Digestive Diseases and Sciences

, Volume 53, Issue 9, pp 2405–2414

Accuracy of Endoscopic Ultrasound in the Diagnosis of Distal and Celiac Axis Lymph Node Metastasis in Esophageal Cancer: A Meta-Analysis and Systematic Review

Authors

  • Srinivas R. Puli
    • Division of Gastroenterology and HepatologyUniversity of Missouri-Columbia
  • Jyotsna B. K. Reddy
    • Division of Gastroenterology and HepatologyUniversity of Missouri-Columbia
  • Matthew L. Bechtold
    • Division of Gastroenterology and HepatologyUniversity of Missouri-Columbia
  • Mainor R. Antillon
    • Division of Gastroenterology and HepatologyUniversity of Missouri-Columbia
    • Division of Gastroenterology and HepatologyUniversity of Missouri-Columbia
Original Paper

DOI: 10.1007/s10620-007-0152-3

Cite this article as:
Puli, S.R., Reddy, J.B.K., Bechtold, M.L. et al. Dig Dis Sci (2008) 53: 2405. doi:10.1007/s10620-007-0152-3

Abstract

Background Published data on the accuracy of endoscopic ultrasound (EUS) for staging distant and celiac axis lymph node (CLN) metastasis in patients with esophageal cancer (ECA) has varied. Aim To evaluate the accuracy of EUS in diagnosing distal and CLN metastasis in ECA patients. Method Study selection: EUS studies confirmed by surgery were selected. Statistical method: Pooling was conducted by both fixed and random-effects models. Results Data were extracted from 25 studies (N = 2029) which met the inclusion criteria. In ECA patients, pooled sensitivity of EUS was 67.2% (95% CI: 62.6–71.6) in diagnosis of distal metastasis and 66.6% (95% CI: 61.9–71.1) in diagnosis of CLN metastasis. EUS had a pooled specificity of 97.9% (95% CI: 97.1–98.6) for distal metastasis and 98.1% (95% CI: 97.3–98.7) for CLN metastasis. Conclusions Although EUS has excellent specificity in accurately diagnosing distal and CLN metastasis in patients with ECA, the sensitivity is low.

Keywords

Meta-analysisSystematic reviewEsophageal cancerDistal metastasisCeliac axis lymphadenopathy

Introduction

Esophageal cancer (ECA) is a devastating disease with a significant impact on patients’ lives and the worldwide health-care system. The incidence of ECA is rising, and currently affects 1–2% of people in the United States and up to 15% of people undergoing endoscopy for gastroesophageal reflux disease (GERD) [15]. Although the overall incidence of ECA is rare, the impact of this disease is significant throughout the world because of its increasing incidence and significant mortality (five-year mortality is >80%) [6].

Based upon the increasing incidence and devastating consequences of ECA, increasing amounts of resources have been evaluated and implemented in an effort to prevent, stage, and treat this terrible disease. Once the diagnosis is made, staging has a significant impact on treatment decisions. With improved staging, the most efficacious treatment may be initiated, potentially leading to prolonged quantity and quality of life.

Staging of ECA is extremely important since it helps differentiate treatment options based upon patient survival. Based on the 1996 national cancer data base, the five-year survival rate of the stages for ECA is: stage 0, 52%; stage I, 42%; stage II, 29%; stage III, 15%; and stage IV, 3% [7]. To improve survival, many treatment modalities have been utilized for ECA, including surgery, radiation therapy, chemotherapy, and combinations of these [7].

ECA patients with distal or CLN metastases have poor survival and high risk of recurrence if treated with surgery alone [812]. Although multiple treatment regimens exist and there is overlap of treatment regimes for each stage, the stage of disease is very important in guiding treatment and predicting outcomes.

Many staging modalities have been utilized to detect distal metastasis of ECA, including CT of the chest or abdomen, magnetic resonance imaging (MRI), positron emission tomography (PET), and endoscopic ultrasound (EUS). CT of the chest or abdomen provides important information regarding size of tumor, lymph node involvement, and potential metastatic lesions. However, CT of the abdomen alone for detection of celiac axis involvement has a sensitivity of 8% and a specificity of 100% [13]. Even helical CT of the abdomen is not overly reliable. By helical CT, 25% of patients thought to have resectable disease had either metastatic involvement of celiac lymph nodes or T4 disease by EUS/FNA [14]. MRI has been shown to be useful in preoperative evaluation and as accurate as CT in the staging of ECA, although studies vary [15]. MRI staging has been shown to have an accuracy of 40% with very low sensitivity and specificity [16, 17]. An alternative to CT or MRI is PET. PET is a noninvasive test which has shown to be beneficial in detection of metastatic disease; detection of locoregional metastases is limited, however, [18]. Because of limitations of CT, MRI, and PET and advances in technology, other modalities, for example EUS, were initiated and reviewed.

