Demographics
A total of 119 eligible patients with a first-time diagnosis of CVST in our institution were enrolled consecutively. After initial assessment, 32 patients were excluded. Among them, 19 patients failed to get CSF opening pressure or refused lumbar puncture after admission; 2 patients were with contraindications for lumbar puncture due to severe brain herniation including cerebral ventricular compression and a midline shift; 10 patients were with non-thrombotic cerebral venous sinus stenosis (CVSS) by giant arachnoid granulations. A flow chart of the detailed enrollment process was showed in Fig. 2. Accordingly, 87 patients (50 males and 37 females) were included into the final analysis. The average age was 39.03 ± 13.08 years (range, 15–66 years). The median onset-to-door time was 3.50 (IQR, 1–11) months. The average BMI was 25.19 ± 3.84 kg/m2. Headache (88.51%) was the most common symptom, followed by nausea or vomiting (63.22%), visual impairment (33.33%), tinnitus (26.44%), dizziness (26.44%), double vision (14.94%), details in Table 1.
Table 1
Clinical features of CVST
Clinical features
|
|
Demographics
|
|
No. of patients
|
87
|
Gender (male/female)
|
50/37
|
Age (years) (Mean ± SD)
|
39.03 ± 13.08
|
Onset-to-door time (months) (Median, IQR)
|
3.5 (IQR, 1–11)
|
BMI (kg/m2)
|
25.19 ± 3.84
|
Clinical symptoms (No., %) n = 87
|
|
Headache
|
77 (88.51)
|
Nausea or vomiting
|
55 (63.22)
|
Visual impairment
|
29 (33.33)
|
Tinnitus
|
23 (26.44)
|
Dizziness
|
23 (26.44)
|
Double vision
|
13 (14.94)
|
Abbreviations: CVST, cerebral venous sinus thrombosis; IQR, inter quartile range; SD, standard deviation; BMI, body mass index. |
Imaging Presentations
Among the 87 patients, 17 (19.54%) had concomitant cerebral infarction by CVST but without imaging findings of severe brain herniation such as obvious cerebral ventricular compression and a midline shift. The anatomic sites of the thrombi distribution were as follows: superior sagittal sinus (60.92%), straight sinus (26.44%), torcular herophili (21.26%), right transverse sinus (80.46%), right sigmoid sinus (63.22%), right jugular vein (58.05%), left transverse sinus (83.33%), left sigmoid sinus (47.13%) and left jugular vein (40.80%).
Correlation Between Cvst-score And Icp
There was a high reliability between the two readers’ scores. The ICC for the observers was 0.99 (95% CI 0.98 to 0.99) with the novel CVST-Score, 0.93 (95% CI 0.90 to 0.96) with the CVES method, and 0.91 (95% CI 0.86 to 0.94) with the CVOS method. Due to the measuring range of the CSF opening-pressure manometer, the ICP > 330mmH2O could not be obtained. So the patients were divided into three subgroups by the range of ICP being <250mmH2O, 250-330mmH2O and >330mmH2O, and CVST-Score of each subgroup was 4.29 ± 2.87, 11.36 ± 3.86 and 14.99 ± 3.15, respectively (p < 0.001). Details in Fig. 3 (panel a). Correspondingly, the score of each subgroup was 2.72 ± 0.92, 4.10 ± 1.27 and 4.98 ± 1.07 with the CVES method (p < 0.001). And the score of each subgroup was 0 (IQR, 0-0.5), 1 (IQR, 0-1.25) and 2 (IQR, 0-2.25) with the CVOS method (p < 0.001). Details in Fig. S1 (panel a) and Fig. S2 (panel a). For patients with ICP being ≤ 330mmH2O, CVST-Score was linearly and positively correlated with ICP. The univariate linear regression analysis was as follows: yICP = 8.44×xCVST-Score + 171.77 (R2 = 0.53, p < 0.001). Scatterplot of CVST-Score vs. ICP was shown in Fig. 3 (panel b).
We divided the patients into different subgroups by their ICP ≥ 250 mmH2O or >330 mmH2O. By plotting the ROC curves, the optimal cut-off values to predict ICP ≥ 250 mmH2O or >330 mmH2O were obtained with the novel CVST-Score, the CVES method and the CVOS method. The optimal threshold of CVST-Score to predict ICP ≥ 250 mmH2O was 7.15 (AUC = 0.956, 95% CI: 0.919–0.994, Youden index = 0.805, P < 0.001), and the sensitivity was 94.0%, specificity 86.5%, positive predictive value (PPV) 90.4%, and negative predictive value (NPV) 91.4%. The optimal threshold of CVST-Score to predict ICP >330 mmH2O was 11.62 (AUC = 0.895, 95% CI: 0.830–0.961, Youden index = 0.708, P < 0.001), and the sensitivity was 90.5%, specificity 80.3%, PPV 59.4% and NPV 96.4%. Details in Fig. 4 (panel b). Furthermore, when using the conventional CVES method, the optimal threshold to predict ICP ≥ 250 mmH2O was 3.25 (AUC = 0.863, 95% CI: 0.788–0.938, Youden index = 0.584, P < 0.001), and the sensitivity was 80.0%, specificity 78.4%, PPV 83.3% and NPV 74.4%. The optimal threshold of the conventional CVES method to predict ICP >330 mmH2O was 3.75 (AUC = 0.834, 95% CI: 0.748–0.921, Youden index = 0.556, P < 0.001), and the sensitivity was 90.5%, specificity 65.2%, PPV 45.2% and NPV 95.6%. Details in Fig. S1 (panel b). Additionally, when assessing with the conventional CVOS method, the optimal threshold to predict ICP ≥ 250 mmH2O was 0.75 (AUC = 0.725, 95% CI: 0.620–0.830, Youden index = 0.384, P < 0.001), and the sensitivity was 60.0%, specificity 78.4%, PPV 78.9% and NPV 59.2%. The optimal threshold of the conventional CVOS method to predict ICP >330 mmH2O was 1.75 (AUC = 0.728, 95% CI: 0.586–0.871, Youden index = 0.465, P < 0.001), and the sensitivity was 57.1%, specificity 89.4%, PPV 63.2% and NPV 86.8%. Details in Fig. S2 (panel b). A summary of the results in ROC analysis was given in Table S2.
Finally, we analyzed the ability of the novel CVST-Score to predict ICP ≥ 250mmH2O with logistic regression models by adjusting for potential confounding factors. In the univariate logistic regression analysis, sex, clinical course and CVST-Score were associated with ICP ≥ 250 mmH2O (p = 0.02, p = 0.03 and p < 0.001, respectively). In the multivariate logistic regression analysis, only CVST-Score remained an independent predictor for ICP ≥ 250mmH2O after adjusting for sex, clinical course, age and BMI (p < 0.001). The ORs and 95% CIs for all the variables were detailed in Fig. 5. Furthermore, considering the possible impact of cerebral infarction on ICP and the sample size in this study, a second multivariate logistic regression model was set up with other five variables including, sex, clinical course, CVST-Score, BMI and cerebral infarction. Intriguingly, CVST-Score was still the only independent predictor for ICP ≥ 250mmH2O (p < 0.001), details in Fig. S3.