Evaluating Pancreatic and Biliary Neoplasms with Small Biopsy-Based Next Generation Sequencing (NGS): Doing More with Less

Simple Summary Pancreatic cancer and cholangiocarcinoma are aggressive diseases mostly diagnosed at an advanced and inoperable stage. This review presents the value of next-generation sequencing (NGS) when performed on small biopsies—including fine-needle aspiration/biopsy samples, brushings, pancreatic juice and bile, and also blood—in the field of pancreatobiliary neoplasia. NGS could guide physicians while evaluating pancreatic solid and cystic lesions or suspicious biliary strictures, performing surveillance in high-risk individuals, or monitoring the disease and assessing prognosis in already diagnosed cancer patients. Evidence suggests that NGS performed on small biopsies is a robust tool for the diagnosis and pre-operative risk stratification of pancreatic and biliary lesions, whereas it also carries significant prognostic and therapeutic value. However, effective standardization of the pre-analytical and analytical assay parameters used for each clinical scenario is needed to fully implement NGS into routine practice and provide more personalized management in patients with suspected or established pancreatobiliary neoplasia. Abstract Pancreatic cancer and cholangiocarcinoma are lethal diseases mainly diagnosed at an inoperable stage. As pancreatobiliary surgical specimens are often unavailable for further molecular testing, this review aimed to highlight the diagnostic, prognostic, and therapeutic impact of next-generation sequencing (NGS) performed on distinct small biopsies, including endoscopic ultrasound fine-needle aspirations and biopsies of pancreatic solid and cystic lesions, biliary duct brushings, and also “liquid biopsies” such as the pancreatic juice, bile, and blood. NGS could clarify indeterminate pancreatic lesions or biliary strictures, for instance by identifying TP53 or SMAD4 mutations indicating high-grade dysplasia or cancer. It could also stratify pancreatic cystic lesions, by distinguishing mucinous from non-mucinous cysts and identifying high-risk cysts that should be excised in surgically fit patients, whereas the combination of cytology, elevated cystic CEA levels and NGS could improve the overall diagnostic accuracy. When NGS is performed on the pancreatic juice, it could stratify high-risk patients under surveillance. On the plasma, it could dynamically monitor the disease course and response to therapy. Notably, the circulating tumor DNA (ctDNA) levels have been associated with staging, grading, and survival. Lastly, NGS has shown potential in identifying potentially actionable molecular alterations. In conclusion, NGS applied on small biopsies could carry significant diagnostic, prognostic, and therapeutic value.


