (range = 0%–31%) or declines (range = 0%–36%) over time. Higher radiation dose to the parietal-occipital region (partial R² = 0.462) and cerebellum (partial R² = 0.276) was independently associated with slower visuomotor processing speed. Higher dose to the hippocampi was associated with increases in depression (partial R² = 0.367) and trait anxiety (partial R² = 0.236).

Conclusions

Meningioma survivors experience neurocognitive deficits and psychological symptoms many years after diagnosis, and a proportion of them decline over time. This study offers proof of concept that incidental radiation to brain regions beyond the tumor site may contribute to these sequelae. Future investigations should include radiation dosimetry when examining risk factors that contribute to the quality of survivorship in this growing population.

Meningioma is the most commonly occurring primary central nervous system tumor.¹ Survival rates are high; 82% at 5 years,² with some deaths unrelated to the tumor.³ Historically, survivors who lived ≥5 years following diagnosis and treatment were considered “cured.”⁴ Yet, recent evidence suggests that many survivors experience long-term neurocognitive and psychological symptoms.^5–7 Little is known about how these symptoms evolve because patients are often discharged after routine post-surgical follow-up.⁸ Most studies that investigate long-term outcomes are conducted at the group level, obscuring individual variability. However, in some survivors symptoms improve, whilst in others symptoms worsen 1 year post-surgery.⁹

Radiation is the standard of care for meningioma that are not safely amenable to surgery, or continue to grow after surgical resection.⁴ Neurocognitive decline occurs months to years after radiation in other brain tumor populations^10,11; whether this occurs in meningioma survivors is largely unknown. In the few studies examining neurocognitive effects of incidental radiation to brain regions beyond the tumor site, dose to hippocampi, frontal cortex, and corpus callosum was associated with neurocognitive outcomes in heterogeneous brain tumor populations including meningioma.^12,13 In our cross-sectional meningioma cohort study, we found that dose to hippocampi, parietal-occipital, and dorsal frontal regions was associated with neurocognitive outcomes including memory and visuomotor processing speed.¹⁴ Those studies did not examine neurocognitive change over time, or control for individual change over time. For this reason, there is limited data to inform risk estimates of neurocognitive outcomes following radiation.

We are unaware of any studies exploring associations between radiation dosimetry and psychological symptoms. Anxiety and depression have been associated with frontal tumors,¹⁵ likely due to ventromedial prefrontal cortex involvement in emotional processing.¹⁶ Radiation to these regions may disrupt frontal-subcortical circuitry, potentially increasing the risk of psychological symptoms.

Identifying risk factors for long-term neurocognitive and psychological symptoms is important to inform appropriate and efficient use of scarce supportive care resources for meningioma survivors.^17,18 The aims of this study are to (1) investigate long-term changes in neurocognitive performance and psychological symptoms after meningioma treatment, and (2) explore associations between radiation dosimetry to circumscribed brain regions and neurocognitive and psychological outcomes, while controlling for individual change over time. We expect variability in individual longitudinal change in these outcomes, with some survivors improving, declining, or remaining stable, over time. We also expect associations between high radiation dose to left hippocampus and a decline in verbal memory,^12,19 high dose to the parietal-occipital region and declines in visuomotor processing speed and executive function,¹⁴ high dose to dorsal frontal lobe and decline in executive function,^12,14,20 and high dose to ventral frontal and subcortical brain regions and increased symptoms of depression and anxiety.¹⁶

Methods and Materials

Participants and Procedure

This was a retrospective study of adults (age ≥ 18 years) with radiologically presumed or confirmed diagnoses of meningioma who underwent more than 1 neurocognitive assessment at the Princess Margaret Cancer Centre in Toronto, Canada for clinical purposes between 2005 and 2020. Reasons for referral included establishing a new neurocognitive baseline, providing recommendations and/or accommodations as appropriate prior to returning to school or work, or addressing neurocognitive complaints. Participants were excluded if they had a history of psychiatric or neurological disorder that prevented them from completing testing or invalid neurocognitive test results. The validity of neurocognitive test results was examined when participants were found to perform below cut-off scores on 2 embedded validity measures: the Reliable Digits Span²¹ and the Trail Making Test Part A.²²

Tumor and treatment-related information were retrieved from medical records. Neurocognitive tests and psychological questionnaires were administered by trained personnel supervised by a licensed neuropsychologist (KE); data were retrieved from clinical files. Radiation dosimetry characteristics for each brain region were delineated from Pinnacle or RayStation planning system software (described subsequently). This study was approved by the University Health Network and University of Toronto research ethics boards (REB #20-5711 and 38523, respectively).

Neurocognitive Assessment

We included tests of attention, executive function, language, memory, and processing speed (Supplementary Table 1) as these domains are most affected in meningioma survivors.^9,23 Neurocognitive test scores were standardized using normative data, accounting for age, sex, and level of education when appropriate, and converted into z-scores (M = 0, SD = 1). Higher scores indicate better performance.

Psychological Symptoms

The 21-item Beck Depression Inventory Second Edition (BDI-II)²⁴ rates depressive symptom severity during the past 2 weeks on a four-point scale. A cut-off point of ≥20 was used to detect clinically significant symptoms.²⁴ The 40-item State-Trait Anxiety Inventory (STAI)²⁵ rates anxiety symptoms on a four-point scale. Raw scores were standardized using normative data and converted into T-scores (M = 50, SD = 10). A cut-off point of T ≥ 65 was used to detect clinically significant anxiety symptoms.

