Final pathological diagnoses in 69 resected intraductal papillary mucinous neoplasm (IPMN) specimens.
Wiesenauer CA, Schmidt CM, Cummings OW, Yiannoutsos CT, Howard TJ, Wiebke EA, Goulet RJ, McHenry L, Sherman S, Lehman GA, Cramer H, Madura JA. Preoperative Predictors of Malignancy in Pancreatic Intraductal Papillary Mucinous Neoplasms. Arch Surg. 2003;138(6):610-618. doi:10.1001/archsurg.138.6.610
Malignant intraductal papillary mucinous neoplasms (IPMNs) can be predicted before surgery.
Retrospective review of a prospectively collected database.
Academic, urban, tertiary care hospital.
Sixty-four consecutive patients with a pathological diagnosis of IPMN.
All 64 patients underwent surgical intervention for IPMN between December 8, 1988, and October 16, 2002.
Main Outcome Measures
Reliable predictors of malignancy.
The 64 patients underwent 69 operations: 39 pancreaticoduodenectomies, 18 distal pancreatectomies, 7 total pancreatectomies, 4 neck and/or body pancreatectomies, and 1 cystgastrostomy with pancreatic biopsy. Twenty-three of 69 specimens were malignant—12 in situ (high-grade dysplasia) and 11 invasive. In a univariate analysis of 12 clinical signs or symptoms recorded, diabetes mellitus and jaundice showed a significant association with malignancy of IPMN. Of 24 serum chemistry studies, hematologic studies, and tumor marker analyses (in serum, bile, and pancreatic fluid), elevation of serum alkaline phosphatase and glucose levels showed correlation with malignancy. Computed tomography, ultrasound, and endoscopic retrograde cholangiopancreatography findings did not distinguish between benign and malignant tumors. Atypia on preoperative cytologic analysis was specific for malignancy (93%) but lacked the same degree of sensitivity (40% in situ, 91% invasive, and 67% overall).
Malignancy of IPMNs is suggested by new-onset diabetes mellitus, jaundice, and elevations in serum glucose or alkaline phosphatase levels. Atypia on preoperative cytologic testing is the finding most predictive of malignancy. The absence of these features does not predict benign disease. These findings may help guide patient and physician decision making.
PANCREATIC CANCER is the fourth most common cause of death due to gastrointestinal malignancy. In the United States alone, approximately 27 000 new cases and an equal number of deaths were reported in 2001.1 As the demographics indicate, available treatments will cure only a handful of patients with this disease. The only possibility of cure is by surgical resection. In centers that perform large volumes of pancreatic surgery, approximately 15% of patients who undergo curative resection will survive 5 years.2 Only approximately 15% of patients who have pancreatic cancer are even candidates for surgical therapy. For patients with locally advanced disease, combined adjuvant chemoradiotherapy has been shown to minimally prolong survival over that of control patients who did not receive this treatment.3 With little hope from available therapies, intensive efforts should be focused on discovery of new treatment modalities, but emphasis should also be placed on screening and prevention.
Ninety percent of patients with pancreatic adenocarcinoma have spontaneous lesions.4,5 The remaining 10% are secondary to known hereditary effects, that is, hereditary pancreatic cancer, hereditary pancreatitis, and Lynch syndrome. Developing effective screening and prevention programs for families at high risk, even in a small portion of patients, if successful, could significantly improve survival rates. In addition, some patients have precursors of pancreatic adenocarcinoma. These precursors include pancreatic intraepithelial neoplasia, mucinous cystadenoma, and intraductal papillary mucinous neoplasms (IPMNs). Surgical therapy at present is the only effective mechanism of prevention for pancreatic adenocarcinoma in these patients. The focus of the present investigation is on IPMNs, which have been classified as such only since 1994, with the term "intraductal papillary mucinous neoplasm" being accepted by the World Health Organization in 1996.6,7 Before that time, IPMNs were grouped with mucinous cystic neoplasms, first being described in the Japanese literature in 19828 and in the English literature in 1986.9
Intraductal papillary mucinous neoplasms can be benign or malignant. Patients often have local symptoms of pain or recurrent pancreatitis, and IPMNs can be mistaken for idiopathic pancreatitis when patients have a large, dilated pancreatic duct. What often distinguishes IPMNs from pancreatitis is that the pancreatic duct is filled with mucin. On endoscopic retrograde cholangiopancreatography (ERCP), these lesions are often suggested by findings such as a dilated pancreatic duct, a gaping ampulla, extrusion of mucin, mucous plugging or obstruction, and a cystic lesion in communication with the main pancreatic duct or a side branch. Current treatment involves resection of the involved pancreas, with intraoperative frozen section to prove that the margin of resection is negative for any stage of IPMN.10 Despite surgical resection, however, many of these lesions will recur in the remnant pancreas (7%-42%).11- 15 Histologic examination of these lesions sometimes reveals that there is multifocality and several different stages of progression of IPMNs within a single specimen, suggesting that IPMNs may represent a field cancerization of the pancreas.16 Patients may require long-term surveillance even after complete resection of the clinically apparent lesion. Furthermore, some physicians advocate intensive monitoring of these patients in lieu of surgical extirpation because sometimes they occur in elderly patients with comorbidities, which make surgical resection less desirable.10 Such an approach might be applied with more confidence if reliable preoperative indicators of malignant vs benign tumors existed.
