Pancreatic cystic lesions are being detected with increasing ORIGINAL ARTICLES

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2005;3: ORIGINAL ARTICLES The Role of Pancreatic Cyst Fluid Molecular Analysis in Predicting Cyst Pathology ASIF KHALID,*, KEVIN M. MCGRATH, MALIHA ZAHID, MATT WILSON, DEBRA BRODY, PATRICIA SWALSKY, ARTHUR J. MOSER, KENNETH K. LEE, ADAM SLIVKA, DAVID C. WHITCOMB, and SYDNEY FINKELSTEIN *VA Pittsburgh Health Care System; Division of Gastroenterology and Hepatology, Department of Medicine, and Department of Surgery, University of Pittsburgh; and Redpath Integrated Pathology, Pittsburgh, Pennsylvania Background & Aims: Current methods to detect malignancy in mucinous cystic neoplasms of the pancreas remain inadequate. The role of detailed molecular analysis in this context was investigated. Methods: Endoscopic ultrasound guided pancreatic cyst aspirates were prospectively collected during a period of 19 months and studied for cytology, carcinoembryonic antigen level, and molecular analysis. Molecular evaluation incorporated DNA quantification (amount and quality), k-ras point mutation, and broad panel tumor suppressor linked microsatellite marker allelic loss analysis by using fluorescent capillary electrophoresis. The sequence of mutation acquisition was also calculated on the basis of a clonal expansion model, and comparison was made to the final pathology. Results: Thirty-six cysts with confirmed histology were analyzed. There were 11 malignant, 15 premalignant, and 10 benign cysts. Malignant cysts could be differentiated from premalignant cysts on the basis of fluid carcinoembryonic antigen level (P.034), DNA quality (P.009), number of mutations (P.002), and on the sequence of mutations acquired (P <.001). Early k-ras mutation followed by allelic loss was the most predictive of a malignant cyst (sensitivity, 91%; specificity, 93%). Conclusions: Malignant cyst fluid contains adequate DNA to allow mutational analysis. A first hit k-ras mutation followed by allelic loss is most predictive of the presence of malignancy in a pancreatic cyst. This approach should serve as an ancillary tool to the conventional work-up of pancreatic cysts. Cumulative amount and timing of detectable mutational damage can assist in diagnosis and clinical management. Pancreatic cystic lesions are being detected with increasing frequency because of the widespread use of high quality imaging tests. These might be inflammatory (pseudocysts), benign (serous cystadenoma), premalignant (mucinous cysts), and malignant (cystadenocarcinoma). The majority of pancreatic cysts detected today are mucinous cystic neoplasms (MCNs). 1 MCNs are considered premalignant and encompass intraductal papillary mucinous neoplasia (IPMN) and mucinous cystadenomas. Resection of MCNs is recommended, given the concern for malignant degeneration. The natural history of MCN, however, remains unknown, and the frequency and timing of malignant change are unclear. Current methods to evaluate pancreatic cysts use imaging and cyst aspirate analysis. In the absence of an associated mass, there are no reliable radiologic criteria distinguishing benign and premalignant from malignant cysts. 2 4 The sensitivity of cytologic analysis of a pancreatic cyst aspirate is suboptimal because of the acellular nature of the specimen. 4 Cyst fluid CEA level is considered the most reliable indicator of a cyst of mucinous origin; however, it cannot predict the presence or absence of cancer. 4 Better tools to assess the presence of malignancy are desirable. This is especially important in patients not deemed optimal candidates for surgery. The molecular pathogenesis underlying the progression of normal pancreatic duct cells to pancreatic cancer is characterized by the accumulation of genetic mutations, gene silencing, and chromosomal deletions Although the cellular content of pancreatic cyst aspirates is frequently suboptimal, polymerase chain reaction (PCR) amplification of DNA from whole or lysed cells shed into the fluid from the cyst lining might be predictive of the cyst pathology. A high level of accumu- Abbreviations used in this paper: CEA, carcinoembryonic antigen; CI, confidence interval; CT, cycle threshold; EUS, endoscopic ultrasound; FNA, fine-needle aspiration; IPMN, intraductal papillary mucinous neoplasia; LOH, loss of heterozygosity; MCN, mucinous cystic neoplasm; OD, optical density; PanIN, pancreatic intraductal neoplasia; PCR, polymerase chain reaction; SD, standard deviation by the American Gastroenterological Association /05/$30.00 PII: /S (05)00409-X

