Due to its mechanical and optical properties, echoendoscope advancement may perforate the gastrointestinal wall, particularly at: (1) areas of angulation (e.g., hypopharynx, hiatal hernia, tip of the duodenal bulb, and rectosigmoidal junction); (2) in the presence of unexpected anatomical alterations (e.g., esophageal or duodenal diverticula); and (3) in luminal obstruction (e.g., gastrointestinal cancer). Several investigations of EUS staging have documented non-transversable malignant stenosis in 15% to 42% of patients with esophageal cancer[19-25], and in 7% to 23% (mean 12%) of patients with rectal cancer[26].

Esophageal perforation: In a survey of members of the American Endosonography Club published in 2002, 86 members reported 16 cervical esophageal perforations among 43 852 upper gastrointestinal EUS procedures performed using radial echoendoscopes (0.03%)[27]. Almost all of these patients were elderly; seven of the 16 (44%) had a history of difficult intubation during previous endoscopy. Large cervical osteophytes were identified in three cases. In nine of 16 cases (56%), the endoscopist had less than one year of experience in EUS. Two patients required surgery. For 13 patients (81%), the complication was managed successfully with conservative treatment. However, one patient died[27].

Another early multicenter survey including 34 centers from Europe, Japan and the United States[28] reported 13 esophageal perforations in 37 915 radial EUS examinations (0.03%); esophageal perforations accounted for 68% of all complications. Most of these patients (77%) had undergone pre-EUS dilatation of esophageal strictures to allow complete esophageal cancer staging, including celiac lymph-node station. Nine of these patients underwent surgery, four were managed conservatively, and one patient died. Further major complications in this survey were two pharyngeal and one duodenal perforation and two cases of bleeding[28].

Only two studies have investigated the frequency of esophageal perforations prospectively. A single-center study in the United States enrolled all patients who underwent upper gastrointestinal EUS by the same experienced endosonographer over a 7-year period. Cervical esophageal perforations occurred in three of 4844 patients (0.03%). Longitudinal echoendoscopes had been used in all three patients[29]. The second prospective study, performed in a department of surgical endoscopy in Denmark, evaluated EUS complications of 3324 consecutive patients. Esophageal perforation occurred in 5 of 10 patients with complications (0.15% of all EUS examinations; 0.94% of all patients with esophageal cancer). Balloon dilation had been performed prior to EUS in two cases. All five patients recovered fully with conservative (n = 4) or surgical treatment (n = 1)[30].

From 2004 to 2006, we conducted a survey in German centers performing EUS[31]. Of 67 centers responding to the questionnaire, 32 registered EUS complications prospectively. Esophageal perforation occurred in only eight of 85 084 reported diagnostic EUS procedures (0.009%). None of the perforations were associated with previous dilation of esophageal strictures. Stenosing esophageal cancer was present in five of eight cases[31].

Duodenal perforations: In contrast to the international multicenter survey conducted from 1982 to 1992[28], the German retrospective survey showed that duodenal perforations occurred significantly more often (19 additional cases) compared to esophageal perforations (0.022%)[31]. In 10 of 19 cases (47.4%), duodenal diverticula (n = 4), duodenal stenosis (n = 3), duodenal ulcer (n = 1), duodenal scarring (n = 1), or acute pancreatitis (n = 1) were reported as potentially contributing factors. Twenty-seven of 28 gastrointestinal perforations were managed surgically, and all the patients survived[31] (Table 1). In a prospective EUS online registry of the German Society of Ultrasound in Medicine, participants reported 10 cases of gastrointestinal perforation in 13 988 diagnostic EUS procedures (0.07%). Again, duodenal perforation was the most common type of perforation, accounting for six out of ten cases[32]. The increasing proportion of duodenal perforations in recent multicenter studies compared to the older surveys may partially reflect changing trends in indications for EUS[7,33].

In a large series of 233 EUS-FNA biopsies in patients with presumed pancreatic cancer, Raut and colleagues reported two cases of duodenal perforation requiring surgical intervention (0.86%)[34]. There was no luminal narrowing of the duodenum in either case[34]. One published case report describes iatrogenic duodenal perforation during EUS, which was managed successfully by endoscopic closure using hemoclips, followed by conservative treatment[35]. In a series of 224 EUS-FNAs, one duodenal perforation accounted for one of five severe complications[36]. A large single-center series of 1034 pancreatic EUS-FNAs found one case of fatal duodenal perforation in a 63-year-old woman with an advanced neuroendocrine tumor of the duodenal wall; the perforation likely resulted from mechanical injury of the duodenal wall by the echoendoscope, rather than from biopsy[37].

