Int. J. Cancer: 70, 661–667 (1997) r 1997 Wiley-Liss, Inc. Publication of the International Union Against Cancer Publication de l’Union Internationale Contre le Cancer Ki-ras MUTATIONS IN EXOCRINE PANCREATIC CANCER: ASSOCIATION WITH CLINICO-PATHOLOGICAL CHARACTERISTICS AND WITH TOBACCO AND ALCOHOL CONSUMPTION Núria MALATS1, Miquel PORTA,1* Josep M.a COROMINAS1, Josep L. PIÑOL1, Juli RIFÀ2 and Francisco X. REAL1 for the PANK-ras I Project Investigators 1Institut Municipal d’Investigació Mèdica, Hospital del Mar, Universitat Autònoma de Barcelona, Barcelona, Spain 2Hospital Son Dureta, Palma de Mallorca, Spain The aims of this study were (i) to assess the prevalence and spectrum of codon 12 Ki-ras mutations in patients diagnosed with exocrine pancreatic cancer (EPC) in 2 general hospitals between 1980 and 1990, (ii) to analyze the association of this genetic alteration with clinical and pathological characteristics, and (iii) to determine the association of Ki-ras mutations with tobacco and alcohol consumption. DNA was amplified from paraffin-embedded tissue samples and mutations in codon 12 of Ki-ras were detected using the artificial RFLP technique. Cox proportional-hazards regression and unconditional logistic regression were applied. Codon 12 Ki-ras mutations were detected in 30 of 51 cases for which molecular results were available. The amino-acid substitutions were Asp (8), Val (6), and Arg (3). A double mutation, including always a Val, was detected in 5 cases. None of the 4 nonductal pancreatic neoplasms were mutated. The mutation prevalence was 79% in metastases and 54% in primary tumors. The risk of a mutated tumor was 3 times higher in alcohol drinkers than in non-drinkers, and a linear trend was apparent. When age, gender, hospital, and tobacco and alcohol consumption were taken into account, a high risk for mutations was detected in patients who only smoked and in patients who only drank, but less so in patients who both smoked and drank. These results raise novel hypotheses regarding the role of tobacco and alcohol in EPC. Int. J. Cancer, 70:661–667, 1997. r 1997 Wiley-Liss, Inc. Advances in the genetic analysis of human cancer provide a basis for its molecular classification, the development of novel diagnostic and therapeutic strategies, and the identification of putative etiological clues. Exocrine pancreas cancer (EPC) is the tumor for which fewer therapeutic advances have been made in the last 50 years, and its mortality remains similar to its incidence (Warshaw and Fernandez del Castillo, 1992; Urrutia and DiMagno, 1996). After Almoguera et al. (1988) reported a high prevalence of mutations in codon 12 of Ki-ras in EPC, about 50 reports have described Ki-ras mutations in this tumor, with prevalence ranging from 47% (Tabata et al., 1993) to 100% (Tada et al., 1990; Caldas et al., 1994). The basis for such variation is unclear, and individual studies are not strictly comparable due to significant differences in strategies used to select patients, number of cases studied, type of sample analyzed, and mutation-detection techniques. Only 7 of the studies analyzed more than 40 patients (Grünewald et al., 1989; Capellà et al., 1991; Hruban et al., 1993; Motojima et al., 1993; Finkelstein et al., 1994; Scarpa et al., 1994; Pellegata et al., 1994). Furthermore, most published series provide scanty information on the selection criteria used, and only one described the population study base and analyzed the differences between cases with and without molecular results (Hruban et al., 1993). While selection biases are generally important in molecular studies, they may be of even greater relevance in EPC, since the proportion of patients undergoing histological confirmation is often low and varies widely. Thus, EPC may be one of the tumor types with a higher rate of diagnostic misclassification (Porta et al., 1994). Moreover, few studies have assessed the associations between Ki-ras mutations and the clinical and pathological characteristics of cases (Grünewald et al., 1989; Nagata et al., 1990; Hruban et al., 1993; Motojima et al., 1993; Scarpa et al., 1994; Pellegata et al., 1994; Finkelstein et al., 1994). It is plausible that lifestyle and environmental factors may cause EPC through the activation of Ki-ras. The molecular abnormalities of tumors may reflect exposure to carcinogens involved in their development, as described for acute myeloid leukemia (Taylor et al., 1992). Similarly, Ki-ras-mutated EPC and wild-type EPC may result from different etiologic agents. The actual causes of EPC remain largely unknown (Boyle et al., 1989; Haddock and Carter, 1990; Warshaw and Fernández del Castillo, 1992). Tobacco has been consistently found to be a risk factor for EPC (Boyle et al., 1989; Haddock and Carter, 1990; Bueno de Mesquita et al., 1991; Howe et al., 1991; Fujji et al., 1990), but only 2 studies have analyzed the relationship between smoking and Ki-ras mutations in EPC, and their results are contradictory (Nagata et al., 1990; Hruban et al., 1993). Although alcohol consumption has been reported to be a risk factor for EPC, its role remains controversial (Boyle et al., 1989; Haddock and Carter, 1990; Bueno de Mesquita et al., 1992). The possible association between alcohol drinking and Ki-ras mutations in EPC has not been studied. The aims of this study were (i) to determine the prevalence and the spectrum of codon-12 Ki-ras mutations in all patients