2 968 KHALID ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 3, No. 10 Table 1. k-ras-2 Gene and Tumor Suppressor Genes (With Associated Markers) With Chromosomal Location and Mutation Type Proximity cancer gene a Mutation type Locus b Marker 1 Marker 2 k-ras Point mutation c 12p12 CMM/RIZ Allelic imbalance d 1p36 1p34 D1S407 MYCL VHL Allelic imbalance d 3p26 3p25 D3S1539 D3S2303 APC Allelic imbalance d 5q23 5q23 D5S592 D5S615 P16 Allelic imbalance d 9p21 9p23 D9S251 D9S254 PTCH e Allelic imbalance d 9q22 D9S252 PTEN Allelic imbalance d 10q23 10q23 D10S520 D10S1173 P53 Allelic imbalance d 17p13 17p13 D17S974 D17S1289 a The proximity cancer gene is listed here; however, the analysis does not assert that this particular gene is affected. Although it might be altered, other proximity known and as yet unreported genes might be causally related to neoplastic progression. b Locus defined according to Southampton database c k-ras-2 point mutation is limited to those affecting the first exon of the gene. d Methods used here report allelic imbalance, which in the case of tumor suppressor genes most likely reflects genomic deletion. e Only one marker is used for 9q22 in proximity to PTCH. lated mutational damage would reflect an underlying malignancy, and similar alterations would not be seen in benign cysts. Because cyst epithelial cells with the highest turnover would contribute relatively more DNA, malignant cyst fluid would likely be enriched with DNA from the most malignant cells. Thus analysis of pancreatic cyst fluid DNA might be capable of detecting malignancy, even if such an alteration affects a relatively small portion of the pancreatic cystic lining. We previously demonstrated that detection of loss of heterozygosity (LOH) by using microsatellite markers closely linked to key tumor suppressor genes can serve as a surrogate marker for gene inactivation and mutation. 19 The purpose of the current study was to evaluate endoscopic ultrasound (EUS) guided pancreatic cyst aspirates by using a panel of LOH markers together with k-ras exon-1 activation mutation detection to predict the underlying biologic behavior as compared with the corresponding surgical histology. Materials and Methods One hundred sixteen patients with pancreatic cysts underwent EUS fine-needle aspiration (FNA) between November 2002 and May Of these, 31 patients underwent surgery, and surgical pathology was available for study. An additional 5 patients reached a final diagnosis of a malignancy based on cytology and were also included. The study was approved by the Institutional Review Board at the University of Pittsburgh Medical Center. EUS was performed with a curvilinear echo-endoscope (FG-36UA; Pentax Precision Instrument Corp, Montvale, NJ). FNA was performed with a 25-, 22-, or 19-gauge needle (EchoTip; Wilson Cook Medical, Winston-Salem, NC). Intravenous antibiotics (levofloxacin, 500 mg) were administered at the time of cyst aspiration. The aspirated fluid underwent cytologic examination, measurement of carcinoembryonic antigen (CEA) level, and molecular analysis. On the basis of the surgical pathology, the cysts were classified as (1) benign (including pseudocysts and cysts with no malignant potential), (2) premalignant (adenoma and borderline histology according to World Health Organization classification), and (3) malignant (carcinoma with or without invasion according to World Health Organization classification). Cyst aspirate cytology evaluation was performed in routine fashion. Cytopathologic criteria for malignancy included nuclear enlargement, pleomorphism (minimum of 3- to 4-fold variation in nuclear size), elevated nuclear/cytoplasm ratio, nuclear membrane irregularity, and coarse chromatin. 19 Cases diagnosed as inconclusive fulfilled some but not all criteria for malignancy. CEA level was measured according to routine laboratory protocol by using serum curves (Advia Centaur Immunoassay System; Bayer Healthcare LLC, West Haven, CT). Molecular analysis was performed as follows. DNA was extracted from 200 L of cyst fluid by column separation according to manufacturer s directions (Qiagen kit; Qiagen, Valencia, CA). The extracted DNA was resuspended in 50 L of dilute Tris buffer (ph 7.0). The concentration of DNA was obtained according to optical density (OD) at 260/280 wavelength to document quantity and purity of extraction. Onemicroliter aliquots were removed for PCR amplification of individual microsatellite markers and direct sequencing of the first exon of the k-ras-2 gene. Nucleic acid amplification was carried out according to manufacturer s instructions (Gene- Amp kit; Applied Biosystems, Foster City, CA). Fluorescentlabeled oligonucleotide primers were used for quantitative determination of allelic imbalance on the basis of the peak height ratio of polymorphic microsatellite alleles. The genomic position of the microsatellite loci and putative associated tumor suppressor genes are described in Table 1. The microsatellite marker D17S1289 was used in quantitative PCR reactions to assess the amount of amplifiable DNA from each specimen. On the basis of the OD measurement, each sample was normalized to 5 ng/ L and then amplified on a quantitative thermocycler system (Icycler; Bio-Rad Laborato-

