Etiology- and region-specific characteristics of transependymal cerebrospinal fluid flow

Abstract

OBJECTIVE

Transependymal flow (TEF) of CSF, often delineated as T2-weighted hyperintensity adjacent to the lateral ventricles on MRI, is a known imaging finding, usually in the setting of CSF flow disturbances. Specific radiological features of TEF and their relationships with clinical markers of hydrocephalus and underlying disease pathology are not known. Here, the authors describe the radiological features and clinical associations of TEF with implications for CSF circulation in the setting of intracranial pathology.

METHODS

After obtaining IRB review and approval, the authors reviewed the radiological records of all patients who underwent intracranial imaging with CT or MRI at St. Louis Children’s Hospital, St. Louis, Missouri, between 2008 and 2019 to identify individuals with TEF. Then, under direct review of imaging, TEF pattern, degree, and location and underlying pathology and other radiological and clinical features pertaining to CSF circulation and CSF disturbances were noted.

RESULTS

TEF of CSF was identified in 219 patients and was most prevalent in the setting of neoplasms (72%). In 69% of the overall cohort, TEF was seen adjacent to the anterior aspect of the frontal horns and the posterior aspect of the occipital horns of the lateral ventricles, and nearly half of these patients also had TEF dorsal to the third ventricle near the splenium of the corpus callosum. This pattern was independently associated with posterior fossa medulloblastoma when compared with pilocytic astrocytoma (OR 4.75, 95% CI 1.43–18.53, p = 0.0157). Patients with congenital or neonatal-onset hydrocephalus accounted for 13% of patients and were more likely to have TEF circumferentially around the ventricles without the fronto-occipital distribution. Patients who ultimately required permanent CSF diversion surgery were more likely to have the circumferential TEF pattern, a smaller degree of TEF, and a lack of papilledema at the time of CSF diversion surgery.

CONCLUSIONS

CSF transmigration across the ependyma is usually restricted to specific periventricular regions and is etiology specific. Certain radiological TEF characteristics are associated with tumor pathology and may reflect impaired or preserved ependymal fluid handling and global CSF circulation. These findings have implications for TEF as a disease-specific marker and in understanding CSF handling within the brain.

Disorders of CSF circulation and CSF flow are common, with a global hydrocephalus prevalence of 85 per 100,000 children and 11 per 100,000 adults.¹ Although CSF is typically confined to the ventricles and subarachnoid, subdural, and peri-/paravascular spaces, under certain conditions it may cross the ventricular ependyma into the brain parenchyma; in this case, is detectable on intracranial imaging. This phenomenon is often referred to as transependymal flow (TEF).

TEF was initially described as blurring of the anterior and lateral margins of the frontal horns of the lateral ventricles on CT in patients with hydrocephalus. The now more common term, TEF, was first described as periventricular lucency.² Since then, the pathophysiology of TEF has been speculated to involve disruption of the ependymal layer of the ventricular wall and subsequent extravasation of CSF into the periventricular space in certain pathological conditions.^3,4 Despite TEF being a commonly recognized phenomenon, especially in acute obstructive hydrocephalus, it may be seen in other clinical conditions and present in a variety of heterogeneous radiological patterns, thereby making TEF a poorly understood finding as it relates to CSF circulation and disease pathology.

Early animal studies of obstructive hydrocephalus showed that TEF of CSF was present in the acute hypertensive state and that albumin labeled with radioactive iodine crossed the periventricular ependyma.^3,5–7 At a cellular level, CSF movement across the ventricular ependyma may occur in either intact or disrupted ependyma.⁴ However, ependymal disruption and periventricular edema is less common in patients with chronic hydrocephalus, as the ependyma in this setting may not allow for visualized movement of CSF into the surrounding parenchyma.^4,5,7 Mechanistic explanations of the pathophysiology of TEF have primarily focused on biomechanical forces and favor TEF of CSF as an absorptive process, rather than as CSF extravasation.⁸ Finally, more recent clinical associations of TEF include acute or subacute obstructive hydrocephalus resulting from tumors, the presence of papilledema, and altered sensorium;⁹ however, TEF has also been noted in patients with nonobstructive hydrocephalus.¹⁰

In this retrospective cohort study, we hypothesized that the radiological features of TEF are associated with underlying brain pathology. Here, we have described the radiological features of the pattern, degree, and location of transependymal CSF flow and relate them to clinically important variables such as diagnosis, ventricular obstruction, intracranial pressure, papilledema, and need for permanent CSF diversion. We show that TEF, rather than being a singular radiological entity, has regional and etiological specificity and may be useful as a radiological adjunct in evaluations of patients with CSF disorders. Furthermore, certain TEF features may have implications for CSF handling at the ependymal surface and, more broadly, relate to altered intracranial CSF circulation in neurological disease.