diagnosed with EPC in 2 general hospitals between 1980 and 1990, (ii) Abbreviations: EPC, exocrine pancreatic cancer; PCR, polymerase chain reaction; RFLP, restriction-fragment-length polymorphism; OR, odds ratio; CI, confidence interval. Results of this study were presented to the II Congresso Ibero-Americano de Epidemiologı́a (Salvador de Bahia, Brazil, April 24–28, 1995), the 27th. Meeting of the European Pancreatic Club (Barcelona, June 28–30, 1995) (Digestion 1995; 56(4):302), the 6th. Meeting of the Spanish Association for Cancer Research (Barcelona, September 24–27, 1995) (Oncology Reports 1995; 2 [Suppl]: 929), and the 14th. International Scientific Meeting of the International Epidemiological Association (Nagoya, Japan, August 27–30, 1996). Contract grant sponsor: Fondo de Investigación Sanitaria (FIS), contract grant numbers 91/0595, 92/0007; Contract grant sponsor: Fundación Salud 2.000 (Grant Serono); Contract grant sponsor: Institut Municipal d’Investigació MeTdica (IMIM), contract grant number 875138/9; Contract grant sponsor: Comissió Interdepartamental de Recerca i Innovació tecnològica (CIRIT), contract grant numbers GRQ93-9301, SGR 434. PANK-ras I Project Investigators: M. Porta and F.X. Real (Principal Investigators), N. Malats, J.L. Piñol, J.M. Corominas, J. Rifà, M. Andreu, M. Conangla, T. Thomson, M. Gallén, R. Solà and F. Tous. *Correspondence to: Institut Municipal d’Investigació Mèdica, Hospital del Mar, Universitat Autònoma de Barcelona, Carrer del Dr. Aiguader #80, E-08003 Barcelona, Spain. Fax: 34-3-2213237. Received 28 June 1996; revised 25 November 1996 MALATS ET AL. 662 to analyze the association between this genetic alteration and clinical and pathological characteristics of the cases, and (iii) to assess the association of Ki-ras mutations with tobacco and alcohol consumption. MATERIAL AND METHODS Patients and information The methodology of the PANK-ras I study has been described elsewhere (Porta et al., 1994; Malats et al., 1995). Briefly, cases diagnosed with EPC (n 5 149) between 1980 and 1990 were identified through the cancer registries at Hospital del Mar (Barcelona) and Hospital Son Dureta (Palma de Mallorca), Spain. Sociodemographic, lifestyle, clinical and pathological information was obtained from clinical and pathological records through a structured data form. Age was calculated from date of birth to date of diagnosis. Prior medical history of diabetes mellitus, acute and chronic pancreatitis, gallstones, duodenal ulcer, and psychiatric disorders was recorded as present, if mention was made in clinical records, and as absent otherwise. Tumor site was coded according to the International Classification of Diseases for Oncology (ICD-O). Disease extent was classified as local when the tumor was localized within the pancreas, regional when it was locally advanced or showed regional lymph-node involvement, and metastatic when distant metastases were demonstrated. Tumors were graded as well-, moderately, and poorly differentiated following ICD-O criteria. Treatment intention was categorized as curative, palliative, and symptomatic. Survival was determined as the period between date of most-likely diagnosis and death/last control. Smoking history was recorded as a dichotomous variable (ever/never). Alcohol consumption was registered in 4 categories: no consumption, light (,30 g per day), moderate (30–80 g per day), and heavy (.80 g per day). Tissue specimens An exhaustive search for histological material from all patients included in the study was carried out. Samples of primary and of metastatic lesions were obtained. Paraffin-embedded tissue was available from 56 EPC patients (64.4% of the 87 cases with histological confirmation). Tissue blocks from the remaining 31 patients could not be retrieved. From each specimen, 20 sections of 5 µ thickness were cut, and placed on glass slides. The first, 10th and 20th sections were stained with hematoxylin/eosin and used for histological evaluation. All cases were reviewed by a pathologist (J.M.C.) who was unaware of the original diagnosis and of the results of the molecular analysis. The pathologist defined tumor (T) and non-tumor (N) areas by microscopic examination. The material corresponding to the T and N areas was scraped, always starting from the N area. The percentage of tumor cells present in the T area was assessed by 2 independent investigators (J.M.C. and F.X.R.). In general, the area analyzed was 1 cm2. Detection of c-Ki-ras mutations Care was taken to avoid contamination during all steps of amplification and analysis. Tissue scrapings were de-paraffinized in xylene and washed with 95% ethanol; tissue pellets were desiccated and heated at 95°C for 10 min. DNA was purified using a commercial kit (LINUS, Cultek, Madrid, Spain) and amplified in 2 steps by nested PCR. The primers and conditions for DNA amplification used in the study have been described (Berrozpe et al., 1994; Malats et al., 1995). This technique was able to detect one homozygous mutated cell in the presence of 102 normal cells. DNA was successfully amplified from 51 of 56 cases from which tissue blocks were available. The primers used in the second, nested, PCR introduced an artificial restriction-endonuclease site for BstNI. The 103-bp products of this amplification reaction were digested overnight. Wild-type sequences were cleaved, resulting in 2 fragments of 82 and 21 bp, whereas codon-12 