- Top of page
- MATERIALS AND METHODS
- FUNDING SUPPORT
- CONFLICT OF INTEREST DISCLOSURES
Pancreatic cancer is the fourth leading cause of cancer death in the United States. The very high mortality rate is the result of late presentation of most patients, a high morbidity associated with pancreatic surgical resection, and a lack of other effective treatment options. The majority of pancreatic masses are ductal adenocarcinomas. However, nonneoplastic entities including primarily chronic pancreatitis and its sequelae, especially autoimmune pancreatitis, can demonstrate overlap with pancreatic neoplasms with regard to both imaging characteristics and pathologic diagnosis, creating a diagnostic challenge from both a clinical and pathologic standpoint. Cystic neoplasms also arise in the pancreas including mucinous cystic neoplasm (MCN) and intraductal papillary mucinous neoplasm (IPMN) with or without associated invasive carcinoma, serous cystadenoma, and solid pseudopapillary neoplasm. These cystic neoplasms, along with nonneoplastic entities, carry various prognoses and also have different associated treatment options, thereby making the correct diagnosis extremely important for both prognosis and management decisions.
Endoscopic ultrasound (EUS) has become one of the preferred imaging techniques in evaluating pancreatic lesions and has high sensitivity as well as the ability to perform staging and tissue sampling at the time of the procedure.[3-5] In fact, primary diagnosis of the majority of pancreatic tumors is now based on EUS-guided fine-needle aspiration (EUS-FNA) biopsy and bile duct brushing. Cytologic diagnosis has reduced patient morbidity and needle tract seeding compared with computed tomography (CT) or ultrasound-guided core needle biopsy. In addition, EUS-FNA had demonstrated high diagnostic accuracy for the majority of pancreatic solid lesions[6-8] and has added value to imaging alone for cystic lesions. A generally accepted false-positive rate for an EUS-FNA diagnosis of gastrointestinal lesions is 0% to 1%.[8, 10-15] These studies predominantly used both clinical follow-up information and surgical resection as the gold standard, with too few resected specimens for confirmation and without regard to the use of neoadjuvant therapy.[12, 14] However, this has been recently challenged by 2 studies.[12, 15] Using the surgical resection as a gold standard and including cases with a diagnosis of “suspicious for malignancy” in the positive group, the false-positive rate was 2.2% and 3.8%, respectively.[12, 15]
The diagnosis of malignancy in cystic lesions is often complicated because of sampling difficulty and reactive changes associated with significant inflammation and atypia. In addition, dysplastic lesions can arise in the pancreatic ductal system or in an underlying mucinous neoplasm and can be confused with malignancy depending on the degree of cytologic atypia.[6, 13, 16, 17] Both false-positive and false-negative diagnoses lead to significant consequences for clinical management because a positive diagnosis often leads to major surgery (pancreatectomy) or radiotherapy/chemotherapy and a false-negative diagnosis delays patient treatment and may result in the loss of surgical resectability.
Cytologic diagnosis with EUS-FNA has been studied extensively, with widely varying sensitivity.[19-24] The sensitivity has been reported to range from 14% to 100%, whereas specificity has ranged from 69% to 100%.[13, 21, 23, 25, 26] However, as mentioned, only rare studies have used the surgical resection as the gold standard to evaluate the cytopathologic diagnostic error and clinical sampling error separately for the diagnosis of pancreatic cystic and solid lesions. In the current study, we reviewed our pancreatic cytology cases with matched surgical specimens for a period of 6 years. The objective was to evaluate both cytopathologic diagnosis and clinical sampling error for both solid and cystic pancreatic lesions with EUS-FNA.