Methods

Patient Selection and Clinical Variables

This retrospective cohort study was approved and need for informed consent was waived by the Human Research and Protection Office Institutional Review Board at Washington University in St. Louis, Missouri, and in accordance with the relevant guidelines and regulations. Institutional imaging records (MRI and CT) of patients aged 0 to 21 years from 2008 to 2019 were queried for the following search terms in the radiology report: “transependymal flow,” “transependymal edema,” “periventricular edema,” “periventricular lucency,” “interstitial edema,” and “white matter edema.” At our institution the term “transependymal flow” is typically used to describe periventricular CSF.

The collected demographic and clinical variables included age, sex, neurological diagnosis, tumor pathology, presence of ventricular obstruction (defined as ventricular distortion due to local mass effect), location of primary mass effect, type of neurosurgical intervention, qualitative description of opening pressure (brisk, spontaneous, and sluggish), and presence of papilledema on funduscopic examination, if performed. Information was obtained using 1 or more hospital visits to capture all surgical procedures and ophthalmological evaluations.

All patients with papilledema had documentation within 14 days of the date of initial intracranial imaging, so patients who underwent funduscopic examination more than 14 days from the initial scan date were excluded to reduce false negatives. Neurosurgical procedures were counted only after TEF diagnosis on imaging and included craniotomy, endoscopic transsphenoidal tumor resection, stereotactic needle biopsy, and CSF diversion. Surgery for temporary CSF diversion included placement of an external ventricular drain (EVD) and lumbar drain, whereas surgical procedures for permanent CSF diversion included shunt insertion, shunt revision, endoscopic third ventriculostomy, and endoscopic fenestration.

Transependymal Flow Imaging Features

Periventricular TEF was determined, in nearly all cases, by using T2-weighted MRI sequences or T2-weighted FLAIR MRI sequences if the T2-weighted image was unavailable. TEF was categorized by pattern, degree, and location (Fig. 1) on the basis of group consensus among the coauthors. TEF pattern was grouped into 5 categories: fronto-occipital (FO), FO plus splenium of the corpus callosum, circumferential around the lateral ventricles, left/right asymmetry, and anterior/posterior asymmetry. The FO pattern referred to T2-weighted hyperintensity along the anterior aspect of the frontal horns and the posterior aspect of the occipital horns in a symmetric fashion (Fig. 1A), and FO plus splenium referred to the FO pattern plus T2-weighted hyperintensity dorsal to the third ventricle near the splenium of the corpus callosum (Fig. 1B). The circumferential pattern described T2-weighted hyperintensity that encompassed the entire ventricular system without preference for the anterior frontal and posterior occipital regions (Fig. 1C). Left/right asymmetry (Fig. 1D) and anterior/posterior asymmetry (Fig. 1E) described T2-weighted hyperintensity asymmetry in their respective directions. Patients who underwent only CT were included and had corresponding periventricular hypodensities that were evaluated in a similar fashion.

FIG. 1.

Radiological characteristics and patterns of TEF on T2-weighted MRI (left) and FLAIR (right). A: FO pattern (arrows indicate hyperintensity anterior to the frontal horns and posterior to the occipital horns). B: FO plus splenium of the corpus callosum pattern (arrows indicates splenium). C: Circumferential pattern. D: Left/right asymmetry pattern. E: Anterior/posterior asymmetry pattern. F: Mild TEF on T2-weighted MRI (grade 1). G: Moderate TEF on T2-weighted MRI (grade 2). H: Severe TEF on T2-weighted MRI (grade 3). I: TEF with hyperintensity (arrow) superior to the lateral ventricles on T2-weighted MRI. J: TEF with hyperintensity (arrow) around the temporal horn. K: TEF with hyperintensity (arrow) around the third ventricle. L: TEF with hyperintensity (arrow) around the fourth ventricle.

TEF degree was determined on axial-plane images by measuring the linear distance of the T2-weighted or T2-weighted FLAIR hyperintensity from the ependymal surface on the line orthogonal to the apex of the frontal horns or occipital horns. TEF degree was considered 1) mild if this distance was < 2 mm, 2) moderate if between 2 mm and 4 mm, and 3) severe if > 4 mm (Fig. 1F–H). TEF location was classified as superior to the lateral ventricles (Fig. 1I), circumferential around the temporal horns (Fig. 1J), circumferential around the third ventricle (Fig. 1K), and/or circumferential around the fourth ventricle (Fig. 1L). In cases in which tumors or peritumoral edema in the posterior fossa completely or mostly obscured the fourth ventricle, TEF around the fourth ventricle was determined to be absent.