mutated sequences were not. Products were analyzed by acrylamide-gel electrophoresis and ethidium-bromide staining. Analyses were restricted to codon 12 because mutations in other codons are very rare in pancreatic cancer (reviewed in Caldas and Kern, 1995). To characterize the nucleotide substitution in codon 12, all mutated samples were further analyzed using a similar RFLP-based approach, as described elsewhere (Berrozpe et al., 1994; Malats et al., 1995). GGT = AGT (Gly = Ser), GGT = GCT (Gly = Ala), and GGT = TGT (Gly = Cys) mutations were not studied because of their low prevalence in EPC (reviewed in Caldas and Kern, 1995). DNA from oral mucosal scrapings was used as normal control, and DNA from pancreas-cancer cell lines or tumors (Berrozpe et al., 1994) was used as control for the Asp, Val and Arg mutations. Interpretation of electrophoresis of digestion products’ was performed by consensus of 2 investigators (NM and AS). When the results were discordant, the analysis was repeated and the results were evaluated again. The reliability of the digestion results was further assessed by an independent investigator (F.X.R.). Agreement was high ($95%) for all enzyme digestions. Statistical methods Comparison of 2 qualitative variables was performed using Pearson’s x2 test or, alternatively, with Fisher’s exact test when $20% of cells had expected counts of less than 5. Odds ratios (OR) were used to estimate the magnitude of associations between variables; when the observed number of cases was zero in one cell of the contingency table, Woolf-Haldane’s correction was applied. The logit estimator of the OR was calculated with precision-based confidence intervals (CI). Adjusted OR and CI were estimated by unconditional logistic regression. Student’s t-test or MannWhitney’s U-test were used to analyze the relationship between a categorical variable with 2 levels, and a normally or non-normally distributed quantitative variable, respectively. Kaplan-Meier plots and the log-rank test were used to evaluate the association between mutations and survival. Cox proportional-hazards regression was applied to analyze survival, taking into account possible confounder variables. All p values are 2-sided. Statistical analysis was performed using Macintosh statistical packages STATVIEW 4.01 (Abacus Concepts, Berkeley, CA) and SPSS 4.0 (SPSS, Chicago, IL), and IBM PC packages EGRET (Statistics and Epidemiologic Research, Seattle, WA) and SAS (SAS Institute, Cary, NC). RESULTS Prevalence and spectrum of Ki-ras mutations The characteristics of the 51 patients from whom results on Ki-ras mutations were available were similar to those of patients from whom molecular results were not available. As might be expected, Ki-ras mutations could be determined in a larger proportion of cases treated with curative intention ( p , 0.001). Codon-12 Ki-ras mutations were detected in 30 of 51 EPC cases (59%; 95% CI, 44.9–70.0). The proportion of tumor cells in the tissue area analyzed was similar for mutated and wild-type tumors (data not shown). The spectrum of mutations was as follows: 8 cases (27%) had an Asp substitution, 6 cases (20%) had a Val substitution, 3 cases (10%) had an Arg substitution, and in 8 cases the type of mutation could not be identified. A double mutation was detected in 5 cases (17%), of which 3 corresponded to primary tumors and 2 to metastases (Table I). A Val substitution was found in all 5 cases. Pathological and clinical characteristics Regarding the pathological characteristics of tumors (Table II), frequency of mutations was similar among cancers of the head, body, and tail of the pancreas. Whereas 29 of 46 ductal-type adenocarcinomas harboured Ki-ras mutations (63%), none of 4 non-ductal pancreatic neoplasms (1 acinar-cell carcinoma and 3 anaplastic carcinomas) were mutated ( p 5 0.026). Among ductaltype adenocarcinomas, the prevalence of mutations was 79% in metastatic-tissue samples and 54% in primary-tumor samples (OR, 3.21, p 5 0.082) (Table II). No association was observed between Ki-ras MUTATIONS IN EXOCRINE PANCREATIC CANCER differentiation degree and Ki-ras mutations ( p 5 0.255). In 4 cases, samples from primary tumor and from a metastatic lesion were analyzed. No consistent pattern was apparent (lower half of Table I). Although not statistically significant, a higher frequency of mutations was observed in cases from Hospital del Mar ( p 5 0.123). Patients less than 56 years old were 3 times more likely to have a mutated tumor than older patients (OR, 3.00; p 5 0.128) (Table III). Mutations were not significantly associated with gender, past medical history, family history of cancer, and interval from first symptom to diagnosis. The mutation prevalence was similar across tumor-stage and treatment-intention strata. Survival was also similar in cases with mutated and wild-type tumors ( p 5 0.116) (Figure 1). After adjustment for possible confounders such as gender, hospital, stage, and first-symptom diagnosis interval through Cox’s proportional-hazards regression analysis, Ki-ras mutations did not show a prognostic value in these patients. In order to explore the degree of certainty with which EPC cases were diagnosed, a diagnostic-certainty classification was developed (Porta et al., 1994). It was based mainly on the following 2 criteria: (i) presence of a discrete tumor mass in the pancreas vs. infiltration to adjacent sites and/or distant metastases, and (ii) availability of pathological sample from pancreas vs. from other sites. Cases were classified in 2 groups: a group with a higher probability of having EPC, and a group with a lower probability. TABLE I – DESCRIPTION OF CASES WITH A BI-MUTATIONAL PATTERN (SECTION A) AND PATIENTS FROM WHOM PRIMARY AND METASTATIC SAMPLES WERE ANALYZED (SECTION B)1 Section A A, B B Patient Primary-tumor tissue 45602 sample 1 sample 2 256201 sample 1 sample 2 261068 NA 287127 NA Mutation frequency was similar in the 2 diagnostic-certainty groups (Table III). Tobacco and alcohol consumption In the crude analysis, the risk of having a mutated EPC was 1.5 times higher for smokers than for non-smokers ( p 5 0.511). The risk was also 3 times higher for alcohol drinkers (.30 g per day) than for non-drinkers (#30 g per day) ( p 5 0.079) (Table III). A similar risk was observed when ever-drinkers were compared with never-drinkers (OR, 2.67; p 5 0.132). When alcohol consumption was classified in 4 categories, a direct relationship was also observed ( p 5 0.071). This pattern was evidenced across strata of all other study variables except for tobacco: moderate and heavy alcohol consumption was positively associated to Ki-ras mutations only among non-smokers. In comparison with patients who did not smoke and did not drink, cases who only smoked had an OR of 5.00 ( p 5 0.170), those who only drank had an OR of 16.00 ( p 5 0.024), and patients who both smoked and drank had an OR of 3.67 ( p 5 0.102) (Table IV). This pattern remained substantially unaltered when light drinkers were not included in the reference group. When age, gender and hospital were adjusted for, the association between alcohol and Ki-ras mutations was again detected in patients who only smoked (OR, 12.92; p 5 0.079) and in patients who only drank (OR, 22.08; p 5 0.025), but less so in patients who both smoked and drank (OR, 8.38; p 5 0.106). The tobaccoalcohol interaction term in the logistic regression model was statistically significant ( p 5 0.044); the calibration (HosmerLemeshow test, p 5 0.319) and discrimination tests (ROC 5 77%) of this multiplicative model indicated that it fitted the data well. DISCUSSION Metastatic tissue [Val] [Val, Arg] [Val] [Asp] 663 NA NA sample 1 [Val, Asp] sample 1 [Val] sample 2 [Arg] sample 4 275846 sample 1 [Val, Arg] sample 2 sample 3 47420 sample 1 sample 2 [unknown] 75964 sample 1 [unknown] sample 2 282223 sample 1 sample 2 1Brackets indicate that the sample was mutated and, if known, the specific amino-acid substitution. A blank space indicates that no mutation was detected. NA, not available. Although the prevalence of Ki-ras mutations in EPC has been the object of several reports, this study provides valuable additional information for the following reasons: (i) it precisely defined the study base by including all patients diagnosed of EPC in a specified period of time in 2 general hospitals that are not referral centers for this tumor; (ii) it compared demographical and clinicopathological characteristics of patients with and without molecular results; and (iii) it assessed the prevalence of Ki-ras mutations among diagnosticcertainty categories, an important issue given the difficulties in conclusively establishing the diagnosis of EPC (Porta et al., 1994, 1996). Of 149 potentially eligible patients, 87 (58.4%) were histologically confirmed. The figure is not particularly low for EPC (Carriaga and Henson, 1995; Porta et al., 1996). Ki-ras mutations could be determined for 51 patients, representing 58.6% of the 87 cases with histological confirmation. The proportion of cases from which tumor tissue is available largely depends on the origin and TABLE II – PATHOLOGICAL CHARACTERISTICS AND Ki-ras MUTATIONS Total n (%) Total Sub-site head body tail unspecified Histology non-ductal type ductal type Differentiation degree2 well-differentiated moderately differentiated poorly differentiated Sample site2 primary metastasis Mutated n (%) OR [95% CI] p value [0.14–6.91] [0.11–16.39] [0.21–2.89] 0.9601 51 30 (58.8) 30 (58.8) 5 (9.8) 3 (5.9) 13 (25.5) 18 (60.0) 3 (60.0) 2 (66.7) 7 (53.8) 1 1.00 1.33 0.78 4 (8.0) 46 (92.0) 0 29 (63.0) 1 15.173 [0.77–299] 0.0261 4 (9.1) 27 (61.4) 13 (29.5) 4 (100) 15 (55.6) 9 (69.2) 1 0.143 0.233 [0.01–2.81] [0.01–5.36] 0.2551 0.7014 26 (57.8) 19 (42.2) 14 (53.8) 15 (78.9) 1 3.21 [0.84–12.3] 0.0825 Percentages are shown in parentheses. 1Fisher’s exact test.–2Ductal-type EPC only.–3Woolf-Haldane’s correction.–4Mantel-Haenszel x2 test for linear trend.–5Pearson’s x2. MALATS ET AL. 664 TABLE III – PATIENT, CLINICAL AND TOXICOLOGICAL CHARACTERISTICS AND DISTRIBUTION OF Ki-ras MUTATIONS Total n (%) Total Gender of patients Female Male Age (years) ,56.4 56.4–63.2 63.3–73.2 .73.2 Hospital3 HSD HM Stage I II III IV Treatment intention radical palliative-symptomatic Survival (months) median Diagnostic certainty lower higher Tobacco No Yes Unspecified Alcohol No/light Moderate/heavy Unspecified Alcohol No Light Moderate Heavy Mutated n (%) OR [95% CI] p value 51 30 (58.8) 20 (39.2) 31 (60.8) 13 (65.0) 17 (54.8) 1.0 0.65 [0.20–2.08] 0.4721 13 (25.5) 12 (23.5) 13 (25.5) 13 (25.5) 10 (76.9) 6 (50.0) 7 (53.8) 7 (53.8) 1.0 0.30 0.35 0.35 [0.05–1.67] [0.06–1.89] [0.06–1.89] 0.4921 0.2832 18 (35.3) 33 (64.7) 8 (44.4) 22 (66.7) 1.0 2.50 [0.77–8.12] 0.1231 17 (33.3) 3 (5.9) 8 (15.7) 23 (45.1) 11 (64.7) 1 (33.3) 4 (50.0) 14 (60.9) 1.0 0.27 0.55 0.85 [0.02–3.67] [0.10–3.00] [0.23–3.11] 0.7534 0.8712 10 (19.6) 41 (80.4) 7 (70.0) 23 (56.1) 1.0 0.55 [0.12–2.42] 0.4954 2.83 3.16 22 (44.9) 26 (54.2) 13 (59.1) 15 (57.7) 1.0 0.94 [0.30–2.99] 0.9221 21 (41.2) 24 (47.0) 6 (11.8) 12 (57.1) 16 (66.7) 2 (33.3) 1.0 1.50 [0.45–5.04] 0.5111 19 (37.2) 26 (52.0) 6 (11.8) 9 (47.4) 19 (73.1) 2 (33.3) 1.0 3.02 [0.86–10.5] 0.0791 15 (33.3) 4 (8.9) 16 (35.6) 10 (22.2) 7 (46.7) 2 (50.0) 11 (68.8) 8 (80.0) 1.0 1.14 2.51 4.57 [0.13–10.4] [0.58–10.9] [0.72–29.1] 0.3364 0.0712 0.1165 Percentages are shown in parentheses. OR, odds ratio (an OR of 1 denotes the reference category). 