A national survey in Israel which investigated mortality associated with diagnostic EUS[38] showed that 13 of 18 reported fatal complications (seven in Israel and six from outside the country) resulted from duodenal tears which led to retroperitoneal perforations. Two of the fatalities were secondary to esophageal perforation. At least four of six cases of duodenal perforation reported from Israel involved patients with duodenal diverticula. Five of eight fatal complications in Israel occurred during examinations by endoscopists who had performed fewer than 300 EUS procedures[38].

Other gastrointestinal perforations: EUS-related gastric and rectal perforation seems very rare. There is only one case of rectal perforation reported in the prospective German EUS registry[32]. One study of 2490 endorectal ultrasound examinations reported no procedure-related perforations[39]. One case of gastric perforation occurred in each of the German retrospective study[31] and the German prospective registry[32]. In all three cases of rectal or gastric perforations the indication for EUS was staging of a stenosing tumor[31,32]. A few pharyngeal and hypopharyngeal perforations have been reported as well[28,32]. Possible risk factors for EUS-related gastrointestinal perforations are summarized in Table 2.

For appropriate acoustic coupling of the scanner to the gastrointestinal wall, EUS textbooks suggest instillation of water into the gastric lumen for some indications (e.g., gastrointestinal cancer staging, examination of subepithelial tumors). In a Danish prospective study, tracheal suction was needed in 15 of 293 patients (5.1%), and one patient aspirated during the procedure (0.3%). However, the "water-in-stomach" method was only used in nine patients (3.1%)[30]. Only five cases of aspiration pneumonia have been reported following EUS[32,38,41].

Few data are available concerning the frequency of bacteremia following diagnostic EUS without EUS-FNB. Janssen et al[42] investigated the incidence of bacteremia in 100 consecutive patients undergoing longitudinal diagnostic EUS using paired blood cultures: immediately before and again 5 min after EUS. Significant bacteremia (Streptococcus viridans and Propionibacterium species; Streptococcus mitis) occurred in only two patients (2%), both of whom underwent endosonographic staging of esophageal cancer (seven of nine patients with esophageal cancer did not experience bacteremia)[42]. Levy et al[43] noted transient bacteremia in one of 52 patients who underwent diagnostic EUS with a radial-scanning echoendoscope (1.9%). None of these three patients with bacteremia following diagnostic non-interventional EUS in the two studies developed symptoms of infection[42,43].

Hematogenous tumor cell dissemination has been documented in 11 of 45 patients (24%) following transrectal ultrasound (TRUS)-staging of rectal cancer. In 17 other patients (38%), circulating tumor cells were detected in peripheral blood samples before and after TRUS[44]. The clinical and prognostic significance of this observation is presently unclear, although data from the same group suggest that the detection of circulating tumor cells in blood samples of patients with stage II colorectal cancer correlates with poor outcome[44]. However, a recent retrospective analysis of metastatic rates of locally advanced esophageal cancer showed no difference between those who required and those who did not require esophageal dilation to complete EUS staging[25]. Prospective data are needed to determine whether passage of an endoscope beyond a stenosing gastrointestinal tumor causes clinically relevant hematogenous tumor cell dissemination.

EUS is reported to be safe in pediatric and elderly populations. No complications were reported in five case series of EUS and EUS-guided procedures performed in a total of 166 children[45-49] plus several case reports. A retrospective analysis of EUS and EUS-FNB in 265 consecutive patients aged 80 years or older showed that diagnostic EUS is feasible and safe in geriatric patients[50]. Another study compared complications of endoscopic retrograde cholangiopancreatography (ERCP) and EUS in elderly (≥ 75 years, n = 184) and non-elderly (< 75 years, n = 816) patient cohorts. The EUS complication rate of 4.8% among elderly patients did not statistically differ from the 3.1% complication rate among non-elderly patients[51].

