EUS for the staging of gastric cancer: a meta-analysis

EUS for the staging of gastric cancer: a meta-analysis

ORIGINAL ARTICLE: Clinical Endoscopy EUS for the staging of gastric cancer: a meta-analysis Simone Mocellin, MD, PhD, Alberto Marchet, MD, Donato Nit...

4MB Sizes 1 Downloads 10 Views

Recommend Documents

No documents
ORIGINAL ARTICLE: Clinical Endoscopy

EUS for the staging of gastric cancer: a meta-analysis Simone Mocellin, MD, PhD, Alberto Marchet, MD, Donato Nitti, MD Padova, Italy

Background: The role of EUS in the locoregional staging of gastric carcinoma is undefined. Objective: We aimed to comprehensively review and quantitatively summarize the available evidence on the staging performance of EUS. Design: We systematically searched the MEDLINE, Cochrane, CANCERLIT, and EMBASE databases for relevant studies published until July 2010. Setting: Formal meta-analysis of diagnostic accuracy parameters was performed by using a bivariate randomeffects model. Patients: Fifty-four studies enrolling 5601 patients with gastric cancer undergoing disease staging with EUS were eligible for the meta-analysis. Main Outcome Measurements: EUS staging accuracy across eligible studies was measured by computing overall sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR). Results: EUS can differentiate T1-2 from T3-4 gastric cancer with high accuracy, with overall sensitivity, specificity, PLR, NLR, and DOR of 0.86 (95% CI, 0.81-0.90), 0.91 (95% CI, 0.89-0.93), 9.8 (95% CI, 7.5-12.8), 0.15 (95% CI, 0.11-0.21), and 65 (95% CI, 41-105), respectively. In contrast, the diagnostic performance of EUS for lymph node status is less reliable, with overall sensitivity, specificity, PLR, NLR, and DOR of 0.69 (95% CI, 0.63-0.74), 0.84 (95% CI, 0.81-0.88), 4.4 (95% CI, 3.6-5.4), 0.37 (95% CI, 0.32-0.44), and 12 (95% CI, 9-16), respectively. Results regarding single T categories (including T1 substages) and Bayesian nomograms to calculate posttest probabilities for any target condition prevalence are also provided. Limitations: Statistical heterogeneity was generally high; unfortunately, subgroup analysis did not identify a consistent source of the heterogeneity. Conclusions: Our results support the use of EUS for the locoregional staging of gastric cancer, which can affect the therapeutic management of these patients. However, clinicians must be aware of the performance limits of this staging tool. (Gastrointest Endosc 2011;73:1122-34.)

Despite its decreasing incidence in Western countries, gastric carcinoma remains the second leading cause of cancer deaths worldwide.1,2 In the United States, 21,000 new cases of this malignancy are estimated to occur in 2011, leading to 10,500 expected deaths.3

Radical surgery still represents the mainstay of treatment with curative intent.4,5 However, new approaches are gaining importance in the therapeutic management of these patients. For instance, EMR is proposed as an alternative to surgery for patients with early gastric cancer in

Abbreviations: DOR, diagnostic odds ratio; NLR, negative likelihood ratio; PLR, positive likelihood ratio; QUADAS, Quality Assessment of Diagnostic Accuracy Studies; SROC, summary receiver-operating characteristic.

doi:10.1016/j.gie.2011.01.030

DISCLOSURE: The authors disclosed no financial relationships relevant to this publication.

Received December 8, 2010. Accepted January 13, 2011. Current affiliations: Meta-Analysis Unit, Department of Oncological and Surgical Sciences, University of Padova, Padova, Italy.

See CME section; p. 1254.

Reprint requests: Simone Mocellin, MD, PhD, Department of Oncological and Surgical Sciences, University of Padova, Via Giustiniani 2, 35128 Padova, Italy.

Copyright © 2011 by the American Society for Gastrointestinal Endoscopy 0016-5107/$36.00

If you would like to chat with an author of this article, you may contact Dr. Mocellin at [email protected]

1122 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 6 : 2011

www.giejournal.org

Mocellin et al

the presence of favorable prognostic features (eg, histologically well-differentiated carcinoma limited to the mucosa, diameter ⬍2 cm, absence of ulceration).6-10 Moreover, different adjuvant and neoadjuvant chemotherapy (with or without with radiotherapy) regimens have been shown to provide significant survival advantage to patients with advanced gastric cancer.11-15 These strategies require reliable staging procedures to ensure the most appropriate (ie, with the highest therapeutic index, the ratio between efficacy and toxicity) treatment for each patient, according to the principles of personalized medicine. In this regard, one of the most studied tools for the locoregional staging of gastric carcinoma is EUS.16,17 This endoscopy-based diagnostic device, which was introduced in the clinical setting in the 1980s, can both distinguish the different layers that compose the gastric wall and visualize the perigastric lymph nodes by means of a miniaturized US probe. Based on numerous reports published over more than 2 decades, EUS is often advocated as the best available method for the locoregional staging of gastric cancer; however, findings are heterogeneous (eg, sensitivity and specificity values range from 50% to 100%), and the comparison with more recent diagnostic techniques (eg, CT, magnetic resonance imaging) is difficult to perform because of the paucity of comparative studies conducted to date.18-25 In addition, although thousands of gastric cancer patients have been enrolled in EUS-based studies, no formal quantitative review of the available evidence has been published that comprehensively examines the staging performance of EUS by using the most appropriate statistical tools for the meta-analysis of diagnostic accuracy data. In this work, we tried to fill this gap in the international literature by performing a formal systematic review and meta-analysis of the available evidence on this subject. Our aim was to provide readers with both a comprehensive overview and a quantitative analysis of the published data regarding the ability of EUS to identify primary tumor depth and regional lymph node status in patients with gastric carcinoma.

MATERIALS AND METHODS Search strategy We performed a comprehensive search of the Englishlanguage literature to identify articles that examined the diagnostic accuracy of EUS (the index test) in the evaluation of primary tumor depth of invasion (according to the American Joint Cancer Committee/Union Internationale Contre le Cancer T categories [T1, T2, T3, T4] as well as in terms of T1 subcategories mucosa [T1m] and submucosa [T1sm]) and regional lymph node status (metastatic [N⫹] vs disease free [N0]) by using histopathology as the reference standard. www.giejournal.org

EUS for gastric cancer

Take-home Message ●

The results of this bivariate meta-analysis support the use of EUS for the locoregional staging of gastric cancer, which can affect the therapeutic management of these patients. However, clinicians must be aware of the performance limits of this staging tool.

