Pancreatic cancer

Epidemiology 

In the United States there are 28,000 to 30,300 newly diagnosed cases of pancreatic cancer and approximately an equal number of deaths per year from pancreatic cancer.1, 2 Ductal adenocarcinoma is the most common form of pancreatic cancer. The annual incidence rate for all types of pancreatic cancer is approximately 9 new cases per 100,000 people, ranking it 11th among cancers.2 The peak incidence occurs in the seventh and eighth decades of life, with the average age at diagnosis being 60 to 65 years.3 The incidence rate is slightly higher in men than women (relative risk 1.35), and 30% to 40% higher in black men.3 This racial discrepancy persists in spite of similar cigarette smoking rates in black men (27%) and white men (25%).2

Survival in untreated patients with pancreatic cancer is poor. For all stages combined, the 1-year survival rate is 19% and the 5-year survival rate is 4%.2 Survival is low because of the late development of clinical symptoms, such that 80% of pancreatic cancers are metastatic at the time of diagnosis. Surgical resection (when margin negative and node negative) offers the best possibility for cure in pancreatic cancer, with the 5-year survival rate improving to 40%, when performed at specialized, high-volume major medical institutions.4 The perioperative mortality rate is considerably higher when resection is performed at low-volume institutions.4, 5

In the United States the incidence rates of pancreatic cancer increased threefold between 1920 and 1978, an increase that has also been observed in other developed countries.6 Since 1978, incidence rates for men and women have declined slightly and appear to have stabilized at the current levels. A portion of the increased incidence from 1920 to 1978 may have been attributable to more accurate disease diagnosis and less misclassification of the disease. Additionally, improved surveillance may partially account for the increased incidence. Many studies have found a relationship between certain environmental exposures and cases of pancreatic cancer, including personal cigarette smoking, environmental tobacco smoke, and chemical exposures.7 Cigarette smoking in the United States and in other countries increased greatly in the first half of the 20th century, such that 40% of adult Americans smoked in 1965. A large portion of the increased incidence of pancreatic cancer is also likely attributable to increased smoking through the 1960s. By 1990 the prevalence of smoking had decreased to 25%, with modest declines again noted in 1999.8 It remains to be seen if this will translate into lower pancreatic cancer incidence rates in the future.

Risk factors 

A number of demographic risk factors have been associated with the development of pancreatic cancer (Table 1). These include older age, black race, male sex, low socioeconomic status, and Ashkenazi Jewish heritage.7 Host etiologic factors associated with an increased risk of pancreatic cancer include diabetes mellitus, pancreatitis, and prior cholecystectomy.2, 3, 9 The association between diabetes, pancreatitis, and the development of pancreatic cancer is complex because pancreatic cancer, as it destroys pancreatic parenchyma, can itself cause diabetes and pancreatitis.

Table 1. Risk factors associated with cancer of the pancreas

	Life-style factors
	Cigarette smoking (dose-response relationship)
	Lack of exercise
	Race/ethnic factors
	Black men
	Native female Hawaiians
	Ashkenazi Jewish heritage (may be due to BRCA2 gene found in some Jewish families)
	Inherited predispositions:
	Hereditary breast and ovarian cancer (BRCA2)
	Familial atypical multiple mole melanoma
	Peutz-Jehghers syndrome
	HNPCC
	Hereditary pancreatitis
	Ataxia-telangiectasia
	Medical conditions
	Cirrhosis
	Diabetes mellitus
	Chronic pancreatitis
	Dietary factors
	High fat/cholesterol
	Obesity
	Nitrosamines in food
	Occupational exposure to carcinogens
	2-naphthylamine
	Benzidine
	Gasoline products
	Polychlorinated biphenyls
	Dry cleaning agents
	DDT (dichloro-diphenyl-trichloroethane)
	Selected high-risk occupations
	Dry cleaning
	Chemical plant work
	Sawmills
	Electrical equipment manufacturers
	Mines
	Metal working
	Height
	Relative risk 1.81; (CI 1.31 to 2.52) when comparing tallest and shortest height categories for men and women
	Previous surgery
	Cholecystectomy

Environmental factors, particularly tobacco smoking, also play a significant role in the development of pancreatic cancer.7 Smokers have a 2- to 3-fold increased risk of pancreatic cancer. It has been estimated that tobacco smoking contributes to the development of almost 30% of pancreatic cancers. Importantly, smoking cessation can reduce this risk. Indeed, Mulder et al12 have estimated that moderate reduction in smoking in Europe could save almost 68,000 lives that would otherwise be lost to pancreatic cancer by the year 2020.

Other risk factors associated with the development of pancreatic cancer are a high-fat/cholesterol diet, cirrhosis, chronic pancreatitis, diabetes mellitus, and a history of cholecystectomy.13, 14 Everhardt and Wright14 performed a metaanalysis of 20 epidemiologic studies on diabetes and pancreatic cancer. The pooled relative risk (RR) of pancreatic cancer in persons with diabetes for 5 years was double (RR 2.0; confidence interval [CI] 1.3 to 2.2) the risk of persons without diabetes. Gapstur et al15 reported a correlation between elevated plasma glucose levels and pancreatic cancer risk, further suggesting that impaired glucose tolerance, insulin resistance, and hyperinsulinemia are involved in the cause of pancreatic cancer. In a recent cohort study of more than 160,000 health care professionals, Michaud et al16 found that high body mass index (a measure of obesity), height, and reduced physical activity increased pancreatic cancer risk. Activity levels and weight were ascertained before pancreatic cancer detection. They found an excess risk of pancreatic cancer among obese men and women and a direct association between height and the risk of pancreatic cancer. An inverse relationship between moderate physical activity and pancreatic cancer was observed. Walking or hiking 1.5 hours or more per week was associated with a 50% reduction in pancreatic cancer. Physical activity had no effect on pancreatic cancer risk for nonoverweight participants. Likewise body mass index had no effect if the participant was a moderate exerciser. For cigarette smoking, the strongest associations with pancreatic cancer were observed when the pack-years smoked were within the previous 15 years. These findings suggest that smoking cessation, weight loss, and exercise may all reduce pancreatic cancer risk. Factors that have been studied and appear to have no association with the development of pancreatic cancer include moderate alcohol intake, nonhereditary and acute pancreatitis, and coffee intake.

Genetic predisposition 

A growing body of research evidence now suggests that genetic predisposition plays a significant role in pancreatic cancer risk. Most case-control studies have found that patients with pancreatic cancer are more likely to have a relative with pancreatic cancer than are healthy control subjects.17 Similarly, several cohort studies have shown an increased risk of developing pancreatic cancer among individuals who report a family history of pancreatic cancer. Tersmette et al18 have shown that this risk may increase with the number of affected members in the family.19 Segregation analyses suggest that aggregation of pancreatic cancer in families has a genetic rather than an environmental basis.20 Klein et al20 performed complex segregation analysis on 287 families ascertained through an index case diagnosed with pancreatic cancer at The Johns Hopkins Medical Institutions between January 1, 1994, and December 31, 1999. Nongenetic transmission models were rejected (P < .0001) and the most parsimonious model included autosomal dominant inheritance of a rare allele, (still to be identified). Klein et al20 were further able to estimate that approximately 0.5% of the population carry this allele.

Inherited syndromes 

Although accounting for less than 10% of the familial aggregation of pancreatic cancer, at least 6 genetic syndromes caused by germline mutations in genes that are associated with an increased risk of pancreatic cancer have been identified. These are summarized in Table 2 and include the following:

Table 2. Genetic disorders and germ-line genetic alterations associated with familial pancreatic cancer (high-risk syndromes)

	Disorder	Gene (location)	Increased risk of pancreatic cancer
	Hereditary breast and ovarian cancer	BRCA2 (13q12-q13)	×3.5 to 10
	Familial atypical multiple mole melanoma syndrome	p16 (9p21)	×12-20
	Peutz-Jeghers syndrome	STK11/LKB1 (19p13)	×130
	Hereditary nonpolyposis colorectal cancer (Lynch II variant)	hMSH2, hMLH1, others	?
	Hereditary pancreatitis	PRSS1 (7q35)	×50
	Ataxia-telangiectasia	ATM (11q22-23)	Rare

Modified from CJ Yeo, JL Cameron, Pancreatic cancer. Curr Prob Surg 1999;36:57–152, Table 5.

A seventh syndrome, that of pancreatic cancer, pancreatic insufficiency and diabetes mellitus, has recently been described in a family (termed Family X), and the phenotype has been linked to chromosome 4q32-34.32

Data from the national familial pancreas tumor registry 

Most of the cases in which there is a familial aggregation of pancreatic cancer are not related to these high risk inherited syndromes. The National Familial Pancreas Tumor Registry (NFPTR*) was established at Johns Hopkins, with the hope of identifying the causes for the aggregation of pancreatic cancer in families. To date, nearly 1000 families have enrolled in this registry. Early analyses of the kindreds enrolled in the NFPTR have shown that the risk of pancreas cancer is 18-fold greater in first-degree relatives of familial pancreatic cancer cases (at least 2 first-degree relatives with pancreatic cancer in the family), than it is in first-degree relatives of sporadic pancreatic cancer cases (families in which there has been only one member with pancreatic cancer).18 In addition, the increased risk of pancreatic cancer in familial pancreatic cancer kindreds extended to second-degree relatives, as a significantly increased proportion of pancreatic cancer existed in second-degree relatives of familial cases compared with sporadic pancreatic cases (3.7% vs 0.6%, P < .0001).18 When 3 or more first-degree family members in the familial kindred subset were affected, the risk of pancreatic cancer increased 57-fold, with an incidence rate of 301/100,000, compared with the reported U.S. incidence rate of 8.8/100,000.18 Thus, although not fully defined, there is a clear familial predisposition to pancreatic cancer when 2 or more first-degree relatives are affected.

Ductal adenocarcinoma 

Infiltrating ductal adenocarcinoma accounts for almost three fourths of all malignant pancreatic neoplasms. These neoplasms are white-yellow, poorly defined, firm masses that often obstruct and dilate the distal common bile and pancreatic ducts. Microscopically, infiltrating ductal adenocarcinomas are composed of infiltrating epithelial cells forming glands of various shapes and sizes surrounded by dense, reactive, fibrous connective tissue (Fig 2,A).

View Large Image Download to PowerPoint

Fig. 2. Four different types of neoplastic pancreatic disease. Microscopic appearance of infiltrating ductal adenocarcinoma (A), microscopic appearance of PanIN-3 (B), microscopic appearance of a mucinous cystadenoma (C), and gross appearance of an intraductal papillary mucinous neoplasm with in situ carcinoma. Note the ovarian-like stroma in the mucinous cystadenoma (C). The intraductal papillary mucinous neoplasm (D) fills the main pancreatic duct.

There are several histologic variants of infiltrating ductal adenocarcinoma. One of these variants, the so-called mucinous non-cystic adenocarcinoma or colloid carcinoma, shows prominent extracellular mucin production. This finding is more common in infiltrating duct adenocarcinomas arising in association with either IPMNs or MCNs. (See below.) Thus the finding of a colloid carcinoma should alert the pathologist to search for one of these precursor lesions. Such a precursor lesion is important to find because it may have an impact on the complete surgical excision of that neoplasm.52

Another important variant of infiltrating ductal adenocarcinoma is medullary carcinoma. These neoplasms are poorly differentiated adenocarcinomas with a syncytial growth pattern, pushing borders, and an associated lymphoid infiltrate. They may have a better outcome than conventional ductal adenocarcinomas. They may also signal an inherited susceptibility to development of pancreatic carcinoma, for example, through HNPCC syndrome.53 These carcinomas, unlike conventional ductal adenocarcinomas, often show a particular genetic trait called microsatellite instability, may harbor Epstein-Barr virus RNA, and usually are K-ras wild type.53, 54

Molecular data such as these may form the basis for better diagnostic techniques, not only for medullary carcinomas but also for all ductal adenocarcinomas. For instance, we have demonstrated that mutant K-ras shed from carcinomas of the pancreas can be detected in duodenal fluid and in stool of patients with pancreatic carcinoma.55 We have also recently developed an immunohistochemical assay for the DPC4 protein, the product of a gene inactivated in approximately 55% of pancreatic ductal adenocarcinomas that may help establish the diagnosis of cancer in morphologically challenging lesions.56, 57 As our understanding of the molecular genetic alterations in carcinomas of the pancreas improves, the application of molecular biology will help in the analysis of histologically ambiguous lesions and in the detection of early pancreatic neoplasms.

