- Email: [email protected]
Accuracy of magnetic resonance venography in diagnosing cerebral venous sinus thrombosis
Accepted Manuscript Accuracy of magnetic resonance venography in diagnosing cerebral venous sinus thrombosis
Liansheng Gao, Weilin Xu, Tao Li, Xiaobo Yu, Shenglong Cao, Hangzhe Xu, Feng Yan, Gao Chen PII: DOI: Reference:
S0049-3848(18)30353-0 doi:10.1016/j.thromres.2018.05.012 TR 7033
To appear in:
Thrombosis Research
Received date: Revised date: Accepted date:
2 March 2018 8 May 2018 13 May 2018
Please cite this article as: Liansheng Gao, Weilin Xu, Tao Li, Xiaobo Yu, Shenglong Cao, Hangzhe Xu, Feng Yan, Gao Chen , Accuracy of magnetic resonance venography in diagnosing cerebral venous sinus thrombosis. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Tr(2017), doi:10.1016/j.thromres.2018.05.012
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Accuracy of magnetic resonance venography in diagnosing cerebral venous sinus thrombosis Liansheng Gao1M.D. Weilin Xu1 M.D. Tao Li1 M.D. Xiaobo Yu1 M.D. Shenglong Cao1 M.D. Hangzhe Xu1 M.D. Feng Yan1 M.D. Gao Chen1 M.D, Ph.D. 1. Department of Neurosurgery, Second Affiliated Hospital, School of Medicine,
PT
Zhejiang University, Hangzhou, Zhejiang, China. Running title: Identification of CVST with MRV
RI
Corresponding Author Gao Chen M.D, Ph.D.
SC
Department of Neurosurgery, Second Affiliated Hospital, School of Medicine, Zhejiang University, 88 Jiefang Rd, Hangzhou, Zhejiang 310009, China. Tel:
NU
+8613805716226. E-mail: [email protected].
Liansheng Gao, Weilin Xu and Tao Li have equally contributed to this work as
MA
co-first authors. Conflict of interests:
D
The authors report no conflict of interests to declare.
PT E
Source of funding:
This research did not receive any specific grant from funding agencies in the public,
Highlights:
CE
commercial, or not-for-profit sectors.
CVST is a rare but lethal cerebrovascular disease. Early diagnosis and proper
AC
treatment could remarkably improve prognosis.
MRV has excellent diagnostic performance and is effective in confirming CVST.
The elliptic centric MRV has high spatial resolution and definition for clinical use.
1
ACCEPTED MANUSCRIPT Abstract Objectives: The non-specific clinical manifestations and lack of effective diagnostic techniques have made cerebral venous sinus thrombosis (CVST) difficult to recognize and easy to misdiagnose. Several studies have suggested that different types of magnetic resonance venography (MRV) have advantages in diagnosing CVST. We
PT
conducted this meta-analysis to assess the accuracy of MRV in identifying CVST. Material and Methods: We searched the Embase, PubMed, and Chinese Biomedical
RI
(CBM) databases comprehensively to retrieve eligible articles up to Mar 31, 2018. The methodological quality of each article was evaluated individually. The summary
SC
diagnostic accuracy of MRV for CVST was obtained from pooled analysis with random-effects models. Sensitivity analysis, subgroup analysis, and meta-regression
NU
were used to explore the sources of heterogeneity. A trim and fill analysis was conducted to correct the funnel plot asymmetry.
MA
Results: The meta-analysis synthesized 12 articles containing 27 cohorts with a total of 1933 cases. The pooled sensitivity and specificity were 0.86 (95% CI: 0.83, 0.89)
D
and 0.94 (95% CI: 0.93, 0.95), respectively. The pooled diagnostic odds ratio (DOR)
PT E
was 75.24 (95% CI: 38.33, 147.72). The area under the curve (AUC) was 0.9472 (95% CI: 0.9229, 0.9715). Subgroup analysis and meta-regression analysis revealed the technical types of MRV and the methods of counting cases contributing to the
CE
heterogeneity. The trim and fill method confirmed that publication bias has little effect on our results.
AC
Conclusions: MRV has excellent diagnostic performance and is accurate in confirming CVST. Key words: Cerebral venous thrombosis; cerebral venous sinus thrombosis; magnetic resonance venography; digital subtraction angiography; meta-analysis Abbreviations: Cerebral venous thrombosis (CVT); cerebral venous sinus thrombosis (CVST); digital subtraction angiography (DSA); computed tomography (CT); magnetic resonance imaging (MRI); time-of-flight (TOF); phase-contrast (PC);magnetic resonance venography (MRV); dimensional (D); elliptic centric (ec); deep venous 2
ACCEPTED MANUSCRIPT system (DVS); contrast-enhanced (CE); inconsistency index (I2); diagnostic odds ratio (DOR); sensitivity (SEN); confidence interval (CI); specificity (SPE);positive likelihood ratio (LR+); negative likelihood ratio (LR-); Likelihood ratio (LR); area under the curve (AUC); summary receiver-operating characteristic (SROC);Chinese Biomedical (CBM); true positive(TP); false positive(FP); false negative(FN); and true
PT
negative(TN); Quality Assessment Tool for Diagnostic Accuracy Studies version 2
AC
CE
PT E
D
MA
NU
SC
RI
(QUADAS-2);effective sample sizes (ESS).
3
ACCEPTED MANUSCRIPT Introduction Cerebral venous thrombosis (CVT) is a rare but life-threatening cerebrovascular disease, which causes a series of serious neurologic symptoms and potential fatal outcomes [1, 2]. CVT accounts for about 0.5% of all strokes. The annual incidence is about 5 per million, far less frequent than ischemic or hemorrhagic stroke [3]. CVT
PT
includes cerebral superficial venous thrombosis, deep cerebral venous thrombosis and cerebral venous sinus thrombosis (CVST), among which CVST appears to be more
RI
common, more severe, and with higher risks [4]. Thrombosis of the cerebral sinuses most often affects younger adults, which may present with a variety of clinical
SC
symptoms, ranging from subtle to severe, while they rarely present with stroke syndrome, which is considered to be an atypical form of cerebral stroke [5, 6].
NU
Headache is the most common complaint, appearing in 71.3% of patients and can be accompanied by vomiting [7].
MA
Diverse clinical presentations and a lack of accurate diagnostic techniques make it difficult to diagnose CVST, which contributes to high misdiagnosis and mortality
D
rates [8]. Early diagnosis and proper treatment may reduce the risk of a fatal outcome
PT E
or severe disability [9]. Many radiologic methods have been used to visualize the cerebral venous system, while it remains a diagnostic challenge to determine CVST accurately. The gold standard for diagnosing CVST is digital subtraction angiography
CE
(DSA). However, the invasive and radioactive characteristics of DSA limit its widespread application [9]. Conventional computed tomography (CT) and magnetic
AC
resonance imaging (MRI) have also been investigated to increase our ability to detect this disease [10]. However, the appearance of thrombosis varies greatly according to stage [11].
Currently, many MRI techniques have been investigated for diagnosing CVST. The most commonly applied MRI technique is magnetic resonance venography (MRV), which enables the visualization of the cerebral venous system. MRV has high spatial resolution and has the advantage of being non-invasive and nonradioactive, which makes it one of the most common imaging methods for diagnosing CVST [12, 13]. There are many types of MRV used for the diagnosis of CVST. The most common 4
ACCEPTED MANUSCRIPT are time-of-flight (TOF) sequence, phase-contrast (PC) MRV, and elliptic centric (ec) MRV. TOF sequence is widely used to assess the cerebral venous sinuses because it has a short examination time and does not require contrast administration, which makes it convenient for clinical use. However, various image artefacts mimicking CVST have been reported for 2D TOF MRV [14]. PC MRV is easily affected by the
PT
velocity of blood flow and turbulence, which makes it unsuitable for visualization of the intracranial venous system, because it will lead to obvious intravascular signal
RI
loss. Moreover, PC MRV is more susceptible to motion artifact and the examination time for the entire venous system is long [15].
SC
Recently, many new MRV techniques have been introduced to increase the diagnostic accuracy of CVST. The ec MRV overcomes the shortcomings of TOF and
NU
PC MRV. It has high spatial resolution and fewer flow artifacts. After 3D reconstruction, it increases the likelihood that the thrombi are revealed. In addition,
MA
the examination time is acceptable for clinical use. Diverse ec MRV techniques, especially the contrast-enhanced (CE) ones, are probably superior for the assessment
D
of the deep venous system (DVS) and cortical veins [16-18].
PT E
In summary, numerous studies focus on the diagnosis of CVST using MRV. However, no comprehensive analysis has been conducted until now. The purpose of
CVST.
CE
this meta-analysis was to comprehensively assess the diagnostic accuracy of MRV for
Materials and methods
AC
Ethical Review
The study was approved by the ethics committee of Second Affiliated Hospital of Zhejiang University School of Medicine. Search strategy Databases including Embase, PubMed and Chinese Biomedical database (CBM) were searched to access eligible published articles (up to Mar 31, 2018). The keywords used are listed as follows: MRV or MR venography or MR angiography or magnetic resonance venography; CVST or cerebral venous sinus thrombosis or cerebral venous thrombosis or cranial venous sinus thrombosis or dural sinus 5
ACCEPTED MANUSCRIPT thrombosis. The detailed search strategy is shown in Table 1. Furthermore, the references of all eligible publications were also examined for possible relevant articles for inclusion in this study. Inclusion and exclusion criteria Inclusion criteria: (1) MRV of any type was applied to diagnose CVST in
PT
suspected patients; (2)the gold standard was DSA and/or clinical comprehensive analysis including other imaging methods and clinical follow-up; (3) each article
RI
included at least 10 cases; (4) the four values: true positive (TP), false positive (FP), false negative (FN), and true negative (TN) could be directly obtained or calculated
SC
from the primary data; (5) the language was limited to English and Chinese. Exclusion criteria: (1) repeated publication data; (2) letters, reviews, case reports,
NU
editorials, conference papers, abstracts or proceedings; (3) unpublished grey literature; (4) animal experiments.
MA
Three neurosurgeons (Tao Li, Feng Yan, and Hangzhe Xu) separately assessed the search results by reading the titles, while another three reviewing authors (Liansheng
D
Gao, WeilinXu, and Xiaobo Yu) independently reviewed the abstracts of the
PT E
underlying eligible articles after initially filtering according to the inclusion and exclusion criteria mentioned above. Any disagreements were settled through consultation and determined by the senior author (Gao Chen).
CE
Data extraction and quality assessment We conducted this meta-analysis according to the PRISMA guidelines
AC
(Supplemental TableⅠ). Three authors (Liansheng Gao, Tao Li and Shenglong Cao) separately reviewed and extracted data with a unified standard before an agreement was reached. The data extracted from each article contained the following information: names of authors, year of publication, country, study design, number of patients and segments involved, gender and age of patients, stages of CVST, type of MRV techniques, field strength, enhancement or not, counting method of the numbers of thrombosis and the figures of TP, FP, FN, and TN. There are two issues should be noted in our study. First, there were two methods of counting existing in the included articles. Some took count of the diseased cerebral 6
ACCEPTED MANUSCRIPT venous sinus segments to access statistical data, while others counted the number of patients. In this review, we combined these two methodologies to acquire data and named it “case”; further subgroup analysis divided this combination. Second, some articles may contain several cohorts, which were distinguished by different letters in our review. All data was re-checked by a third author (Weilin Xu) and any
PT
disagreements were disposed by another reviewer (Gao Chen). The document quality included was assessed using the Quality Assessment Tool for
RI
Diagnostic Accuracy Studies version 2 (QUADAS-2) recommended by Cochrane [19]. The quality assessment was performed and the bias risk map was drawn with
SC
Review Manager 5.3 [19]. Statistical analysis
NU
This meta-analysis was conducted following the standard methods recommended for diagnostic accuracy meta-analysis [20, 21].
MA
First, the heterogeneity level among cohorts due to the threshold effect was determined using threshold analysis. Spearman correlation coefficient between the
D
logit of sensitivity and the logit of (1−specificity) was used to indicate the threshold
PT E
effect. A high correlation with p< 0.05 means an obvious threshold effect. Second, the inconsistency index (I2) of the diagnostic odds ratio (DOR) was used to assess the degree of heterogeneity among cohorts due to the non-threshold effect. I2
CE
ranging from 0% to 40% represents no heterogeneity; 30% to 60% represents moderate heterogeneity; 50% to 90% represents substantial heterogeneity; and 75% to
AC
100% represents considerable heterogeneity according to the recommendations of Cochrane. The fixed-effects coefficient binary regression model was used if no obvious heterogeneity was discovered [19, 21]; otherwise, the random-effects coefficient binary regression model was applied to summarize the data. Meta-regression, subgroup analysis, and sensitivity analysis were also used to explore probable sources of heterogeneity. Each subgroup was supposed to comprise at least three cohorts with homogeneous characteristics following the same parameters. A Z-test was performed to compare subgroups and a value of p<0.05 indicates substantial differences. 7
ACCEPTED MANUSCRIPT Third, fixed- or random-effects models were used to compute the pooled data of the sensitivity (SEN), specificity (SPE), positive likelihood ratio (LR+), negative likelihood ratio (LR-), and DOR with 95% confidence intervals (CIs) according to the analysis mentioned above. A value of 0.5 was added to each cell in the table if any of the figures of TP, FP, FN, TN appeared to be zero, avoiding the SENs or SPEs being
PT
100%. Fourth, the summary receiver-operating characteristic curve (SROC) was drawn,
RI
and the area under the curve (AUC) and Q* index was calculated accordingly. The diagnostic accuracy was evaluated as follows: the value of AUC between 51% and 70%
SC
indicates low accuracy, while 71% to 90% indicates moderate accuracy and 90% or greater indicates high accuracy.