Radiation Dosimetry

Treatment-planning computerized tomography scans were co-registered with contrast-enhanced 3D T1-weighted magnetic resonance imaging on Pinnacle or RayStation planning system software. We segmented the brain into broad regions of interest (ROI’s; cerebellum, dorsal frontal, hippocampus, parietal-occipital, subcortical, temporal, and ventral frontal), manually contoured in the treatment-planning system according to standardized contouring anatomy modules (Supplementary Table 2 and Figure 1). Tumor and surgical beds were excluded from ROI contours (ie, the contoured gross tumor volume); brain tissue surrounding the tumor and surgical beds were included in ROI contours. Contouring was performed by a single observer (AS) and reviewed by a radiation oncologist (DST). Clinical radiation dose plans were used to calculate dose volume histograms for the delineated structures. Dosimetric endpoints of interest were chosen a priori and included mean dose and dose received by 50% (D50) of the brain target volume.

Statistical Analyses

Changes in Neurocognitive Performance and Psychological Symptoms

A reliable change index (RCI) with correction for practice effects²⁶ was calculated to analyze changes in neurocognitive functions and psychological symptoms from the initial assessment to most recent follow-up, and the number of survivors with improved (ie, z ≥ +1.645), stable (z between −1.645 and +1.645), and declined (z ≤ −1.645) performance were counted for each test. RCI’s assess whether the score difference of a patient at 2 time points is more likely explained by measurement error, or whether it is a significant change. RCI’s take into account the improvements in cognitive test performance due to repeated evaluation with the same test materials that typically occur in healthy populations, and therefore provide a measure of reliable changes in performance in individual patients, compared with changes in the performance of healthy controls (ie, normative data). As a result, this index identifies changes in test scores that are both clinically and statistically meaningful. The index is derived from the standard error of measurement of each measure and represents the 90% confidence interval for the difference in performance between 2 assessments that are expected if no real change occurred. The 90% confidence interval is obtained by multiplying the standard deviation of the initial assessment—follow-up assessment difference by a z-score cut-off value of ±1.645 (corresponding with a 2-tailed alpha of 0.10%, 90% confidence interval).

Predictors of Group Neurocognitive Performance and Psychological Symptoms

We used principal components analyses to create neurocognitive factors, as previously described(¹⁴; Supplementary Table 1). Multivariable linear mixed effects analyses were conducted to evaluate predictors of change in each neurocognitive factor and psychological measure, with age, sex, education, surgery, tumor laterality, tumor location, time between assessments, and radiation delivered to brain ROI’s as fixed effects, and individual-specific baseline neurocognitive factors and psychological symptoms as random intercepts. A 2-level model was examined, with neurocognitive factors and psychological symptoms (level-1) nested within individuals (level-2). We first evaluated the intraclass coefficient to determine whether a nesting structure was appropriate, with neurocognitive factors and psychological symptoms as dependent variables and participant ID as the random intercept. We then evaluated demographic and medical variables (Model 1). Surgery and tumor location were confounded (ie, all patients that had frontal tumors received surgery) and entered into separate models (Model 1a and 1b, respectively). The time between assessments was included to quantify the longitudinal nature of this multilevel model. Because assessments completed shortly after surgery or radiation may be influenced by acute sequelae of treatment, the time between most recent surgery or radiation treatment and the first assessment was included in a separate model. In addition, depression (ie, BDI-II) and anxiety (ie, STAI-S) were included in the neurocognitive outcomes models as psychological variables may influence neurocognitive outcomes. Model 1a is represented below.

Model 1a: Outcome = β₀ + β₁ Age + β₂ Sex + β₃ Education + β₄ Surgery + β₅ Tumor Laterality + β₆ Time + β₇ Depression + β₈ Anxiety + (1 | Participant ID)

In Model 2, significant predictors from Model 1 were retained and dosimetry to circumscribed brain regions (cerebellum, dorsal frontal, hippocampi, parietal-occipital, subcortical, and ventral frontal regions) was individually evaluated. Dosimetric variables (mean dose or D50) that improved model fit according to maximum likelihood indices were retained. Survivors who did not receive radiation were included as having zero dose to any brain structures, ensuring robust modeling of baseline function and nonradiation variables (including the model intercepts and fixed effect coefficients). Time between assessments was included in Model 2 even if not significant in Model 1 to quantify the longitudinal nature of this multilevel model. An example of Model 2 is included below, although this varied based on Model 1 results for each of the outcome variables.

Model 2: Visuomotor Processing Speed = β₀ + β₁ Time + β₂ ROI Dose + (1 | Participant ID)

ROI’s significantly contributing to the neurocognitive factor models were retained in exploratory post hoc analyses, where effects of significant predictors on individual neurocognitive tests within neurocognitive factors were evaluated. In all models, estimates of normality were computed. Variables that fell outside of the −1/1 range for skewness were log-transformed. Because a high dose to the left hippocampus has been associated with worse verbal memory,¹⁹ the effects of left hippocampus dose on verbal memory were also explored.

Statistical significance was defined as P < .05. P-values are 2-sided and are reported as unadjusted values. Effect sizes (Partial R²) demonstrating the strength of the predictor variable in the model are interpreted as small (0.02), medium (0.13), and large (0.26).²⁷ To reduce risk of family-wise type I errors, we reduced the number of comparisons by using neurocognitive factors rather than individual test scores in our models. Exploratory post hoc analyses were only conducted with significant omnibus tests. All analyses were performed using R version 1.2.5042.