We systematically searched the MEDLINE, Cochrane, CANCERLIT, and EMBASE databases for studies published until July 2010, by using the following search terms: “gastric” (or “stomach”), “cancer” (or “carcinoma”), “endoscopic” (or “endoscopy”), “ultrasound” (or “ultrasonography”), “endosonographic,” and “EUS.” We searched for additional references by cross-checking bibliographies of retrieved full-text papers.

Study selection and data extraction We included studies that met all of the following inclusion criteria: (1) a minimal sample size of 10 patients with histologically proven primary carcinoma of the stomach; (2) evaluation of EUS compared with histopathology of primary tumor and lymph nodes; and (3) sufficient data to construct a 2 ⫻ 2 confusion matrix such that the cells in the table could be labeled as true positive, false positive, true negative, and false negative. We excluded studies having possible overlap with the selected studies (ie, studies from the same study group, institution, and period of inclusion). The index test was evaluated for the following diagnostic categories: (A) primary tumor depth: T1 (tumor limited to the mucosa and submucosa) versus non-T1 categories, T2 (tumor invading the muscularis propria ⫾ subserosa) versus non-T2 categories, T3 (tumor invading the serosa) versus non-T3 categories, T4 (tumor infiltrating adjacent structures) versus non-T4 categories, T1-2 (tumor invading the gastric wall up to the muscularis propria/subserosa) versus T3-4 (tumor infiltrating the serosa or adjacent organs), T1m (T1 tumor limited to the mucosa layer) versus T1sm (T1 tumor involving the submucosa layer); (B) regional lymph node status: N⫹ (metastatic lymph nodes) versus N0 (disease-free lymph nodes). Because the TNM staging classification changed over time, it was not possible to assess the diagnostic accuracy of EUS across different N categories (N1, N2, N3). The selection of studies for this meta-analysis was done by 2 independent reviewers (S.M., A.M.). Differences between the 2 reviewers were resolved by having a third reviewer (D.N.) assess the full article: the decision about whether to include the article was made by consensus. Relevant data were extracted from the articles selected for inclusion in the meta-analysis. Data quality was assessed by using a standard procedure according to the Quality Volume 73, No. 6 : 2011 GASTROINTESTINAL ENDOSCOPY 1123

EUS for gastric cancer

Assessment of Diagnostic Accuracy Studies (QUADAS) criteria.26 In addition to these data, we also recorded the following information for each study: overall study characteristics (including the first author, country, language, and date of publication) and characteristics regarding the index test (including type of echoendoscope [radial vs linear array transducer technology], US frequency, and EUS criteria for tumor depth and lymph node status). In case of missing data, we requested information from the authors of the study. When raw data were presented in 3 ⫻ 3 or 4 ⫻ 4 tables (eg, when the tumor depth [T] or lymph node stage was defined by ⬎2 categories), we reconstructed 2 ⫻ 2 contingency tables by considering a given T (or any N-positive category) as the positive state to be distinguished from the other T categories (or from the N-negative cases).

Statistical analysis As currently recommended for meta-analysis of diagnostic accuracy studies,27-29 we used a bivariate randomeffects approach to obtain weighted overall estimates of the sensitivity (rate of cases correctly classified as diseased or true positive rate) and specificity (rate of cases correctly classified as healthy or true negative rate) of EUS in diagnosing both tumor depth infiltration and lymph node status. This approach assumes a bivariate distribution for the logit-transformed values of paired sensitivity and specificity. Overall sensitivity and specificity and their 95% confidence intervals were calculated based on the binominal distributions of the true positives and true negatives. Besides accounting for study size and between-study heterogeneity, the bivariate model adjusts for the frequently observed negative correlation between the sensitivity and the specificity of the index test (threshold effect). An additional advantage of using the bivariate model is that the bivariate nature of the original data can be maintained throughout the analysis, allowing the generation of reliable summary estimates of sensitivity and specificity. A summary receiver-operating characteristic (SROC) curve was constructed as a way to summarize the truepositive and false-positive rates from different studies. The SROC summary point represents the combination of the average sensitivity and specificity and is provided along with a 95% confidence interval as well as prediction interval. The clinical (or patient-relevant) utility of EUS was evaluated by using likelihood ratios (which state how many times more likely particular test results are in patients with disease than in those without disease) to calculate posttest probability based on the Bayesian theorem: this concept was graphically displayed by means of the Fagan nomogram. To formally quantify the extent of between-study variation (ie, heterogeneity), we used the I2 statistic and the 1124 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 6 : 2011

Mocellin et al

Cochrane Q test: the former describes the proportion of total variation in study estimates that is caused by heterogeneity, whereas the latter formally indicates whether heterogeneity is significant.30 Because it is common in diagnostic accuracy studies, we anticipated that there would be substantial between-study variation in reported pairs of sensitivity and specificity. Therefore, to investigate the potential sources of heterogeneity, we used subgroup analysis and meta-regression by including covariates in the bivariate model, which enabled us to assess the effect of the following factors on the diagnostic accuracy of EUS: year of publication, country (Western vs Eastern), EUS technical features (US frequency: higher frequency allows higher definition at the cost of lower penetration), definition of target condition (ie, criteria for T stage and N stage), gastric tumor site (any vs cardia), and prevalence of target condition. The between-study variation possibly caused by the threshold effect (a source of heterogeneity unique to diagnostic meta-analysis) was visually assessed by using the SROC curve and calculated as the squared coefficient of correlation between logit sensitivity and specificity estimated by the bivariate model. Furthermore, formal testing for publication bias (more generally known as small study effect) was conducted by a regression of diagnostic odds ratio (DOR), which describes the odds of positive test results in participants with disease compared with the odds of positive test results in those without disease, on a natural logarithm scale against 1/公ESS (effective sample size), weighting by ESS: P ⬍ .10 for the slope coefficient indicates significant asymmetry of the funnel plot.31 All statistical analyses were conducted by using STATA 11.0 (StataCorp LP, College Station, Tex). All statistical tests were 2 sided, and the significance level was set at 5% (except for funnel plot asymmetry).