NU
Finally, Deek’s funnel plot and linear regression methods were used to assess the publication bias, visually with the X-axis indicating DOR and the Y-axis indicating
MA
1/root (ESS) [22]. The value of p<0.05 suggests significant asymmetry, which indicates publication bias [23]. The trim and fill method was applied to rectify the
D
funnel plot asymmetry caused by publication bias [24].
PT E
The statistical analysis was performed with Stata statistical software 14.0 (StataCorp LP, College Station, TX) and Meta-Disc statistical software version 1.4
Results
CE
[19, 21].
Article selection and the characteristics
AC
The retrieved records from Embase, PubMed and Chinese Biomedical databases included 378, 650 and 439 articles, respectively. No additional results were acquired from other sources. A total of 1116 articles remained after duplicates were removed. After titles and abstracts were screened, 101 potentially eligible records were chosen for further full text evaluation. After removal of reviews, case reports, articles whose data could not be extracted and all overlapping data, 12 articles containing 27 cohorts with 1933 cases satisfied all inclusion and exclusion criteria and were enrolled in the final meta-analysis [11, 14, 25-34] The document selection procedure following the PRISMA flow diagram is shown 8
ACCEPTED MANUSCRIPT in Fig 1. The characteristics of included articles are summarized in Table 2. The age span of patients was 12 to 80 years. The number of male patients was fewer than female patients (Male/Female=183/220). Twenty three cohorts were retrospective while 4 were prospective. The sample size of each cohort ranged from 16 to 300 cases. 12 cohorts counted the diseased cerebral venous sinus segments to access statistical
PT
data, while other 15 cohorts counted the number of patients. Three types of MRV techniques were analyzed in our review. There were 11 cohorts using 3D
RI
contrast-enhanced (CE) MRV to diagnose CVST, while 13 used TOF MRV and 3 used PC MRV. The field strength included 0.35T, 1.5T and 3.0T with or without
SC
enhancement. DSA was used as gold standard in 12 cohorts while other 15 cohorts used combination of multiple imaging methods and clinical follow up as reference
NU
standard. The phase of CVST could not be distinguished in most studies. The methodological quality graph and methodological quality summary graph are
MA
shown in Fig 2. Most studies had low or unclear risk of bias which meant the study qualities were acceptable.
D
Quantitative synthesis
PT E
The Spearman correlation coefficient was 0.121 (p=0.547) and the I2 of the diagnostic odds ratio (DOR) was 58.8% in the overall analysis. Meta-regression analysis revealed that two parameters, technical types of MRV and the methods of
CE
counting cases, contribute to the heterogeneity due to the non-threshold effect with p-values equaling 0.011 and 0.013, respectively.
AC
Among all 27 cohorts, the pooled SEN was 0.86 (95% CI: 0.83, 0.89); the pooled SPE was 0.94 (95% CI: 0.93, 0.95); the pooled LR+ was 8.64 (95% CI: 4.93, 15.14); the pooled LR- was 0.17 (95% CI: 0.12, 0.25); the pooled DOR was 75.24 (95% CI: 38.33, 147.72). The forest plots of SEN and SPE of the 27 included cohorts are shown in Fig3. The AUC for the SROC was 0.9472 (95% CI: 0.9229, 0.9715) with a Q* index equal to 0.8868, which is shown in Fig 4. Subgroup analysis was conducted according to five parameters and the results are shown in Table 3. Twelve cohorts took count of the number of diseased cerebral venous sinus segments to calculate diagnostic indexes. The pooled SEN and SPE 9
ACCEPTED MANUSCRIPT were 0.84 (95% CI: 0.80, 0.88) and 0.96 (95% CI: 0.95, 0.97), respectively. The corresponding LR+ and LR- were 19.00 (95% CI: 9.45, 38.21) and 0.14 (95% CI: 0.06, 0.31), respectively. The summary DOR was 198 (95% CI: 82.74, 473.82) and the AUC was 0.9790. Fifteen other cohorts counted the number of patients to calculate diagnostic indexes. The pooled SEN and SPE were 0.88 (95%CI: 0.83, 0.92)
PT
and 0.83 (95% CI: 0.79, 0.87), respectively. The corresponding LR+ and LR- were 3.69 (95% CI: 2.19, 6.23) and 0.21 (95% CI: 0.14, 0.34), respectively. The summary
RI
DOR was 23.61 (95% CI: 10.81, 51.55) and the AUC was 0.9096.The Z-test revealed that there was a significant statistical difference between these two parameters
SC
(Pinteraction= 0.011).
3D CE MRV was applied in 11 cohorts. The pooled SEN and SPE were: 0.85 (95%
NU
CI: 0.80, 0.89) and 0.98 (95% CI: 0.97, 0.99), respectively. The corresponding LR+ and LR- were 32.13 (95% CI: 12.18, 84.79) and 0.12 (95% CI: 0.05, 0.31),
MA
respectively. The summary DOR was 328.92 (95% CI: 113.38, 954.26) and the AUC was 0.9858. TOF MRV was applied in 13 cohorts. The pooled SEN and SPE were:
D
0.85 (95% CI: 0.79, 0.89) and 0.88 (95% CI: 0.84, 0.91), respectively. The
PT E
corresponding LR+ and LR- were 4.12 (95% CI: 2.34, 7.25) and 0.26 (95% CI: 0.15, 0.44), respectively. The summary DOR was 26.22 (95% CI: 11.22, 61.27) and the AUC was 0.9061. PC MRV was applied in 3 cohorts. The pooled SEN and SPE were:
CE
0.89 (95% CI: 0.82, 0.95) and 0.88 (95% CI: 0.83, 0.92), respectively. The corresponding LR+ and LR- were 4.62 (95% CI: 1.02, 21.04) and 0.13 (95% CI: 0.08,
AC
0.23), respectively. The summary DOR was 35.01 (95% CI: 7.06, 173.00) and the AUC was 0.9525.The Z-test revealed that there was a significant statistical difference between 3D CE MRV and TOF MRV (Pinteraction=0.006), and between 3D CE MRV and PC MRV (Pinteraction=0.045), with no statistical significant difference between TOF MRV and PC MRV (Pinteraction=0.132). The results of subgroup analysis based on study design (prospective versus retrospective), reference standard (DSA only versus not DSA) and field strength (3.0T versus 1.5T) are shown in Table 3. Other subgroups were ineligible for subgroup analysis because of insufficient data. 10
ACCEPTED MANUSCRIPT The results of the sensitivity analysis showed that no matter which single article was excluded, the heterogeneity was not obviously decreased, and no obvious changes were found in the pooled DOR (Table 4). In addition, after removing the four articles with relatively high risk of bias [14, 25, 26, 32], the heterogeneity did not significantly decrease (I2=60.2%) and there was no significant change in the pooled
PT
DOR (DOR=47.86, 95% CI: 16.86-135.92, p=0.475). Heterogeneity analysis
RI
Moderate heterogeneity among cohorts was found in the summary analysis of DOR with I2=58.8% and the results of meta-regression analysis, subgroup analysis and
SC
sensitivity analysis are displayed above. Publication bias
NU
The results indicate a publication bias among included cohorts (p=0.02, Fig 5A). The results of the trim and fill method indicated that there was no significant
MA
statistical difference in the pooled effect value and their CIs before and after five additional cohorts were appended (The p-values were both equal to 0). The filled
D
funnel plot shown in Fig 5B was symmetric, which also indicated no publication bias.
PT E
Discussion
CVST is a rare but lethal cerebrovascular disease with unclear etiology. The clinical manifestations of CVST are complex and nonspecific. Reasonable selection of the
CE
examination methods is essential for the early diagnosis and proper treatment of CVST, which may remarkably improve the prognosis [35]. MRV is non-invasive and
AC
could display the configuration of cerebral veins and sinuses better than conventional CT and MRI, which make it a promising imaging modality for the diagnosis of CVST [36].
This meta-analysis summarizes the diagnostic performance of commonly used MRV types in identifying CVST. The result of a threshold analysis implied no obvious threshold effect in our study. The pooled SEN was 0.86 and a higher value of pooled SPE was 0.94, which meant that MRV had low rate of missed diagnosis and an even lower rate of misdiagnosis in confirming CVST. Furthermore, the DOR synthesized the sensitivity and specificity data into a single value to better reflect the 11
ACCEPTED MANUSCRIPT diagnostic accuracy [37]. Our results showed that the DOR was 75.24 for the overall analysis, which indicated high diagnostic accuracy, while the I2 of the DOR was 58.8%, which revealed moderate heterogeneity among the cohorts. LR is another meaningful index used in the evaluation of the accuracy of diagnostic tests. It combines the advantages of sensitivity, specificity, positive predictive value, and
PT
negative predictive value, and is not affected by the prevalence rate. LR > 10 or LR < 0.1 indicates a high level of accuracy [19]. The pooled positive and negative LR of
RI
our study was 8.64 and 0.17, respectively, indicating good diagnostic accuracy. The AUC is an intuitive plot to report the ability of a diagnostic test and could be used to
SC
compare the diagnostic accuracy among the subgroups. Our results showed that the AUC was 0.9472, suggesting great diagnostic performance independent from diverse
NU
parameters such as type of MRV, computing methods, study design, and reference standards.
MA
Meta-regression found that the heterogeneity mainly originated from two parameters, the technical types of MRV (p=0.011) and the method of counting cases
D
(p=0.013). Different technical types of MRV had different diagnostic abilities, which
PT E
caused the heterogeneity in the pooled analysis. As mentioned above, two methods of counting existed in the included cohorts. Because one patient may have several cerebral venous sinus thrombi, the number of diseased venous segments was much
CE
greater than the number of patients. Combining these two kinds of data may induce relatively large differences among the sample sizes and yield the heterogeneity. The
AC
pooled SENs and SPEs show obvious differences between these two subgroups, which is possibly due to the different constituent ratios of 3D CE MRV and non-3D CE MRV between the two subgroups. Sensibility analysis is another method to find the origin of heterogeneity; in addition, it can also evaluate the stability of the meta-analysis [38]. In our study, sensibility analysis suggested that the results of the meta-analysis were not over-reliant on some articles and the conclusion was stable. Three types of MRV were used to identify CVST and all showed high accuracy, but with some differences. The DOR of 3D CE MRV was higher than those of the 12
ACCEPTED MANUSCRIPT other two. The Z-test revealed the diagnostic performance of 3D CE MRV was better than that of TOF MRV (Pinteraction=0.006) and PC MRV (Pinteraction=0.045), while TOF MRV and PC MRV showed no statistical differences (Pinteraction=0.132).These results were coincident with previous studies [14, 15]. 3D CE MRV has high spatial resolution and contrast and thus can clearly show the normal intracranial venous
PT
system and easily diagnose CVST [39-41]. The LR+ of 3D CE MRV was 32.13, which was much higher than those of the other two types of MRV, which suggested
RI
that TOF and PC MRV may suffer from false positives and have higher rate of misdiagnosis than 3D CE MRV, which was coincident with previous studies [14, 15].
SC
Ayanzen et al. reported TOF has a variety of image artifacts [42]. Liauw et al. thought PC MRV was more susceptible to motion artifacts [43]. This may be the main cause
NU
of misdiagnosis. However, the LR− seemed similar among these subgroups, indicating that they had a similar rate of missed diagnosis. As a result, when these
MA
three types of MRV are negative for the diagnosis of CVST, a negative conclusion cannot be made. In this situation, the DSA should be used so as not to miss the
D
diagnosis. Above all, 3D CE MRV leads to fewer false positives and shows obvious
PT E
advantages. The contrast agent is rich in the blood while absent in thrombosis, which increases the contrast between the blood and thrombus. However, the enhanced MRV is inappropriate for critical patients and TOF or PC MRV could be an alternative.
CE
Moreover, it should be indicated that MRV seems to be powerless in diagnosing small cortical vein thrombosis and partial occlusion [42].