Ventricular size was approximated using the FO horn ratio (FOHR) at the time of the index (preintervention) scan (CT or MRI) with previously described methods.¹¹ In all cases in which a neurosurgical intervention was performed after CT and prior to MRI, an EVD was placed for patients with acute presentation of hydrocephalus related to mass effect from a tumor. Post-EVD MRI was typically performed immediately afterward or on the same day. Due to the time interval between EVD placement and MRI, we noticed that ventricular size (FOHR) and degree of TEF, but not patterns/locations of TEF, were reduced on MRI when compared with the CT administered for the treatment for hydrocephalus. Therefore, we used pre-EVD CT to determine FOHR and degree of TEF and higher-resolution post-EVD MRI to characterize TEF patterns and locations.

T2-weighted and/or T2-weighted FLAIR MRI sequences were independently reviewed and evaluated by the authors (P.H.Y., A.A.B., and E.B.D.), who were otherwise blinded to patient clinical histories. Discrepancies were resolved after discussion with the senior author (J.M.S.).

Statistical Analysis

Statistical analysis of clinical and radiological characteristics (continuous and categorical) and determination of statistical significance were performed using linear regression, the 2-tailed Student t-test, 1-way ANOVA, the chi-square test, and the 2-tailed Fisher exact test. Post hoc analysis using Bonferroni and Fisher least significant difference comparisons were performed where appropriate. Statistical tests were performed using StatPlus Excel addin version 7 (AnalystSoft, Inc.) and SAS software version 9.4 (SAS Institute, Inc.). A p value less than 0.05 was taken to indicate a statistically significant difference.

Results

Patterns of TEF in the Overall Cohort

In total, 219 patients had TEF on intracranial imaging (Table 1 and Supplementary Table 1). Seven patients did not undergo brain MRI and underwent only head CT, whereas the remaining 212 patients had MRI available for direct review. The average age was 8.2 years, and 45.4% were female. TEF varied in pattern, location, and degree but was always seen in the anterior aspect of the frontal horns and the posterior aspect of the occipital horns. The FO and FO plus splenium patterns were the most prevalent TEF patterns observed (total 69.4%). TEF was most commonly seen in patients with neoplasms (72.1%) Ventricular obstruction was observed in 83.6% of the entire cohort and 98.1% of patients with neoplasms. In the overall cohort, when TEF extended beyond the typical locations, it was more likely to be present in the temporal horn (67.6%) rather than the superior aspect of the lateral ventricles (4.6%), third ventricle (37.4%), and fourth ventricle (20.1%).

TABLE 1.

Demographic, radiological, and clinical characteristics and treatments of the cohort