Precursors to ductal adenocarcinoma 

PanINs 

Most ductal adenocarcinomas of the pancreas are believed to arise from PanINs. PanINs are lesions composed of mucin-producing epithelia with varying degrees of cytologic and architectural atypia that involve the small ducts of the pancreas (Fig 2,B). PanINs can be flat (PanIN-1A), papillary without atypia (PanIN-1B), papillary with atypia (PanIN-2), or may even meet histopathologic criteria for carcinoma in situ (PanIN-3).49 We believe that just as there is progression from adenoma to adenoma with high-grade dysplasia to infiltrating adenocarcinoma in the colon, so too is there progression from PanIN-1 to PanIN-2 to PanIN-3 to infiltrating adenocarcinoma in the pancreas. The histologic criteria used to grade PanINs are explained in detail on the World Wide Web at http://pathology.jhu.edu/pancreas_panin.

Several lines of evidence suggest that PanINs are the precursors to infiltrating ductal adenocarcinoma. First, PanINs are frequently found in pancreata with infiltrating cancers. Furthermore, higher-grade PanINs are more common in pancreata from patients with pancreatic cancer than they are in pancreata from patients without pancreatic cancer.58 Second, some PanINs have been observed to progress to infiltrating cancer over time. Brat et al59 reported 3 patients in whom infiltrating ductal adenocarcinomas developed months to years after PanINs were identified in their pancreata. Third, and most convincingly, PanINs display some of the same genetic changes as infiltrating adenocarcinomas. For example, activating point mutations in codon 12 of the K-ras gene, common in infiltrating adenocarcinoma, have been demonstrated in both early and late PanINs. PanINs also harbor mutations in tumor-suppressor genes, namely p16, p53, BRCA2, and DPC4.60 We recently showed that p16 inactivation first occurs primarily in PanINs-2 and –3 and that DPC4 first occurs very late in neoplastic progression, in PanINs-3.61 This work on genetic alterations in PanINs has led us to formulate a histologic-genetic progression model for pancreatic duct adenocarcinoma (Fig 3).

View Large Image Download to PowerPoint

Fig. 3. Histologic-genetic progression model of infiltrating pancreatic ductal adenocarcinoma from PanIN. Printed with permission from Wilentz RE et al. Cancer Res 2000;60:2002–6.

Importantly, this progression model suggests that the molecular detection of precursor lesions and early cancers is indeed possible. For example, like mutant K-ras genes originating in infiltrating ductal adenocarcinomas, mutant K-ras genes shed from PanINs have also been identified in stool, duodenal fluid, and pancreatic juice samples.62

IPMNs 

IPMNs are mucinous, often villous, neoplasms of the pancreas that grow within the larger native ducts of the pancreas and that do not contain ovarian-like stroma (Fig 2,D). Like the other forms of incipient neoplasia in the pancreas, IPMNs can show varying degrees of atypia and can progress. “Intraductal papillary mucinous adenomas” are IPMNs without significant cytologic or architectural atypia. “Borderline IPMNs” show a moderate amount of dysplasia. “Intraductal papillary mucinous neoplasm with carcinoma in situ” is the designation given to those tumors in which the intraductal lesion displays severe cytologic and architectural atypia. The term infiltrating adenocarcinoma arising in association with an IPMN is reserved for those cases in which an invasive cancer is identified.51 As with MCNs, IPMNs should be completely resected, submitted, and histologically examined because of the heterogeneity of these tumors.

IPMNs must be distinguished from both PanINs and MCNs. By definition, IPMNs are grossly visible lesions that involve the larger pancreatic ducts and ductules, whereas PanINs grow within the smaller ducts and are microscopic lesions. Mucinous cystic neoplasms are independent masses that do not grow within the native pancreatic duct system. In addition, IPMNs do not contain the ovarian stroma frequently seen in mucinous cystic tumors. IPMNs occur with approximately equal frequency among men and women, whereas almost all MCNs arise in women. In addition, unlike PanINs and MCNs, patients with IPMNs have a particular endoscopic feature—mucin oozing from the ampulla of Vater during endoscopy.51 Table 4 summarizes the differences between PanINs, IPMNs, and MCNs.

Table 4. PanINs, IPMNs, and MCNs

		PanINs	IPMNs	MCNs
	Age	60s–70s	60s–70s	40s–50s
	Sex	Both men and women	Both men and women	Mainly women
	Location	Head-body-tail	Head	Body-tail
	Native duct system involvement	Small ducts	Large ducts	No involvement
	Endoscopy/radiology	No findings (lesions are microscopic)	Mucin oozing out of ampulla of Vater	Pancreatic tail mass
	Ovarian stroma	No	No	Yes

Not surprisingly, recent genetic studies support a progression model for IPMNs. Activating point mutations in codon 12 of the K-ras oncogene occur in approximately 60% of IPMNs, and these mutations seem to occur more frequently in the more atypical areas of IPMNs.65 Fujii et al66 identified numerous allelic losses in IPMNs; again, these allelic losses occurred more frequently in the more atypical areas of the IPMNs. Iacobuzio-Donahue et al67 recently studied DPC4 expression in IPMNs and found that, in contrast to ductal adenocarcinomas and mucinous cystic neoplasms, virtually all IPMNs express DPC4.

As with mucinous cystic neoplasms, this progression model implies that noninvasive IPMNs follow benign courses. Indeed, current evidence supports this assertion. In contrast, invasive adenocarcinomas arising in the setting of an IPMN can recur and metastasize.51

Other pancreatic neoplasms 

Summary 

Most pancreatic neoplasms arise from histologically recognizable precursors. The 3 most common precursors to invasive cancer—pancreatic intraepithelial neoplasias, mucinous cystic neoplasms, and intraductal papillary mucinous neoplasms—all contain tall columnar mucin-producing cells. In PanIn lesions, these cells line small native pancreatic ductules; in IPMNs, they form a villiform mass within large native pancreatic ducts. In contrast, MCNs contain de novo cysts not connected to the native pancreatic duct system.

The recognition of incipient neoplasia in the pancreas is important, because it provides the opportunity to identify a curable disease. For example, the complete removal of a mucinous cystadenoma can cure a patient of a neoplasm that, if not removed, can progress to an incurable infiltrating cancer. Similarly, the application of modern molecular screening tests has the potential to lead to the identification of early-stage, and therefore surgically treatable, ductal adenocarcinomas and their precursors.

Molecular biology may lead not only to the earlier diagnosis of pancreatic tumors but also to the more accurate diagnosis of these neoplasms. The molecular characterization of pancreaticoblastomas, acinar cell carcinomas, and solid-pseudopapillary neoplasms has revealed that there are separate genetic pathways of neoplastic progression for pancreatic ductal and nonductal neoplasms. Genetic changes may, in conjunction with histologic study, provide important information on tumor type and thus on patient treatment and prognosis.

Oncogenes 

Oncogenes are derived from normal cellular genes called protooncogenes. When activated by mutations or amplifications, these genes possess transforming properties. The names, locations, and mutational frequencies of oncogenes commonly mutated in pancreatic cancer are shown in Table 5.

Table 5. Commonly mutated genes in pancreatic ductal adenocarcinoma

	Gene name	Chromosomal location	Mutational frequency
	Oncogenes
	K-ras	12p	95%
	AKT2	19q	10%–20%
	MYB	6q	10%
	Tumor-suppressor genes
	p16/RB1	9p/13q	95%
	p53	17p	75%
	DPC4	18q	55%
	LKB1/SKT11	19p	5%
	MKK4	17p	4%
	ALK4	12q	2%
	Genome maintenance genes
	BRCA2	13q	7%–10%
	MSI+/TGFBR2	3p	3%
	MLH1	3p	3%

Modified from Su GH, Kern SE. Molecular genetics of ductal pancreatic neoplasia. Curr Opin Gastroenterol 2000;16:419–25.

Tumor-suppressor genes 

Tumor-suppressor genes normally function to restrain cell proliferation. Loss of function of these genes occurs as a result of mutations, deletions, chromosomal rearrangements, methylation, or mitotic recombination. For most tumor suppressor genes, biallelic inactivation is required to cause loss of function, which results in abnormally increased or unregulated cell proliferation. The names, locations, and mutational frequencies of tumor-suppressor genes commonly mutated in pancreatic cancer are shown in Table 1.

Genome-maintenance genes 

Genome-maintenance genes encode for proteins that correct many errors that normally occur when DNA is replicated. When these mismatch repair genes are dysfunctional, errors in DNA replication are not repaired. There are at least 2 types of DNA replication checkpoints that are important for maintaining chromosome stability. The first is the spindle checkpoint, which ensures that the chromatids do not separate until they are properly aligned along the mitotic spindle. The second is the DNA-damage checkpoint, a G2-M checkpoint, which prevents cells with DNA damage from entering mitosis. The names, locations, and mutational frequencies of genome-maintainence genes commonly mutated in pancreatic cancer are shown in Table 5.

Current approaches for detecting pancreatic cancer 

The vast majority of pancreatic cancers are diagnosed at a late, incurable stage. However, early detection of small, resectable cancers may improve the outcome of this deadly disease.105 The optimal approach for early detection of pancreatic cancer is unknown and still under study. An ideal approach would use an imaging test along with molecular markers of neoplastic disease to diagnose a benign, dysplastic precancerous lesion or an early, localized cancer. Ongoing research at Johns Hopkins related to early detection of pancreatic neoplasia in patients with Peutz-Jeghers syndrome and at-risk relatives from familial pancreatic cancer kindreds will provide additional data on the potential benefits of a screening program with this approach. Currently, 2 imaging modalities for screening and early detection include magnetic resonance imaging (MRI)/magnetic resonance cholangiopancreatography, and endoscopic ultrasonography (EUS). Standard radiologic imaging tests such as transabdominal ultrasonography, magnetic resonance imaging, and spiral (helical) computed tomography (CT) scanning used in the community are not sufficiently sensitive for detecting small early cancers, even in patients with symptoms. However, CT imaging continues to improve with the development of multidetector technique and 3-dimensional reconstruction, which allow improved resolution.106 For early detection, EUS may be the imaging modality of choice because it detects smaller pancreatic lesions than those detected with thin section dual phase spiral CT.107, 108 The accuracy of diagnosis of pancreatic cancer approaches 100% for EUS and 92% for dual-phase CT.109 Furthermore, EUS can readily discriminate between solid and cystic lesions (unlike CT scanning) and provide a cytologic diagnosis of minute lesions as small as 2 to 5 mm not visualized by CT scanning, ultrasonography, or magnetic resonance imaging. Fine-needle aspiration (FNA) performed during an EUS procedure also help to establish a diagnosis of malignancy with an accuracy of 90%. Endoscopic retrograde cholangiopancreatography (ERCP)is less likely to detect small tumors, because earlier lesions are associated with minimal effects on the pancreatico-biliary tree. Other investigative tools for the detection of pancreatic cancers include positron emission tomography (PET) and intraductal ultrasound scanning. PET may be helpful for identifying occult metastases.110 Intraductal ultrasonography may be performed as an extension of an ERCP procedure to detect small pancreatic tumors and IPMNs.111

CA19-9, the only commercially available, widely used tumor marker for pancreatic cancer, can be valuable for monitoring the therapeutic response of patients with pancreatic cancer that have elevated serum CA19-9 levels.112 But CA19-9 is of limited value as a screening marker, because approximately 10% to 15% of individuals do not secrete CA19-9 because of their Lewis antigen status. In addition, CA19-9 levels may be within the normal range while the cancer is small and asymptomatic and can be elevated in benign biliary or pancreatic conditions.

Current approaches to survey at-risk individuals use EUS of the pancreas, multidetector CT with thin sections of the pancreas, and serum CA 19-9 measurements. ERCP, EUS-FNA, and other investigations may be performed if abnormalities are found on EUS. Recently, Brentnall et al113 reported their experience with screening of high-risk families. Of 14 patients from 3 families surveyed primarily by EUS, 7 were found to have EUS and ERCP abnormalities suggestive of pancreatic ductal lesions. Pathologic analysis of pancreatectomy resection specimens revealed high-grade PanIN, but no cancers. Because total removal of the pancreas (total pancreatectomy) is a major surgical procedure that results in both endocrine and exocrine insufficiency (requiring insulin and enzyme replacement), its application to patients for prophylaxis of pancreatic cancer requires careful consideration and counseling.