Results

Of 18 survivors who met eligibility criteria, 1 was excluded due to a prior history of comorbid neurological and psychiatric disease. The final sample consisted of 17 participants (53% men), mean age of 51.5 (SD = 6.73) years at the first assessment. The mean time between the first and last assessment was 4.19 years (SD = 2.78, range 0.5–7.8 years). Participant demographic and medical characteristics are shown in Table 1 and Supplementary Table 3.

Table 1.

Participant Demographic, Medical, and Neurocognitive and Psychological Test Score Characteristics

	Mean (SD)	Range
Demographic and medical variables
Age at diagnosis (y)	48.29 (8.73)	29–63
Age at first assessment (y)	51.53 (6.73)	41–64
Age at most recent assessment (y)	56.00 (6.85)	43–71
Education (y)	16.23 (3.17)	11–21
	N (%)
Sex (male)	9 (53.94)
Tumor location
Frontal	3 (17.65)
Suprasellar	6 (35.30)
Temporal	6 (35.30)
Parietal	1 (5.88)
Occipital	1 (5.88)
Tumor laterality
Left	10 (58.82)
Right	5 (29.41)
Bilateral	2 (11.76)
Surgery (yes)	13 (76.47)
WHO grade
Grade I	9 (53.94)
Grade II	3 (17.65)
Unresected	5 (29.41)
Radiation dose
5000 cGy	6 (35.30)
5400 cGy	4 (23.53)
6000 cGy	3 (17.65)
No radiation	4 (23.53)
	First assessment mean (SD)	Follow-up assessment mean (SD)
Neurocognitive factor scores
Visuomotor processing speed	−0.38 (1.55)	−0.46 (2.00)
Executive function	−0.24 (0.60)	0.03 (0.69)
Memory	−0.57 (1.11)	−0.57 (1.15)
Attention/working memory	−0.36 (0.79)	−0.37 (0.72)
Neurocognitive test scores
TMT A	−0.17 (1.44)	−0.59 (2.58)
TMT B	−0.62 (2.84)	−0.77 (4.06)
WAIS coding	−0.31 (1.14)	−0.20 (1.20)
WAIS symbol search	−0.29 (1.38)	−0.19 (1.24)
WCST errors	−0.35 (0.89)	0.01 (0.84)
FAS	−0.46 (0.57)	−0.10 (0.72)
WAIS arithmetic	0.15 (1.05)	0.12 (1.00)
RCFT delayed recall	−1.25 (1.10)	−0.87 (1.29)
CVLT-II LDFR	−0.19 (1.24)	−0.13 (1.34)
WAIS digit forward	−0.72 (0.88)	−0.18 (0.92)
WAIS digit backward	−0.31 (0.94)	−0.56 (0.76)
Psychological test scores
BDI-II	15.65 (10.80)	12.88 (10.91)
STAI state	54.88 (11.93)	57.06 (16.93)
STAI trait	58.81 (12.44)	60.25 (19.15)

Changes in Neurocognitive Performance and Psychological Symptoms

Mean neurocognitive performance across time points is presented in Table 1. As a group, performance ranged from average to below average across neurocognitive variables and time points, although there was significant variability within individuals and over time. RCI analyses indicated that a higher proportion of survivors demonstrated stable neurocognitive trajectories (range = 56%–93% across tests) more often than improvements (range = 0%–31%) and declines (range = 0%–36%; Figure 1). The largest proportion of improvements was found on tests of processing speed (WAIS Coding, 31%), visual memory (RCFT Delayed Recall, 25%), and working memory (WAIS Digits Backward, 21%). The largest proportion of declines was exhibited on tests of attention (WAIS Digit Forward, 36%), processing speed (WAIS Coding, 13%), and working memory (WAIS Arithmetic, 7%). More variability was identified among psychological symptoms, with 41%–73% exhibiting no change, 25%–29% improving, and 0%–31% worsening (Figure 1). Individual neurocognitive and psychological test trajectories are shown in Supplementary Figure 2.

Figure 1.

Proportions of patients with improved, stable, or declined neurocognitive performance and psychological symptoms. Note: Changes were evaluated between first and most recent assessment using Reliable Change Indices with correction for practice effects.¹ The number of patients with improved (ie, filled dark gray square +1.645), stable (between −1.645 and +1.645), and declines (light gray square −1.645) in test performance were counted for each test.

Predictors of Group Neurocognitive Performance and Psychological Symptoms

Baseline demographic and medical model.—The mean intraclass correlation coefficient for the baseline models was 0.514 (Table 2; range = 0.258–0.857), illustrating nonindependence of responses within participants and justifying use of multilevel modeling. There were no differences between the models evaluating the time between assessments and the time between most recent surgery or radiation treatment and the first assessment. As such, only the time between assessment models are included here. The time between most recent surgery or radiation treatment and the first assessment are shown in Supplementary Table 4. Older age was associated with decline in Memory (β = −0.054, P = .035, partial R² = 0.215) and Attention/Working Memory (β = −0.072, P = .016, partial R² = 0.432) factors. Higher anxiety was associated with a decline in Executive Function (β = −0.028, P = .009*, partial R² = 0.228). Greater time between assessments was associated with improvements in depressive symptoms (β = −0.107, P = .036, partial R² = 0.144). The remaining demographic and medical predictors were not significantly associated with neurocognitive or psychological outcomes and were not retained in Model 2 analyses. However, moderate to large effects were identified among several additional demographic and medical variables, with the largest effects found between older age and higher state anxiety (partial R² = 0.345), right-sided tumors and higher trait anxiety (partial R² = 0.331), and frontal tumors and worse executive function (partial R² = 0.285).