The Role of NGS Performed on Pancreatic Small Biopsies
The summary of the published studies reporting on the role of small biopsy-based NGS in the evaluation and management of pancreatic lesions is shown in Table 1. Most studies highlighted its value in diagnosis (e.g., indeterminate case clarification or pre-operative stratification) or targeted therapy selection. Habib, 2021 [54] FNA; plasma cfDNA Lesions suspicious for PDAC 9 gene panel FNA-based NGS identified 16/16 of the KRAS mutations found in their paired histological specimens, in contrast to 6/8 identified by the plasma-based molecular analysis; mutations in the KRAS and TP53 genes were the most common ones detected Dupain, 2020 [55] CT or EUS-FNA and EUS-FNB Pancreatic cancer metastases 87 gene panel Among the metastatic tumors (e.g., from pancreas, breast, and colon) prospectively tested, FNA-based was highly concordant with the CNB-based NGS; potentially actionable alterations were also identified De Biase, 2020 [56] FNAs and direct fluid samples Solid and cystic pancreatic lesions 22 gene panel KRAS p.G12V and p.G12D were the most common mutations detected in the 42 pancreatic lesions tested Carrara, 2020 [57] EUS-FNA and EUS-FNB PDACs 161 gene panel In this clinical trial, NGS was successful in almost all samples tested and exhibited higher diagnostic yield (94%) than histology (91%) or cytology (88%); at least two mutations were found in the majority of PDAC cases, whereas KRAS mutations were the most common ones detected Fulmer, 2020 [58] EUS-FNA Solid and cystic pancreatic lesions 143 gene panel DNA of high quality was retrieved from most samples; NGS revealed clinically significant mutations in 10/14 mucinous cysts (e.g., KRAS, GNAS, TP53 mutations) and 13/15 PDACs (KRAS mutations in 10 and TP53 in 9 samples), whereas it did not exhibit any mutation in the 4 PanNETs tested Plougmann, 2020 [59] EUS-FNA Solid pancreatic lesions 19  Mutations were found in all tested IPMNs (n = 12), most often in the KRAS and GNAS genes, whereas none of the tested pseudocysts (n = 3) showed any KRAS/GNAS mutations; cellular fraction exhibited superior results than the liquid fraction molecular analysis Pancreatic juice Pancreatic solid and cystic lesions, also non-neoplastic controls 9 gene panel PDAC patients showed higher mutation concentrations than IPMNs or controls; mutations in the TP53 and SMAD4 genes were found most often in PDACs, whereas they were also detected in 15/57 and 1/57 of the IPMNs tested, respectively, albeit in none of the controls; KRAS mutations were also found in 10/24 of the controls; two high-risk patients under surveillance showed TP53 or SMAD4 mutations in the pancreatic juice-based molecular analysis, more than a year before their cancer diagnosis EUS-FNA or direct collection from the resected tissue specimens Pancreatic cystic lesions 11 gene panel KRAS and GNAS mutations were the most common ones found in IPMNs (78% and 58% of the cases, respectively); KRAS mutations were the most common ones found in MCNs (6/12 cases tested); CTNNB1 mutations were found in SPNs, whereas VHL mutations were found in SCAs Wang, 2015 [80] EUS-FNA Pancreatic cystic lesions Non-coding RNA sequencing miRNA expression profiling was used to distinguish low-grade from high-grade/malignant pancreatic cystic lesions; the latter showed enrichment of 13 and depletion of two miRNAs

Preoperative Evaluation of Pancreatic Cysts
A few of the published studies aimed to unravel the value of NGS in the preoperative evaluation of pancreatic cysts, in order to reduce unnecessary surgical procedures. This challenging task emerges more often in recent years, as more incidental cysts are detected, following the prevalent use of enhanced imaging technology [86]. To manage pancreatic cysts effectively, physicians should generally decipher if: (a) the cysts are mucinous or non-mucinous (the latter can be either non-neoplastic or benign with minimal malignant potential which can safely be managed conservatively), and (b) there is presence of at least HGD within the cysts; these would be classified as high-risk cysts, which are triaged for surgery [42]. In accordance with the literature, this review also found that the presence of KRAS mutations supported the diagnosis of a mucinous (IPMN or MCN) over a nonmucinous cyst (e.g., pseudocyst or SCA), whereas GNAS mutations favored IPMN over MCN [25,29,65,66,76,79]. NGS enhanced the diagnostic accuracy of EUS-FNA cytology to detect neoplastic mucinous cysts and differentiate them from the non-mucinous ones [49]. Notably, evidence indicated that NGS was more sensitive than the cytologic examination or elevated CEA cystic fluid levels (≥192 ng/mL), which are the two modalities traditionally used to evaluate pancreatic cysts [25,48,52,68]. For instance, Ren et al. showed the combination of cytologic examination, elevated CEA cystic fluid levels, and NGS reached a sensitivity of 94.1% and a specificity of 100% for the detection of neoplastic mucinous cysts [48]. Apart from discriminating between mucinous and non-mucinous pancreatic cysts, NGS was able to robustly identify high-risk cysts. A few studies indicated that specific mutations detected with NGS were associated with cystic neoplasms exhibiting HGD or invasion [25,76,87]. Rosenbaum et al. examined 113 pancreatic cystic fluid lesions from 105 patients and reported that SMAD4, TP53, CDKN2A, or NOTCH1 mutations indicated the presence of IPMN with high-grade dysplasia or cancer. Of interest, NGS combined with cytology improved the overall diagnostic accuracy to detect IPMNs and identify the high-risk IPMNs [25]. Similarly, Jones et al. also found that the presence of SMAD4, TP53, or CDKN2A alterations, discovered with NGS, indicate IPMNs with high-grade dysplasia or invasion [76].