1Pearson’s x2.–2Mantel-Haenszel x2 test for trend.–3HSD, Hospital Son Dureta, Palma de Mallorca; HM, Hospital del Mar, Barcelona.–4Fisher’s exact test.–5Log-rank test. TABLE IV – ASSOCIATION BETWEEN Ki-ras MUTATIONS AND ALCOHOL AND TOBACCO CONSUMPTION IN PATIENTS WITH EXOCRINE PANCREATIC CANCER (N 5 45) Alcohol Tobacco No/light No/light Moderate/heavy Moderate/heavy No Yes No Yes Mutations Yes No 4 5 8 11 8 2 1 6 OR p value1 [95% CI] 1 5.00 0.170 [0.65–38.15] 16.00 0.024 [1.45–176.45] 3.67 0.102 [0.77–17.43] OR, Crude (unadjusted) odds ratio. exact test. 1Fisher’s which was also a study based on hospital tumor registries. Of the 124 patients with available tissue, Hruban et al. (1993) were able to determine Ki-ras mutations in 82 (66%), a proportion only slightly above our own. FIGURE 1 – Kaplan-Meier survival curves for mutated and wild-type exocrine-pancreas-cancer cases. selection of patients. Unfortunately, seldom has this been specified in the area of Ki-ras and EPC. An exception is provided by Hruban et al. (1993), who found no significant differences between cases with and without molecular results; this finding agrees with ours, On the prevalence and spectrum of Ki-ras mutations The overall prevalence of Ki-ras mutations found in this series was 59%, a figure that is in the lower range of the literature (Caldas and Kern, 1995). This result was essentially unaltered when the single acinar-type tumor was excluded from the analysis (60%). When 3 additional anaplastic carcinomas were excluded, the mutation frequency was 63%. Several facts could explain the slightly lower prevalence found in our study: the inclusion of a wide spectrum of cases, in contrast to other more selected case Ki-ras MUTATIONS IN EXOCRINE PANCREATIC CANCER series; geographical variation (Scarpa et al., 1994), especially if we consider that the hospital from which cases showed the lowest frequency of mutations (44%) is located in an island (Mallorca); and less sensitive mutation-detection techniques. The latter is an unlikely factor, since the sensitivity of the RFLP technique was high (1%) and, when applied to the study of bile-system cancer, this method yielded a mutation prevalence higher than previously described in these neoplasms (Malats et al., 1995). The mutation spectrum found was similar to that described in other studies (Caldas and Kern, 1995). Asp and Val were the most frequent amino-acid substitutions. A bi-mutational pattern was detected in 5 tumors, and it was not associated with clinical or pathological characteristics. In 3 of them, the 2 mutations were present in a single tissue section. Because there is currently no technique available for detecting Ki-ras mutations in situ, microdissection will be necessary to establish whether both mutations occur in the same cell population or not. Other authors have described more than one mutation in samples from the same individual with EPC (Grünewald et al., 1989; Mariyama et al., 1989; Nagata et al., 1990; Motojima et al., 1993; Schaeffer et al., 1994; Caldas et al., 1994; Iguchi et al., 1996), but this pattern appears to be more prevalent in our series. In this regard, it is noteworthy that a Val substitution was identified in the 5 cases with a bi-mutational pattern. Similarly, a review of the literature shows that 14 of the 17 cases in which 2 mutations were present harbored a Val, a proportion well above that observed in cases with a single mutation (approximately 30%). The biological significance of this finding is currently unknown, although it has been proposed that different Ki-ras mutations may be associated with lesions of distinct malignant potential (Tada et al., 1996; Berrozpe et al., 1994). If Val mutations confer on cells a low capacity of progression, additional mutations in Ki-ras could provide a selective growth advantage. On the association of Ki-ras mutations with pathological and clinical characteristics This analysis yielded 2 significant associations. First, mutations were not detected in the 4 non-ductal-type tumors. There is now ample evidence that pancreatic acinar- and islet-cell tumors do not harbour Ki-ras mutations (Hoorens et al., 1993), but little is known about anaplastic tumors. Second, the probability of having a mutated sample when specimens were obtained from metastases was 3-fold that of primary tumors, in agreement with the results of Finkelstein et al. (1994). If confirmed, this finding could be clinically relevant, because it has been proposed that Ki-ras mutation detection might aid in the diagnosis of EPC (Urrutia and Dimagno, 1996), and metastases are often biopsied to avoid complications due to normal pancreatic-tissue damage. Ki-ras mutations have been reported in ductal proliferative lesions in patients with and without EPC (Yanagisawa et al., 1993a,b; Tabata et al., 1993; Tada et