We sought to determine whether there were any preoperative predictors of malignancy in pancreatic IPMNs that would support an approach of observation in potentially benign IPMNs. We therefore conducted a retrospective review of a prospectively collected database at Indiana University Medical Center, an academic, urban, tertiary care hospital. Sixty-four consecutive patients with a pathological diagnosis of IPMN underwent operative intervention between December 8, 1988, and October 16, 2002. This study examines symptoms, signs, serum chemistry studies, hematologic studies, tumor markers, radiographic studies, ERCP, and cytologic evaluations to discern whether there are reliable preoperative predictors of malignancy.
All data in this study were collected and reported in strict compliance with patient confidentiality guidelines put forth by the Indiana University institutional review board. Patients included in this retrospective review were referred to the Indiana University School of Medicine between December 8, 1988, and October 16, 2002. All patients underwent a history and physical examination, baseline serum chemistry tests, and necessary radiographic or endoscopic procedures to confirm the diagnosis of IPMN. All patients underwent surgery for IPMN.
Data were prospectively collected and analyzed from the principal investigator's (J.A.M.) database. Serum chemistry studies, hematologic studies, and tumor markers were analyzed from routine preoperative phlebotomy samples. Pancreatic tumor markers were analyzed from routine ERCP sampling. In selected cases, pancreatic fluid, if insufficient, was provoked with intravenous secretin. Bile tumor markers were analyzed from aspirations of the gallbladder at the time of surgery. Computed tomography (CT) and ultrasound (US) examinations were read by the surgeon (C.M.S., T.J.H., E.A.W., R.J.G., or J.A.M.) and the staff radiologist. The ERCPs were interpreted by the gastroenterology staff. Cytologic evaluations were obtained from pancreatic duct lavage and brushings via ERCP. Fine-needle aspiration (FNA) samples were obtained at the time of endoscopic US (EUS) or ERCP.
Patients underwent either pylorus-preserving or classic pancreaticoduodenectomy, distal pancreatectomy, neck and/or body pancreatectomy, or total pancreatectomy. One patient in the first year of this review underwent a cystgastrostomy and pancreatic biopsy. Pancreaticoduodenectomy was performed via a chevron or midline incision. After thorough abdominal exploration and a generous Kocher maneuver, the gall bladder was removed. Before removal, bile was aspirated from the gall bladder into a sterile container for tumor marker and cytologic analyses. The common bile duct was then transected, and the anterior aspect of the portal vein was dissected free of the overlying pancreatic neck. Subsequently, the duodenum was transected in the pylorus-preserving pancreaticoduodenectomy, or the stomach was transected in the classic pancreaticoduodenectomy, followed by transection of the pancreatic neck, uncinate process, and jejunum distal to the ligament of Treitz. Reconstruction was undertaken with an isoperistaltic limb of jejunum in retrocolic fashion and anastomosed with an end-to-side pancreaticojejunostomy, followed by an end-to-side choledochojejunostomy, followed by an antecolic or retrocolic end-to-side duodenojejunostomy or gastrojejunostomy. Distal pancreatectomy involved exploration of the lesser sac, mobilization of the inferior and superior margins of the pancreas, and takedown of the splenocolic and renocolic ligaments. This procedure was followed by takedown of the short gastric vessels, removal of the diaphragmatic attachments to the spleen, and individual ligation of the splenic artery and vein. The pancreas was transected and oversewn in 2 layers. Total pancreatectomy combined both of these procedures, and the neck and/or body pancreatectomy in all cases followed a previous distal pancreatectomy. Frozen histologic section of remaining pancreatic tissue was performed before any operative reconstruction. Evidence of IPMNs prompted further resection. Drains placed next to pancreatic resections or anastomoses were routine.