Evaluation of High-Risk Patients under Surveillance with Pancreatic Juice-Based NGS
Likewise, some teams utilized pancreatic juice-based NGS to recognize HGD or cancer while evaluating solid or cystic pancreatic lesions. For instance, TP53 or multiple KRAS mutations were associated with invasive IPMNs [24,67]. Furthermore, Suenaga et al. tested the pancreatic juice from a mixture of pancreatic cancer and precursors (with both LGD and HGD) under surveillance, in addition to normal controls. They found that patients with HGD or cancer exhibited higher numbers and concentration of mutations other than KRAS/GNAS (also a higher overall mutation concentration) in their pancreatic juice. Mutations in TP53 and/or SMAD4 or a high SMAD4/TP53 mutation score were associated with HGD or cancer, whereas none of them were detected in the controls. Thus, NGS facilitated the stratification of high-risk patients under pancreatic surveillance, by identifying the patients harboring at least HGD [23]. Yu et al. applied pancreatic juicebased NGS in a cohort of 115 pancreatic solid and cystic lesions (34 PDACs, 57 IPMNs, and 24 non-neoplastic controls). They reported that PDAC patients showed higher mutation concentrations than IPMNs or controls. Although TP53 and SMAD4 mutations were associated with PDACs, they were also detected in 15/57 and 1/47 of IPMNs, respectively, but in none of the controls. Notably, two high-risk patients of the cohort under surveillance showed TP53/SMAD4 mutations more than a year before their cancer diagnosis [26].

Identification of Potentially Actionable Mutations in PDAC Patients
Apart from its use in diagnosis and preoperative risk stratification of pancreatic solid and cystic lesions, small biopsy-based NGS also showed potential in identifying potentially actionable alterations in PDAC patients. Takano et al. found such alterations in 22.4% of the cases tested [50], whereas Elhanafi et al. identified actionable mutations in the BRAF, MET, ERBB2, ARID1A, and BRCA1 genes in a few of the PDACs tested [70]. Lastly, Valero et al. reported at least one mutation in 17/19 of PDAC patients of their cohort, whereas KRAS, TP53, SMAD4, and ARID1A mutations were the ones most commonly detected. Notably, actionable alterations (e.g., in ATM or mTOR genes) were also found in some samples [46].

Evaluation of Neoplasms Other Than PDAC and Its Precursors
Whereas most studies focused on PDAC and its precursors, small biopsy-based NGS was also used to evaluate the molecular profile of other pancreatic neoplasms, pointing to a specific diagnosis or providing additional prognostic and therapeutic information. Gleeson et al. tested 90 primary and 32 metastatic PanNETs from the liver and reported that the former most often harbored MEN1, DAXX, ATRX, and TSC2 mutations. In addition, they found that alterations in TSC2, KRAS, and TP53 genes were associated with poor prognosis, whereas they also identified potentially actionable alterations in some members of the mTOR pathway (PTEN, TSC2, and PIK3CA) in 10% of primary and 12.5% metastatic NETs tested [74]. Whereas KRAS mutations were often in PDACs and IPMNs, they were not detected in the PanNET cases tested in two studies [58,83]. VHL mutations indicated a diagnosis of SCA in some studies. Of interest, Vestrup Rift et al. found that although mutations in KRAS and GNAS genes were the most common ones found in IPMNs, they were not detected in the three SCAs tested [66,68,76]. Furthermore, the presence of a CTNNB1 mutation indicated SPN; Kubota et al. found a CTNNB1 mutation in all seven SPNs, yet in just 1/11 NETs and in none of the PDACs, acinar cell carcinomas and pancreatitis cases of their cohort [81].