AC
There were 8 cohorts utilizing 3.0T MRV and 14 others adopting 1.5T MRV. AUCs for the SROC of these two techniques were 0.9653 and 0.9159, respectively, which implied that they both have a high level of diagnostic accuracy. Compared to 1.5T MRV, 3.0T MRV has clearer resolution of the image and shorter scanning time, which together increase the efficiency in clinical use [44]. 3.0T MRV is good at identifying subtle lesions and anatomical structures, especially for minor cortical veins [45]. Although the Z-test revealed no significant difference in diagnostic performance between these two subgroups (Pinteraction=0.117), the 3.0T MRV has better performance than 1.5T MRV in clinical application. Further evaluation is needed. 13
ACCEPTED MANUSCRIPT Two reference standards were used to assess CVST, one was DSA only and the other was the combination of multiple imaging and clinical follow up. Whiting et al. thought the adoption of inconsistent reference standards may potentially overestimate the diagnostic accuracy of the test [19]. Conventionally, the gold standard for diagnosing CVST is DSA; however, its invasiveness, which may increase the risk of
PT
mortality, and the radioactivity, which may be a contraindication for special persons like pregnant woman, limits its application. Fortunately, there was no statistically
RI
significant difference in diagnostic performance between these two subgroups (Pinteraction=0.932). Thus, the use of a combination of multiple imaging and clinical
SC
follow up are an acceptable alternative which partly overcome the shortcomings of DSA.
NU
Deek’s funnel plot showed obvious asymmetry suggesting possible publication bias (p=0.02), which means the diagnostic value may be overemphasized. This is possibly
MA
due to our exclusion of unpublished grey literature. However, further trim and fill analysis indicated that the pooled effect quantities estimate changed little after five
D
cohorts were filled, which indicated our results were stable and the publication bias
PT E
had little effect on our results. The trim and fill analysis was in fact a type of sensitivity analysis method. Peters et al. thought trim and fill analysis was conducive
Limitations
CE
to reducing the bias of pooled effect quantities [46].
Despite the excellent diagnostic performance and effectiveness of MRV in
AC
confirming CVST, there were some limitations in our meta-analysis. First, the cohorts included showed great diversity in study design, computing method, technical types of MRV, reference standard, patient selection, and sample sizes. In addition, some included articles also show relatively high risk of bias and these factors may potentially contribute to heterogeneity. As a result, it should be prudent to interpret the results. Second, a subgroup analysis of important clinical significance could not be conducted due to insufficient data extracted from the included cohorts for several parameters, such as the phases of CVST, other reference standards such as CTV, and 14
ACCEPTED MANUSCRIPT different segments of the cerebral venous sinus. More studies using these parameters in the comparison are needed. Finally, the data of our meta-analysis mainly came from countries in Asia and Europe, and only published articles with full text in English and Chinese were included, which may omit some eligible articles either unpublished or reported in
PT
other languages from other countries. Conclusion
RI
In summary, our meta-analysis suggested that MRV has a high level of diagnostic accuracy for identifying CVST, although the results should still be interpreted and
SC
promoted prudently. In addition, more prospective studies with high quantity and large sample sizes are needed to confirm our conclusions. With the continued
NU
improvement of imaging techniques and our understanding of CVST, MRV will have
MA
broad application prospects for diagnosing CVST.
References
D
[1] Stam J. Thrombosis of the cerebral veins and sinuses. N Engl J Med.
PT E
2005;352:1791-8.
[2] Coutinho JM. Cerebral venous thrombosis. J Thromb Haemost. 2015;13 Suppl 1:S238-44.
CE
[3] Bousser MG, Ferro JM. Cerebral venous thrombosis: an update. Lancet Neurol. 2007;6:162-70.
AC
[4] Weimar C. Diagnosis and treatment of cerebral venous and sinus thrombosis. Curr Neurol Neurosci Rep. 2014;14:417. [5] Martinelli I, Bucciarelli P, Passamonti SM, Battaglioli T, Previtali E, Mannucci PM. Long-term evaluation of the risk of recurrence after cerebral sinus-venous thrombosis. Circulation. 2010;121:2740-6. [6] Saposnik G, Barinagarrementeria F, Brown RD, Jr., Bushnell CD, Cucchiara B, Cushman M, et al. Diagnosis and management of cerebral venous thrombosis: a statement
for
healthcare
professionals
from
the
American
Association/American Stroke Association. Stroke. 2011;42:1158-92. 15
Heart
ACCEPTED MANUSCRIPT [7] Sassi SB, Touati N, Baccouche H, Drissi C, Romdhane NB, Hentati F. Cerebral Venous Thrombosis: A Tunisian Monocenter Study on 160 Patients. Clinical and applied thrombosis/hemostasis : official journal of the International Academy of Clinical and Applied Thrombosis/Hemostasis. 2017;23:1005-9. [8] Wei Y, Deng X, Sheng G, Guo XB. A rabbit model of cerebral venous sinus
PT
thrombosis established by ferric chloride and thrombin injection. Neuroscience letters. 2018;662:205-12.
RI
[9] Masuhr F, Mehraein S, Einhaupl K. Cerebral venous and sinus thrombosis. Journal of neurology. 2004;251:11-23.
thrombosis. Clinical radiology. 2002;57:449-61.
SC
[10] Connor SE, Jarosz JM. Magnetic resonance imaging of cerebral venous sinus
NU
[11] Sari S, Verim S, Hamcan S, Battal B, Akgun V, Akgun H, et al. MRI diagnosis of dural sinus - Cortical venous thrombosis: Immediate post-contrast 3D GRE
MA
T1-weighted imaging versus unenhanced MR venography and conventional MR sequences. Clinical neurology and neurosurgery. 2015;134:44-54.
PT E
neurology. 2000;247:252-8.
D
[12] Bousser MG. Cerebral venous thrombosis: diagnosis and management. Journal of
[13] Sun Y, Zheng DY, Ji XM, Weale P, Wu H, Jiang LD, et al. Diagnostic performance of magnetic resonance venography in the detection of recanalization in
CE
patients with chronic cerebral venous sinus thrombus. Chinese medical journal. 2009;122:2428-32.
AC
[14] Klingebiel R, Bauknecht HC, Bohner G, Kirsch R, Berger J, Masuhr F. Comparative evaluation of 2D time-of-flight and 3D elliptic centric contrast-enhanced MR venography in patients with presumptive cerebral venous and sinus thrombosis. Eur J Neurol. 2007;14:139-43. [15] Pui MH. Cerebral MR venography. Clinical imaging. 2004;28:85-9. [16] Farb RI, Scott JN, Willinsky RA, Montanera WJ, Wright GA, terBrugge KG. Intracranial venous system: gadolinium-enhanced three-dimensional MR venography with auto-triggered elliptic centric-ordered sequence--initial experience. Radiology. 2003;226:203-9. 16
ACCEPTED MANUSCRIPT [17] Wetzel SG, Law M, Lee VS, Cha S, Johnson G, Nelson K. Imaging of the intracranial venous system with a contrast-enhanced volumetric interpolated examination. European radiology. 2003;13:1010-8. [18] Mermuys KP, Vanhoenacker PK, Chappel P, Van Hoe L. Three-dimensional venography of the brain with a volumetric interpolated sequence. Radiology.
PT
2005;234:901-8. [19] Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al.
RI
QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Annals of internal medicine. 2011;155:529-36.
SC
[20] Deville WL, Buntinx F, Bouter LM, Montori VM, de Vet HC, van der Windt DA,
medical research methodology. 2002;2:9.
NU
et al. Conducting systematic reviews of diagnostic studies: didactic guidelines. BMC
[21] Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc: a
MA
software for meta-analysis of test accuracy data. BMC medical research methodology. 2006;6:31.
D
[22] Kato T, Shinoda J, Nakayama N, Miwa K, Okumura A, Yano H, et al. Metabolic
11C-choline
PT E
assessment of gliomas using 11C-methionine, [18F] fluorodeoxyglucose, and positron-emission
2008;29:1176-82.
tomography.
AJNR
Am
J
Neuroradiol.
CE
[23] Tsuyuguchi N, Takami T, Sunada I, Iwai Y, Yamanaka K, Tanaka K, et al. Methionine positron emission tomography for differentiation of recurrent brain tumor
AC
and radiation necrosis after stereotactic radiosurgery--in malignant glioma. Annals of nuclear medicine. 2004;18:291-6. [24] Duval S, Tweedie R. Trim and fill: A simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56:455-63. [25] Meckel S, Reisinger C, Bremerich J, Damm D, Wolbers M, Engelter S, et al. Cerebral venous thrombosis: diagnostic accuracy of combined, dynamic and static, contrast-enhanced 4D MR venography. AJNR Am J Neuroradiol. 2010;31:527-35. [26] Ye J, Zhang XJ, Li J, Zhong Q, Hong JF, Wang SR. Case-control study of cerebral venous sinus thrombosis between MRV and DSA. Chinese Journal of 17
ACCEPTED MANUSCRIPT Neurosurgical Disease Research. 2016;15(6):485-9. [27] Jalli R, Zarei F, Farahangiz S, Khaleghi F, Petramfar P, Borhani-Haghighi A, et al. The Sensitivity, Specificity, and Accuracy of Contrast-Enhanced T1-Weighted Image, T2*-Weighted Image, and Magnetic Resonance Venography in Diagnosis of Cerebral Venous Sinus Thrombosis. Journal of stroke and cerebrovascular diseases :
PT
the official journal of National Stroke Association. 2016;25:2083-6. [28] Li XY, Yang J, Sun JB, Ni L. Comparison of TOF MRV and CE MRV in the
RI
diagnosis of cerebral venous sinus thrombosis. Chinese Community Doctors. 2012;14(317):244.
SC
[29] Ho JS, Rahmat K, Ramli N, Fadzli F, Chong HT, Tan CT. Cerebral venous sinus thrombosis: Comparison of multidetector computed tomography venogram (MDCTV)
NU
and magnetic resonance venography (MRV) of various field strengths. Neurology Asia. 2012;17(4):281-91.
MA
[30] Yi BN, Ba HT, Wang J, Zhang DQ. The value of magnetic resonance imaging in the diagnosis of intracranial venous sinus thrombosis. Guide of Chinese Medicine.
D
2012;10(21):149-50.
PT E
[31] Ju KJ, Wang LJ, Cao H. Diagnostic value of three-dimensional enhanced magnetic resonance angiography in patients with intracranial venous sinus thrombosis. Guangdong Medical Journal. 2011;32(19):2579-81.
CE
[32] Yang ML, Li T, Huang YH, Qu H. Early imaging features of cerebral venous sinus thrombosis: a report of 62 cases. Chinese Journal of Postgraduates of Medicine.
AC
2009;32(10):59-61.
[33] Ma BW, Huang LY, Du YH, Chen GS, Li YJ, Hou XL. Clinical Characters and Misdiagnosis Analysis in Patients with Cerebral Vein and Sinus Thrombosis. Journal of Ningxia Medical College. 2008;30(6):717-9. [34] Liang L, Korogi Y, Sugahara T, Onomichi M, Shigematsu Y, Yang D, et al. Evaluation of the intracranial dural sinuses with a 3D contrast-enhanced MP-RAGE sequence: prospective comparison with 2D-TOF MR venography and digital subtraction angiography. AJNR Am J Neuroradiol. 2001;22:481-92. [35] Qu H, Yang M. Early imaging characteristics of 62 cases of cerebral venous 18
ACCEPTED MANUSCRIPT sinus thrombosis. Exp Ther Med. 2013;5:233-6. [36] Zafar A, Ali Z. Pattern of magnetic resonance imaging and magnetic resonance venography changes in cerebral venous sinus thrombosis. Journal of Ayub Medical College, Abbottabad : JAMC. 2012;24:63-7. [37] Nakajima T, Kumabe T, Kanamori M, Saito R, Tashiro M, Watanabe M, et al.
PT
Differential diagnosis between radiation necrosis and glioma progression using sequential proton magnetic resonance spectroscopy and methionine positron emission
RI
tomography. Neurologia medico-chirurgica. 2009;49:394-401.
[38] Wang J, Gao J, He J. [Diagnostic value of ProGRP and NSE for small cell lung
SC
cancer: a meta-analysis]. Zhongguo fei ai za zhi = Chinese journal of lung cancer. 2010;13:1094-100.
NU
[39] Rollins N, Ison C, Reyes T, Chia J. Cerebral MR venography in children: comparison of 2D time-of-flight and gadolinium-enhanced 3D gradient-echo
MA
techniques. Radiology. 2005;235:1011-7.
[40] Kirchhof K, Welzel T, Jansen O, Sartor K. More reliable noninvasive
D
visualization of the cerebral veins and dural sinuses: comparison of three MR
PT E
angiographic techniques. Radiology. 2002;224:804-10. [41] Lettau M, Sartor K, Heiland S, Hahnel S. 3T high-spatial-resolution contrast-enhanced MR angiography of the intracranial venous system with parallel
CE
imaging. AJNR Am J Neuroradiol. 2009;30:185-7. [42] Ayanzen RH, Bird CR, Keller PJ, McCully FJ, Theobald MR, Heiserman JE.
AC
Cerebral MR venography: normal anatomy and potential diagnostic pitfalls. AJNR Am J Neuroradiol. 2000;21:74-8. [43] Liauw L, van Buchem MA, Spilt A, de Bruine FT, van den Berg R, Hermans J, et
al.
MR
angiography
of
the
intracranial
venous
system.
Radiology.