EUS utilizes an echoendoscope, which is passed directly into the esophagus, with the ability to visualize the individual histologic layers of the esophagus [19]. This approach is particularly useful in evaluating invasion of local disease, especially ECA. EUS has been shown to detect more locoregional node involvement than CT or PET, with sensitivity approaching 90% and specificity approaching 97% [13, 20, 21]. The accuracy of EUS to determine tumor depth has also been estimated to be quite accurate, approaching 80–92% in some reports [2123]. However, studies vary with regard to the accuracy of EUS in detecting both distant and CLN metastasis of ECA.

With EUS emerging as a very useful staging tool, its role in staging ECA continues to be re-addressed. Multiple studies have identified the potential benefits of EUS with ECA staging, especially identification of distant spread of the disease to the celiac lymph nodes; however, results regarding the extent of its benefits have been inconsistent. We conducted a meta-analysis to examine the role of EUS in the staging of ECA with distal and celiac lymph node metastases.

This meta-analysis and systematic review was written in accordance with the proposal for reporting by the QUOROM (Quality of Reporting of Meta-analyses) statement [24]. Because this manuscript looks at the diagnostic accuracy of a test, the study design for this meta-analysis and systematic review conformed to the guidelines of the Standards for Reporting of Diagnostic Accuracy (STARD) initiative [25].

Methods

Study selection criteria

Only EUS studies confirmed by surgery were selected. Distal metastasis was defined as metastasis to peritoneum, liver, cervical lymph nodes, celiac axis lymph nodes, or abdominal lymph nodes. EUS criteria used for nodal invasion were: larger than 1 cm, hypoechoic, and round instead of elliptical. Only studies from which a 2 × 2 table could be constructed for true positive, false negative, false positive and true negative values were included.

Data collection and extraction

Articles were sought in Medline, Pubmed, Ovid journals, Cumulative Index for Nursing and Allied Health Literature, ACP journal club, DARE, International Pharmaceutical Abstracts, old Medline, Medline nonindexed citations, OVID Healthstar, and Cochrane Controlled Trials Registry. The search terms used were endoscopic ultrasound, EUS, ultrasound, endosonography, esophageal cancer, oesophageal cancer, celiac axis lymph nodes, metastasis, tumor staging, nodal invasion, staging, surgery, sensitivity, specificity, positive predictive value, and negative predictive value. The data extracted from each study were used to construct 2 × 2 tables. Two authors (SP and JR) independently searched and extracted the data into an abstraction form. Any differences were resolved by mutual agreement.

Quality of studies

A clinical trial with a control arm can be assessed for the quality of the study. A number of criteria have been used to assess this quality of a study (e.g. randomization, selection bias of the arms in the study, concealment of allocation, and blinding of outcome) [26, 27]. There is no consensus on how to assess studies without a control arm. Hence, these criteria do not apply to studies without a control arm [27]. Therefore, for this meta-analysis and systematic review, studies were selected based on completeness of data and inclusion criteria.

Statistical methods

Meta-analysis for the accuracy of EUS in diagnosing distal and CLN lymphadenopathy was performed by calculating pooled estimates of sensitivity, specificity, likelihood ratios, and diagnostic odds ratios. EUS studies were grouped into time periods to standardize the change in EUS technology and EUS criteria for lymph node involvement [28]. These periods were 1986–1995, 1996–1999, and 2000–2006. Pooling was conducted using both the Mantel–Haenszel method (fixed-effects model) and the DerSimonian Laird method (random-effects model). The confidence intervals were calculated using the F distribution method [29]. Forrest plots were drawn to show the point estimates in each study in relation to the summary pooled estimate. The width of the point estimates in the Forrest plots indicates the weight assigned to that study. For 0 value cells, 0.5 was added as described by Cox [30]. The heterogeneity of the sensitivities and specificities were tested by applying the likelihood ratio test [31]. The heterogeneity of likelihood ratios and diagnostic odds ratios were tested using Cochran’s Q test based upon inverse variance weights [32]. Heterogeneity among studies was also tested by using summary receiver operating characteristic (SROC) curves. SROC curves were used to calculate the area under the curve (AUC). The effect of publication and selection bias on the summary estimates was tested by use of the Egger bias indicator [33] and the Begg–Mazumdar bias indicator [34]. Also, funnel plots were constructed to evaluate potential publication bias using the standard error and diagnostic odds ratio [35, 36].