Developing biomarkers for early detection 

Biomarkers of pancreatic cancer are needed for both early diagnosis of individuals with symptoms whose initial workup fails to yield a diagnosis and for use as a screening test to permit the early detection of pancreatic cancer in symptom-free individuals at high risk for development of the disease. Ideally, a screening test should be easy to perform, such as serum testing for PSAlevels to identify prostate cancer or testing for fecal occult blood to screen for colorectal neoplasias. The potential high concentration of DNA and proteins, and the relative lack of other normal constituents like in the serum, make pancreatic juice a potentially optimal specimen to use when screening patients at high risk for pancreatic cancer. Such a practice would be analogous to using sputum to screen for lung cancer114 or nipple aspirates to screen for breast cancer.115 Furthermore, pancreatic juice can be collected during routine upper gastrointestinal endoscopy after secretin stimulation, without the need for ERCP. Even when pancreatic cancer is suspected, imaging tests sometimes fail to identify a pancreatic mass. In this setting, molecular markers could facilitate early diagnosis by aiding in the interpretation of inconclusive cytology specimens obtained by sampling the pancreatic duct or from fine-needle aspirates obtained during EUS.

Biomarkers can be divided into 3 biochemical targets: DNA-based, RNA-based, and protein-based. DNA-based techniques aim to detect cancer-specific DNA alterations. The diagnostic potential of both DNA- and RNA-based markers has improved with the use of quantitative polymerase chain reaction (PCR).Pancreatic cancer DNA can be detected in pancreatic juice samples, in fine-needle aspirates of the pancreas, in the serum, and in the stool of affected patients. For example, mutant K-ras is found in the pancreatic juice or pancreatic duct brushings and in stool of many patients with pancreatic cancer. However, it is not a specific marker, being found among symptom-free smokers and in pancreatic juice from patients with chronic pancreatitis without cancer. In addition, mutant K-ras in plasma is observed late in the natural history of pancreatic cancer, correlating with advanced disease.116, 117 Of the DNA-based abnormalities described in pancreatic cancer, DNA methylation abnormalities may be particularly suitable for use in early detection strategies. First, numerous aberrant methylation events occur during carcinogenesis (eg, methylation of hMLH1 and p16).118 Second, aberrant methylation can be detected in secondary sources with the very sensitive methylation-specific PCR technique. Third, aberrant methylation patterns of affected genes are typically consistent across cancers from a number of different patients and thus can be targeted for diagnostic tests.119 Methylation-specific PCR has successfully been used to identify methylated DNA in blood, saliva, prostatic fluid, and sputum of patients with cancers. Pancreatic carcinomas harbor aberrant methylation of a number of cancer-related genes (ppENK, p16, TSLC1, and others).120, 121 Of these, methylation of ppENK occurs in more than 90% of pancreatic cancers and is not seen in normal pancreas, suggesting that it may be a valuable molecular marker of pancreatic cancer.

As with detection of pancreatic cancer DNA, detection of pancreatic cancer mRNA, may be less useful for serum-based diagnosis and may be more applicable for the analysis of pancreatic juice or fine-needle aspirates. The main RNA-based marker investigated to date has been hTERT.About 90% of cancers express the telomerase hTERT subunit, and <90% of patients with pancreatic cancer have detectable telomerase activity in their pancreatic juice.122 The detection of telomerase enzymatic activity or the hTERT subunit may be helpful in differentiating pancreatic cancer from benign pancreatic disease. However, because telomerase is expressed in inflammatory cells, it may not be sufficiently specific for use as a cancer-screening marker. Several recent studies of gene expression profiling by cDNA and oligonucleiotide microarrays and SAGE, which permit the expression levels of tens of thousands of genes to be monitored simultaneously and rapidly, have generated a long list of genes that are overexpressed at the RNA level in pancreatic cancers compared with normal pancreas.123, 124 Mesothelin, expressed at the RNA and protein level in almost all pancreatic cancers but not in normal pancreas, is one of several such overexpressed genes likely to be tested as a marker of pancreatic cancer.125

Protein-based markers ultimately may have the most application for pancreatic cancer diagnostics. The ultimate goal of such a marker would be a “PSA-test” for pancreatic cancer. In addition to the use of genes overexpressed at the RNA level to identify new protein markers, large-scale analysis of proteins in biologic fluids or cells is an attractive approach for identifying cancer markers. “Proteomics” is the term now used for profiling proteins in biologic samples. One such proteomics technique is SELDI (surface-enhanced laser desorption ionization mass spectrometry), which analyzes protein profiles of samples applied to protein chips.126 SELDI profiling of pancreatic juice led to the identification of markedly elevated HIP/PAP levels in pancreatic juice samples from patients with pancreatic cancer compared with patients with other pancreatic diseases (Fig 4).126

View Large Image Download to PowerPoint

Fig. 4. Representative spectrum examples of SELDI analysis of pancreatic juice samples bound to IMAC-3 cupper ProteinChip array. A peak of <16,570 Da (arrow) was present in the 4 pancreatic juice samples from patients with pancreatic adenocarcinoma (PC4, PC8, PC18, PC24) but absent in the 4 patients with other pancreatic diseases [IPMN; islet cell tumor (ICT); serous cystadenoma (SC)].

The recent advances in proteomics, gene expression analysis and DNA methylation suggest that molecular markers of pancreatic cancer may soon prove useful in the early detection and diagnosis of this disease. As additional advances are made, their most valuble application would appear to be screening of populations at high risk for pancreatic cancer.

Clinical presentation 

Most patients with pancreatic cancer are initially seen with the development of jaundice, which usually occurs as a result of the neoplasm, arising in the head of the pancreas, obstructing the intrapancreatic portion of the common bile duct. Accompanying the jaundice often are dark urine, light stools, weight loss, abdominal pain, and pruritus. Weakness and anorexia may also be present.

At times, pancreatic cancer may present in an unusual manner. New-onset diabetes may be the first clinical feature in approximately 10% of patients.132 Occasionally acute pancreatitis may be the first signal of a pancreatic neoplasm, related to partial obstruction of the pancreatic duct. Such a presentation is being reported frequently in patients with IPMN.133 It is important to consider the diagnosis of pancreatic neoplasia in these uncommon patients diagnosed with pancreatitis, particularly when there is no obvious cause for the acute pancreatitis (such as gallstones or alcohol abuse).

Other symptoms found in a small percentage of patients include nausea and vomiting related to gastroduodenal obstruction. Mechanical obstruction of the proximal duodenum by right-sided neoplasms, or at the ligament of Treitz by cancers of the midbody of the pancreas are often later findings of pancreatic cancer and suggest relatively advanced disease.

The most common physical finding at initial presentation is jaundice. Often, patients with deep jaundice will exhibit cutaneous signs of scratching, related to the pruritis. Hepatomegaly and a palpable gallbladder may also be found. Physical findings in patients with disseminated cancer may include palpable hepatic metastases, left supraclavicular adenopathy (Virchow's node), periumbilical lymphadenopathy (Sister Mary Joseph's nodes), and the finding of drop metastases in the pelvis encircling the perirectal region (Blumer's shelf). Patients with advanced disease may also exhibit cachexia and muscle wasting.

Patients with cancer of the head of the pancreas typically have laboratory study results marked by elevated serum total bilirubin, alkaline phosphatase and γ-glutamyl transpeptidase, with mild elevations of the hepatic aminotransferases. Hepatitis serologic study result are often assessed as part of the workup for jaundice, and they are typically negative. Normochromic anemia and hypoalbuminemia may reflect the chronic nature of the neoplastic process and its nutritional sequelae. In patients with localized cancer of the body and tail of the pancreas, standard laboratory values are usually normal. For these tumors, when liver function test abnormalities do occur, they typically indicate diffuse metastatic disease with involvement of the liver or porta hepatis. It is uncommon for patients with standard ductal adenocarcinoma of the pancreas to have either hyperamylasemia or hyperlipasemia. Patients with IPMN and associated cancer may, however, have elevations of either amylase or lipase. In patients with deep jaundice, the coagulation parameters should be checked, because prolonged exclusion of the bile from the gastrointestinal tract leads to a malabsorption of the fat-soluble vitamins and decreased hepatic production of vitamin K–dependent clotting factors. This can result in prologation of the prothrombin time.

Many different tumor markers have been studied in an attempt to facilitate an early diagnosis of pancreatic cancer. One of the most pressing needs for this disease is a specific and sensitive early tumor marker. At present there are no accurate and reliable serum markers that can be used for the diagnosis of pancreatic cancer. The best of the commercially available markers is the carbohydrate antigen 19-9 (CA19-9).134–136 CA19-9 is a Lewis blood group–related mucin that has been extensively studied in the diagnosis, prognosis, and monitoring of pancreatic cancer. With an upper limit of normal of 37 units/mL, CA19-9 only approaches an 80% accuracy in identifying patients with pancreatic cancer. The combined use of CA19-9 and either ultrasonography, CT scanning, or ERCP improves the diagnostic accuracy of each individual test, to the point that accuracy can approach, but not achieve, 100% (Table 10).

Table 10. The combination of CA19-9 and imaging tests for diagnosis of pancreatic cancer

		Positive predictive value
		Without CA 19-9	CA 19-9 >40 U/mL	CA19-9 >100 U/mL
	Abdominal ultrasonography	62%	100%	100%
	Abdominal CT	71%	89%	100%
	ERCP	62%	80%	100%

From Ritts R et al. Pancreas 1994;9:707–16.

As has been discussed in previous sections of this monograph, one hope for the future involves new developments in the area of early detection, by use of molecular strategies. With gene expression data138 and data from other molecular strategies,139, 140 it can be anticipated that earlier detection of pancreatic cancer will one day be possible.

Diagnostic imaging 

Currently, state-of-the-art CT scanning of the pancreas uses multidetector CT acquisition,141 a technology introduced in the late 1990s. Before this, high-quality spiral (or helical) CT was the preferred noninvasive imaging modality for the diagnosis and staging of pancreatic cancer.142 Multidetector CT incorporates dual-phase imaging in both the arterial and venous phases of enhancement. Water is used as the oral contrast agent of choice. Nonionic contrast medium (120 mL) is administered via a peripheral intravenous catheter at a rate of 3 mL/sec. Slices through the pancreas are obtained every 1.25 mm, with all images being acquired during one 20-second breath hold. For visualizing the study on film, 3- to 5-mm slices are printed on film. However, the 1.25-mm slices are reviewed at a 3-dimensional workstation, by use of a software platform (we use Siemens Virtuoso, Siemens, Iselin, NJ). The addition of 3-dimensional viewing of the data sets improves the detection, staging, and surgical planning.141 Adenocarcinoma of the pancreas typically appears as a low-density (hypodense) mass within the pancreas (Figs 5–7), often best seen on the venous phase of enhancement.

View Large Image Download to PowerPoint

Fig. 5. Late arterial phase of spiral CT scan, with contrast used as oral agent. The kidneys and aorta are contrast enhanced, as is the inferior vena cava. Dilated bile ducts are seen in the liver, and the gallbladder is distended. A large (5 cm) hypodense mass is seen in the head of the pancreas, and the superior mesenteric vein (SMV) is not seen. Additional caudal images confirmed occlusion of the SMV, with numerous mesenteric venous collaterals. This tumor was deemed unresectable, on the basis of the advanced local disease.

View Large Image Download to PowerPoint

Fig. 6. Arterial phase of multidetector CT scan, with water used as oral agent. The kidneys and aorta are contrast enhanced. A 3-cm hypodense tumor mass is seen in the pancreatic uncinate process, anterior to the aorta and inferior vena cava. The tumor abuts the right lateral aspect of the superior mesenteric vein. The superior mesenteric artery is contrast-enhanced, patent and not approached by tumor. This tumor was resected via pancreaticoduodenectomy, with negative resection margins.

View Large Image Download to PowerPoint

Fig. 7. Multidetector CT images from patient with small cancer in head of pancreas. (Top) Sagital 3-dimensional reconstruction shows normal aorta, celiac axis, and superior mesenteric artery. (Bottom) Coronal 3-dimensional reconstruction shows normal liver, gastric fundus, portal vein, as well as intact superior mesenteric artery and vein.