The ERCP procedure involved placement of a video duodenoscope (Olympus America Inc, Melville NY) orally after intravenous sedation with meperidine hydrochloride and midazolam hydrochloride or diazepam. The main pancreatic duct was cannulated, and pancreatic fluid was aspirated into a sterile container for tumor marker and cytologic analyses.
Similarly, EUS involved placement of a flexible echo endoscope (Olympus America Inc and Pentax, Englewood, Colo), flexible radial scanning, and flexible linear arrayo. Fine-needle aspiration of the IPMN was performed through the operative port for cytologic analyses.
All specimens were primarily reviewed by the pathologists at the Indiana University School of Medicine. A single pathologist (O.W.C.) retrospectively reviewed all of the specimens and reports to ensure their validity. Intraductal papillary mucinous neoplasm was defined as a mucinous lesion of the pancreas in communication with the pancreatic duct that lacked ovarian-type stroma (characteristic of cystadenomas). Staging of IPMNs was defined as normal (no feature of IPMN appreciated), adenoma (dilated pancreatic duct lined by mucinous epithelium, with ≤1 criteria for low-grade dysplasia; also called duct ectasia), low-grade dysplasia (≥2 of the following criteria: epithelial tufting, nuclear pseudostratification, nuclear atypia, and mitotic figures; also called borderline), high-grade dysplasia (cribriform or solid growth usually associated with high-grade nuclear atypia; also called noninvasive intraductal carcinoma or carcinoma in situ), and invasive. Malignant IPMNs were defined as IPMNs with high-grade dysplasia or invasive staging. Criteria for cytologic atypia included at least 1 of the following: increased nuclear-cytoplasmic ratio, increased nuclear size, nuclear crowding, or hyperchromasia.
Statistical associations between categorical factors (positive signs, symptoms, ERCP findings, etc) with the presence of malignant IPMNs were assessed via the Fisher exact test. In addition, the association of continuous variables (age, serum chemistry values, etc) with the presence of malignant IPMNs were analyzed using a univariate logistic regression model and tested via the Wald test. Multivariate analyses assessing the simultaneous impact of various categorical and continuous factors were also undertaken. To determine the optimal subset of markers associated with the presence of a malignant IPMN out of the many indices measured in this study, a stepwise model-selection procedure was used. A candidate factor was included in the model if it had an associated Wald P = .05. The model with each added factor was refitted, and if another previously entered factor became nonsignificant (Wald P>.10), it was removed from the model. The selection procedure stopped if it was impossible to add or subtract any additional factors. The predictive accuracy of each set of predictors was assessed via receiver operating characteristic curves17 fitted using a maximum-likelihood approach.18 The incremental diagnostic improvement resulting from the addition of 1 or more predictors to a baseline predictor set was assessed using the method of DeLong et al.19
This is a retrospective review of a prospectively collected database of patients with IPMNs diagnosed and treated operatively between December 8, 1988, and October 16, 2002. All patients underwent surgical treatment at the Indiana University School of Medicine. The purpose of the review was to determine whether malignant IPMNs could be predicted before surgery.
Each preoperative study or observation was analyzed to determine whether it correlated with the ultimate tissue diagnosis, whether it be benign or malignant. In the 64 patients studied, there were 69 operations: 39 pancreaticoduodenectomies (57%), 18 distal pancreatectomies (26%), 7 total pancreatectomies (10%), 4 neck and/or body pancreatectomies (6%), and 1 pancreatic biopsy (1%). There was no correlation between the operation performed or the location of the tumor and the ultimate tissue diagnosis.