RESULTS Eligible studies and quality assessment As shown in Figure 1, the literature search identified 54 eligible studies published between 1988 and 20103285: their main characteristics are reported in Table 1. Overall, 5601 patients were enrolled in 16 different countries, with a mean of 104 patients per study (range 14-388 patients). The quality of the eligible studies, as assessed according to the QUADAS criteria, is reported in Figure 2. The percentage of high-quality studies (ie, those for which a yes response applied) varied between 13% and 100% for each of the 14 items. For most QUADAS items (8/14), all studies were classified as high quality, whereas for 3 of them (ie, uncertain results reporting, reference standard interpreted blind to index test result, and time between index test and reference standard description), the proportion of high-quality studies was less than 50%. www.giejournal.org

Mocellin et al

EUS for gastric cancer

Figure 1. Flow chart of the systematic literature search. GIST, GI stromal tumor.

Primary tumor depth (T Stage) Except for the 6 reports that did not explicitly describe the criteria,50,68,69,71,75,80 the definition of T stage for the index test was uniform across studies and relied on the fact that EUS visualizes the gastric wall as a 5-layer structure: accordingly, T1, T2, T3, and T4 tumors are defined based on the destruction of layers 1-3, 1-4, and 1-5 and infiltration of adjacent organs, respectively. We first performed a meta-analysis of the eligible studies reporting data on T stages grouped into T1-2 versus T3-4 tumors to evaluate the ability of EUS to discriminate between early to intermediate (T1-2) and advanced (T3-4) gastric carcinoma. To this aim, 41 studies were available, for a total of 3510 patients. The sensitivity and specificity of the single studies as well as their combined values are displayed in Figure 3. The SROC curve along with the summary point and the 95% confidence and prediction intervals are shown in Figure 4. Overall, the sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and DOR were 0.86 (95% CI, 0.81-0.90), 0.91 (95% CI, 0.890.93), 9.8 (95% CI, 7.5-12.8), 0.15 (95% CI, 0.11-0.21), and 65 (95% CI, 41-105), respectively. The Fagan plot (Fig. 5) shows that EUS can be clinically informative because it increases the previous probability of being classified as T1-2 from 50% (average prevalence of T1-2 cases) to 91% when positive, and it lowers the same probability to 13% when negative. Between-study heterogeneity was substantial both for sensitivity (I2: 84.9%, Q test P ⬍ .001) and specificity (I2: 70.4%, Q test P ⬍ .001). www.giejournal.org

Subgroup analysis showed that only publication year appears to have a significant impact on EUS diagnostic performance, with studies conducted before the year 2000 reporting on average higher sensitivity (0.93 [95% CI, 0.900.97] vs 0.80 [95% CI, 0.74-0.86]) and specificity (0.94 [95% CI, 0.91-0.96] vs 0.89 [95% CI, 0.86-0.92]). The proportion of heterogeneity likely caused by the threshold effect was very low (0.1%). Regression testing for funnel plot asymmetry showed no evidence of statistically significant publication bias (P ⫽ .39). With regard to the single T categories (T1, T2, T3, and T4), a summary of the meta-analysis findings is reported in Table 2. Fagan plot analysis showed that EUS can be clinically informative because it increases the previous probability of being classified as T1, T2, T3, and T4, respectively, from 35%, 20%, 35%, and 10% (average prevalence of corresponding T categories) to 91%, 63%, 75%, and 76% when positive; moreover, it lowers the above probabilities, respectively, to 9%, 9%, 8%, and 4% when negative. Similarly, with regard to the ability of EUS to differentiate T1 mucosal (T1m) from submucosal (T1sm) tumors, a summary of the meta-analysis findings is provided in Table 2. Compared with the diagnostic accuracy for T1 tumors as a whole, Fagan plot analysis showed that EUS is less likely to be clinically useful to identify T1m tumors because it increases the previous probability of being classified as T1m from 75% (average T1m prevalence) to 92% when positive, and it lowers this probability to 39% when negative. Of note, publication bias could not be excluded for these data (regression test P value ⫽ .06). Volume 73, No. 6 : 2011 GASTROINTESTINAL ENDOSCOPY 1125

EUS for gastric cancer

Mocellin et al

TABLE 1. Main features of the 54 eligible studies (N ⴝ 5601) included in the meta-analysis