al., 1996), in small or pre-invasive EPC (Schaeffer et al., 1994), and in cystadenomas (Berrozpe et al., 1994), suggesting that they are an early event in pancreatic carcinogenesis. These findings are not necessarily in conflict with our results, since Ki-ras activation might play a role both early and late in the course of tumor progression. It is surprising that very little information is available in the literature for comparison: few studies provide details on Ki-ras mutations in primary and metastatic tumors; primary and metastatic lesions from a given patient have been studied only in a very small number of cases (Almoguera et al., 1988; Mariyama et al., 1989). The analysis of multiple tissue samples from the same patient, including preneoplastic ductal lesions, primary tumor and metastasis, remains a priority. Of even greater interest and complexity is the analysis of serial tissue samples obtained from the same individual at different time points. In agreement with other reports (Grünewald et al., 1989; Shibata et al., 1990; Nagata et al., 1990; Hruban et al., 1993), Ki-ras mutations were weakly or not at all associated with the clinical characteristics of the patients, or with survival. The hypothesis of a 665 higher frequency of mutations among cases with higher diagnostic certainty (Porta et al., 1994) was not confirmed here. Probably, this was due to the fact that all patients with available tissue had a solid diagnostic basis. On the association of Ki-ras mutations with tobacco and alcohol consumption Only 2 studies have previously assessed the relationship between tobacco and Ki-ras mutations in EPC. Nagata et al. (1990) observed a negative association, with 86% of mutated cases among ever-smokers and 96% among never-smokers (OR, 0.25; p 5 0.289). Conversely, Hruban et al. (1993) found a positive association, with 88% of mutated cases among ever-smokers and 68% among never-smokers (OR, 3.53; p 5 0.046). In our series, the prevalence of mutated cases was 67% in smokers and 57% in non-smokers. ras mutations have been described in other tobacco-related neoplasms such as lung, upper aerodigestive tract, and bladder cancer, although the prevalence and ras gene affected in each of these locations differ. In lung cancer, Ki-ras mutations appear to be strongly associated with smoking, and a large proportion are G-to-T transversions (Husgafvel-Pursiainen et al., 1993). This nucleotide substitution is consistently induced by benzopyrene, a tobacco carcinogen, in experimental studies (Fujji et al., 1990). We found no significant association between tobacco and mutation spectrum, regardless of alcohol consumption (data not shown). Little is known concerning Ki-ras mutations and alcohol drinking. The prevalence of ras mutations in alcohol-related cancers (stomach, liver, and upper-aero-digestive-tract tumors) is reportedly low. We here report an association between Ki-ras mutations and alcohol consumption. Our results suggested an interaction between tobacco and alcohol. The possibility of such interaction has received little attention until now. A recent study by Silverman et al. (1995) indicated that moderate alcohol drinking does not increase the risk of EPC. However, black non-smokers who had heavy alcohol consumption did have an increased risk; such increase was not observed among smokers, suggesting that a detrimental effect of alcohol drinking might be stronger among non-smokers than among smokers. There is currently no information on a possible interaction among tobacco, alcohol consumption and Ki-ras mutations at any tumor site in humans. Two studies have described the relationship between tobacco, alcohol and mutations in the p53 gene in patients with upper-aero-digestive-tract cancers. Brennan et al. (1995) observed a higher risk for p53 mutations in patients who smoked and drank in comparison with those who only smoked, suggesting a synergistic interaction. Franceschi et al. (1995) did not find such interaction, although in their study only immunohistochemical techniques were used. The only report on p53 alterations and tobacco in EPC showed that nuclear accumulation of p53 was more common among non-smokers than among smokers; alcohol drinking was not assessed (Lee et al., 1995). The antagonistic interaction between tobacco and alcohol consumption suggested by our results must be interpreted cautiously. If it is confirmed by subsequent studies, a number of theoretical issues would need to be addressed: what is the role of the pancreas in the metabolism of carcinogens involved in EPC? can alcohol consumption modify the metabolism of tobacco-derived carcinogens potentially involved in the development of EPC? do tobacco and alcohol compounds compete for the same metabolic pathways? Foster et al. (1993) found higher levels of P450 enzymes in pancreas and liver tissue from patients with chronic pancreatitis and pancreas cancer than in organ donors. Recent work has confirmed that CYP2E1, which is alcohol-inducible, and CYP1A1, which is involved in the metabolism of aromatic hydrocarbons, are present in the exocrine pancreas (Chassagne et al., 1995). Limitations and concluding remarks This study is an example of how the molecular classification of tumors may be of help to better assess risk estimates for specific 666 MALATS ET AL. exposures. Nevertheless, several study limitations must be noted. First, although this is one of the largest published