- Top of page
- MATERIALS AND METHODS
- FUNDING SUPPORT
- CONFLICT OF INTEREST DISCLOSURES
Using the surgical resection as the gold standard for diagnosis, we systematically studied the sensitivity, specificity, positive predictive value, negative predictive value, false-positive and false-negative rates, and accuracy of EUS-FNA diagnosis of pancreatic solid and/or cystic lesions performed at the University of Rochester from 2004 through 2010. We have confirmed previous studies that indicated that pancreatic EUS-FNA has a very low false-positive rate (0%) (Table 5). The false-negative rate was relatively higher, with the majority of false-negative results found to be due to SE, but some were the result of IE (IE rate in group 2: 3% vs SE rate in group 3: 23%). The false-negative rate, sensitivity, and accuracy were significantly different when considering IE alone (group 2) compared with SE alone (group 3). In addition, we also confirmed that the frequency of SE (33%) in EUS-FNA of cystic lesions (group 4) was significantly higher than that in solid lesions (12%; Group 5). The sensitivity (47%) and accuracy (67%) for cystic lesions (group 4) were significantly lower than those in solid lesions (85% and 87%, respectively; group 5). The results of the current study demonstrate that a false-negative diagnosis is the primary concern in pancreatic EUS-FNA procedures.
As shown in previous studies, EUS has become one of the preferred imaging techniques in evaluating pancreatic solid lesions and has high sensitivity and specificity with a very low false-positive rate.[3-5] The majority of previous studies used clinical follow-up and radiologic imaging combined with surgical resection as the gold standard. Recently, Gleeson et al used direct surgical resection after EUS-FNA as a gold standard to study the false-positive rate of EUS-FNA. They found that the false-positive and false-suspicious cases represented 7.2% of gastrointestinal EUS-FNA cases and the incidence of discordance was higher in nonpancreatic FNA cases (22 of 147 cases; 15%) than pancreatic FNA cases (5 of 230 cases; 2.2%). This study tried to correct the broadly accepted false-positive rate for EUS-FNA of approximately 0% to 1%. Siddiqui et al further showed that the false-positive rate of EUS-FNA diagnoses for solid pancreatic lesions was 3.8% (14 of 367 cases) when both suspicious and positive cytology findings were interpreted as malignant. Using the surgical resection as the gold standard for diagnosis, we found no false-positive results in the current study (none of 101 cases). This result agrees with the previously accepted false-positive rate of 0% to 1%. However, we noted that our pancreatic EUS-FNA cases had a higher false-negative rate 25% (95% CI, 16%-36%) in both solid and cystic lesions. The current study has further investigated the false-negative rate of EUS-FNA.
Studies of false-negative rates of pancreatic EUS-FNA are challenged by the lack of clinical follow-up and a way to prove negative EUS-FNA cases. In the current study, many benign and malignant pancreatic surgical resection cases were identified. This gave us the opportunity to use the surgical resection diagnosis as a gold standard to explore the false-negative rate and other parameters. However, limitations remain because many cases performed during the studied time period did not undergo resections. Based on this gold standard, the false-negative rate among solid lesions is 15% (group 5), which is lower than the 35% rate (range, 8%-84%) calculated from the previous meta-analysis of pancreatic solid lesions.[5, 8, 27, 28] To further study the causes of false-negative results at the study institution, the false-negative rate was calculated in 2 groups. In group 2, only those cases with an IE were used for analysis; in group 3, only cases with an SE were evaluated. We found that the false-negative rate (3%) caused by IE was significantly lower than that caused by SE (23%) (P = .003). This suggests that SE, rather than IE, is a major cause of higher false-negative findings. Because we included both solid and cystic lesions, we further examined the false-negative rate in these 2 groups separately (solid lesions in group 5 and cystic lesions in group 4). The false-negative rate among cystic lesions was significantly higher than that in solid lesions (53% vs 15%; P = .005). However, as mentioned earlier, there were 403 total negative pancreatic EUS-FNA cases during this time period. Only 19 cases were proven to be false-negative cases and 20 cases were found to be true-negative findings by comparing the surgical resection diagnosis. The true diagnoses of the remaining cases were unknown because the majority of negative cases do not go on to surgical resection. We attempted to find follow-up information for these negative EUS-FNA cases. Unfortunately, in most cases, no clinical follow-up information was available, which made the accurate examination of the true false-negative rate impossible. However, the real frequency of false-negative findings has to be within a maximum of 49% (19 of 39 cases) and minimum of 5% (19 of 403 cases). Although all cases with matched pancreatic EUS-FNA specimens and surgical resections were included in the current study, we can argue that the data still represent the relative ratio of IE and SE. Cytologic IE is likely less common because of accepted stringent criteria to make a positive cytology diagnosis.