2000;214:678-82. [44] Hu HH, Haider CR, Campeau NG, Huston J, 3rd, Riederer SJ. Intracranial contrast-enhanced magnetic resonance venography with 6.4-fold sensitivity encoding at 1.5 and 3.0 Tesla. Journal of magnetic resonance imaging : JMRI. 2008;27:653-8. [45] Tomasian A, Salamon N, Krishnam MS, Finn JP, Villablanca JP. 3D 19
ACCEPTED MANUSCRIPT high-spatial-resolution cerebral MR venography at 3T: a contrast-dose-reduction study. AJNR Am J Neuroradiol. 2009;30:349-55. [46] Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Performance of the trim and fill method in the presence of publication bias and between-study heterogeneity.
PT
Statistics in medicine. 2007;26:4544-62.
Acknowledgements
RI
None.
SC
Conflict of interests
NU
The authors have no conflict of interests to declare.
Source of funding
MA
This research did not receive any specific grant from funding agencies in the public,
PT E
Authors’ contributions
D
commercial, or not-for-profit sectors.
This meta-analysis was designed by Liansheng Gao and Gao Chen, data collection and analysis were conducted by Weilin Xu, Tao Li, Feng Yan, Shenglong Cao,
CE
Xiaobo Yu and Hangzhe Xu, this article was checked by Gao Chen. This article was
AC
written by Liansheng Gao, Weilin Xu and Tao Li.
Figure Legends
Fig 1. Flow diagram of the article selection process. Fig 2. Methodological quality graph (A) and methodological quality summary graph (B) of each article using the QUADAS-2. Fig 3. Forest plots of the sensitivity (A) and specificity (B) with the 95% confidence interval (CI) of each cohort represented by the horizontal line through each circle of this meta-analysis. The pooled sensitivity and specificity for this meta-analysis is represented by diamonds. df=degrees of freedom. 20
ACCEPTED MANUSCRIPT Fig 4. Summary receiver-operating characteristic (SROC) curve of MRV for the diagnosis of CVST. AUC = area under the curve, SE = standard error. Fig 5. Deek’s funnel plot of publication bias, as determined by linear regression of the inverse root of effective sample sizes (ESS) on log diagnostic odds ratios (A) and filled funnel plot with pseudo 95% confidence limits (B). The filled cohorts for this
PT
meta-analysis are represented by squares. Table 1. Search terms and strategies of the accuracy of MRV in diagnosing CVST in
RI
different databases.
Table 2. Characteristics of cohorts included in the meta-analysis of MRV for the
SC
diagnosis of CVST.
Table 3. Subgroup analyses of diagnostic accuracy variables.
AC
CE
PT E
D
MA
accuracy of MRV in diagnosing CVST.
NU
Table 4. Sensibility analysis of each article included in the meta-analysis of the
21
ACCEPTED MANUSCRIPT Table 1 Search terms and strategies of the accuracy of MRV in diagnosing CVST in different databases Database
Search strategy
PubMed
((((((CVST) OR cerebral venous sinus thrombosis) OR cerebral venous thrombosis) OR cranial venous sinus thrombosis) OR dural sinus thrombosis)) AND (MRV or MR venography or MR angiography or magnetic resonance venography)
CBM
(((((cerebral venous sinus thrombosis) OR cranial venous sinus thrombosis) OR cerebral sinus thrombosis) OR CVST)) AND (MRV or MR venography or MR angiography or magnetic resonance venography)
Embase
('cerebral sinus thrombosis':ab,ti OR 'cvst':ab,ti OR 'sinus thrombosis':ab,ti) AND ('magnetic resonance venography':ab,ti OR 'mrv':ab,ti OR 'magnetic resonance angiography':ab,ti)
PT
Abbreviations: CBM, Chinese Biomedical databases; MRV, magnetic resonance venography; CVST, cerebral venous sinus
AC
CE
PT E
D
MA
NU
SC
RI
thrombosis.
22
ACCEPTED MANUSCRIPT
Table 2 Characteristics of cohorts included in the meta-analysis of MRV for the diagnosis of CVST Computi Author
Year
Study
No. of
No. of
design
patients
cases*
Country
Age(mean)
Final diagnosis
Reference
(CVST)
standard
M/F
Phase
ng
Field Techniques
T P
methods Ye J (a) [26]
1
16
10
Ye J (b)
1
16
10
6
80
Ye J (c) 2016
China
acute
22
R
31.8
C S U DSA
Ye J (d)
6
48
16
Ye J (e)
8
128
25
Ye J (f)
8
32
11
Jalli R (a) [27]
46
46
22
chronic
N A
General MRI
TP
FP
FN
TN
segment
3D CE MRV
3.0T+
10
0
0
6
segment
TOF MRV
3.0T-
10
2
0
4
segment
3D CE MRV
3.0T+
22
1
0
57
segment
TOF MRV
3.0T-
16
4
0
28
segment
3D CE MRV
3.0T+
14
0
11
103
segment
TOF MRV
3.0T-
11
2
0
19
patient
PC MRV
1.5T-
20
12
2
12
patient
TOF MRV
1.5T-
9
5
0
8
segment
PC MRV
3.0/1.5T-
60
19
7
214
3.0/1.5T+
62
0
5
233
I R
subacute
9/6
strength (enhanced+)
and CT and CTV
2016
Iran
R
Jalli R (b)
22
Sari S (a) [11] 2015
Turkey
Li XY (a) [28]
30
300
30
300
China
Li XY (b)
16
2012
Malaysia
C A
R
27
16 16
27
D E
PT
E C
P
M 9
40.1
16 2012
14/54
22
R
Sari S (b)
Ho JS [29]
35
31.2
acute
and clinical follow up
67
17/13
DSA
NA
67
3D-Volumetric segment GRE MRV
11
DSA or clinical
6/10
patient
3D CE MRV
3.0T+
10
1
1
4
patient
TOF MRV
3.0T-
9
3
2
2
NA
patient
TOF MRV
0.35/1.5/3.0-
10
3
0
14
NA
patient
PC MRV
1.5-
12
1
2
5
NA
patient
3D CE MRV
NA+
17
9
1
101
NA 11
follow up General MRI or
40.9
19/8
10
CT or clinical follow up
Yi BN [30]
2012
China
R
20
20
31
8/12
14
Ju KJ [31]
2011
China
R
128
128
42.6
58/70
18
DSA General MRI or DSA
23
ACCEPTED MANUSCRIPT
Meckel S (a) [25]
39
39
20
patient
Meckel S (b)
39
39
20
patient
TOF MRV
1.5T-
16
4
4
15
1.5T+
18
0
2
19
1.5T-
28
5
6
117
1.5T+
33
1
1
121
1.5T-
10
1
9
58
1.5T+
7
1
12
58
dynamic/static 3D MRV(4D)
General MRI or Meckel S (c)
39
Meckel S (d)
39
156
34
156
CTV or clinical
34
I R
NA
follow up Meckel S (e)
2010
Switzerland
R
Meckel S (f)
39
78
39
78
49.1
17/22
2009
China
R
21
21
NA
NA
Ma BW [33]
2008
China
R
51
51
NA
NA
25 2007
Germany
25
Liang LX (a) [34]
35 Japan
D E
25
T P E
P
35
49
Liang LX (b)
35
35
19/6
U N
19
A M 51
7
40.6
Klingebiel R (b)
2001
25
R
SC
19
Yang ML [32]
Klingebiel R (a) [14]
19
7
DSA
segment segment
3D MRV(4D) TOF MRV
dynamic/static
segment 3D MRV(4D) patient
TOF MRV
1.5T-
19
1
1
0
All phases
patient
TOF MRV
NA-
48
0
3
0
patient
3D CE MRV
1.5T+
6
0
1
18
patient
TOF MRV
1.5T-
5
8
2
10
patient
TOF MRV
1.5T-
6
2
6
21
1.5T+
10
0
2
23
General MRI or DSA
DSA and CTA and clinical
NA
follow up
DSA 12
TOF MRV
dynamic/static
NA
12
16/19
T P
segment
CT or CTA or
NA
3D MP RAGE patient MRV
C C
Abbreviations: R, respective; P, prospective; M, male; F, female; DSA, digital subtraction angiography; MRI, magnetic resonance imaging; DSA, digital subtraction angiography; MRV, magnetic resonance venography; CTA, computed tomography angiography; CVST, cerebral venous and sinus thrombosis; NA, not available; D, dimensional; CE, contrast-enhanced; TOF, time of flow; GRE, gradient echo; MP RAGE,
A
magnetization-prepared rapid acquisition with gradient echo.
* There were two methods of counting existing in the included articles. Some took count of the diseased cerebral venous sinus segments to access statistical data, while others counted the number of patients. In this review, we combined these two methodologies to acquire data and named it “case”.
24
ACCEPTED MANUSCRIPT
Table 3 Subgroup analyses of diagnostic accuracy variables Threshold Category
Cohorts (n)
Cases
Effects
SEN (95 % CI)
SPE (95 % CI)
LR+ (95 % CI)
LR- (95 % CI)
DOR (95 % CI)
0.86(0.83,0.89)
0.94(0.93,0.95)
8.64(4.93,15.14)
0.17(0.12,0.25)
75.24(38.33,147.72)
T P
(p-value) Overall
27
1933
0.547
I R
Computing Methods segment
12
1388
0.301
0.84(0.80,0.88)
0.96(0.95,0.97)
19.00(9.45,38.21)
patient
15
545
0.484
0.88(0.83,0.92)
0.83(0.79,0.87)
3.69(2.19,6.23)
3D CE MRV(A)
11
1001
0.670
0.85(0.80,0.89)
0.98(0.97,0.99)
32.13(12.18,84.79)
TOF MRV(B)
13
566
0.667
0.85(0.79,0.89)
0.88(0.84,0.91)
4.12(2.34,7.25)
PC MRV(C)
3
366
0.667
0.89(0.82,0.95)
0.88(0.83,0.92)
Retrospective
23
1831
0.451
0.86(0.83,0.89)
Prospective
4
102
1.000
0.76(0.61,0.87)
C S U
Study design
PT
Reference Standard DSA only
12
481
0.812
not DSA
15
1452
0.832
Field strength 3.0T
8
352
1.5T
14
775
E C
C A
D E
0.94(0.92,0.95) 0.88(0.77,0.95)
Pinteraction
0.9472(0.0124)
0.011
198(82.74,473.82)
0.9790(0.0096)
0.21(0.14,0.34)
23.61(10.81,51.55)
0.9096(0.0256)
0.12(0.05,0.31)
328.92(113.38,954.26)
0.9858(0.0088)
0.006 vs B
0.26(0.15,0.44)
26.22(11.22,61.27)
0.9061(0.0274)
0.132 vs C
4.62(1.02,21.04)
0.13(0.08,0.23)
35.01(7.06,173.67)
0.9525(0.0141)
0.045 vs A
8.87(4.90,16.04)
0.15(0.09,0.26)
84.83(40.40,178.12)
0.9539(0.0127)
4.65(1.03,21.05)
0.32(0.14,0.73)
17.40(3.31,91.45)
0.8641(0.0812)
Techniques
N A
M
0.14(0.06,0.31)
AUC (SE)
0.275
0.923
0.84(0.77,0.89)
0.93(0.89,0.95)
7.70(3.27,18.11)
0.20(0.10,0.38)
65.80(19.96,216.84)
0.9477(0.0235)
0.86(0.83,0.90)
0.94(0.92,0.95)
8.44(4.03,17.68)
0.16(0.08,0.31)
70.39(28.94,171.22)
0.9452(0.0174) 0.117
0.691
0.86(0.79,0.92)
0.94(0.90,0.97)
7.95(2.71,23.28)
0.11(0.03,0.38)
97.53(21.38,444.94)
0.9653(0.0196)
0.404
0.80(0.74,0.85)
0.92(0.89,0.94)
7.56(3.53,17.07)
0.25(0.15,0.41)
41.32(16.40,104.10)
0.9159(0.0274)
Abbreviations: SEN, sensitivity; SPE, specificity; LR, likelihood ratio; CI, confidence interval; DOR, diagnostic odds ratio; AUC, area under curve; SE, standard error; 3D, three-dimensional; CE, contrast-enhanced; MRV, magnetic resonance venography; TOF, time-of-flight; PC, phase-contrast; DSA, digital subtraction angiography.
25
ACCEPTED MANUSCRIPT Table 4 Sensibility analysis of each article included in the meta-analysis of the accuracy of MRV in diagnosing CVST I2 (%)
Author
DOR
Ye J [26]
59.66
28.05
126.89
64.4
Jalli R [27]
86.75
43.13
174.50
58.1
Sari S [11]
63.41
31.39
128.11
54.6
Li XY [28]
87.15
44.12
172.14
57.1
Ho JS [29]
75.04
37.52
150.08
60.3
Yi BN [30]
78.10
38.98
156.48
60.1
Ju KJ [31]
72.42
36.05
145.50
59.8
Meckel S [25]
59.68
25.66
138.78
56.3
Yang ML [32]
78.75
39.74
156.06
59.7
Ma BW [33]
77.83
39.19
154.54
60.0
Klingebiel R [14]
84.79
43.77
Liang LX [34]
80.57
39.98
Overall
75.24
38.33
SC
RI
PT
95 % CI
164.26
54.0
162.38
58.9
147.72
58.8
Abbreviations: MRV, magnetic resonance venography; CVST, cerebral venous and sinus thrombosis; DOR, diagnostic odds ratio;
AC
CE
PT E
D
MA
NU
CI, confidence interval.
