Efficacy, Feasibility, and Safety of Endoscopic Ultrasound-guided Fine-needle Biopsy for the Diagnosis of Gastrointestinal Subepithelial Lesions

A Systematic Review and Meta-analysis

Abstract

Background: 

Endoscopic ultrasound (EUS) fine-needle biopsy (FNB) has become an efficient method for diagnosing gastrointestinal (GI) subepithelial lesions (SELs). However, recent guidelines have not regarded FNB as the primary strategy for diagnosing GI SELs. We performed this study to systematically measure the efficacy, feasibility, and safety of EUS-FNB in diagnosing GI SELs.

Materials and Methods: 

Relevant studies were searched in PubMed and EMBASE and published after January 2015 were included. The overall rates of diagnostic yield, technical success, and adverse events were calculated as outcome measures. The Jadad scale and the Newcastle-Ottawa scale were used to evaluate the quality of the trials, funnel plots and Egger’s test were used to measure publication bias, and sensitivity and subgroup analyses were performed to explore the variance of heterogeneity and sensitivity, respectively.

Results: 

Sixteen studies analyzing 969 patients between 2015 and 2020 were included. Studies showed little change in sensitivity, and 13 were considered high quality. A certain degree of publication bias existed in the diagnostic accuracy rate. The overall rates of diagnostic yield, technical success, and adverse events were [85.69% (95% confidence interval (CI): 82.73-88.22, I²=41.8%), 98.83% (95% CI: 96.73-99.97, I²=54.3%), and 1.26% (95% CI: 0.35-2.54, I²=0.0%)]. No clinical influencing factors were identified in the subgroup analysis.

Conclusions: 

EUS-FNB is a promising technology with a relatively superior diagnostic yield, technical success, and security, which is an optimal option for the diagnosis of SELs.

Gastrointestinal (GI) subepithelial lesions (SELs) include various non-neoplastic and neoplastic lesions. The most common SELs are lipomas, followed by GI stromal tumors and leiomyomas. Minimally invasive tissue sampling is a better method to prevent unwarranted resections of truly benign SELs and delayed therapy for malignant SELs.¹ Minimally invasive methods include endoscopic ultrasound (EUS)-guided fine-needle aspiration (FNA) and fine-needle biopsy (FNB).

FNA was first reported in 1992² and has become the primary method for acquiring tissue from pancreatic lesions.³ However, the ability of this approach to diagnose SELs has been challenged because of some limitations. First, the diagnostic accuracy rates range from 34% to 91%.^4,5 The rate would even be decreased without an on-site cytopathologist.⁶ Furthermore, the ability to obtain sufficient tissue is limited,⁷ and tissue cores were obtained in only approximately half of patients.⁵ However, histologic core tissues are required to correctly diagnose SELs through analyses of the histologic architecture and immunohistochemistry (IHC).^8,9 Based only on cytologic features, a definitive diagnosis of SELs,¹⁰ such as gastrointestinal stromal tumors (GIST), schwannoma, and leiomyoma, is difficult to achieve. Furthermore, histologic tissue is also becoming important for molecular analysis in the current personalized medicine era.¹¹ In contrast, FNB appears to be the best option for facilitating a superior definitive SEL diagnosis.¹² FNB has refined the needle that features a beveled side hole or cutting-tip geometry. Thus, this needle can obtain both cytologic aspirate and histologic samples. Therefore, FNB might provide more specific information on malignancy, including metastatic origin, differentiation degree, and proliferation rate.¹¹

However, recent guidelines have not chosen FNB as the optimal SEL diagnostic strategy,¹³ potentially because of the paucity of evidence for the use of FNB needles for SEL sampling. Although some meta-analyses of FNB have been published, most did not exclusively focus on SELs and included multiple lesions, such as pancreatic masses and lymph nodes. Many included studies were not performed within 5 years, which may lead to inaccuracy because of the development of novel FNB needles and technical procedures. Therefore, an updated meta-analysis was performed to systemically summarize current studies on the efficacy, feasibility, and safety of FNB for diagnosing GI SELs.

MATERIALS AND METHODS

Search Strategy

Articles were searched in PubMed and EMBASE using the following keywords: FNB, endoscopy, ultrasound, and GI SELs. In addition, the references of articles and the related article functions were also used to search for eligible articles.

