Family History and the Risk of Liver, Gallbladder, and Pancreatic Cancer'1

Vol. 3, 209-2 12, April/May 1994 Cancer Epidemiology, Biomarkers & Prevention 209

Family History and the Risk of Liver, Gallbladder, and Pancreatic Cancer’1

Esteve Fernandez, Carlo La Vecchia,2 Barbara D’Avanzo, Introduction Eva Negri, and Silvia Franceschi Although several case reports and some formal epidemio- Institut Municipal d’lnvestigacio M#{232}dica(IMIM), Universitat AutOnoma de logical studies have addressed the issue of familial aggre- Barcelona (UAB), Dr. Aiguader 80, 08003 Barcelona, Catalonia, Spain gation ofpancreas (1-6), liver(7-1 1), and gallbladder cancer IE. F.]; Istituto di Ricerche Farmacologiche “Mario Negri,” Via Eritrea 62, (1 2, 1 3), few of them have provided relative or population 20157 Milano, Italy ]E. F., C. L. V., B. D., E. N.]; Istituto di Biometria e attributable risk estimates according to selected indicators of Statistica Medica, Universit#{224} di Milano, Via Venezian 1, 20133 Milano, Italy IC. L. V.]; and Centro di Riferimento Oncologico, family history of cancer, after allowance for major recog- 33081 Aviano (Pordenone), Italy IS. F.l nized potential confounding factors. Further, since there is consistent evidence that a few digestive tract cancers share some common risk factors, ei- Abstract ther environmental (14) or potentially genetic (15), com- The relationship between family history of selected bined analyses ofthe pattern of risk for different cancers may provide useful information. neoplasms in first-degree relatives and the risk of We decided therefore to analyze family histories of se- pancreatic, liver, and gallbladder cancer was lected neoplasms in a case-control study of pancreatic, liver, investigated using data from a case-control study and gallbladder cancer conducted in northern Italy. conducted in northern Italy on 320 histologically confirmed incident cases of liver cancer, 58 of gallbladder cancer, 362 of pancreatic cancer, and 1408 Subjects and Methods controls admitted to the hospital for acute, The data were derived from an ongoing series of case-control nonneoplastic, nondigestive tract disorders. Significant studies, on the basis of a network of teaching and general associations were observed between family history of hospitals in the Greater Milan area (i.e., the largest urban hepatocellular carcinoma and primary liver cancer area in northern Italy, with approximately 4 million inhab- itants). Recruitment of cases of several digestive tract neo- [relative risk (RR) = 2.4; 95% confidence interval (Cl), 1 .3 to 4.4], between family history of pancreatic cancer plasms and of the corresponding controls began in January 1983, and the present analysis is based on data collected and pancreatic cancer (RR = 3.0; 95% Cl, 1.4 to 6.6), before December 1992. and between family history of gallbladder cancer and The general design of this investigation has been pre- gallbladder cancer (RR = 13.9; 95% Cl, 1.2 to 1619). viously described (1 6, 1 7). In summary, trained interviewers The elevated risk of liver cancer associated with family identified and questioned cases of selected cancer sites and history was not materially modified by adjustment for controls admitted to hospital using a structured question- tobacco, alcohol, and personal history of cirrhosis and naire, including information on sociodemographic factors, hepatitis (RR = L9; 95% CI, 15 to 5.3) Similarly, personal characteristics, and lifestyle habits (such as smok- the risk for pancreatic cancer did not appreciably ing, alcohol, coffee, and other methylxanthine-contain ing change after allowance for tobacco, alcohol, dietary beverage consumption), frequency of consumption of a few factors, and medical history of diabetes and pancreatitis selected indicator foods, a problem-oriented medical his- (RR = 2.8; 95% Cl, 13 to 6.3). This pattern of risk tory, and history of use of oral contraceptives and other fe- would support the existence of a genetic component in male hormone preparations. A family history of cancer, in- the familial aggregation of liver and pancreatic cancer. cluding major digestive tract neoplasms (esophageal, In terms of population attributable risk, approximately stomach, large bowel, liver, gallbladder, and pancreatic), 3% of the newly diagnosed liver and pancreatic cancers was specifically investigated for first-degree relatives. would be related to this familial component. Cases The cases included in the present analysis were patients under 75 years of age with histologically confirmed incident (i.e., diagnosed within the year before interview) cancers of the liver (n = 320: 235 males and 85 females), gallbladder(n = 58: 27 males and 31 females), and pancreas Received 7/8/93; revised 9/28/93; accepted 9/28/93. (n = 362: 229 males and 1 33 females). They were admitted 1 This work was conducted within the framework of the CNR (Italian National to the National Cancer Institute, to several university hos- Research Council) Applied Project “Clinical Applications of Oncological Re- search” (Contract No.92.02384.PF39) and with the contributions ofthe Italian pitals, and to the Ospedale Maggiore of Milan, which in- Association for Cancer Research, the Italian League Against Tumours, Milan, cludes the four largest teaching and general hospitals in and A. Marchegiano Borgomainerio. The stay of E. Fernandez at Istituto Mario Milan. All the interviews were conducted in hospital and Negri (Milan) was partly supported by a grant ofthe Commission for Research restricted to identified surviving patients, in the absence of and Technology (CIRIT) of the Government of Catalonia (Grant EE93-1 28). 2 To whom requests for reprints should be addressed, at Istituto di Ricerche search for proxies for deceased ones. No information was Farmacologiche “Mario Negri,” via Eritrea, 62-201 57 Milano, Italy. collected on patients without pathological confirmation, or