Bacteremia following upper gastrointestinal tract EUS-FNB appears to be rare. Its frequency has been investigated in three prospective studies of 202 patients[42,43,62]. In one study of 100 patients, all blood cultures obtained 30 min and 60 min after the procedure were negative, except in six patients in whom the culture was positive for coagulase-negative Staphylococcus, which the investigators regarded as contaminating and non-pathogenic[62]. Among 102 patients studied by Janssen et al[42] and Levy et al[43], five had bacteremia following EUS-FNB [three with Streptococcus viridans, one with Streptococcus group F (1), and one with Gram-negative bacilli]. No signs or symptoms of infection developed in these five patients. Large case series of EUS-FNB for solid pancreatic mass showed that 0.4% to 1.0% of patients experienced febrile episodes following the procedure[63,64]. In one patient undergoing EUS-FNA of a large pancreatic head mass, pyrexia and abdominal pain emerged followed by extensive acute portal vein thrombosis[65]. In addition to two cases of fever after lymph node EUS-FNA which resolved with antibiotics[54,66], there have been 12 published reports of serious infectious complications following EUS-FNA or EUS-TCB of mediastinal lymph nodes. In one case, a mediastinal abscess developed after EUS-guided aspiration biopsy of a mediastinal lymph-node metastasis from hepatocellular carcinoma (HCC). The iatrogenic abscess was treated successfully by EUS-guided drainage and subsequent endoscopic closure of the resulting esophagomediastinal fistula[67]. Another mediastinal abscess developed following EUS-FNA of a mediastinal metastasis of malignant teratoma; the lesion required computed tomography (CT)-assisted drainage and subsequent thoracotomy[68]. Infectious mediastinitis has also occurred as a result of transesophageal EUS-FNA of an enlarged necrotic lymph node[69] and of a small benign lymph node in the aortopulmonary window[70]. Another case of mediastinitis after EUS-FNB for mediastinal lymphadenopathy developed as a consequence of perforation due to technical problems with the needle[71]. In one patient, EUS-FNA of a tuberculous subcarinal lymph node led to multiple symptomatic mediastinal-esophageal fistulae, which resolved in the course of tuberculostatic treatment[72]. Only one case has been published describing infectious endocarditis (Streptococcus salivarius) secondary to EUS-FNA of a mediastinal lymph node in an oncology patient[73]. Recently, five cases of mediastinal abscess have been reported following EUS-FNA in patients with sarcoidosis; four required surgical treatment[74].

There are six published reports of infection after EUS-FNA or EUS-TCB of subepithelial tumors. In one case, EUS-FNA of an esophageal leiomyoma caused severe intramural infection which resulted in esophagectomy[75]. In another case, fever developed 2 h after EUS-FNA of a large mesenchymal esophageal tumor, presumably due to left lobar pneumonia[41]. In the third case a large duodenal gastrointestinal stromal tumor (GIST) became infected with Enterobacter cloacae following EUS-FNA. The resulting abscess resolved following endoscopic transduodenal drainage and antibiotic treatment[76]. In a series of 52 consecutive EUS-TCBs of gastric subepithelial tumors, two cases of severe septic complications (Streptococcus sepsis and gastric wall abscess) occurred (3.9%)[77]. In our own series of 46 EUS-guided biopsies of hypoechoic subepithelial gastric tumors using a 19 G aspiration-needle, one extramural abscess and septic infection (Morganella morganii) developed in a patient with a large gastric GIST who was receiving 15 mg of prednisolone per day[78]. Despite the high colonization rate of the lower digestive tract lumen with pathogenic bacteria, EUS-FNB of (peri-)rectal and (peri-)colonic solid masses appears relatively safe. Only two cases of infectious complication following transrectal or transcolonic EUS-FNB of solid mass lesions has been published[79,80]. In a prospective risk assessment of bacteremia in 100 patients undergoing EUS-FNB of rectal and perirectal lesions, six patients developed positive blood cultures[81]. However, four of these cases were considered to be contaminants (coagulase-negative Staphylococcus, Moraxella, Peptostreptococcus stomatis). Two patients had true-positive blood cultures (Bacteroides fragilis and Gemella morbillorum). No immediate or delayed infectious complications occurred in any patient[81]. Five other studies (476 patients) showed no complications in patients who underwent transrectal or transcolonic EUS-FNB of solid masses[82-85].

In contrast, EUS-FNB of cystic lesions appears to carry a significant risk of severe infectious complications. In a large multicenter study of EUS-FNA, two of 18 patients undergoing EUS-FNA for cystic pancreatic lesions without peri-procedural antibiotics developed cyst infection[58]. Despite the large number of EUS-guided diagnostic procedures performed for cystic pancreatic lesions, there have been only four reports of infection after EUS-FNB performed with prophylactic antibiotics[52,86,87]. Of the 909 patients who underwent EUS-FNB of pancreatic cystic lesions included in the systematic review of Wang et al[61], 93.7% received prophylactic antibiotics (as recommended by the American Society of Gastrointestinal Endoscopy[88]), and the rate of cyst infection was estimated at 0.55%. However, there is no randomized prospective study supporting the utility of peri-procedural antimicrobial therapy. One recent retrospective study comparing the incidence of infectious complications after 263 EUS-FNBs of cystic pancreatic lesions for 88 cases with and 178 cases without antibiotic prophylaxis showed no significant difference[86]. The authors of one experimental study proposed to implement the combination of systemic antibiotics and local 5% povidone iodine during EUS-FNB of cystic lesions[89].

