- Email: [email protected]
Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS Digestive and Liver Disease xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld
Review Article
Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis Tseng Yu-Jen a , Zeng Xiao-qing a , Chen Jie a , Li Na a , Xu Peng-ju b , Chen Shi-yao c,∗ a b c
Department of Gastroenterology, Zhongshan Hospital, Fudan University, Shanghai, China Department of Radiology, Zhongshan Hospital, Fudan University, Shanghai, China Department of Gastroenterology, Endoscopy Center, Evidence-based Medicine Center, Zhongshan Hospital, Fudan University, Shanghai, China
a r t i c l e
i n f o
Article history: Received 8 January 2016 Accepted 18 February 2016 Available online xxx Keywords: Computed tomography Diagnosis Gastroesophageal varices Meta-analysis
a b s t r a c t Aims: Gastroesophageal varices (GOV) is a common complication in patients with portal hypertension. We conducted a meta-analysis in attempt to evaluate the diagnostic accuracy of computed tomography (CT) as a noninvasive imaging tool for identifying GOV in reference to esophagogastroduodenoscopy (EGD). Methods: A systemic literature search of multiple databases were conducted to identify articles that investigated the diagnostic performance of CT for GOV, while employing EGD as reference standard. A 2 × 2 table was conducted according to the available published data for both esophageal varices (EV) and gastric varices (GV) as individual subgroups. The following indices were calculated: pooled sensitivity and specificity, positive and negative likelihood ratio, diagnostic odds ratio, and area under receiver operating characteristics. All statistical analyses were conducted via STATA13.0 and RevMan5.3. Results: A total of 11 studies were included in this meta-analysis, 10 articles evaluated the diagnostic accuracy of CT for EV (807 subjects) and 7 articles for GV (583 subjects). The pooled sensitivity and specificity for identifying EV were 0.896 (95% CI, 0.841–0.934) and 0.723 (95% CI, 0.644–0.791), respectively, with an AUROC of 0.86. The pooled sensitivity and specificity for identifying GV were 0.955 (95% CI, 0.903–0.980) and 0.658 (95% CI, 0.433–0.829), respectively, with an AUROC of 0.95. A subgroup analysis suggested varying CT technology could serve as a potential source of heterogeneity between included studies. A Deek’s funnel plot indicated a low probability for publication bias. Conclusion: Computed tomography could potentially replace EGD as a primary screening tool for diagnosing GOV, however results should be interpreted with caution given its suboptimal specificity. © 2016 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
1. Introduction Portal hypertension is a progressive complication secondary to intra-hepatic, pre-hepatic, or post-hepatic aetiology [1]. Liver cirrhosis being the more common intra-hepatic cause, affects roughly 1% of the population worldwide, with Asian and African countries heavily weighing on disease prevalence [2]. Portal hypertension is often associated with a series of complications including ascites, hepatic encephalopathy, and gastroesophageal varices. Among which, gastroesophageal varix haemorrhage is the most common gastroenterological emergency [3]. Approximately 50% of patients
∗ Corresponding author at: Department of Gastroenterology, Endoscopy Center, Evidence-based Medicine Center, Zhongshan Hospital, Fudan University, 180 Fenglin Road, Shanghai 0086200032, China. Tel.: +86 13601767310/15000223468; fax: +86 2164038472. E-mail address: [email protected] (S.-y. Chen).
with cirrhosis develop gastroesophageal varices and their presence is often correlated with disease severity [4]. Gastric and esophageal varices can occur concurrently or in solitary, with esophageal varices more prevalent in patients with portal hypertension. The prevalence of gastroesophageal haemorrhage is approximately 10–30% per year [5]. Esophageal varices occur in 30 to 40% of cirrhotic patients, while gastric varices occur in approximately 20%. However, GV rupture is associated with a higher mortality rate of up to 45% [1,6]. Risk factors for gastroesophageal variceal haemorrhage include size of varix (large >10 mm, medium 5–10 mm, and small <5 mm), Child Pugh’s score, and presence of red-spots [7]. The high morbidity and mortality rates associated with gastroesophageal variceal haemorrhage demands an early detection and prophylactic treatment for patients at risk for disease development. The American Association for the Study of Liver Diseases (AASLD) recommends patients to undergo screening esophagogastroduodenoscopy (EGD) for detecting esophageal and gastric varices
http://dx.doi.org/10.1016/j.dld.2016.02.007 1590-8658/© 2016 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS
2
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
when the diagnosis of cirrhosis is made. All varices are graded as small or large (>5 mm) and the presence or absence of red spots is duly noted. Patients with gastroesophageal varices should be assessed for risk of variceal haemorrhage and treated accordingly with prophylactic medication (-blockers) or minimally invasive preventive therapy such as (EVL or sclerotherapy) [4]. However, the point prevalence of esophageal varices requiring prophylaxis ranges from 15 to 25%, and even lower for gastric varices. The majority of patients who undergo EGD screening at the time of cirrhosis diagnosis either have no varices, or have small varices that do not require treatment [8]. Moreover, EGD is an invasive and expensive procedure that requires sedation and is poorly tolerated by patients due to associated discomfort during and after the procedure [9]. Computed tomography imaging could potentially replace EGD as a non-invasive, more tolerable, and inexpensive test in the accurate diagnosis and risk assessment of gastroesophageal varices. With advancements in radiological imaging techniques, CT application now allows for multiple rendering models, such as but not limited to, multidetector computed tomography (MDCT), volume rendering (VR), minimum intensity projection (CT-MIP) and shade surface display (SSD), which could provide a more infallible identification and assessment of gastric varices [9–11]
identify relevant articles on diagnostic accuracy of computed tomography for gastroesophageal varices in reference to esophagogastroduodenoscopy (EGD). A combination of the following search terms were used: (esophageal varices OR gastric varices OR gastroesophageal varices) AND (CT OR computed tomography OR angiography). Refer to Appendix 1 for detailed search strategy. 2 reviewers (YT and XZ) independently reviewed the title and abstract of studies in attempt to eliminate irrelevant articles, based on a priori established inclusion and exclusion criteria. 2.2. Inclusion and exclusion criteria Inclusion criteria are as follows: (1) diagnostic accuracy of CT imaging was assessed in reference to EGD, (2) data provided was sufficient to conduct a 2 × 2 table for analysis, (3) absence or presence of either gastric or esophageal varices were assessed. Exclusion criteria includes: (1) study population was limited to patients with gastroesophageal varices initially confirmed with EGD, (2) studies that only provided diagnostic rate, without calculable sensitivity and specificity. No language or article type restrictions were imposed. Additional reference articles were acquired through a manual search of computerized databases. 2.3. Quality assessment
2. Methods 2.1. Search strategy This meta-analysis was conducted according to the PRISMA statement [12]. A systematic search of MEDLINE, Embase, Web of Science, and Scopus was performed through August 2015 to
All included studies were subjected to the Quality Assessment for Studies Diagnostic Accuracy-2 (QUADAS-2) guideline and rated according to the 4 domains (Patient Selection, Index Test, Reference Standard, Flow and Timing) for risk of bias and sources of variation (applicability) assessment. Risk of bias was judged as “low”, “high”, or “unclear”, under the guidance of a series of 10 signalling
Fig. 1. Study Identification flowchart.
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
questions. Similarly, applicability was rated as “low”, “high”, or “unclear” [13].