Eligibility Criteria

The eligibility criteria included the following: (1) studies published after January 2015 and written in English; (2) SELs were the only lesion or contained other lesions, but cases of SELs were able to be separated; and (3) studies reported the following outcome measures: diagnostic yield, technical success, and adverse events.

The exclusion criteria included the following: (1) reviews, letters, case reports, and comments; (2) duplicate studies; and (3) studies with insufficient data.

Quality Assessment

Publication bias was analyzed using a visual inspection with Egger’s test and funnel plots.¹⁴ The quality of the retrospective trials was measured with the Newcastle-Ottawa scale, and the quality of prospective studies was measured with the Jadad scale. Studies that scored >6 points on the Newcastle-Ottawa scale and 4 points on the Jadad scale were considered high-quality trials. Study quality was evaluated by 2 authors (Y.T. and X.T.), and disagreements were resolved by discussion with a senior reviewer (R.L.).

Data Extraction

Two authors (Y.T. and X.T.) independently collected the information, including author name, publication year, region, design and centers of studies; number, mean age, ratio of male to female patients; number, mean size, location and depth of lesions; whether or not only SELs and stomach lesions were evaluated; types and passes of needles; and the rates of diagnostic yield, technical success, and adverse events.

Outcome Measures

The outcome measures include: (1) diagnostic yield, which is defined as the ability to make a definitive final diagnosis; (2) technical success, which is defined as the ability to obtain sufficient tissue and cytologic samples; and (3) adverse events, which are defined as any deviation from the clinical course after FNB procedure, such as bleeding, abdominal pain, and fever. And bleeding was excessive and postprocedural bleeding at the site of puncture or perforation.

Statistical Analyses

Outcome measures are reported as weighted proportions and 95% confidence interval (CI). Heterogeneity was assessed using the I² and χ² tests. The random-effects model was used when significant heterogeneity was present (I²≥50%, P>0.5); otherwise, the fixed-effects model was used. Furthermore, sensitivity analyses were performed to measure the effect of each study on outcome measures, and subgroup analyses were conducted to explore the variance of heterogeneity. Egger’s test and a funnel plot were used to assess publication bias. All analyses were conducted using R Statistics 4.0.3 (Shanghai, China), and a P-value <0.05 was considered statistically significant.

RESULTS

Studies and Selection

A flow diagram of the literature search is shown in Figure 1. We retrieved 3472 studies from the initial search of PubMed and EMBASE. Following an initial screen of the titles, 2369 records were excluded. After the abstract review, 945 records were further excluded for a variety of reasons, and the remaining 158 articles underwent the full-text review. After the final application of the inclusion and exclusion criteria stated above, 16 studies published between 2015 and 2020 were selected for the final analysis.^1,15–29

FIGURE 1: Flow chart of the study screening process.

Study Characteristics

The detailed characteristics of the studies are shown in Table 1. Of the 16 papers, 8 (50.0%) studies were prospective studies and 8 (50.0%) were retrospective trials. Six (37.5%) were conducted in multiple centers and 10 (62.5%) were conducted in single centers. Four (25%) were conducted in the United States, 3 (18.8%) were conducted in Europe and 9 (56.3%) were conducted in Asia. Fifteen (93.75%) evaluated SELs only. Of the 969 patients, 487 (50.26%) were male, 482 (49.74%) were female, and the mean age was 61 years. Of the 780 lesions, 640 (82.05%) were located in the stomach, and the mean size was 25.4 mm. The EUS-FNB procedure was performed with 19, 20, 21, 22, and 25 G needles [ProCore (Cook Medical), Acquire (Boston Scientific), or SharkCore (Medtronic)], and the mean number of passes was 2.29.

