Article Text
PDF
Abstract
Background and purpose Few studies have examined collateral formation in patients with cerebral venous thrombosis (CVT). The aim of this study was to analyse the impact of baseline intracranial venous collaterals on the clinical outcome and imaging features of patients with acute CVT.
Material and methods MRIs from consecutive patients with acute CVT were retrospectively analysed. The category system described by Qureshi was used to assess the pattern of venous collaterals. Clinical and imaging features and outcomes were analysed using bivariate and multivariate models to assess the association of collateral patterns with the type of parenchymal lesions and clinical outcome (modified Rankin Scale) at 30 and 90 days.
Results One hundred patients were included (77 women; median age 32 years; and median of 18 months of follow-up). Venous collaterals were present in 88% of the patients; type I collaterals in 3 patients, type II collaterals in 27 patients, and type III collaterals in 58 patients. Twelve patients did not exhibit any collaterals. Cohen's κ coefficient between evaluators was 0.86. In the bivariate analysis, type III collaterals were associated with isolated intracranial hypertension and complete recovery, whereas type I collaterals were associated with encephalopathy. However, in the multivariate regression analysis, the collateral pattern was not associated with clinical presentation, type of brain lesion or outcome.
Conclusions Intracranial venous collaterals are frequently found in patients with CVT during the acute phase. However, they do not have an independent effect on the type of brain damage, clinical manifestations or prognosis.
- STROKE
Statistics from Altmetric.com
Introduction
The underlying mechanism of symptomatic recovery in patients with cerebral venous thrombosis (CVT) remains poorly understood. It is conceivable that the role of intracranial venous collaterals is to prevent venous stasis, thrombus propagation, cerebral oedema, venous infarct and intracerebral haemorrhage. Parenchymal lesions are an unfavourable prognostic sign and will influence recovery.1–7 However, only a few studies have examined collateral formation in patients with CVT.5 ,8 ,9 Therefore, the aim of this study was to analyse the influence of the baseline intracranial collateral pattern on the presence and severity of parenchymal lesions (infarcts of haemorrhages) and on the clinical outcome, controlling for known prognostic factors.
Methods
This study was a retrospective analysis of data collected prospectively in our CVT registry. Consecutive patients with the first-ever confirmed CVT treated at the National Institute of Neurology and Neurosurgery in Mexico City were enrolled from November 2003 to October 2013. Inclusion criteria consisted of an acute phase CVT confirmed by MRI and three-dimensional (3D) gadolinium-enhanced MR venography (MRV) and complete follow-up as an outpatient for at least 90 days (for those who did not die during the acute phase).
All patients underwent a standardised diagnostic follow-up and treatment protocol. Patient data, including demographics, medical history, risk factors, CVT onset time, arrival time at the hospital, time to confirmation of CVT, complications, treatments and procedures, inhospital mortality and ambulatory status at discharge, were registered for each patient.
Clinical manifestations
The clinical presentation was categorised according to the main clinical signs and symptoms that led to the emergency department consultation, which included the isolated intracranial hypertension (iICH) syndrome, encephalopathy, focal syndrome or seizures at onset. The Glasgow Coma Score (GCS) at onset was dichotomised into <8 points or ≥9 points. All patients were treated with anticoagulation therapy, regardless of the clinical presentation, in line with standardised treatment procedures.10 ,11 The modified Rankin Scale (mRs) was used to assess functional outcomes at hospital discharge, as well as at 30 days and 3 months of follow-up. We classified the patients as exhibiting good (mRs of 0–2) or poor (mRs 3–6) functional outcomes; as a secondary outcome, we explored complete recovery (mRs 0–1). Remote seizures (defined as the presence of any type of seizure after the CVT event under the setting of lesional epilepsy) were also evaluated during the follow-up period.
Since this investigation was a purely observational study, based on current practice with no special investigations or procedures, it was exempt from informed consent. However, this study was approved by the Institutional Ethics Committee.