3
3. Results 3.1. Characteristics of included studies
2.4. Statistical analysis This meta-analysis was conducted based on a bivariate binomial mixed model [14]. All statistical analyses were conducted with the midas module in STATA 13.0 (StataCorp, College Station, Texas) and RevMan5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). All studies were initially classified into two subgroups: esophageal varices or gastric varices, with 10 and 7 applicable studies respectively. A 2 × 2 contingency table was constructed for esophageal and gastric varices independently to calculate the following diagnostic indices: pooled sensitivity and specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), positive predictive value (PPV), negative predictive value (NPV) and diagnostic odds ratio (DOR). Summary receiver operating characteristics (SROC) and its corresponding area under the curve (AUC) were also calculated to estimate the overall diagnostic performance of computed tomography. Heterogeneity between studies were assessed using the I2 inconsistency test. I2 > 50% indicated presence of substantial heterogeneity, a subsequent subgroup analysis was carried out in attempt to identify potential covariates. Post-test probability was calculated with a presumed prior probability of 35% for EV and 20% for GV via Fagan’s monogram. A Deek’s funnel plot was used to detect publication bias. All the above mentioned parameters were conducted separately for the diagnostic accuracy of esophageal varices and gastric varices. P-value less than 0.05 were considered statistically significant.
A total of 11 publications were included in this meta-analysis, 10 studies were appropriate diagnostic test accuracy studies for esophageal varices and 7 for gastric varices. Six studies explored the diagnostic abilities of computed tomography for both esophageal and gastric varices [11,15–19]. A detailed flowchart of study identification is shown in Fig. 1. Ten studies for identifying esophageal varices included 807 subjects, in which 555 were reference positive subjects, 252 were negative subjects. On the other hand, 7 studies for identifying gastric varices included 583 subjects, with 255 positive subjects and 328 negative subjects. A detailed summary of study characteristics are shown in Table 1. All included studies were subjected to quality assessment via the QUADAS-2 tool. Four studies were rated low risk for both risk of bias and applicability concerns, these studies were considered high quality studies [16,20–22]. The remaining studies were judged as suboptimal for evident high or unclear risk in the following domains: index test, reference standard, flow and timing [11,15,17–19,23,24]. A detailed QUDADS-2 assessment is shown in Fig. 2. 3.2. Diagnostic accuracy The overall pooled diagnostic parameters for esophageal varices were: pooled sensitivity 0.896 (95% CI, 0.841–0.934), specificity 0.723 (95% CI, 0.644–0.791), PLR 3.241 (95% CI, 2.489–4.220),
Fig. 2. QUDADS 2 assessment for risk bias (a) and applicability concerns (b).
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS
4
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
Table 1 Study characteristics. Study
Year
Country
No. patients
Chen
2006
China
60
Gulati
2000
India
50
He
2012
China
92
51 ± 12 y
73
Jin
2007
China
57
53.0 ± 12.6 y
41
Kim
2007
Korea
90
54.8 y
65
Kim YJ
2007
Korea
67
56.2 y
39
Moftah
2013
Egypt
54
56.84 ± 7.52 y
40
Perri
2008
USA
57.5
64
Wang
2014
China
96
56.38 ± 12.32 y
59
Zhu
2010
China
127
45.2 y
96
Yu
2011
USA
109
55.9 y
60
101
Mean age
Male
Aetiology of PH
Index test
Time interval between index and reference tests
Child pugh score
51.9 ± 13.7 y
45
Viral cirrhosis 42; schistosomiasis cirrhosis 7; alcoholic cirrhosis 3; pancreatitis or pancreatic pseudocyst 5; pancreatic tumour 2 Extrahepatic portal venous obstruction (EHO) 50
GE LightSpeed 16 multislice spiral CT scanner (24 detector)
N/A
N/A
Siemens Somatom Plus 4 subsecond helical CT Scanner (single detector) GE LightSpeed VCT 64 (64 detector)
6 weeks (mean 31.6d)
N/A
4 weeks
Child A 33, Child B 44, Child C 15
GE LightSpeed 16 CT scanner (24 detector)
0–3d
N/A
Siemens sensation 16 detector CT scanner
4h
Child A 73; Child B 17; Child C 0
GE HighSpeed single-detector or GE LightSpeed 4 detector CT GE 8 detector lightspeed or GE 4 detector lightspeed CT scanner Multidetector CT scan (4 detector or higher)
4 weeks
Child A 16; Child B 25; Child C 26
N/A
N/A
2 days
Child A 45; Child B 40; Child C 16
GE LightSpeed 16 slice spiral CT (24 detector) GE LightSpeed 4 detector CT scanner
3 weeks
Child A 32; Child B 36; Child C 28 Child A 48; Child B 47; Child C 32
16- or 64-MDCT Scanner (Sensation, Siemens Healthcare)
10 weeks
7.5 y
36
NLR 0.143 (95% CI, 0.093–0.222), and DOR 22.599 (95% CI, 12.734–40.108). While that of gastric varices were: pooled sensitivity 0.955 (95% CI, 0.903–0.980), specificity 0.658 (95% CI, 0.433–0.829), PLR 2.790 (95% CI, 1.550–5.024), NLR 0.069 (95% CI, 0.035–0.134), DOR 40.585 (95% CI, 17.458–94.349). The corresponding forest plot for sensitivity and specificity are shown in Figs. 3 and 4. Heterogeneity between studies were substantial in both the esophageal and gastric varices subgroup according to the corresponding I2 , which mandated a subgroup analysis to identify potential source of heterogeneity.
Viral cirrhosis 78; alcoholic cirrhosis 12; pancreatic cirrhosis 1; biliary cirrhosis 1 Viral cirrhosis 36; schistosomiasis cirrhosis 10; alcoholic cirrhosis 2; portal vein cavernous transformation 2; splenectomy 5; unknown aetiology 2 HBV 66; HCV 19; alcohol 2; cryptogenic 1; allagille syndrome 1 HBV 15; HCV 24; HBV + HCV 6; alcohol 15; PBC 5; cryptogenic 2 Cirrhosis 54
Alcohol 19; cholestasic 18; NASH 15; viral 22; other 27 N/A
Alcohol 13; HBV 95; HCV 6; Budd-Chiari syndrome 8; Cryptogenic 5 HBV 7; HCV 51; alcohol 19; cryptogenic or miscellaneous causes 32
4 weeks (2-23d)
N/A
The SROC and corresponding AUC of each subgroup are shown in Figs. 5 and 6. The AUC for esophageal varices is 0.86 (95% CI, 0.82–0.88) while that of gastric varices is 0.95 (95% CI, 0.93–0.97), indicating desirable diagnostic abilities of computed tomography in identifying gastroesophageal varices in patients with portal hypertension. The Fagan’s plot prior probability is established according to disease prevalence, with 35% for esophageal varices and 20% for gastric varices. A positive CT reading for diagnosing esophageal varices will increase the post-test probability to 64%, while a negative reading will decrease the post-test probability to 7%. Similarly,
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
5
Fig. 3. (a) Sensitivity forest plot for diagnosing esophageal varices with computed tomography. (b) Specificity forest plot for diagnosing esophageal varices with computed tomography.
Fig. 4. (a) Sensitivity forest plot for diagnosing gastric varices with computed tomography. (b) Specificity forest plot for diagnosing gastric varices with computed tomography.
Fig. 5. SROC for esophageal varices diagnosis with confidence and predictive ellipses.
a positive diagnosis for gastric varices on a CT scan will increase the post-test probability to 41%, while a negative result with reduce the post-test probability to 2%. An additional Fagan’s plot for large varices have been included for clinical reference, since variceal
Fig. 6. SROC for gastric varices diagnosis with confidence and predictive ellipses.
size is an essential indicator for prophylactic treatment. The point prevalence for large varices is approximately 15–25% at the time of EGD screening [4], thus a positive CT reading for identifying large varices will increase the post-test probability to 45% and a negative reading will decrease the probability to 3% (Figs. 7–9).
