Is endoscopic ultrasound examination necessary in the management of esophageal cancer?

Table 1 Baseline characteristics of meta-analyses on endoscopic ultrasound in esophageal carcinoma

Ref.	Timeframe	Patients (No. studies; P/R)	EUS types (MHz)	Study criteria
Puli et al[52], 2008	1986-2005	2020 (25; 10/15)	NR	EUS accuracy confirmed by surgery in distal and celiac axis lymph node metastasis
van Vliet et al[29], 2008	1985-2005	4713 (84; NA1)	NR	Comparison of diagnostic staging performance of EUS, CT and PET
Puli et al[32], 2008	1986-2005	2558 (49; 16/33)	NR	EUS studies on T and N staging confirmed by surgery
Thosani et al[30], 2012	1988-2008	1019 (19; 12/7)	Radial and/or mini-probe (7.5-30)	EUS in T1a vs T1b lesions compared to histology by EMR or surgery/excluded studies on < 15 patients, or with suspicious lymph nodes (> 1 cm)
Sun et al[76], 2015	1992-2013	724 (16; 10/6)	Radial, linear and/or mini-probe (5-20)	EUS staging accuracy after neoadjuvant chemotherapy. Surgery was confirmatory test in all included studies.
Qumseya et al[36], 2015	1994-2012	656 (11; 4/7)	Radial, linear and/or mini-probe (NR)	EUS in BE and HGD, or esophageal adenocarcinoma (EAC)/excluded studies on advanced esophageal cancer

¹Did not report retrospective or prospective nature of studies. References[29,30,32,36,52,76]. P/R: Prospective to retrospective ratio; NR: Not reported; BE: Barrett’s esophagus; HGD: High-grade dysplasia; EAC: Esophageal adenocarcinoma; EUS: Endoscopic ultrasonography; CT: Computed tomography; PET: Positron emission tomography; NA: Not applicable.

Table 2 Outcomes of meta-analyses on endoscopic ultrasound in esophageal carcinoma

Ref.	Sensitivity (95%CI)	Specificity (95%CI)	Heterogeneity	Conclusion/interpretation
Puli et al[52], 2008	Celiac N = 66% (62-71); M = 67% (63-72)	Celiac N = 98% (97-99); M = 98% (97-99)	Insignificant: P > 0.10 for all estimates	EUS has low sensitivity and utility for staging metastases to celiac lymph nodes and distant sites.
van Vliet et al[29], 2008	N staging: EUS = 80% (75-84); CT = 50% (41-60); PET 57% (43-70)	N staging: EUS = 70% (65-75); CT = 83% (77-89); PET = 85% (76-95)	NR	EUS, CT, and PET have distinctive roles in staging. For distant metastases, PET probably has higher sensitivity than CT. No evidence of publication bias in CT vs EUS studies; other analyses too small to test.
Puli et al[32], 2008	T1 = 82% (78-85); T4 = 92% (89-95); w/o FNA N = 85% (83-86); w/ FNA N = 97% (92-99)	T1 = 99.4% (99-100); T4 = 97% (97-98); w/o FNA N = 85% (83-86); w/ FNA N = 96% (91-98)	Insignificant: P > 0.10 for all estimates	EUS has excellent accuracy, with better performance in T4 over T1 disease (AUC 0.94-0.98). N staging is improved with FNA use (AUC 0.99 vs 0.89).
Thosani et al[30], 2012	T1a = 85% (82-88); T1b = 86% (82-89)	T1a = 87% (84-90); T1b = 86% (83-89)	Significant; P < 0.05 by χ2	EUS has good accuracy for T1a and T1b lesions; AUC ≥ 0.93. Technical factors can affect the diagnostic accuracy of EUS.
Sun et al[76], 2015	T1 = 23% (16-32); T2 = 29% (19-41); T3 = 81% (72-88); T4 = 43% (31-56); N = 69% (58-79)	T1 = 95% (93-97); T2 = 84% (77-88); T3 = 42% (33-52); T4 = 96% (94-97) N = 52% (42-62)	Significant; I2 = 0%-75% depending on stage (table presented in article)	EUS has modest accuracy after neoadjuvant therapy; AUC for T staging ranges from 0.64 to 0.84, while AUC for N-staging was 0.64.
Qumseya et al[36], 2015	≥ T1sm = 56% (47-65)	>/-T1sm = 89% (85-92)	Significant; I2 = 82%; Q = 56, P < 0.0001	Advanced disease detected in 14% (95%CI: 8%-22%; P < 0.0001). The NNT (performing EUS) to identify 1 case of advanced disease was 7 (95%CI: 5-13). EUS significantly changes therapeutic approach.

NR: Not reported; EUS: Endoscopic ultrasonography; CT: Computed tomography; PET: Positron emission tomography; AUC: Area under the curve.