Table 1 Distribution of 320 cases of liver cancer, 58 cases of gallbladder Table 2 Distribution of selected cancers in first degree relatives of 320 cancer, 362 cases of pancreatic cancer, and 1408 controls according to cases of liver cancer, 58 of gallbladder cancer, 362 of pancreatic cancer, sex and age. Milan, Italy 1983-1 992 and 1408 controls.” Milan, Italy 1983-1 992

Liver Gallbladder Pancreatic Liver Gallbladder Pancreatic Controls . ‘ Controls cancer cancer cancer e 0 cancer cancer cancer cancer in relatives No. (“1#{176}) No. (%,) No. (%) No. (%) No. (%) No. (%) No. (%) No. (%)

Sex Esophagus 5 (1.6) 4 (1.1) 7 (0.5) Males 235 (73.4) 27 (46.6) 229 (63.3) 1031 (73.2) Stomach 16 (5.1) 3 (5.4) 14 (3.9) 38 (2.8) Females 85 (26.6) 31 (53.4) 133 (36.7) 377 (26.8) Colorectum 9(2.9) 6(1.7) 22(1.6) Liver 19(6.1) 2(3.6) 9(2.5) 34(2.5) Age group (yr) Gallbladder 1 (1.8) 4(1.1) 2(0.1) <45 40 (12.5) 5 (8.6) 24 (6.6) 264 (18.8) Pancreas 1 (1.8) 14(3.9) 15(1.1) 45-54 56 (17.5) 1 1 (19.0) 84 (23.2) 382 (27.1) a Information is missing for some individuals. 55-64 1 40 (43.8) 20 (34.5) 1 36 (37.6) 459 (32.6) 65-74 84 (26.2) 22 (37.9) 1 1 8 (32.6) 303 (21 .5)

Total 320)100) 58(100) 362(100) 1408(100) Table 3 RR” and 95% Cl of liver, gallbladder and pancreatic cancer according to history of selected cancers in first-degree relatives. Milan, Italy 1983-1992

Liver cancer Gallbladder cancer Pancreatic cancer outside the network of participant hospitals. This, however, Site of cancer in relatives RR RR RR includes the major public hospitals in the area under inves- (95% Cl) (95%, Cl) (95% Cl) tigation, where the large majority of patients with serious neoplasms is admitted. The median age was 60 years for Esophagus 2.4 (0.7-7.6) 2.4 (0.7-8.3) Stomach 1 .7 (0.9-3.2) 1 .9 (0.5-6.5) 1 .4 (0.7-2.7) liver, 61 years for gallbladder, and 60 years for pancreatic Intestines 2.0 (0.9-4.5) 1 .2 (0.5-3.0) cancer cases. Liver 2.4 (1 .3-4.4) 1 .2 (0.3-5.4) 0.9 (0.4-1 .9) Controls. The comparison group included 1408 patients Gallbladder 13.9 (1.2-163.9) 9.9 (2.4-41.8) (1031 males and 377 females) younger than 75 years (me- Pancreas 1 .4 (0.2-1 1 .7) 3.0 (1 .4-6.6)

