Table 2.
Summary of recent meta-analysis of diagnostic EUS in pancreatic cancer.
| SRMA (Author, year) | Number of Studies and Total Number of Patients (n) | Modality/Comparison | Type of Lesion | Main Outcome Measure | Sensitivity (%) | Specificity (%) | PPV | NPV | Adequacy (%) | Accuracy (%) | Contamination (%) | Other Parameters (%) | Adverse Events (%) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Role of Diagnostic EUS | |||||||||||||
| Rahman MIO et al., 2020 [96] | 2 studies, n = 77 | EUS vs. CECT in pancreatic protocol | Neoplastic pancreatic lesions | Diagnostic accuracy for pancreatic cancer resectability | 87 | 63 | - | - | - | - | - | Similar diagnostic OR (p > 0.05) | - |
| Krishna SG et al., 2017 [14] | 4 studies | EUS after an indeterminate MDCT | Suspected pancreatic malignancies | Diagnostic performance for detection of pancreatic malignancies | 85 | 58 | 77 | 66 | - | 75 | - | - | - |
| Li Y et al., 2019 [97] | 16 studies, n = 1325 | CE EUS for pancreatic masses | Pancreatic masses | Diagnostic performance of CE EUS for the differentiation of pancreatic masses | 93 | 84 | - | - | - | - | - | LR+ 5.58 LR− 0.09 DOR 72.5% |
- |
| Yamashita Y et al., 2019 [98] | 9 studies, n = 887 | CE EUS | Pancreatic cancer | Diagnostic performance for diagnosing pancreatic cancer | 93 | 80 | - | - | - | - | - | LR+ 4.56 LR− 0.09 DOR 59.89 |
- |
| Shin CM et al., 2023 [99] | 6 studies, n = 430 | Combined CE EUS and EUS elastography in solid pancreatic lesions | Solid pancreatic lesions | Diagnostic performance in detecting pancreatic malignancies | 84 | 85 | - | - | - | - | - | LR+ 5.31 LR− 0.15 DOR 67.72 |
- |
| Facciorusso A et al., 2021 [100] | 6 studies, n = 701 | CE EUS-guided vs. standard EUS FNA in pancreatic masses | Solid pancreatic lesions | Diagnostic outcome | 84.6 vs. 75.3 (p < 0.001) | 100% both | - | - | 95.1 vs. 89.4 (p = 0.02) | 88.8 vs. 83.6 (p = 0.05) | - | Histological core procurement p = 0.08, number of needle passes p = 0.29 | - |
| EUS tissue acquisition (EUS TA) | |||||||||||||
| Banafea O et al., 2016 [101] | 20 studies, n = 2761 | EUS FNB | Pancreatic mass | Diagnsotic accuracy | 90.8 | 96.5 | - | - | - | 91 | - | LR+ 14.8 LR− 0.12 DOR 142.47 |
35 of 1760 patients in 15 studies |
| Guedes HG et al., 2018 [56] | 4 studies, n = 504 | 22G versus 25G needles in EUS FNB for solid pancreatic mass |
Solid pancreatic masses | Diagnostic performance | 91 vs. 93 p>0.05 |
83 vs. 87 p>0.05 |
- | - | - | - | - | LR+ 4.26 vs. 4.57 LR− 0.13 vs. 0.08 p > 0.05 |
- |
| Xu MM et al. [102] | 11 studies, n = 837 | 22G vs. 25G EUS FNA needle | Solid pancreatic lesions | Diagnostic performance | 88 vs. 92 p = 0.046 |
100 vs. 100 p = 0.842 |
- | - | - | - | - | LR+ 12.61 vs. 8.44 LR− 0.16 vs. 0.13 AUSROC 0.97 vs. 0.96 |
- |
| Tian G et al., 2018 [103] | 16 studies, n = 1824 | 22G vs. 25G EUS FNA needle | Masses with suspicion of pancreatic cancer | Diagnostic yield for the detection of pancreatic cancer | 89 vs. 90 p = 0.02 |
100 vs. 99 p = 0.15 |
- | - | - | - | - | LR+ 485.28 vs. 59.53 LR− 0.11 vs. 0.10 AUROC 0.97 for both |
- |
| Yang Y et al., 2016 [104] | 16 studies, n = 828 | EUS FNB | Solid malignant pancreatic lesions | Diagnostic accuracy | 84 | 98 | LR+ 8 LR− 0.17 DOR 64 AUROC 0.96 |
- | |||||
| Reneleus et al., 2021 [63] |
11 studies, n = 1365 | EUS FNB vs. FNA | Solid pancreatic lesions | Diagnostic accuracy and safety | - | - | - | - | - | Diagnostic accuracy of 87 vs. 81 (p = 0.005). Cytopathological accuracy of 89 vs. 82 (p = 0.04). Histological accuracy of 81 vs. 74 (p = 0.39) |
- | Mean TSR was 99% in both. Mean needle passes required for adequate tissue was 2.3 vs. 1.6 (mean difference was 0.71) (p < 0.0001) |
2.3 vs. 1.8 (p = 0.64) |