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS
6
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
Fig. 7. Fagan’s plot for esophageal varices diagnosis post-test probability.
Fig. 8. Fagan’s plot for gastric varices diagnosis post-test probability.
3.3. Subgroup analysis and publication bias
4. Discussion
Significant heterogeneity was identified for both subgroups, the overall I2 for esophageal varices was 70.87% (95% CI, 35.20–100.00) and 91.97% (95% CI, 84.52–99.42) for gastric varices. Three subgroup analysis including study quality (suboptimal or high quality), CT technology (<16 detector CT or ≥16 detector CT), and sample size (<80 subjects or ≥80 subjects) were carried out in attempt to identify the source of heterogeneity. Results of the subgroup analysis were conducted with the STATA 13.0 (StataCorp, College Station, Texas) midas Module is shown in Table 2. High quality studies were signified if the QUADAS-2 Assessment confirmed low risk for both risk of bias and applicability concerns. If either category indicated a high or unclear risk, the study was classified as suboptimal. The CT technology subgroup was classified according to the CT scanner employed in the study. One study described the use of more than one CT scanner (4 detector or more), was then classified in the <16 detector subgroup [11]. Studies that employed a CT scanner with 16 or more detectors had a higher sensitivity and specificity compared to those with less than 16 detectors (0.904 vs. 0898 and 0.808 vs. 0.663, respectively). The use of different CT technology served as a potential source of heterogeneity between studies. Publication bias of included studies were investigated through the use of Deek’s funnel plot, the P-value associated with the slope coefficient for diagnosis of esophageal varices and gastric varices were 0.457 and 0.090 respectively, indicating low probability of publication bias (Figs. 10 and 11).
The recent BAVENO VI Consensus focused on the use of non-invasive methods for the screening and surveillance of gastroesophageal varices and of portal hypertension. Currently, the primary gastroesophageal variceal screening method remains as esophagogastroduodenoscopy (EGD) with varying recommended surveillance intervals. However, new considerations regarding cost for screening and surveillance programmes have been raised, urging future studies and long-term data [25]. The overall sensitivity and specificity for computed tomography in detecting esophageal varices were 0.896, and 0.715, respectively, with an acceptable missed diagnosis rate of 10.4% and high misdiagnosis rate of 28.5%. The overall PLR is 3.241 and NLR is 0.143 suggesting the likelihood of an accurate positive CT diagnosis for esophageal varices is 3-fold higher in patients with the condition compared to patients without. A PLR of 10 and NLR of 0.1 are considered substantial evidence to rule in or rule out a diagnosis, respectively [26]. The diagnostic odds ratio combined both sensitivity and specificity, presenting as a single diagnostic indicator with a higher value indicating higher accuracy. The DOR for diagnosis of esophageal varices was 22.599, which indicated high overall accuracy. The AUC 0.86 also supported the former index proving good overall diagnosis performance. On the other hand, the overall sensitivity and specificity for computed tomography in diagnosing gastric varices were 0.955 and 0.658, respectively. Although the missed diagnosis was an ideal 4.5%, but the high misdiagnosis rate of
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model
ARTICLE IN PRESS
YDLD-3073; No. of Pages 8
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
7
Table 2 Subgroup analysis. Study subgroup
Sensitivity (95% CI)
Specificity (95% CI)
+LR (95% CI)
−LR (95% CI)
DOR (95% CI)
AUROC
Esophageal varices (10) Gastric varices (7) High quality studies (4) Suboptimal studies (7) ≥16 detector CT (6) <16 detector CT (5) ≥80 subjects (6) <80 subjects (5)
0.896 (0.841–0.934) 0.955 (0.903–0.980) 0.879 (0.746–0.947) 0.912 (0.864–0.944) 0.904 (0.851–0.939) 0.898 (0.788–0.954) 0.904 (0.848–0.941) 0.900 (0.790–0.955)
0.723 (0.644–0.791) 0.658 (0.433–0.829) 0.746 (0.580–0.862) 0.765 (0.668–0.841) 0.808 (0.702–0.882) 0.663 (0.562–0.751) 0.747 (0.614–0.846) 0.781 (0.669–0.863)
3.241 (2.489–4.220) 2.790 (1.550–5.024) 3.459 (1.911–6.262) 3.881 (2.685–5.612) 4.701 (2.927–7.552) 2.668 (2.010–3.543) 3.579 (2.226–5.753) 4.105 (2.610–6.457)
0.143 (0.093–0.222) 0.069 (0.035–0.134) 0.162 (0.069–0.381) 0.115 (0.073–0.180) 0.119 (0.074–0.190) 0.153 (0.071–0.330) 0.128 (0.076–0.217) 0.129 (0.058–0.285)
22.599 (12.734–40.108) 40.585 (17.458–94.349) 21.323 (5.881–77.320) 33.780 (17.474–65.301) 39.545 (18.037–86.600) 17.412 (7.058– 42.958) 27.894 (11.631–66.899) 31.940 (11.158–91.432)
0.86 0.95 0.89 0.91 0.93 0.72 0.92 0.80
Fig. 11. Publication Bias for Gastric Varices Diagnosis.
Fig. 9. Fagan’s plot for large varices diagnosis post-test probability.
34.2% raises concerns. The overall PLR was 2.790 and NLR was 0.069. The DOR was optimal at 40.585 with a concurring AUC of 0.95. Substantial heterogeneity was identified among the 11 included studies, which was investigated with a subsequent subgroup analysis. Three subgroup analyses were carried out with CT technology presented as the most likely source of heterogeneity. A threshold effect was not indicated in this particular meta-analysis due to a lack of uniformed standard for imaging diagnosis. This metaanalysis possesses several limitations. Merely 10 and 7 studies were included in each subgroup, which may hinder the statistical power necessary to draw a definite conclusion. Computed tomography is a non-invasive imaging modality available for the detection gastrointestinal lesions. Radiological advancements have allowed for high-quality multiplanar reformation and three-dimensional reconstruction of the abdomen using multidetector computed tomography (MDCT), along with multiple rending models (VR, MIP, SSD), providing optimal visualization for portosystemic collaterals [10,27–29]. The role of CT in detecting gastroesophageal varices could potentially replace EGD, given its diagnostic accuracy is at par. Due to the numerous rendering techniques and with aid of contrast dye, CT could also identify submucosal lesions, often missed in conventional EGD. The more tolerable and less expensive imaging modality could potentially replace EGD as a primary screening method for gastroesophageal varices in patients with portal hypertension. Given the low point-prevalence of the condition, the necessity of an invasive and expensive EGD screening during the time of diagnosis of cirrhosis, or other aetiology of portal hypertension remains debatable. Conflict of interest None declared. Acknowledgements
Fig. 10. Publication bias for esophageal varices diagnosis.
The study was supported by the Innovation Fund of Shanghai Scientific Committee (No. 15411950501).
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS
8
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
Appendix 1. Detailed search strategy Ovid MEDLINE(R) 1946 to present with daily updates Number
Searches
Results
Search type
1
esophag* or esophag* gastr* or gastr* esophag* or gastr* oesophag* or gastroesophag* or gastrooesophag* or oesophag* or oesophag* gastr* or gastr* 1 and (varic* or varix).mp. Exp “Esophageal and Gastric Varices”/ 2 or 3 angiogra* or arteriogr* or portogra* Exp tomography, X-ray computed/ ct OR compute* tomograph* OR cta OR compute* tomograph* angiogra* 5 and (6 or 7) 4 and 8 Limit 9 to yr = “2000–2015”
668,044
Advanced
16,155 11,724
Advanced Advanced
16,155 252,273
Advanced Advanced
328,503
Advanced
331,482
Advanced
52,593 269 180
Advanced Advanced Advanced
2 3 4 5 6 7
8 9 10 Exp, explode.