Reference

Patients

Murata et al, 198832 Tio et al, 198933 199134

146

Country

Reference test

EUS type

EUS MHz

Stomach site

Japan

Pathology of surgical specimen

Radial

7.5-10

Any

80

the Netherlands

Pathology of surgical specimen

Radial

7.5-12

Any

74

Japan

Pathology of surgical specimen

Radial

12

Any

Botet et al,

199135

50

United States

Pathology of surgical specimen

Radial

7.5-12

Any

Saito et al,

199136

110

Japan

Pathology of surgical specimen

Radial

7.5-12

Any

Italy

Pathology of surgical specimen

Radial

7.5-12

Any

Akahoshi et al,

199337

35

Dittler et al, 199338

254

Germany

Pathology of surgical specimen

Radial

7.5-12

Any

Grimm et al, 199339

147

Germany

Pathology of surgical specimen

Radial

7.5

Any

Ziegler et al, 199340

108

Germany

Pathology of surgical specimen

Radial

7.5-20

Any

Shimizu et al, 199441

128

Japan

Pathology of surgical specimen

Radial

7.5-12

Any

199642

29

France

Pathology of surgical specimen

Radial

7.5-12

Cardia

65

Italy

Pathology of surgical specimen

Radial

7.5-12

Any

69

Taiwan

Pathology of surgical specimen

Radial

7.5-12

Any

149

Japan

Pathology of surgical specimen

Radial

7.5-20

Any

104

Japan

Pathology of EMR or specimen

Radial, linear

20

Any

Akahoshi et al, 199847

75

Japan

Pathology of EMR or specimen

Radial

15

Any

Hunerbein et al, 199848

22

Germany

Pathology of surgical specimen

Radial

12.5

Any

119

Taiwan

Pathology of surgical specimen

Radial

7.5-12

Any

31

Japan

Pathology of surgical specimen

NR

NR

Any

46

Japan

Pathology of EMR or specimen

Radial

20

Any

49

Japan

Pathology of EMR or specimen

Radial, linear

20

Any

79

Italy

Pathology of surgical specimen

Radial

7.5-12

Any

Tseng et al, 200054

74

Taiwan

Pathology of surgical specimen

Radial

7.5-12

Any

Willis et al, 200055

116

Germany

Pathology of surgical specimen

Radial

7.5-12

Any

Chen et al, 200256

57

Taiwan

Pathology of surgical specimen

Radial

7.5-20

Any

220

Japan

Pathology of EMR or specimen

Radial

12-20

Any

32

China

Pathology of surgical specimen

Radial

7.5-20

Any

48

Korea

Pathology of EMR or specimen

Radial

20

Any

51

Germany

Pathology of surgical specimen

Radial

7.5-12

Any

49

Germany

Pathology of EMR or specimen

NR

12.5

Any

India

Pathology of surgical specimen

Radial

7.5

Any Any

Caletti et al,

Francois et al, Massari et al, Perng et al,

199643

199644

Hamada, et al,

199745

Yanai et al, 199746

Wang et al, 199849 Nakamura et al, Okamura et al, Yanai et al,

199950

199951

199952

Mancino et al,

200053

Hizawa et al, 200257 Xi et al,

200358

Bhandari et al,

200459

Habermann et al, Hunerbein et al,

200460

200461

Javaid et al, 200462

112

Polkowski et al, 200463

88

Poland

Pathology of surgical specimen

Radial

7.5-12

Shimoyama et al, 200464

45

Japan

Pathology of surgical specimen

Linear

7.5

Cardia

Pedrazzani et al, 200565

51

Italy

Pathology of surgical specimen

Linear

7.5

Cardia

293

Japan

Pathology of EMR or specimen

Radial

7.5-20

Any

267

Japan

Pathology of EMR or specimen

NR

12-20

Any

Singapore

Pathology of surgical specimen

Radial

7.5-12

Any

Yoshida et al, Akashi et al, Ang et al,

200566

200667

200668

57

1126 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 6 : 2011

www.giejournal.org

Mocellin et al

EUS for gastric cancer

TABLE 1. Continued Reference

Patients

Arocena et al,

200669

Ganpathi et al,

200670

17 102

Country

Reference test

EUS type

EUS MHz

Stomach site

Spain

Pathology of surgical specimen

Linear

12.5

Any

Singapore

Pathology of surgical specimen

Radial

7.5-12

Any

Potrc et al, 200671

82

Slovenia

Pathology of surgical specimen

Radial

7.5-12

Any

Tsendsuren et al, 200672

41

China

Pathology of surgical specimen

Linear

5-7.5

Any

Barbour et al, 200773

206

USA

Pathology of surgical specimen

Radial

7.5-12

Cardia

Bentrem et al, 200774

218

United States

Pathology of surgical specimen

NR

7.5-12

Any

63

China

Pathology of surgical specimen

Radial

7.5-20

Any

44

UK

Pathology of surgical specimen

Radial

7.5-12

Cardia

Tan et al,

200775

Blackshaw et al,

200876

200877

75

Hong Kong

Pathology of surgical specimen

Radial

12-20

Any

200878

40

Korea

Pathology of surgical specimen

Radial

7.5-12

Any

Ahn et al, 200979

71

Korea

Pathology of surgical specimen

Radial

5-12

Any

Heye et al, 200980

14

Germany

Pathology of surgical specimen

NR

NR

Any

Mouri et al, 200981

222

Japan

Pathology of EMR or specimen

Radial

12-20

Any

Choi et al, 201082

388

Korea

Pathology of EMR or specimen

Radial

12

Any

277

Korea

Pathology of EMR or specimen

Radial

5-20

Any

176

Korea

Pathology of EMR or specimen

Radial

20

Any

36

Spain

Pathology of surgical specimen

Radial

7.5-20

Any

Lok et al,

Park et al,

Hwang et al, Kim et al,

201083

201084

Repiso et al,

201085

NR, Not reported.

Figure 2. The quality of the eligible studies as assessed according to the 14 items included in the Quality Assessment of Diagnostic Accuracy Studies criteria. Dark blue bars indicate yes, light blue bars no, and green bars uncertain/unclear.

www.giejournal.org

Volume 73, No. 6 : 2011 GASTROINTESTINAL ENDOSCOPY 1127

EUS for gastric cancer

Mocellin et al

Figure 3. EUS diagnostic performance to distinguish T1-2 from T3-4 tumors: forest plot of the studies included in the meta-analysis.

Lymph node status (N Stage) We then performed a meta-analysis of the eligible studies reporting data on N stage (positive vs negative) to evaluate the ability of EUS to diagnose regional lymph node status of patients with gastric carcinoma. To this aim, 39 studies were available, for a total of 3315 patients. Except for the 6 reports that did not describe the criteria,50,58,59,71,75,80 the EUS definition of N stage varied across studies; in particular, some investigators relied exclusively on lymph node size (⬎8 mm),44,49,54,60,78,83 whereas most relied on lymph node morphology (eg, sharp margin and hypoechoic pattern) coupled or not with size value. Sensitivities and specificities of the single studies as well as their combined values are displayed in Figure 6. The SROC curve along with the summary point and its 95% confidence and prediction intervals are shown in Figure 7. Overall sensitivity, specificity, PLR, NLR, and DOR were 0.69 (95% CI, 0.63-0.74), 0.84 (95% CI, 0.81-0.88), 4.4 (95% CI, 3.6-5.4), 0.37 (95% CI, 0.32-0.44), and 12 (95% CI, 9-16), respectively. The Fagan plot (Fig. 8) shows that EUS could be clinically informative because it increases the previous probability of being classified as N⫹ from 55% (average preva1128 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 6 : 2011

lence of N⫹ cases) to 84% when positive, and it lowers the same probability to 31% when negative. Between-study heterogeneity was significant both for sensitivity (I2: 81.5%, Q test P ⬍ .001) and specificity (I2: 72.4%, Q test P ⬍ .001). The proportion of heterogeneity likely caused by the threshold effect was substantial (37%). Subgroup analysis showed higher sensitivity (0.72 [95% CI, 0.66-0.78] vs 0.58 [95% CI, 0.47-0.70]) and lower specificity (0.83 [95% CI, 0.78-0.87] vs 0.88 [95% CI, 0.83-0.93]) in studies with higher disease prevalence (lymph node positivity ⬎50%); in contrast, studies using high-frequency EUS reported on average lower sensitivity (0.60 [95% CI, 0.49-0.71] vs 0.72 [95% CI, 0.67-0.78]) and higher specificity (0.88 [95% CI, 0.83-0.94] vs 0.83 [95% CI, 0.78-0.87]). Regression testing for funnel plot asymmetry showed no evidence of statistically significant publication bias (P ⫽ .93).