3 October 2005 PANCREATIC CYST FLUID MOLECULAR ANALYSIS AND CYST PATHOLOGY 969 Figure 1. The pancreatic cyst DNA shows a point mutation (12 valine substitution for glycine) at codon 12. Fluorescent sequencing has been performed by using a downstream primer for the anti-sense strand of amplicon DNA. Both the normal sequence (GGT) and the mutant sequence (GTT) are represented (arrow), in keeping with hemizygous sequence mutation. ries, Hercules, CA), and the cycle threshold (CT) value was noted. This provided a measure of the amount of amplifiable DNA in each sample. Because the samples had been normalized to a set level, qpcr CT values provided a measure of DNA integrity reflected by the degree of amplifiability. All remaining amplifications were performed in standard thermocyclers (Promega USA, Madison, WI) by using standard cycle profiles optimized for individual markers. Post-amplification products were electrophoresed, and relative fluorescence was determined for individual alleles (GeneScan ABI3100; Applied Biosystems). The ratio of peaks was calculated by dividing the value for the shorter sized allele by that of the longer sized allele. Thresholds for significant allelic imbalance were determined by using normal (non-neoplastic) specimens for every marker used in the panel. Peak height ratios falling outside of 2 standard deviations (SDs) beyond the mean for each polymorphic allele pairing were assessed as showing significant allelic imbalance. In each case, a buccal brush or alternative source of non-neoplastic DNA was available to establish informativeness status and then to determine the exact pattern of polymorphic marker alleles. Having established significant allelic imbalance, it was then possible to calculate the proportion of cellular DNA that was subject to hemizygous loss. For example, a polymorphic marker pairing whose peak height ratio was ideally 1.00 with an SD of.23 in normal tissue could be inferred to have 50% of its cellular content affected by hemizygous loss if the peak height ratio was.5 or 2.0, as previously described. 19 This requires that a minimum of 50% of the DNA in a given sample is derived from cells possessing deletion of the specific microsatellite marker. The deviation from ideal normal ratio of 1.0 indicated which specific allele was affected. In a similar fashion, allele ratios below.5 or above 2.0 could be mathematically correlated with the proportion of cells affected by genomic loss. The order of mutation acquisition was then calculated on the basis of a clonal expansion model for carcinogenesis in which acquired mutations are causally associated with phenotypic overgrowth of precursor cells. For example, 2 equivalent aliquots of pancreatic cyst fluid DNA showing 90% and 50% mutated DNA, respectively, for 2 specific markers could then be arranged in a timeline in which the 90% mutation preceded the 50% mutation. Assuming a model of allelic loss with minimal non-neoplastic cell DNA inclusion, the percentage of mutated DNA was determined for each marker. Determination of k-ras-2 point mutation was accomplished by fluorescent-based direct sequencing of the amplified first exon of the gene as previously reported. 19 Involvement of 100% of cells was considered to be present when the fluorescent peak height of the mutated base on sequencing was equal to or greater than the normal sequence base (Figure 1). Because k-ras-2 point mutation needs to only affect one allele, 100% microdissected cell involvement was considered to be present when normal and mutant peak heights were equal. Additional experiments were performed to validate our approach of calculating the mutation acquisition pattern from the ratio of allele peak heights on electropherograms comparing with microdissection-based genotyping of multiple sites within a tumor (results not shown). Prior studies 20,21 have shown that mutations concordant in different topographic sites of a tumor are expected to have occurred earlier in the process of carcinogenesis than those that are discordant. Statistical Analysis Continuous variables are presented as mean SD. Differences across groups were compared by using a one-way analysis of variance. Multiple comparisons were corrected for by using the post hoc Bonferroni test. In cases in which the variance across groups was not homogenous, as assessed by the Levene statistic, data were analyzed by using a nonparametric analysis of variance or the Kruskal Wallis test. If this was significant, further limited, hypothesis driven two-group nonparametric comparisons were undertaken by using the Mann