Recent advances in hardware and software have allowed MRI to improve its ability to diagnose and stage pancreatic cancer. These recent advances include high-resolution imaging, fast imaging, volume acquisitions, magnetic resonance cholangiopancreatography, and functional imaging.143, 144 Sequences with high temporal resolution are used for the detection and staging of pancreatic cancer. Arterial and venous phase studies are used to evaluate arterial and venous patency. Because most pancreatic cancers have sparse vascularity and dense cellularity, the tumors appear with low-signal intensity on T1-weighted fat-suppressed images, and diminished enhancement on dynamic contrast-enhanced images (Figs 8 and 9).144

View Large Image Download to PowerPoint

Fig. 8. Two T1-weighted MR images with contrast enhancement with gadolinium. A mass in the head of the pancreas appears as a hypointense area. (From Yeo CJ, Cameron JL. Pancreatic Cancer. Curr Prob Surg 1999;36:57–152, Fig 17).

View Large Image Download to PowerPoint

Fig. 9. Single shot, spin echo MRI-cholangiopancreatogram in patient with obstructive jaundice. Both the common bile duct and the pancreatic duct are dilated, and a hypointense area of tumor is apparent in the periampullary region. (From Yeo CJ, Cameron JL. Pancreatic Cancer. Curr Prob Surg 1999;36:57–152, Figure 18).

ERCP allows direct imaging of the pancreatic duct, the site of origin of most pancreatic cancers. The sensitivity of ERCP for the diagnosis of pancreatic cancer is quite high, with the finding of long, irregular stricture in an otherwise normal pancreatic duct being virtually pathognomonic in the appropriate clinical setting (Fig 10).

View Large Image Download to PowerPoint

Fig. 10. ERCP in patient with obstructive jaundice reveals classic “double-duct” sign. There is evidence of tumor at the genu of the common bile duct and pancreatic duct. (From Yeo CJ, Cameron JL. Pancreatic Cancer. Curr Prob Surg 1999;36:57–152, Figure 19).

EUS is a recently developed and now established technique for obtaining images of the pancreas.145, 146, 147, 148, 149, 150 The close proximity of the transducer to the region of interest permits the use of higher ultrasound frequencies than do transcutaneous techniques, thereby improving image resolution (Fig 11).

View Large Image Download to PowerPoint

Fig. 11. EUS image with linear array echoendoscope, revealing a mass in the head of the pancreas with no vascular invasion of the superior mesenteric artery (SMA), superior mesenteric vein (SMV), or portal vein (portal). (From Yeo CJ, Cameron JL. Pancreatic Cancer. Curr Prob Surg 1999;36:57–152, Figure 24).

EUS-FNA appears to offer some advantage over other techniques in cases where a tissue diagnosis of pancreatic cancer is required before treatment. (Of note, unless protocol-based neoadjuvant chemotherapy or chemoradiation therapy is planned, in most patients with a presumed resectable tumor seen by imaging, such a tissue diagnosis is not necessary). Gress et al153 evaluated EUS-FNA in 102 patients with suspected pancreatic cancer and previously-negative CT- or ERCP-guided cytologic examinations. The sensitivity for EUS-FNA cytologic study in this difficult setting was 93%, the specificity was 83%, and the complication rate was only 3%. Although these results clearly demonstrate that EUS-FNA is able to accurately yield a tissue diagnosis of pancreatic cancer (when CT- and ERCP-guided cytologic evaluation has failed), it must be stressed that patients with resectable lesions suspicious for pancreatic cancer do not require such a tissue diagnosis before surgical resection.

Although CT, MRI, and EUS remain the mainstays of obtaining images of patients with suspected pancreatic cancer, the newer technique of PET may play an increasing role in the future.154, 155 PET uses the increased metabolism of glucose by pancreatic cancer cells as the basis of imaging. Current PET scanning for pancreatic cancer uses fluorine-18 (a positron-emitting tracer) as a glucose-like substrate in vivo. Fluorine-18 has a two hour physical half-life. When labeled to flurodeoxyglucose (FDG), it is rapidly taken up by malignant tumor cells. FDG-PET has been reported to be highly sensitive and specific for pancreatic cancer, with results that surpass CT: sensitivity (PET vs CT) 92% vs 65%, specificity 85% vs 62%.156 Importantly, FDG localizes not only at tumor sites, but at sites of inflammation and infection. One way to differentiate benign from malignant lesions by PET uses a semiquantitative index called the “standard uptake value” (SUV). The SUV is an index derived from the decay-corrected dose, corrected for the patient's body weight. Generally, lesions with an SUV greater than 2.5 correlate with malignancy. FDG-PET has been shown, in patients with pancreatic cancer, to give information relevant to prognosis (based on high vs low SUV), and to add to the diagnostic accuracy of CT and ERCP in detecting tumor dissemination.155 Unfortunately, currently Medicare does not reimburse for PET imaging in patients with pancreatic cancer, thus its use has been limited.

Histopathologic diagnosis 

The use of pancreatic biopsy (percutaneous or endoscopic) in the diagnostic workup of a patient with a suspected pancreatic cancer has advocates, as well as vocal opponents. Although percutaneous biopsy is generally safe, serious complications such as hemorrhage, pancreatitis, fistula, abscess, and death have been reported. Additionally, there have been reports of tumor seeding along the subcutaneous tract of the needle, and concerns regarding tumor dissemination by the act of capsular puncture of the neoplasm. Numerous studies have evaluated the results of percutaneous core biopsy or FNA cytologic study in the assessment of patients with pancreatic masses. Although a diagnosis of cancer may be highly specific (approaching 99%), the sensitivity is generally lower, ranging from 50% to 70%. This is because, although a histologic diagnosis of adenocarcinoma is quite reliable, malignancy cannot be excluded with certainty when the pathologist cannot find malignant cells in the specimen (Fig 12).

View Large Image Download to PowerPoint

Fig. 12. FNA of pancreatic adenocarcinoma: cluster of malignant cells with coarse nuclear chromatin, variable nuclear size and shape, and disorderly nuclear crowding. (Original magnification × 400.) (From Tsiotos GG, Sarr MG. Diagnosis and clinical staging of pancreatic cancer. In: Howard JM, Idezuki Y, Ihse I, Prinz RA, editors. Surgical diseases of the pancreas. 3rd ed. Baltimore: Williams and Wilkins; 1998. p. 504, Fig 52.5).

Laparoscopy 

The role of diagnostic staging laparoscopy in patients with pancreatic cancer remains controversial.157 Proponents believe that laparoscopy can identify a substantial number of unresectable patients (with advanced disease) and therefore recommend that it be applied to all patients. Opponents believe that the costs of such a practice (money, patient risk, resources) outweigh the benefit derived by the small number of patients for whom diagnostic laparoscopy is useful. The rationale for the use of laparoscopy comes from data that indicate that between 20% and 40% of patients staged by other modalities (CT, MRI, EUS) will be determined to have unanticipated peritoneal or small liver metastases by laparoscopy. However, part of the rationale for using laparoscopy involves a presumed equivalence of nonoperative palliation with operative palliation in patients with pancreatic cancer. Thus routine laparoscopic staging only makes sense if the percentage of patients discovered to have disseminated or unresectable disease remains high (20% to 40%) in the era of current multidetector CT, and if patients who are deemed to be unresectable and spared laparotomy can be optimally palliated without operation. Diagnostic staging laparoscopy can be performed with minimal morbidity and mortality on an outpatient basis, by use of a 30-degree angled laparoscope and evaluating the entire peritoneal surface, the pericolic gutters, the hemidiaphragms, the pelvis, as well as the surfaces of the liver. Specimens of peritoneal or omental nodules or liver implants undergo biopsy under direct vision. Enlarged lymph nodes can also be sampled successfully via biopsy or by fine-needle aspiration. There are varying degrees of expertise in the application of laparoscopy, with some highly experienced groups performing a more extensive laparoscopic evaluation including examination of the hilus of the liver and mesenteric and celiac vessels, and adding the use of laparoscopic ultrasonography for the detection of nonsurface hepatic metastases, vascular invasion, or deep-seated lymphadenopathy.158–163

Patients diagnosed with obstructive jaundice caused by tumors in the head of the pancreas typically have only a 15% to 20% incidence of unexpected intraperitoneal metastases after modern staging studies. In contrast, patients with cancer of the body and tail of the pancreas have unexpected peritoneal metastases in up to 50% of patients. On the basis of these data, staging laparoscopy appears to be best supported for patients with cancer of the body or tail of the pancreas. In these patients the primary tumor does not typically cause biliary or gastric outlet obstruction, and therefore patients do not routinely require palliation of biliary or gastric obstruction. Thus in this group of patients laparoscopy can spare the patient an unnecessary laparotomy because there is little role for operative palliation. However, the role of routine preoperative staging laparoscopy is controversial in patients with localized right-sided tumor by modern imaging, who present with obstructive jaundice, symptoms of gastric outlet obstruction, and tumor-related abdominal and back pain. Many surgeons believe that such patients are best managed via resection if possible, or operative palliation to include biliary-enteric bypass, gastrojejunostomy and alcohol celiac nerve block. Preoperative staging laparoscopy would serve no purpose in such a setting.

A recent report by Barriero et al157 underscores this practice. A retrospective review of 188 patients with pancreatic or periampullary malignancy who underwent both high-quality CT and laparotomy over a 3-year period was performed. The overall respectability rate for all periampullary cancers was 67%, compared with only 18% for left-sided pancreatic cancers. After patients undergoing operative palliation were excluded, a nontherapeutic laparotomy would have been avoided by the use of diagnostic laparoscopy in only 2% of patients with periampullary tumors. In contrast, for left-sided pancreatic tumors, 53% of patients would have benefited from laparoscopy, and 35% of patients could have avoided an unnecessary laparotomy.

Nonoperative palliation 

Nonoperative management is appropriate in patients with pancreatic cancer who are determined to have distant metastases, unresectable local disease, or disseminated intraabdominal tumor, or in patients with acute or chronic debilitating diseases that make anesthesia and surgery prohibitive. The exception to these indications for nonoperative management are those patients with symptomatic upper gastrointestinal obstruction (from tumors that obstruct the duodenal C loop or the ligament of Treitz), where nonoperative palliation may not be reliable, and where gastrojejunostomy may be appropriate. In patients who are to be managed without operation, a tissue diagnosis can be obtained via biopsy of distant metastases or local disease. As noted previously, preoperative biopsy is avoided in patients with localized, apparently resectable tumors, because the pathologic information gained by such an invasive procedure does not influence the decision to explore the patient.

Biliary obstruction 

Jaundice is present in most patients with pancreatic cancer. If untreated, obstructive jaundice can result in progressive liver dysfunction, liver failure, and early death. Furthermore, the pruritus associated with obstructive jaundice can be symptomatically debilitating and rarely responds to medications. Biliary decompression can now be achieved either by endoscopic or percutaneous transhepatic techniques in nearly all patients who are not candidates for surgical intervention.

The technique of endoscopic biliary stent insertion for palliation of malignant obstructive jaundice was first described in 1980.164 The method and equipment are now well standardized, and a technical success rate exceeding 90% should be expected from those performing such endoscopic stenting regularly. Typically, after the endoscopic visual inspection, deep biliary cannulation is attempted with or without the assistance of a guide wire or sphincterotome. Occasionally, precut sphincterotomy with a needle knife may be necessary. Once deep biliary cannulation has been accomplished, a guide wire is manipulated above the malignant stricture (Fig 13) and a 7F or 10F plastic endoprosthesis is secured in position by being pushed over the guide wire (Fig 14).

View Large Image Download to PowerPoint

Fig. 13. Image obtained at ERCP, with guide wire passed cephalad through distal common bile duct obstruction caused by cancer in head of pancreas. (From Yeo CJ, Cameron JL. Pancreatic Cancer. Curr Prob Surg 1999;36:57–152, Figure 26).

View Large Image Download to PowerPoint

Fig. 14. Image obtained immediately after successful ERCP. A 10F plastic endoprosthesis has been placed across a malignant distal bile duct obstruction. A markedly dilated intrahepatic biliary tree is visualized. (From Yeo CJ, Cameron JL. Pancreatic Cancer. Curr Prob Surg 1999;36:57–152, Figure 27).