EUS-FNB of bronchogenic and other mediastinal cysts may cause cyst infection, mediastinitis and severe sepsis. Foregut duplication cysts account for 6% to 15% of primary mediastinal masses[90]. The differentiation between esophageal bronchogenic cysts and subepithelial tumors or paraesophageal masses may be difficult because of the hypoechoic and inhomogeneous mucoid content[91,92]. One study of 19 patients with bronchogenic cysts diagnosed by EUS demonstrated that 13 were anechoic, but the other six were hypoechoic, which suggested, erroneously in these cases, a solid mass lesion. CT or magnetic resonance imaging were diagnostic for a cyst in only four cases[93]. In another study, 9 of 27 benign mediastinal cysts were misdiagnosed as solid masses by CT[94]. Eight cases have been reported of mediastinitis following EUS-FNA or EUS-TCB of mediastinal mass lesions that were hypoechoic on EUS and had high CT attenuation values, but were later shown to be bronchogenic cysts. Because solid lesions were suspected, 5 of these patients did not receive antibiotics[36,92,93,95,96]. In three other cases, severe infection requiring surgical intervention occurred despite pre- and post-interventional intravenous antibiotics[97]. Moreover, one case of asymptomatic contamination of a mediastinal foregut duplication cyst with Candida albicans developed following pre-procedural antibiotic prophylaxis[98]. However, no infectious complications were seen among 22 patients who underwent transesophageal EUS-FNA of suspected mediastinal cysts using peri-interventional antibiotic prophylaxis[91].

In one recently published study on the utility of EUS-guided biopsy of extramural pelvic masses, two patients with cystic pelvic masses developed abscesses (7%)[99]. Also EUS-FNA of ascites and pleural fluid carries a risk of infection. Despite levofloxacin prophylaxis, one of 25 patients undergoing transgastric or transduodenal EUS-FNA of suspected malignant ascites developed bacterial peritonitis. Treatment with intravenous antibiotics was successful[100]. In another study, two of 60 patients (3%) developed self-limited fever following EUS-guided paracentesis[101]. Three other studies (47 patients) did not report any complications following EUS-FNA of ascites or peritoneal nodules in ascites in patients who did not receive prophylactic antibiotics[102-104].

Biliary peritonitis is a severe complication which may develop secondary to penetration or perforation of the common bile duct during EUS-FNB of a pancreatic mass in patients with biliary obstruction[105]. During a study to identify patients with microlithiasis as a cause of idiopathic pancreatitis, EUS-guided gallbladder bile duct aspiration resulted in biliary peritonitis in two of the three patients enrolled; these complications prompted the investigators to discontinue the study[106]. However, a recent systematic review of nine studies (284 patients) showed EUS-FNB of gallbladder masses or bile duct strictures to be safe, with no reported complications[107]. During an international survey of the indications, yield, and safety of EUS-FNB of liver lesions, one death due to septic cholangitis associated with EUS-FNB of a liver metastasis was reported. The patient who died had an occluded biliary stent, with biliary obstruction secondary to pancreatic cancer[108]. Mechanical irritation caused by the tip of the scope or by a water-filled balloon during pancreatic EUS may cause proximal or distal biliary stent migration, possibly followed by cholangitis[109].

Hyperamylasemia was common following pancreatic EUS-FNB in one recent prospective study, occurring in 11 of 100 patients. However, only two patients developed mild acute pancreatitis[110]. Nine large studies have reported iatrogenic acute pancreatitis following EUS-FNB of pancreatic lesions, with rates ranging from 0.19% to 2.0%[37,41,52,53,60,111-113]. A pooled analysis of 4909 solid pancreatic mass EUS-FNBs from 19 US centers identified 14 cases of iatrogenic post-procedural acute pancreatitis (0.29%)[114]. Among them, 71% were mild, 21% moderate, and 7% severe. One patient with significant co-morbidity died secondary to severe procedure-related pancreatitis. The frequency of acute pancreatitis at individual centers ranged from 0% to 2.35%. Only two centers (413 cases of pancreatic EUS-FNB, two cases of acute pancreatitis) assessed EUS-FNB complications prospectively. Some centers with retrospective complication assessment reported no cases of iatrogenic acute pancreatitis, suggesting that acute pancreatitis may have been underreported for the retrospective cohort. Interestingly, the survey showed that half of patients who developed acute pancreatitis after EUS-FNB had benign disease, suggesting that these patients may carry a higher risk. Another possible risk factor for iatrogenic acute pancreatitis following EUS-FNB is a history of acute recurrent pancreatitis[114]. In a German retrospective survey, eight of nine patients (86%) with reported acute pancreatitis after pancreatic mass EUS-FNB had benign disease[31]. In a large retrospective series of EUS-FNB for pancreatic cystic lesions, acute pancreatitis occurred in six of 603 cases (1%)[52]. A retrospective analysis of pancreatic EUS-FNB detected pancreatitis in three of 114 (2.6%) patients with cystic lesions, but none in the 134 patients with solid lesions[41].