TABLE 1 - Main Characteristics of Included Studies

References	Region	Study Design	Centers	No. of Patients	Mean Age (y)	Male/Female	No. of Lesion	Mean Lesion Size (mm)	Lesion Depth	Stomach Lesions Only	Evaluate SELs Only	Mean Needle Passes (n)	Type Needle	Adverse Event
Antonini et al27	Europe	Retrospective	Multiple-center	50	61.5	22/28	70	43.1	3, 4	No	Yes	2.2	20	0
Han et al15	Asia	Prospective	Single-center	22	59.5	10/12	22	25.5	2, 4	Yes	Yes	1.0	22	0
Lee et al16	Asia	Retrospective	Single-center	78	53.5	38/40	77	28.0	4	Yes	Yes	3.0	22	1 bleeding
Hedenstrom et al1	Europe	Prospective	Single-center	83	68	42/41	89	30.0	Unknown	No	Yes	2.0	19, 22	1 bleeding
El et al28	USA	Retrospective	Single-center	15	65	9/6	10	25.5	Unknown	No	Yes	1.0	19, 22, 25	0
Lee et al17	Asia	Retrospective	Single	43	49	21/22	43	25.0	3, 4	Yes	Yes	3.0	22	0
Schlag et al18	Europe	Prospective	Single-center	20	58.4	11/9	20	16.0	3, 4	No	Yes	2.6	22	0
Inoue et al29	Asia	Retrospective	Multiple-center	57	66	30/27	57	20.0	Unknown	No	Yes	2.0	19, 22, 25	2 bleeding
DiMaio et al19	USA	Retrospective	Multiple-center	226	66	113/113	28	29.0	4	No	No	2.0	22, 25	0
Iwai et al20	Asia	Prospective	Single-center	23	67.6	8/15	23	24.0	4	Yes	Yes	2.0	22, 19	0
Kim et al21	Asia	Prospective	Multiple-center	36	62.5	19/17	36	25.0	3, 4	No	Yes	1.5	20	2 bleeding
Kim et al22	Asia	Retrospective	Single-center	107	46.8	55/52	107	26.0	2, 3, 4	Yes	Yes	Unknown	Unknown	1 fever and 2 bleeding
Park et al23	Asia	Prospective	Single-center	39	58.3	18/21	28	21.0	3, 4	No	Yes	2.2	22, 20	2 bleeding
de Moura et al24	USA	Retrospective	Multiple-center	114	62.07	57/57	114	21.2	2, 3, 4	No	Yes	2.9	19, 20, 21, 22, 25	0
Sanaei et al25	USA	Prospective	Multiple-center	26	68.7	15/11	26	23.5	2, 3, 4	No	Yes	4.0	22	2 abdominal pain
Okuwaki et al26	Asia	Prospective	Single-center	30	67	19/11	30	24.0	Unknown	No	Yes	3.0	22.0	2 bleeding

SEL indicates subepithelial lesion.

Quality Assessment

Newcastle-Ottawa scores and Jadad scores were used to measure the quality of the 8 retrospective and 8 prospective studies, respectively. As reported in Table 2, 13 studies were considered high quality, and 3 studies were considered low quality.

TABLE 2 - Quality Assessment

Newcastle-Ottowa Scores
References	Selection	Comparability	Outcome		Scores
Antonini et al27	***	*	**		6
Lee et al16	**	*	**		5
El et al28	***	**	**		7
Lee et al17	**	*	**		5
Inoue et al29	***	*	***		7
DiMaio et al19	**	*	*		4
Kim et al22	**	**	**		6
de Moura et al24	***	*	**		6
Jadad scores
Study	Randomization	Allocation Concealment	Double Blind	Dropouts	Score
Han et al15	1	U	1	1	3
Hedenstrom et al1	1	1	1	1	4
Schlag et al18	1	1	1	1	4
Iwai et al20	1	2	1	1	5
Kim et al21	1	U	2	1	4
Park et al23	1	1	2	1	5
Sanaei et al25	2	2	1	1	4
Okuwaki et al26	1	1	2	1	5

U indicates unknown.

Overall Meta-analysis

Fifteen studies with 969 patients were included to analyze the diagnostic yields, technical outcomes and safety of FNB for diagnosing GI SELs. The overall aggregated rates for the diagnostic yield was 85.69% (95% CI: 82.73-88.22, I²=41.8%), the technical success was 98.83% (95% CI: 96.73-99.97, I²=54.3%), and the adverse events was 1.26% (95% CI: 0.35-2.54, I²=0.0%), which, respectively, are outlined in Figure 2.

FIGURE 2: Forest plots for the overall aggregated rates. A, Diagnostic yield. B, Technical success. C, Adverse events. CI indicates confidence interval.