After stent placement, serial liver function tests are performed to confirm a decline in the serum bilirubin. Complications of endoprosthesis placement can be categorized as early or late complications. Early complications include cholangitis, pancreatitis, bleeding, and bile duct or duodenal perforation. Late complications include stent occlusion, cholecystitis, delayed duodenal perforation and stent migration. The major late complication remains stent occlusion, which may occur anywhere from days to many months after placement. The occurrence of plastic stent occlusion, with its principal complication of cholangitis and recurrent jaundice, has led most endoscopists to favor planned stent removal and replacement. Although the optimal time interval for stent exchange has not been determined, in practice this has typically been scheduled at 3- to 6-month intervals. A report concerning stents 10F to 11.5F in size noted occlusion rates of 4.2% and 10.8% at 3 and 6 months, respectively, suggesting that an exchange interval of up to 6 months may be appropriate.165 Metallic expandable endoprostheses have been developed by a number of manufacturers, all first deployed through the percutaneous transhepatic route, but with recent modifications to allow endoscopic placement. Once fully deployed, metallic endoprostheses become embedded in the wall of the bile duct and should be considered permanent (although they can be removed at surgery), effectively eliminating the problem of stent migration, and reducing the incidence of stent occlusion. However, tumor ingrowth remains a problem with metallic endoprostheses, causing late stent occlusion. The typical remedy for occlusion of a metallic stent is the insertion of a second metallic or plastic endoprosthesis through the existing lumen of the occluded stent, to allow adequate biliary drainage.

The use of a percutaneous transhepatic biliary drainage technique was first reported in 1974 as an extension of diagnostic transhepatic cholangiography, allowing bile diversion into the duodenum in patients with biliary obstruction.166 The techniques of transhepatic cholangiography and percutaneous biliary drainage are now well standardized and widely disseminated.167 Diagnostic cholangiography first defines the site of bile duct obstruction (Fig 15) and serves as a road map for biliary drainage.

View Large Image Download to PowerPoint

Fig. 15. Percutaneous transhepatic cholangiogram in patient with obstructive jaundice. There is a malignant obstruction of the biliary tree at the level of the junction of the common hepatic duct and cystic duct. Staples are visible from a remote laparotomy for peptic ulcer disease. (From Yeo CJ, Cameron JL. Pancreatic Cancer. Curr Prob Surg 1999;36:57–152, Figure 28).

In most cases, biliary drainage with an internal-external catheter serves as the initial management, with passage of the drainage catheter through the obstruction into the duodenum possible in more than 90% of patients. Initial stiff drainage catheters can subsequently be exchanged for larger diameter, softer catheters. Subsequent management involves maintenance of internal-external drainage catheters (Fig 16), or the percutaneous placement of a totally indwelling endoprosthesis.

View Large Image Download to PowerPoint

Fig. 16. Cholangiogram obtained after placement of internal-external percutaneous transhepatic biliary drainage catheter. The catheter transverses the obstruction in the head of the pancreas. The tip of the catheter resides in the duodenum, distal to the ampulla. (From Yeo CJ, Cameron JL. Pancreatic Cancer. Curr Prob Surg 1999;36:57–152, Figure 29).

Overall, percutaneous transhepatic biliary drainage can be successfully performed in approximately 95% of patients diagnosed with biliary obstruction. Complications of percutaneous transhepatic catheter drainage include stent occlusion, hemobilia related to the transhepatic route, bile peritonitis, bile pleural effusion, cholangitis, pancreatitis and acute cholecystitis.167

The available data support the use of an endoscopic method as the primary method for nonoperative palliation of jaundice in patients with pancreatic cancer (Table 11).

Table 11. Results of percutaneous and endoscopic stent placement in patients with malignant bile duct obstruction

		Percutaneous stent (n = 490)		Endoscopic stent (n = 689)
		Range	Mean	Range	Mean
	30-day mortality rate (%)	6–33	9	37518	14
	Hospital stay (days)	37638	14	37340	7
	Success rate (%)	76–100	92	82–100	90
	Early complication (%)	4–67	16	8–34	21
	Late complication (%)	7–38	28	13–45	28

From Watanapa P, Williamson RCN. Br J Surg 1992;79:8–20, p. 14, Table 5.

A prospective randomized trial,168 and numerous other comparisons reviewed by Watanapa and Williamson169 have shown a comparable success rate for the endoscopic approach, generally associated with a lower degree of procedure-related death and morbidity.170

Operative palliation 

Palliative surgery for pancreatic cancer is appropriate in patients with unresectable disease discovered at the time of potentially curative laparotomy, or in patients at good risk whose tumor-related symptoms are poorly alleviated via nonoperative means. Palliative surgery is designed to relieve biliary obstruction, avoid or treat duodenal obstruction, palliate tumor-associated pain, and improve quality of life.

The surgical options for palliation of obstructive jaundice all include some form of an internal biliary bypass. The 3 most common techniques currently used include choledochoduodenostomy, cholecystojejunostomy, or hepatico- (choledocho) jejunostomy. Although choledochoduodenostomy provides effective relief of obstructive jaundice in a number of benign conditions, it has generally been avoided in patients with pancreatic cancer because of concerns regarding the proximity of the anastomosis to the tumor, with the possibility of recurrent jaundice. Although cholecystojejunostomy has been advocated by some surgeons because it can be performed quickly (and can be done laparoscopically) and does not require dissection of the extrahepatic tree, the data do not support its use. Rather, the preferred technique is choledocho- (or hepatico-) jejunostomy, with the gallbladder being removed before mobilization of the biliary tree (Fig 17).

View Large Image Download to PowerPoint

Fig. 17. Illustration depicts anatomy after one method of palliative intervention. The biliary-enteric anastomosis is shown as a retrocolic end-to-side hepaticojejunostomy with a jejunal loop. A jejunojejunostomy is performed below the transverse mesocolon, to divert the enteric stream away from the biliary-enteric anastomosis. Also shown is a retrocolic gastrojejunostomy. (From Cameron JL, Atlas of Surgery, Volume 1, B.C. Decker, Toronto, 1990, p 427, Image V).

At the time of diagnosis of pancreatic cancer, approximately one third of patients will have symptoms of nausea and vomiting. Although true mechanical obstruction of the duodenum occurs much less frequently, as the malignant disease progresses, duodenal obstruction may occur in a number of patients. Obstruction can occur at either the duodenal C loop by cancer in the head of the pancreas, or at the ligament of Treitz by a cancer of the body of the pancreas. Over the years, information has accrued regarding the natural history of duodenal obstruction associated with pancreatic cancer. Sarr and Cameron172 reviewed more than 8000 surgically managed patients and found that 13% of patients who did not undergo gastrojejunostomy at their initial operation required a gastrojejunostomy before their death, and an additional 20% of patients died with symptoms of duodenal obstruction. A review by Singh and Reber173 found that 21% of patients required a gastrojejunostomy late in the course of their disease. Additionally, an analysis of more than 1600 cases found that 17% of patients who underwent biliary bypass alone had development of duodenal obstruction at a mean of 8.6 months after operation and required subsequent gastric bypass.169

To date, only one prospective, randomized trial by Lillemoe et al174 has evaluated the role of prophylactic gastrojejunostomy in patients found at laparotomy to have unresectable disease. Eighty-seven patients without evidence of preoperative duodenal obstruction or intraoperative tumor encroachment around the duodenum were randomized to receive either a prophylactic retrocolic gastrojejunostomy (n = 44) or no gastrojejunostomy (n = 43). The postoperative mortality rate (0%), morbidity rate (32% to 33%) and length of hospital stay (8 days) were similar, as were the mean survival (8 months). However, 8 of 43 patients (19%) without gastrojejunostomy developed late gastric outlet obstruction requiring intervention (gastrojejunostomy in 7, duodenal stent in 1), whereas no patient in the prophylactic gastrojejunostomy group required such intervention (P < .01). On the basis of these data and the results of previous retrospective analyses, we recommend the performance of a retrocolic gastrojejunostomy in patients who are found at laparotomy to have unresectable right-sided pancreatic cancer.

We typically perform a retrocolic, isoperistaltic loop gastrojejunostomy, using the jejunum 20 to 30 cm beyond the ligament of Treitz, and placing the horizontal gastrotomy somewhat posterior, in the most dependent portion of the gastric greater curvature (Fig 17). Using this technique, the incidence of early delayed emptying appears to be low and hospital discharge is not delayed.174 Importantly, vagotomy is not performed for the palliation of pancreatic cancer, because it may further contribute to delayed gastric emptying. Instead, routine gastric acid secretory inhibition agents are used to prevent marginal ulceration, such as histamine H2-receptor antagonists or proton pump inhibitors.

The abdominal and back pain associated with an unresectable pancreatic cancer can be the major debilitating symptom for the patient. At the time of palliative surgery, this symptom can be addressed by intraoperative chemical splanchnicectomy. The use of intraoperative chemical splanchnicectomy for this indication was first introduced by Copping et al175 in 1969 and was reported in 41 patients in 1978. In this uncontrolled study, 88% of patients with pain caused by pancreatic cancer were reported to experience relief of pain after chemical splanchnicectomy. Only one prospective, randomized placebo-controlled study of intraoperative chemical splanchnicectomy has been reported. In this study at Johns Hopkins, chemical splanchnicectomy was performed by injection of either 20 mL of 50% alcohol or a saline solution placebo on either side of the aorta at the level of the celiac axis (Fig 18).

View Large Image Download to PowerPoint

Fig. 18. Technique of alcohol celiac nerve block. Twenty milliliters of 50% alcohol are injected on each side of the aorta (Ao) at the level of the celiac axis. IVC, Inferior vena cava. (From Lillemoe KD et al. Ann Surg 1993;217:447–57).

A recent review describes the Johns Hopkins experience with surgical palliation of unresectable pancreatic and periampullary adenocarcinoma.177 Over a 6-year period, 256 patients underwent operative palliation (Table 12).

Table 12. The Johns Hopkins experience with surgical palliation (December 1991–December 1997; n = 256 patients)

	Age	64 years
	Sex	57% male
	Presenting symptoms
	Abdominal pain	64%
	Jaundice	57%
	Procedures
	Chemical splanchnicectomy	75%
	Biliary and gastric bypass	51%
	Gastric bypass	19%
	Operative time	3.9 hours
	Transfusions (mean)	0
	Operative mortality rate	3.1%
	Overall morbidity rate	22%
	Postoperative length of stay	10 days
	Median survival	6.5 months
	1-year survival rate	25%
	2-year survival rate	9%

From Sohn TA, et al. J Am Coll Surg 1999;188:658–69.

Pancreaticoduodenectomy for tumors of the head, neck, or uncinate process 

In 1912 Kausch, a German surgeon from Berlin, reported the first successful resection of the duodenum and a portion of the pancreas for an ampullary cancer.178 More than 2 decades later, Whipple et al179 in New York City reported 3 cases of pancreaticoduodenal resection. Although the early reports of Kausch and Whipple describe pancreaticoduodenal resections that spared the pylorus and retained the entire stomach, by the mid 1970s pancreaticoduodenectomy was most commonly performed in combination with a distal gastrectomy. In 1978 Traverso and Longmire180 repopularized the concept of pylorus preservation during pancreaticoduodenectomy. Several reports have reviewed large experiences with pylorus preservation, and this operation is the most typical variant of pancreaticoduodenectomy performed currently (Fig 19).

View Large Image Download to PowerPoint

Fig. 19. Pylorus-preserving pancreaticoduodenectomy. Top left: The structures resected include the duodenum (except for the initial 1 to 2 cm beyond the pylorus); head, neck and uncinate process of the pancreas, with tumor (black); gallbladder; and distal extrahepatic biliary tree. Top right: The structures retained include the entire stomach, pylorus, proximal 1 to 2 cm of duodenum, body and tail of the pancreas, proximal biliary tree, and jejunum distal to the ligament of Treitz. Bottom: The reconstruction is shown as a proximal end-to-end pancreaticojejunostomy, hepaticojejunostomy decompressed via a percutaneous transhepatic catheter and a distal duodenojejunostomy. (From Yeo CJ, Cameron JL. The pancreas. In: Hardy JD, editor. Hardy's textbook of surgery. 2nd ed. Philadelphia: JB Lippincott; 1988. p 718, Fig 28-9).