Embase 1974 to 2015 September 09 Number
Searches
Results
Search type
1
esophag* or esophag* gastr* or gastr* esophag* or gastr* oesophag* or gastroesophag* or gastrooesophag* or oesophag* or oesophag* gastr* or gastr* 1 and (varic* or varix).mp. Exp oesophagus varices/ Exp stomach varices/ 3 or 4 2 or 5 angiogra* or arteriogr* or portogra* Exp computer assisted tomography/ ct OR compute* tomograph* OR cta OR compute* tomograph* angiogra* 7 and (8 or 9) 6 and 10 Limit 11 to yr = “2000–2015”
1,029,640
Advanced
24,799 16,585 2323 17,835 24,910 334,353
Advanced Advanced Advanced Advanced Advanced Advanced
680,197
Advanced
574,361
Advanced
94,138 639 543
Advanced Advanced Advanced
2 3 4 5 6 7 8 9
10 11 12
Exp, explode. Web of Science (Advanced Search) TS = ((esophag* or esophag* gastr* or gastr* esophag* or gastr* oesophag* or gastroesophag* or gastrooesophag* or oesophag* or oesophag* gastr* or gastr*) AND (varic* or varix)) AND TS = ((Angiogra* or arteriogra* or portogra*) AND (ct OR compute* tomograph* OR cta OR compute* tomograph* angiogra*)) 146. Indexes = SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH, CCR-EXPANDED, IC Timespan = 2000–2015 Scopus (Advanced Search) (TITLE-ABS-KEY ((esophag* OR esophag* gastr* OR gastr* esophag* OR gastr* oesophag* OR gastroesophag* OR gastrooesophag* OR oesophag* OR oesphag* gastr* OR gastr*) AND (varic* OR varix)) AND TITLE-ABS-KEY ((angiogra* OR arteriogra* OR portogra*) AND (ct OR compte* tomograph* OR cta OR compute* tomograph* angiogra*)) AND PUBYEAR > 1999) 27.
References [1] Triantafyllou M, Stanley AJ. Update on gastric varices. World Journal of Gastrointestinal Endoscopy 2014;6:168–75.
[2] Schuppan D, Afdhal NH. Liver cirrhosis. Lancet 2008;371:838–51. [3] Marshall JP, Smith PD, Hoyumpa AM. Gastric varices. American Journal of Digestive Diseases 1977;22:947–55. [4] Garcia-Tsao G, Sanyal AJ, Grace ND, et al. Prevention and management of gastroesophageal varices and variceal hemorrhage in cirrhosis. Hepatology 2007;46:922–38. [5] Garcia-Tsao G. Current management of the complications of cirrhosis and portal hypertension: variceal hemorrhage, ascites, and spontaneous bacterial peritonitis. Gastroenterology 2001;120:726–48. [6] Bosch J, Abraldes J, Berzigotti A, et al. Portal hypertension and gastrointestinal bleeding. Seminars in Liver Disease 2008;28:3–25. [7] Kim T, Shijo H, Kokawa H, et al. Risk factors for hemorrhage from gastric fundal varices. Hepatology 1997;25:307–12. [8] Colli A, Gana JC, Turner D, et al. Capsule endoscopy for the diagnosis of oesophageal varices in people with chronic liver disease or portal vein thrombosis. Cochrane Database of Systematic Reviews 2014;10:D8760. [9] Dessouky BA, Abdel AES. Multidetector CT oesophagography: an alternative screening method for endoscopic diagnosis of oesophageal varices and bleeding risk. Arab Journal of Gastroenterology 2013;14:99–108. [10] Nakayama Y, Imuta M, Funama Y, et al. CT portography by multidetector helical CT: comparison of three rendering models. Radiation Medicine 2002;20: 273–9. [11] Perri RE, Chiorean MV, Fidler JL, et al. A prospective evaluation of computerized tomographic (CT) scanning as a screening modality for esophageal varices. Hepatology 2008;47:1587–94. [12] Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Annals of Internal Medicine 2009;151:264–9. W64. [13] Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Annals of Internal Medicine 2011;155:529–36. [14] Reitsma JB, Glas AS, Rutjes AW, et al. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. Journal of Clinical Epidemiology 2005;58:982–90. [15] Jin Y, Wu YL, Wang L, et al. A comparative study on the diagnosis of esophageal and gastric varices between multislice spiral CT portography and endoscopy. Chinese Journal of Digestion 2007;27:78–82. [16] Zhu K, Meng X, Pang P, et al. Gastric varices in patients with portal hypertension: evaluation with multidetector row CT. Journal of Clinical Gastroenterology 2010;44:e108–15. [17] Gulati MS, Paul SB, Arora NK, et al. Esophageal and gastric vasculature in children with extrahepatic portal hypertension: evaluation by intravenous CT portography. Clinical Imaging 2000;24:351–6. [18] He X, Huang ZK, Long LL, et al. Comparison of multi-slice spiral CT portography and endoscopy in the diagnosis of esophageal and gastric varices in liver cirrhosis. China: Chinese Medical Association; 2012. p. 1092–5 [42 Dongsi Xidajie, Beijing 100710, China]. [19] Chen LN, Qu Q, Sun B, et al. Values of endoscopy and computed tomographic angiography for portal vein in diagnosis of portal hypertension. China: WJC Press; 2006. p. 3421–5. [20] Wang AM, Pan KH, Cao GQ, et al. Value of multi-detector CT in the diagnosis of gastric varices in patients with cirrhosis. Radiologic Practice 2014;07:805–7. [21] Kim SH, Kim YJ, Lee JM, et al. Esophageal varices in patients with cirrhosis: multidetector CT esophagography – comparison with endoscopy. United States: Radiological Society of North America Inc.; 2007. p. 759–68. [22] Kim YJ, Raman SS, Yu NC, et al. Esophageal varices in cirrhotic patients: evaluation with liver CT. American Journal of Roentgenology 2007;188:139–44. [23] Moftah SG, Kamal S, Hanna ATK. CT esophagography: non invasive screening and grading of esophageal varices in cirrhosis. The Egyptian Journal of Radiology and Nuclear Medicine 2014;45:263–70. [24] Yu NC, Margolis D, Hsu M, et al. Detection and grading of esophageal varices on liver CT: comparison of standard and thin-section multiplanar reconstructions in diagnostic accuracy. American Journal of Roentgenology 2011;197: 643–9. [25] de Franchis R. Expanding consensus in portal hypertension: report of the Baveno VI Consensus Workshop: stratifying risk and individualizing care for portal hypertension. Journal of Hepatology 2015;63:743–52. [26] Akobeng AK. Understanding diagnostic tests 2: likelihood ratios, pre- and post-test probabilities and their use in clinical practice. Acta Paediatrica 2007;96:487–91. [27] Ba-Ssalamah A, Prokop M, Uffmann M, et al. Dedicated multidetector CT of the stomach: spectrum of diseases. Radiographics 2003;23:625–44. [28] Lee JY, Kim TY, Jeong WK, et al. Clinically severe portal hypertension: role of multi-detector row CT features in diagnosis. Digestive Diseases and Sciences 2014;59:2333–43. [29] Matsumoto A, Kitamoto M, Imamura M, et al. Three-dimensional portography using multislice helical CT is clinically useful for management of gastric fundic varices. American Journal of Roentgenology 2001;176:899–905.