DISCUSSION This meta-analysis, the first to include such a large cohort of patients (N ⫽ 5601) and conducted with modern statistical methods, quantitatively summarizes the availwww.giejournal.org

Mocellin et al

EUS for gastric cancer

Figure 4. EUS diagnostic performance to distinguish T1-2 from T3-4 tumors: hierarchical summary receiver-operating characteristic (HSROC) curve along with the summary point and the 95% confidence and prediction intervals. Gray circles, which represent the single study estimates of sensitivity and specificities, are proportional to the number of patients enrolled.

able evidence of the diagnostic performance of EUS for the locoregional staging of gastric cancer. We found that EUS evaluation is associated with clinically relevant high performance rates (Figs. 3-5) for differentiating between early to intermediate (T1-2) and advanced (T3-4) primary gastric tumors. In particular, the average positive and negative likelihood ratios of EUS (9.8 and 0.15, respectively) approach the desirable values (ⱖ10 and ⱕ0.1, respectively) for optimal diagnostic tests.86 These figures, along with the Bayesian nomogram provided in Figure 5, support the use of EUS in the clinical setting, with special regard to the therapeutic management of patients who are to be selected for neoadjuvant treatments before surgery, as mentioned previously. The diagnostic accuracy for the single T categories (T1, T2, T3, and T4) is overall less satisfactory (Table 2), although one could pinpoint that the identification of each single category is less important than the ability to differentiate T1-2 from T3-4 tumors from a practical viewpoint (ie, the therapeutic management of patients). In the T1 category, a clinically relevant issue is the differentiation of T1m from T1sm lesions, which would enable clinicians to better select patients who might benefit from EMR. Unfortunately, the average EUS performance in this subset of www.giejournal.org

Figure 5. EUS diagnostic performance to distinguish T1-2 from T3-4 tumors: Fagan plot (Bayesian nomogram).

patients does not appear to be sufficiently informative on practical grounds, although the data on this subject are from a smaller number of studies, and publication bias could not be excluded in our meta-analysis (Table 2). With regard to regional lymph node status, the ability of EUS to distinguish between positive and negative cases is unsatisfactory. However, although all diagnostic performance parameters are worse than those described for the T1-2/T3-4 staging issue, the Fagan plot (Fig. 8) shows that EUS could still be clinically informative because it increases the previous probability of being classified as N⫹ from 55% (average prevalence of N⫹ cases) to 84% when positive, and it lowers the same probability to 31% when negative. Accordingly, EUS can provide clinicians with an estimate of the risk of patients with lymph node metastatic disease; although this information is not as reliable as that supplied for primary tumor depth, it can still represent a further clue in selecting patients for preoperative treatments. Overall, a common feature of our findings is that statistical heterogeneity was generally high, a very frequent observation in meta-analysis of diagnostic studies. UnforVolume 73, No. 6 : 2011 GASTROINTESTINAL ENDOSCOPY 1129

EUS for gastric cancer

Mocellin et al

TABLE 2. Meta-analysis of EUS diagnostic performance for T1, T2, T3, and T4 as well as T1 mucosal (T1m) gastric carcinoma T1

T2

T3

T4

T1m

41

39

39

34

15

3866

3475

3444

3017

2318

Sensitivity (95% CI)

0.83 (0.77-0.88)

0.65 (0.57-0.72)

0.86 (0.83-0.89)

0.66 (0.52-0.77)

0.83 (0.76-0.89)

Specificity (95% CI)

0.96 (0.93-0.97)

0.91 (0.88-0.92)

0.85 (0.80-0.89)

0.98 (0.97-0.98)

0.79 (0.65-0.88)

PLR (95% CI)

19.8 (12.7-31.1)

6.9 (5.4-8.9)

5.7 (4.3-7.5)

28.1 (18.5-42.5)

3.9 (2.4-6.3)

NLR (95% CI)

0.18 (0.13-0.24)

0.39 (0.31-0.48)

0.16 (0.13-0.2)

0.35 (0.24-0.51)

0.21 (0.16-0.28)

DOR (95% CI)

112 (70-179)

35 (24-52)

80 (41-153)

No. of studies Patients

18 (12-27)

19 (13-27)

Heterogeneity I 2, % Q test P value Threshold effect, %

99

95

97

93

99

⬍.001

⬍.001

⬍.001

⬍.001

⬍.001

44

1

1

5

96

.80

.16

.92

.30

.06

Publication bias test P value

CI, Confidence interval; PLR, positive likelihood ratio; NLR, negative likelihood ratio; DOR, diagnostic odds ratio.

Figure 6. EUS diagnostic performance to distinguish lymph node–positive from lymph node–negative tumors: forest plot of the studies included in the meta-analysis.

1130 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 6 : 2011

www.giejournal.org

Mocellin et al

EUS for gastric cancer

Figure 7. EUS diagnostic performance to distinguish lymph node– positive from lymph node–negative tumors: hierarchical summary receiver-operating characteristic (HSROC) curve along with the summary point and the 95% confidence and prediction intervals. Gray circles, which represent the single study estimates of sensitivity and specificities, are proportional to the number of patients enrolled.

tunately, subgroup analysis did not identify a consistent source of heterogeneity, which appears to be linked at least in part to year of publication, a combination of threshold effects, disease prevalence and EUS frequency, and publication bias, depending on the type of diagnostic issue addressed (T1-2 vs T3-4, N⫹ vs N⫺, and T1m vs T1sm, respectively). Another potential source of heterogeneity is operator experience, an aspect often emphasized for US-based diagnostic imaging; however, this issue could not be computationally addressed in our metaanalysis because of the scarcity (or absence) of data reported by single studies on this subject. Based on these findings, no specific recommendation can be made on how to improve the standardization (and thus the consistency) of the EUS results. With regard to the already existing literature on this subject, 2 systematic reviews (without meta-analysis) have been published on EUS and gastric cancer locoregional staging between 2007 and 2009. In the former,22 which was dedicated to primary tumor depth, the authors conclude that EUS remains the first-choice imaging modality in T staging compared with magnetic resonance imaging and CT; in the latter,18 which addressed the issue of lymph node status, the same investigators conclude that EUS (like www.giejournal.org

Figure 8. EUS diagnostic performance to distinguish lymph nodepositive from lymph node–negative tumors: Fagan plot (Bayesian nomogram).