26
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Liansheng Gao, Weilin Xu, Tao Li, Xiaobo Yu, Shenglong Cao, Hangzhe Xu, Feng Yan, Gao Chen PII: DOI: Reference:
S0049-3848(18)30353-0 doi:10.1016/j.thromres.2018.05.012 TR 7033
To appear in:
Thrombosis Research
Received date: Revised date: Accepted date:
2 March 2018 8 May 2018 13 May 2018
Please cite this article as: Liansheng Gao, Weilin Xu, Tao Li, Xiaobo Yu, Shenglong Cao, Hangzhe Xu, Feng Yan, Gao Chen , Accuracy of magnetic resonance venography in diagnosing cerebral venous sinus thrombosis. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Tr(2017), doi:10.1016/j.thromres.2018.05.012
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Accuracy of magnetic resonance venography in diagnosing cerebral venous sinus thrombosis Liansheng Gao1M.D. Weilin Xu1 M.D. Tao Li1 M.D. Xiaobo Yu1 M.D. Shenglong Cao1 M.D. Hangzhe Xu1 M.D. Feng Yan1 M.D. Gao Chen1 M.D, Ph.D. 1. Department of Neurosurgery, Second Affiliated Hospital, School of Medicine,
PT
Zhejiang University, Hangzhou, Zhejiang, China. Running title: Identification of CVST with MRV
RI
Corresponding Author Gao Chen M.D, Ph.D.
SC
Department of Neurosurgery, Second Affiliated Hospital, School of Medicine, Zhejiang University, 88 Jiefang Rd, Hangzhou, Zhejiang 310009, China. Tel:
NU
+8613805716226. E-mail: [email protected].
Liansheng Gao, Weilin Xu and Tao Li have equally contributed to this work as
MA
co-first authors. Conflict of interests:
D
The authors report no conflict of interests to declare.
PT E
Source of funding:
This research did not receive any specific grant from funding agencies in the public,
Highlights:
CE
commercial, or not-for-profit sectors.
CVST is a rare but lethal cerebrovascular disease. Early diagnosis and proper
AC
treatment could remarkably improve prognosis.
MRV has excellent diagnostic performance and is effective in confirming CVST.
The elliptic centric MRV has high spatial resolution and definition for clinical use.
1
ACCEPTED MANUSCRIPT Abstract Objectives: The non-specific clinical manifestations and lack of effective diagnostic techniques have made cerebral venous sinus thrombosis (CVST) difficult to recognize and easy to misdiagnose. Several studies have suggested that different types of magnetic resonance venography (MRV) have advantages in diagnosing CVST. We
PT
conducted this meta-analysis to assess the accuracy of MRV in identifying CVST. Material and Methods: We searched the Embase, PubMed, and Chinese Biomedical
RI
(CBM) databases comprehensively to retrieve eligible articles up to Mar 31, 2018. The methodological quality of each article was evaluated individually. The summary
SC
diagnostic accuracy of MRV for CVST was obtained from pooled analysis with random-effects models. Sensitivity analysis, subgroup analysis, and meta-regression
NU
were used to explore the sources of heterogeneity. A trim and fill analysis was conducted to correct the funnel plot asymmetry.
MA
Results: The meta-analysis synthesized 12 articles containing 27 cohorts with a total of 1933 cases. The pooled sensitivity and specificity were 0.86 (95% CI: 0.83, 0.89)
D
and 0.94 (95% CI: 0.93, 0.95), respectively. The pooled diagnostic odds ratio (DOR)
PT E
was 75.24 (95% CI: 38.33, 147.72). The area under the curve (AUC) was 0.9472 (95% CI: 0.9229, 0.9715). Subgroup analysis and meta-regression analysis revealed the technical types of MRV and the methods of counting cases contributing to the
CE
heterogeneity. The trim and fill method confirmed that publication bias has little effect on our results.
AC
Conclusions: MRV has excellent diagnostic performance and is accurate in confirming CVST. Key words: Cerebral venous thrombosis; cerebral venous sinus thrombosis; magnetic resonance venography; digital subtraction angiography; meta-analysis Abbreviations: Cerebral venous thrombosis (CVT); cerebral venous sinus thrombosis (CVST); digital subtraction angiography (DSA); computed tomography (CT); magnetic resonance imaging (MRI); time-of-flight (TOF); phase-contrast (PC);magnetic resonance venography (MRV); dimensional (D); elliptic centric (ec); deep venous 2
ACCEPTED MANUSCRIPT system (DVS); contrast-enhanced (CE); inconsistency index (I2); diagnostic odds ratio (DOR); sensitivity (SEN); confidence interval (CI); specificity (SPE);positive likelihood ratio (LR+); negative likelihood ratio (LR-); Likelihood ratio (LR); area under the curve (AUC); summary receiver-operating characteristic (SROC);Chinese Biomedical (CBM); true positive(TP); false positive(FP); false negative(FN); and true
PT
negative(TN); Quality Assessment Tool for Diagnostic Accuracy Studies version 2
AC
CE
PT E
D
MA
NU
SC
RI
(QUADAS-2);effective sample sizes (ESS).
3
ACCEPTED MANUSCRIPT Introduction Cerebral venous thrombosis (CVT) is a rare but life-threatening cerebrovascular disease, which causes a series of serious neurologic symptoms and potential fatal outcomes [1, 2]. CVT accounts for about 0.5% of all strokes. The annual incidence is about 5 per million, far less frequent than ischemic or hemorrhagic stroke [3]. CVT
PT
includes cerebral superficial venous thrombosis, deep cerebral venous thrombosis and cerebral venous sinus thrombosis (CVST), among which CVST appears to be more
RI
common, more severe, and with higher risks [4]. Thrombosis of the cerebral sinuses most often affects younger adults, which may present with a variety of clinical
SC
symptoms, ranging from subtle to severe, while they rarely present with stroke syndrome, which is considered to be an atypical form of cerebral stroke [5, 6].
NU
Headache is the most common complaint, appearing in 71.3% of patients and can be accompanied by vomiting [7].
MA
Diverse clinical presentations and a lack of accurate diagnostic techniques make it difficult to diagnose CVST, which contributes to high misdiagnosis and mortality
D
rates [8]. Early diagnosis and proper treatment may reduce the risk of a fatal outcome
PT E
or severe disability [9]. Many radiologic methods have been used to visualize the cerebral venous system, while it remains a diagnostic challenge to determine CVST accurately. The gold standard for diagnosing CVST is digital subtraction angiography
CE
(DSA). However, the invasive and radioactive characteristics of DSA limit its widespread application [9]. Conventional computed tomography (CT) and magnetic
AC
resonance imaging (MRI) have also been investigated to increase our ability to detect this disease [10]. However, the appearance of thrombosis varies greatly according to stage [11].
Currently, many MRI techniques have been investigated for diagnosing CVST. The most commonly applied MRI technique is magnetic resonance venography (MRV), which enables the visualization of the cerebral venous system. MRV has high spatial resolution and has the advantage of being non-invasive and nonradioactive, which makes it one of the most common imaging methods for diagnosing CVST [12, 13]. There are many types of MRV used for the diagnosis of CVST. The most common 4
ACCEPTED MANUSCRIPT are time-of-flight (TOF) sequence, phase-contrast (PC) MRV, and elliptic centric (ec) MRV. TOF sequence is widely used to assess the cerebral venous sinuses because it has a short examination time and does not require contrast administration, which makes it convenient for clinical use. However, various image artefacts mimicking CVST have been reported for 2D TOF MRV [14]. PC MRV is easily affected by the
PT
velocity of blood flow and turbulence, which makes it unsuitable for visualization of the intracranial venous system, because it will lead to obvious intravascular signal
RI
loss. Moreover, PC MRV is more susceptible to motion artifact and the examination time for the entire venous system is long [15].
SC
Recently, many new MRV techniques have been introduced to increase the diagnostic accuracy of CVST. The ec MRV overcomes the shortcomings of TOF and
NU
PC MRV. It has high spatial resolution and fewer flow artifacts. After 3D reconstruction, it increases the likelihood that the thrombi are revealed. In addition,
MA
the examination time is acceptable for clinical use. Diverse ec MRV techniques, especially the contrast-enhanced (CE) ones, are probably superior for the assessment
D
of the deep venous system (DVS) and cortical veins [16-18].
PT E
In summary, numerous studies focus on the diagnosis of CVST using MRV. However, no comprehensive analysis has been conducted until now. The purpose of
CVST.
CE
this meta-analysis was to comprehensively assess the diagnostic accuracy of MRV for
Materials and methods
AC
Ethical Review
The study was approved by the ethics committee of Second Affiliated Hospital of Zhejiang University School of Medicine. Search strategy Databases including Embase, PubMed and Chinese Biomedical database (CBM) were searched to access eligible published articles (up to Mar 31, 2018). The keywords used are listed as follows: MRV or MR venography or MR angiography or magnetic resonance venography; CVST or cerebral venous sinus thrombosis or cerebral venous thrombosis or cranial venous sinus thrombosis or dural sinus 5
ACCEPTED MANUSCRIPT thrombosis. The detailed search strategy is shown in Table 1. Furthermore, the references of all eligible publications were also examined for possible relevant articles for inclusion in this study. Inclusion and exclusion criteria Inclusion criteria: (1) MRV of any type was applied to diagnose CVST in
PT
suspected patients; (2)the gold standard was DSA and/or clinical comprehensive analysis including other imaging methods and clinical follow-up; (3) each article
RI
included at least 10 cases; (4) the four values: true positive (TP), false positive (FP), false negative (FN), and true negative (TN) could be directly obtained or calculated
SC
from the primary data; (5) the language was limited to English and Chinese. Exclusion criteria: (1) repeated publication data; (2) letters, reviews, case reports,
NU
editorials, conference papers, abstracts or proceedings; (3) unpublished grey literature; (4) animal experiments.
MA
Three neurosurgeons (Tao Li, Feng Yan, and Hangzhe Xu) separately assessed the search results by reading the titles, while another three reviewing authors (Liansheng
D
Gao, WeilinXu, and Xiaobo Yu) independently reviewed the abstracts of the
PT E
underlying eligible articles after initially filtering according to the inclusion and exclusion criteria mentioned above. Any disagreements were settled through consultation and determined by the senior author (Gao Chen).
CE
Data extraction and quality assessment We conducted this meta-analysis according to the PRISMA guidelines
AC
(Supplemental TableⅠ). Three authors (Liansheng Gao, Tao Li and Shenglong Cao) separately reviewed and extracted data with a unified standard before an agreement was reached. The data extracted from each article contained the following information: names of authors, year of publication, country, study design, number of patients and segments involved, gender and age of patients, stages of CVST, type of MRV techniques, field strength, enhancement or not, counting method of the numbers of thrombosis and the figures of TP, FP, FN, and TN. There are two issues should be noted in our study. First, there were two methods of counting existing in the included articles. Some took count of the diseased cerebral 6
ACCEPTED MANUSCRIPT venous sinus segments to access statistical data, while others counted the number of patients. In this review, we combined these two methodologies to acquire data and named it “case”; further subgroup analysis divided this combination. Second, some articles may contain several cohorts, which were distinguished by different letters in our review. All data was re-checked by a third author (Weilin Xu) and any
PT
disagreements were disposed by another reviewer (Gao Chen). The document quality included was assessed using the Quality Assessment Tool for
RI
Diagnostic Accuracy Studies version 2 (QUADAS-2) recommended by Cochrane [19]. The quality assessment was performed and the bias risk map was drawn with
SC
Review Manager 5.3 [19]. Statistical analysis
NU
This meta-analysis was conducted following the standard methods recommended for diagnostic accuracy meta-analysis [20, 21].
MA
First, the heterogeneity level among cohorts due to the threshold effect was determined using threshold analysis. Spearman correlation coefficient between the
D
logit of sensitivity and the logit of (1−specificity) was used to indicate the threshold
PT E
effect. A high correlation with p< 0.05 means an obvious threshold effect. Second, the inconsistency index (I2) of the diagnostic odds ratio (DOR) was used to assess the degree of heterogeneity among cohorts due to the non-threshold effect. I2
CE
ranging from 0% to 40% represents no heterogeneity; 30% to 60% represents moderate heterogeneity; 50% to 90% represents substantial heterogeneity; and 75% to
AC
100% represents considerable heterogeneity according to the recommendations of Cochrane. The fixed-effects coefficient binary regression model was used if no obvious heterogeneity was discovered [19, 21]; otherwise, the random-effects coefficient binary regression model was applied to summarize the data. Meta-regression, subgroup analysis, and sensitivity analysis were also used to explore probable sources of heterogeneity. Each subgroup was supposed to comprise at least three cohorts with homogeneous characteristics following the same parameters. A Z-test was performed to compare subgroups and a value of p<0.05 indicates substantial differences. 7
ACCEPTED MANUSCRIPT Third, fixed- or random-effects models were used to compute the pooled data of the sensitivity (SEN), specificity (SPE), positive likelihood ratio (LR+), negative likelihood ratio (LR-), and DOR with 95% confidence intervals (CIs) according to the analysis mentioned above. A value of 0.5 was added to each cell in the table if any of the figures of TP, FP, FN, TN appeared to be zero, avoiding the SENs or SPEs being
PT
100%. Fourth, the summary receiver-operating characteristic curve (SROC) was drawn,
RI
and the area under the curve (AUC) and Q* index was calculated accordingly. The diagnostic accuracy was evaluated as follows: the value of AUC between 51% and 70%
SC
indicates low accuracy, while 71% to 90% indicates moderate accuracy and 90% or greater indicates high accuracy.