Heterogeneity Among Studies

For the rates of diagnostic yield (I²=42%; P=0.51) and adverse events (I²=0%; P=0.63), significant heterogeneity was not observed between the studies. However, studies reporting the technical success rate displayed heterogeneity (I²=54%; P<0.01), and we further investigated potential sources by performing a subgroup analysis. We preidentified several potential sources of heterogeneity for the diagnostic yield, technical success and the adverse event rates including study design, sample size, geographical location, needle size, lesion size, stomach lesions only, lesion depth, and number of passes.

The diagnostic yield rates of studies with sample sizes of <30 was significantly higher than studies with sample sizes of more than 30 [0.92 (95% CI: 086-0.95), N=8, I²=45.40%, P=0.01 vs. 0.84 (95% CI: 0.80-0.87), N=8, I²=0). The other subgroup analyses showed little significant difference. However, the rate was numerically higher in the United States than in Europe and Asia [0.92 (95% CI: 0.82-0.96), N=4, I²=8.10%; 0.84 (95% CI: 0.76-0.89), N=3, I²=0; and 0.85 (95% CI 0.82-0.88), N=9, I²=0, P=0.59]. The highest rate was observed for lesions located at depths of 2, 3, and 4 layers, followed by those located at a depth of 4 and at depths of 3, 4/3, 4, and 5 layers [0.88 (95% CI: 0.84-0.91), N=8, I²=50.70%; 0.84 (95% CI: 0.76-0.89), N=3, I²=0; and 0.82 (95% CI: 0.76-0.87), N=5, I²=43.60%, respectively, P=0.18]. Two to 3 mean passes tended to decrease the diagnostic yield rate compared with more than 3 passes and <2 passes [0.82 (95% CI: 0.77-0.86), N=8, I²=5.90%; 0.88 (95% CI: 0.82-0.92), N=4, I²=88.00%; and 0.88 (95% CI: 0.83-0.92), N=4, I²=0, P=0.12]. The use of 22 and 20 G needles produced a higher diagnostic yield rate than the use of 19 G mixed and mixed/unknown needles [0.88 (95% CI: 0.79-0.94), N=2, I²=0; 0.88 (95% CI: 0.83-0.91), N=6, I²=79.50%; 0.85 (95% CI: 0.78-0.89), N=5, I²=0; and 0.81 (95% CI: 0.67-0.90), N=3, I²=56.0%, P=0.42] (Table 3).

TABLE 3 - Pooled Estimation of Diagnostic Yield and Subgroup

		Random-effect Model
Subgroup	Arms (N)	DYR	95% CI	I 2 (%)	P
Overall	16	0.8569	0.8273-0.8822	41.80
Study design
Retrospective study	8	0.85	0.8105-0.8825	0
Prospective study	8	0.8675	0.8194-0.9042	77.60	0.54
Geographical location
Europe	3	0.837	0.7649-0.8902	0	0.32
USA	4	0.9143	0.8221-0.9610	8.10
Asia	9	0.8538	0.8168-0.8843	59.00
Sample size
>30	8	0.8362	0.7997-0.8672	0	0.01
≤30	8	0.9152	0.8618-0.9491	45.40
Stomach lesions only
Yes	5	0.8608	0.8145-0.8970	0	0.81
No	11	0.8539	0.8133-0.8869	57.90
lesion size
>25 mm	7	0.8535	0.7972-0.8963	79.10	0.87
≥25 mm	9	0.8585	0.8223-0.8883	0
Lesion depth
2, 3, 4/UC	8	0.881	0.8418-0.9115	50.70	0.18
3, 4/3, 4, 5	5	0.8249	0.7617-0.8740	43.60
4	3	0.8362	0.7574-0.8930	0
Needle size
20	2	0.8837	0.7972-0.9363	0	0.42
22	6	0.8761	0.8254-0.9137	79.50
19 mixed	5	0.8457	0.7844-0.8920	0
Mixed/unknown	3	0.8102	0.6708-0.8994	56.00
Passes (n)
<2	4	0.8833	0.8277-0.9227	0	0.12
≥2, <3	8	0.8242	0.7744-0.8649	5.90
≥3	4	0.8807	0.8239-0.9209	0.88

CI indicates confidence interval; DYR, diagnostic yield rate.