Of the 69 pathological specimens reviewed, 23 were malignant (33%), 45 were benign (65%), and 1 did not contain IPMNs (1%) (Figure 1). Of the malignant IPMNs, 12 were in situ (high-grade dysplasia) and 11 were invasive cancers. Four patients with invasive cancers had metastases to lymph nodes, and an additional patient with invasive cancer had a distant metastasis to the diaphragm. Of the benign IPMNs, 2 were staged as adenomas, and 43 were staged as low-grade dysplasia (borderline). The patient listed with no IPMNs on pathological specimens had IPMNs at the margin on permanent pathological analyses at a previous surgery in addition to persistent symptoms of recurrent pancreatitis. Subsequent completion pancreatectomy showed pancreatitis but no IPMNs in the specimen.
There were 33 men and 31 women in this series. There was a trend suggesting that male sex was significantly associated with malignant IPMNs (P = .06) (Table 1). Of the 20 patients with a diagnosis of malignant IPMNs, 14 (70%) were men. In other words, 14 (42%) of 33 male patients had malignancy overall compared with 6 (19%) of 31 female patients. Patient age did not differ statistically between malignant IPMNs and nonmalignant IPMNs (Table 2).
Twelve patient symptoms and signs were recorded in the prospective database (Table 1 and "weight loss" in Table 2). The association between these factors and the presence of malignancy was assessed using the Fisher exact test (logistic regression was used for weight loss), as described previously. Of these symptoms and signs, diabetes mellitus (P = .003) and jaundice (P = .01) showed a significant association with malignancy of IPMNs. The presence of new-onset diabetes mellitus was defined as the existence of glucose intolerance necessitating diet modification, hypoglycemic medication use, or insulin use within 2 years prior to diagnosis of IPMN without another obvious explanation. The diagnosis of jaundice was based on the presence of scleral icterus on examination. There was a trend marginally associating weight loss (P = .06) and diarrhea (P = .10) with malignant IPMNs. Weight loss was quantified by the patient at the time of the preoperative visit. The diagnosis of diarrhea was made if patients experienced 3 or more bowel movements per day or if the volume of their movements in a day was greater than or equal to 0.5 L. Pancreatitis and pain symptoms did not correlate with malignancy.
In the prospective database, we also analyzed 9 serum chemistries, 6 hematologic studies, and 3 tumor markers in each of serum, bile, and pancreatic fluid (Table 2). The analysis involved a univariate logistic regression model as described in the "Methods" section. Of these 24 tests, serum glucose (P = .008) and alkaline phosphatase (P = .01) showed correlation with malignancy. The elevations of the serum alkaline phosphatase and glucose levels in patients with malignancy are consistent with our findings of jaundice and diabetes mellitus. Serum CA19-9 levels (as a continuous variable) did not reach statistical significance in the logistic regression model. Nonetheless, the presence of elevated serum CA19-9 levels (>70 U/mL) was significantly associated with malignancy. Only 3% of benign IPMNs had an elevated preoperative serum CA19-9 level (1 of 36 patients with benign IPMNs tested). By comparison, 7 (37%) of the 19 malignant samples had elevated serum CA 19-9 levels (Fisher exact test P = .002).
Next, we examined radiographic and invasive procedures, which included CT, US, EUS, and ERCP. Of the 69 patient encounters, CT was performed in 61, ERCP in 62, EUS in 33, and transabdominal US in 25. Criteria considered consistent with IPMN on CT or either US technique were a dilated main or branch duct, a mass, a cystic lesion, or a lesion in continuity with the pancreatic duct. Preoperative CT findings were consistent with IPMN in 51 (84%) of 61 patients tested (17 [81%] or 21 patients with malignant tumors and 34 [85%] of 40 patients with benign tumors). Preoperative transabdominal US findings were consistent with IPMN in 17 (68%) of 25 patients tested overall. Nine (90%) of 10 patients with malignant tumors had US findings consistent with IPMN vs only 8 (53%) of 15 patients with benign tumors; this difference did not reach statistical significance (P = .09). None of the CT or US findings that were tracked predicted malignancy. All patients undergoing ERCP had findings consistent with IPMN, including visualization of mucin, filling defects of the pancreatic duct, a dilated pancreatic duct, a cyst in the pancreas, and a gaping or patulous ampulla. Additional ERCP findings that were tracked included the presence of stricture, obstruction, and pancreas divisum (Table 3). No ERCP findings that we recorded were predictive of malignancy.