NGS Performed on FNA vs. Tissue Biopsy Samples
Evidence has shown that FNA-based NGS was highly concordant with its matched tissue-based molecular analysis, where it often revealed additional alterations, modifying the management plan of the patients [54,61,73,75,85]. In addition, it exhibited superior sensitivity than PCR or Sanger sequencing [83]. Rapid on-site evaluation (ROSE), besides improving the diagnostic accuracy of EUS-FNA, facilitated the acquisition of material for subsequent NGS testing, sparing the patients from additional invasive procedures [60]. Of interest, FNB was more likely to result in adequate material for subsequent NGS testing than FNA (OR: 4.95; 95% CI: 1.11-22.05; p = 0.04) [70], whereas larger gauge biopsy needles were more likely to result in successful NGS findings [64].

The Role of NGS Performed on Biliary Small Biopsies
The summary of the published studies reporting on the role of small biopsy-based NGS in the evaluation and management of suspicious biliary strictures is shown in Table 2.  NGS was performed on biliary tract brushings, forceps biopsies, and bile, whereas the authors utilized LBC samples or material directly collected for molecular evaluation [78,[88][89][90][91][92]. Mutations in the KRAS, TP53, SMAD4, and CDKN2A genes were the most common ones detected [78]. Diagnostic accuracy was found to be relatively high. Rosenbaum et al. examined 96 strictures from 88 patients and reported that NGS exhibited higher sensitivity than cytology, whereas the presence of TP53, SMAD4, and CDKN2A mutations was 100% specific to detect HGD or cancer [90]. Furthermore, Singhi et al. examined 346 benign, premalignant, and malignant strictures from 252 patients and reported a sensitivity of 73% and specificity of 100% for malignancy. NGS exhibited an enhanced performance compared with the CA19-9 serum levels or the pathologic evaluation (performed on biliary brushings, biopsies, or both), whereas it also improved the overall diagnostic accuracy when combined with the pathologic evaluation. Notably, it also revealed potentially actionable alterations, such as the ERBB2 amplification in a few patients [92]. Two other studies additionally reported an improvement in the diagnostic accuracy, when the results of NGS were combined with the cytomorphologic evaluation [78,91]. Furthermore, NGS was found to be more sensitive, specific, and accurate than FISH, an already established method used to triage indeterminate biliary tract specimens [78].
Two research groups evaluated the potential of bile-based NGS in the evaluation of suspicious biliary strictures. NGS was more sensitive to detect malignancy, compared with the initial pathomorphological evaluation, performed either with FNA or FNB [88]. Notably, results were highly concordant with the molecular analysis performed in the matched tissue specimens, as 96.2% of the alterations present in the tissues were detected with bile-based NGS [89].

The Role of NGS Performed on Blood-Based Liquid Biopsies
The published evidence concerning the role of blood liquid biopsy-based NGS in the evaluation of pancreatic neoplasms is summarized in Table 3. Most studies highlighted its value in monitoring patients already diagnosed with locally advanced or metastatic PDAC, besides assessing prognosis or selecting the most appropriate targeted therapy in this clinical setting.  Multi-analyte liquid biopsy (EV-derived mRNA/miRNA, cfDNA concentration, KRAS MAF, and CA19-9 levels) exhibited superior diagnostic accuracy to detect and stage PDACs than CA19-9 and imaging, respectively; this approach also spotted metastases missed by imaging at baseline, which were later discovered during surgery or follow-up imaging, exhibiting the potential to identify suitable surgical candidates