The sensitivity, specificity, and accuracy of pancreatic EUS-FNA were also analyzed in various subgroups. The sensitivity (47%) and accuracy (67%) for cystic lesions (group 4) were significantly lower than those for solid lesions (group 5; 85% and 87%, respectively). A preoperative diagnosis of pancreatic cystic lesions by EUS-FNA remains a challenge. Multiple clinical and laboratory methods are usually used to improve the sensitivity and accuracy. Cystic fluid sent for amylase and carcinoembryonic antigen (CEA) measurement has shown variable and inconsistent accuracy rates.[8, 20, 25] Performing CEA levels on pancreatic cysts has demonstrated accuracy rates ranging from 42.9% to 86%.[8, 24, 29, 30] The cytologic diagnosis of cystic lesions with EUS-FNA has also been studied extensively with widely variable sensitivity reported.[13, 19-24] The sensitivity has been reported to range from 23% to 100% and specificity has been reported to range from 71% to 100%.[8, 13, 23] One meta-analysis by Thosani et al showed that the pooled sensitivity and specificity in diagnosing mucinous cystic lesions were 63% and 88%, respectively, in 11 studies and 54% and 92%, respectively, in 4 prospective studies. The sensitivity in the current study (47%) was lower than these meta-analysis results because we restricted our cases to those with subsequent surgical resections. Brugge et al concluded that a CEA level from pancreatic mucinous cysts was more accurate than the cytology diagnosis (86% vs 58%), but less accurate in the setting of a malignant cyst (62% vs 75%). Recently, Rogart et al reported that cyst wall puncture and FNA improved the diagnostic yield for mucinous cysts. In addition, cytologic classification with high-grade epithelial atypia in cystic lesion FNA specimens demonstrated a higher prediction for malignancy and added value for the clinical evaluation of cystic lesions.[9, 32] However, the overall sensitivity of the EUS-FNA cytologic diagnosis of pancreatic cystic lesions is still low. More sensitive and specific techniques are needed and should be developed as new technologies emerge, such as molecular tests and confocal laser endomicroscopy.
Confocal laser endomicroscopy is a novel imaging technology in which a low-power laser illuminates and scans a single focal plane of the tissue.[33-35] This technique allows for the detection of the microscopic detail of the surface epithelium in pancreatic cysts. Further studies and development of this new technique may facilitate sampling the most suspicious area of a cyst.
Recently, molecular tests were reported that aid in the detection of malignant pancreatic cystic lesions.[36-41] In 1 small case study, the sensitivity, specificity, and positive predictive value of molecular diagnosis were 83%, 100%, and 100%, respectively, for a malignant cyst and 86%, 93%, and 95%, respectively, for a benign mucinous cyst. Cystic fluids from 19 patients were studied for genetic mutations in 169 genes commonly altered in human cancers. In addition, a larger number (113 cases) of additional IPMNs were then analyzed to determine the prevalence of KRAS and GNAS mutations. In total, GNAS mutations were present in 66% of IPMNs and either KRAS or GNAS mutations could be identified in 96%. It is interesting to note that all 44 serous cystadenomas were GNAS and KRAS wild-type. GNAS/KRAS mutations reliably distinguished IPMNs from serous cystadenomas with a sensitivity of 96% and a specificity of 100%. From the review, KRAS had a low sensitivity (range, 11%-57%) and a high specificity (range, 93%-100%) for diagnosing mucinous cysts. The specificity (range, 71%-93%) and sensitivity (range, 20%-53%) decreased further when KRAS mutations were used for the determination of malignancy. Emerging molecular tests, including next-generation sequencing and droplet polymerase chain reaction, will hopefully shed light on improving the sensitivity, specificity, and false-negative rate.
In the current study, all analyses were based on matched cytology and surgical resection data. We found that a pancreatic EUS-FNA cytologic diagnosis had a high specificity and accuracy for solid lesions and lower sensitivity, specificity, and accuracy for cystic lesions. The major cause of a false-negative result on EUS-FNA was due to SE than IE. We suggest that combined cytologic diagnosis with new sensitive molecular tests, CEA measurement, or advanced confocal laser endomicroscopy could dramatically improve the false-negative rate for cystic lesions in the future.