magnetic resonance imaging and CT) cannot reliably be used to confirm or exclude lymph node metastasis in gastric cancer. These conclusions are similar to those that we have drawn, although our meta-analysis provides formal evidence to sustain these hypotheses and more importantly provides readers with a quantification of the average performance of this endoscopic tool. The latter aspect, along with the Bayesian nomograms, allows clinicians to get a precise sense of the risk of making errors (both in terms of false-positive and false-negative predictions) while using EUS, which ultimately can help them optimize the therapeutic management of patients based on probabilities and not generic dichotomic (works/does not work) expert opinions. In this respect, we strongly believe that the much-cited ”personalized medicine” passes also through the use of the available tests based on the awareness of their predictive values and not just only on the knowledge of their ”raw” binary response (positive/negative). Finally, between 2001 and 2008, 2 meta-analyses were published on both T staging and N staging of gastric Volume 73, No. 6 : 2011 GASTROINTESTINAL ENDOSCOPY 1131

EUS for gastric cancer

cancer with EUS. In the first, the authors conclude that EUS is more accurate for the staging of advanced rather than early disease24; in the second, which also included esophageal carcinomas, the investigators state that EUS is highly effective in differentiating T1-2 from T3-4 tumors.19 However, these findings are not comparable to ours because of the fact that the methods used in that work (based on the Moses-Littenberg model87) are no longer considered scientifically sound for the meta-analysis of diagnostic accuracy studies27-29 and because the number of included studies (n ⫽ 13 and 22, respectively) was much lower than what we could retrieve and analyze (n ⫽ 54). Taken together, our results support the use of EUS for the locoregional staging of gastric cancer, provided that clinicians are aware of the diagnostic performance limits of this tool, as outlined earlier. Technological improvements (in particular in the definition of lymph node status) such as those recently proposed88-90 may lead to better EUS performance rates and thus to the optimization of gastric cancer staging, which should ultimately ameliorate the clinical management of these patients. ACKNOWLEDGMENTS The authors thank Salvatore Correra for his help in retrieving the relevant articles. They also thank Dr. Marta Briarava for setting up the dedicated database used to perform this meta-analysis. REFERENCES 1. Shah MA, Kelsen DP. Gastric cancer: a primer on the epidemiology and biology of the disease and an overview of the medical management of advanced disease. J Natl Compr Canc Netw 2010;8:437-47. 2. Crew KD, Neugut AI. Epidemiology of gastric cancer. World J Gastroenterol 2006;12:354-62. 3. Jemal A, Siegel R, Xu J, et al. Cancer statistics, 2010. CA Cancer J Clin 2010;60:277-300. 4. Dicken BJ, Bigam DL, Cass C, et al. Gastric adenocarcinoma: review and considerations for future directions. Ann Surg 2005;241:27-39. 5. Jackson C, Cunningham D, Oliveira J. Gastric cancer: ESMO clinical recommendations for diagnosis, treatment and follow-up. Ann Oncol 2009;20(Suppl 4):34-6. 6. Wang KK, Prasad G, Tian J. Endoscopic mucosal resection and endoscopic submucosal dissection in esophageal and gastric cancers. Curr Opin Gastroenterol 2010;26:453-8. 7. Deprez PH, Bergman JJ, Meisner S, et al. Current practice with endoscopic submucosal dissection in Europe: position statement from a panel of experts. Endoscopy. Epub 2010 Sep 23. 8. Yamamoto H. Technology insight: endoscopic submucosal dissection of gastrointestinal neoplasms. Nat Clin Pract Gastroenterol Hepatol 2007;4:511-20. 9. Bennett C, Wang Y, Pan T. Endoscopic mucosal resection for early gastric cancer. Cochrane Database Syst Rev 2009:CD004276. 10. Gotoda T. Endoscopic resection of early gastric cancer. Gastric Cancer 2007;10:1-11. 11. Ajani JA, Barthel JS, Bekaii-Saab T, et al. Gastric cancer. J Natl Compr Canc Netw 2010;8:378-409. 12. Paoletti X, Oba K, Burzykowski T, et al. Benefit of adjuvant chemotherapy for resectable gastric cancer: a meta-analysis. JAMA 2010;303:1729-37.

1132 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 6 : 2011

Mocellin et al

13. Mezhir JJ, Tang LH, Coit DG. Neoadjuvant therapy of locally advanced gastric cancer. J Surg Oncol 2010;101:305-14. 14. House MG, Brennan MF. Neoadjuvant therapy for gastric cancer. Adv Surg 2008;42:151-68. 15. Das P, Fukami N, Ajani JA. Combined modality therapy of localized gastric and esophageal cancers. J Natl Compr Canc Netw 2006;4:375-82. 16. Polkowski M. Endosonographic staging of upper intestinal malignancy. Best Pract Res Clin Gastroenterol 2009;23:649-61. 17. Byrne MF, Jowell PS. Gastrointestinal imaging: endoscopic ultrasound. Gastroenterology 2002;122:1631-48. 18. Kwee RM, Kwee TC. Imaging in assessing lymph node status in gastric cancer. Gastric Cancer 2009;12:6-22. 19. Puli SR, Batapati Krishna Reddy J, Bechtold ML, et al. How good is endoscopic ultrasound for TNM staging of gastric cancers? A meta-analysis and systematic review. World J Gastroenterol 2008;14:4011-9. 20. Kwee RM, Kwee TC. The accuracy of endoscopic ultrasonography in differentiating mucosal from deeper gastric cancer. Am J Gastroenterol 2008;103:1801-9. 21. Reddy NK, Markowitz AB, Abbruzzese JL, et al. Knowledge of indications and utilization of EUS: a survey of oncologists in the United States. J Clin Gastroenterol 2008;42:892-6. 22. Kwee RM, Kwee TC. Imaging in local staging of gastric cancer: a systematic review. J Clin Oncol 2007;25:2107-16. 23. Weber WA, Ott K. Imaging of esophageal and gastric cancer. Semin Oncol 2004;31:530-41. 24. Kelly S, Harris KM, Berry E, et al. A systematic review of the staging performance of endoscopic ultrasound in gastro-oesophageal carcinoma. Gut 2001;49:534-9. 25. Messmann H, Schlottmann K. Role of endoscopy in the staging of esophageal and gastric cancer. Semin Surg Oncol 2001;20:78-81. 26. Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3:25. 27. Harbord RM, Whiting P, Sterne JA, et al. An empirical comparison of methods for meta-analysis of diagnostic accuracy showed hierarchical models are necessary. J Clin Epidemiol 2008;61:1095-103. 28. Leeflang MM, Deeks JJ, Gatsonis C, et al. Systematic reviews of diagnostic test accuracy. Ann Intern Med 2008;149:889-97. 29. Reitsma JB, Glas AS, Rutjes AW, et al. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol 2005;58:982-90. 30. Higgins JP, Thompson SG. Quantifying heterogeneity in a metaanalysis. Stat Med 2002;21:1539-58. 31. Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol 2005;58:882-93. 32. Murata Y, Suzuki S, Hashimoto H. Endoscopic ultrasonography of the upper gastrointestinal tract. Surg Endosc 1988;2:180-3. 33. Tio TL, Coene PP, Schouwink MH, et al. Esophagogastric carcinoma: preoperative TNM classification with endosonography. Radiology 1989; 173:411-7. 34. Akahoshi K, Misawa T, Fujishima H, et al. Preoperative evaluation of gastric cancer by endoscopic ultrasound. Gut 1991;32:479-82. 35. Botet JF, Lightdale CJ, Zauber AG, et al. Preoperative staging of gastric cancer: comparison of endoscopic US and dynamic CT. Radiology 1991; 181:426-32. 36. Saito N, Takeshita K, Habu H, et al. The use of endoscopic ultrasound in determining the depth of cancer invasion in patients with gastric cancer. Surg Endosc 1991;5:14-9. 37. Caletti G, Ferrari A, Brocchi E, et al. Accuracy of endoscopic ultrasonography in the diagnosis and staging of gastric cancer and lymphoma. Surgery 1993;113:14-27. 38. Dittler HJ, Siewert JR. Role of endoscopic ultrasonography in gastric carcinoma. Endoscopy 1993;25:162-6. 39. Grimm H, Binmoeller KF, Hamper K, et al. Endosonography for preoperative locoregional staging of esophageal and gastric cancer. Endoscopy 1993;25:224-30.