54-73 gene panel
During the advanced setting, before or after surgery TP53 and KRAS mutations were the most common ones found, whereas potentially actionable mutations were also identified in most advanced PDACs; advanced PDACs also showed higher number of aberrations and ctDNA amount (% ctDNA) than the resectable ones; concordance between plasma and tissue NGS was 61% and 52% for TP53 and KRAS mutations, respectively; increased total % ctDNA was associated with shorter OS Although not detecting all mutations found in the tissue-based NGS, liquid biopsy identified a much higher number of alterations not detected in its paired biopsies, reflecting more efficiently the intratumoral heterogeneity; cfDNA collected during progression revealed additional mutations not identified at the pre-operative cfDNA samples Pietrasz, 2017 [120] Plasma cfDNA Patients with resectable, locally advanced, or metastatic PDAC

gene panel
Before the first cycle of chemotherapy (after surgery for the resectable patients); serial sampling for 8 patients KRAS, TP53, and SMAD4 mutations were the most common ones detected; the presence of ctDNA was associated with tumor grade and stage (higher detection rates in high-grade and metastatic PDACs); ctDNA presence and quantity was associated with shorter OS in advanced PDACs, whereas its absence conferred longer OS and DFS in resected PDACs Adamo, 2017 [121] Plasma cfDNA Patients with PDAC or CP, and healthy controls 50 gene panel Before therapy PDACs exhibited higher cfDNA yields than CPs and controls; KRAS mutations were the most common ones detected and were associated with poor prognosis; when both plasma and tissue biopsy were available, plasma NGS failed to detect any mutations detected in their paired tissue biopsies Before (baseline) and during chemotherapy, also with each CT ctDNA was found in 93.7% of the patients at baseline, even in cases where CA19-9 was undetectable; the combination of ctDNA and CA19-9 increased sensitivity; ctDNA quantity was higher in stage IV than III PDACs, whereas higher ctDNA amount was associated with disease progression and shorter TTP and OS at baseline, being a more significant prognostic marker than serum CA19-9; ctDNA quantity changes at the longitudinal plasma samples predicted response to therapy in most patients Takai In this phase II clinical trial, ctDNA was detected in most patients, whereas mutations in KRAS, TP53, ATM, and CDKN2A were the most common ones found at baseline; when paired plasma and tissue biopsy were available in the same patient KRAS mutation detection was 100% concordant between them; most mutations detected at baseline were also found at the follow-up samples, whereas relative ctDNA quantity was linked with the serum CA19-9 levels and tumor burden Zill, 2015 [127] Plasma cfDNA Patients with advanced PDAC or biliary carcinoma 54 gene panel Baseline; serial sampling for 8 patients (monitoring) Plasma NGS exhibited high sensitivity, specificity, and diagnostic accuracy, whereas it even detected additional alterations from its paired tissue-based NGS; KRAS and TP53 mutations were the most common ones found, whereas actionable alterations (e.g., BRAF or EGFR mutations) were also identified; in the serial samples, changes in ctDNA quantity correlated with the tumor marker (e.g., CA19-9) altered levels, reflecting disease progression or therapy response