We did not prospectively analyze tumor size or pancreatic duct diameter as a preoperative predictor of malignancy. In part, this is because IPMNs take on so many different forms (combinations of main and side duct dilation, absence of dilation, discrete cystic lesions, etc). Retrospective examination of preoperative CT, ERCP, transabdominal US, and EUS images revealed that in the small subset of patients assigned quantitative tumor data, the mean largest diameter of the benign tumors was 2.74 cm and of the malignant tumors was 4.44 cm. The data, although suggestive, are not sufficient to make any statistical conclusions.
Table 4 provides cytologic findings obtained through pancreatic ductal lavage, brushing, or FNA. All cytologic specimens were obtained via ERCP, except FNA specimens, which were derived by ERCP and/or EUS; FNAs were not doubly counted in this analysis. For example, for each preoperative encounter, FNA was considered 1 positive result if it was present in the ERCP FNA alone, the EUS FNA alone, or both. Cytologic diagnoses of normal epithelial cells, extracellular mucin, and mucinous epithelium were not associated with malignancy. Atypical or malignant cells, however, were associated with malignancy when collected via ductal lavage (P = .008) and FNA (P<.001) and were marginally associated with malignancy when collected via brushings (P = .06).
Table 5 gives the predictive values of various sources of preoperative cytologic evaluation on a malignant outcome. Presence of atypia was the single factor most strongly associated with malignancy (Fisher exact test 2-sided P<.001). Twelve (75%) of 16 samples with malignancy had atypical cytologic findings compared with 11 (21%) of 53 benign samples. Atypical cytologic findings were observed in 12 (36%) of 33 male patients but only in 3 (10%) of 31 female patients (P = .02). The confounding of the effect of atypia and sex may at least partially explain the statistically significant trend associating sex with the likelihood of malignancy. The only factors, other than sex, that were significantly associated with the presence of atypia (and malignancy) were glucose level (Wald P = .01) and diabetes mellitus (Fisher exact test P = .007).
Using all cytologic sources, the presence of atypia is 91% sensitive in detecting malignancy in patients with invasive IPMN carcinoma. The presence of atypia is, however, only 40% sensitive in detecting malignancy in patients with in situ (high-grade dysplasia) IPMN carcinoma. Therefore, the presence of malignancy will not be discovered before surgery in 9% of patients with invasive IPMN carcinoma and in 60% of patients with in situ (high-grade dysplasia) carcinoma. Using all cytologic sources, the presence of atypia is 93% specific for malignancy. In examining individual cytologic sources, pancreatic ductal lavage seems to be the most specific test (100%) for malignancy since there were no false-positive determinations. In addition, FNA, when the results of ERCP and EUS are combined, is 100% sensitive at detecting invasive (but not high-grade dysplasia) IPMN carcinoma.
For the multivariate analysis, we used a model-selection process, as described in the "Statistical Analysis" subsection. By restricting inclusion into the multivariate logistic regression model to variables that simultaneously (ie, adjusted for each other) reached significance at the 5% level, only 2 factors were included: atypia (Wald test P<.001) and alkaline phosphatase level (P = .02). The direction of the association was consistent with the univariate analyses (ie, patients with atypia and those with elevated alkaline phosphatase levels were significantly more likely to have malignant IPMNs). Diabetes mellitus and glucose level, although significant on univariate analysis, did not add to the predictability of malignancy in multivariate analysis.
To assess the predictive accuracy of various factors, we constructed a diagnostic index based on the logistic regression model. Higher values of this index were suggestive of malignancy, and lower values were not. The single factor with the highest predictive index was the presence of atypical findings, with overall predictive accuracy just greater than 82%, estimated via the maximum-likelihood procedure described in the "Statistical Analysis" subsection. Overall predictive ability means that of 2 randomly chosen patients, 1 IPMN benign and 1 malignant, the patient with malignancy will have atypical findings more than 82% of the time. Expanding on the best multivariate model, containing alkaline phosphatase level and the presence of atypia, the overall predictive ability of the model increased to greater than 84%. This increase, however, was not statistically significant.
We conducted exploratory analyses by adding other factors to the predictive model, but no factor combination produced a predictive index with significantly higher predictive ability. These observations suggest that, at least with respect to the statistical model, any information about the likelihood of malignancy furnished by other factors was subsumed in the information provided by the presence of atypia. Seen from another perspective, after considering the information on malignancy supplied by the presence or absence of atypia, no significant additional information was provided by other factors examined.