Characteristic	Total*	Neoplasm	Aqueductal Stenosis†	IVH/PHH	Congenital Ventriculomegaly	Postinfectious	Arachnoid Cyst	MMC
Total	219 (100.0)	158 (72.1)	10 (4.6)	10 (4.6)	10 (4.6)	9 (4.1)	5 (2.3)	4 (1.8)
Age, yrs	8.2 (0.03–19.7)	8.9 (7.5–19.7)	10.1 (5.5–19.2)	2.3 (1–9.8)	3.6 (3.4–11.4)	6.8 (2.9–19)	3.9 (1.6–9.2)	3.8 (8.3–11)
Female sex	88 (45.4)	66 (41.8)	0 (0)	5 (50)	4 (40)	5 (55.6)	0 (0)	2 (50)
Radiologic
Average FOHR	0.448	0.439	0.481	0.415	0.490	0.551	0.579	0.455
Ventricular obstruction
Absent	36 (16.4)	3 (1.9)	0 (0)	10 (100)	8 (80)	4 (44.4)	0 (0)	4 (100)
Present	183 (83.6)	155 (98.1)	10 (100)	0 (0)	2 (20)	5 (55.6)	5 (100)	0 (0)
Single focus	174 (79.5)	147 (93.0)	10 (100)	0 (0)	2 (20	5 (55.6)	5 (100)	0 (0)
TEF pattern
FO	75 (34.2)	60 (38.0)	6 (60)	1 (10)	1 (10)	2 (22.2)	0 (0)	0 (0)
FO plus splenium	77 (35.2)	74 (46.8)	1 (10)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)
Circumferential	35 (16.0)	8 (5.1)	0 (0)	3 (30)	6 (60)	4 (44.4)	5 (100)	3 (75)
Lt/rt asymmetry	11 (5.0)	6 (3.8)	1 (10)	4 (40)	0 (0)	0 (0)	0 (0)	1 (25)
Anterior/posterior asymmetry	21 (9.6)	10 (6.3)	2 (20)	2 (20)	3 (30)	3 (33.3)	0 (0)	0 (0)
TEF degree
1 (mild)	50 (22.8)	31 (19.6)	3 (30)	3 (30)	3 (30)	1 (11.1)	4 (80)	2 (50)
2 (moderate)	115 (52.5)	84 (53.2)	5 (50)	5 (50)	7 (70)	5 (55.6)	1 (20)	1 (25)
3 (severe)	54 (24.7)	43 (27.2)	2 (20)	2 (20)	0 (0)	3 (33.3)	0 (0)	1 (25)
TEF location
Superior to lateral ventricles	10 (4.6)	5 (3.2)	0 (0)	1 (10)	1 (10)	2 (22.2)	1 (20)	0 (0)
Temporal horns	148 (67.6)	97 (61.4)	9 (90)	8 (80)	9 (90)	8 (88.9)	3 (60)	4 (100)
3rd ventricle	82 (37.4)	56 (35.4)	3 (30)	4 (40)	4 (40)	5 (55.6)	1 (20)	3 (75)
4th ventricle	44 (20.1)	23 (14.6)	0 (0)	4 (40)	2 (20)	6 (66.7)	2 (40)	3 (75)
Clinical
Funduscopic exam	42 (19.2)	35 (22.2)	3 (30)	0 (0)	2 (20)	1 (11.1)	0 (0)	0 (0)
Papilledema	28 (66.7)	25 (71.4)	2 (66.7)		0 (0)	0 (0)
No papilledema	14 (33.3)	10 (28.6)	1 (33.3)		2 (100)	1 (100)
Opening pressure reported	107 (62.6)	78 (62.9)	4 (40)	6 (60)	4 (40)	4 (44.4)	4 (80)	2 (50)
Brisk	72 (67.3)	55 (70.5)	4 (100)	5 (83.3)	1 (25)	2 (50)	4 (100)	0 (0)
Spontaneous	28 (26.2)	17 (21.8)	0 (0)	1 (16.7)	3 (75)	2 (50)	0 (0)	2 (100)
Sluggish	7 (6.5)	6 (7.7)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)
Treatment
No neurosurgery after TEF	18 (8.2)	8 (5.1)	0 (0)	1 (10)	5 (50)	0 (0)	0 (0)	0 (0)
Underwent neurosurgery after TEF	201 (91.8)	150 (94.9)	10 (100)	9 (90)	5 (50)	9 (100)	5 (100)	4 (100)
Surgery for CSF diversion	171 (85.1)	124 (82.7)	10 (100)	9 (100)	5 (100)	8 (88.9)	5 (100)	4 (100)
Temporary CSF diversion	57 (33.3)	53 (42.7)	0 (0)	0 (0)	0 (0)	1 (12.5)	0 (0)	0 (0)
EVD	56 (32.7)	53 (42.7)				1 (12.5)
Lumbar drain	1 (0.6)	0 (0.0)				0 (0)
Permanent CSF diversion	114 (66.7)	71 (57.3)	10 (100)	9 (100)	5 (100)	7 (87.5)	5 (100)	4 (100)
Shunt insertion	51 (29.8)	38 (30.6)	1 (10)	1 (11.1)	3 (60)	3 (37.5)	3 (60)	1 (25)
Shunt revision	19 (11.1)	4 (3.2)	1 (10)	6 (66.7)	1 (20)	2 (25)	1 (20)	3 (75)
ETV‡	37 (21.6)	25 (20.2)	8 (80)	1 (11.1)	1 (20)	0 (0)	1 (20)	0 (0)
Endoscopic fenestration‡	7 (4.1)	4 (3.2)	0 (0)	1 (11.1)	0 (0)	2 (25)	0 (0)	0 (0)

MMC = myelomeningocele.

Values are shown as number (%) or mean (range).

^*Included all 219 patients. Categories with 3 or fewer patients each are included in Supplementary Table 1.

^†Aqueductal stenosis without evidence of neoplastic process.

^‡If a patient underwent ETV and endoscopic fenestration in the same surgery, it was counted as ETV.

Papilledema was noted in 66.7% of patients who underwent funduscopic examination. Among patients who underwent CSF diversion surgery, 93.5% had a “brisk” or “spontaneous” opening pressure documented. In total, 85.1% of patients underwent surgery for CSF diversion, 66.7% of whom underwent permanent CSF diversion surgical procedures.

Overall, younger age at time of TEF was associated with an asymmetrical/circumferential TEF pattern (p = 0.0017) (Fig. 2A, Supplementary Fig. 1). Severe degree of TEF was significantly associated with larger FOHR compared with mild (p = 0.0002) and moderate (p < 0.0001) TEF degrees (Fig. 2B).

FIG. 2.