4 970 KHALID ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 3, No. 10 Table 2. Patient Demographics, Clinical Characteristics, and EUS Features by Pathologic Categories Are Presented Mean age (y) Sex Pain without pancreatitis Weight loss Pancreatitis Mean cyst size (cm) Cyst morphology Malignant 69 Male, 5; female, 6 7/11 6/11 0/ Simple, 3; complex, 8 Pre-malignant 63 Male, 6; female, 9 6/15 0/15 1/ Simple, 6; complex, 9 Benign 43 Male, 4; female, 6 1/10 0/10 6/ Simple, 7; complex, 3 Whitney U test. A two-tailed P value.05 was considered significant. Sensitivity and specificity were calculated as appropriate. Data were analyzed by using the statistical package SPSS version 12.0 for Windows (SPSS Inc, Chicago, IL). Results Thirty-six patients with pancreatic cysts were eligible for analysis. This was based on the presence of final surgical pathology (31 cases) or cytologically proven cancer from the FNA sample (5 cases). All benign and premalignant diagnoses were based on the surgical pathology. Eleven cystic lesions were malignant (8 invasive cancer, 3 carcinoma in situ), 15 were premalignant (9 borderline, 6 no dysplasia), and 10 cystic lesions were benign (6 pseudocysts, 1 lymphoepithelial cyst, 1 mesothelial cyst, 1 retention cyst, and 1 oligocystic serous cyst adenoma). Patient demographics, clinical presentation, and cyst morphology are discussed in Table 2. Six of the malignant cysts were diagnosed as a result of the presence of unequivocally malignant cells on cytologic evaluation of FNA of an associated solid component in the cyst. Five of these patients were not deemed surgical candidates, and therefore pathologic confirmation is not available. Three cysts with invasive cancer and 3 with carcinoma in situ underwent surgery (4 IPMNs, 2 mucinous cystadenocarcinomas). Of these only 1 cyst was diagnosed as malignant on the basis of cytology. There were 10 IPMNs and 5 mucinous cystadenomas in the premalignant category. FNA cytology was diagnostic in only 1 case; 2 premalignant cysts had inconclusive cytology, and the others (12 cysts) had negative cytology. Of the 10 benign cysts, 1 had inconclusive cytology, and the remaining 9 had negative cytology. Cyst fluid CEA level was available in 27 cases (9 malignant, 10 premalignant, and 8 benign cysts). CEA analysis could not be performed in 9 (25%) cases because of insufficient fluid to run the test. Mean CEA levels for the benign, premalignant, and malignant groups were 13 ng/ml (SD 19) [95% confidence interval (CI), 3 30], 5237 ng/ml (SD 10,494) [95% CI, ,744], and 108,360 ng/ml (SD 251,860) [95% CI, 85, ,957], respectively. This difference was significant between the benign and premalignant groups (P.001) and the premalignant and malignant groups (P.05). Extreme values, however, were encountered in the premalignant and malignant categories. DNA analysis included calculation of the total amount of DNA present in the cyst fluid (OD) and its quality (amplifiable DNA reflected by the CT). The OD was available on 35 cyst aspirates. Mean OD for the 3 groups was as follows: benign, 6.5 (SD 5.9) [95% CI, ]; premalignant, 3.7 (SD 2.6) [95% CI, ]; and malignant group, 16.5 (SD 15.7) [95% CI, ]. The CT value was available for 35 cysts. Of these, 5 cysts (3 premalignant and 2 benign) had no amplifiable DNA onto 40 cycles on qpcr and hence were assumed to be 40 for sake of analysis. One malignant sample did not undergo qpcr. These results are as follows: benign, 33 (SD 4.9) [95% CI, ]; premalignant, 31.2 (SD 5.0) [95% CI, ]; and malignant group, 24.5 (SD 4.5) [95% CI, ]. The difference between the premalignant and malignant groups was significant for the OD (Bonferroni P.008) and the CT values (Bonferroni P.007). Receiver operator area under the curves for a CT value of 27 (optimal cutoff) was.80 for the presence of malignancy in an MCN (not shown). Mutational analysis (LOH and k-ras point mutation) was available on all cyst aspirates. No mutations were detected in any of the 10 cysts in the benign group. Six of 15 cysts in the premalignant group carried mutations. The mean number of mutations in the premalignant group was.9 (SD 1.2) [95% CI,.2 1.6]. Ten of the 11 malignant cysts carried multiple mutations. No mutations were detected in 1 cyst with high-grade dysplasia. The mean number of mutations in the malignant group was 2.8 (SD 1.3) [95% CI, ]. The number of mutations differed significantly between the malignant and premalignant (P.003), malignant and benign (P.001), and premalignant and benign categories (P.036). The sequence of mutation accumulation was significantly different between the premalignant and malignant categories. Ten of the 11 malignant cysts had acquired a k-ras mutation first, followed by allelic loss. In contrast, 2 of 14 premalignant cysts acquired a k-ras mutation as