The technical success rates of studies analyzing only stomach lesions were significantly higher than studies analyzing lesions located in the esophagus, stomach, duodenum, rectum and colon [1.00 (95% CI: 0.99-1.00), N=5, I²=56.20% vs. 0.98 (95% CI: 0.94-1.00), N=11, I²=0, P=0.03]. The other subgroup analyses did not show significant differences. However, the rate for studies with sample sizes of <30 was slightly higher than studies with sample sizes of more than 30 [1.00 (95% CI: 0.98-1.00), N=8, I²=0, P=0.33 vs. 0.98 (95% CI: 0.93-1.00), N=8, I²=76.80%). Lesions with a size >25 mm tended to improve the technical success rate compared with lesions <25 mm in size [0.97 (95% CI: 0.91-1.00), N=7, I²=61.10% vs. 1.00 (95% CI: 0.99-1.00), N=9, I²=0, P=0.07] (Table 4). The subgroup analyses of the adverse events did not show significant difference (Table 5).

TABLE 4 - Pooled Estimation of Technical Success Rate and Subgroup

		Random-effect Model
Subgroup	Arms (N)	TSR	95% CI	I 2 (%)	P
Overall	16	0.9883	0.9673-0.9997	54.30
Study design
Retrospective study	8	0.9930	0.9595-1.0000	69.70	0.44
Prospective study	8	0.9808	0.9535-0.9978	13.50
Geographical location
Europe	3	0.9971	0.9732-1.0000	0	0.61
USA	4	0.9801	0.9393-1.0000	73.10
Asia	9	0.9957	0.9569-1.0000	0
Sample size
>30	8	0.9750	0.9325-0.9987	76.80	0.33
≤30	8	0.9991	0.9806-1.0000	0
Stomach lesions only
Yes	5	0.9999	0.9897-1.0000	56.20	0.03
No	11	0.9779	0.9430-0.9985	0
lesion size
>25 mm	7	0.9689	0.9102-0.9996	61.10	0.07
≥25 mm	9	0.9979	0.9875-1.0000	0
Lesion depth
2, 3, 4/UC	8	0.9871	0.9448-1.0000	68.40	0.71
3, 4/3, 4, 5	5	0.9828	0.9331-1.0000	58.90
4	3	0.9965	0.9706-1.0000	0
Needle size
20	2	0.9843	0.8924-1.0000	67.80	0.81
22	6	0.9966	0.9782-1.0000	0
19 mix	5	0.9777	0.9047-1.0000	68.10
Mixed/unknown	3	0.9855	0.8972-1.0000	80.10
Passes (n)
<2	4	0.9973	0.9631-1.0000	40.00	0.63
≥2, <3	8	0.9800	0.9334-1.0000	67.00
≥3	4	0.9940	0.9713-1.0000	0

CI indicates confidence interval; TSR, technical success rate.

TABLE 5 - Pooled Estimation of Adverse Event Rate and Subgroup

		Random-effect Model
Subgroup	Arms (N)	AER	95% CI	I 2	P
Overall	16
Study design
Retrospective study	8	0.0055	0.0000-0.0195	0	0.31
Prospective study	8	0.0153	0.0016-0.0376	0
Geographical location
Europe	3	0.0022	0.0000-0.0229	0	0.28
USA	4	0.0000	0.0053-0.0357	0
Asia	9	0.0180		0
Sample size			0.0046-0.0303
>30	8	0.0153		0	0.69
≤30	8	0.0028	0.0000-0.0242	0
Stomach lesions only
Yes	5	0.0058	0.0000-0.0231	0	0.36
No	11	0.0122	0.0015-0.0294	0
lesion size
>25 mm	7	0.0187	0.0016-0.0472	0	0.18
25≥ mm	9	0.0063	0.0000-0.0192	0
Lesion depth
2, 3, 4/UC	8	0.0106	0.0006-0.0284	0	0.77
3, 4/3, 4, 5	5	0.0112	0.0000-0.0437	31.10
4	3	0.0035	0.0000-0.0294	0
Needle size
20	2	0.0157	0.0000-0.1076	67.80	0.95
22	6	0.0055	0.0000-0.0258	0
19 mix	5	0.0068	0.0000-0.0303	0
Mixed/unknown	3	0.0170	0.0007-0.0461	0
Passes (n)
<2	4	0.0127	0.0000-0.0407	0	0.87
≥2, <3	8	0.0074	0.0000-0.0246	0
≥3	4	0.0094	0.0000-0.0378	15.30

AER indicates adverse event rate; CI, confidence interval.