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
ARTICLE IN PRESS Digestive and Liver Disease xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld
Review Article
Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis Tseng Yu-Jen a , Zeng Xiao-qing a , Chen Jie a , Li Na a , Xu Peng-ju b , Chen Shi-yao c,∗ a b c
Department of Gastroenterology, Zhongshan Hospital, Fudan University, Shanghai, China Department of Radiology, Zhongshan Hospital, Fudan University, Shanghai, China Department of Gastroenterology, Endoscopy Center, Evidence-based Medicine Center, Zhongshan Hospital, Fudan University, Shanghai, China
a r t i c l e
i n f o
Article history: Received 8 January 2016 Accepted 18 February 2016 Available online xxx Keywords: Computed tomography Diagnosis Gastroesophageal varices Meta-analysis
a b s t r a c t Aims: Gastroesophageal varices (GOV) is a common complication in patients with portal hypertension. We conducted a meta-analysis in attempt to evaluate the diagnostic accuracy of computed tomography (CT) as a noninvasive imaging tool for identifying GOV in reference to esophagogastroduodenoscopy (EGD). Methods: A systemic literature search of multiple databases were conducted to identify articles that investigated the diagnostic performance of CT for GOV, while employing EGD as reference standard. A 2 × 2 table was conducted according to the available published data for both esophageal varices (EV) and gastric varices (GV) as individual subgroups. The following indices were calculated: pooled sensitivity and specificity, positive and negative likelihood ratio, diagnostic odds ratio, and area under receiver operating characteristics. All statistical analyses were conducted via STATA13.0 and RevMan5.3. Results: A total of 11 studies were included in this meta-analysis, 10 articles evaluated the diagnostic accuracy of CT for EV (807 subjects) and 7 articles for GV (583 subjects). The pooled sensitivity and specificity for identifying EV were 0.896 (95% CI, 0.841–0.934) and 0.723 (95% CI, 0.644–0.791), respectively, with an AUROC of 0.86. The pooled sensitivity and specificity for identifying GV were 0.955 (95% CI, 0.903–0.980) and 0.658 (95% CI, 0.433–0.829), respectively, with an AUROC of 0.95. A subgroup analysis suggested varying CT technology could serve as a potential source of heterogeneity between included studies. A Deek’s funnel plot indicated a low probability for publication bias. Conclusion: Computed tomography could potentially replace EGD as a primary screening tool for diagnosing GOV, however results should be interpreted with caution given its suboptimal specificity. © 2016 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
1. Introduction Portal hypertension is a progressive complication secondary to intra-hepatic, pre-hepatic, or post-hepatic aetiology [1]. Liver cirrhosis being the more common intra-hepatic cause, affects roughly 1% of the population worldwide, with Asian and African countries heavily weighing on disease prevalence [2]. Portal hypertension is often associated with a series of complications including ascites, hepatic encephalopathy, and gastroesophageal varices. Among which, gastroesophageal varix haemorrhage is the most common gastroenterological emergency [3]. Approximately 50% of patients
∗ Corresponding author at: Department of Gastroenterology, Endoscopy Center, Evidence-based Medicine Center, Zhongshan Hospital, Fudan University, 180 Fenglin Road, Shanghai 0086200032, China. Tel.: +86 13601767310/15000223468; fax: +86 2164038472. E-mail address: [email protected] (S.-y. Chen).
with cirrhosis develop gastroesophageal varices and their presence is often correlated with disease severity [4]. Gastric and esophageal varices can occur concurrently or in solitary, with esophageal varices more prevalent in patients with portal hypertension. The prevalence of gastroesophageal haemorrhage is approximately 10–30% per year [5]. Esophageal varices occur in 30 to 40% of cirrhotic patients, while gastric varices occur in approximately 20%. However, GV rupture is associated with a higher mortality rate of up to 45% [1,6]. Risk factors for gastroesophageal variceal haemorrhage include size of varix (large >10 mm, medium 5–10 mm, and small <5 mm), Child Pugh’s score, and presence of red-spots [7]. The high morbidity and mortality rates associated with gastroesophageal variceal haemorrhage demands an early detection and prophylactic treatment for patients at risk for disease development. The American Association for the Study of Liver Diseases (AASLD) recommends patients to undergo screening esophagogastroduodenoscopy (EGD) for detecting esophageal and gastric varices
http://dx.doi.org/10.1016/j.dld.2016.02.007 1590-8658/© 2016 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS
2
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
when the diagnosis of cirrhosis is made. All varices are graded as small or large (>5 mm) and the presence or absence of red spots is duly noted. Patients with gastroesophageal varices should be assessed for risk of variceal haemorrhage and treated accordingly with prophylactic medication (-blockers) or minimally invasive preventive therapy such as (EVL or sclerotherapy) [4]. However, the point prevalence of esophageal varices requiring prophylaxis ranges from 15 to 25%, and even lower for gastric varices. The majority of patients who undergo EGD screening at the time of cirrhosis diagnosis either have no varices, or have small varices that do not require treatment [8]. Moreover, EGD is an invasive and expensive procedure that requires sedation and is poorly tolerated by patients due to associated discomfort during and after the procedure [9]. Computed tomography imaging could potentially replace EGD as a non-invasive, more tolerable, and inexpensive test in the accurate diagnosis and risk assessment of gastroesophageal varices. With advancements in radiological imaging techniques, CT application now allows for multiple rendering models, such as but not limited to, multidetector computed tomography (MDCT), volume rendering (VR), minimum intensity projection (CT-MIP) and shade surface display (SSD), which could provide a more infallible identification and assessment of gastric varices [9–11]
identify relevant articles on diagnostic accuracy of computed tomography for gastroesophageal varices in reference to esophagogastroduodenoscopy (EGD). A combination of the following search terms were used: (esophageal varices OR gastric varices OR gastroesophageal varices) AND (CT OR computed tomography OR angiography). Refer to Appendix 1 for detailed search strategy. 2 reviewers (YT and XZ) independently reviewed the title and abstract of studies in attempt to eliminate irrelevant articles, based on a priori established inclusion and exclusion criteria. 2.2. Inclusion and exclusion criteria Inclusion criteria are as follows: (1) diagnostic accuracy of CT imaging was assessed in reference to EGD, (2) data provided was sufficient to conduct a 2 × 2 table for analysis, (3) absence or presence of either gastric or esophageal varices were assessed. Exclusion criteria includes: (1) study population was limited to patients with gastroesophageal varices initially confirmed with EGD, (2) studies that only provided diagnostic rate, without calculable sensitivity and specificity. No language or article type restrictions were imposed. Additional reference articles were acquired through a manual search of computerized databases. 2.3. Quality assessment
2. Methods 2.1. Search strategy This meta-analysis was conducted according to the PRISMA statement [12]. A systematic search of MEDLINE, Embase, Web of Science, and Scopus was performed through August 2015 to
All included studies were subjected to the Quality Assessment for Studies Diagnostic Accuracy-2 (QUADAS-2) guideline and rated according to the 4 domains (Patient Selection, Index Test, Reference Standard, Flow and Timing) for risk of bias and sources of variation (applicability) assessment. Risk of bias was judged as “low”, “high”, or “unclear”, under the guidance of a series of 10 signalling
Fig. 1. Study Identification flowchart.
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
questions. Similarly, applicability was rated as “low”, “high”, or “unclear” [13].