www.giejournal.org

Mocellin et al

40. Ziegler K, Sanft C, Zimmer T, et al. Comparison of computed tomography, endosonography, and intraoperative assessment in TN staging of gastric carcinoma. Gut 1993;34:604-10. 41. Shimizu S, Tada M, Kawai K. Endoscopic ultrasonography for early gastric cancer. Endoscopy 1994;26:767-8. 42. Francois E, Peroux J, Mouroux J, et al. Preoperative endosonographic staging of cancer of the cardia. Abdom Imaging 1996;21:483-7. 43. Massari M, Cioffi U, De Simone M, et al. Endoscopic ultrasonography for preoperative staging of gastric carcinoma. Hepatogastroenterology 1996;43:542-6. 44. Perng DS, Jan CM, Wang WM, et al. Computed tomography, endoscopic ultrasonography and intraoperative assessment in TN staging of gastric carcinoma. J Formos Med Assoc 1996;95:378-85. 45. Hamada S, Akahoshi K, Chijiiwa Y, et al. Relationship between histological type and endosonographic detection of regional lymph node metastases in gastric cancer. Br J Radiol 1997;70:697-702. 46. Yanai H, Matsumoto Y, Harada T, et al. Endoscopic ultrasonography and endoscopy for staging depth of invasion in early gastric cancer: a pilot study. Gastrointest Endosc 1997;46:212-6. 47. Akahoshi K, Chijiwa Y, Hamada S, et al. Pretreatment staging of endoscopically early gastric cancer with a 15 MHz ultrasound catheter probe. Gastrointest Endosc 1998;48:470-6. 48. Hunerbein M, Ghadimi BM, Haensch W, et al. Transendoscopic ultrasound of esophageal and gastric cancer using miniaturized ultrasound catheter probes. Gastrointest Endosc 1998;48:371-5. 49. Wang JY, Hsieh JS, Huang YS, et al. Endoscopic ultrasonography for preoperative locoregional staging and assessment of resectability in gastric cancer. Clin Imaging 1998;22:355-9. 50. Nakamura K, Kamei T, Ohtomo N, et al. Gastric carcinoma confined to the muscularis propria: how can we detect, evaluate, and cure intermediate-stage carcinoma of the stomach? Am J Gastroenterol 1999;94:2251-5. 51. Okamura S, Tsutsui A, Muguruma N, et al. The utility and limitations of an ultrasonic miniprobe in the staging of gastric cancer. J Med Invest 1999; 46:49-53. 52. Yanai H, Noguchi T, Mizumachi S, et al. A blind comparison of the effectiveness of endoscopic ultrasonography and endoscopy in staging early gastric cancer. Gut 1999;44:361-5. 53. Mancino G, Bozzetti F, Schicchi A, et al. Preoperative endoscopic ultrasonography in patients with gastric cancer. Tumori 2000;86:139-41. 54. Tseng LJ, Mo LR, Tio TL, et al. Video-endoscopic ultrasonography in staging gastric carcinoma. Hepatogastroenterology 2000;47:897-900. 55. Willis S, Truong S, Gribnitz S, et al. Endoscopic ultrasonography in the preoperative staging of gastric cancer: accuracy and impact on surgical therapy. Surg Endosc 2000;14:951-4. 56. Chen CH, Yang CC, Yeh YH. Preoperative staging of gastric cancer by endoscopic ultrasound: the prognostic usefulness of ascites detected by endoscopic ultrasound. J Clin Gastroenterol 2002;35:321-7. 57. Hizawa K, Iwai K, Esaki M, et al. Is endoscopic ultrasonography indispensable in assessing the appropriateness of endoscopic resection for gastric cancer? Endoscopy 2002;34:973-8. 58. Xi WD, Zhao C, Ren GS. Endoscopic ultrasonography in preoperative staging of gastric cancer: determination of tumor invasion depth, nodal involvement and surgical resectability. World J Gastroenterol 2003;9: 254-7. 59. Bhandari S, Shim CS, Kim JH, et al. Usefulness of three-dimensional, multidetector row CT (virtual gastroscopy and multiplanar reconstruction) in the evaluation of gastric cancer: a comparison with conventional endoscopy, EUS, and histopathology. Gastrointest Endosc 2004;59:619-26. 60. Habermann CR, Weiss F, Riecken R, et al. Preoperative staging of gastric adenocarcinoma: comparison of helical CT and endoscopic US. Radiology 2004;230:465-71. 61. Hunerbein M, Handke T, Ulmer C, et al. Impact of miniprobe ultrasonography on planning of minimally invasive surgery for gastric and colonic tumors. Surg Endosc 2004;18:601-5.