Clinical and radiological features of TEF. A: Older patients were more likely to have the FO and FO plus splenium TEF patterns (p = 0.0017). B: FOHR was greater in patients with severe TEF degree (3) than those with mild TEF degree (1) (p = 0.0002) and those with moderate TEF degree (2) (p < 0.0001). C: The FO and FO plus splenium TEF patterns were more likely to be present in the setting of ventricular obstruction (obs.) (p < 0.0001). D: Patients with papilledema (pap.) were more likely to have the FO and FO plus splenium TEF patterns (p = 0.0094). E: Patients without papilledema were more likely to undergo surgery for permanent (Perm.) CSF diversion (p = 0.0076). F: A greater proportion of patients with the asymmetrical/circumferential TEF patterns underwent permanent CSF diversion (p = 0.0033). Median (middle line), interquartile range (box), and minimum and maximum (whiskers) are shown. Temp. = temporary.

Papilledema was more likely in patients with the FO and FO plus splenium TEF patterns rather than the asymmetrical/circumferential TEF patterns (p = 0.0094) (Fig. 2D), and patients with TEF and without papilledema were more likely to undergo permanent rather than temporary CSF diversion (p = 0.0076) (Fig. 2E). Patients with the FO and FO plus splenium TEF patterns were more likely to have ventricular obstruction (p < 0.0001) (Fig. 2C).

Among all patients who underwent surgery for CSF diversion, a larger proportion of patients with the asymmetrical/circumferential TEF patterns was more likely to undergo permanent CSF diversion (p = 0.0033) (Fig. 2F). Of the 8 patients with neoplasms who did not undergo surgery after TEF was noted, 2 were lost to follow-up and the remaining 6 underwent only chemotherapy and/or radiation therapy. Of the 10 patients without neoplasms who did not undergo surgery, 5 patients with macrocephaly and ventriculomegaly were managed conservatively and the mild circumferential TEF pattern resolved, 3 patients had TEF discovered on an outpatient MRI and did not establish care with the department of neurosurgery at our institution, 1 patient had a previously resected cavernoma that developed TEF on postoperative MRI and resolved, and 1 medically sick inpatient had an incidental spontaneous intraparenchymal hemorrhage on MRI and died of unrelated causes 2 days later. To explore the clinical significance of TEF, we next evaluated the TEF patterns on the basis of underlying disease processes.

Patterns of TEF in Patients With Neoplasms

Neoplasms associated with TEF were seen in 158 patients (72.1%) of the entire cohort (Table 1) and most commonly included pilocytic astrocytoma (n = 52), medulloblastoma (27), WHO grade III ependymoma (9), and craniopharyngioma (8) (Supplementary Table 2). FO and FO plus splenium TEF patterns were seen in 85% of these patients. Moderate to severe degrees of TEF were seen in 80.4% of patients. TEF was often seen around the temporal horns (61.4%) but rarely superior to the lateral ventricles (3.2%). Patients with severe degree of TEF had larger ventricles than patients with mild or moderate degrees of TEF (all p < 0.0001) (Fig. 3A). Patients with moderate and severe degrees of TEF were more likely to have the FO plus splenium TEF pattern rather than the FO TEF pattern (p = 0.0397) (Fig. 3B). Patients with papilledema were more likely to undergo temporary rather than permanent CSF diversion surgery (p = 0.0373) (Fig. 3C).

FIG. 3.

TEF in pediatric patients with intracranial neoplastic disease. A: FOHR was greater in patients with severe TEF degree (3) than those with mild TEF degree (1) and those with moderate TEF degree (2) (p < 0.0001). B: Patients with the FO plus splenium TEF pattern were more likely to have greater TEF degree (p = 0.0397). C: Patients with papilledema were less likely to require permanent CSF diversion than those without papilledema (p = 0.0373). D: A greater proportion of patients with single infratentorial (Infratent.) tumors had typical TEF patterns than those with single supratentorial (Supratent.) tumors (p = 0.0014). Median (middle line), interquartile range (box), and minimum and maximum (whiskers) are shown.

To further determine the role of tumor location on TEF pattern, we analyzed a subset of 148 patients with singular intracranial tumors delineated on the basis of supratentorial versus infratentorial location. Despite this broad categorization, we found that infratentorial tumors were more likely to have the FO and FO plus splenium TEF patterns rather than asymmetrical/circumferential TEF patterns (p = 0.0014) (Fig. 3D).

We next asked whether TEF pattern was related to underlying disease pathology independent of tumor location. To test this hypothesis, we performed a subset analysis of the 2 posterior fossa tumor types with the greatest numbers of patients: pilocytic astrocytoma (n = 32) and medulloblastoma (25) (Table 2). In the univariate logistic regression model that used TEF pattern (FO vs FO plus splenium) as the dependent variable, patients with pilocytic astrocytoma were more likely to have the FO TEF pattern, whereas those with medulloblastoma were more likely to have the FO plus splenium TEF pattern (OR 4.0, 95% CI 1.27–14.4, p = 0.0198). Smaller tumor volume (OR 0.98, 95% CI 0.96–0.997, p = 0.0252) and presence of leptomeningeal disease (OR 7.69, 95% CI 1.31–147.1, p = 0.0337) were associated with the FO plus splenium TEF pattern. Patients who ultimately underwent shunt surgery were also more likely to have the FO plus splenium TEF pattern at presentation (p = 0.0279). Age, sex, tumor location within the posterior fossa (fourth ventricle, cerebellar hemisphere, and vermis), TEF degree, opening pressure, and papilledema were not associated with TEF pattern in this subset analysis.