3
3. Results 3.1. Characteristics of included studies
2.4. Statistical analysis This meta-analysis was conducted based on a bivariate binomial mixed model [14]. All statistical analyses were conducted with the midas module in STATA 13.0 (StataCorp, College Station, Texas) and RevMan5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). All studies were initially classified into two subgroups: esophageal varices or gastric varices, with 10 and 7 applicable studies respectively. A 2 × 2 contingency table was constructed for esophageal and gastric varices independently to calculate the following diagnostic indices: pooled sensitivity and specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), positive predictive value (PPV), negative predictive value (NPV) and diagnostic odds ratio (DOR). Summary receiver operating characteristics (SROC) and its corresponding area under the curve (AUC) were also calculated to estimate the overall diagnostic performance of computed tomography. Heterogeneity between studies were assessed using the I2 inconsistency test. I2 > 50% indicated presence of substantial heterogeneity, a subsequent subgroup analysis was carried out in attempt to identify potential covariates. Post-test probability was calculated with a presumed prior probability of 35% for EV and 20% for GV via Fagan’s monogram. A Deek’s funnel plot was used to detect publication bias. All the above mentioned parameters were conducted separately for the diagnostic accuracy of esophageal varices and gastric varices. P-value less than 0.05 were considered statistically significant.
A total of 11 publications were included in this meta-analysis, 10 studies were appropriate diagnostic test accuracy studies for esophageal varices and 7 for gastric varices. Six studies explored the diagnostic abilities of computed tomography for both esophageal and gastric varices [11,15–19]. A detailed flowchart of study identification is shown in Fig. 1. Ten studies for identifying esophageal varices included 807 subjects, in which 555 were reference positive subjects, 252 were negative subjects. On the other hand, 7 studies for identifying gastric varices included 583 subjects, with 255 positive subjects and 328 negative subjects. A detailed summary of study characteristics are shown in Table 1. All included studies were subjected to quality assessment via the QUADAS-2 tool. Four studies were rated low risk for both risk of bias and applicability concerns, these studies were considered high quality studies [16,20–22]. The remaining studies were judged as suboptimal for evident high or unclear risk in the following domains: index test, reference standard, flow and timing [11,15,17–19,23,24]. A detailed QUDADS-2 assessment is shown in Fig. 2. 3.2. Diagnostic accuracy The overall pooled diagnostic parameters for esophageal varices were: pooled sensitivity 0.896 (95% CI, 0.841–0.934), specificity 0.723 (95% CI, 0.644–0.791), PLR 3.241 (95% CI, 2.489–4.220),
Fig. 2. QUDADS 2 assessment for risk bias (a) and applicability concerns (b).
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS
4
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
Table 1 Study characteristics. Study
Year
Country
No. patients
Chen
2006
China
60
Gulati
2000
India
50
He
2012
China
92
51 ± 12 y
73
Jin
2007
China
57
53.0 ± 12.6 y
41
Kim
2007
Korea
90
54.8 y
65
Kim YJ
2007
Korea
67
56.2 y
39
Moftah
2013
Egypt
54
56.84 ± 7.52 y
40
Perri
2008
USA
57.5
64
Wang
2014
China
96
56.38 ± 12.32 y
59
Zhu
2010
China
127
45.2 y
96
Yu
2011
USA
109
55.9 y
60
101
Mean age
Male
Aetiology of PH
Index test
Time interval between index and reference tests
Child pugh score
51.9 ± 13.7 y
45
Viral cirrhosis 42; schistosomiasis cirrhosis 7; alcoholic cirrhosis 3; pancreatitis or pancreatic pseudocyst 5; pancreatic tumour 2 Extrahepatic portal venous obstruction (EHO) 50
GE LightSpeed 16 multislice spiral CT scanner (24 detector)
N/A
N/A
Siemens Somatom Plus 4 subsecond helical CT Scanner (single detector) GE LightSpeed VCT 64 (64 detector)
6 weeks (mean 31.6d)
N/A
4 weeks
Child A 33, Child B 44, Child C 15
GE LightSpeed 16 CT scanner (24 detector)
0–3d
N/A
Siemens sensation 16 detector CT scanner
4h
Child A 73; Child B 17; Child C 0
GE HighSpeed single-detector or GE LightSpeed 4 detector CT GE 8 detector lightspeed or GE 4 detector lightspeed CT scanner Multidetector CT scan (4 detector or higher)
4 weeks
Child A 16; Child B 25; Child C 26
N/A
N/A
2 days
Child A 45; Child B 40; Child C 16
GE LightSpeed 16 slice spiral CT (24 detector) GE LightSpeed 4 detector CT scanner
3 weeks
Child A 32; Child B 36; Child C 28 Child A 48; Child B 47; Child C 32
16- or 64-MDCT Scanner (Sensation, Siemens Healthcare)
10 weeks
7.5 y
36
NLR 0.143 (95% CI, 0.093–0.222), and DOR 22.599 (95% CI, 12.734–40.108). While that of gastric varices were: pooled sensitivity 0.955 (95% CI, 0.903–0.980), specificity 0.658 (95% CI, 0.433–0.829), PLR 2.790 (95% CI, 1.550–5.024), NLR 0.069 (95% CI, 0.035–0.134), DOR 40.585 (95% CI, 17.458–94.349). The corresponding forest plot for sensitivity and specificity are shown in Figs. 3 and 4. Heterogeneity between studies were substantial in both the esophageal and gastric varices subgroup according to the corresponding I2 , which mandated a subgroup analysis to identify potential source of heterogeneity.
Viral cirrhosis 78; alcoholic cirrhosis 12; pancreatic cirrhosis 1; biliary cirrhosis 1 Viral cirrhosis 36; schistosomiasis cirrhosis 10; alcoholic cirrhosis 2; portal vein cavernous transformation 2; splenectomy 5; unknown aetiology 2 HBV 66; HCV 19; alcohol 2; cryptogenic 1; allagille syndrome 1 HBV 15; HCV 24; HBV + HCV 6; alcohol 15; PBC 5; cryptogenic 2 Cirrhosis 54
Alcohol 19; cholestasic 18; NASH 15; viral 22; other 27 N/A
Alcohol 13; HBV 95; HCV 6; Budd-Chiari syndrome 8; Cryptogenic 5 HBV 7; HCV 51; alcohol 19; cryptogenic or miscellaneous causes 32
4 weeks (2-23d)
N/A
The SROC and corresponding AUC of each subgroup are shown in Figs. 5 and 6. The AUC for esophageal varices is 0.86 (95% CI, 0.82–0.88) while that of gastric varices is 0.95 (95% CI, 0.93–0.97), indicating desirable diagnostic abilities of computed tomography in identifying gastroesophageal varices in patients with portal hypertension. The Fagan’s plot prior probability is established according to disease prevalence, with 35% for esophageal varices and 20% for gastric varices. A positive CT reading for diagnosing esophageal varices will increase the post-test probability to 64%, while a negative reading will decrease the post-test probability to 7%. Similarly,
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
5
Fig. 3. (a) Sensitivity forest plot for diagnosing esophageal varices with computed tomography. (b) Specificity forest plot for diagnosing esophageal varices with computed tomography.
Fig. 4. (a) Sensitivity forest plot for diagnosing gastric varices with computed tomography. (b) Specificity forest plot for diagnosing gastric varices with computed tomography.
Fig. 5. SROC for esophageal varices diagnosis with confidence and predictive ellipses.
a positive diagnosis for gastric varices on a CT scan will increase the post-test probability to 41%, while a negative result with reduce the post-test probability to 2%. An additional Fagan’s plot for large varices have been included for clinical reference, since variceal
Fig. 6. SROC for gastric varices diagnosis with confidence and predictive ellipses.
size is an essential indicator for prophylactic treatment. The point prevalence for large varices is approximately 15–25% at the time of EGD screening [4], thus a positive CT reading for identifying large varices will increase the post-test probability to 45% and a negative reading will decrease the probability to 3% (Figs. 7–9).
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS
6
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
Fig. 7. Fagan’s plot for esophageal varices diagnosis post-test probability.
Fig. 8. Fagan’s plot for gastric varices diagnosis post-test probability.