www.giejournal.org

EUS for gastric cancer

62. Javaid G, Shah OJ, Dar MA, et al. Role of endoscopic ultrasonography in preoperative staging of gastric carcinoma. ANZ J Surg 2004;74:108-11. 63. Polkowski M, Palucki J, Wronska E, et al. Endosonography versus helical computed tomography for locoregional staging of gastric cancer. Endoscopy 2004;36:617-23. 64. Shimoyama S, Yasuda H, Hashimoto M, et al. Accuracy of linear-array EUS for preoperative staging of gastric cardia cancer. Gastrointest Endosc 2004;60:50-5. 65. Pedrazzani C, Bernini M, Giacopuzzi S, et al. Evaluation of Siewert classification in gastro-esophageal junction adenocarcinoma: what is the role of endoscopic ultrasonography? J Surg Oncol 2005;91:226-31. 66. Yoshida S, Tanaka S, Kunihiro K, et al. Diagnostic ability of highfrequency ultrasound probe sonography in staging early gastric cancer, especially for submucosal invasion. Abdom Imaging 2005;30:518-23. 67. Akashi K, Yanai H, Nishikawa J, et al. Ulcerous change decreases the accuracy of endoscopic ultrasonography diagnosis for the invasive depth of early gastric cancer. Int J Gastrointest Cancer 2006;37:133-8. 68. Ang TL, Ng TM, Fock KM, et al. Accuracy of endoscopic ultrasound staging of gastric cancer in routine clinical practice in Singapore. Chin J Dig Dis 2006;7:191-6. 69. Arocena MG, Barturen A, Bujanda L, et al. MRI and endoscopic ultrasonography in the staging of gastric cancer. Rev Esp Enferm Dig 2006; 98:582-90. 70. Ganpathi IS, So JB, Ho KY. Endoscopic ultrasonography for gastric cancer: does it influence treatment? Surg Endosc 2006;20:559-62. 71. Potrc S, Skalicky M, Ivanecz A. Does endoscopic ultrasound staging already allow individual treatment regimens in gastric cancer. Wien Klin Wochenschr 2006;118(Suppl 2):48-51. 72. Tsendsuren T, Jun SM, Mian XH. Usefulness of endoscopic ultrasonography in preoperative TNM staging of gastric cancer. World J Gastroenterol 2006;12:43-7. 73. Barbour AP, Rizk NP, Gerdes H, et al. Endoscopic ultrasound predicts outcomes for patients with adenocarcinoma of the gastroesophageal junction. J Am Coll Surg 2007;205:593-601. 74. Bentrem D, Gerdes H, Tang L, et al. Clinical correlation of endoscopic ultrasonography with pathologic stage and outcome in patients undergoing curative resection for gastric cancer. Ann Surg Oncol 2007;14: 1853-9. 75. Tan SY, Wang JY, Shen L, et al. Relationship between preoperative staging by endoscopic ultrasonography and MMP-9 expression in gastric carcinoma. World J Gastroenterol 2007;13:2108-12. 76. Blackshaw G, Lewis WG, Hopper AN, et al. Prospective comparison of endosonography, computed tomography, and histopathological stage of junctional oesophagogastric cancer. Clin Radiol 2008;63:1092-8. 77. Lok KH, Lee CK, Yiu HL, et al. Current utilization and performance status of endoscopic ultrasound in a community hospital. J Dig Dis 2008;9: 41-7. 78. Park SR, Lee JS, Kim CG, et al. Endoscopic ultrasound and computed tomography in restaging and predicting prognosis after neoadjuvant chemotherapy in patients with locally advanced gastric cancer. Cancer 2008;112:2368-76. 79. Ahn HS, Lee HJ, Yoo MW, et al. Diagnostic accuracy of T and N stages with endoscopy, stomach protocol CT, and endoscopic ultrasonography in early gastric cancer. J Surg Oncol 2009;99:20-7. 80. Heye T, Kuntz C, Dux M, et al. CT and endoscopic ultrasound in comparison to endoluminal MRI: preliminary results in staging gastric carcinoma. Eur J Radiol 2009;70:336-41. 81. Mouri R, Yoshida S, Tanaka S, et al. Usefulness of endoscopic ultrasonography in determining the depth of invasion and indication for endoscopic treatment of early gastric cancer. J Clin Gastroenterol 2009;43: 318-22. 82. Choi J, Kim SG, Im JP, et al. Comparison of endoscopic ultrasonography and conventional endoscopy for prediction of depth of tumor invasion in early gastric cancer. Endoscopy 2010;42:705-13. 83. Hwang SW, Lee DH, Lee SH, et al. Preoperative staging of gastric cancer by endoscopic ultrasonography and multidetector-row computed tomography. J Gastroenterol Hepatol 2010;25:512-8.

Volume 73, No. 6 : 2011 GASTROINTESTINAL ENDOSCOPY 1133

EUS for gastric cancer

84. Kim GH, Park do Y, Kida M, et al. Accuracy of high-frequency catheterbased endoscopic ultrasonography according to the indications for endoscopic treatment of early gastric cancer. J Gastroenterol Hepatol 2010;25:506-11. 85. Repiso A, Gomez-Rodriguez R, Lopez-Pardo R, et al. Usefulness of endoscopic ultrasonography in preoperative gastric cancer staging: diagnostic yield and therapeutic impact. Rev Esp Enferm Dig 2010; 102:413-20. 86. Deeks JJ. Systematic reviews in health care: Systematic reviews of evaluations of diagnostic and screening tests. BMJ 2001;323:157-62. 87. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic ap-

1134 GASTROINTESTINAL ENDOSCOPY Volume 73, No. 6 : 2011

Mocellin et al

proaches and some additional considerations. Stat Med 1993; 12:1293-316. 88. Hawes RH. The evolution of endoscopic ultrasound: improved imaging, higher accuracy for fine needle aspiration and the reality of endoscopic ultrasound-guided interventions. Curr Opin Gastroenterol 2010;26:436-44. 89. Kanick SC, van der Leest C, Aerts JG, et al. Integration of single-fiber reflectance spectroscopy into ultrasound-guided endoscopic lung cancer staging of mediastinal lymph nodes. J Biomed Opt 2010;15: 017004. 90. Saftoiu A, Gheonea DI. Tridimensional (3D) endoscopic ultrasound - a pictorial review. J Gastrointest Liver Dis 2009;18:501-5.

www.giejournal.org