Patients and methods

This was a retrospective 20-year cohort study conducted in a major referral hospital covering 14 provinces in southern Thailand. The inclusion criteria were as follows: 1) 18 years old or older, 2) admitted in the inpatient department, and 3) primary neuroimaging available for review (contrast-enhanced magnetic resonance imaging [MRI] of the brain, magnetic resonance venography [MRV] of the brain, or cerebral angiography). Although the CT venography of the brain is considered as a reliable method for CVT diagnosis, the visualization of venous collateral is problematic owing to the limited resolution of the CT venography. Therefore, the CT venography was not set in the inclusion criteria.

Inpatient records of the patients diagnosed with spontaneous CVT between January 2002 and December 2022 were screened using the International Classification of Diseases, 10th revision (ICD-10), using the following codes: G08 (intracranial and intraspinal phlebitis and thrombophlebitis), I636 (cerebral infarction due to CVT), I676 (non-pyogenic thrombosis of the intracranial venous system), and O225 (CVT in pregnancy). Cases with wrong diagnoses, incomplete medical records, or lacking primary neuroimaging data were excluded. Patients with cavernous sinus thrombosis were also excluded.

Data collection

Baseline characteristics, clinical presentations, initial functional score, neuroimaging characteristics, treatment, and clinical outcome were collected from the electronic medical record. Venous collaterals were independently reviewed by two neuroradiologists. The interrater reliability of venous collateral grading was calculated by using percentage agreement. Any discordant radiographic interpretation was discussed until a consensus was reached. The relevant data were accessed for research purposes by the authorized research team members between May 1^st, 2023 and June 30^th, 2023. The personal identity data of the individual participants in the research was protected and cannot be accessed according to the Personal Data Protection Act (2019).

Definition of terms

To unify ambiguous terms that are often used in the literature, we objectively classified clinical presentations into four distinctive neurological syndromes as follows: 1) seizure (any type of seizure according to the International League Against Epilepsy (ILAE) 2017 Classification [5]) 2) focal neurological deficits (any abnormal focal neurological deficits), 3) isolated increased intracranial pressure (the collective clinical symptoms of headache, blurred vision, with or without papilledema and objective evidence of increased intracranial pressure proven by the CSF pressure over 20 cm H2O without any focal neurological deficits), and. 4)isolated headache (headache without abnormal neurologic signs and the CSF pressure under 20 cm H2O). Venous collaterals were classified according to the Qureshi classification [6] as follows: in grade 1, the collateral bypassed the occluded segment and connected it within the same dural sinus; in grade 2, the collateral bypassed the occluded segment and connected it to a different dural sinus in the same circulatory pathway; and in grade 3, the collateral bypassed the occluded segment and connected it to a different sinus in a different circulatory pathway. If multiple grades were identified, the highest grade was selected (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Brain imaging showing venous collaterals according to the Qureshi classification.

(A) Venous collateral grade 1. Sagittal magnetic resonance venography (MRV) of the brain demonstrated thrombosis of the superior sagittal sinus (asterisk) and the venous collateral bypassing the occlusion site and connecting with the distal part of the same sinus (arrowhead). (B) Venous collateral grade 2. Coronal cerebral angiography revealed extensive thrombosis from the distal superior sagittal sinus to the left proximal transverse sinus (asterisk) and the venous collateral connecting the middle superior sagittal sinus with the transverse-sigmoid sinus junction (arrowhead). (C) Venous collateral grade 3. Gadolinium-enhanced sagittal magnetic resonance imaging of the brain demonstrated thrombosis of the great vein of Galen and straight sinus (asterisk) and venous collateral connecting the inferior sagittal sinus with the distal superior sagittal sinus (arrowhead). (D) MRV of the brain revealed thrombosis of the superior sagittal sinus (asterisk) and two venous collaterals as follows: (1) venous collateral grade 2, connecting the proximal superior sagittal sinus connecting with the Sylvian vein (arrow) and (2) venous collateral grade 3, connecting the proximal superior sagittal sinus with the inferior sagittal sinus. The patient with multiple venous collaterals was graded according to the highest grade of the Qureshi classification.