3.3. Subgroup analysis and publication bias
4. Discussion
Significant heterogeneity was identified for both subgroups, the overall I2 for esophageal varices was 70.87% (95% CI, 35.20–100.00) and 91.97% (95% CI, 84.52–99.42) for gastric varices. Three subgroup analysis including study quality (suboptimal or high quality), CT technology (<16 detector CT or ≥16 detector CT), and sample size (<80 subjects or ≥80 subjects) were carried out in attempt to identify the source of heterogeneity. Results of the subgroup analysis were conducted with the STATA 13.0 (StataCorp, College Station, Texas) midas Module is shown in Table 2. High quality studies were signified if the QUADAS-2 Assessment confirmed low risk for both risk of bias and applicability concerns. If either category indicated a high or unclear risk, the study was classified as suboptimal. The CT technology subgroup was classified according to the CT scanner employed in the study. One study described the use of more than one CT scanner (4 detector or more), was then classified in the <16 detector subgroup [11]. Studies that employed a CT scanner with 16 or more detectors had a higher sensitivity and specificity compared to those with less than 16 detectors (0.904 vs. 0898 and 0.808 vs. 0.663, respectively). The use of different CT technology served as a potential source of heterogeneity between studies. Publication bias of included studies were investigated through the use of Deek’s funnel plot, the P-value associated with the slope coefficient for diagnosis of esophageal varices and gastric varices were 0.457 and 0.090 respectively, indicating low probability of publication bias (Figs. 10 and 11).
The recent BAVENO VI Consensus focused on the use of non-invasive methods for the screening and surveillance of gastroesophageal varices and of portal hypertension. Currently, the primary gastroesophageal variceal screening method remains as esophagogastroduodenoscopy (EGD) with varying recommended surveillance intervals. However, new considerations regarding cost for screening and surveillance programmes have been raised, urging future studies and long-term data [25]. The overall sensitivity and specificity for computed tomography in detecting esophageal varices were 0.896, and 0.715, respectively, with an acceptable missed diagnosis rate of 10.4% and high misdiagnosis rate of 28.5%. The overall PLR is 3.241 and NLR is 0.143 suggesting the likelihood of an accurate positive CT diagnosis for esophageal varices is 3-fold higher in patients with the condition compared to patients without. A PLR of 10 and NLR of 0.1 are considered substantial evidence to rule in or rule out a diagnosis, respectively [26]. The diagnostic odds ratio combined both sensitivity and specificity, presenting as a single diagnostic indicator with a higher value indicating higher accuracy. The DOR for diagnosis of esophageal varices was 22.599, which indicated high overall accuracy. The AUC 0.86 also supported the former index proving good overall diagnosis performance. On the other hand, the overall sensitivity and specificity for computed tomography in diagnosing gastric varices were 0.955 and 0.658, respectively. Although the missed diagnosis was an ideal 4.5%, but the high misdiagnosis rate of
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model
ARTICLE IN PRESS
YDLD-3073; No. of Pages 8
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
7
Table 2 Subgroup analysis. Study subgroup
Sensitivity (95% CI)
Specificity (95% CI)
+LR (95% CI)
−LR (95% CI)
DOR (95% CI)
AUROC
Esophageal varices (10) Gastric varices (7) High quality studies (4) Suboptimal studies (7) ≥16 detector CT (6) <16 detector CT (5) ≥80 subjects (6) <80 subjects (5)
0.896 (0.841–0.934) 0.955 (0.903–0.980) 0.879 (0.746–0.947) 0.912 (0.864–0.944) 0.904 (0.851–0.939) 0.898 (0.788–0.954) 0.904 (0.848–0.941) 0.900 (0.790–0.955)
0.723 (0.644–0.791) 0.658 (0.433–0.829) 0.746 (0.580–0.862) 0.765 (0.668–0.841) 0.808 (0.702–0.882) 0.663 (0.562–0.751) 0.747 (0.614–0.846) 0.781 (0.669–0.863)
3.241 (2.489–4.220) 2.790 (1.550–5.024) 3.459 (1.911–6.262) 3.881 (2.685–5.612) 4.701 (2.927–7.552) 2.668 (2.010–3.543) 3.579 (2.226–5.753) 4.105 (2.610–6.457)
0.143 (0.093–0.222) 0.069 (0.035–0.134) 0.162 (0.069–0.381) 0.115 (0.073–0.180) 0.119 (0.074–0.190) 0.153 (0.071–0.330) 0.128 (0.076–0.217) 0.129 (0.058–0.285)
22.599 (12.734–40.108) 40.585 (17.458–94.349) 21.323 (5.881–77.320) 33.780 (17.474–65.301) 39.545 (18.037–86.600) 17.412 (7.058– 42.958) 27.894 (11.631–66.899) 31.940 (11.158–91.432)
0.86 0.95 0.89 0.91 0.93 0.72 0.92 0.80
Fig. 11. Publication Bias for Gastric Varices Diagnosis.
Fig. 9. Fagan’s plot for large varices diagnosis post-test probability.
34.2% raises concerns. The overall PLR was 2.790 and NLR was 0.069. The DOR was optimal at 40.585 with a concurring AUC of 0.95. Substantial heterogeneity was identified among the 11 included studies, which was investigated with a subsequent subgroup analysis. Three subgroup analyses were carried out with CT technology presented as the most likely source of heterogeneity. A threshold effect was not indicated in this particular meta-analysis due to a lack of uniformed standard for imaging diagnosis. This metaanalysis possesses several limitations. Merely 10 and 7 studies were included in each subgroup, which may hinder the statistical power necessary to draw a definite conclusion. Computed tomography is a non-invasive imaging modality available for the detection gastrointestinal lesions. Radiological advancements have allowed for high-quality multiplanar reformation and three-dimensional reconstruction of the abdomen using multidetector computed tomography (MDCT), along with multiple rending models (VR, MIP, SSD), providing optimal visualization for portosystemic collaterals [10,27–29]. The role of CT in detecting gastroesophageal varices could potentially replace EGD, given its diagnostic accuracy is at par. Due to the numerous rendering techniques and with aid of contrast dye, CT could also identify submucosal lesions, often missed in conventional EGD. The more tolerable and less expensive imaging modality could potentially replace EGD as a primary screening method for gastroesophageal varices in patients with portal hypertension. Given the low point-prevalence of the condition, the necessity of an invasive and expensive EGD screening during the time of diagnosis of cirrhosis, or other aetiology of portal hypertension remains debatable. Conflict of interest None declared. Acknowledgements
Fig. 10. Publication bias for esophageal varices diagnosis.
The study was supported by the Innovation Fund of Shanghai Scientific Committee (No. 15411950501).
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007
G Model YDLD-3073; No. of Pages 8
ARTICLE IN PRESS
8
Y.-J. Tseng et al. / Digestive and Liver Disease xxx (2016) xxx–xxx
Appendix 1. Detailed search strategy Ovid MEDLINE(R) 1946 to present with daily updates Number
Searches
Results
Search type
1
esophag* or esophag* gastr* or gastr* esophag* or gastr* oesophag* or gastroesophag* or gastrooesophag* or oesophag* or oesophag* gastr* or gastr* 1 and (varic* or varix).mp. Exp “Esophageal and Gastric Varices”/ 2 or 3 angiogra* or arteriogr* or portogra* Exp tomography, X-ray computed/ ct OR compute* tomograph* OR cta OR compute* tomograph* angiogra* 5 and (6 or 7) 4 and 8 Limit 9 to yr = “2000–2015”
668,044
Advanced
16,155 11,724
Advanced Advanced
16,155 252,273
Advanced Advanced
328,503
Advanced
331,482
Advanced
52,593 269 180
Advanced Advanced Advanced
2 3 4 5 6 7
8 9 10 Exp, explode.