5 October 2005 PANCREATIC CYST FLUID MOLECULAR ANALYSIS AND CYST PATHOLOGY 971 Table 3. Summary Data of the Aspirate CEA Level, DNA OD, CT on qpcr, and the Number and Sequence of Mutations for the Premalignant and Malignant Cysts Pathology Patients CEA OD CT Number of mutations k-ras followed by LOH Malignant , , /11 Premalignant , /15 P value CEA values are given in ng/ml. the first step of DNA damage. Of these, only 1 cyst contained an allelic loss. The occurrence of k-ras point mutation first with and without subsequent allelic loss was significantly associated with a malignant cyst (P.001) (Table 3). For the presence of malignancy in an MCN, the sensitivity and specificity of a k-ras mutation occurring as the first hit were 91% and 86%, respectively, whereas the presence of allelic loss after k-ras mutation increased sensitivity and specificity to 91% and 93%, respectively. Discussion The inability to differentiate between pancreatic cysts that are benign and precancerous or indeed cancerous has led to the practice of resecting all. Thus, a significant number of patients might undergo major surgery for a benign cyst, and some might opt to defer surgery if the behavior of a malignant cyst remains unrecognized. A recent report showed the cyst aspirate CEA level to be the most accurate predictor of MCN. However, the CEA level lacks the ability to accurately diagnose the presence of malignancy and is confounded by extreme values in the malignant category. Furthermore, aspirating a sufficient sample to run the CEA analysis is not always possible. This was the case in one fourth of our cases. Cytologic analysis of pancreatic cysts in the absence of a solid component remained insensitive (1/6), a reflection of the sparse cellularity in the cyst aspirate. Revelations from the Human Genome Project raise the possibility of unique diagnostic approaches, which might be applied to these small but representative samples from pancreatic cysts. We now know that the altered morphology of malignant cells reflects underlying genetic changes. The progressive morphologic changes seen in the development of pancreatic ductal adenocarcinoma from normal cells have been modeled in the now familiar pancreatic intraductal neoplasia (PanIN) system. 5 This system includes standardized pathologic criteria for each progressive stage in tumor development linked with the underlying genetic aberrations K-ras codon 12 mutations, for example, represent one of the earliest genetic changes in the development of pancreatic ductal cancer. 5,6 Sequential inactivation of tumor suppressor gene is also seen. This progression can occur through a variety of processes including gene mutation, hypermethylation, or loss of a chromosome or chromosomal segment containing the tumor suppressor gene. Any combination of these events will lead to loss of tumor suppressor gene activity. We have proven the utility of the detection of these molecular changes as an accurate and objective marker of malignancy in brushings from biliary strictures. 19 Our approach of initial evaluation of the presence and quality of DNA followed by DNA mutational analysis has not been reported to date. It stands to reason that benign pancreatic cysts (eg, pseudocysts) and low-grade MCN will have a low rate of cellular turnover and by extension scant cyst fluid DNA. In contradistinction, malignant pancreatic cystic neoplasm should have uncontrolled cell growth and constant release of high quality, albeit mutated, DNA into the cyst fluid bathing the malignant lining. This is apparent from our results showing significantly different cyst aspirate OD and CT Table 4. Mutation Acquisition Pattern of 4 Selected Malignant Specimens Pathology Path mutation order Aspirate mutation order Mucinous cystadenocarcinoma kras-9p,9p-17p,17p-9q-1p kras-9p,9p-17p,17p Mucinous cystadenocarcinoma kras-9p-17p-3p kras-9p-17p Mucinous cystadenocarcinoma kras-1p-10q-5q 1 kras-1p34 IPMN with carcinoma in situ kras-9q-9p-5q-17p kras-9q-5q NOTE. The middle column provides the mutation sequence in microdissected tissue from the surgical specimen (first 3 cases) or from a positive cytology slide prepared from FNA of a solid component of a malignant cyst. The right column provides the mutation sequence from the corresponding cyst fluid aspirate.