Table 2.

Effects of Demographic, Medical, and Radiation Dosimetry Variables on Neurocognitive Factors and Psychological Symptoms

	β	SE	df	t-value	p-value	Partial R2†	χ2 (p-value)
Model 1a: Effects of Demographic, Medical, and Psychological Variables on Neurocognitive Factors and Effects of Demographic and Medical Variables on Psychological Symptoms
Visuomotor and processing speed
Intercept	0.103	0.541	10.130	0.191	0.852	---
Age	0.003	0.069	10.812	0.041	0.968	0.000
Sex	−0.486	0.473	10.429	−1.026	0.328	0.094
Education	0.087	0.155	10.071	0.562	0.586	0.031
Time	−0.001	0.007	20.136	−0.070	0.945	0.000
Surgery	−0.847	0.533	10.547	−1.590	0.141	0.197
Tumor Laterality	−0.255	0.534	10.831	−0.478	0.642	0.021
Depression	−0.016	0.024	14.110	−0.660	0.520	0.030
Anxiety	−0.020	0.018	15.474	−1.077	0.298	0.067
Executive function
Intercept	−0.084	0.124	6.559	−0.675	0.523	−−−
Age	0.007	0.016	7.101	0.451	0.665	0.019
Sex	0.092	0.112	7.062	0.825	0.436	0.063
Education	0.023	0.035	6.130	0.665	0.530	0.048
Time	0.002	0.003	20.197	0.752	0.461	0.024
Surgery	−0.120	0.130	8.585	−0.926	0.380	0.067
Tumor laterality	0.061	0.131	7.797	0.469	0.652	0.019
Depression	−0.009	0.014	19.125	−0.595	0.559	0.015
Anxiety	−0.028	0.010	22.939	−2.831	0.009*	0.228
Memory
Intercept	−0.447	0.242	10.293	−1.846	0.094	---
Age	−0.054	0.032	11.022	−1.687	0.035*	0.215
Sex	0.171	0.217	10.932	0.786	0.449	0.057
Education	0.039	0.068	9.827	0.573	0.580	0.035
Time	0.005	0.006	21.562	0.802	0.431	0.028
Surgery	0.040	0.252	12.413	0.158	0.877	0.002
Tumor Laterality	−0.255	0.254	11.808	−1.006	0.335	0.082
Depression	0.016	0.026	20.102	0.621	0.542	0.017
Anxiety	−0.033	0.018	22.963	−1.784	0.088	0.105
Attention and working memory
Intercept	−0.365	0.194	9.946	−1.875	0.090	---
Age	−0.072	0.025	10.750	−2.846	0.016*	0.432
Sex	0.213	0.172	10.479	1.241	0.242	0.130
Education	0.058	0.055	9.787	1.047	0.320	0.103
Time	0.006	0.003	23.000	1.875	0.074	0.131
Surgery	−0.027	0.194	10.870	−0.138	0.892	0.002
Tumor laterality	−0.095	0.196	11.200	−0.487	0.636	0.021
Depression	0.002	0.013	15.416	0.177	0.862	0.002
Anxiety	−0.001	0.010	18.362	−0.120	0.906	0.001
Depression
Intercept	15.109	2.046	9.778	7.383	0.000*	---
Age	0.408	0.262	11.277	1.556	0.147	0.185
Sex	−3.155	1.905	12.580	−1.656	0.122	0.187
Education	0.273	0.620	12.239	0.441	0.667	0.016
Time	−0.111	0.047	31.881	−2.361	0.024*	0.144
Surgery	−3.265	2.054	11.113	−1.590	0.140	0.194
Tumor laterality	2.450	1.972	10.785	1.243	0.240	0.132
State anxiety
Intercept	55.524	3.228	9.203	17.198	0.000*	---
Age	0.840	0.417	10.200	2.013	0.071	0.291
Sex	−3.428	2.914	10.942	−1.176	0.265	0.116
Education	0.541	0.957	10.854	0.565	0.584	0.029
Time	−0.087	0.065	29.940	−1.349	0.188	0.056
Surgery	−0.823	3.210	10.131	−0.256	0.803	0.007
Tumor laterality	5.178	3.148	9.718	1.645	0.132	0.224
Trait anxiety
Intercept	58.083	3.221	8.585	18.034	0.000*	---
Age	0.701	0.424	10.160	1.653	0.129	0.221
Sex	−4.624	3.001	11.494	−1.541	0.150	0.178
Education	0.170	0.985	11.553	0.173	0.866	0.003
Time	−0.041	0.079	29.836	−0.521	0.606	0.009
Surgery	−2.913	3.261	10.151	−0.893	0.392	0.077
Tumor laterality	6.683	3.174	9.495	2.105	0.063	0.331