Results

The initial search identified 4,013 reference articles; of these, 436 relevant articles were selected and reviewed. Twenty-five studies (N = 2029), which met the inclusion criteria were included in this analysis [3761]. For distal metastasis there were 25 studies [3761]. There were 23 studies for CLN metastasis [3759]. Figure 1 shows the search results and Table 1 shows the characteristics of EUS studies included in this meta-analysis. All 25 studies included in this analysis were published as full-text articles in peer-review journals. The pooled estimates given are estimates calculated by the fixed-effect model.
https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig1_HTML.gif
Fig. 1

Search results

Table 1

Table showing the characteristics of studies included in this analysis

 

Author

Year of publication

Type of enrolment

Confirmatory test

1

Catalano et al.

1999

Prospective

Surgery or FNA

2

Binmoeller et al.

1995

Prospective

Surgery

3

Natsugoe et al.

1996

Consecutive

Surgery

4

Giovannini et al.

1999

Prospective

Surgery or FNA

5

Grimm et al.

1993

Prospective

Surgery

6

Salminen et al.

1999

Consecutive

Surgery

7

Vikers et al.

1998

Consecutive

Surgery

8

Williams et al.

1999

Consecutive

Surgery

9

Tio et al.

1989

Prospective

Surgery

10

Tio et al.

1989

Prospective

Surgery

11

Tio et al.

1990

Consecutive

Surgery

12

Tio et al.

1989

Prospective

Surgery

13

Eloudeibi et al.

2001

Consecutive

Surgery

14

Tio et al.

1986

Prospective

Surgery

15

Reed et al.

1998

Consecutive

Surgery

16

Wallace et al.

2000

Consecutive

Surgery

17

DeWitt et al.

2005

Prospective

Surgery

18

Browney et al.

1999

Prospective

Surgery

19

Eloudeibi et al.

2001

Consecutive

Surgery

20

Krasna et al.

1999

Consecutive

Surgery

21

Vazquez-Sequeiros et al.

2001

Consecutive

Surgery

22

Kallimanis et al.

1995

Consecutive

Surgery

23

Shimizu et al.

1997

Consecutive

Surgery

24

Nishimaki et al.

1999

Consecutive

Surgery

25

Fok et al.

1992

Consecutive

Surgery

Accuracy of EUS in the diagnosis of CLN metastasis

Pooled sensitivity of EUS in diagnosing CLN involvement by ECA was 66.6% (95% CI: 61.9–71.1). EUS had a pooled specificity of 98.1% (95% CI: 97.3–98.7). The Forrest plot in Fig. 2 shows the sensitivity and specificity of EUS in the diagnosis of CLN metastasis. The positive likelihood ratio of EUS was 20.4 (95% CI: 10.6–39.1) and the negative likelihood ratio was 0.3 (95% CI: 0.2–0.5). Figure 3 shows the Forrest plot of the positive and negative likelihood ratio of EUS in the diagnosis of CLN metastasis. The diagnostic odds ratio, the odds of having CLN metastasis in positive as compared with negative EUS studies was 74.6 (95% CI: 36.6–151.9). All the pooled estimates calculated by fixed and random-effect models were similar. SROC curves showed an area under the curve of 0.91. Figure 4 shows the SROC curves for EUS in the diagnosis of CLN metastasis. The P for chi-squared heterogeneity for all the pooled accuracy estimates was >0.10.
https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig2_HTML.gif
Fig. 2

Forrest plot showing the sensitivity and specificity of EUS in the diagnosis of CLN metastasis

https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig3_HTML.gif
Fig. 3

Forrest plot showing the positive and negative likelihood ratio of EUS in the diagnosis of CLN metastasis