Publication Bias and Sensitivity Analyses

Funnel plots of the technical success rate and adverse event rate showed good symmetry and were confirmed by the Egger test (P=0.78 and 0.95, respectively), all suggesting minimal publication bias. Regarding the diagnostic accuracy rate, the funnel plot of the diagnostic yield rate was slightly asymmetrical and confirmed by Egger’s test (P=0.02), all suggesting a certain degree of publication bias (Fig. 3). The forest plot showed little change in the sensitivity of the technical success rate after systematically removing each study (Fig. 4).

FIGURE 3: Funnel plot and Egger’s test showing publication bias. Funnel plot: A, Diagnostic yield. B, Technical success. C, Adverse events. Egger’s test: D, Diagnostic yield. E, Technical success. F, Adverse events.

FIGURE 4: Forest plot indicating results of sensitive analysis for the technical success rate. CI indicates confidence interval.

Complications

Among the 16 studies included in this review, no serious complications were reported. Twelve (75%) cases of bleeding were reported in 7 studies, 2 (12.5%) cases of abdominal pain were reported in 1 study, and 1 (6.3%) case of fever was reported in 1 study. Regarding bleeding treatments, 2 (16.7%) patients with bleeding improved with conservative management, 8 (66.7%) were completely managed endoscopically with hemoclips or an epinephrine injection, and 2 (16.7%) had an unclear outcome. Patients with fever were treated with an intravenous infusion of antibiotics, and those with abdominal pain were managed conservatively. All results supported the safety of EUS-FNB procedures.

DISCUSSION

When comparing the efficacy between FNA and FNB, FNB outperformed FNA in all outcomes.^28,30 When conclusive diagnoses of GISTs cannot be achieved, resection would be performed, and FNB appeared to be more reliable in excluding a suspected neoplastic lesion.²⁹ Thus, FNB also resulted in a significantly lower unwarranted resection rate than FNA. FNB without rapid on-site evaluation (ROSE) showed a similar diagnostic rate to FNA with ROSE.¹¹ A meta-analysis revealed that the diagnostic yield of FNB was not impacted by the use of ROSE,³¹ and the algorithm also did not suggest the routine use of ROSE.³² Therefore, FNB may eliminate ROSE utilization, which may be beneficial in some regions where cytopathologists are not available because of the cost.²⁰ In these regions, the usefulness of visual assessment of the sample adequacy has been assessed. Junbum reported that a nonbloody white core and bloody core with white spots had a high positive predictive value as compared with that of a bloody core or gelatinous white material.³³ Iwashita et al³⁴ reported that the diagnostic yields were significantly higher for macroscopic visible core ≥4 mm than <4 mm. However, Iwai indicated that the diagnostic rate for a yellowish core and a bloody core did not differ significantly, but tissue that can be immunohistochemically stained is present in specimens that appear to be a bloody core.²⁰ As for the correlation between sonographic feature variables and the diagnostic yield of FNB for gastric SET, no significant difference was found, such as the presence of cystic foci, hyperechoic spots and border irregularity.¹⁶ However, some studies found that some sonographic features are independent predictive factors of GISTs among gastric SETs, such as irregular borders, mucosal ulceration, inhomogeneous echogenicity, presence of cystic foci, and nonoval shape.^22,35

Core tissue for IHC is necessary for differentiating GISTs from other SELs, and the tissue architecture is important to diagnose, especially when lesions are located in the muscularis propria layer. In the studies included in the present analysis, 94.4% to 96.8% of lesions were able to be obtained from core biopsy tissue.^16,21 IHC was accomplished in 69.3% to 100% of samples obtained using FNB compared with 40.0% to 73.9% of samples obtained using FNA.^{15,20,24,28,29} Compared with using FNB, the number of needle passes required to obtain adequate samples was significantly greater when using FNA,[20, 31] and the procedure time was also longer when using unroofing biopsy.²³ Furthermore, these advantages of FNB result in less risk to patients, as well as increased efficiency for endoscopy and cytopathology doctors.²⁸