https://doi.org/10.1371/journal.pone.0302162.g001

The modified Rankin scale (mRS) was used to evaluate the functional status upon initial admission and discharge. In addition, the hospital clinical outcome was categorized into non-dependency (mRS 0–2) and dependency or death (mRS 3–6). Patients who died of an intracranial condition, especially brain herniation, were marked as CVT-related deaths; all other causes were noted as non-CVT-related.

Data analysis

Discrete data are described as numbers and percentages. Continuous data are presented as either mean ± standard deviation (SD) or median and interquartile range, depending on normality. The data were compared between patients with and without venous collaterals using the univariate chi-squared test, Mann–Whitney U test, or independent t-test. Factors with P-values <0.05 in the univariate analysis were recruited into the multivariate logistic regression analysis. Factors with P-values <0.05 in the multivariate logistic regression were determined as independently associated factors. This research study was approved by the Ethics Committee of the Faculty of Medicine, Prince of Songkla University.

Comparison of baseline characteristics, clinical presentations, and radiographic findings

Patients with venous collaterals were significantly younger than those without (37.0±13.9 years vs. 44.9±17.4 years, t = -2.041, P = 0.048). There was no significant difference in terms of sex (with and without collaterals: 60.0% female vs. 64.4% female, respectively, P = 0.724) or smoking exposure (5.0% vs. 5.1%, P = 0.976) between groups.

No significant differences were found between groups in terms of clinical presentations (with and without collaterals: focal neurological deficit, 60.0% vs. 37.3%, P = 0.076; isolated increased intracranial pressure, 30.0% vs. 17.0%, P = 0.209; isolated headache, 30.0% vs. 42.4%, P = 0.327; seizure, 45.0% vs. 40.7%, P = 0.735). The clinical severity at admission was also comparable between groups (with and without collaterals: mRS 3–5, 45.0% vs. 37.3%, P = 0.542).

To confirm the CVT diagnosis and identify venous collaterals, at least one gold-standard neuroimaging method (either brain MRI or MRV) was performed in all patients. The number of venous thrombosis sites was quantified to evaluate the clot burden. However, there was no significant difference between groups in terms of patients with more than three thrombosis sites (with and without collaterals: 50.0% vs. 30.5%, P = 0.115). Patients with venous collaterals had a significantly higher prevalence of CVT in the superior sagittal sinus than those without venous collaterals (80.0% vs. 54.2%, P = 0.041). The other CVT sites were comparable between groups (with and without collaterals: superficial cortical vein, 30.0% vs. 27.1%, P = 0.804; transverse sinus, 60.0% vs. 49.2%, P = 0.401; deep cerebral vein, 5.0% vs. 5.1%, P = 0.988; sigmoid sinus, 40.0% vs. 45.8%, P = 0.654; straight sinus, 10.0% vs. 5.1%, P = 0.435; and jugular vein, 25.0% vs. 20.3%, P = 0.661). The prevalence of intracranial hemorrhage was also similar between groups (50.0% vs. 35.6%, P = 0.254) (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Comparison of the baseline characteristics and radiographic findings between patients with cerebral venous thrombosis with and without venous collaterals.

https://doi.org/10.1371/journal.pone.0302162.t001

Characteristics of venous collaterals

The venous collateral grading was independently assessed by 2 neuroradiologists. The interrater reliability calculated by percentage agreement was 84.8%. The 12 discordant venous collateral findings were concluded by consensus. In this cohort, the venous collaterals were found in 20 patients (25.3%) according to the Qureshi classification [6]. Grade-3 collaterals were the most common (11 patients), followed by grades 1 (5 patients) and 2 (4 patients).