series on Ki-ras mutations in EPC, the number of cases analyzed is small, and risk estimates lack precision. Second, information on tobacco and alcohol consumption was obtained from medical records and could be subject to a misclassification bias leading to underestimation of the association; however, the interaction between alcohol and tobacco was observed regardless of the cut-off used to categorize alcohol drinking. Finally, the interaction could be an effect of the statistical model used. Yet, the calibration and discrimination tests of the multiplicative model, including the interaction term, indicated that the model fitted the data well. Moreover, similar results were obtained using an additive model (data not shown). Whereas our findings need to be confirmed in larger, preferably prospective studies of unselected groups of patients, they provide new leads on the clinico-pathological correlates of Ki-ras mutations, and on the role of tobacco and alcohol in pancreatic diseases. ACKNOWLEDGEMENTS The study was partly supported by the Fondo de Investigación Sanitaria (FIS) (grants 91/0595 and 92/0007), by Fundación Salud 2.000 (grant Serono), by the Institut Municipal d’Investigació Mèdica (IMIM) (grant 875138/9), and by the Comissió Interdepartamental de Recerca i Innovació Tecnològica (CIRIT) (grants GRQ93-9301 and SGR 434). We are grateful to Dr. C. Balagué, Dr. G. Berrozpe, Ms. E. Carrillo, Mr. D.J. MacFarlane, Dr. A. Salas, Ms. A. Serrat, Dr. T. Thomson and Dr. M. Torà for valuable contributions, and to Mr. L. Español, Ms. P. Barbas and Ms. H. Martinez for secretarial assistance. REFERENCES ALMOGUERA, C., SHIBATA, D., FORRESTER, K., MARTIN, J., ARNHEIM, N. and PERUCHO, M., Most human carcinomas of the exocrine pancreas contain mutant c-Ki-ras genes. Cell, 53, 549–554 (1988). BERROZPE, G., SCHAEFFER, J., PEINADO, M.A., REAL, F.X. and PERUCHO, M., Comparative analysis of mutations in p53 and Ki-ras genes in pancreas cancer. Int. J. Cancer, 58, 185–191 (1994). BOYLE, P., HSIEH, C.C., MAISONNEUVE, P., LA VECCHIA, C., MACFARLANE, G.J., WALKER, A.M. and TRICHOPOULOS, D., Epidemiology of pancreas cancer (1988). Int. J. Pancreatol., 5, 327–346 (1989). BRENNAN, J.A., BOYLE, J.O., KOCH, W.M., GOODMAN, R.H., EBY, Y.J., COUCH, M.J., FORASTIERE, A.A. and SIDRANSKY, D., Association between cigarette smoking and mutation of the p53 gene in squamous-cell carcinoma of the head and neck. New Engl. J. Med., 332, 712–717 (1995). BUENO DE MESQUITA, H.B., MAISONNEUVE, P., MOERMAN, C.J., RUNIA, S. and BOYLE, P., Life-time history of smoking and exocrine carcinoma of the pancreas: a population-based case-control study in The Netherlands. Int. J. Cancer, 49, 816–822 (1991). BUENO DE MESQUITA, H.B., MAISONNEUVE, P., MOERMAN, C.J., RUNIA, S. and BOYLE, P., Lifetime consumption of alcoholic beverages, tea and coffee and exocrine carcinoma of the pancreas: a population-based case-control study in The Netherlands. Int. J. Cancer, 50, 514–522 (1992). CALDAS, C., HAHN, S.A., HRUBAN, R.H., REDSTON, M.S., YEO, C.J. and KERN, S.E., Detection of K-ras mutations in the stool of patients with pancreatic adenocarcinoma and pancreatic ductal hyperplasia. Cancer Res., 54, 3568–3573 (1994). CALDAS, C. and KERN, S.E., K-ras mutation and pancreatic adenocarcinoma. Int. J. Pancreatol., 18, 1–6 (1995). CAPELLÀ, G., CRONAUER-MITRA, S., PEINADO, M.A. and PERUCHO, M., Frequency and spectrum of mutations at codons 12 and 13 of the c-K-ras gene in human tumors. Environ. Hlth Perspect., 93, 125–131 (1991). CARRIAGA, M.T. and HENSON, D.E., Liver, gallbladder, extrahepatic bile ducts, and pancreas. Cancer, 75, 171s–190s (1995). CHASSAGNE, P., DAUJAT, M., MAUREL, P., MICHEL, H. and LARREY, D., Cytochromes p-450 1A1 and 2E1 are present in human pancreas. Evidence by molecular biology. Gastroenterology, 108, A348 (1995). FINKELSTEIN, S.D., PRYGODZKI, R., PRICOLO, V.E., SAYEGH, R., BAKKER, A., SWALSKY, P.A. and KELLER, G., Ki-ras-2 topographic genotyping of pancreatic adenocarcinoma. Arch. Surg., 129, 367–372 (1994). FOSTER, E.C., IDLE, J.R., HARDWICK, J.P., BARS, R., SCOTT, P. and BRAGANZA, J.M., Induction of drug-metabolizing enzymes in human pancreatic cancer and chronic pancreatitis. J. Pathol., 169, 457–463 (1993). FRANCESCHI, S., GLOGHINI, A., MAESTRO, R., BARZAN, L., BIDOLI, E., TALAMINI, R., VUKOSAVLJENIC, T., CARBONE, A. and BOIOCCHI, M., Analysis of the p53 gene in relation to tobacco and alcohol in cancers of the upper aero-digestive tract. Int. J. Cancer, 60, 872–876 (1995). FUJJI, H., EGAMI, H., CHANEY, W., POUR, P. and PELLIG, J., Pancreatic ductal adenocarcinomas induced in Syriam hamsters by N-nitrosobis (2oxopropyl) amine contain a c-Ki-ras oncogene with point-mutated codon 12. Mol. Carcinogen., 3, 296–301 (1990). GRÜNEWALD, K., LYONS, J., FRÖHLICH, A., FEICHTINGER, H., WEGER, R.A., SCHWAB, G., JANSSEN, J.W.G. and BAETRAM, C.R., High frequency of Ki-ras codon-12 mutations in pancreatic adenocarcinomas. Int. J. Cancer, 43, 1037–1041 (1989). HADDOCK, G. and CARTER, D.C., Aetiology of pancreatic cancer. Brit. J. Surg., 77, 1159–1166 (1990). HOORENS, A., LEMOINE, N.R., MCLELLAN, E., MOROHOSHI, T., KAMISAWA, T., HEITZ, P.U., STAMM, B., RUSCHOFF, J., WIEDENMANN, B. and KLOPPEL, G., Pancreatic acinar-cell carcinoma. An analysis of cell lineage markers, p53 expression, and Ki-ras mutation. Amer. J. Pathol., 143, 685–698 (1993). HOWE, G.R., JAIN, M., BURCH, J.D. and MILLER, A.B., Cigarette smoking