dian age, 56 years) admitted for a wide spectrum of acute, .‘ Adjusted for age, sex, area of residence, and education by multiple logistic nonneoplastic, nondigestive disorders to the Ospedale Mag- regression. giore and the same teaching hospitals where cases were re- cruited. Thirty-four % had traumatic conditions, 1 7% had nontraumatic orthopedic disorders, 37% had acute surgical conditions, and 1 2% had other miscellaneous diseases, such reported a family history of stomach cancer, 34 (2.5%) of as ear, nose, and throat, skin, or dental disorders. The catch- liver cancer, and 22 (1 .6%) of large bowel cancer. Among ment area of cases and controls was comparable: more cases, the highest proportions were observed for family his- than 80% of cases and controls resided in the same re- tory ofliver cancer among patients with primary liver cancer gion, Lombardy, and more than 9O% came from Northern (6.1 %), for family history of stomach cancer among cases of Italy. Participation rate was over 95% for both cases and liver cancer (5.1 %), gallbladder cancer (5.4%), and pancre- controls. As for cases, all the data were collected by direct atic cancer (3.9%), and for family history of pancreatic can- interview in hospital. The distribution of cases of various cer among pancreatic cancer cases (3.9%). cancer sites and controls according to sex and age group The corresponding RRs are given in Table 3. Statistically is shown in Table 1. significant increased risks were observed between family Data Analysis. RR and the corresponding 95% CI of pan- history of hepatocellular carcinoma and family history of

creatic, liver and gallbladder cancer, according to history of liver cancer (RR = 2.4) and between family history of pan- selected cancers among first-degree relatives were first com- creatic cancer and pancreatic cancer (RR = 3.0). A tendency puted using unconditional multiple logistic regression (18), of excess risk was observed for a family history of a number including terms for age, sex, area of residence, and educa- of other digestive sites, specifically for liver cancer with RRs tion. In addition, when significant associations were ob- of 1 .7, 2.0, and 2.4 for family history of stomach, intestine, served, further allowance was also made for selected van- and esophageal cancer, respectively. None of these esti- ables (smoking habits, alcohol consumption, and selected mates, however, was significant. History of gallbladder can- medical history). cer was reported by one case of gallbladder cancer and two controls, yielding a RR of 1 3.9, which, however, had a very wide confidence interval. The relationships between family Results history of liver cancer and the risk of hepatocellular canci- The distribution ofcases ofliver, pancreatic, and gallbladder noma and between family history of pancreatic cancer and cancer, and of the comparison group, according to history the risk of pancreatic cancer are further analyzed in terms of in first-degree relatives of cancers of the esophagus, stom- simultaneous allowance for several known or potential con- ach, intestines, liver, gallbladder, and pancreas is shown in founding factors and in separate strata of sex and age (Table Table 2. Thirty-eight (2.8%) individuals in the control group 4). With reference to liver cancer, adjustment for smoking, alcohol consumption, history of cirrhosis, and hepatitis did not appreciably modify the risk estimate (RR = 2.9). No significant interaction emerged with sex or age. Similarly, for

I The abbreviations used are: RR, relative risk estimate; Cl, confidence pancreatic cancer, adjustment for smoking habits, alcohol interval. consumption, history of pancreatitis, and diabetes did not

Table 4 RR (and 95% Cl) of liver cancer and pancreatic cancer was restricted to first-degree relatives and collected through according to family history of liver and, respectively, pancreatic cancer, interview only, in the absence of a clinical on histological sex, and age group. Milan, Italy 1983-1 992 evidence of the diagnosis of cancer among relatives. Addi-