6 972 KHALID ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 3, No. 10 values for the malignant and nonmalignant categories. The OD value was higher in the benign group than in the premalignant group. This might be due to the presence of DNA released from inflammatory/necrotic cells in the benign group, whereas scant DNA is expected in an indolent MCN with slow cell turnover. The DNA quality, however, was significantly higher (manifested by a lower CT) in the premalignant group than in the benign group, reflective of the respective cellular events. Also seen was the exponentially increasing mutational damage from benign cysts (no mutations) to premalignant cysts to malignant cysts, supporting a process parallel to the PanIN system. The number of mutations served as an independent tool to differentiate between these groups (Table 3). The most sensitive and specific predictor for the presence of malignancy, however, was the pattern of mutation accumulation. A particular aspect of pancreatic carcinogenesis that lacks clarity is the pattern and rate of mutation accumulation and the uncharted territory of the unique weight or significance of each individual mutation and whether this weight varies with the timing of occurrence in relation to other mutations. An exhaustive analysis of the molecular alteration in PanIN lesions would be required to answer some of these questions, although an interesting mutation acquisition pattern did arise in our series. Having defined the presence and cumulative amount of mutational damage in pancreatic cyst fluid, the timeline of individual mutation acquisition was calculated (see materials and methods). This time course derived from analysis of the cyst fluid was compared in select cases with that derived from genotyping microdissected tissue samples from the pancreatic resection specimen (Table 4). A near perfect correlation with respect to time course of mutational acquisition was found for the earlier occurring mutations, supporting the validity of the methods used to detect and characterize mutational change. Of note was the finding of k-ras-2 point mutational change as a first event, followed by allelic loss in all patients but one ultimately demonstrated to have a malignant MCN. Only one patient in the premalignant group exhibited this pattern. This pattern of mutational damage was found to be very sensitive and specific for the presence of malignancy in MCN (91% and 93%, respectively) (Table 3). This predictive power of a specific mutation pattern, however, has not been reported previously in pancreatic cyst aspirates or other hypocellular specimens. An additional advantage of the molecular analysis is the relatively small amount of fluid required for its performance. CEA analysis failed in 25% of the samples because of insufficient fluid. This scenario is not uncommon as a result of either viscous fluid in MCN that cannot be aspirated in sufficient volume through the EUS needle or a small cyst. In comparison, the molecular analysis described here can be run on a few drops of fluid (.4 ml) only. Despite our significant results, we do believe that these findings need to be validated in a larger series; a multicenter study is currently underway. A shortcoming of studies like ours is that the performance of such molecular analysis requires expertise that is not widely available. In summary, pancreatic MCNs are a challenge to treat because of the inability of currently used methods to detect malignant change. Accurate assessment of a cyst s biologic behavior is desirable, so that surgical or conservative approach can be appropriately directed. The detection of surrogate markers of the cyst pathology in the form of molecular alterations adds a new dimension to our clinical approach. We show that pancreatic cyst fluid contains adequate DNA for analysis, and the DNA amount and mutational damage reflect the pathology of the cyst. An initial k-ras mutation followed by allelic loss is