NU
Finally, Deek’s funnel plot and linear regression methods were used to assess the publication bias, visually with the X-axis indicating DOR and the Y-axis indicating
MA
1/root (ESS) [22]. The value of p<0.05 suggests significant asymmetry, which indicates publication bias [23]. The trim and fill method was applied to rectify the
D
funnel plot asymmetry caused by publication bias [24].
PT E
The statistical analysis was performed with Stata statistical software 14.0 (StataCorp LP, College Station, TX) and Meta-Disc statistical software version 1.4
Results
CE
[19, 21].
Article selection and the characteristics
AC
The retrieved records from Embase, PubMed and Chinese Biomedical databases included 378, 650 and 439 articles, respectively. No additional results were acquired from other sources. A total of 1116 articles remained after duplicates were removed. After titles and abstracts were screened, 101 potentially eligible records were chosen for further full text evaluation. After removal of reviews, case reports, articles whose data could not be extracted and all overlapping data, 12 articles containing 27 cohorts with 1933 cases satisfied all inclusion and exclusion criteria and were enrolled in the final meta-analysis [11, 14, 25-34] The document selection procedure following the PRISMA flow diagram is shown 8
ACCEPTED MANUSCRIPT in Fig 1. The characteristics of included articles are summarized in Table 2. The age span of patients was 12 to 80 years. The number of male patients was fewer than female patients (Male/Female=183/220). Twenty three cohorts were retrospective while 4 were prospective. The sample size of each cohort ranged from 16 to 300 cases. 12 cohorts counted the diseased cerebral venous sinus segments to access statistical
PT
data, while other 15 cohorts counted the number of patients. Three types of MRV techniques were analyzed in our review. There were 11 cohorts using 3D
RI
contrast-enhanced (CE) MRV to diagnose CVST, while 13 used TOF MRV and 3 used PC MRV. The field strength included 0.35T, 1.5T and 3.0T with or without
SC
enhancement. DSA was used as gold standard in 12 cohorts while other 15 cohorts used combination of multiple imaging methods and clinical follow up as reference
NU
standard. The phase of CVST could not be distinguished in most studies. The methodological quality graph and methodological quality summary graph are
MA
shown in Fig 2. Most studies had low or unclear risk of bias which meant the study qualities were acceptable.
D
Quantitative synthesis
PT E
The Spearman correlation coefficient was 0.121 (p=0.547) and the I2 of the diagnostic odds ratio (DOR) was 58.8% in the overall analysis. Meta-regression analysis revealed that two parameters, technical types of MRV and the methods of
CE
counting cases, contribute to the heterogeneity due to the non-threshold effect with p-values equaling 0.011 and 0.013, respectively.
AC
Among all 27 cohorts, the pooled SEN was 0.86 (95% CI: 0.83, 0.89); the pooled SPE was 0.94 (95% CI: 0.93, 0.95); the pooled LR+ was 8.64 (95% CI: 4.93, 15.14); the pooled LR- was 0.17 (95% CI: 0.12, 0.25); the pooled DOR was 75.24 (95% CI: 38.33, 147.72). The forest plots of SEN and SPE of the 27 included cohorts are shown in Fig3. The AUC for the SROC was 0.9472 (95% CI: 0.9229, 0.9715) with a Q* index equal to 0.8868, which is shown in Fig 4. Subgroup analysis was conducted according to five parameters and the results are shown in Table 3. Twelve cohorts took count of the number of diseased cerebral venous sinus segments to calculate diagnostic indexes. The pooled SEN and SPE 9
ACCEPTED MANUSCRIPT were 0.84 (95% CI: 0.80, 0.88) and 0.96 (95% CI: 0.95, 0.97), respectively. The corresponding LR+ and LR- were 19.00 (95% CI: 9.45, 38.21) and 0.14 (95% CI: 0.06, 0.31), respectively. The summary DOR was 198 (95% CI: 82.74, 473.82) and the AUC was 0.9790. Fifteen other cohorts counted the number of patients to calculate diagnostic indexes. The pooled SEN and SPE were 0.88 (95%CI: 0.83, 0.92)
PT
and 0.83 (95% CI: 0.79, 0.87), respectively. The corresponding LR+ and LR- were 3.69 (95% CI: 2.19, 6.23) and 0.21 (95% CI: 0.14, 0.34), respectively. The summary
RI
DOR was 23.61 (95% CI: 10.81, 51.55) and the AUC was 0.9096.The Z-test revealed that there was a significant statistical difference between these two parameters
SC
(Pinteraction= 0.011).
3D CE MRV was applied in 11 cohorts. The pooled SEN and SPE were: 0.85 (95%
NU
CI: 0.80, 0.89) and 0.98 (95% CI: 0.97, 0.99), respectively. The corresponding LR+ and LR- were 32.13 (95% CI: 12.18, 84.79) and 0.12 (95% CI: 0.05, 0.31),
MA
respectively. The summary DOR was 328.92 (95% CI: 113.38, 954.26) and the AUC was 0.9858. TOF MRV was applied in 13 cohorts. The pooled SEN and SPE were:
D
0.85 (95% CI: 0.79, 0.89) and 0.88 (95% CI: 0.84, 0.91), respectively. The
PT E
corresponding LR+ and LR- were 4.12 (95% CI: 2.34, 7.25) and 0.26 (95% CI: 0.15, 0.44), respectively. The summary DOR was 26.22 (95% CI: 11.22, 61.27) and the AUC was 0.9061. PC MRV was applied in 3 cohorts. The pooled SEN and SPE were:
CE
0.89 (95% CI: 0.82, 0.95) and 0.88 (95% CI: 0.83, 0.92), respectively. The corresponding LR+ and LR- were 4.62 (95% CI: 1.02, 21.04) and 0.13 (95% CI: 0.08,
AC
0.23), respectively. The summary DOR was 35.01 (95% CI: 7.06, 173.00) and the AUC was 0.9525.The Z-test revealed that there was a significant statistical difference between 3D CE MRV and TOF MRV (Pinteraction=0.006), and between 3D CE MRV and PC MRV (Pinteraction=0.045), with no statistical significant difference between TOF MRV and PC MRV (Pinteraction=0.132). The results of subgroup analysis based on study design (prospective versus retrospective), reference standard (DSA only versus not DSA) and field strength (3.0T versus 1.5T) are shown in Table 3. Other subgroups were ineligible for subgroup analysis because of insufficient data. 10
ACCEPTED MANUSCRIPT The results of the sensitivity analysis showed that no matter which single article was excluded, the heterogeneity was not obviously decreased, and no obvious changes were found in the pooled DOR (Table 4). In addition, after removing the four articles with relatively high risk of bias [14, 25, 26, 32], the heterogeneity did not significantly decrease (I2=60.2%) and there was no significant change in the pooled
PT
DOR (DOR=47.86, 95% CI: 16.86-135.92, p=0.475). Heterogeneity analysis
RI
Moderate heterogeneity among cohorts was found in the summary analysis of DOR with I2=58.8% and the results of meta-regression analysis, subgroup analysis and
SC
sensitivity analysis are displayed above. Publication bias
NU
The results indicate a publication bias among included cohorts (p=0.02, Fig 5A). The results of the trim and fill method indicated that there was no significant
MA
statistical difference in the pooled effect value and their CIs before and after five additional cohorts were appended (The p-values were both equal to 0). The filled
D
funnel plot shown in Fig 5B was symmetric, which also indicated no publication bias.
PT E
Discussion
CVST is a rare but lethal cerebrovascular disease with unclear etiology. The clinical manifestations of CVST are complex and nonspecific. Reasonable selection of the
CE
examination methods is essential for the early diagnosis and proper treatment of CVST, which may remarkably improve the prognosis [35]. MRV is non-invasive and
AC
could display the configuration of cerebral veins and sinuses better than conventional CT and MRI, which make it a promising imaging modality for the diagnosis of CVST [36].
This meta-analysis summarizes the diagnostic performance of commonly used MRV types in identifying CVST. The result of a threshold analysis implied no obvious threshold effect in our study. The pooled SEN was 0.86 and a higher value of pooled SPE was 0.94, which meant that MRV had low rate of missed diagnosis and an even lower rate of misdiagnosis in confirming CVST. Furthermore, the DOR synthesized the sensitivity and specificity data into a single value to better reflect the 11
ACCEPTED MANUSCRIPT diagnostic accuracy [37]. Our results showed that the DOR was 75.24 for the overall analysis, which indicated high diagnostic accuracy, while the I2 of the DOR was 58.8%, which revealed moderate heterogeneity among the cohorts. LR is another meaningful index used in the evaluation of the accuracy of diagnostic tests. It combines the advantages of sensitivity, specificity, positive predictive value, and
PT
negative predictive value, and is not affected by the prevalence rate. LR > 10 or LR < 0.1 indicates a high level of accuracy [19]. The pooled positive and negative LR of
RI
our study was 8.64 and 0.17, respectively, indicating good diagnostic accuracy. The AUC is an intuitive plot to report the ability of a diagnostic test and could be used to
SC
compare the diagnostic accuracy among the subgroups. Our results showed that the AUC was 0.9472, suggesting great diagnostic performance independent from diverse
NU
parameters such as type of MRV, computing methods, study design, and reference standards.
MA
Meta-regression found that the heterogeneity mainly originated from two parameters, the technical types of MRV (p=0.011) and the method of counting cases
D
(p=0.013). Different technical types of MRV had different diagnostic abilities, which
PT E
caused the heterogeneity in the pooled analysis. As mentioned above, two methods of counting existed in the included cohorts. Because one patient may have several cerebral venous sinus thrombi, the number of diseased venous segments was much
CE
greater than the number of patients. Combining these two kinds of data may induce relatively large differences among the sample sizes and yield the heterogeneity. The
AC
pooled SENs and SPEs show obvious differences between these two subgroups, which is possibly due to the different constituent ratios of 3D CE MRV and non-3D CE MRV between the two subgroups. Sensibility analysis is another method to find the origin of heterogeneity; in addition, it can also evaluate the stability of the meta-analysis [38]. In our study, sensibility analysis suggested that the results of the meta-analysis were not over-reliant on some articles and the conclusion was stable. Three types of MRV were used to identify CVST and all showed high accuracy, but with some differences. The DOR of 3D CE MRV was higher than those of the 12
ACCEPTED MANUSCRIPT other two. The Z-test revealed the diagnostic performance of 3D CE MRV was better than that of TOF MRV (Pinteraction=0.006) and PC MRV (Pinteraction=0.045), while TOF MRV and PC MRV showed no statistical differences (Pinteraction=0.132).These results were coincident with previous studies [14, 15]. 3D CE MRV has high spatial resolution and contrast and thus can clearly show the normal intracranial venous
PT
system and easily diagnose CVST [39-41]. The LR+ of 3D CE MRV was 32.13, which was much higher than those of the other two types of MRV, which suggested
RI
that TOF and PC MRV may suffer from false positives and have higher rate of misdiagnosis than 3D CE MRV, which was coincident with previous studies [14, 15].
SC
Ayanzen et al. reported TOF has a variety of image artifacts [42]. Liauw et al. thought PC MRV was more susceptible to motion artifacts [43]. This may be the main cause
NU
of misdiagnosis. However, the LR− seemed similar among these subgroups, indicating that they had a similar rate of missed diagnosis. As a result, when these
MA
three types of MRV are negative for the diagnosis of CVST, a negative conclusion cannot be made. In this situation, the DSA should be used so as not to miss the
D
diagnosis. Above all, 3D CE MRV leads to fewer false positives and shows obvious
PT E
advantages. The contrast agent is rich in the blood while absent in thrombosis, which increases the contrast between the blood and thrombus. However, the enhanced MRV is inappropriate for critical patients and TOF or PC MRV could be an alternative.
CE
Moreover, it should be indicated that MRV seems to be powerless in diagnosing small cortical vein thrombosis and partial occlusion [42].
AC
There were 8 cohorts utilizing 3.0T MRV and 14 others adopting 1.5T MRV. AUCs for the SROC of these two techniques were 0.9653 and 0.9159, respectively, which implied that they both have a high level of diagnostic accuracy. Compared to 1.5T MRV, 3.0T MRV has clearer resolution of the image and shorter scanning time, which together increase the efficiency in clinical use [44]. 3.0T MRV is good at identifying subtle lesions and anatomical structures, especially for minor cortical veins [45]. Although the Z-test revealed no significant difference in diagnostic performance between these two subgroups (Pinteraction=0.117), the 3.0T MRV has better performance than 1.5T MRV in clinical application. Further evaluation is needed. 13
ACCEPTED MANUSCRIPT Two reference standards were used to assess CVST, one was DSA only and the other was the combination of multiple imaging and clinical follow up. Whiting et al. thought the adoption of inconsistent reference standards may potentially overestimate the diagnostic accuracy of the test [19]. Conventionally, the gold standard for diagnosing CVST is DSA; however, its invasiveness, which may increase the risk of
PT
mortality, and the radioactivity, which may be a contraindication for special persons like pregnant woman, limits its application. Fortunately, there was no statistically
RI
significant difference in diagnostic performance between these two subgroups (Pinteraction=0.932). Thus, the use of a combination of multiple imaging and clinical
SC
follow up are an acceptable alternative which partly overcome the shortcomings of DSA.