Pancreatic cancer continues to be a significant threat in the United States. Less than 5% of patients who have a diagnosis of pancreatic adenocarcinoma undergo curative resection and survive for 5 years. Early detection and prevention, and better treatments in established tumors, need to be emphasized. Intraductal papillary mucinous neoplasms represent a premalignant precursor to pancreatic ductal adenocarcinoma. Thirty-two percent of IPMNs in this series harbored malignancy. In this series, which matches some of the largest series reported to date, we conducted a retrospective review of a prospectively collected database in a single hospital of 64 patients undergoing surgery for IPMNs. Malignancy of IPMNs was associated with the presence of new-onset diabetes mellitus and jaundice, the levels of serum glucose and alkaline phosphatase, and the presence of elevated levels of serum CA19-9. Notable trends included the association of male sex, weight loss, and diarrhea with malignancy. Most important, a finding of atypical cells on preoperative cytologic analysis was virtually diagnostic of malignancy in IPMNs.
In a variety of univariate analyses, we observed that diabetes mellitus and elevated glucose levels were predictive of malignancy. It is not clear why the existence of new-onset diabetes mellitus and correspondingly elevated preoperative glucose levels were more common in patients with malignancy. We hypothesize that this may be a marker of more extensive disease. However, thus far, this finding is not supported by our current data regarding the pathological findings and the extent of surgery. Furthermore, these patients did not have a higher incidence of pancreatic duct obstruction or recurrent or chronic pancreatitis to explain the pancreatic endocrine dysfunction. More extensive disease also could not be substantiated in patients with preoperative diarrhea. These findings are consistent with what other researchers20 have published, that chronic pancreatitis is a not a predictor of malignant disease.
Weight loss, jaundice, and correspondingly elevated alkaline phosphatase levels seem to be signs of a more advanced state of IPMN. Male sex was weakly associated with malignant IPMN. Male patients were also more likely to have atypical cytologic test results, which was the single most predictive factor of malignancy (P<.001). Males may indeed be predisposed to more aggressive IPMNs. Alternatively, male patients' duration of symptoms may have been more prolonged before they sought medical attention compared with their female counterparts.
Atypia was the most predictive factor of malignancy (P<.001). This holds true in the multivariate analysis as well. The presence of atypical cells on preoperative cytologic analysis was virtually diagnostic of malignancy associated with the IPMN. The absence of atypical cells on preoperative cytologic testing, however, was not a reliable indicator of benign IPMNs.
Several researchers have described preoperative variables common in patients with IPMNs, and some have attempted to correlate them with subsequent histologic findings or survival outcomes. These variables have included tumor size, sex, diabetes mellitus, pancreatitis, steatorrhea, abdominal mass, weight loss, serum CA19-9 level, and serum carcinoembryonic antigen level. Few studies,20- 23 however, have shown any statistically significant preoperative predictors of malignancy. Loftus et al20 showed, in a series of 15 patients with IPMNs, that a history of preoperative chronic pancreatitis is predictive of benign disease. A similarly sized study by Uehara et al24 estimated size of main pancreatic duct/ampullary orifice and lesion size from preoperative imaging and showed a correlation of increasing size with malignancy. A main pancreatic duct greater than 9 mm or a lesion greater than 20 mm was associated with malignancy. The same study correlated cytologic findings of malignancy in pancreatic fluid with an ultimate diagnosis of malignancy, although the definition of malignancy (79% of patients) is unclear in their study. Some studies have shown that expression of p53 and Ki-67 proteins is significantly greater in surgical specimens from malignant than benign IPMNs.25 K-ras mutations have also been examined as predictors of IPMN malignancy, but they fell short of statistical significance.26 Although potentially useful, these molecular studies were performed on postsurgical specimens and have not been shown, to our knowledge, to be useful in the preoperative setting. Similarly, main and side branch types of IPMN pathological abnormalities have been correlated with malignant outcome, but preoperative discrimination of these types is not straightforward.27