Monitoring Disease Course and Response to Therapy in PDAC Patients
Most studies used plasma cfDNA and targeted gene panels for NGS analysis, whereas the blood collection point ranged from treatment-naive patients (before chemotherapy or surgery), and also from patients during therapy and at disease progression. As it allows serial sampling, plasma-based liquid biopsy has shown great potential in the dynamic monitoring of the disease course and response to therapy of PDAC patients. Berger et al.
performed NGS using the plasma cfDNA from 20 patients with metastatic PDAC and reported that their mutational landscape was often altered from baseline to the first, second, and third lines of treatment. Of interest, ctDNA quantity dropped from the baseline levels (before treatment initiation) during chemotherapy, whereas it surged during progression. In treatment-naive patients, the decrease in ctDNA quantity during therapy was associated with longer progression-free survival (PFS) [117]. Similarly, Park et al. tested 69 plasma cfDNA samples from 69 PDAC patients and found that the lowest ctDNA levels were associated with complete/partial disease response; thus, ctDNA levels were successful to monitor tumor burden, response to therapy, and disease progression [116]. Another study monitored 189 stage III and IV PDAC patients before, during chemotherapy, and together with each CT scan. They reported that ctDNA quantity was higher in stage IV than III PDACs, whereas higher ctDNA quantity was associated with disease progression and shorter overall survival (OS) at baseline, being a more significant prognostic marker than serum CA19-9 levels. CtDNA quantity changes during the sequential sampling predicted response to therapy in most of these patients [122]. In addition, Bachet et al. performed a randomized phase 2b trial enrolling 122 advanced PDAC patients, and showed that the presence of ctDNA at the first chemotherapy cycle was associated with shorter OS and PFS. Additionally, patients who responded to therapy exhibited negative or low ctDNA levels, whereas the ctDNA quantity alterations detected during sequential plasma sampling were associated with the overall response rate (ORR), OS, and PFS [101]. Notably, the presence of CTCs or ctDNA could be used to monitor PDAC patients receiving neoadjuvant therapy (NAT). A study by Yin et al. extracted the ctDNA and CTCs in a cohort composed of patients with pathologic complete response (pCR) after NAT. They found ctDNA in 7/16 and CTCs in 5/5 of the patients tested, suggesting their presence could indicate recurrence and worse survival [97].

Assessing Prognosis of PDAC Patients
As also displayed in the aforementioned studies, evidence suggests that the presence of plasma ctDNA and/or its levels are associated with the PDAC burden, staging, grading, and prognosis. Strijker et al. tested 77 plasma cfDNA samples from 58 metastatic PDAC patients and reported that ctDNA was most often found in patients with larger tumors and liver metastases. In addition, the ctDNA quantity was associated with 3D tumor volume (as measured by imaging), whereas it also predicted OS [108]. In another study, a higher tumor fraction was correlated with liver metastasis, shorter OS, and higher CA19-9 serum levels [100]. Pietratz et al. utilized plasma-based NGS in a cohort composed of resectable, locally advanced, and metastatic PDACs, and demonstrated that the presence of ctDNA was associated with tumor grade and stage (higher detection rates in high-grade and metastatic PDACs). Additionally, ctDNA presence and quantity were associated with shorter OS in advanced PDACs, whereas its absence conferred longer OS and DFS in resected PDACs [120].

Identifying Potentially Actionable Mutations in PDAC Patients
Besides its ability to monitor the disease course and response to therapy and its prognostic value, plasma-based NGS could also identify potentially targetable alterations in PDAC patients. In one study, actionable alterations were detected in 14/48 patients (e.g., in ALK, ATM, EGFR, and PIK3CA) [123], whereas in another one, such alterations (e.g., in BRCA1, EGFR, MET, BRAF, PIK3CA, and ERBB2) were also identified [104]. In the study by Li et al., two patients were successfully treated with immune checkpoint and PARP inhibitors (PARPi), based on the detection of MLH1 and BRCA1 mutations, respectively [103]. Lastly, Vidula et al. detected BRCA1/2 mutations in the plasma cfDNA samples tested, tailoring patients for treatment with PARPi therapy, whereas they also identified mechanisms of PAPRi resistance, such as BRCA1/2 reversion mutations [47].

NGS Performed on Blood-Based Liquid Biopsy vs. Tissue Biopsy Samples
Although blood-based NGS exhibits low concordance compared with tissue biopsybased molecular analysis [112,118], it could identify additional alterations not detected in its paired biopsies, thus reflecting intratumoral heterogeneity more efficiently [119,127]. Notably, cfDNA collected during progression could also reveal new mutations, indicating tumor evolution [119].