Comparison of etiology, treatment, and hospital outcome

Hormonal exposure (35.0%) was the most common cause of CVT in patients with venous collaterals, while idiopathic CVT (32.2%) was the most common in patients without venous collaterals. Patients with venous collaterals had a significantly higher prevalence of hormonal exposure than those without (35.0% vs. 6.8%, P = 0.002). However, no significant differences were found between groups in terms of infection (with and without collaterals: 5.0% vs. 6.8%, P = 0.778), cancer (10.0% vs. 25.4%, P = 0.147), pregnancy and puerperium (10.0% vs. 1.7%, P = 0.093), hypercoagulable state (20.0% vs. 17.0%, P = 0.757), miscellaneous etiologies (10.0% vs. 27.1%, P = 0.115), and idiopathic CVT (30.0% vs. 32.2%, P = 0.855).

Most patients in both groups underwent anticoagulant therapy (either unfractionated or low-molecular-weight heparin). The numbers of patients treated with anticoagulants (with and without collaterals: 95.0% vs. 89.8%, P = 0.482) and decompressive surgery (with and without collaterals: 10.0% vs. 6.8%, P = 0.638) were comparable between groups. There was no difference in poor outcome (mRS 3–6) at discharge between groups (with and without collaterals: 60.0% vs. 71.2%, P = 0.353). Of the 79 patients in this cohort, The overall mortality in the study cohort was 6.3%. Furthermore, the numbers of CVT-related deaths were similar between groups (100.0% vs. 75.0%, P = 0.439) (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Comparison of the etiology, treatment, and outcome between patients with cerebral venous thrombosis with and without venous collaterals.

https://doi.org/10.1371/journal.pone.0302162.t002

Factors associated with venous collaterals in patients with CVT

Significant factors in the univariate analysis included age (with vs. without venous collaterals: 37.0±13.9 years old vs. 44.9±17.4 years old, P = 0.048), occlusion in the superior sagittal sinus (80.0% vs. 54.2%, P = 0.041), and hormonal exposure (35.0% vs. 6.8%, P = 0.002). After including these factors in the final multivariate logistic regression analysis model, hormonal exposure (adjusted odds ratio [aOR] 7.276, 95% confidence interval [CI] 1.606–32.966, P = 0.010) and occlusion in the superior sagittal sinus (aOR 3.581, 95% CI 1.941–13.626, P = 0.044) were identified as independently associated factors for venous collaterals in CVT (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multivariate analysis of factors associated with venous collaterals in patients with cerebral venous thrombosis.

https://doi.org/10.1371/journal.pone.0302162.t003

Prevalence and characteristics of venous collaterals and limitations of collateral evaluation

As venous collaterals were not assessed in previous studies, the available data regarding their prevalence in patients with CVT is limited. Our study revealed the prevalence of venous collaterals in this population to be 25.3%, which was low compared to previous reports ranging between 66% and 89% [1, 3, 4].

Among these studies, different methods were used to quantify and classify the existence of venous collaterals, which led to difficulty in data interpretation. Sheth et al. [1] used the venous collateral scale (VCS) to grade patients in an American population in which cortical venous draining without anastomosis to a patent sinus was graded as VCS 1, while our study and Barboza et al. [3] used the Qureshi classification, which does not consider non-anastomotic veins. Therefore, the prevalence of venous collaterals might have been overestimated by Sheth et al. [1]. In addition, Zhao et al. [4] categorized venous collaterals into one of two groups based on the number of well-visualized collateral pathways: poor collaterals (0–1 well-visualized pathway) and good collaterals (2 or more well-visualized pathways). The number of patients with poor collaterals might have been overestimated by Zhao et al. [4], as patients without any visualized venous collaterals were included in this group. Barboza et al. [3] used the Qureshi classification, but we determined the prevalence of venous collaterals to be one-third of that reported by them.

Although Qureshi grade 3 was predominant in our study and that by Barboza et al. [3], we found almost eight times as many Qureshi grade 1 collaterals. However, caution is recommended when interpreting these findings due to the small sample sizes in both studies.