| van Riet PA et al., 2021 [62] | 18 RCTs, n = 2695 | EUS FNB vs. FNA for sampling | Solid pancreatic and non-pancreatic lesions | Diagnostic accuracy, adequacy, number of passes, presence of tissue cores, and adverse events | - | - | - | - | 90 vs. 88 (p = 0.76) | 85 vs. 80 (p = 0.03) High-quality studies 82 vs. 74 (p = 0.002) |
- | Mean number of passes was lower in FNB (mean difference −0.54) p = 0.03. Presence of tissue cores: 79 vs. 63 (p = 0.11) |
0.8 vs. 1.0 (p = 0.8) |
| Hassan GM et al., 2022 [105] | 9 RCTs, n NA | EUS FNB vs. EUS FNA | Solid pancreatic masses | Diagnostic accuracy for the diagnosis of pancreatic cancer | - | - | - | - | - | FNB had a superior accuracy compared to FNA (OR 1.87) | - | - | - |
| Bang JY et al., 2016 [106] | 9 studies, n = 576 | Procore vs. standard EUS FNA needle in solid lesions | All solid lesions | Diagnostic adequacy, diagnostic accuracy, acquisition of histological core tissue, and mean number of passes | - | - | - | - | 75.2 vs. 89.0; OR 0.39 (p = 0.23) | 85.8 vs. 86.2; OR 0.88 (p = 0.53) | - | Rate of histological core specimen acquisition (77.7% vs. 76.5%; OR 0.94, p = 0.85). Lower mean number of passes required for diagnosis with the ProCore needle (SMD—1.2, p < 0.001). |
- |
| Li Z et al., 2022 [107] | 18 studies, n = 2718 | EUS FNB vs. EUS FNB | Pancreatic and non-pancreatic solid lesions (only solid pancreatic lesions are mentioned in the subgroup analysis) | Diagnostic accuracy, number of needle passes, adequacy, presence of tissue cores, and adverse events | - | - | - | - | FNB had a higher adequacy (RR = 0.93) p = 0.004 | Similar pooled accuracy (RR = 0.97) p = 0.13 | - | Fewer number of passes for adequate sampling in FNB group (MD 0.57) p < 0.00001. Presence of tissue core was similar (RR 0.60) p = 0.16 |
Similar (RR 1.27) p = 0.97 |
| Facciorusso A et al., 2020 [108] | 11 trials, 833 patients | 22G FNB vs. 22G FNA needle | Solid pancreatic lesions | Diagnostic outcome and tissue adequacy | 93.1 vs. 90.4 | 100 in both | - | - | Slightly in favour of FNB (p = 0.61) | - | - | No difference in histological core procurement (p = 0.86). Similar number of passes in FNB (MD -0.32, p = 0.07) |
Six adverse events in FNA group and one in FNB group reported |
| Gkolfakis P et al., 2022 [69] | RCT 16, n = 1934 | Different FNB needles | Solid pancreatic masses | Diagnostic accuracy (network meta-analysis) | 94.6% with Franseen needle, 93.9% with Fork-tip needle, 90.4% with Menghini-tip needle, 82% with reverse-bevel needle, and 87.4% with FNA needle | Pooled specificity 100% with all needles tested | - | - | Franseen needle was better than FNA and reverse-beveled needles. Fork-tip needles were superior to reverse-beveled needle. None was superior when compared to FNA with ROSE. Both 22G and 25G Franseen needles followed by the 22G fork-tip needle showed the highest SUCRA scores concerning sample adequacy |
Franseen needle was better than FNA and reverse-beveled needles. Fork-tip needles were superior to reverse-beveled needle. None was superior when compared to FNA with ROSE. The 22G Franseen needle ranked as the best FNB needle in terms of diagnostic accuracy (SUCRA score of 0.81) |
- | - | Pooled rate was 2.7% with Franseen needle; 2% with Fork-tip needle; 1.3% with Menghini-tip needle; 0.8% with reverse-bevel needle, and 1.9% with the FNA needle. |
| Facciorusso A et al., 2019 [109] | 24 studies, n = 6641 | Franseen vs. fork-tip EUS FNB needles | Pancreatic and non-pancreatic solid lesions (only solid pancreatic lesions are mentioned in the subgroup analysis) | Sample adequacy | Similar sensitivity (95.3 vs. 93.4) | Similar specificity [100] | - | - | 97 vs. 92.6 (p = 0.006) | 96.8 vs. 95.2 (p = 0.8) | - | Histological core procurement was 94 vs. 93.1 (p = 0.7). Fewer number of passes compared to standard FNA needles (MD for Franseen was -0.44 and for Fork-tip was −1.82) |
- |