and cancer of the pancreas: evidence from a population-based case-control study in Toronto, Canada. Int. J. Cancer, 47, 323–328 (1991). HRUBAN, R.H., VAN MANSFELD, A.D., OFFERHAUS, G.J., VAN WEERING, D.H., ALLISON, D.C., GOODMAN, S.N., KENSLER, T.W., BOSE, K.K., CAMERON, J.L. and BOS, J.L., Ki-ras oncogene activation in adenocarcinoma of the human pancreas. A study of 82 carcinomas using a combination of mutant-enriched polymerase-chain-reaction analysis and allele-specific oligonucleotide hybridization. Amer. J. Pathol., 143, 545–554 (1993). HUSGAFVEL-PURSIAINEN, K., HACKMAN, P., RINDANPÄÄ, M., ANTTILA, S., KARJALAINEN, A., PARTANEN, T., TAIKINA-AHO, O., HEIKKILA, L. and VAINIO, H., Ki-ras mutations in human adenocarcinoma of the lung: association with smoking and occupational exposure to asbestos. Int. J. Cancer, 53, 250–256 (1993). IGUCHI, H., SUGANO, K., FUKAYAMA, N., OHKURA, H., SADAMOTO, K., OHKOSHI, K., SEO, Y., TOMODA, H., FUNAKOSHI, A. and WAKASUGI, H., Analysis of Ki-ras codon-12 mutations in the duodenal juice of patients with pancreatic cancer. Gastroenterology, 110, 221–226 (1996). LEE, E.Y., CIBULL, M.L., O’DANIEL-PEARCE, E., STRODEL, W.E. and JENNINGS, C.D., Expression of p53 protein in pancreatic adenocarcinoma. Relationship to cigarette smoking. Int. J. Pancreatol., 17, 237–242 (1995). MALATS, N., PORTA, M., PIÑOL, J.L., COROMINAS, J.M., RIFÀ, J. and REAL, F.X., FOR THE PANKRAS I PROJECT INVESTIGATORS, Ki-ras mutations as a prognostic factor in extra-hepatic-bile-system cancer. J. clin. Oncol., 13, 1679–1686 (1995). MARIYAMA, M., KISHI, K., NAKAMURA, K., OBATA, H. and NISHIMURA, S., Frequency and types of point mutation at the 12th codon of the c-Ki-ras gene found in pancreatic cancers from Japanese patients. Jap. J. Cancer Res., 80, 622–626 (1989). MOTOJIMA, K., URANO, T., NAGATA, Y., SHIKU, H., TSURIFUNE, T. and KANEMATSU, T., Detection of point mutations in the Kirsten-ras oncogene provides evidence for the multicentricity of pancreatic carcinoma. Ann. Surg., 217, 138–143 (1993). NAGATA, Y., ABE, M., MOTOSHIMA, K., NAKAYAMA, E. and SHIKU, H., Frequent glycine-to-aspartic-acid mutations at codon 12 of c-Ki-ras in human pancreatic cancer in Japanese. Jap. J. Cancer Res., 81, 135–140 (1990). PELLEGATA, N.S., SESSA, F., RENAULT, B., BONATO, M., LEONE, B.E., SOLCIA, E. and RANZANI, G.N., K-ras and p53 gene mutations in pancreatic cancer: ductal and non-ductal tumors progress through different genetic lesions. Cancer Res., 54, 1556–1560 (1994). PORTA, M., MALATS, N., PIÑOL, J.L., RIFÀ, J., ANDREU, M. and REAL, F.X., FOR THE PANKRAS I PROJECT INVESTIGATORS, Diagnostic certainty and potential for misclassification in exocrine pancreatic cancer. J. clin. Epidemiol., 47, 1069–1079 (1994). PORTA, M., MALATS, N., PIÑOL, J.L., REAL, F.X. and RIFÀ, J., Relevance of misclassification of disease status in epidemiologic studies of exocrine pancreatic cancer. J. clin. Epidemiol., 49, 602–603 (1996). Ki-ras MUTATIONS IN EXOCRINE PANCREATIC CANCER SCARPA, A., CAPELLI, P., VILLANUEVA, A., ZAMBONI, G., LLUÍS, F., ACCOLA, R., MARIUZZI, G. and CAPELLÀ, G., Pancreatic cancer in Europe: Ki-ras-genemutation pattern shows geographical differences. Int. J. Cancer, 57, 167–171 (1994). SCHAEFFER, B.K., GLASNER, S., KUHLMANN, E., MYLES, J.L. and LONGNECKER, D.S., Mutated c-K-ras in small pancreatic adenocarcinomas. Pancreas, 9, 161–165 (1994). SHIBATA, D., ALMOGUERA, C., FORRESTER, K., DUNITZ, J., MARTIN, S.E., COSGROVE, M.M., PERUCHO, M. and ARNHEIM, N., Detection of c-K-ras mutations in fine-needle aspirates from human pancreatic adenocarcinomas. Cancer Res., 50, 1279–1283 (1990). SILVERMAN, D.T. and 12 OTHERS, Alcohol and pancreatic cancer in blacks and whites in the United States. Cancer Res., 55, 4899–4905 (1995). TABATA, T., FUJIMORI, T., MAEDA, S., YAMAMOTO, M. and SAITOH, Y., The role of ras mutations in pancreatic cancer, pre-cancerous lesions and chronic pancreatitis. Int. J. Pancreatol., 14, 237–244 (1993). TADA, M., OHASHI, M., SHIRATORI, Y., OKUDAIRA, T., KOMATSU, Y., KAWABE, T., YOSHIDA, H., MACHINAMI, R., KISHI, K. and OMATA, M., Analysis of K-ras gene mutations in hyperplastic duct cells of the pancreas without pancreatic disease. Gastroenterology, 110, 227–231 (1996). 667 TADA, M., YOKOSUKA, O., OMATA, M., OHTO, M., ISONO, K., Analysis of ras gene mutations in biliary and pancreatic tumors by polymerase chain reaction and direct sequencing. Cancer, 66, 930–935 (1990). TAYLOR, J.A., SANDLER, D.P., BLOOMFIELD, C.D., SHORE, D.L., BALL, E.D., NEUBAUER, A., MCINTYRE, O.R. and LIU, E., ras oncogene activation and occupational exposures in acute myeloid leukemia. J. nat. Cancer Inst., 84, 1626–1632 (1992). URRUTIA, R. and DIMAGNO, E.P., Genetic markers: the key to early diagnosis and improved survival in pancreatic cancer? (Editorial) Gastroenterology, 110, 306–310 (1996). WARSHAW, L.A. and FERNÁNDEZ DEL CASTILLO, C., Pancreatic carcinoma. New Engl. J. Med., 326, 455–465 (1992). YANAGISAWA, A., OHASHI, K., HORI, M., TAKAGI, K., KITAGAWA, T., SUGANO, H. and KATO, Y., Ductectatic-type mucinous cystadenoma and cystadenocarcinoma of the human pancreas: a novel clinicopathological entity. Jap. J. Cancer Res., 84, 474–479 (1993a). YANAGISAWA, A., OHTAKE, K., OHASHI, K., HORI, M., KITAGAWA, T., SUGANO, H. and KATO, Y., Frequent c-Ki-ras oncogene activation in mucous-cell hyperplasias of pancreas suffering from chronic inflammation. Cancer Res., 53, 953–956 (1993b).