Some limitations of the Qureshi classification should be mentioned. Although the Qureshi classification is practical for acute or subacute CVT, using this classification in chronic CVT is problematic. The presence of multiple venous collaterals in chronic CVT generates a chaotic venous network in MRI and MRV. Therefore, they might be difficult to accurately grade, and some individuals with chronic CVT might have features of all three grades. To solve this problem, we propose defining multiple venous networks as grade 4 in the Qureshi classification, which should be validated in future research. Another issue is that the evaluation of venous collaterals by static neuroimaging methods (MRI and MRV) does not demonstrate the venous flow direction, affirming the collateral drainage. Recently, novel MRI techniques have been introduced to assess cerebral venous hemodynamics. Venous blood flow, velocity, and pulsatility can be systematically measured by several techniques, including phase-contrast MRI at 7 T and magnetic resonance velocity mapping. However, these parameters have not been validated for CVT [7]. In a recent study, Dempfle et al. [8] quantified the venous volume using Statistical Parametric Mapping 8 and Advanced Normalization Tools software in susceptibility-weighted imaging. Although the measurements produced by this method were reproducible and reliable, their clinical implications should be studied.

Hemodynamic assessment of the cerebral venous system by transcranial color-coded duplex sonography (TCCS) could demonstrate venous flow [9]. TCCS identified four distinctive pathologic patterns of venous drainage in the setting of CVT: 1) drainage diversion to the cavernous sinus and the deep cerebral vein, 2) drainage reversal in the basal veins, 3) compensatory flow to the transverse sinus, and 4) flow reversal in the transverse sinus [2]. Although normalization of the initially pathologic venous TCCS was a predictor of a favorable outcome, the clinical implications of this parameter need to be studied further considering the small sample size and the use of an operator-dependent tool.

Association between the superior sagittal sinus and venous collaterals

Our study is the first to identify an association between the superior sagittal sinus and venous collaterals in CVT. This association could be explained in terms of anatomy. Three parts of the superior sagittal sinus receive venous drainage from different structures. The anterior third, in which the border is the coronal suture, receives venous drainage from the central veins of the prefrontal medial aspect of the brain. The middle third of the superior sagittal sinus is the main drainage site of the superficial cortical vein from the dorsolateral prefrontal region of the brain, in combination with drainage from the anterior third. The posterior third receives venous blood from the straight sinus [10]. As the superior sagittal sinus is the longest venous channel to receive drainage from most of the superficial cortical veins, occlusion of the superior sagittal sinus directly increases the cerebral venous volume and pressure in the superficial cortical veins. Apart from the superior sagittal sinus and the Trolard vein, the superficial cortical vein also drains into two other interconnecting channels: the inferior cerebral vein (Labbe vein) to the transverse sinus and the superficial Sylvian vein to the cavernous sinus [10]. Previous studies of venous collaterals focused mainly on visualizing these interconnecting drainages [2, 4].

Apart from the classic major venous channel, other venous collaterals have occasionally been reported. In cases of a meningioma invading the superior sagittal sinus, the diploic veins, which connect the inner layer with the outer layer of the cortical bone, were illustrated as another venous collateral system in cerebral venous occlusion [11]. These diploic veins could be anatomically normal structures or compensatory venous collaterals. This question was indirectly answered via neurosurgery. The subsequent brain edema and venous congestion in patients who had the diploic veins removed during craniotomy emphasized their venous collateral role. Diploic veins are classified into four groups according to their anatomic course through the skull: 1) frontal diploic veins, which mainly drain to the anterior temporal vein; 2) anterior temporal diploic veins, which drain to the middle meningeal vein, deep temporal vein, cavernous sinus, and sphenoparietal sinus; 3) posterior temporal diploic veins, which drain to either the mastoid emissary vein or transverse sinus; and 4) occipital diploic veins, which drain to either the occipital vein or transverse sinus [12].