| Facciorusso A et al., 2022 [110] | 8 studies, n = 2147 | EUS FNB with and without ROSE | Solid pancreatic lesions | Sample adequacy | 94.3 vs. 91.5 | - | - | - | EUS FNB with ROSE is not superior to EUS FNB alone (95.5 vs. 88.9, p = 0.07) especially when end-cutting needles (compared to reverse-bevel needles) are used | Superior in the EUS FNB + ROSE group (OR = 2.49, p = 0.03) | Number of needle passes needed to obtain diagnostic samples was not significantly different (mean difference 0.07; p = 0.62) | - | Only one study reported (Crino et al.) |
| Kong F et al., 2016 [111] | 7 studies, n = 1299 | EUS FNB with ROSE vs. EUS FNB without ROSE | Pancreatic masses | Diagnostic adequacy, yield, number of needle passes, pooled sensitivity, and specificity | 91 vs. 85 | 100 in both | - | - | No significant difference in cytological adequacy | No significant difference in diagnostic yield | - | LR+ 28.15 vs. 29.08 LR− 0.1 vs. 0.16. Fewer needle passes in ROSE group (4 vs. 7, p < 0.0001) |
- |
| Lisotti A et al., 2020 [112] | 12 studies, n = 505 | Repeat EUS FNB for the diagnosis of solid pancreatic masses | Solid pancreatic masses | Diagnostic performance of repeat EUS FNB in case of negative or inconclusive first FNA | 77 (83% with ROSE) | 98 | 99 | 61 | - | - | - | LR+ 38.9 LR− 0.23 |
- |
| Han S et al., 2021 [113] | 26 studies, n = 3398 (in primary NMA) | Various EUS TA needles | Solid pancreatic masses | Diagnostic accuracy compared to 22G Echotip (Cook) EUS FNA needle (NMA) | - | - | - | - | - | Performance score-wise: 22 G SharkCore FNB needle (Medtronic) > 22G EZ Shot 3 FNB needle (Olympus) > 22G Acquire FNB needle (Boston Scientific) |
- | Diagnostic accuracy was not significantly different between needles with or without suction except 20G FNB needle with suction which performed significantly worse than the 22G FNA needle with suction | - |
| Suction Techniques in EUS TA | |||||||||||||
| Facciorusso A et al., 2023 [42] | 9 RCTs, n = 756 | Various EUS FNB techniques | Solid pancreatic masses | Rates of sample adequacy, blood contamination, and tissue integrity (NMA) | Modified wet suction was most sensitive (SUCRA score, 0.85) followed by slow-pull techniques and no stylet technique (SUCRA scores, 0.66 and 0.48, respectively) | - | - | - | Modified wet-suction technique was best for adequacy (SUCRA score of 0.90) followed by dry-suction and slow-pull techniques (SUCRA scores of 0.59 and 0.50, respectively) | - | Higher level of blood contamination seen with dry-suction than slow-pull technique; no-suction technique ranked as the best strategy (SUCRA score of 0.99) followed by the slow-pull technique (SUCRA score of 0.65). Modified wet-suction (SUCRA score of 0.32) and dry-suction (SUCRA score of 0.12) techniques showed poor performance in terms of blood contamination of the sample |
Regarding tissue integrity, modified wet-suction technique was ranked as the best strategy (SUCRA score of 0.89) followed by slow-pull (SUCRA score of 0.66) and no-suction (SUCRA score of 0.42) techniques | Uncommon and usually mild, without significant impact on patient outcomes (abdominal pain and bleeding) |
| Ramai D et al., 2021 [76] | 6 studies, n = 418 | Wet vs. dry suction techniques | Solid pancreatic masses | Adequacy, sample contamination, and histological accuracy | - | - | - | - | Wet-suction technique has superior tissue adequacy (pooled adequacy rate of 91.9 vs. 77.32 (OR 3.18, p < 0.001)) | Wet-suction technique is superior in histological diagnosis (OR of 3.68, pooled rate of 84.06 vs. 68.87, p < 0.001). Wet suction has superior sample quality, and accuracy | Wet-suction technique has comparable blood contamination (OR of 1.18, contamination rate of 58.33 and 54.6, p = 0.256) | - | - |