Complications 

The operative mortality rate for pancreaticoduodenectomy is currently less than 3% in major surgical centers with significant experience with the procedure.181, 182, 191, 192 Leading causes of postoperative in-hospital death include postoperative sepsis, hemorrhage, and cardiovascular events. In contrast to the low mortality rate, the incidence of postoperative complications can approach 40% to 50%.192–194 The leading causes of morbidity include early delayed gastric emptying, disruption or failure of healing of the pancreatic anastomosis (pancreatic fistula), intraabdominal abscess, hemorrhage and others (Table 13).

Table 13. Complications after pancreaticoduodenectomy

	Common
	Delayed gastric emptying
	Pancreatic fistula
	Intraabdominal abscess
	Hemorrhage
	Wound infection
	Metabolic
	Diabetes
	Pancreatic exocrine insufficiency
	Uncommon
	Fistula
	Biliary
	Duodenal
	Gastric
	Organ failure
	Heart
	Liver
	Lung
	Kidney
	Pancreatitis
	Marginal ulceration

From Yeo CJ. Surg Clin North Am 1995;75:913–24.

These complications may have minimal impact on postoperative length of hospital stay or they may prolong hospitalization and require either interventional radiologic techniques194 or reoperation.

Distal pancreatectomy for tumors of the body and tail 

A minority of patients with pancreatic cancer have tumors arising in the body and tail of the gland. Such primary tumors in the left side of the pancreas do not obstruct the bile duct and thus do not present with early jaundice. Typically, diagnosis of these tumors is delayed. Tumors of the body and tail are often larger than tumors of the head and are associated with a much higher incidence of metastatic disease. As a result, the likelihood that curative resection will be possible is lower for left-sided, as compared with right-sided primaries. However, if the diagnosis is made when the tumor is localized and not encasing the celiac axis, the superior mesenteric vessels, or the portal vein, then resection remains a surgical option. Importantly, involvement of either the splenic artery or vein does not alone render the patient unresectable, because the entirety of these vessels can be resected en bloc with the tumor. As has been mentioned previously, in addition to routine staging studies to include abdominal CT or MRI, there appears to be an important role for staging laparoscopy in patients with left-sided tumors. Should staging studies fail to reveal evidence of disseminated tumor or unresectable local disease, then exploration is appropriate.

The entire abdomen is explored to search for metastatic disease. A careful search of the liver for metastatic deposits is undertaken, as well as a thorough evaluation of all the peritoneal surfaces. The lesser omentum is opened to allow assessment of the celiac axis and periaortic region. The ligament of Treitz is carefully evaluated, because tumors in the body of the pancreas may invade the fourth portion of the duodenum at the ligament. Additionally, the gastrocolic ligament should be opened to allow assessment of the body and tail of the pancreas and better assessment of the tumor's proximity to the ligament of Treitz and to the superior mesenteric vessels.

Localized tumors without extensive vascular or retroperitoneal involvement are appropriate for resection. Involvement of the splenic artery and vein does not indicate unresectability. Splenic preservation is not indicated when the resection is being performed for pancreatic adenocarcinoma, therefore the spleen is mobilized out of the retroperitoneum, often with early ligation of the splenic artery. The short gastric vessels along the gastric greater curvature require division, as do the vessels within the splenocolic ligament. Mobilization of the spleen from the retroperitoneum facilitates dissection of the tail of the pancreas from the retroperitoneum (Fig 20).

View Large Image Download to PowerPoint

Fig. 20. Illustration near completion of distal pancreatectomy and splenectomy for large tumor in body of pancreas. The spleen and tail of the pancreas have been mobilized out of the retroperitoneum. The pancreatic parenchyma is being divided by use of electrocautery. (From Cameron JL. Atlas of surgery. Vol 1. Toronto: BC Decker; 1990. p. 435, Image H).

The resectability rates for adenocarcinoma of the body and tail of the pancreas in the era before routine staging laparoscopy were approximately 10%. The routine use of staging laparoscopy to identify metastases not visualized by CT or MRI have improved the resectability rates. Overall patients undergoing left-sided pancreatic resection have median survival rates ranging from 7 to 14 months, with 5-year survival rates of approximately 15% or less.182, 199, 200 A comparison between results for right-sided pancreatic resection (pancreaticoduodenectomy) and left-sided pancreatic resection (distal pancreatectomy) is shown in Table 14.

Table 14. Right-sided versus left-sided pancreatic resection: recent Johns Hopkins Experience (1984–1999)

		Right-sided (Pancreaticoduodenectomy) (n = 564)	Left-sided (Distal pancreatectomy) (n = 52)	P value
	Tumor diameter	3.1 cm	4.7 cm	<.001
	Positive resection margins	30%	20%	NS
	Positive lymph node status (N1)	73%	59%	0.03
	Postoperative mortality rate	2.3%	1.9%	NS
	Overall complications	31%	25%	NS
	Median postoperative length of hospital stay	11 days	7 days	NS
	Survival
	1 year	64%	50%	NS
	5 years	17%	15%
	Median	18 mo	12 mo

From Sohn TA et al. J Gastrointestinal Surg 2000;4:567–79.

Combined modality therapy in the adjuvant setting 

The current standard of 5-fluorouracil (5-FU)–based combined modality chemoradiotherapy has evolved from in vitro data, animal studies, and a series of human studies, most notable being those from the Gastrointestinal Tumor Study Group (GITSG). This study, with split-course irradiation used in modest doses with concurrent bolus 5-FU followed by maintenance 5-FU, demonstrated a survival advantage for the therapy in comparison to surgery alone.201 Although criticized for slow and limited accrual, this study was the first to document that adjuvant therapy after surgical resection for pancreatic surgery prolonged survival. Additional studies by the GITSG demonstrated the benefit of combined chemoradiotherapy versus chemotherapy alone or radiation therapy alone for patients with local regionally-advanced unresectable disease.202–204

A number of groups have further developed this approach (Table 15).

Table 15. Recent adjuvant studies in pancreatic cancer

	Study	Regimen	No. of patients	Median survival (mos)
	UCLA	5FU, MMC, LV, dipyridamole	38	15.5
	Johns Hopkins	Standard regimen	99	21
		Intensive regimen	21	17.5
		Observation	53	13.5
	Johns Hopkins	50 Gy, 5FU, MMC, LV, dipyridamole	39	16
	Stanford	54 Gy and 5FU CI	52	32
	Virginia Mason	54 Gy and 5FU CI, cisplatin, IFN	33	45
	EORTC	40 Gy in 2 modules and 5FU vs observation	218	24.5 vs 19
	ESPAC-1	Chemotherapy vs observation	541	19.7 vs 14 15.5 vs 16.1
		Chemoradiation vs observation
	RTOG-9704	5FU CI, 5FU/XRT (50 Gy), 5FU × 2 vs Gem, 5FU/XRT, Gem × 3	518	Study to close July 2002

5FU, 5-Fluorouracil; LV, leucovorin; CI, continuous infusion; MMC, mitomycin C; IFN, interferon-α; Gy, Gray; EORTC, European Organization for Research and Treatment of Cancer.

Attempting to enhance the activity of chemotherapy in pancreatic cancer, other agents have been examined in combination with 5-FU. Mitomycin-C (MMC) is an antitumor antibiotic with activity in several gastrointestinal cancers, including pancreatic cancer. The UCLA group has published their experience with MMC (10 mg/m2 administered intravenously every 6 weeks) and 5-FU (200 mg/m2/d administered via continuous infusion), in combination with leukovorin (30 mg/m2 weekly)and dipyridamole (75 mg orally daily) in 38 patients with locally advanced pancreatic carcinoma.207 There were noted to be 14 partial responders with 1 complete response. The median survival for all patients was 15.5 months, which is an improvement over historical data for local-regionally advanced disease. This regimen has subsequently been applied to pancreatic cancer in combination with radiotherapy.208 The Johns Hopkins group has recently presented data on 39 patients with resected pancreatic cancer treated with combined radiotherapy (50 Gy in 25 fractions with a planned 2-week break after 25 Gy) and chemotherapy consisting of 5-FU 400 mg/m2 D1-3, MMC 10 mg/m2 D1, leukovorin 20 mg/m2 D1-3 and dipyridamole 75 mg administered orally 4 times daily D0-4, administered on weeks 1 and 4. One month after combined chemoradiotherapy, patients received 4 additional cycles (4 months) of the same chemotherapy alone. At 12.6 months median follow-up, median survival was 16 months.208

The Stanford group has recently published their experience in 52 patients with resected pancreatic cancer, using combined radiotherapy (45 Gy to tumor bed and nodes in 1.8 Gy fractions with boost to total of 54 Gy if surgical margins were positive) and chemotherapy (5-FU 200-250 mg/m2/d administered without break throughout radiation therapy). All patients were able to complete therapy without grade IV toxicities. With median follow-up of 24 months, the median survival was reported to be 32 months.209

Recently the Virginia Mason Clinic published their experience in 33 patients with resected pancreatic adenocarcinoma who received combined radiotherapy (external beam at a dose of 4500 to 5400 cGy in standard fractions d1-35) and chemotherapy (5-FU 200 mg/m2/d as continuous infusion, weekly cisplatin 30 mg/m2 intravenous bolus, and interferon-α 3 million units administered subcutaneously every other day) during radiation.210 After combined modality chemoradiotherapy, chemotherapy alone was administered (5-FU 200 mg/m2/d as continuous infusion) in two 6-week courses during weeks 9 to 14 and 17 to 22. There were significant grade III/IV gastrointestinal toxicities, including vomiting, mucositis, diarrhea, and gastrointestinal bleeding. With a median follow-up of 26 months, the 2-year survival rate was 84%, with a median survival of 45 months.210 These encouraging data await further confirmation.

In July 2002 the Radiation Therapy Oncology Group (RTOG) closed study R97-04. This phase III study of 518 patients with pancreatic cancer randomized between 5-FU continuous infusion (250 mg/m2/d for 3 weeks) followed by 5-FU continuous infusion (250 mg/m2/d) during radiation therapy (50.4 Gy in 1.8 Gy/fractions) followed by 2 cycles of 5-FU continuous infusion, versus gemcitabine 1000 mg/m2 weekly × 3 followed by 5-FU continuous infusion during radiation therapy followed by 3 cycles gemcitabine alone. The experimental question being asked was whether gemcitabine before and after 5-FU–based chemoradiotherapy would be more efficacious than continuous infusion of 5-FU before and after the same 5-FU–based chemoradiotherapy. In 1997, when this study was designed, there was inadequate knowledge regarding how to safely administer gemcitabine concurrently with irradiation to allow for concurrent gemcitabine and radiotherapy. This study was the first North American cooperative group trial since the GITSG trial. Although the survival results for this trial will not be known until sometime in 2003, a number of important observations have already resulted. These include that neither arm was observed to have unacceptable acute toxicity during the trial, that accrual was quite rapid (12 to 14 patients per month), reflecting both the support of the Eastern Cooperative Oncology Group (ECOG) and the Southwest Oncology Group, and the willingness of patients and their physicians to participate in adjuvant trials for pancreatic cancer.

In spite of a growing body of literature supporting the benefit of adjuvant combined modality therapy after resection in patients with pancreatic cancer, adjuvant chemoradiation has not been universally accepted as standard of care. One of the criticisms has been that none of these studies included an observation-only arm. Two European studies have demonstrated contrasting conclusions.