	β	SE	df	t-value	p-value	Partial R2†	χ2 (p-value)
Model 1b: Effects of demographic, medical, and psychological variables on neurocognitive factors and effects of demographic and medical variables on psychological symptoms
Visuomotor and processing speed
Intercept	0.208	0.647	9.861	0.322	0.754	---
Age	0.046	0.072	10.557	0.638	0.537	0.037
Sex	−0.800	0.546	10.017	−1.465	0.174	0.176
Education	0.245	0.168	9.788	1.454	0.177	0.177
Time	−0.004	0.008	19.783	−0.475	0.640	0.011
Tumor laterality	0.109	0.596	10.389	0.182	0.859	0.003
Tumor location	0.937	0.745	9.916	1.257	0.237	0.137
Depression	−0.007	0.023	14.187	−0.314	0.758	0.007
Anxiety	−0.024	0.019	15.036	−1.291	0.216	0.095
Executive function
Intercept	−0.321	0.131	0.146	−2.446	0.023	---
Age	0.004	0.015	0.017	0.264	0.794	0.007
Sex	0.183	0.112	0.119	1.641	0.114	0.222
Education	0.006	0.033	0.035	0.167	0.869	0.003
Time	0.004	0.003	0.013	1.245	0.226	0.060
Tumor laterality	−0.015	0.127	0.144	−0.116	0.908	0.001
Tumor location	−0.301	0.152	0.165	−1.978	0.060	0.285
Depression	−0.008	0.013	0.017	−0.649	0.523	0.015
Anxiety	−0.028	0.009	0.010	−2.957	0.007*	0.245
Memory
Intercept	−0.471	0.281	10.538	−1.675	0.123	---
Age	−0.056	0.032	11.260	−1.738	0.038*	0.224
Sex	0.193	0.240	10.541	0.806	0.438	0.063
Education	0.029	0.072	9.972	0.405	0.694	0.018
Time	0.005	0.006	22.278	0.812	0.425	0.027
Tumor Laterality	−0.280	0.271	11.501	−1.030	0.324	0.090
Tumor Location	−0.076	0.326	10.790	−0.233	0.820	0.005
Depression	0.013	0.025	21.949	0.534	0.599	0.011
Anxiety	−0.031	0.018	22.980	−1.702	0.102	0.096
Attention and working memory
Intercept	−0.420	0.224	10.053	−1.875	0.090	---
Age	−0.072	0.025	10.900	−2.869	0.015*	0.434
Sex	0.236	0.190	10.300	1.241	0.242	0.133
Education	0.054	0.058	9.862	0.924	0.378	0.082
Time	0.006	0.003	22.995	1.896	0.071	0.132
Tumor laterality	−0.115	0.210	11.039	−0.547	0.595	0.027
Tumor location	−0.072	0.258	10.181	−0.279	0.786	0.008
Depression	0.002	0.013	16.593	0.174	0.864	0.002
Anxiety	−0.001	0.010	18.447	−0.101	0.920	0.001
Depression
Intercept	12.449	2.579	10.470	4.828	0.001*	---
Age	0.467	0.283	9.903	1.652	0.130	0.208
Sex	−2.675	2.272	11.252	−1.177	0.263	0.106
Education	0.433	0.698	10.641	0.621	0.548	0.033
Time	−0.104	0.050	31.917	−2.100	0.043*	0.117
Tumor laterality	2.536	2.302	10.256	1.102	0.296	0.102
Tumor location	−1.522	3.101	10.670	−0.491	0.634	0.021
State anxiety
Intercept	56.606	3.699	10.003	15.303	0.000*	---
Age	0.913	0.406	9.697	2.250	0.059	0.345
Sex	−4.285	3.168	10.672	−1.353	0.204	0.148
Education	0.860	0.975	10.339	0.882	0.398	0.071
Time	−0.096	0.066	30.000	−1.470	0.152	0.065
Tumor laterality	6.060	3.342	9.738	1.813	0.101	0.255
Tumor location	2.597	4.307	10.326	0.603	0.560	0.034
Trait anxiety
Intercept	56.571	3.927	9.721	14.406	0.000*	---
Age	0.792	0.428	9.097	1.850	0.097	0.270
Sex	−4.653	3.395	10.761	−1.370	0.199	0.147
Education	0.450	1.040	10.184	0.433	0.674	0.018
Time	−0.042	0.082	29.743	−0.513	0.612	0.008
Tumor laterality	6.942	3.534	9.335	1.965	0.080	0.291
Tumor location	−0.195	4.591	10.077	−0.043	0.967	0.002