Our study did not identify significant effects of different needle sizes on the rate, although the use of 22 and 20 G needles resulted in numerically higher rates than the use of 19 G mixed and mixed/unknown needles. However, a previous meta-analysis discovered that 22 G needles were a predictor related to a higher diagnostic yield than 19, 20, and 25 G needles.³⁶ This discrepancy may be attributed to the targeted objects, as other studies included all the lesions around the digestive tract, while our study only included SELs. One of our included studies showed that 22 G needles and 25 G needles both had relatively high diagnostic yields (89% and 86%, respectively). Another meta-analysis of GI SELs also indicated that different types and sizes of needles do not result in detectable disparity in the diagnostic yield rate. A potential explanation for this finding is that the smaller needle is technically demanding but can obtain more sufficient samples.³⁷

The management of lesions with a size <20 mm remains controversial. Major guidelines state that lesions ≥20 mm could be biopsied and excised.^38,39 However, The NCCN guideline indicate that the 20 mm cut off is arbitrary,³⁸ and Canadian guidelines report that even lesions <10 mm could be resected with the risk of metastasis.⁴⁰ Six studies with lesions smaller than 20 mm were included in our meta-analysis.^{1,15,18,20,23,29} One study showed that only 25% of lesions with a mean size of 16 mm were obtained from core biopsy specimens.¹⁸ Two studies showed that a lesion size of 20 mm or smaller was a positively associated factor affecting the diagnosis rate.^23,29 However, the number of lesions with a size smaller than 20 mm in our analysis was only 83 (13.20%), the diagnostic yield of which was 69.88%. However, 83 was too small to analyze, we chose a lesion size of 25 mm or smaller for the subgroup analysis, which showed no significant differences among all outcome measures. However, this result does not mean that lesion size is not an influencing factor. Further analyses should be performed to ascertain the ability of FNB to diagnose small SEL lesions.

The complications occurring among 969 patients only included bleeding, abdominal pain, and fever. Bleeding was the most common complication (75%) because SELs are often hypervascular and may be injured by the shape of the tip. Hemorrhage is usually treated with endoscopic hemostasis and conservation. The adverse event rate for FNB has not been reported to be higher than the rate for FNA.⁴¹ Therefore, FNB represents a technically safe procedure.

Our meta-analysis had some limitations. First, Egger’s test and the funnel plot indicated the existence of publication bias in our study. The source of the bias may be that we only included English studies, the results of which appeared more likely to be positive results. Second, the subgroup analysis showed that sample sizes had a significant effect on the diagnostic yield rate, and thus we were unable to perform an accurate analysis of trials with sample sizes of <30. Finally, some included studies may induce biases, such as single-center studies, retrospective studies, and clinical studies lacking a control group, and studies from various regions used heterogeneous definitions and standards. However, we only included studies published after January 2015. To our knowledge, the narrower the time frame is, the greater the significance of the results. Furthermore, our study focused on GI SELs and performed a precise analysis of the feasibility and effectiveness of FNB for diagnosing GI SELs. The diagnostic accuracy may be different between GI SELs and other lesions when diagnosed using FNB. Pancreatic adenocarcinomas are sufficiently diagnosed by a cytologic evaluation, while IHC evaluations are required for SELs. The heterogeneity in our study was also relatively low. However, larger-scale and better-designed studies are warranted to further assess the efficacy of FNB.

As for novel EUS scope, the forward-viewing EUS scope (FV-EUS) was used to overcome the lack of a forward view, which makes the puncture site readily visible, facilitating the puncture of lesions which are located at the fornix or are associated with strictures. Moreover, FV-EUS overcomes the difficulty in fixing the target, which expand the capabilities of the endoscope in small lesions.^42,43 Matsuzaki indicated that procedure time and tissue sample area using FV-EUS were superior than those using the oblique-viewing echoendoscope in GISTs.⁴³ Larghi also showed that FV-EUS has a high diagnostic accuracy and safety in GISTs, even in a <20 mm small lesions.⁴⁴

CONCLUSIONS

In conclusion, for GI SELs, EUS-FNB represents a superior safe, feasible, and effective tool for procuring adequate specimens and ensuring diagnostic accuracy and further facilitates proper treatment and avoids unwarranted resection.