A European Organization for Research and Treatment of Cancer trial randomized 218 patients with pancreatic and nonpancreatic periampullary adenocarcinoma 2 to 8 weeks after potentially curative resection to either observation or to combined radiotherapy (40 Gy with a 3 or 4 field technique in 2 Gy fractions with a 2-week break at mid-treatment) and chemotherapy (5-FU administered as a continuous infusion 25 mg/kg/d during the first week of each 2 week radiation therapy module only).211 No postradiation chemotherapy was administered. Median survival was 19 months in the observation group versus 24.5 months in the treatment group, but this difference was not statistically significant (P = .737). For the subgroup of patients with pancreatic adenocarcinoma (n = 114), the median survival was 12.6 months in the observation group versus 17.1 months in the treatment arm, again not statistically significant (P = .099). Of note, 21 of 104 patients randomized to the treatment arm were not treated. In addition, although the original dose of 5-FU was already modest, 35 patients in the treatment arm received only 3 days of 5-FU during the second module of radiotherapy, secondary to grade I/II toxicities. Therefore, although controversial, this study may be better described as an underpowered but trending toward positive study.211

Recently, the European Study Group for Pancreatic Cancer (ESPAC) randomized 541 patients with pancreatic adenocarcinoma in a 4-arm design, on the basis of a 2 × 2 factorial design: (1) observation, (2) concomitant chemoradiotherapy alone (20 Gy in 10 fractions over 2 weeks with 500 mg/m2 5-FU intravenous bolus during the first 3 days of radiation therapy; the module is repeated after a planned 2-week break) followed by no additional chemotherapy, (3) chemotherapy alone (leukovorin 20 mg/m2 bolus followed by 5-FU 425 mg/m2 administered for 5 consecutive days repeated every 28 days for 6 cycles), or (4) chemoradiotherapy followed by chemotherapy.212 There was no significant difference in survival between patients assigned to chemoradiotherapy (median survival 15.5 months) versus observation (median survival 16.1 months; P = .24). The survival data were similar in the subset (n = 285 patients) randomized through the 2 × 2 design. In contrast, there was a survival advantage for those patients treated with chemotherapy alone (median survival 19.7 months) versus observation (median survival 14 months, P = .0005). For the same subset randomized through the original 2 × 2 design, survival demonstrated a trend toward improved survival for chemotherapy alone (median survival 17.4 months) versus observation alone (15.9 months), but it was not statistically significant (P = .19). Multivariate analysis for known prognostic factors including margin status, lymph node involvement, and tumor grade and size did not alter the effect for chemoradiotherapy treatment. The ESPAC-1 authors concluded that there was no survival benefit for adjuvant chemoradiotherapy. In addition, the authors concluded that a potential benefit existed for adjuvant chemotherapy alone after surgical resection.

Although this ESPAC-1 trial was a randomized study consisting of more than 500 patients, the conclusions of the study should be carefully measured. To encourage maximal patient recruitment, the study was modified in that 68 patients were assigned separately and randomized to either chemoradiotherapy or observation alone. In addition, 188 patients were assigned separately and randomized to either chemotherapy alone or observation. In a sense, 3 randomizations were possible for inclusion into the same study. Also, patients in the additional 2 randomizations could have potentially received “background chemotherapy or chemotherapy,” which was not specifically defined. The background treatment was not known in 82 eligible patients. Of note, these patients were still assigned into an arm of the study in spite of the lack of definitive knowledge of prior therapy. Finally, 25 of the eligible 541 patients refused to accept their randomization, and an additional 25 patients withdrew as a result of treatment toxicities.

As the debate regarding the optimal adjuvant therapy for pancreatic cancer continues, several studies have recently opened or have been proposed by either the cooperative groups or through single institutions. Table 16 summarizes some open or planned studies in the adjuvant setting.

Table 16. Selected active or planned adjuvant studies

	Study	Regimen
	RTOG1091	Gem 1000 mg/m2 × 3 w, XRT 50 Gy 1.8 Gy fx vs Gem 600 mg/m2 weekly, followed by gem × 3 cycles
	ACOSOG Z5006	Arm 1: 5FU CI/LV/DPM/MMC, 5FU/XRT (50 Gy), 5FU CI × 2 cycles Arm 2: 5FU/LV/DPM/MMC × 6 cycles Arm 3: XRT (50 GY)/ 5FU CI/cisplatin/IFN, 5FU CI × 2 cycles
	Johns Hopkins	GM-CSF allo vaccine, 5-FU CI/LV/DPM/MMC, 5FU/XRT, 5-FU CI × 2 cycles followed by GM-CSF allogeneic vaccine × 3

Gem, Gemcitabine; LV, leucovorin; IFN, interferon-α; CI, continuous infusion; MMC, mitomycin; DPM, dipyridamole; GM-CSF, granulocyte-macrophage colony-stimulating factor. All are planned as phase II studies.

Combined modality therapy for locally advanced disease 

In patients with metastatic pancreatic cancer, the current standard of care is single-agent gemcitabine. There has therefore been significant interest in using gemcitabine either earlier in treatment or in combination with radiation therapy; however, gemcitabine is also a potent radiosensitizer.213 Because of this, studies combining radiotherapy with gemcitabine have found it necessary to proceed cautiously. Blackstock et al214 examined in a phase I study, gemcitabine (starting at 20 mg/m2) administered twice weekly in combination with radiation therapy (total dose 50.4 Gy in 1.8 Gy fractions) in 19 patients with locally advanced pancreatic adenocarcinoma. Thrombocytopenia, neutropenia and nausea/vomiting were dose-limiting toxicities. Of the 15 patients assessable for response, 3 partial responses were identified. A dose of 40 mg/m2 administered twice weekly in combination with radiotherapy (to a total dose of 50.4 Gy) was subsequently examined by the Cancer and Leukemia Group B in a phase II study of 38 patients with locally advanced pancreatic cancer.215 After chemoradiotherapy, patients without disease progression received gemcitabine alone 1000 mg/m2 weekly × 3 every 4 weeks for 5 additional cycles. Grade III/IV hematologic toxicity was significant and was identified in 60% of patients. In addition, grade III/IV gastrointestinal toxicity was identified in 42% of patients. With median follow-up of 10 months, median survival was 7.9 months.

Two groups have published phase I experiences with gemcitabine dose escalation, starting at 300 mg/m2 with radiation therapy (5040 cGy in standard fractions) in patients with resectable or locally advanced disease. Dose-limiting toxicities were identified at a gemcitabine dose of 700 mg/m2. In addition, late toxicities were identified in 2 of 6 patients at 600 mg/m2. Of note, partial responses were seen in 3 of 6 patients with locally advanced disease at a dose of 600 mg/m2.216–218

The MD Anderson Cancer Center (MDACC) has since published a corollary phase I study of 18 patients with locally advanced disease by use of rapid fractionation external beam radiation.219 Patients received dose escalation gemcitabine from 350 mg/m2 to 500 mg/m2weekly × 7, with concurrent rapid fractionation 3000 cGy external beam radiation therapy during the first 2 weeks of therapy. Hematologic and nonhematologic toxicities were significant in all 3 patient cohorts. There were 8 responses (4 minor and 4 partial). One of 2 patients who were subsequently explored had a curative resection. The recommended phase II testing dose of gemcitabine was 350 mg/m2.219

Recently, the University of Michigan has also described an alternative approach, using standard doses of gemcitabine at 1000 mg/m2 weekly × 3 every 4 weeks and administering radiation therapy as dose escalation beginning at 24 Gy (1.6 Gy fractions in 15 fractions) in 37 patients with locally advanced disease.220 Most patients received postchemoradiotherapy chemotherapy at the discretion of the treating physician. Seventy-five percent of the patients received at least 85% of the planned gemcitabine. Two of 6 assessable patients experience dose-limiting toxicity at the final planned radiation dose of 42 Gy in 2.8 Gy fractions. An additional 2 patients had development of late gastrointestinal toxicities at this dose level. Six patients were documented to have a partial response, with a complete radiographic response in 2 patients. In addition, 4 patients with documented stable disease at the time of study entry experienced objective responses (2 partial and 2 complete reponses). Definitive resection was achieved in 1 of 3 patients. With median follow-up of 22 months, median survival for the entire group was 11.6 months. The recommended phase II radiation dose was 36 Gy in 2.4 Gy fractions.

There have also been attempts to optimize chemotherapy when combined with radiation therapy. The ECOG published a phase I study of 7 patients with locally advanced disease by use of combination chemotherapy consisting of radiation therapy to a maximum 59.4 Gy in 1.8 Gy fractions.221 The 5-FU (200 mg/m2/d as continuous infusion throughout radiation therapy) was administered with weekly gemcitabine dose escalation beginning at 100 mg/m2. Because of dose-limiting toxicities seen in 2 of the first 3 patients, the study was amended to lower the initial dose of gemcitabine to 50 mg/m2. However, dose-limiting toxicities were subsequently seen in 3 of 4 patients at the 50 mg/m2 dose. Three of the 5 dose-limiting toxicities occurred at radiation doses less than 36 Gy. The study was subsequently closed.

A more promising combination may be gemcitabine and cisplatin. This combination has demonstrated synergy in a variety of human tumor cell lines and has been demonstrated to have clinical benefit when used in the metastatic setting.222–224 This combination is currently being further evaluated in clinical studies.

Given the current published data, would 5-FU or gemcitabine be better suited to be used concurrently with radiation therapy for either resected or locally advanced disease? The MDACC retrospectively examined their database of 114 patients with locally advanced disease treated with combination radiation therapy (rapid fractionation 30 Gy in 10 fractions) and either 5-FU continuous infusion 200 to 300 mg/m2 (61 patients) or gemcitabine 250 to 500 mg/m2 weekly × 7 (53 patients).225 Patients receiving gemcitabine developed a significantly higher incidence of severe acute toxicity: defined as toxicity requiring a hospital stay of more than 5 days, mucosal ulceration with bleeding, more than 3 dose deletions of gemcitabine or discontinuation of 5-FU, or toxicity resulting in surgical intervention or death, compared with those patients receiving 5-FU (23% vs 2%, P <.0001). Five of 53 patients treated with gemcitabine/radiation therapy subsequently underwent surgical resection compared with 1 of 61 patients treated with 5-FU/radiation therapy. However, with a short median follow-up, median survival was similar (11 months vs 9 months, P = .19).225

Neoadjuvant therapy 

Neoadjuvant therapy is a potentially attractive alternative to current adjuvant therapies for several reasons: (1) radiation is more effective on well-oxygenated cells that have not been devascularized by surgery, (2) contamination and subsequent seeding of the peritoneum with tumor cells resulting from surgery could theoretically be reduced, (3) patients with metastatic disease on restaging after neoadjuvant therapy would not need to undergo exploration, and (4) the risk of delaying adjuvant therapy after resection would be eliminated, because it would be delivered in the neoadjuvant setting.

The MDACC published their experience of 142 patients with localized resectable pancreatic adenocarcinoma who were either treated (1) before operation with radiation therapy (50.4 Gy in either standard 1.8 Gy fractions or consisting of 30 Gy rapid fractionation in 3 Gy/fraction) combined with 5-FU continuous infusion 300 mg/m2/d followed by surgical resection or (2) by resection followed by postoperative chemotherapy (5-FU continuous infusion 300 mg/m2/d) and radiation therapy (50.4 Gy in standard fraction). There were no delays to surgery in the neoadjuvant group, but there were noted to be delays to initiate postoperative therapy in 6 of 25 patients who underwent surgical resection first. At a median follow-up of 19 months, no significant differences in survival were noted.226

The Fox Chase Cancer Center published their experience of 53 patients with localized resectable pancreatic cancer who were treated before operation with radiation therapy (5040 cGy in 180 cGy fractions) and chemotherapy (MMC 10 mg/m2 on day 2 with 5-FU 1000 mg/m2/d continuous infusion on days 2 to 5 and 29 to 32). Forty-one patients subsequently underwent exploratory laparotomy at the conclusion of preoperative chemoradiation. From this group of patients, 17 were not resectable (11 patients with hepatic or peritoneal metastases and 6 patients with local extension that precluded resection). Twenty-four patients underwent resection. There were significant treatment-related hematologic and nonhematologic toxicities identified, including one patient with treatment-related toxicities that precluded reexploration. Median survival for the entire group was 9.7 months, and 15.7 months for the group that underwent surgical resection.227

The Fox Chase group has since published a follow-up study of 30 patients with localized left-sided resectable pancreatic cancer, of which 26 received preoperative radiation therapy (50.4 Gy) with 5-FU continuous infusion. Ten patients who received such preoperative therapy subsequently underwent resection. Median survival was 34 months for the resected group.228

Other potential radiation sensitizers have also been examined in the preoperative setting. The MDACC have used paclitaxel 60 mg/m2 over 3 hours weekly with 30 Gy radiation therapy rapid fractionation.229 Of note, if patients could undergo surgical resection, they could also receive intraoperative radiation therapy. Grade III hematologic and nonhematologic toxicities were identified in 16 patients. No delays in surgery were attributable to preoperative therapy. Twenty of 25 patients who underwent exploratory laparotomy underwent resection. There were no histologic complete responders. With a median follow-up of 45 months, the 3-year survival rate for those patients after potentially curative resection was 28%, with an overall median survival of 19 months.