Discussion
This review aimed to highlight the impact of small biopsy-based NGS in the evaluation of pancreatic and biliary neoplasms, guiding clinicians to provide personalized management for their patients. Evidence has shown that NGS could be applied with success in distinct small biopsies-including FNAs and FNBs of pancreatic solid and cystic lesions, pancreatic and biliary duct brushings, and liquid biopsies such as pancreatic juice, bile, and blood-providing answers to common clinical scenarios (Figure 1). Firstly, NGS could help clarify indeterminate pancreatic or biliary cases by microscopy. According to the PDAC progression model, KRAS mutations are found early, whereas TP53 and DPC4 mutations occur later during the PDAC carcinogenesis [18,19]. Thus, although KRAS mutations could be detected in PanINs and IPMNs/MCNs of any grade or even in non-neoplastic cases, TP53 and SMAD4 alterations indicate the presence of HGD or cancer, triaging eligible patients for surgery [21][22][23][24][25][26][27]. Notably, Hosoda et al. selected 23 isolated HG-PanIN cases characterized by the absence of concurrent PDAC to perform molecular analysis. They reported that TP53 mutations were found in just 2/23 cases, whereas they did not find any non-synonymous SMAD4 alterations, suggesting that both mutations arise mostly at invasion [128]. Apart from its value assessing equivocal pancreatic lesions, NGS could be used in the evaluation and management of suspicious biliary strictures, as it has shown higher sensitivity than cytology or elevated serum CA19-9 levels to detect malignancy and an enhanced performance compared with FISH [78,[88][89][90][91][92].
Furthermore, NGS could enhance the stratification of pancreatic cystic lesions-being an effective tool to distinguish mucinous from non-mucinous cysts-and identify high-risk cysts that should be excised in surgically fit patients [42]. Whereas the presence of KRAS mutations supports the diagnosis of a mucinous over a non-mucinous cyst, GNAS mutations favor the diagnosis of IPMN over MCN [25,29,65,66,76,79]. Evidence suggests that NGS has higher sensitivity than cytology or elevated cystic fluid CEA levels [25,48,52,68], whereas the combination of cytology, high CEA levels and NGS has shown the highest diagnostic accuracy to detect neoplastic mucinous cysts [48]. In addition, the presence of specific mutations-such as the ones in the SMAD4, TP53, CDKN2A, or NOTCH1 genes-have been linked with high-risk cysts [25,76,87]. To manage pancreatic cysts, clinicians most often use specific criteria described by organizations such as the International Association of Pancreatology (Fukuoka guidelines) [129] and the Americal Gastroenterological Association (AGA) [130]. For instance, the Fukuoka guidelines enlist distinct "high-risk stigmata" and "worrisome features", which should be considered before deciding to surgically excise a pancreatic cystic lesion or recommend close follow-up. The presence of suspicious or positive cytology (microscopic features consistent with HGD or invasion) also triages eligible patients for surgery [131,132]; however, although pancreatic cyst cytology has a high specificity, its sensitivity is considered suboptimal [129]. Of interest, a recent meta-analysis on the Fukuoka and AGA guidelines found they both exhibited an inadequate diagnostic accuracy to distinguish between low-and high-risk pancreatic cysts [133]. Considering the findings presented in this review, the potential inclusion of NGS testing in these guidelines may enhance their diagnostic potential.
In the field of liquid biopsies, NGS has shown promising results when testing blood, pancreatic juice, and bile from patients with pancreatobiliary neoplasia. As blood-based liquid biopsy allows serial sampling (e.g., before chemotherapy or surgery, during therapy, and at disease progression), it has shown great potential in the dynamic monitoring of the PDAC disease course and response to therapy [97,101,117]. This is of great importance, especially when considering that traditional modalities used to monitor PDAC, such as the CA 19.9 serum levels and radiology, could exhibit suboptimal accuracy [134]. In addition, the presence of plasma ctDNA and/or its quantity have been associated with the PDAC burden, staging, grading, and prognosis [100,108,120], whereas they could often reflect heterogeneity more efficiently that tissue biopsy-revealing new mutations that indicate evolution-and potentially affect prognosis and response to therapy [119,127]. Of interest, a recent meta-analysis showed that the presence of ctDNA was associated with poor OS both at baseline and post-operatively (HR = 2.27; 95%CI (1.13-4.56) vs. HR = 3.66; 95%CI (1.45-9.28), respectively) in patients with resectable PDAC. Additionally, another metaanalysis compared the diagnostic accuracy of liquid with tissue-based molecular analysis, reporting that the former exhibited a pooled sensitivity and specificity of 70% and 86%, respectively, yet concordance was just 31.9% (as shown with a Venn diagram), when all mutations were considered [135].