Imaging and assessment of the collateral formation
The patients underwent MRI and 3D-MRV within 24 h of the emergency consultation. The institutional MRI protocol for CVT includes T1-weighted and T2-weighted images in the sagittal, coronal and axial planes; fluid-attenuated inversion recovery images; T2* in gradient echo; diffusion-weighted and apparent diffusion coefficient images; and a 3D time of flight MRV. MRIs were obtained with a 1.5 T or 3.0 T GE Signa HDxt system (General Electric Company; Fairfield, Connecticut, USA).12
MRI and MRV images were evaluated jointly by a board-certified neuroradiologist (CM, with 6 years of neuroimaging review experience) and two stroke neurologists (AA, with 20 years of neuroimaging review experience, and MAB, with 5 years of neuroimaging review experience) on a high-resolution monitor. Except for the treatment, the readers were blinded to the clinical data. The MRIs were analysed first, followed by the MRV source images. CM and MAB separately analysed the MRV collateral patterns and then a comparison among the results was performed; AA analysed any patterns that lacked agreement between the first two evaluators.
Parenchymal lesions were classified as non-haemorrhagic lesions (venous infarctions) and haemorrhagic lesions (haemorrhagic infarctions, subarachnoid haemorrhage and intracranial haemorrhage). The lesion size (whether non-haemorrhagic or haemorrhagic) was evaluated through direct measurement of the largest diameter (centimetres) of the venous infarction or haemorrhage by tracing on the MRI that visualised the slice with the largest area of the parenchymal lesion.
Mixed-venous system thrombosis was classified as a combination of CVT of any superficial (superior sagittal sinus (SSS), lateral and sigmoid sinus and/or jugular vein) and deep (Rosenthal’s basal vein, inferior longitudinal sinus, straight sinus, or thalamostriate vein) venous system.
A three-component method proposed by Qureshi and coworkers8 ,13 was used to classify the collateral formation as follows: grade I, collaterals bypass the occluded segment of the dural venous sinus but connect within the same sinus; grade II, collaterals bypass the occluded segment but connect within a different sinus; and grade III, collaterals bypass the occluded segment and connect within a different circulation. This classification recognises the superficial and deep venous circulations as distinct pathways that can be linked through collateral pathways (see figure 1 for reference).
Types of collaterals according to Qureshi’s classification. (A) Grade I collaterals: sagittal MR venography (MRV) with thrombosis of the superior sagittal sinus (SSS) with a loop that bypasses the occluded segment and connects to a proximal segment within the same sinus (white arrows). (B) Grade II collaterals on a sagittal view of an MRV that connects the SSS with the transverse sinus (white arrows). (C) Thrombosis of the SSS on a sagittal view of an MRV with a grade III collateral that bypasses the occluded segment of the dural sinus and connects to a different circulation (white arrows).
Statistical analysis
Data analysis was performed using the statistical package SPSS (V.20.0, IBM Inc., Armonk, New York, USA). The following baseline variables were included: age, sex, clinical manifestations at admission, vascular risk factors and comorbidities. Continuous values are expressed as the means±SD or as the median and IQR. Nominal variables are expressed as counts and percentages. Cohen’s κ coefficients were calculated to evaluate the interobserver agreement for collateral score grading.
Parenchymal lesions (haemorrhagic and non-haemorrhagic) and clinical outcome (mRs) were established as dependent variables for the statistical comparison among the different collateral patterns. We performed bivariate (two-tailed χ2 statistics with Yates correction) and multivariate (logistic regression; p in 0.05, p out 0.1) analyses adjusted for potential modifiers such as gender, age, GCS, malignancy and central nervous system (CNS) infection. The fitness of the model was evaluated using the Hosmer-Lemeshow goodness-of-fit test, which was considered reliable if p>0.2. The results are presented as ORs with the 95% CI. A p value of <0.05 was considered statistically significant.
Results
In total, 126 patients with confirmed CVT were identified during the study period: 19 patients were excluded due to failure to follow-up at our clinic after discharge, and 7 were excluded due to incomplete radiological studies. A total of 100 patients were analysed (77 females and 23 males; median age of 32 years, IQR=23–41 years). The main risk factors found for CVT were as follows: pregnancy and puerperium (24%), oral contraceptives (17%), hereditary thrombophilia (known and newly diagnosed, 22%), hormonal replacement therapy (3%), malignancy (3%) and CNS infection (2%). The median number of days from first symptoms and signs until diagnosis was 8 (IQR=4–18 days). In 24 patients, the diagnosis was made 15 days after the first symptoms; 45.8% of these patients presented with iICH, and 30% were men. No significance was found between the time of diagnosis and the development of collateral vessels (OR=0.94, 95%CI=0.23 to 3.79, p=0.93) or the individual collateral patterns. Table 1 describes the main demographic data of the patients.