Two studies have investigated the value of EUS-FNA and EUS-TCB in the diagnosis of chronic pancreatitis. Mild acute pancreatitis developed in two of 27 patients (7.4%) following EUS-FNA[115] and in one of 16 patients (6%) following EUS-TCB[116].

This rare complication was reported only recently in two cases following EUS-FNA of pancreatic mass lesions. Both cases resolved after antibiotic treatment, transpapillary duct drainage and percutaneous drainage[117,118].

Self-limiting mild intraluminal bleeding due to EUS-FNB has been reported in up to 4% of cases[64] (Figure 2). A prospective study evaluating the risk of extraluminal hemorrhage caused by EUS-FNA found only three cases among 227 patients (1.3%)[119]. Extraluminal bleeding following EUS-FNB can be visualized clearly using EUS. Bleeding appears as an expanding hyperechoic or hypoechoic region adjacent to the sampled lesion[119]. Occasionally, color-coded Doppler may demonstrate blood flow within a “patent needle track” (Figure 3). In a retrospective survey of German EUS centers (13 223 EUS-FNB; Table 3), bleeding was the most common complication of EUS-FNB (0.15% of cases)[31]. Mild and self-limited bleeding was reported in 15 patients, whereas five patients suffered from severe hemorrhages requiring transfusion, and two of those patients required surgical intervention. One patient was treated with low-molecular-weight heparin (LMWH) immediately after the procedure and developed a large mediastinal hematoma and hemothorax; thoracoscopy and transfusion of seven units of packed red cells were required. Another patient required transfusion of four units of packed red cells following EUS-FNB of pancreatic head carcinoma with associated portal hypertension[31]. The prospective German EUS registry also showed that bleeding was the most frequent complication, reported in 1.4% of cases of EUS-FNB. However, 29 of 30 cases were judged to be mild, requiring no endoscopic intervention, surgery, or transfusion[32]. Patients with highly vascularized lesions (e.g., subepithelial mesenchymal tumors, neuroendocrine tumors, and some metastases), as well as those with cystic lesions, may be at greater risk. An Italian multicenter retrospective study of EUS complications showed that three out of seven EUS-FNA complications were intracystic bleeding[59]. Three of 50 patients (6%) in a retrospective study developed acute intracystic hemorrhage following EUS-FNB of pancreatic cystic lesions[120]. Two patients experienced transient abdominal pain, and none of the three patients required blood transfusion. Intracystic bleeding episodes were readily recognized during the investigation as expanding hyperechoic areas surrounding needle puncture sites within the target cysts. Contributing risk factors could not be identified for the bleeding events[120]. In a large Italian single-center experience, bleeding was the most frequent pancreatic EUS-FNA complication (0.96%; n = 1034), with seven of ten cases occurring in cystic lesions[37].

Open in New Tab   Full Size Figure   Download Figure

Figure 2 Intraluminal bleeding after endoscopic ultrasonography-guided fine-needle aspiration. Marked intraluminal bleeding occurred following 19 G endoscopic ultrasonography-guided fine-needle aspiration of a large, hypervascularized gastrointestinal stromal tumor in the stomach. Hemorrhage stopped without intervention.

A prospective analysis compared the risk of bleeding after EUS-FNA or EUS-TCB in patients taking non-steroidal anti-inflammatory drugs (NSAIDs), aspirin (ASA) or prophylactic LMWH to those who did not receive those medications[121]. Two of six patients (33.3%) receiving prophylactic LMWH developed extraluminal bleeding, compared to 0 of 26 patients receiving platelet inhibitors, and 7 of 190 patients (3.7%) not receiving those medications. There appears to be no increased bleeding risk during EUS-FNB for patients taking ASA or NSAIDs, but a possible risk for patients receiving prophylactic LMWH[121]. There are no published data regarding the risk of EUS-FNB in patients treated with clopidogrel and other thienopyridines. However, one study of patients undergoing transbronchial lung biopsy showed that clopidogrel, especially when combined with ASA, greatly increased bleeding risk[122].