Pancreatic juice could also be used to monitor PDAC patients after their surgery or individuals with high-risk to develop cancer. Suenaga et al. reported that mutations in TP53 and/or SMAD4 or a high SMAD4/TP53 mutation score were associated with HGD or cancer, whereas both were not detected in the control samples of the study. Thus, NGS facilitated the stratification of high-risk patients under pancreatic surveillance, through the identification of the patients harboring at least HGD [23].
As most PDAC patients are not eligible for surgery, another emerging application of NGS could be to identify potentially actionable alterations, such as in the BRAF, MET, ERBB2, ARID1A, BRCA1, ATM and mTOR genes [46,70]. This could even be carried out at the level of plasma-based NGS. For instance, the detection of mutations in the MLH1 or BRCA genes could tailor patients for treatment with immune checkpoint and PARP inhibitors, respectively [47,103]. Such findings could shift the direction of PDAC management away from the "one size fits all" chemotherapy approach towards precision oncology, as PDAC is not a single disease, albeit exhibiting molecular heterogeneity [136]. In addition, according to the alterations detected with NGS, patients could be selected for the most suitable clinical trials [137].
Lastly, small biopsy-based NGS could also identify alterations associated with other pancreatic lesions, pointing to a specific diagnosis or providing prognostic and therapeutic information. For instance, mutations in the MEN1, DAXX, and ATRX genes have been associated with PanNETs [74,138]. Additionally, VHL mutations indicate a diagnosis of SCA or a metastasis from a renal cell carcinoma [68,76,139,140], whereas CTNNB1 mutations a diagnosis of SPN [81].
To be successful, several pre-analytical and analytical parameters associated with any small-tissue based NGS need to be optimized. For instance, as PDAC generally contains abundant desmoplastic stroma, cellularity and tumor fraction could be low, negatively impacting the sensitivity of the reaction; thus, sample adequacy needs to be assessed before running an NGS reaction [141][142][143][144]. Some studies in our review showed that FNB was more likely than FNA to result in adequate material for subsequent NGS testing [64,70]. However, the application of ROSE could facilitate the acquisition of cytologic material to be further processed for NGS testing, sparing the patients from additional invasive procedures [60]. The findings of this review should be interpreted with caution, as there was significant variation in the clinical setting of the included studies (e.g., PDACs of various stages or various percentages of different disease entities), whereas some studies recruited small patient numbers to draw meaningful results. In addition, there was substantial heterogeneity in the preanalytical-for instance, regarding FNAs, NGS was performed on directly collected material, cytology slide scraping, supernatants from post-centrifuged or residual LBC samples-or analytical parameters of the NGS assays applied (e.g., diversity of gene panels or depth of coverage). Thus, it is imperative to validate the most robust assays for each pancreatobiliary small-biopsy application described in this review. Future research in the form large prospective studies or randomized clinical trials may strengthen the aforementioned findings.

Conclusions
Evidence suggests that NGS performed on small biopsies is a robust tool for the diagnosis and risk stratification of pancreatic and biliary lesions, whereas it also carries significant prognostic and therapeutic value. However, effective standardization of the pre-analytical and analytical assay parameters used for each clinical scenario is needed to fully implement NGS into routine practice.