https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig4_HTML.gif
Fig. 4

SROC curves for EUS in the diagnosis of CLN metastasis

Accuracy of EUS in the diagnosis of distal metastasis

Pooled sensitivity of EUS in diagnosing distal metastasis by ECA was 67.2% (95% CI: 62.6–71.6). EUS had a pooled specificity of 97.9% (95% CI: 97.1–98.6). Figure 5 depicts the Forrest plots of sensitivity and specificity for EUS in the diagnosis of distal metastasis. The positive likelihood ratio of EUS was 18.5 (95% CI: 10.1–33.9) and the negative likelihood ratio was 0.3 (95% CI: 0.2–0.5). The Forrest plots in Fig. 6 show the positive and negative likelihood ratio of EUS in the diagnosis of distal metastasis. The diagnostic odds ratio, the odds of having distal metastasis in positive as compared with negative EUS studies was 68.9 (95% CI: 35.4–134.1). All the pooled estimates calculated by fixed and random-effect models were similar. SROC curves showed an area under the curve of 0.91. Figure 7 shows SROC curves of EUS in the diagnosis of distal metastasis. The P for chi-squared heterogeneity for all the pooled accuracy estimates was >0.10.
https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig5_HTML.gif
Fig. 5

Forrest plot showing the sensitivity and specificity of EUS in the diagnosis of distal metastasis

https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig6_HTML.gif
Fig. 6

Forrest plot showing the positive and negative likelihood ratio of EUS in the diagnosis of distal metastasis

https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig7_HTML.gif
Fig. 7

SROC curves for EUS in the diagnosis of distal metastasis

Effect of technology

EUS studies were grouped into three time periods to standardize the change in EUS technology and to standardize the change in EUS criteria for tumor staging [28]. These periods of time were 1986–1994, 1995–1999, and 2000–2006. For CLN metastasis, there were eight studies in the period 1989–1995. During 1996–1999 ten studies met the inclusion criteria. Five studies met the inclusion criteria between 2000 and 2006. All the pooled estimates during the three periods of time are given in Table 2. The P for chi-squared heterogeneity for all the pooled accuracy estimates was >0.10.
Table 2

Pooled diagnostic accuracy estimates of EUS in the diagnosis of CLN metastasis for different periods of time with 95% confidence intervals

Time period

No. of studies

Pooled sensitivity

Pooled specificity

Pooled LR+

Pooled LR−

Pooled DOR

1989–1995

8

57.1% (47.4–66.5)

97.5% (95.1–98.9)

14.2 (6.1–33.0)

0.4 (0.3–0.6)

46.0 (14.9–142.0)

1996–1999

10

77.2% (70.4–83.0)

98.3% (97.3–99.1)

22.2 (6.3–78.5)

0.2 (0.1–0.6)

103.5 (27.3–392.1)

2000–2006

5

59.5% (50.4–68.2)

98.1% (96.0–99.3)

23.4 (11.7–46.6)

0.3 (0.1–97.8)

85.7 (36.3–202.4)

LR+: positive likelihood ratio; LR−: negative likelihood ratio; DOR: diagnostic odds ratio

To diagnose distal metastasis, during the periods 1986–1994 and 1995–1999 there were seven and thirteen studies, respectively. Four studies met the inclusion criteria for the period between 2000 and 2006. The P for chi-squared heterogeneity for all the pooled accuracy estimates was >0.10. The pooled estimates of studies during these periods are shown in Table 3.
Table 3

Pooled diagnostic accuracy estimates of EUS in the diagnosis of distal metastasis in different periods, with 95% confidence intervals

Time period

No. of studies

Pooled sensitivity

Pooled specificity

Pooled LR+

Pooled LR−

Pooled DOR

1986–1994

7

59.1% (49.6–68.2)

98.3% (96.0–99.4)

20.2 (7.1–57.9)

0.4 (0.2–0.6)

67.8 (17.9–257.7)

1995–1999

13

76.6% (70.1–82.3)

97.7% (96.6–98.6)

15.2 (6.1–37.5)

0.3 (0.1–0.6)

67.8 (23.5–195.7)

2000–2006

4

77.2% (66.4–85.9)

98.0% (95.7–99.3)

24.3 (11.9–49.5)

0.3 (0.2–0.4)

99.6 (40.5–245.1)

LR+: positive likelihood ratio; LR−: negative likelihood ratio; DOR: diagnostic odds ratio

Bias estimates

The publication bias calculated by use of the Begg–Mazumdar bias indicator gave a Kendall’s tau b value of −0.2, P = 0.21; use of the Egger bias indicator gave a value of −0.56 (95% CI = −2.28–1.16, P = 0.50). The funnel plots in Figs 8 and 9 show there was no effect of publication bias on the pooled estimates calculated for distal or CLN metastasis.
https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig8_HTML.gif
Fig. 8

Funnel plot looking at publication bias in studies included for CLN metastasis

https://static-content.springer.com/image/art%3A10.1007%2Fs10620-007-0152-3/MediaObjects/10620_2007_152_Fig9_HTML.gif
Fig. 9

Funnel plot looking at publication bias in studies included for distal metastasis

A subgroup analysis was performed removing the studies in which the last or the first author was the same (e.g. Tio et al.). This was done to make sure the same data were not used by the studies i.e. to avoid duplication. In the subgroup analysis, there was no significant change in the pooled estimates. Separate accuracy estimated for radial vs. linear EUS technology could not be performed, because most of the studies did not make a distinction and did not give separate accuracy values for radial or linear EUS technology.