There have been several case reports of severe bleeding associated with EUS-FNB, including hemosuccus pancreaticus in cystic pancreatic lesions[123], left adrenal gland hemorrhage[124], retroperitoneal bleeding following EUS-FNA of solid and cystic pancreatic lesions[59,125,126], mesenteric bleeding[127], and intramural bleeding following EUS-TCB of subepithelial tumors[78,96,128]. Three fatal cases of extraluminal bleeding have been reported following pancreatic mass EUS-FNA. One patient developed uncontrolled bleeding from a pseudoaneurysm and died following EUS-FNB performed using a radial-scanning echoendoscope[129]. Another patient died in the consequence of massive gastrointestinal bleeding 6 h after EUS-FNA of a pancreatic cancer[31]. The third fatal outcome was observed after a 19 G EUS-FNA and intracystic brushing of a cystic pancreatic lesion in a patient receiving anticoagulation therapy. The patient died one month later due to a subacute retroperitoneal bleed related to the EUS-guided procedure and further complications[130].

Several studies have described abdominal pain (following EUS-FNB of pancreatic lesions and abdominal lymph nodes) or thoracic pain (following EUS-FNB of mediastinal lymph nodes and mediastinal mass lesions) without clinical, laboratory, or radiological findings suggestive of any specific complication[30,53,60]. One case of massive pneumoperitoneum due to transduodenal EUS-FNB of a solid pancreatic lesion resolved without any intervention other than antibiotics[131]. In the absence of clinical and radiological evidence of bowel perforation, the authors explained that the pneumoperitoneum resulted from insufflated air tracking into the peritoneal cavity through the EUS-FNB site. They postulate that some patients with abdominal pain following EUS-FNB have unrecognized pneumoperitoneum[131].

Pneumothorax following EUS-FNB has been reported as one of three complications in a series of 159 EUS-FNBs for transesophageal biopsy of mediastinal masses[132].

Needle-tract seeding is a rare complication following FNB of intra-abdominal tumors. Based on large retrospective surveys, its frequency is estimated at 0.003% to 0.009%[133]. However, one prospective comparative study reported the incidence of needle-tract implantation of HCC and pancreatic carcinoma after ultrasound-guided percutaneous puncture to be 1.5%[134]. A systematic review and a meta-analysis of percutaneous biopsy of HCC reported tumor seeding frequencies of 2.29%[135] and 2.7%[136], respectively. After percutaneous biopsy of colorectal cancer liver metastases, the risk of needle-track metastases seems even higher, occurring in 10% to 19% of cases[133].

The actual frequency of EUS-FNB-related needle-tract seeding is unknown, but six published case reports have raised concern that the substantial rise of EUS-guided biopsies used to diagnose pancreatic masses and lymphadenopathy may increase the incidence of this complication. In one case, a large gastric wall implantation metastasis with retroperitoneal invasion (50 mm × 30 mm) was diagnosed 16 mo after potentially curative resection of a small (8 mm) T1N0M0 pancreatic tail carcinoma. The tumor had previously been sampled using transgastric EUS-FNA (with five needle passes)[137]. Two similar cases have been reported recently. In both cases, gastric wall metastases occurred at 26 mo and nearly 4 years after transgastric EUS-FNB and pancreatic surgery with curative intent for solid-cystic pancreatic cancer[138,139]. In the fourth case, transgastric EUS-FNB (one needle pass) of a melanoma metastasis to a perigastric lymph node was complicated by a large (30 mm) gastric implantation metastasis along the needle tract. It was detected and resected 6 mo later during surgical resection of lymph node metastasis following neoadjuvant chemotherapy[140]. A recent case report describes needle tract implantation metastasis of the esophageal wall following 19 G EUS-FNA of a large metastatic subcarinal lymph node metastasis of gastric cancer. The esophageal implantation metastasis developed despite perioperative systemic chemotherapy and disappearance of the subcarinal metastatic lymph node; fortunately, it resolved after 2 mo of radiotherapy[141]. The potential risk of peritoneal carcinomatosis due to transgastric EUS-FNA of pancreatic malignancies was highlighted by the case of peritoneal dissemination of an intraductal papillary mucinous neoplasm of the pancreas following EUS-FNA of the tumor[142]. Nevertheless, the risk of peritoneal carcinomatosis appears to be significantly lower with EUS-FNB compared to percutaneous needle biopsy under computed-tomographic or ultrasonographic guidance[143].