TABLE 2.

Univariate logistic regression model of the FO and FO plus splenium TEF patterns in patients with posterior fossa pilocytic astrocytoma and medulloblastoma

Variable	FO (n = 21)	FO Plus Splenium (n = 36)	OR (95% CI)*	p Value*
Age, yrs†	6.9 ± 3.6	8.6 ± 4.2	1.12 (0.98–1.30)	0.1073
Male sex	13 (61.9)	20 (55.6)	0.77 (0.25–2.29)	0.6388
Tumor location
4th ventricle	6 (28.6)	16 (44.4)	Ref
Cerebellar hemisphere	12 (57.1)	17 (47.2)	0.53 (0.15–1.72)	0.8144
Vermis	3 (14.3)	3 (8.3)	0.38 (0.05–2.51)	0.4458
Pathology
Pilocytic astrocytoma	16 (76.2)	16 (44.4)	Ref
Medulloblastoma	5 (23.8)	20 (55.6)	4.0 (1.27–14.44)	0.0198
Tumor vol, cm3†	63.4 ± 28.1	44.6 ± 30.6	0.98 (0.96–0.997)	0.0252
LM disease	1 (4.8)	10 (27.8)	7.69 (1.31–147.08)	0.0337
FOHR‡	0.45 ± 0.05	0.43 ± 0.07	0.95 (0.87–1.04)	0.2767
TEF degree
Mild	2 (9.5)	4 (11.1)	Ref
Moderate	12 (57.1)	18 (50)	0.75 (0.09–4.50)	0.6407
Severe	7 (33.3)	14 (38.9)	1.00 (0.12–6.56)	0.8277
Opening pressure
Brisk	9 (42.9)	17 (47.2)	1.89 (0.07–51.68)	0.7546
Spontaneous	2 (9.5)	4 (11.1)	2.00 (0.06–73.68)	0.7414
Sluggish	1 (4.8)	1 (2.9)	Ref
Unknown/missing	9 (42.9)	14 (38.9)
Shunt	2 (9.5)	13 (36.1)	5.37 (1.27–37.12)	0.0279
Papilledema
Yes	5 (23.8)	5 (13.9)	0.27 (0.04–1.55)	0.1432
No	3 (14.3)	11 (30.6)	Ref
Unknown/missing	13 (61.9)	20 (55.6)

LM = leptomeningeal.

Values are shown as number (%) or mean ± SD unless indicated otherwise. Boldface type indicates statistical significance (p = 0.05).

^*ORs with 95% CIs and p values were determined with univariate logistic regression models to assess the effect of the risk factor on the FO plus splenium pattern.

^†The ORs for age and tumor volume were calculated per 1-unit increase.

^‡The OR for FOHR was calculated per 0.01-unit increase.

To explore the independent associations between the clinical and radiological variables and TEF pattern, we carried out a multivariate logistic regression model analysis of the FO plus splenium TEF pattern (Table 3). In the backward selection model that used a stay significance of 0.1, medulloblastoma remained significantly associated with the FO plus splenium TEF pattern (OR 4.75, 95% CI 1.43–18.53, p = 0.0157). There was a trend toward this pattern being associated with older age, but this was not statistically significant. No other variables listed in Table 2 reached the stay significance threshold to be considered in the multivariate analysis.

TABLE 3.

Multivariable logistic regression model of the FO plus splenium TEF pattern

Risk Factor (Ref)	OR (95% CI)	p Value*
Age (per 1-yr increase)	1.15 (0.99–1.36)	0.0718
Medulloblastoma pathology (pilocytic astrocytoma)	4.75 (1.43–18.53)	0.0157

^*Stay significance was defined as 0.1.