The current management of IPMNs involves surgical resection to margins free of IPMNs. Since IPMN in some situations may represent a field defect, this treatment may not be the best way to eradicate any possibility that the patient will ultimately develop malignancy. Some surgeons have advocated total pancreatectomy to remove all pancreatic tissue at risk.28,29 This is similar to the approach in patients with ulcerative colitis. Other groups think that this is a drastic approach and that patients should undergo resection of the lesion to clean margins and then be closely followed with magnetic resonance cholangiopancreatography and CT.10,30 Although this approach is the most commonly practiced at present, patients have been known to return with recurrences of IPMNs and sometimes with invasive adenocarcinoma not amenable to resection. If total pancreatectomy is not the answer, then we need to come up with a better marker of malignant disease not only so that patients who opt for expectant management or novel chemopreventive strategies can be followed with better accuracy, but also for patients who have undergone resection so that they can be screened for recurrence. This study reports reliable diagnosis of invasive IPMNs by preoperative cytologic examination. At the invasive stage of disease, these patients are seldom curable. An earlier stage of IPMN carcinoma (IPMN with high-grade dysplasia) could not be reliably diagnosed by preoperative cytologic examination in this study. In the future, techniques for obtaining preoperative cytologic test results may be able to be modified to increase the yield of cells sampled and thereby increase the chances of discovering atypical cells if in fact they exist. Peroral pancreatoscopy or intraductal ultrasonography may be methods to not only improve the cytologic sampling but also to help rule out the possibility of skip lesions and multifocality.31 Although our retrospective review identifies factors associated with malignancy in the preoperative setting, we cannot reliably predict which patients will have benign disease. Nonetheless, these findings help guide patient and physician decision making, and prospective cytologic studies looking at novel tumor markers are warranted. These findings also reinforce an aggressive surgical approach in patients with IPMNs who are surgical candidates.
Corresponding author and reprints: C. Max Schmidt, MD, PhD, Departments of Surgery and Biochemistry and Molecular Biology, Indiana University School of Medicine, 1044 W Walnut St, R4-039, Indianapolis, IN 46202 (e-mail: firstname.lastname@example.org).
Accepted for publication February 8, 2003.
This study was presented at the 110th Scientific Session of the Western Surgical Association, Vancouver, British Columbia, November 18, 2002, and is published after peer review and revision. The discussions that follow this article are based on the originally submitted manuscript and not the revised manuscript.
Drs Wiesenauer and Schmidt contributed equally to the preparation of the manuscript.
Rahim Moossa, MD, San Diego, Calif: IPMN of the pancreas was only described in 1986 by Itai and his colleagues from Japan, and this clinical-pathological entity masqueraded under multiple names that we all cannot remember until 1996, when the World Health Organization nomenclature of IPMN became accepted. We do not know whether this is a new tumor or whether this is something that existed but was not recognized. There are only 3 undisputed facts about IPMN:
The resected specimen always exhibits a distinct spectrum of histologic malignancy ranging from small papilloma through carcinoma in situ to lethal invasive cancer and even metastatic disease.
The cancer is definitely multifocal.
Survival after resection is, in general, much better than for the ductal adenocarcinoma of the pancreas.
The authors have demonstrated through statistical analysis that there are several preoperative predictors of malignancy, but they appropriately emphasize that the absence of any or all of these factors does not exclude malignancy. Of the 65 cases reported, only 11 had invasive cancer and only 12 had carcinoma in situ. Unfortunately, no survival data are provided. So the key questions we have for the authors are:
How much of the pancreas should we resect?
Can we rely on the pathologist about the status of the surgical margins on frozen section histology?
How do we safely monitor the pancreatic remnant? Can we rely on [magnetic resonance cholangiopancreatography] or tumor markers or endoscopic ultrasound? Are they that reliable in the postoperative state?
There is a distinct impression by some surgeons—and I am one of them—that survival data for benign IPMN and for the "malignant counterpart" are about the same. Does this mean that our intraoperative assessment and decision making is imperfect and we leave cancer behind or that the pathologist has not examined enough sections of the resected specimen or that as a result of the field defect a new malignancy has arisen in the pancreatic remnant?
How far are you from recommending a routine total pancreatectomy, since many of your patients are already diabetic?
What is the role of postoperative radiotherapy and chemotherapy?
Finally, since we are seeing more and more IPMN recently, is it a true increase in incidence, or are we simply reclassifying what we did not recognize before?
Richard A. Prinz, MD, Chicago, Ill: Because IPMN is a newly recognized entity, our knowledge of its natural history, especially following surgical intervention, is somewhat limited. I wonder if you could share with us the rate of recurrence that you have had both in those patients whom you thought had benign disease and those who had frank malignancy. How was this recurrence recognized and treated?