Recent data demonstrate vital tumor cell presence within gastrointestinal luminal fluid in 48% of patients with luminal cancers (resulting from tumor cell sloughing), as well as in 10% of patients who underwent EUS-FNB of pancreatic cancer lesions[144]. The importance of this finding with regard to potential tumor cell dissemination by EUS-guided transmural biopsy is speculative at this time. Whereas pancreatic head cancer EUS-FNBs are performed typically through the duodenal wall (which is later resected with the pancreatic head in cases of resectable cancers), pancreatic body and tail tumor EUS-FNBs are performed transgastrically, crossing the peritoneal structures of the Bursa omentalis. Due to the risk of tumor cell seeding along the needle tract, which is proven in principle by the above described case reports, the indication for preoperative EUS-FNB of suspected distal pancreatic cancer is controversial[145-147]. To date, prospective comparative studies of the long-term impact of pre-operative EUS-FNB of suspected pancreatic malignancies are lacking. A recent retrospective study investigated the long-term outcomes of patients undergoing distal pancreatectomy for primary pancreatic neoplasias, comparing patients who received (n = 179) and who did not receive pre-operative EUS-FNB (n = 51)[112]. After a 16-mo median follow-up, no local recurrences were detected within the gastric wall among 91 patients with adenocarcinoma or neuroendocrine tumors who underwent pre-operative EUS-FNB. Median disease-specific survival and recurrence-free survival in patients with pancreatic adenocarcinoma were not statistically worse in 57 patients with pre-operative EUS-FNB as compared with six patients without pre-operative transgastric biopsy. However, partial gastrectomy due to suspected infiltration of the gastric wall by pancreatic adenocarcinoma was performed only in patients with pre-operative EUS-FNB. The study authors expressed concern that surgical pathology showed direct transmural gastric invasion from the pancreatic primary site in six of these seven patients[112].

Scarce information exists regarding the potential impact of pre-operative EUS-FNB on complications of subsequent surgery. One study showed no correlation with overall morbidity, specific complications, or technical difficulties during distal pancreatectomy, when analyzed according to specific pancreatic pathology sub-classifications. However, subgroup analysis revealed a higher incidence of postoperative complications in patients with cystic neoplasms who underwent pre-operative EUS-FNB[112].

A false-positive diagnosis obtained from EUS-FNB may result in inappropriate treatment decisions, which could negatively impact the patient’s clinical outcome. Therefore, we regard false-positive results as a major EUS-FNB complication in this review. The overwhelming majority of studies report EUS-FNB specificity and positive predictive value for detecting cancer at 100%. A thorough review of EUS-FNB studies conducted in 2008 identified only ten studies (1750 cases) plus one case study that altogether reported 27 false-positive diagnoses using EUS-FNB[8]. In most of these cases, benign pancreatic diseases (chronic pancreatitis, intrapancreatic splenic heterotopia, mucinous cystic adenoma) were misdiagnosed as ductal adenocarcinoma or neuroendocrine pancreatic tumor[64,66,145,148-154]. However, few studies have correlated EUS-FNB findings with surgical pathology[149,150,155-158]. Therefore, the “true false-positive rate” of EUS-FNB was unknown until a remarkable 2010 study presented a more realistic picture[159]. A cohort of 337 consecutive patients with positive or suspicious EUS-FNB results underwent surgery in the absence of neoadjuvant therapy; there were 19 false-positive cases (5.6%) (based on “positive” cytopathological interpretations) and 25 false-positive cases (7.3%) (based on “positive” plus “suspicious” cytopathological findings), respectively. For pancreatic lesions, the authors reported a 2.2% rate of discordance between pre-operative cytological diagnosis and subsequent surgical pathology. In this study, most discordant cases resulted from malignant cell contamination by luminal cancers (primarily esophageal) when performing EUS-FNB of nearby lymph nodes. Retrospectively, only 11 cases were identified as resulting from cytopathological interpretive error[159]. These results are supported by a retrospective cohort study of 367 patients who underwent surgical resection without previous neoadjuvant treatment[160]. This study showed a 1.1% false-positive rate of EUS-FNB for solid pancreatic lesions. When suspicious cytological findings were considered, the false-positive rate was estimated at 3.8%. On review of discordant cases, almost all were caused by pathological misinterpretation in the setting of chronic pancreatitis[160]. Another possible explanation for false-positive diagnosis by EUS-FNB may be inadvertent aspiration of contaminated luminal fluid or traversal of mucosal high-grade dysplasia or malignant infiltration of the gastrointestinal wall[144,161]. Malignant cells are identified within gastrointestinal luminal fluid not only in 48% of cases of luminal cancer, but also in 10% of cases of extraluminal (pancreatic) cancer undergoing EUS-FNB[144].