Besides the diploic veins, the falcine venous plexus may play a role in venous drainage. The falcine sinus is formed and functions in the fetal period. The venous drainage function of the falcine sinus gradually subsides in childhood and adulthood but remains as a remnant structure. However, recent studies in healthy adult controls using either electron microscopy or MRI discovered the function of the falcine venous plexus, which connects the superior sagittal sinus with the deep venous system. The falcine venous plexus of the posterior third of the superior sagittal sinus is a consistently dense network communicating with the inferior sagittal sinus, while those of the anterior and middle thirds are variable [13, 14]. The clinical significance of the falcine and parafalcine venous plexi has been demonstrated in superior sagittal sinus-invading meningiomas [15]. According to the anatomy of venous channels, we hypothesized that some venous structures may not be visualized under normal conditions, as the main drainage pathway could function normally. However, the venous collateral can be visualized when it starts to drain venous congestion. The superior sagittal sinus is the longest venous channel and is therefore the path of least resistance compared to a more complicated interconnecting pathway, which may support the venous collateral.

Association between hormonal exposure and venous collaterals

Our study is the first to demonstrate an association between hormonal exposure and venous collaterals in CVT. The major source of hormonal exposure in patients with CVT is oral contraception, which mainly contains estrogen and its derivatives. The association between estrogen exposure and venous thrombotic risk has been demonstrated in clinical studies and molecular research [16]. Besides venous thromboembolism, estrogen also affects angiogenesis by mediating several chemokines, including nitric oxide and vascular endothelial growth factor, to stimulate vascular endothelial cell proliferation and migration [17]. Although the relationship between estrogen and angiogenesis was studied mainly with reference to arterial vasculature [18], the molecular mechanism of estrogen-directed angiogenesis via estrogen receptors (ERs) expressed on the vascular endothelium should be similar between arteries and veins [19].

Currently, three ER isoforms have been identified: ERα, ERβ, and G protein-coupled ER (GPER). There are no direct studies of ERs in CVT; however, a study on chronic venous disease confirmed significant increases in ERα, ERβ, and GPER expression in the venous endothelium according to disease severity [20]. At the cellular level, ER activation reduces calcium-dependent venous contraction. In addition, ER agonists facilitate venous relaxation [21]. We postulate that exposure to estrogen activates these two effects, leading to enhanced distensibility in veins, and that estrogen may play a role in venous collaterals in CVT.

Strengths, limitations, and suggestions

The heterogeneous clinical symptoms and signs of CVT in previous studies are major drawbacks in comparing and interpreting results. We addressed this problem by categorizing the clinical presentations into clearly defined clinical syndromes. Another strength of this study was the use of reliable radiological data, which were independently validated by two neuroradiologists with the good interrater reliability. However, there are some limitations. First, the neuroimaging technic in this cohort was not uniform which might provide discrepant findings between the different method. Second, the radiological data in this study were derived mainly from either contrast-enhanced MRI or MRV of the brain, which are static neuroimaging methods and cannot demonstrate the direction of collateral flow. Assessing cerebral venous flow diversion via dynamic neuroimaging methods (like cerebral angiography, advanced MRI techniques, or TCCS) may be an area of research in the future. Another issue was the retrospective nature of this study, as some data was not retrievable; however, we tried to control for this by selecting only patients with complete data, resulting in a small sample size.

Our study had expanded the limited current knowledge of venous collaterals in CVT. We suggested the clinicians and neuroradiologists to meticulously searching for venous collaterals in particularly the CVT patient who had superior sagittal sinus thrombosis or hormonal exposure. However, our experience found a difficulty to identify venous collateral accurately from the CT venography due to it’s limited image resolution. We suggested to use contrast-enhanced MRI or MRV as they provide the better visualization of venous collateral. Future research should focus on validating the Qureshi classification of venous collaterals in terms of clinical outcome, the association with secondary intracranial hemorrhage, and the correlation with recanalization.