Kelly McMasters, MD, Louisville, Ky: This presentation focused on identifying preoperative predictors of malignancy in patients with IPMN. The problem I have with this newly popularized diagnosis is deciding which patients have IPMN in the first place. I've operated on a few patients who I was told had IPMN only to find chronic pancreatitis. What are the criteria for diagnosing IPMN preoperatively?
James R. DeBord, MD, Peoria, Ill: One of the clinical manifestations of these patients was significant diarrhea. The assumption was that it was from exocrine insufficiency, yet there was really no correlation with ductal obstruction. Did you treat these patients with replacement therapy? Did they respond to that? Or is diarrhea from some other cause?
Mark Talamonti, MD, Chicago: You presented greatly detailed information about the results of your ERCP, and our gastroenterologists will frequently give us information from their ERCP about the so-called mapping of the changes within the pancreatic duct in an attempt at guiding us for the operation to perform. How did you choose a Whipple procedure, how did you choose the distal pancreatectomy vs a Whipple procedure, and did the ERCP help you at all to choose the operation that you performed?
Dr Madura: I would like to especially thank all of our coauthors, who worked so hard at analyzing these data. Dr Moossa, thanks for your comments. You bring out the things that we have been struggling with since we made our first diagnosis. That is, is this truly a new disorder or is this something we have seen since ERCP has been utilized? We misdiagnosed the first patient we did, and the patient came back a number of years later with a huge cystic malignancy that was treated at another hospital with a cystgastrostomy. Is the disease multifocal? Is this a field defect? Can you be comfortable in following the patients after a simple resection? We are still trying to find that out as this disease evolves.
You asked how much do we resect? What we have done is to resect back to histologically clean pancreas, whether proximal or distal, and when we did many of these cases I went to the pathology suite with the pathologist and looked at the frozen section. If there was any evidence of mucinous epithelium or IPMN, then we resected more pancreas until we had clean duct.
How many of the tumors returned? Well, I think that 12% of these patients came back with disease in the remaining remnant, and following them, as you point out, is difficult. It's easy if they have a distal pancreatectomy because we have an excellent group of ERCP gastroenterologists and they can brush the pancreatic duct and get cytology, which is exceedingly reliable in our hands. The Whipple patients are problematic because you don't have access to that duct endoscopically, so we have been using all of the modalities, the CT scan, MRCP; and if we see a dilated duct or the patient begins to have pancreatitis again, then we would go back and re-explore and re-excise the pancreatic remnant.
We didn't present any survival data because that was not the focus of this study, but you are correct. The people who have benign or low-grade dysplasia do well and survive long periods of time. Those who have invasive IPMN don't do as well and many of them do just as poorly as the usual ductal adenocarcinoma patients.
Is this a new malignancy when it comes back or is it something that was there and it just grew? I have no way to know that except that the ducts all look normal on the preoperative ERCPs in the part that we leave behind. What about total pancreatectomy? We have done that in about 10% of our patients, and we have done those because the tumor is either pan-pancreatic—we cannot get a clean margin at the primary operation—or they come back with "new" disease. None of these patients so far, with the exception of frank invasive adenocarcinoma, have had chemotherapy or radiation therapy.
Dr Prinz, our rate of recurrence is probably about 12%, and we are following all of these patients who are still alive annually.
What about the IPMN criteria? I think Dr Schmidt presented those very clearly and our pathologists have been thorough, with all of this material that we have given them. I think they have become pretty good and pretty reliable at making these diagnoses.
The question was asked about diarrhea: we have no idea why these people have diarrhea. We don't think it was due to ductal obstruction as you have heard. We also don't think it was due to pancreatic insufficiency. Some would respond transiently to enzyme therapy, but many did not.
Finally, how do we choose the specific operation? Most of the time it is pretty easy because this is a segmental disease on CT and ERCP and it is in the tail, the body, or the head, and so you have the luxury to plan ahead of time what you are going to do. They are very visible on ERCPs because there are ductal irregularities. The ducts are dilated. You can see intraductal filling defects that lead you to the correct operation. In the patients in whom we have done initial total pancreatectomies, as I said, we could not get a clean margin and we had to go ahead and complete the pancreatectomy.