NU
Deek’s funnel plot showed obvious asymmetry suggesting possible publication bias (p=0.02), which means the diagnostic value may be overemphasized. This is possibly
MA
due to our exclusion of unpublished grey literature. However, further trim and fill analysis indicated that the pooled effect quantities estimate changed little after five
D
cohorts were filled, which indicated our results were stable and the publication bias
PT E
had little effect on our results. The trim and fill analysis was in fact a type of sensitivity analysis method. Peters et al. thought trim and fill analysis was conducive
Limitations
CE
to reducing the bias of pooled effect quantities [46].
Despite the excellent diagnostic performance and effectiveness of MRV in
AC
confirming CVST, there were some limitations in our meta-analysis. First, the cohorts included showed great diversity in study design, computing method, technical types of MRV, reference standard, patient selection, and sample sizes. In addition, some included articles also show relatively high risk of bias and these factors may potentially contribute to heterogeneity. As a result, it should be prudent to interpret the results. Second, a subgroup analysis of important clinical significance could not be conducted due to insufficient data extracted from the included cohorts for several parameters, such as the phases of CVST, other reference standards such as CTV, and 14
ACCEPTED MANUSCRIPT different segments of the cerebral venous sinus. More studies using these parameters in the comparison are needed. Finally, the data of our meta-analysis mainly came from countries in Asia and Europe, and only published articles with full text in English and Chinese were included, which may omit some eligible articles either unpublished or reported in
PT
other languages from other countries. Conclusion
RI
In summary, our meta-analysis suggested that MRV has a high level of diagnostic accuracy for identifying CVST, although the results should still be interpreted and
SC
promoted prudently. In addition, more prospective studies with high quantity and large sample sizes are needed to confirm our conclusions. With the continued
NU
improvement of imaging techniques and our understanding of CVST, MRV will have
MA
broad application prospects for diagnosing CVST.
References
D
[1] Stam J. Thrombosis of the cerebral veins and sinuses. N Engl J Med.
PT E
2005;352:1791-8.
[2] Coutinho JM. Cerebral venous thrombosis. J Thromb Haemost. 2015;13 Suppl 1:S238-44.
CE
[3] Bousser MG, Ferro JM. Cerebral venous thrombosis: an update. Lancet Neurol. 2007;6:162-70.
AC
[4] Weimar C. Diagnosis and treatment of cerebral venous and sinus thrombosis. Curr Neurol Neurosci Rep. 2014;14:417. [5] Martinelli I, Bucciarelli P, Passamonti SM, Battaglioli T, Previtali E, Mannucci PM. Long-term evaluation of the risk of recurrence after cerebral sinus-venous thrombosis. Circulation. 2010;121:2740-6. [6] Saposnik G, Barinagarrementeria F, Brown RD, Jr., Bushnell CD, Cucchiara B, Cushman M, et al. Diagnosis and management of cerebral venous thrombosis: a statement
for
healthcare
professionals
from
the
American
Association/American Stroke Association. Stroke. 2011;42:1158-92. 15
Heart
ACCEPTED MANUSCRIPT [7] Sassi SB, Touati N, Baccouche H, Drissi C, Romdhane NB, Hentati F. Cerebral Venous Thrombosis: A Tunisian Monocenter Study on 160 Patients. Clinical and applied thrombosis/hemostasis : official journal of the International Academy of Clinical and Applied Thrombosis/Hemostasis. 2017;23:1005-9. [8] Wei Y, Deng X, Sheng G, Guo XB. A rabbit model of cerebral venous sinus
PT
thrombosis established by ferric chloride and thrombin injection. Neuroscience letters. 2018;662:205-12.
RI
[9] Masuhr F, Mehraein S, Einhaupl K. Cerebral venous and sinus thrombosis. Journal of neurology. 2004;251:11-23.
thrombosis. Clinical radiology. 2002;57:449-61.
SC
[10] Connor SE, Jarosz JM. Magnetic resonance imaging of cerebral venous sinus
NU
[11] Sari S, Verim S, Hamcan S, Battal B, Akgun V, Akgun H, et al. MRI diagnosis of dural sinus - Cortical venous thrombosis: Immediate post-contrast 3D GRE
MA
T1-weighted imaging versus unenhanced MR venography and conventional MR sequences. Clinical neurology and neurosurgery. 2015;134:44-54.
PT E
neurology. 2000;247:252-8.
D
[12] Bousser MG. Cerebral venous thrombosis: diagnosis and management. Journal of
[13] Sun Y, Zheng DY, Ji XM, Weale P, Wu H, Jiang LD, et al. Diagnostic performance of magnetic resonance venography in the detection of recanalization in
CE
patients with chronic cerebral venous sinus thrombus. Chinese medical journal. 2009;122:2428-32.
AC
[14] Klingebiel R, Bauknecht HC, Bohner G, Kirsch R, Berger J, Masuhr F. Comparative evaluation of 2D time-of-flight and 3D elliptic centric contrast-enhanced MR venography in patients with presumptive cerebral venous and sinus thrombosis. Eur J Neurol. 2007;14:139-43. [15] Pui MH. Cerebral MR venography. Clinical imaging. 2004;28:85-9. [16] Farb RI, Scott JN, Willinsky RA, Montanera WJ, Wright GA, terBrugge KG. Intracranial venous system: gadolinium-enhanced three-dimensional MR venography with auto-triggered elliptic centric-ordered sequence--initial experience. Radiology. 2003;226:203-9. 16
ACCEPTED MANUSCRIPT [17] Wetzel SG, Law M, Lee VS, Cha S, Johnson G, Nelson K. Imaging of the intracranial venous system with a contrast-enhanced volumetric interpolated examination. European radiology. 2003;13:1010-8. [18] Mermuys KP, Vanhoenacker PK, Chappel P, Van Hoe L. Three-dimensional venography of the brain with a volumetric interpolated sequence. Radiology.
PT
2005;234:901-8. [19] Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al.
RI
QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Annals of internal medicine. 2011;155:529-36.
SC
[20] Deville WL, Buntinx F, Bouter LM, Montori VM, de Vet HC, van der Windt DA,
medical research methodology. 2002;2:9.
NU
et al. Conducting systematic reviews of diagnostic studies: didactic guidelines. BMC
[21] Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc: a
MA
software for meta-analysis of test accuracy data. BMC medical research methodology. 2006;6:31.
D
[22] Kato T, Shinoda J, Nakayama N, Miwa K, Okumura A, Yano H, et al. Metabolic
11C-choline
PT E
assessment of gliomas using 11C-methionine, [18F] fluorodeoxyglucose, and positron-emission
2008;29:1176-82.
tomography.
AJNR
Am
J
Neuroradiol.
CE
[23] Tsuyuguchi N, Takami T, Sunada I, Iwai Y, Yamanaka K, Tanaka K, et al. Methionine positron emission tomography for differentiation of recurrent brain tumor
AC
and radiation necrosis after stereotactic radiosurgery--in malignant glioma. Annals of nuclear medicine. 2004;18:291-6. [24] Duval S, Tweedie R. Trim and fill: A simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56:455-63. [25] Meckel S, Reisinger C, Bremerich J, Damm D, Wolbers M, Engelter S, et al. Cerebral venous thrombosis: diagnostic accuracy of combined, dynamic and static, contrast-enhanced 4D MR venography. AJNR Am J Neuroradiol. 2010;31:527-35. [26] Ye J, Zhang XJ, Li J, Zhong Q, Hong JF, Wang SR. Case-control study of cerebral venous sinus thrombosis between MRV and DSA. Chinese Journal of 17
ACCEPTED MANUSCRIPT Neurosurgical Disease Research. 2016;15(6):485-9. [27] Jalli R, Zarei F, Farahangiz S, Khaleghi F, Petramfar P, Borhani-Haghighi A, et al. The Sensitivity, Specificity, and Accuracy of Contrast-Enhanced T1-Weighted Image, T2*-Weighted Image, and Magnetic Resonance Venography in Diagnosis of Cerebral Venous Sinus Thrombosis. Journal of stroke and cerebrovascular diseases :
PT
the official journal of National Stroke Association. 2016;25:2083-6. [28] Li XY, Yang J, Sun JB, Ni L. Comparison of TOF MRV and CE MRV in the
RI
diagnosis of cerebral venous sinus thrombosis. Chinese Community Doctors. 2012;14(317):244.
SC
[29] Ho JS, Rahmat K, Ramli N, Fadzli F, Chong HT, Tan CT. Cerebral venous sinus thrombosis: Comparison of multidetector computed tomography venogram (MDCTV)
NU
and magnetic resonance venography (MRV) of various field strengths. Neurology Asia. 2012;17(4):281-91.
MA
[30] Yi BN, Ba HT, Wang J, Zhang DQ. The value of magnetic resonance imaging in the diagnosis of intracranial venous sinus thrombosis. Guide of Chinese Medicine.
D
2012;10(21):149-50.
PT E
[31] Ju KJ, Wang LJ, Cao H. Diagnostic value of three-dimensional enhanced magnetic resonance angiography in patients with intracranial venous sinus thrombosis. Guangdong Medical Journal. 2011;32(19):2579-81.
CE
[32] Yang ML, Li T, Huang YH, Qu H. Early imaging features of cerebral venous sinus thrombosis: a report of 62 cases. Chinese Journal of Postgraduates of Medicine.
AC
2009;32(10):59-61.
[33] Ma BW, Huang LY, Du YH, Chen GS, Li YJ, Hou XL. Clinical Characters and Misdiagnosis Analysis in Patients with Cerebral Vein and Sinus Thrombosis. Journal of Ningxia Medical College. 2008;30(6):717-9. [34] Liang L, Korogi Y, Sugahara T, Onomichi M, Shigematsu Y, Yang D, et al. Evaluation of the intracranial dural sinuses with a 3D contrast-enhanced MP-RAGE sequence: prospective comparison with 2D-TOF MR venography and digital subtraction angiography. AJNR Am J Neuroradiol. 2001;22:481-92. [35] Qu H, Yang M. Early imaging characteristics of 62 cases of cerebral venous 18
ACCEPTED MANUSCRIPT sinus thrombosis. Exp Ther Med. 2013;5:233-6. [36] Zafar A, Ali Z. Pattern of magnetic resonance imaging and magnetic resonance venography changes in cerebral venous sinus thrombosis. Journal of Ayub Medical College, Abbottabad : JAMC. 2012;24:63-7. [37] Nakajima T, Kumabe T, Kanamori M, Saito R, Tashiro M, Watanabe M, et al.
PT
Differential diagnosis between radiation necrosis and glioma progression using sequential proton magnetic resonance spectroscopy and methionine positron emission
RI
tomography. Neurologia medico-chirurgica. 2009;49:394-401.
[38] Wang J, Gao J, He J. [Diagnostic value of ProGRP and NSE for small cell lung
SC
cancer: a meta-analysis]. Zhongguo fei ai za zhi = Chinese journal of lung cancer. 2010;13:1094-100.
NU
[39] Rollins N, Ison C, Reyes T, Chia J. Cerebral MR venography in children: comparison of 2D time-of-flight and gadolinium-enhanced 3D gradient-echo
MA
techniques. Radiology. 2005;235:1011-7.
[40] Kirchhof K, Welzel T, Jansen O, Sartor K. More reliable noninvasive
D
visualization of the cerebral veins and dural sinuses: comparison of three MR
PT E
angiographic techniques. Radiology. 2002;224:804-10. [41] Lettau M, Sartor K, Heiland S, Hahnel S. 3T high-spatial-resolution contrast-enhanced MR angiography of the intracranial venous system with parallel
CE
imaging. AJNR Am J Neuroradiol. 2009;30:185-7. [42] Ayanzen RH, Bird CR, Keller PJ, McCully FJ, Theobald MR, Heiserman JE.
AC
Cerebral MR venography: normal anatomy and potential diagnostic pitfalls. AJNR Am J Neuroradiol. 2000;21:74-8. [43] Liauw L, van Buchem MA, Spilt A, de Bruine FT, van den Berg R, Hermans J, et
al.
MR
angiography
of
the
intracranial
venous
system.
Radiology.
2000;214:678-82. [44] Hu HH, Haider CR, Campeau NG, Huston J, 3rd, Riederer SJ. Intracranial contrast-enhanced magnetic resonance venography with 6.4-fold sensitivity encoding at 1.5 and 3.0 Tesla. Journal of magnetic resonance imaging : JMRI. 2008;27:653-8. [45] Tomasian A, Salamon N, Krishnam MS, Finn JP, Villablanca JP. 3D 19
ACCEPTED MANUSCRIPT high-spatial-resolution cerebral MR venography at 3T: a contrast-dose-reduction study. AJNR Am J Neuroradiol. 2009;30:349-55. [46] Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Performance of the trim and fill method in the presence of publication bias and between-study heterogeneity.