	β	SE	df	t-value	p-value	Partial R2†	χ2 (p-value)
Model 2: Effects of radiation dosimetry variables on neurocognitive factors and psychological symptoms
Visuomotor processing speed
Intercept	−0.397	0.366	15.102	−1.085	0.295	---	---
Time	0.000	0.004	22.173	0.110	0.913	0.000	---
Cerebellum	−0.001	0.000	15.203	−2.395	0.029*	0.276	5.532 (0.019*)
Dorsal frontal	0.006	0.042	14.999	0.150	0.883	0.001	0.025 (0.874)
Hippocampi	0.000	0.000	14.791	0.449	0.660	0.013	0.227 (0.634)
Parietal-occipital	−0.111	0.031	15.242	−3.591	0.003*	0.462	10.627 (0.001*)
Subcortical	0.000	0.000	10.493	−0.014	0.989	0.000	0.000 (0.988)
Ventral frontal	0.000	0.000	14.989	0.739	0.472	0.035	0.607 (0.436)
Executive function
Intercept	−0.127	0.131	10.459	−0.957	0.354	---	---
Time	0.004	0.002	20.439	1.745	0.094	0.101	---
Anxiety	−0.003	0.000	1.810	−4.230	0.000*	0.440
Cerebellum	0.002	0.003	6.550	0.405	0.700	0.020	0.095 (0.758)
Dorsal frontal	0.002	0.013	14.836	0.163	0.873	0.094	0.028 (0.868)
Hippocampi	0.000	0.000	10.396	0.005	0.996	0.002	0.000 (0.992)
Parietal-occipital	0.007	0.013	14.911	0.541	0.597	0.004	0.324 (0.569)
Subcortical	0.000	0.000	14.600	0.225	0.825	0.004	0.054 (0.817)
Ventral frontal	0.000	0.000	10.478	−0.346	0.734	0.044	0.137 (0.711)
Memory
Intercept	−0.558	0.195	14.323	−2.857	0.124*	---	---
Age	−0.080	0.005	13.952	1.545	0.018*	0.061	---
Time	0.007	0.030	35.462	−2.684	0.131	0.335	---
Cerebellum	0.000	0.000	10.384	−0.093	0.927	0.001	0.013 (0.909)
Dorsal frontal	0.021	0.019	13.688	1.130	0.278	0.084	1.494 (0.222)
Hippocampi	0.000	0.000	12.696	−0.627	0.542	0.030	0.476 (0.490)
Parietal-occipital	0.033	0.018	13.659	1.854	0.085	0.200	3.781 (0.052)
Subcortical	0.000	0.000	10.392	0.120	0.906	0.001	0.012 (0.911)
Ventral frontal	0.000	0.000	13.367	−0.905	0.382	0.055	1.002 (0.317)
Attention/working memory
Intercept	−0.338	0.129	13.318	−2.619	0.021*	---	---
Age	−0.061	0.021	14.504	−2.968	0.010*	0.404	---
Time	0.005	0.003	35.852	1.754	0.088	0.078	---
Cerebellum	0.000	0.000	13.593	−1.449	0.170	0.130	2.396 (0.122)
Dorsal frontal	0.020	0.012	13.604	1.625	0.127	0.159	2.960 (0.085)
Hippocampi	0.000	0.000	10.316	-0.140	0.891	0.001	0.026 (0.872)
Parietal-occipital	0.003	0.014	13.651	0.246	0.809	0.004	0.074 (0.785)
Subcortical	0.000	0.000	14.018	0.828	0.421	0.046	0.812 (0.367)
Ventral frontal	0.000	0.000	13.720	0.128	0.900	0.001	0.018 (0.893)
Depression
Intercept	13.901	1.890	14.190	7.352	0.000*	---	---
Time	−0.085	0.044	25.533	−1.908	0.068	0.108	---
Cerebellum	0.002	0.002	13.605	1.225	0.241	0.089	1.643 (0.199)
Dorsal frontal	−0.274	0.177	13.194	−1.545	0.146	0.139	2.547 (0.110)
Hippocampi	0.004	0.002	11.182	2.717	0.019*	0.367	6.955 (0.008*)
Parietal-occipital	−0.144	0.189	13.585	−0.762	0.459	0.038	0.653 (0.419)
Subcortical	0.004	0.004	12.395	1.057	0.311	0.072	1.249 (0.264)
Ventral frontal	0.002	0.002	13.358	1.020	0.326	0.065	1.160 (0.281)
State anxiety
Intercept	56.169	2.865	13.407	19.606	0.000*	---	---
Time	−0.030	0.056	22.180	−0.548	0.589	0.010	---
Cerebellum	0.002	0.003	13.074	0.496	0.628	0.017	0.279 (0.597)
Dorsal frontal	−0.281	0.276	13.148	−1.018	0.327	0.068	1.146 (0.284)
Hippocampi	0.003	0.003	12.224	0.989	0.342	0.069	1.086 (0.297)
Parietal-occipital	−0.097	0.284	12.859	−0.341	0.738	0.008	0.131 (0.718)
Subcortical	−0.002	0.006	12.743	−0.439	0.668	0.014	0.216 (0.642)
Ventral frontal	0.000	0.002	12.827	0.043	0.966	0.000	0.002 (0.961)
Trait anxiety
Intercept	58.167	3.044	12.651	19.108	0.000*	---	---
Time	−0.010	0.074	23.862	−0.138	0.891	0.000	---
Cerebellum	0.005	0.003	12.290	1.613	0.132	0.015	2.774 (0.096)
Dorsal frontal	−0.396	0.284	12.124	−1.391	0.189	0.122	2.108 (0.146)
Hippocampi	0.006	0.003	11.042	1.940	0.078	0.236	3.924 (0.047*)
Parietal-occipital	0.013	0.304	12.289	0.040	0.969	0.000	0.001 (0.969)
Subcortical	0.003	0.006	11.760	0.474	0.644	0.016	0.270 (0.603)
Ventral frontal	0.001	0.003	12.251	0.246	0.810	0.004	0.076 (0.783)

Notes: β = regression coefficients; SE = standard error; df = degrees of freedom.

Model 1 included time between assessments, demographic variables, and medical variables. Surgery and tumor location were confounding variables and were therefore entered separately into the model (Model 1a and 1b, respectively). Model 2 included time between assessment, significant demographic variables and medical variables from Model 1, and ROI’s. Continuous independent variables were centered around their mean, whereas categorical independent variables were effect-coded before being entered into the model. Effect-coding contrasts each group mean with the grand mean. Sex was coded as −1 = Female and 1 = Male. Surgery was coded as No = −1 and Yes =1. Tumor laterality was coded as −1 = Right, 0 = Bilateral, and 1 = Left. Tumor location was coded as 1 = Frontal and −1 = Not frontal, where −1 represents the contrasting group.

*Significant P-values < .05.