most predictive of the presence of carcinoma with or without invasion in a pancreatic cystic lesion. References 1. Fernandez-del Castillo C, Targarona J, Thayer SP, et al. Incidental pancreatic cysts: clinicopathologic characteristics and comparison with symptomatic patients. Arch Surg 2003;138: Ahmad NA, Kochman ML, Brensinger C, et al. Interobserver agreement among endosonographers for the diagnosis of neoplastic versus non-neoplastic pancreatic cystic lesions. Gastrointest Endosc 2003;58: Ahmad NA, Kochman ML, Lewis JD, et al. Can EUS alone differentiate between malignant and benign cystic lesions of the pancreas? Am J Gastroenterol 2001;96: Brugge WR, Lewandrowski K, Lee-Lewandrowski E, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterology 2004;126: Hruban RH, Adsay NV, Albores-Saavedra J, et al. Pancreatic intraepithelial neoplasia: a new nomenclature and classification system for pancreatic duct lesions. Am J Surg Pathol 2001;25: Apple SK, Hecht JR, Lewin DN, et al. Immunohistochemical evaluation of K-ras, p53, and HER-2/neu expression in hyperplastic, dysplastic, and carcinomatous lesions of the pancreas: evidence for multistep carcinogenesis. Hum Pathol 1999;30: Griffin CA, Hruban RH, Morsberger LA, et al. Tumor suppressive pathways in pancreatic carcinoma. Cancer Res 1997;57: Mizumoto K, Tanaka M. Genetic diagnosis of pancreatic cancer. J Hepatobiliary Pancreat Surg 2002;9: Griffin CA, Hruban RH, Morsberger LA, et al. Consistent chromosome abnormalities in adenocarcinoma of the pancreas. Cancer Res 1995;55: Boschman CR, Stryker S, Reddy JK, et al. Expression of p53 protein in precursor lesions and adenocarcinoma of human pancreas. Am J Pathol 1994;145:

7 October 2005 PANCREATIC CYST FLUID MOLECULAR ANALYSIS AND CYST PATHOLOGY Redston MS, Caldas C, Seymour AB, et al. p53 mutations in pancreatic carcinoma and evidence of common involvement of homocopolymer tracts in DNA microdeletions. Cancer Res 1994; 54: Barton CM, Staddon SL, Hughes CM, et al. Abnormalities of the p53 tumor suppressor gene in human pancreatic cancer. Br J Cancer 1991;64: Biankin AV, Kench JG, Morey AL, et al. Overexpression of p21(waf1/cip1) is an early event in the development of pancreatic intraepithelial neoplasia. Cancer Res 2001;61: Caldas C, Hahn SA, da Costa LT, et al. Frequent somatic mutations and homozygous deletions of the p16 (MTS1) gene in pancreatic adenocarcinoma. Nat Genet 1994;8: Goldstein AM, Fraser MC, Struewing JP, et al. Increased risk of pancreatic cancer in melanoma-prone kindreds with p16ink4 mutations. N Engl J Med 1995;333: Wilentz RE, Iacobuzio-Donahue CA, Argani P, et al. Loss of expression of Dpc4 in pancreatic intraepithelial neoplasia: evidence that DPC4 inactivation occurs late in neoplastic progression. Cancer Res 2000;60: Hahn SA, Schutte M, Hoque AT, et al. DPC4, a candidate tumor suppressor gene at human chromosome 18q21.1. Science 1996;271: Luttges J, Galehdari H, Brocker V, et al. Allelic loss is often the first hit in the biallelic inactivation of the p53 and DPC4 genes during pancreatic carcinogenesis. Am J Pathol 2001;158: Khalid A, Pal R, Sasatomi E, et al. Use of microsatellite marker loss of heterozygosity in accurate diagnosis of pancreaticobiliary malignancy from brush cytology samples. Gut 2004;53: Finkelstein SD, Marsh W, Demetris AJ, et al. Microdissectionbased allelotyping discriminates de novo tumor from intrahepatic spread in hepatocellular carcinoma. Hepatology 2003;37: Rolston R, Sasatomi E, Hunt J, et al. Distinguishing de novo second cancer formation from tumor recurrence: mutational fingerprinting by microdissection genotyping. J Mol Diagn 2001;3: Address requests for reprints to: Asif Khalid, MD, Division of Gastroenterology, Hepatology and Nutrition, University of Pittsburgh Medical Center, M2, C-wing, PUH, 200 Lothrop Street, Pittsburgh, Pennsylvania khalida@upmc.edu; fax: (412) Matt Wilson, Patricia Swalsky, and Sydney Finkelstein are employees of Redpath Integrated Pathology.