PT
Statistics in medicine. 2007;26:4544-62.
Acknowledgements
RI
None.
SC
Conflict of interests
NU
The authors have no conflict of interests to declare.
Source of funding
MA
This research did not receive any specific grant from funding agencies in the public,
PT E
Authors’ contributions
D
commercial, or not-for-profit sectors.
This meta-analysis was designed by Liansheng Gao and Gao Chen, data collection and analysis were conducted by Weilin Xu, Tao Li, Feng Yan, Shenglong Cao,
CE
Xiaobo Yu and Hangzhe Xu, this article was checked by Gao Chen. This article was
AC
written by Liansheng Gao, Weilin Xu and Tao Li.
Figure Legends
Fig 1. Flow diagram of the article selection process. Fig 2. Methodological quality graph (A) and methodological quality summary graph (B) of each article using the QUADAS-2. Fig 3. Forest plots of the sensitivity (A) and specificity (B) with the 95% confidence interval (CI) of each cohort represented by the horizontal line through each circle of this meta-analysis. The pooled sensitivity and specificity for this meta-analysis is represented by diamonds. df=degrees of freedom. 20
ACCEPTED MANUSCRIPT Fig 4. Summary receiver-operating characteristic (SROC) curve of MRV for the diagnosis of CVST. AUC = area under the curve, SE = standard error. Fig 5. Deek’s funnel plot of publication bias, as determined by linear regression of the inverse root of effective sample sizes (ESS) on log diagnostic odds ratios (A) and filled funnel plot with pseudo 95% confidence limits (B). The filled cohorts for this
PT
meta-analysis are represented by squares. Table 1. Search terms and strategies of the accuracy of MRV in diagnosing CVST in
RI
different databases.
Table 2. Characteristics of cohorts included in the meta-analysis of MRV for the
SC
diagnosis of CVST.
Table 3. Subgroup analyses of diagnostic accuracy variables.
AC
CE
PT E
D
MA
accuracy of MRV in diagnosing CVST.
NU
Table 4. Sensibility analysis of each article included in the meta-analysis of the
21
ACCEPTED MANUSCRIPT Table 1 Search terms and strategies of the accuracy of MRV in diagnosing CVST in different databases Database
Search strategy
PubMed
((((((CVST) OR cerebral venous sinus thrombosis) OR cerebral venous thrombosis) OR cranial venous sinus thrombosis) OR dural sinus thrombosis)) AND (MRV or MR venography or MR angiography or magnetic resonance venography)
CBM
(((((cerebral venous sinus thrombosis) OR cranial venous sinus thrombosis) OR cerebral sinus thrombosis) OR CVST)) AND (MRV or MR venography or MR angiography or magnetic resonance venography)
Embase
('cerebral sinus thrombosis':ab,ti OR 'cvst':ab,ti OR 'sinus thrombosis':ab,ti) AND ('magnetic resonance venography':ab,ti OR 'mrv':ab,ti OR 'magnetic resonance angiography':ab,ti)
PT
Abbreviations: CBM, Chinese Biomedical databases; MRV, magnetic resonance venography; CVST, cerebral venous sinus
AC
CE
PT E
D
MA
NU
SC
RI
thrombosis.
22
ACCEPTED MANUSCRIPT
Table 2 Characteristics of cohorts included in the meta-analysis of MRV for the diagnosis of CVST Computi Author
Year
Study
No. of
No. of
design
patients
cases*
Country
Age(mean)
Final diagnosis
Reference
(CVST)
standard
M/F
Phase
ng
Field Techniques
T P
methods Ye J (a) [26]
1
16
10
Ye J (b)
1
16
10
6
80
Ye J (c) 2016
China
acute
22
R
31.8
C S U DSA
Ye J (d)
6
48
16
Ye J (e)
8
128
25
Ye J (f)
8
32
11
Jalli R (a) [27]
46
46
22
chronic
N A
General MRI
TP
FP
FN
TN
segment
3D CE MRV
3.0T+
10
0
0
6
segment
TOF MRV
3.0T-
10
2
0
4
segment
3D CE MRV
3.0T+
22
1
0
57
segment
TOF MRV
3.0T-
16
4
0
28
segment
3D CE MRV
3.0T+
14
0
11
103
segment
TOF MRV
3.0T-
11
2
0
19
patient
PC MRV
1.5T-
20
12
2
12
patient
TOF MRV
1.5T-
9
5
0
8
segment
PC MRV
3.0/1.5T-
60
19
7
214
3.0/1.5T+
62
0
5
233
I R
subacute
9/6
strength (enhanced+)
and CT and CTV
2016
Iran
R
Jalli R (b)
22
Sari S (a) [11] 2015
Turkey
Li XY (a) [28]
30
300
30
300
China
Li XY (b)
16
2012
Malaysia
C A
R
27
16 16
27
D E
PT
E C
P
M 9
40.1
16 2012
14/54
22
R
Sari S (b)
Ho JS [29]
35
31.2
acute
and clinical follow up
67
17/13
DSA
NA
67
3D-Volumetric segment GRE MRV
11
DSA or clinical
6/10
patient
3D CE MRV
3.0T+
10
1
1
4
patient
TOF MRV
3.0T-
9
3
2
2
NA
patient
TOF MRV
0.35/1.5/3.0-
10
3
0
14
NA
patient
PC MRV
1.5-
12
1
2
5
NA
patient
3D CE MRV
NA+
17
9
1
101
NA 11
follow up General MRI or
40.9
19/8
10
CT or clinical follow up
Yi BN [30]
2012
China
R
20
20
31
8/12
14
Ju KJ [31]
2011
China
R
128
128
42.6
58/70
18
DSA General MRI or DSA
23
ACCEPTED MANUSCRIPT
Meckel S (a) [25]
39
39
20
patient
Meckel S (b)
39
39
20
patient
TOF MRV
1.5T-
16
4
4
15
1.5T+
18
0
2
19
1.5T-
28
5
6
117
1.5T+
33
1
1
121
1.5T-
10
1
9
58
1.5T+
7
1
12
58
dynamic/static 3D MRV(4D)
General MRI or Meckel S (c)
39
Meckel S (d)
39
156
34
156
CTV or clinical
34
I R
NA
follow up Meckel S (e)
2010
Switzerland
R
Meckel S (f)
39
78
39
78
49.1
17/22
2009
China
R
21
21
NA
NA
Ma BW [33]
2008
China
R
51
51
NA
NA
25 2007
Germany
25
Liang LX (a) [34]
35 Japan
D E
25
T P E
P
35
49
Liang LX (b)
35
35
19/6
U N
19
A M 51
7
40.6
Klingebiel R (b)
2001
25
R
SC
19
Yang ML [32]
Klingebiel R (a) [14]
19
7
DSA
segment segment
3D MRV(4D) TOF MRV
dynamic/static
segment 3D MRV(4D) patient
TOF MRV
1.5T-
19
1
1
0
All phases
patient
TOF MRV
NA-
48
0
3
0
patient
3D CE MRV
1.5T+
6
0
1
18
patient
TOF MRV
1.5T-
5
8
2
10
patient
TOF MRV
1.5T-
6
2
6
21
1.5T+
10
0
2
23
General MRI or DSA
DSA and CTA and clinical
NA
follow up
DSA 12
TOF MRV
dynamic/static
NA
12
16/19
T P
segment
CT or CTA or
NA
3D MP RAGE patient MRV
C C
Abbreviations: R, respective; P, prospective; M, male; F, female; DSA, digital subtraction angiography; MRI, magnetic resonance imaging; DSA, digital subtraction angiography; MRV, magnetic resonance venography; CTA, computed tomography angiography; CVST, cerebral venous and sinus thrombosis; NA, not available; D, dimensional; CE, contrast-enhanced; TOF, time of flow; GRE, gradient echo; MP RAGE,
A
magnetization-prepared rapid acquisition with gradient echo.
* There were two methods of counting existing in the included articles. Some took count of the diseased cerebral venous sinus segments to access statistical data, while others counted the number of patients. In this review, we combined these two methodologies to acquire data and named it “case”.
24
ACCEPTED MANUSCRIPT
Table 3 Subgroup analyses of diagnostic accuracy variables Threshold Category
Cohorts (n)
Cases
Effects
SEN (95 % CI)
SPE (95 % CI)
LR+ (95 % CI)
LR- (95 % CI)
DOR (95 % CI)
0.86(0.83,0.89)
0.94(0.93,0.95)
8.64(4.93,15.14)
0.17(0.12,0.25)
75.24(38.33,147.72)
T P
(p-value) Overall
27
1933
0.547
I R
Computing Methods segment
12
1388
0.301
0.84(0.80,0.88)
0.96(0.95,0.97)
19.00(9.45,38.21)
patient
15
545
0.484
0.88(0.83,0.92)
0.83(0.79,0.87)
3.69(2.19,6.23)
3D CE MRV(A)
11
1001
0.670
0.85(0.80,0.89)
0.98(0.97,0.99)
32.13(12.18,84.79)
TOF MRV(B)
13
566
0.667
0.85(0.79,0.89)
0.88(0.84,0.91)
4.12(2.34,7.25)
PC MRV(C)
3
366
0.667
0.89(0.82,0.95)
0.88(0.83,0.92)
Retrospective
23
1831
0.451
0.86(0.83,0.89)
Prospective
4
102
1.000
0.76(0.61,0.87)
C S U
Study design
PT
Reference Standard DSA only
12
481
0.812
not DSA
15
1452
0.832
Field strength 3.0T
8
352
1.5T
14
775
E C
C A
D E
0.94(0.92,0.95) 0.88(0.77,0.95)
Pinteraction
0.9472(0.0124)
0.011
198(82.74,473.82)
0.9790(0.0096)
0.21(0.14,0.34)
23.61(10.81,51.55)
0.9096(0.0256)
0.12(0.05,0.31)
328.92(113.38,954.26)
0.9858(0.0088)
0.006 vs B
0.26(0.15,0.44)
26.22(11.22,61.27)
0.9061(0.0274)
0.132 vs C
4.62(1.02,21.04)
0.13(0.08,0.23)
35.01(7.06,173.67)
0.9525(0.0141)
0.045 vs A
8.87(4.90,16.04)
0.15(0.09,0.26)
84.83(40.40,178.12)
0.9539(0.0127)
4.65(1.03,21.05)
0.32(0.14,0.73)
17.40(3.31,91.45)
0.8641(0.0812)
Techniques
N A
M
0.14(0.06,0.31)
AUC (SE)
0.275
0.923
0.84(0.77,0.89)
0.93(0.89,0.95)
7.70(3.27,18.11)
0.20(0.10,0.38)
65.80(19.96,216.84)
0.9477(0.0235)
0.86(0.83,0.90)
0.94(0.92,0.95)
8.44(4.03,17.68)
0.16(0.08,0.31)
70.39(28.94,171.22)
0.9452(0.0174) 0.117
0.691
0.86(0.79,0.92)
0.94(0.90,0.97)
7.95(2.71,23.28)
0.11(0.03,0.38)
97.53(21.38,444.94)
0.9653(0.0196)
0.404
0.80(0.74,0.85)
0.92(0.89,0.94)
7.56(3.53,17.07)
0.25(0.15,0.41)
41.32(16.40,104.10)
0.9159(0.0274)
Abbreviations: SEN, sensitivity; SPE, specificity; LR, likelihood ratio; CI, confidence interval; DOR, diagnostic odds ratio; AUC, area under curve; SE, standard error; 3D, three-dimensional; CE, contrast-enhanced; MRV, magnetic resonance venography; TOF, time-of-flight; PC, phase-contrast; DSA, digital subtraction angiography.
25
ACCEPTED MANUSCRIPT Table 4 Sensibility analysis of each article included in the meta-analysis of the accuracy of MRV in diagnosing CVST I2 (%)
Author
DOR
Ye J [26]
59.66
28.05
126.89
64.4
Jalli R [27]
86.75
43.13
174.50
58.1
Sari S [11]
63.41
31.39
128.11
54.6
Li XY [28]
87.15
44.12
172.14
57.1
Ho JS [29]
75.04
37.52
150.08
60.3
Yi BN [30]
78.10
38.98
156.48
60.1
Ju KJ [31]
72.42
36.05
145.50
59.8
Meckel S [25]
59.68
25.66
138.78
56.3
Yang ML [32]
78.75
39.74
156.06
59.7
Ma BW [33]
77.83
39.19
154.54
60.0
Klingebiel R [14]
84.79
43.77
Liang LX [34]
80.57
39.98
Overall
75.24
38.33
SC
RI
PT
95 % CI
164.26
54.0
162.38
58.9
147.72
58.8
Abbreviations: MRV, magnetic resonance venography; CVST, cerebral venous and sinus thrombosis; DOR, diagnostic odds ratio;
AC
CE
PT E
D
MA
NU
CI, confidence interval.
26
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5