Dose to ROI’s.—Because there were no differences between the effects of mean dose or D50 to ROI’s on neurocognitive and psychological outcomes, only mean dose models are presented here. D50 models are shown in Supplementary Tables 5 and 6. In terms of neurocognitive performance, slower visuomotor processing speed was associated with a higher mean dose to the parietal-occipital region (χ²(1) = 10.627, P = .001, partial R² = 0.462) and cerebellum (χ²(1) = 5.532, P = .019, partial R² = 0.276). Although no ROI doses were significantly associated with the other neurocognitive factors, medium effects were identified between parietal-occipital dose and memory (partial R² = 0.200), and dorsal frontal dose and attention/working memory (partial R² = 0.159).

For psychological symptoms, a higher mean dose to hippocampi was associated with increased depressive symptoms (χ²(1) = 6.955, P = .008, partial R² = 0.367) and trait anxiety (χ²(1) = 3.924, P = .048, partial R² = 0.236). Variable coefficients from Model 2 are shown in Table 2.

Exploratory Analyses

Higher dose to the parietal-occipital region was associated with slower performance on timed tests (TMTA: χ²(1) = 7.247, P = .007, partial R² = 0.363; TMTB: χ²(1) = 11.315, P = .001, partial R² = 0.505; Symbol Search: χ²(1) = 5.666, P = .017, partial R² = 0.288; Coding: χ²(1) = 6.060, P = .014, partial R² = 0.302). Similarly, a higher dose to the cerebellum was associated with slower TMTA (χ² (1) = 7.107, P = .008, partial R² = 0.357) and TMTB (χ²(1) = 7.394, P = .006, partial R² = 0.368). Dose to the left hippocampus was not associated with verbal memory (CVLT-II LDFR). Variable coefficients for these models are shown in Table 3.

Table 3.

Effects of Radiation Dose to Circumscribed Brain Regions as Predictors of Impairment on Individual Neurocognitive Tests

ROI	Neurocognitive test	B	SE	df	t-value	P-value	Partial R2†	χ2 (P-value)
Cerebellum
	TMT A	−0.001	0.000	13.749	−2.786	0.015*	0.357	7.107 (.008*)
	TMT B	−0.002	0.001	13.954	−2.855	0.013*	0.368	7.394 (.006*)
	WAIS coding	0.000	0.000	15.529	−0.622	0.543	0.025	0.431 (.512)
	WAIS symbol search	0.000	0.000	15.370	−1.194	0.251	0.087	1.544 (.214)
Parietal-occipital
	TMT A	−0.106	0.038	13.852	−2.816	0.014*	0.363	7.247 (.007*)
	TMT B	−0.215	0.057	13.992	−3.762	0.002*	0.505	11.315 (.001*)
	WAIS coding	−0.059	0.023	15.534	−2.530	0.023*	0.302	6.060 (.014*)
	WAIS symbol search	−0.062	0.026	15.161	−2.432	0.028*	0.288	5.666 (.017*)
Left hippocampus
	CVLT-II LDFR	0.000	0.000	10.380	−0.053	0.958	0.000	0.003 (.954)

Notes: Dose to cerebellum and parietal-occipital regions significantly predicted the visuomotor processing speed factor, and thus individual neurocognitive tests for this factor are presented. Executive function, memory, and attention/ working memory factors were not predicted by dosimetry variables and therefore individual tests loading onto these factors cannot be interpreted and are not presented.

^†Partial R² values of 0.02, 0.13, and 0.26 represent small, medium, and large effects, respectively (Cohen, 1988).

*Significant P-values < .05.

Discussion

Meningioma survivors experience neurocognitive and psychological symptoms long after treatment,^5,6 which can affect quality of life, including occupational functioning.⁶ However, post-treatment supportive care resources for this population are scarce.^17,18 Identifying which survivors are at risk of long-term neurocognitive and psychological symptoms will inform the appropriate and efficient use of intervention resources. How these symptoms evolve over time, and whether radiation contributes to these changes in this population is largely unknown. Novel findings of our study indicate that incidental radiation to brain regions beyond the meningioma tumor site is associated with long-term changes in neurocognitive and psychological outcomes. This approach provides a proof of concept that should be used in larger prospective trials investigating treatment sequelae in survivors.

Overall, most of our survivors demonstrated stable—albeit often below average—neurocognitive trajectories. The largest proportion of improvements was found on tests of processing speed, visual memory, and working memory, whereas the largest proportion of decline was found on tests of attention, processing speed, and working memory, comparable with findings from Rijnen et al.⁹ More variability was identified in psychological symptoms; approximately one-third of survivors exhibited improvements in depression, one-third exhibited worsening of depression or trait anxiety. Our findings demonstrate variability across neurocognitive and psychological trajectories which may be masked by group-level analyses. Individual-level approaches to identify survivors at risk of adverse neurocognitive and psychological outcomes are recommended.

Radiation dosimetry to brain regions beyond the tumor site was associated with changes in neurocognitive and psychological symptoms; dose to parietal-occipital and cerebellar regions was associated with slowed visuomotor processing speed, and dose to hippocampi was associated with increased symptoms of depression and trait anxiety. Effect size analyses indicated that several demographic, medical, psychological, and radiation dosimetry variables had medium to large effects on these outcomes, underscoring the significance of these associations, and warranting further investigation in studies with larger samples. Contrary to our hypotheses, we did not find relations between dosimetry and other neurocognitive and psychological outcomes including left hippocampus and verbal memory, parietal-occipital and dorsal frontal regions and executive functions, and ventral frontal region and psychological symptoms. This could be due to the study sample size, which limited power to detect weaker associations between radiation dose and psychological endpoints.