Clinical characteristics of the participants
Clinical features
Clinical conditions at onset are described in table 1. Focal syndrome (69%) and seizures at onset (46%) were the most frequent clinical findings at the onset of CVT; 10% of patients presented a GCS<8 points. Good outcomes (mRs 0–2) at 30 and 90 days were observed in 81% (70 patients with any type of collateral and 11 with none) and 88% (77 patients with any type of collateral and 11 with none) of patients, respectively; three patients died during the acute phase (30 day): one patient with type I collaterals, one patient with type III collaterals and one patient without collaterals. Complete recovery (mRs 0–1) was found in 51% and 72% of patients at 30 and 90 days, respectively.
Patients with any type of collateral had a median time (from first signs/symptoms until medical evaluation) of 8 days (IQR 4–18 days), compared with 9.5 days (IQR 4–17) for patients without collaterals (p=0.046). The median follow-up time was 18 (IQR 9–39) months.
Imaging features
In total, 88 patients presented with some type of collateral pattern. The type III pattern was the most commonly identified pattern (58 patients); in 41 patients (70.7%) with this type of collateral, thrombosis was localised in the SSS. The highest proportion of haemorrhagic transformation from a venous infarction was present in 12 (44.4%) patients who presented a type II collateral pattern. Table 2 describes the main radiological patterns of lesion and venous drainage systems affected. The evaluators demonstrated good agreement in terms of collateral patterns (Cohen’s κ=0.86 among the two evaluators (CM and MAB)); for those cases without agreement, a third evaluator (AA) assessed the study and defined the appropriate collateral pattern.
Cerebral venous thrombosis collateral patterns and imaging features
Statistical comparisons
The first model (from the two-tailed χ2 bivariate analysis of the presence of any type of collateral and the presence and severity of parenchymal lesions) revealed no significance for any of the variables evaluated. A second step in this model was performed using the same analysis but on every independent collateral pattern.
In the clinical profile at onset analysis, encephalopathy (OR=22.2, 95% CI=1.81 to 27.2, p=0.001) showed significant association with the type I collateral pattern. iICH (OR=3.22, 95% CI=1.27 to 8.15, p=0.01) and focal syndrome (OR=0.36, 95% CI=0.14 to 0.91, p=0.03) showed significant associations with type III collateral patterns. From the follow-up outcome variables, mRs=3–6 points at the 30 day (OR=9.41, 95% CI=0.81 to 10.9, p=0.03) and 90 day follow-up (OR=17.4, 95% CI=1.44 to 20.9, p=0.003) were significantly associated with type I collaterals. As a secondary outcome, complete recovery (mRs 0–1) showed significant association with the type III collateral pattern (OR=0.34, 95% CI=0.15 to 0.77, p=0.009). No other significant associations were found among the variables and collateral patterns evaluated (table 3).
Bivariate analysis of collateral patterns versus clinical, imaging and outcome variables
Variables with significant associations were included in a multivariate regression model, adjusted for age, gender, malignancy, CNS infection, and GCS<8. None of these variables (type of collateral pattern) showed significance as independent predictors when compared with the presence of parenchymal lesions (haemorrhagic or non-haemorrhagic) and clinical outcome for complete recovery (mRs 0–1), good outcome (mRs 0–2) and bad outcome (mRs 3–6) in the multivariate analysis (Hosmer-Lemeshow test for goodness of fit in the final step of the regression model: x2=3.411, 4 df, p=0.49, adjusted for age, gender, malignancy, CNS infection, and GCS<8 for type I pattern vs encephalopathy, and mRs at 30 and 90 days; and Hosmer-Lemeshow: x2=1.78, 3 df, p=0.62, also adjusted for age, gender, malignancy, CNS infection, and GCS<8 for type III pattern vs ICH and focal syndrome).
Discussion
The utility of intracranial venous collaterals in patients with CVT has been poorly studied. A previous study8 in a smaller sample of patients reported that 38% of patients experienced collateral formation. In our study, according to Qureshi’s classification system,8 ,12 collaterals were present in 88% of patients. To the best of our knowledge, this is the largest sample of patients with CVT who have undergone collateral vessel assessment to determine its relationship with outcome. Our study evaluated the main features that have been associated with the collateral pattern, including the following: brain lesion size, haemorrhagic transformation from venous infarction, brain oedema and iICH.6 ,7 All patients underwent a standardised protocol and follow-up, and image quality was optimal for evaluation of collateral patterns. The interobserver evaluation agreement was good; therefore, analysis of the data according to the type of collateral led to an optimal association with data from variables included in the statistical model.