Discussion

The major advantage of EUS is the ability to perform FNA during the procedure for tissue diagnosis. The procedure is, in comparison with alternative options, safe, less invasive, and does not require general anesthesia or hospitalization [62]. The complication rate is extremely low (0.5–2.3%) with several studies reporting no complications [40, 6365].

This meta-analysis and systematic review shows that EUS has very high specificity (∼99%) in the diagnosis of distal and celiac axis lymph node metastasis of ECA. The sensitivity of EUS in the diagnosis of distant and celiac axis metastasis is not high (∼65–67%). The diagnostic odds ratio is defined as the odds of having a positive test in patients with the true anatomic stage of the disease compared with patients who do not have the disease. In the diagnosis of celiac and distal metastasis, EUS as a diagnostic test has a very high diagnostic odds ratio (∼70). For example, if EUS indicates a patient with ECA has distal or celiac metastasis, the odds of correct diagnosis of the anatomic stage of the disease are 70:1. Another way of looking at this is that EUS is an excellent test for confirming distal or CLN metastasis but does not perform as well at excluding the same. The affect of FNA on the accuracy estimates could not be evaluated as there were not enough studies with accuracy estimated for EUS with FNA.

The positive likelihood ratio of a test is a gauge of how well the test identifies a disease state. The higher the positive likelihood ratio, the better the test performs in identifying the true disease status. A negative likelihood ratio of a test is, on the other hand, a gauge of how well the test performs in excluding a disease state. The lower the negative likelihood ratio, the better the test performs in excluding a disease. For distal or CLN metastasis, EUS has a high positive likelihood ratio and a low negative likelihood ratio. This indicates that EUS performs better at diagnosing than excluding distal or CLN metastasis.

To evaluate the affect of technology, studies were grouped into periods of time. The presumption is that during a period of time, the technology of EUS used might be the same. One of the weaknesses of doing this kind of pooling is that some of the studies might use older equipment even though the paper was published in the most recent time period. But this seems to be an accepted method of looking at affect of technology [28] and no alternative way of looking at this effect exists. Over the last two decades, the specificity of EUS in the diagnosis of distal and CLN metastasis has remained high. In addition, the sensitivity of EUS in the diagnosis of distal metastasis has improved, possible because of improvement of imaging technology or training. In the past two decades the sensitivity of EUS in the diagnosis of CLN metastasis has improved slightly. For CLN metastasis, the increase and then the decrease in sensitivity might be because of the smaller number of studies during the most recent period.

Heterogeneity among different studies was determined by drawing SROC curves and finding the AUC, because different studies might use slightly different criteria for staging. An AUC of 1 for any test indicates the test is excellent. SROC curves for EUS showed that the AUC was very close to 1, indicating that EUS is an excellent diagnostic test for staging esophageal cancers.

Studies with statistically significant results tend to be published and cited. Smaller studies may show larger treatment effects because of fewer case-mix differences (e.g. patients with only early or late disease) than larger trials. This bias can be estimated by use of bias indicators and by construction of funnel plots. This publication and selection bias may affect the summary estimates. Also, bias among studies can affect the shape of the funnel plot. In this meta-analysis and systematic review, bias calculations using the Egger bias indicator [33] and the Begg–Mazumdar bias indicator [34] revealed no statistically significant bias. Furthermore, analysis using funnel plots showed no significant publication bias among the studies included in this analysis.

Conclusions

EUS has excellent specificity in the accurate diagnosis of distal metastasis and CLN metastasis in a patient with ECA. In the diagnosis the sensitivity of EUS is not high. The sensitivity of EUS in the diagnosis of distal metastasis has improved over the last two decades. It is not clear if FNA improves the sensitivity and specificity of EUS in evaluating CLN metastasis of ECA. Further improvements in EUS should be strongly considered, because staging technology is needed to improve sensitivity in the diagnosis of distal and CLN metastasis of ECA.

Copyright information

© Springer Science+Business Media, LLC 2007