| Giri S et al., 2023 [81] | 7 studies, n = 2048 | Various suction techniques in EUS TA | Solid pancreatic and non-pancreatic lesions | Compare the diagnostic yields during EUS TA (NMA) | - | - | - | - | There was no difference between the various modalities. For the SUCRA analysis, WS > SSP > DS > NS | No significant difference in ORs of adequacy when adjusted for either of the needle types. For the SUCRA analysis, WS > NS > DS > SSP | When adjusting for FNA needle, there was no difference between the interventions | No significant difference between the studies with respect to moderate-to-high cellularity of samples | - |
| EUS TA in Presence of Biliary Stents | |||||||||||||
| Facciorusso A et al., 2023 [114] | 7 studies, n = 2458 | EUS TA in presence and absence of biliary stent | Solid pancreatic head masses | Diagnostic accuracy before and after biliary stenting in jaundiced patients with pancreatic head masses | Overall diagnostic sensitivity lower in biliary stent group (82.9 vs. 87.5; OR 0.59; p < 0.001); in SEMS subgroup (p = 0.006) but not in plastic stent group (p = 0.12) | - | - | - | No significant difference in adequacy (p = 0.81) | No overall significant difference 85.4 vs. 88.1 (p = 0.07). No significant difference in plastic stent vs. no stent (p = 0.67). Significant difference in SEMS vs. no SEMS (p = 0.05) |
- | No significant difference in number of needle passes (p = 0.38) | No significant difference (p = 0.75) |
| Giri S et al., 2023 [115] | 9 studies, n = 3257 | EUS TA in presence and absence of biliary stent | Pancreatic masses undergoing EUS TA | Diagnostic accuracy of EUS TA in presence and absence of biliary stent | 79 vs. 88; Using non-strict criteria in patients with stents, the sensitivity was lower with metal stents than with plastic stents (83% vs. 90%) | - | - | - | Comparable in stent vs. non-stent groups and in plastic and SEMS group | Lower accuracy with stent (OR of 0.58) using non-strict criteria and comparable sensitivity between metal stents and plastic stents |
- | Patients with stents required greater number of passes (MD = 0.31) | - |
| Advances of EUS | |||||||||||||
| Chandan S et al., 2020 [116] | 9 studies, n = 1308 | EUS-guided precipitation-based LBC conventional smear | Solid pancreatic masses | Diagnostic yield of EUS-guided conventional smear vs. LBC | Precipitation based LBC higher sensitivity (85.2 vs. 79.7) | Precipitation based LBC comparable specificity (99.5 vs. 99.4) | Precipitation based LBC comparable PPV (99.5 in both) | NPV was found to be higher with filtration-based LBC technique (50.9%) as compared with CS (46.2%) and precipitation-based LBC techniques (35.4%). | - | Precipitation-based LBC had a higher accuracy | - | - | - |
| Prasoppokakorn T et al., 2021 [117] | 8 studies, n = 870 | AI-assisted diagnosis of PDAC by EUS | Pancreatic mass | AI-assisted B-mode EUS sensitivity and specificity 90%, 91% respectively. AI-assisted CE EUS sensitivity and specificity 95%, 95% respectively. AI-assisted EUS elastography sensitivity and specificity 88%, 83% respectively. |
AI-assisted EUS 91% AI-assisted B-mode EUS 91% |
AI-assisted EUS 90% AI-assisted B-mode EUS 90% |
AI-assisted B-mode EUS 94% | AI-assisted B-mode EUS 84% | - | - | - | - | - |
| Dhali A et al., 2023 [118] | 21 studies | AI-assisted vs. conventional EUS for detection of pancreatic SoLs | Diagnostic performance | Higher accuracy of AI-assisted EUS for detection and differentiation | 93.9 | 93.1 | 91.6 | 93.6 | - | 93.6 | - | - | - |
Abbreviations: PDAC—Pancreatic ductal adenocarcinoma; EUS—Endoscopic ultrasound; CE EUS—Contrast-enhanced EUS; EUS TA—EUS-guided tissue acquisition; FNA—Fine-needle aspiration; FNB—Fine-needle biopsy; LBC—Liquid-based cytology; CS—Conventional smear; ROSE—Rapid on-site cytology evaluation; MOSE—Macroscopic on-site cytology evaluation; SEMS—Self-expanding metal stent; WS—Wet-suction method; SSP—Stylate slow-pull method; DS—Dry-suction method; AI: Artificial intelligence; PPV—Positive predictive value; NPV—Negative predictive value; OR—Odd’s ratio; RR—Relative risk; AUROC—Area under receiver operating curve; NMA—Network meta-analysis; LR—Likelihood ratio.