Currently, ECOG is planning to open a prospective neoadjuvant trial, randomizing patients to intensified gemcitabine-based or 5-FU/platinum–based chemoradiotherapy. This trial makes an important distinction between clearly unresectable disease, focusing on the issues of partial versus complete encasement of the superior mesenteric artery and length of superior mesenteric vein involved by tumor at initial presentation.

To date, the current data demonstrate that although neoadjuvant chemoradiotherapy can be administered safely, there is no clear advantage to this strategy compared with postoperative therapy. In the cohort of questionably resectable patients, it remains to be seen whether there are patients for whom this approach may represent an important therapeutic advantage.

Components of the immune system required for cancer immunotherapy 

There are a number of cell types that, when activated, are extremely efficient at recognizing and killing their target. B and T cells each have combinatorial clonally distributed antigen receptors that provide the specificity to recognize foreign antigens and to discriminate self from nonself. Through the recombination of genes encoding subunits of their receptors, both B and T cells can recognize more than a million different antigens whether they are in the form of the product of a new genetic alteration, a reactivated embryonic gene, or an overexpressed gene. The B cell recognizes free antigenic determinants or whole-surface molecules, eliminating the need for special antigen processing. In contrast, the T-cell receptor recognizes fragments of the antigenic protein bound to human leulocyte antigen (HLA) class I and II molecules on tumor cells or specialized cell types known as antigen-presenting cells (APCs), respectively. These peptide-HLA complexes are formed as a result of fragmentation of proteins within specialized cellular compartments and subsequent association with a binding site on the HLA molecule. APCs (macrophages, B cells, and dendritic cells) have the ability to capture extracellular proteins that are released by the tumor through secretion, shedding, or tumor lysis. These proteins are subsequently internalized via endocytosis and processed through the exogenous pathway. Peptide fragments (10 to 25 amino acids in length) then bind to the HLA class II protein, before expression of the complex on the cell surface. This complex is recognized exclusively by CD4+ helper T cells in the context of a second costimulatory molecule such as B7. In contrast, most tumor cells cannot process and present antigen through the exogenous pathway, because they usually do not derive from professional APCs. However, all cells including tumor cells have the ability to process and present antigens that derive from cellular proteins through the endogenous pathway. Any protein within a tumor cell can gain access to the cytosol and undergo enzymatic degradation into 8 to 10 amino acid fragments by specialized machinery (the proteasome) via the endogenous pathway. The peptide fragments are subsequently transported into the endoplasmic reticulum, where they bind to HLA class I molecules and are transported to the cell surface for recognition by CD8+ T cells. In general, CD4+ T cells provide helper or regulatory function whereas CD8+ T cells carry out direct tumor lysis.

Immunotherapy in clinical practice 

Immunotherapy can be broadly divided into passive and active therapeutic approaches (Table 17).

Table 17. Immunotherapy strategies

	Passive immunotherapy
	Unlabeled monoclonal antibodies
	Radioimmunotherapy
	Antibody-directed immunotoxins
	T cell adoptive transfer
	Active nonspecific immunotherapy
	Whole tumor cells/ tumor lysate mixed with bacterial adjuvant
	Systemic cytokines
	Active specific immunotherapy (vaccines)
	Genetically modified whole tumor cells
	Protein/peptide/carbohydrate–*based antigen vaccines
	Dendritic cell-based antigen vaccines
	DNA-based vaccines
	Recombinant viral-based antigen vaccines

*Glycoprotein antigens are recognized by T and B cells in a similar way as protein/peptide antigens.

In contrast to passive immunotherapy, active specific immunotherapy (vaccine therapy) targets specific tumor antigens as a result of the induction of antigen-specific B-cell– or T-cell–mediated immune responses. Active specific therapy can also generate antigen-specific memory T-cell responses that are capable of being reactivated if tumor cells expressing the same antigen profile recur. Furthermore, the induction of cellular immune responses has the added benefit of allowing natural access to the microenvironment of the tumor. Preclinical studies have already shown that T-cell–mediated vaccine therapy can induce antitumor immune responses that are potent enough to eradicate many murine tumors. Translation of these vaccine approaches into therapies for patients with pancreatic cancer are in the early phases of clinical development. Examples of the different vaccine approaches that are currently undergoing clinical testing include peptide- and protein-based vaccines or whole tumor cell vaccines.

Peptide- and protein-based vaccines 

There are at least 2 major advantages to peptide- and protein-based vaccines: they are inexpensive and simple to produce, and they can be given in large quantities, thereby allowing for maximal immunization with relevant antigens. Point mutations in a variety of oncogenes (K-ras) or tumor suppressor genes (p53, p16, DPC4, BRCA2, Her-2/neu) have been associated with different histologically defined precursor lesions, and some are being studied as candidate immune targets. Mutated K-ras is a particularly attractive immune target because it is mutated in >90% of pancreatic adenocarcinomas.231, 232 This antigen has been tested for the induction of antitumor immunity in several trials. However, post vaccination responses have been observed in very few patients and have not correlated with clinical responses.233–235 Peptide- and protein-based vaccines that are currently under development are listed in Table 18.

Table 18. Some current clinical trials testing vaccines in pancreatic adenocarcinoma

	Approach	Sponsor	Stage	Trial phase
	rFowlpox CEA/TRICOM ± GM-CSF	NCI/Fox Chase NCI	Metastatic	Phase I
	CEA peptide/adjuvant + GM-CSF	NCI	Metastatic	Phase II
	Virulizin/Gem vs Gemcitabine	Lorus Therapeutic	Metastatic	Phase III
	G17DT/Gem vs Gemcitabine	Aphton	Metastatic	Phase III
	K-ras/adjuvant ± IL-2 or GM-CSF	NCI	Metastatic	Phase I
	Gem/XRT vs K-ras + Gem/XRT	RTOG	Resected	Phase II
	GM-CSF allogeneic whole tumor cell	Johns Hopkins	Resected	Phase II
	GM-CSF allogeneic whole tumor cell	Johns Hopkins	Metastatic	Phase I

Gem, Gemcitabine; GM-CSF, granulocyte-macrophage colony-stimulating factor.

Heat shock protein (HSP)–based vaccines are a newer approach that have demonstrated promise. HSPs are ubiquitous and highly conserved cellular proteins that have multiple functions, including helping newly synthesized polypeptides fold, assisting in protein transport, and associating with peptides generated during protein degradation. They are also believed to stimulate macrophage and dendritic cell activation and assist in representation of peptides. Preclinical studies have shown that HSPs isolated from tumor cells can serve as potent vaccines by taking advantage of their role as a peptide transporter and as a stimulator of APCs. This approach has been tested in patients with resected pancreatic adenocarcinoma from whom HSP could be obtained and purified. Eligible patients were administered 5 mg of protein (HSP-96) subcutaneously weekly for 4 weeks.236 The vaccine was well tolerated. In addition, an increase in postvaccination CD8+ T cells specific for autologous tumor was observed in one patient.

A look to the future 

There are significant challenges that must be overcome if immune-based therapies are to play an important role in the treatment of pancreatic cancer. First, immune-based strategies must be able to circumvent the genetic alterations within a tumor cell, which allow the tumor cell to evade immunologic recognition and eradication. Typically, genetic alterations result in the loss of either antigen expression or the ability to adequately present antigen to T cells. One possible solution to this problem is to design polyvalent vaccines and antibodies that target immunity against several tumor rejection antigens. Second, additional candidate pancreatic tumor antigens are needed to serve as immune-relevant tumor targets. New and more rapid methods need to be developed to identify these targets. Third, it is unlikely that immunotherapy alone will be able to overcome mechanisms that functionally inactivate tumor-specific T cells, a recognized problem that limits its effectiveness in patients with large tumor burdens (like pancreatic cancer). Combinatorial therapies that provide tumor debulking and immune modulation given in sequence with immune-based therapies will be required to overcome this problem. Consequently, it might be possible to enhance the effects of vaccine based approaches by combining the cytoreductive and immune-modulating elements of chemotherapy with the tumor cell cytotoxic specificity of immunotherapy. Both combinations have been shown to overcome peripheral tolerance to tumor antigens in preclinical models.244–247

MMP inhibitors 

The MMPs are a group of closely related endopeptidases collectively capable of degrading essentially all components of the extracellular matrix.248 These proteins are overexpressed in the majority of human neoplasms and are believed to contribute to the process of tumor invasion, angiogenesis and metastasis. A number of agents whose primary mechanism of action is inhibition of the MMP have been developed in the clinic, and at least 2 of the them have been specifically explored in pancreatic cancer: Bay 12-9566 and Marimastat. Bay 12-9566 is a biphenyl matrix metalloprotease inhibitor (MMPI) that selectively inhibits the proteolytic activity of two of the more relevant members of the MMP family.248, 249 This agent was tested in a randomized clinical trial in 277 patients with advanced pancreatic cancer in comparison with gemcitabine. Unfortunately, in spite of promising activity in preclinical studies, the median survival of patients treated with this agent was 3.2 months, significantly worse than the 6.4-month median survival of patients treated with gemcitabine, leading to a halt in the development of this agent in patients with pancreatic cancer.249 The second MMPI that has been actively developed in pancreatic cancer is Marimastast, a nonselective hydroxanate MMPI. Phase II studies conducted with Marimastat indicated that treatment with the agent reduced the levels of circulating tumor markers in patients with advanced pancreatic cancer, leading to further study.250 In a randomized phase II trial, treatment with 25 mg of Marimastat daily was equivalent to gemcitabine in patients with advanced disease, supporting the activity, albeit modest, of this compound in patients with advanced pancreatic cancer.251 On the basis of these results, the development of Marimastat in patients with pancreatic cancer (either alone or in combination with standard treatments) continues.

Ras pathway inhibitors 

The second group of molecularly targeted agents being explored in pancreatic cancer are the inhibitors of farnesyl transferase (FT), the enzyme that mediates the posttranslational modification of the ras protein required for this oncogene to be activated.252 More than 90% to 95% of pancreatic cancers harbor mutations in this gene, making it theoretically an attractive target. Several FT inhibitors have been tested in clinical studies to date, and one of these agents, R115777 (Zarnestra) has been tested in both phase II and III clinical trials. Treatment of a total of 76 patients with advanced pancreatic cancer in phase II studies did not show any objective response and resulted in a median survival in treated patients of less than 4.5 months.253, 254 Parallel pharmacodynamic studies indicated, however, that the agent inhibited the FT in peripheral blood mononuclear cells.253 Phase III evaluation has been equally disappointing, because the addition of R115777 to gemcitabine did not result in better outcome compared with gemcitabine alone.255 The negative results could, however, be related to several factors. First, it is not clear that the ras oncogene itself is the target of these agents, because tumors with wild type ras are also inhibited by FT inhibitors.252 Second, the most prevalent mutation of ras in pancreatic cancer involves the K-ras oncogene, which is known to be activated by a process different from farnesylation.252

Erb family inhibitors 

The third class of novel targeted compounds that have been studied in pancreatic cancer are the inhibitors of the ERB family of receptors.256 This is a family of transmembrane receptors composed of an extracellular domain, a transmembrane anchoring domain and an intracellular domain with tyrosine kinase activity.256 Most human epithelial tumors, including cancer of the pancreas, are characterized by dysregulation of this receptor family. Two members of the family, ERB-1 (EGFR) and ERB-2 have been targeted therapeutically with either monoclonal antibodies against the extracellular domain of the receptor such as C-225 (ERB-1) and Herceptin (ERB-2), or small molecules that target the intracellular TK domain of the receptor (OSI-774).256 The combination of C-225 and Herceptin with gemcitabine has been tested in phase I-II studies, demonstrating a promising response rate in the 15% to 20% range in patients with advanced disease, with a mild to moderate toxicity profile.257, 258 OSI-774 is currently been explored in combination with gemcitabine (compared with gemcitabine alone) in patients with advanced pancreatic cancer.

In conclusion, a variety of novel targeted agents are being explored in patients with pancreatic cancer. Although some trials have been negative, studies with other compounds appear to be more promising. The development of novel targeted agents in pancreatic cancer will continue over the next few years. It should be emphasized, however, that most of these trials have been conducted with classic clinical trials methods used, with no attempt to determine the relationship between molecular target expression and functionality and modulation by treatment. It is likely that the implementation of such biologic end points in these studies will provide critical information to better understand the pharmacology of these novel agents.