Patterns of TEF in Patients With Nonneoplastic Diagnoses

Nonneoplastic diagnoses associated with TEF were seen in 61 patients (28% of the entire cohort) and included aqueductal stenosis (n = 10), congenital ventriculomegaly (10), intraventricular hemorrhage (IVH)/posthemorrhagic hydrocephalus (PHH) (10), and postinfectious hydrocephalus (9) (Table 1, Supplementary Table 1). Patients with aqueductal stenosis had an average age of 10.1 years, all of whom had evidence of ventricular obstruction, and an average FOHR of 0.481; the FO and FO plus splenium TEF patterns were present in 70% of these patients. Outside of the FO locations, TEF was restricted to around the temporal horns and third ventricle. Two of 3 patients who underwent a funduscopic examination had evidence of papilledema. On the other hand, IVH/PHH patients were younger (mean age 2.3 years), all of whom lacked ventricular obstruction, and had an average FOHR of 0.415; 90% demonstrated the asymmetrical/circumferential TEF patterns. Outside of the FO locations, TEF was also observed in locations other than around the temporal horns and third ventricle. Patients with congenital ventriculomegaly were also young (average age 3.6 years), of whom 80% did not have evidence of ventricular obstruction, and had an average FOHR of 0.490; 90% demonstrated the asymmetrical/circumferential TEF patterns. None of the 2 patients who underwent funduscopic examination had papilledema. In each of these 3 groups, all patients who underwent neurosurgery at our institution underwent surgery for permanent CSF diversion.

Patients with postinfectious hydrocephalus had an intermediate age (mean age 6.8 years) and an average FOHR of 0.551, and ventricular obstruction was noted in approximately half the patients. The asymmetrical/circumferential TEF pattern was seen in 77.7% of patients, and TEF was again found in multiple locations around the ventricular system. In total, 87.5% of patients who underwent neurosurgery had a permanent CSF diversion procedure.

Discussion

TEF is a radiological finding present in patients with a range of neurological diseases, particularly those with disorders of CSF circulation, as seen in the large majority of our cohort who underwent surgery for CSF diversion (85% of all patients required CSF diversion surgery, of whom 67% required permanent CSF diversion). The classic association between ventricular size, intraventricular/intracranial pressure, and TEF may be present in the setting of acute hypertensive hydrocephalus,^4,12,13 but this scenario does not hold true for all cases of TEF. Our results, at minimum, suggest that TEF is heterogeneous and may be associated with disease pathology, CSF circulation, intracranial pressure, and ventricular size. Specifically, the TEF characteristics described in this study may improve our biological understanding of fluid handling by the ventricular ependymal surface and may provide a broader clinical picture of retained or lost global CSF circulation.

A large portion of our cohort demonstrated findings consistent with the classic understanding of acute obstructive hydrocephalus and propensity for TEF in that setting. For example, TEF degree correlated with FOHR. These relationships suggest that there is a fundamental biophysical relationship between ventricular size and degree of TEF, resulting in TEF in specific locations and possibly leading to the downstream effects of local tissue disruption. In other words, when enlarging ventricles reach a size threshold, there may be a tendency for CSF transmigration across the ependyma. CSF handling by the ventricular ependyma may be partially dependent on ventricular morphology, perhaps via pressure-sensing channels.¹⁴ We also noted that TEF was always present in the periventricular regions anterior to the frontal horns and posterior to the occipital horns, which may reflect lesser known circulation paths that occur in a rostral-caudal direction.^15,16 These local circulation paths differ from the bulk CSF flow pathways in that these may be directed toward specialized regions of ependyma to aid in physiological CSF signaling^17–19 and become overwhelmed when bulk flow pathways are obstructed. Patients who fit the classic acute hypertensive hydrocephalus picture (i.e., those with obstructive tumors or aqueductal stenosis) and had a greater degree of TEF, presence of papilledema, an FO or FO plus splenium TEF pattern, and higher opening pressure tended to retain their ability to self-regulate CSF circulation and not require permanent CSF diversion surgery if the source of the ventricular obstruction was removed (i.e., patients with obstructive tumors). Therefore, the FO and FO plus splenium TEF patterns seen in this setting may be a physiological response of healthy ependymal cells to impediment in bulk CSF flow, elevated intraventricular/intracranial pressure, and compensatory ventricular enlargement.

Overall, among all patients with symmetric TEF in the FO regions, we observed that approximately half also had TEF dorsal to the roof of the third ventricle near the splenium of the corpus callosum (FO plus splenium TEF pattern). To investigate the significance of this particular pattern among a relatively homogeneous patient population, we performed a subset analysis of patients with singular posterior fossa pilocytic astrocytomas or medulloblastomas. We found that medulloblastoma diagnosis was the only variable in multivariate analysis that was independently associated with the FO plus splenium TEF pattern. The univariate analysis, however, may have also provided interesting insights, such as the association between the presence of leptomeningeal disease at diagnosis and requirement for shunt surgery in patients with the FO plus splenium TEF pattern. Indeed, leptomeningeal disease is known to be associated with hydrocephalus requiring permanent CSF diversion.²⁰ Our results suggest that the FO plus splenium TEF pattern may be indicative of a nonobstructive hydrocephalus component and may help identify patients more likely to need permanent CSF diversion.