Due to its high accuracy and lower complication rate compared to ERCP for suspected pancreaticobiliary disease (especially for obstructive jaundice), EUS and EUS-FNB should be performed first when possible. Moreover, biliary stent placement prior to EUS is believed to negatively affect the accuracy of endosonographic staging of pancreatic head malignancies[162]. To perform EUS, EUS-FNB and therapeutic ERCP in one session has advantages with regard to cost, total procedure time and workflow. Several studies have shown that single-session EUS and ERCP does not result in higher complication rates[163-168]. However, ERCP is more widely available than EUS and some cases (e.g., obstructive cholangitis) require biliary decompression more urgently than cytological diagnosis. Therefore, biliary stenting often precedes EUS-FNB. A recent retrospective comparative study showed that pre-EUS stenting of biliary obstructions due to pancreatic cancer did not increase the EUS-FNB complication rate. Moreover, the accuracy of tissue diagnosis was not impaired if ERCP and endoscopic stenting were performed more than one day prior to EUS-FNB[169].

Due to the relatively low number of complications associated with EUS-FNB, available studies are statistically underpowered to adequately address the question of risk factors. Endosonographer experience seems to be important. Of five nonfatal complications in a large multicenter EUS-FNA study (n = 457), none occurred during the last 12 mo of the study[58]. Similarly, the largest EUS-TCB study (n = 247) showed that all complications were observed among the first 100 patients[56].

There is no evidence that needle size affects complication rate. Only one retrospective multicenter study reported a complication rate of 2% in 540 patients undergoing 22 G EUS-FNA of solid pancreatic masses, versus 0% in 302 patients undergoing 25 G EUS-FNA[113]. Small prospective studies comparing needle sizes reported no complications in any group[170,171]. EUS-FNB using a 19 G needle, when combined with cytobrushing of pancreatic cystic lesions, conveys a significant complication risk associated primarily with cytobrush trauma to the cyst wall. In a prospective controlled study of 37 patients with 39 cystic pancreatic lesions, seven experienced complications (19%): one experienced severe bleeding and one experienced severe pancreatitis[172]. In another study using the same technique (n = 30), three patients experienced complications, including one case of fatal bleeding[130]. This procedure-related morbidity is considerably higher than the 2.2% to 5.2% complication rate reported in several EUS-FNB studies of cystic pancreatic lesions[54,86,87,173]. Nevertheless, the overall EUS-FNB complication rate seems higher for cystic than solid pancreatic masses[54,58,61]. In a systematic review which included 7337 EUS-FNBs of solid pancreatic masses and 909 EUS-FNBs of pancreatic cystic lesions, complications were reported in 60 patients with solid masses (0.82%; only prospective studies: 2.44%) and 25 patients (2.75%; only prospective studies: 5.07%) with pancreatic cystic lesions[61]. Data from the same systematic review confirmed that EUS-FNB is exceptionally safe for mediastinal lesions (n = 1310; complication rate: 0.38%), abdominal masses (n = 381; complication rate: 0.26%) and adrenal gland biopsies (n = 81; complication rate: 0%). EUS-FNB had an associated overall morbidity of 1.03% for pancreatic lesions (n = 8246; only prospective studies: 2.64%), 2.33% for liver lesions (n = 344), 2.07% for perirectal lesions (n = 193), and 3.53% for ascites (n = 85)[61].

Safe and effective EUS and EUS-FNB performance require adequate sedation. Only a few studies report on the frequency of sedation-related complications in EUS and EUS-FNB. In the prospective German EUS registry, oxygen desaturation (SpO2 < 80%) was reported in 0.39% of patients, but most patients (94.4%) did not require intubation or mechanical ventilation[32]. A retrospective single-center study on the safety of nurse-administered propofol sedation for upper gastrointestinal EUS showed that 6 of 806 patients (0.7%) experienced declines in oxygen saturation (SpO2 < 90%), and four patients (0.5%) required assisted positive pressure ventilation[17]. Comparable results have been reported by other authors[174]. Several studies show that EUS with deep sedation using propofol (administered by an anesthesiologist, by a second gastroenterologist, or by a trained nurse) is as safe as conscious sedation using benzodiazepines (midazolam) and pethidine[174-177].