Embase 1974 to 2015 September 09 Number
Searches
Results
Search type
1
esophag* or esophag* gastr* or gastr* esophag* or gastr* oesophag* or gastroesophag* or gastrooesophag* or oesophag* or oesophag* gastr* or gastr* 1 and (varic* or varix).mp. Exp oesophagus varices/ Exp stomach varices/ 3 or 4 2 or 5 angiogra* or arteriogr* or portogra* Exp computer assisted tomography/ ct OR compute* tomograph* OR cta OR compute* tomograph* angiogra* 7 and (8 or 9) 6 and 10 Limit 11 to yr = “2000–2015”
1,029,640
Advanced
24,799 16,585 2323 17,835 24,910 334,353
Advanced Advanced Advanced Advanced Advanced Advanced
680,197
Advanced
574,361
Advanced
94,138 639 543
Advanced Advanced Advanced
2 3 4 5 6 7 8 9
10 11 12
Exp, explode. Web of Science (Advanced Search) TS = ((esophag* or esophag* gastr* or gastr* esophag* or gastr* oesophag* or gastroesophag* or gastrooesophag* or oesophag* or oesophag* gastr* or gastr*) AND (varic* or varix)) AND TS = ((Angiogra* or arteriogra* or portogra*) AND (ct OR compute* tomograph* OR cta OR compute* tomograph* angiogra*)) 146. Indexes = SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH, CCR-EXPANDED, IC Timespan = 2000–2015 Scopus (Advanced Search) (TITLE-ABS-KEY ((esophag* OR esophag* gastr* OR gastr* esophag* OR gastr* oesophag* OR gastroesophag* OR gastrooesophag* OR oesophag* OR oesphag* gastr* OR gastr*) AND (varic* OR varix)) AND TITLE-ABS-KEY ((angiogra* OR arteriogra* OR portogra*) AND (ct OR compte* tomograph* OR cta OR compute* tomograph* angiogra*)) AND PUBYEAR > 1999) 27.
References [1] Triantafyllou M, Stanley AJ. Update on gastric varices. World Journal of Gastrointestinal Endoscopy 2014;6:168–75.
[2] Schuppan D, Afdhal NH. Liver cirrhosis. Lancet 2008;371:838–51. [3] Marshall JP, Smith PD, Hoyumpa AM. Gastric varices. American Journal of Digestive Diseases 1977;22:947–55. [4] Garcia-Tsao G, Sanyal AJ, Grace ND, et al. Prevention and management of gastroesophageal varices and variceal hemorrhage in cirrhosis. Hepatology 2007;46:922–38. [5] Garcia-Tsao G. Current management of the complications of cirrhosis and portal hypertension: variceal hemorrhage, ascites, and spontaneous bacterial peritonitis. Gastroenterology 2001;120:726–48. [6] Bosch J, Abraldes J, Berzigotti A, et al. Portal hypertension and gastrointestinal bleeding. Seminars in Liver Disease 2008;28:3–25. [7] Kim T, Shijo H, Kokawa H, et al. Risk factors for hemorrhage from gastric fundal varices. Hepatology 1997;25:307–12. [8] Colli A, Gana JC, Turner D, et al. Capsule endoscopy for the diagnosis of oesophageal varices in people with chronic liver disease or portal vein thrombosis. Cochrane Database of Systematic Reviews 2014;10:D8760. [9] Dessouky BA, Abdel AES. Multidetector CT oesophagography: an alternative screening method for endoscopic diagnosis of oesophageal varices and bleeding risk. Arab Journal of Gastroenterology 2013;14:99–108. [10] Nakayama Y, Imuta M, Funama Y, et al. CT portography by multidetector helical CT: comparison of three rendering models. Radiation Medicine 2002;20: 273–9. [11] Perri RE, Chiorean MV, Fidler JL, et al. A prospective evaluation of computerized tomographic (CT) scanning as a screening modality for esophageal varices. Hepatology 2008;47:1587–94. [12] Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Annals of Internal Medicine 2009;151:264–9. W64. [13] Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Annals of Internal Medicine 2011;155:529–36. [14] Reitsma JB, Glas AS, Rutjes AW, et al. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. Journal of Clinical Epidemiology 2005;58:982–90. [15] Jin Y, Wu YL, Wang L, et al. A comparative study on the diagnosis of esophageal and gastric varices between multislice spiral CT portography and endoscopy. Chinese Journal of Digestion 2007;27:78–82. [16] Zhu K, Meng X, Pang P, et al. Gastric varices in patients with portal hypertension: evaluation with multidetector row CT. Journal of Clinical Gastroenterology 2010;44:e108–15. [17] Gulati MS, Paul SB, Arora NK, et al. Esophageal and gastric vasculature in children with extrahepatic portal hypertension: evaluation by intravenous CT portography. Clinical Imaging 2000;24:351–6. [18] He X, Huang ZK, Long LL, et al. Comparison of multi-slice spiral CT portography and endoscopy in the diagnosis of esophageal and gastric varices in liver cirrhosis. China: Chinese Medical Association; 2012. p. 1092–5 [42 Dongsi Xidajie, Beijing 100710, China]. [19] Chen LN, Qu Q, Sun B, et al. Values of endoscopy and computed tomographic angiography for portal vein in diagnosis of portal hypertension. China: WJC Press; 2006. p. 3421–5. [20] Wang AM, Pan KH, Cao GQ, et al. Value of multi-detector CT in the diagnosis of gastric varices in patients with cirrhosis. Radiologic Practice 2014;07:805–7. [21] Kim SH, Kim YJ, Lee JM, et al. Esophageal varices in patients with cirrhosis: multidetector CT esophagography – comparison with endoscopy. United States: Radiological Society of North America Inc.; 2007. p. 759–68. [22] Kim YJ, Raman SS, Yu NC, et al. Esophageal varices in cirrhotic patients: evaluation with liver CT. American Journal of Roentgenology 2007;188:139–44. [23] Moftah SG, Kamal S, Hanna ATK. CT esophagography: non invasive screening and grading of esophageal varices in cirrhosis. The Egyptian Journal of Radiology and Nuclear Medicine 2014;45:263–70. [24] Yu NC, Margolis D, Hsu M, et al. Detection and grading of esophageal varices on liver CT: comparison of standard and thin-section multiplanar reconstructions in diagnostic accuracy. American Journal of Roentgenology 2011;197: 643–9. [25] de Franchis R. Expanding consensus in portal hypertension: report of the Baveno VI Consensus Workshop: stratifying risk and individualizing care for portal hypertension. Journal of Hepatology 2015;63:743–52. [26] Akobeng AK. Understanding diagnostic tests 2: likelihood ratios, pre- and post-test probabilities and their use in clinical practice. Acta Paediatrica 2007;96:487–91. [27] Ba-Ssalamah A, Prokop M, Uffmann M, et al. Dedicated multidetector CT of the stomach: spectrum of diseases. Radiographics 2003;23:625–44. [28] Lee JY, Kim TY, Jeong WK, et al. Clinically severe portal hypertension: role of multi-detector row CT features in diagnosis. Digestive Diseases and Sciences 2014;59:2333–43. [29] Matsumoto A, Kitamoto M, Imamura M, et al. Three-dimensional portography using multislice helical CT is clinically useful for management of gastric fundic varices. American Journal of Roentgenology 2001;176:899–905.
Please cite this article in press as: Tseng Y-J, et al. Computed tomography in evaluating gastroesophageal varices in patients with portal hypertension: A meta-analysis. Dig Liver Dis (2016), http://dx.doi.org/10.1016/j.dld.2016.02.007