Using multivariate regression analysis, we found that tumor type was associated with the TEF pattern. In addition, the FO plus splenium TEF pattern, as a previously unrecognized TEF pattern, may be related to the underlying tumor biology that permits ependymal CSF transmigration. This may have important clinical implications in the pathogenesis of medulloblastoma. Although we could not assess the tumor growth rate and propensity for TEF or specific TEF patterns, we found that smaller tumor volume at diagnosis was associated with the FO plus splenium TEF pattern, while ventricular size was not. This could indicate more rapid medulloblastoma growth that may result in a specific CSF transgression dorsal to the third ventricle. This analysis, however, was limited to those pathologies in posterior fossa locations because of the otherwise relatively small number of patients with each tumor category and tumor location. Nevertheless, our results show that the FO plus splenium TEF pattern, in addition to being a previously unrecognized TEF pattern, may be related to underlying tumor biology that may have important clinical implications.

Although fewer patients without neoplasms were included in this imaging cohort, we drew several potentially important conclusions from this group. First, because we evaluated all individuals at our institution with an imaging finding of TEF over a 10-year interval, we found that nonneoplastic associations of TEF accounted for a minority of patients with TEF. We saw that some patterns fit a nonclassic picture, in which factors independent of ventricular obstruction alone may have governed CSF movement across the ventricular surface. For instance, patients with congenital or neonatal onset of CSF disorders, such as congenital ventriculomegaly, myelomeningocele, and IVH/PHH, had varying degrees of age of presentation, ventricular obstruction, average ventricular sizes, TEF patterns, and TEF locations.

Our study adds to the observation that TEF is predominantly associated with ventricular obstruction and that TEF is a predictor of need for permanent CSF diversion.²¹ In our patient cohort, we observed that patients with obstructive etiologies tended to have FO and FO plus splenium TEF patterns, which were less likely in patients with asymmetrical/circumferential TEF patterns who underwent permanent CSF diversion, thereby revealing that TEF patterns may provide useful clinical information related to disease etiology and CSF regulation. Importantly, the asymmetrical/circumferential TEF patterns were defined by some degree of ependymal asymmetry or generalized ependymal abnormality that may indicate impaired ability to regulate CSF circulation. Thus, these TEF patterns may reflect region-specific variations in ependymal fluid-handling ability leading to a permanent loss in the regulation of global CSF circulation. Indeed, abnormal ependymal cell development and function have been implicated, particularly in the pathophysiology of congenital hydrocephalus by human and animal studies.^22–24

It should be noted that the FO and FO plus splenium TEF patterns were typically seen in older patients (mean approximately 9 years old in the neoplasm and aqueductal stenosis groups), which contrasts the circumferential TEF pattern seen in younger patients with diagnoses such as PHH, congenital ventriculomegaly, or arachnoid cyst (mean age approximately 3 years). These age-associated differences in TEF patterns may reflect the brain’s response to hydrocephalus after normal ependymal development in older patients with obstructive etiologies and abnormal ependymal development in younger patients with nonobstructive etiologies. The fact that TEF patterns were common in patients with a range of congenital or neonatal conditions may suggest that early brain injury/developmental factors significantly and irreversibly alter ependymal fluid handling, thereby leading to global CSF dysregulation, but this mechanism warrants further investigation.

This is the first study, to our knowledge, that broadly defined the TEF features associated with neurological disease and described the clinical significance of these features as they relate to CSF circulation and pathology. However, there were several limitations to this study, including qualitative/subjective grading of TEF characteristics and missing data regarding papilledema and opening pressures. Another factor that may be important to the appearance of TEF is the time course of disease presentation. Patients with new tumor diagnoses may present later than patients with prolonged history of shunted hydrocephalus. However, because it is not universally clear when symptom/disease onset begins and baseline imaging is obtained only at presentation (when symptoms become significant, not when symptoms begin), it is impossible to relate timing/acuity of presentation with TEF characteristics. This study was also limited by the small sample sizes of each diagnosis group, which precluded definitive conclusions based on many disease subtypes. Future directions include the inclusion of larger sample sizes per disease group; utilization of image segmentation techniques and a standard atlas space to quantify and compare TEF patterns, degrees, and locations across pathologies; evaluation of cases with disorders of CSF flow without TEF; and determination of the cellular and biophysical mechanisms of TEF resolution over time.

Conclusions

In a large cohort of pediatric patients, we showed that TEF of CSF across the ventricular ependyma is region specific and seen most often anterior to the frontal horns and posterior to the occipital horns. Furthermore, there were distinct patterns of TEF independent of ventricle size and ventricular obstruction that suggest modulation of CSF TEF that is pathology and etiology specific. Further study is needed to understand the mechanisms underlying the interaction of CSF with the ventricular surface in various disease and developmental states.

ABBREVIATIONS

EVD

external ventricular drain

FO

fronto-occipital

FOHR

FO horn ratio

IVH

intraventricular hemorrhage

PHH

posthemorrhagic hydrocephalus

TEF

transependymal flow