The most common collateral pattern found was type III (65.7% of patients). This pattern tends to send collateral vessels to a separate circulation system, bypassing the occluded segment. We found that this pattern was associated with SSS occlusion together with bilateral transverse lateral thrombosis, which could explain why alternate drainage circulation outside the dural sinus was present. In the bivariate analysis, the type III collateral pattern was associated with ICH (OR 3.2, 95% CI 1.27 to 8.15), whereas the type I collateral pattern was associated with encephalopathy (OR 22.2, 95% CI 1.81 to 27.2). The type I collateral pattern was also associated with a poor prognosis (mRs 3–6), and we failed to find any associations with venous collaterals and parenchymal lesions. However, in the multivariate regression analysis, the collateral pattern was not associated with the clinical presentation, type of brain lesion or outcome. These data indicate that the majority of patients with acute CVT exhibit some type of collateral during the acute phase, which has a poor impact on the clinical variables, including remote seizures in the follow-up.
The poor prognosis in the patients who lacked collateral vessels could be explained by the extent of brain parenchymal damage, local oedema, and venous congestion secondary to obstruction, and failure to achieve decreased pressure through collateral vessels could lead to the haemorrhagic transformation of a venous infarction.5 ,6 Our reverse model failed to identify an association between the absence of venous collaterals and the type of brain lesion; therefore, we believe that stronger predictors could allow the identification of individuals with a poor prognosis, and the collateral pathway could be part of a continuum in secondary lesions that aggravate other factors, such as the site and extension of thrombosis, malignancy and lesion size.14 ,15 Further studies are needed to investigate whether early assessment of venous collateral efficiency can help to choose the optimal aggressiveness of treatments. In particular, the possibility of identifying the patients at higher risk of an unfavourable outcome could contribute to identifying those who might benefit the most from endovascular procedures. Further studies are also necessary to confirm the possible role of the extension, localisation and aetiology of CVT in hindering the activation of collateral pathways and indirectly influencing the clinical evolution of the disease.
The time required for collateral formation is unknown. In our study, patients with no collaterals showed a significantly lower mean time from the first signs/symptoms to disease evolution, which could explain the absence of collateral drainage because time seems to correlate with the likelihood of developing these alternative pathways in venous vessels.
Certain limitations of the present study need to be acknowledged. Although we used a prospectively collected cohort, the impact of baseline collateralisation on early clinical outcome was determined in a retrospective manner. Hence, the study is subject to the drawbacks of an observational design. Our model was intended to identify an association (whether positive or negative) between the absence or presence of any type of collateral, as well as the various collateral patterns, and outcome, which should be considered to be a mixed model. The optimal approach would be a prospective study to determine a good prognostic association with the presence of collaterals and various outcome variables using a long-term follow-up model. We did not analyse recanalisation, as we intended to analyse collateral vessels as an independent factor. However, it appears that recanalisation predominantly occurs within the first few months because in two studies in which multiple follow-up MRI were performed, no difference in the recanalisation rate at 3 months or 1 year was found.16 ,17 It is unknown whether MRI-visualised recanalisation is associated with the probability of recurrence, and no correlation has been found between the degree of recanalisation and clinical outcome.18 ,19 Although our results do not demonstrate a clear cause-and-effect relationship, they provide an understanding of the haemodynamic behaviour of venous vessels in CVT, and research models should therefore be included to assess this factor in future prospective studies.
In conclusion, intracranial venous collaterals are frequently found in patients with CVT during the acute phase. However, this factor does not seem to have an impact on the type of brain damage, clinical manifestations or prognosis.
References
Footnotes
Twitter Follow Carolina Mejias at @caroms8133
Contributors MAB, AA, CM, AQ-C and JC-L were involved in the design of the study. MAB, JC-L and AQ-C collected the data and bibliographic sources. MAB, CM, JC-L and AA evaluated the radiological studies included in this study. MAB and AA were involved in the statistical analysis, and MAB drafted the manuscript. All authors approved the final version of this manuscript.
Competing interests None.
Ethics approval Ethics Committee of the Instituto Nacional de Neurologia.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional data are available.