Associations between parietal-occipital regions and visuomotor ability (eg, transformation of visual information into commands for directing attention and guiding motor output, mental manipulation of concrete shapes and abstract symbols) are well-established.²⁸ High dose to the cerebellum was also associated with slowed visuomotor processing speed. The cerebellum is thought to facilitate the connection of distributed networks involved in visuomotor and processing speed functions.²⁹ Our finding supports the idea that distributed neurocognitive processes that rely on a complex network of structures are also sensitive to radiation. This is consistent with the finding that the most common neurotoxic effect of radiation is not focal necrosis but diffuse cerebral injury.³⁰ Our work provides a rationale for developing novel treatment modalities that reduce radiation dose to normal brain structures, such as proton beam therapy.³¹ Many countries around the world, including Canada, are working to make this technology broadly available to patients with brain tumors,³² including meningioma.

Radiation dose was also associated with psychological symptoms, a finding that has not been previously reported. Specifically, a high dose to the hippocampi was associated with increases in symptoms of depression and anxiety. The hippocampus is an active site of adult neurogenesis, vulnerable to radiation.³³ Research has focused primarily on the effects of radiation on the hippocampus and memory,¹⁹ given the fundamental role of the hippocampus in memory function.³⁴ Yet, various studies have suggested that impaired adult hippocampal neurogenesis can result in depression (for a review, see Liu et al.³⁵). It is plausible that radiation to the hippocampus damages neurogenic stem cells in that region,³³ increasing risk of psychological symptoms. Moreover, the hippocampus, a central component of the limbic system, has strong connections with emotion-related brain regions including the amygdala and prefrontal cortex that may receive similar radiation doses. Thus, it is unclear which structures are driving this finding. What is clear is that our current understanding of the psychological effects of radiation is limited. Psychological symptoms in brain tumor populations are thought to be caused by multiple factors, including neurophysiologic changes due to the tumor and treatment, and psychosocial adjustment to a brain tumor diagnosis.^36–38

Study Limitations

This was a retrospective study from a small sample of patients who completed neuropsychological assessments for clinical purposes. The retrospective design resulted in differences in numbers of assessments, and timing of test administration. We did not have available pre-treatment baseline data for many participants, limiting conclusions related to neurocognitive and psychological change as a result of radiation toxicity. In addition, 76% of participants in our sample received radiation; radiation is only standard of care for meningioma that is not safely amenable to surgery or after incomplete surgical resection⁴ and our results should be considered in that context. The small sample size resulted in limited power to examine factors that may be contributing to neurocognitive and psychological symptoms, including cardiovascular risk factors, anti-epileptic medications, and other concomitant medications and medical factors. Education has been associated with cognitive outcomes in the literature, but this association was not significant in our sample. Most of our participants were highly educated (ie, just 3 of 17 [17.6%] participants obtained a high school education or less), which may be driving this nonsignificant finding. Nonetheless, given that education is associated with greater cognitive reserve, our findings may underestimate the impact of radiation on patients with less cognitive reserve. Although the demographic characteristics of our sample are consistent with reported meningioma tumor diagnoses across Canada,¹ this work was conducted in a large urban tertiary cancer center and generalizability may be a concern. Given that participants were treated in a multidisciplinary brain tumor center with dedicated comprehensive medical and psychosocial care, these results may underestimate the symptoms of those who live in rural areas or are treated at smaller institutions where services are more limited. Nonetheless, our results provide preliminary evidence that radiation delivered to circumscribed brain regions is associated with neurocognitive and psychological outcomes. Notably, although this study examined reliable change using RCI with correction for practice effects, the best way to measure reliable change remains a topic of discussion in the field^37,38. The medium to large effect sizes support using this approach in larger prospective trials. It is important to employ data analytic strategies that identify survivors at risk of decline and elucidate mechanisms underlying these effects. Such studies would aid in the selection of patients who are likely to experience neurocognitive and psychological sequelae following radiation and allow for the development of refined treatment methods to improve the unmet needs of this population.^17,18

Conclusion and Clinical Implications

Although advancements in surgical and radiation techniques for meningioma have improved patient care, little is known about the long-term quality of survivorship in these patients. Our study suggests that many survivors continue to experience neurocognitive and psychological symptoms long after diagnosis, underscoring the need for neuropsychological assessment and psychosocial resources to manage symptoms. Understanding the long-term effects of radiation is important given the use of radiation in meningioma patients, the prevalence of this disease and increasing survival rates, and the potential impact of neurocognitive and psychological factors on functional outcomes.⁶ Given that other brain tumor populations experience similar neurocognitive and psychological symptoms following treatment[39,40],^39,40 further investigation of incidental radiation to the brain for other brain tumor patients is warranted. Continued research in these areas will allow for opportunities to develop and test supportive care interventions (eg, psychotherapy, cognitive rehabilitation) as well as novel radiation modalities (eg, proton beam therapy) to mitigate and reduce the effects of these toxicities on the quality of life.

Supplementary Material

Supplementary material is available online at Neuro-Oncology (https://academic.oup.com/neuro-oncology).

Funding

This work was supported in part by an award from the Strategic Training in Transdisciplinary Radiation Science for the 21st Century [AS]; the Princess Margaret Cancer Foundation; and the Ontario Ministry of Health and Long-Term Care (OMOHLTC). The views expressed do not necessarily reflect those of the OMOHLTC.

Conflict of Interest Statement

DST received travel funding from Mevion Medical Systems and Elekta AB in 2022, outside the present study. The authors have no additional conflicts of interest.

Ethics Approval

This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the University Health Network and University of Toronto research ethics boards (REB #20-5711 and 38523, respectively).