Cases Controls tional sources of information (from other relatives, on clinical RR (95% CI) notes) were searched only in the case of uncertainty on the No. (%) No. (%) presence and type of neoplasms. The absence of a genen- Liver cancer’ alized increased risk between various different cancer sites Total 19(6.1) 34(2.5) 2.9(1.5-5.3) is suggestive ofthe existence of a real association for specific sites. In principle, cancer patients might tend to selectively Sex assign the primary site of their cancer to tumors in relatives. Males 15(6.4) 21 (2.0) 3.1 (1.5-6.3) Females 4(4.7) 13(3.4) 1.6(0.5-5.6) However, most cancer patients in Italy have no precise information on their diagnoses during their hospital stay, and Age group (yr) several patients even after hospital discharge. This potential <60 9(5.8) 22(2.5) 2.3(1.0-5.1) bias, therefore, is unlikely to have played any major role. 60 10(6.0) 12(2.3) 3,0(1.2-7.3) The excess risk for liven and pancreatic cancer might be partly explained by misclassification, due to the difficulties Pancreatic cancer” in diagnosis ofthese neoplasms, and particularly, to the pos- Total 14)3.9) 15(1.1) 2.8(1.3-6.3) sibility of misclassification of liven metastasis as primary liven Sex tumors. For these reasons, too, case identification was re- Males 5 (2.2) 9 (0.9) 2.5 (0.8-7.9) stnicted to pathologically confirmed neoplasms, since seni- Females 9 (6.8) 6 (1 .6) 3.9 (1 .3-1 1 .9) ous misclassification can be induced by inclusion of clinical diagnosis of liven cancer. Among other strengths ofthe study, Age group (yr) there are the practically complete participation, the com- <60 6(3.4) 11 (1.2) 2.2 (0.8-6.5) parable catchment area of cases and controls (i.e., control 60 8 (4.4) 4 (0.8) 4.0 (1.1-1 4.3) individuals would have been referred, if affected by liver, * Adjusted for sex, age, area of residence, education, smoking habits, alcohol gallbladder or pancreatic cancer, to the same hospitals consumption, history ofcirrhosis and hepatitis by multiple logistic regression. where cases were identified), and the comparable setting of ** Adjusted for sex, age, area of residence, education, smoking habits, alcohol consumption, history of pancreatitis and diabetes by multiple logistic regres- data collection (20). With reference to confounding, we sion. were able to adequately allowfon age, and a number of other potentially relevant covaniates. The present findings, therefore, indicate and quantify the presence ofa significantfamily aggregation for both liven appreciably modify the overall estimate (RR = 2.8). Like- and pancreatic cancer. However, using this study design it wise, no interaction was observed with sex or age. was not possible to distinguish clearly between the envi- nonmental or genetic component of such a familial aggre- Discussion gation. For instance, with reference to liver cancer, the familial risk could be partly or largely due to family The present analysis provides additional quantitative infor- aggregation of hepatitis B infection, which is a major de- mation on the existence of a familial component for liver, tenminant ofpnimary liven cancer (1 7, 21 -23). For pancreatic gallbladder, and pancreatic cancer. Liver and pancreatic cancer, at least part of the association could be related to cancers appear to be approximately 3-fold more frequent tobacco smoking, which is a recognized risk factor for the among patients with a positive history of the corresponding disease (24) and tends to aggregate in families, too. cancers. In terms of population attributable risk (1 9), ap- The risk estimates, however, were not materially mod i- proximately a 3% of the pancreatic and liven cancers would fied after adjustment for major identified covaniates, includ- be related to this familial component. There was also some ing tobacco, alcohol, and past medical conditions related to evidence of a familial aggregation for gallbladder cancer, liven cancer (cirrhosis and hepatitis) or pancreatic cancer although it is based on very small absolute numbers. (pancreatitis and diabetes). This would, therefore, support, These findings are consistent with risk estimates from although indirectly, the existence of a genetic component in various other studies. Besides few reports offamilies on clus- the familial aggregation of liven and pancreatic cancer. tens of cancer (9, 1 1 ), a case-control study conducted in Taiwan (7) reported an elevated risk (RR = 4.6) of borderline significance for liven cancer associated with a positive family Acknowledgments history of this cancer; the estimated population attributable The authors wish to thank J. Baggott, I. Garimoldi, H. Martinez, P. Barbas, and risk was 3.5%. For pancreatic cancer, Ghadinian et a!. (1 ) in the G. A. Pfeiffer Memorial Library staff for editorial assistance. a case-control study in Canada found an increased risk for the familial component (RR = 1 5.1 ; 95% CI, 2.0-31 1 .2), in References the absence of apparent differences in environmental risk 1 . Ghadirian, P., Boyle, P., Simard, A., Baillargeon, I., Maisonneuve, P., and factors between familial and nonfamilial cases. Perret, C. Reported family aggregation of pancreatic cancer within a Although an apparent association between the cancer population-based case-control study in the Francophone community in sites studied and family history of the same neoplasms in Montr#{233}al, Canada. Int. J. Pancreatol., 10: 183-196, 1991. first-degree relatives was observed in this as well as in pne- 2. Lynch, H. T., Fitzsimmons, M. L., Smyrk, T. C., Lanspa, S. I., Watson, P., vious studies (1-6), these results may be influenced by van- McClellan, J., and Lynch, J. F. Familial pancreatic cancer: clinicopathologic study of 18 nuclear families. Am. J. Gastroenterol., 85: 54-60, 1990. ous sources of bias. Recall bias may have produced some 3. Friedman, J. M., and Fialkow, P. J. Familial carcinoma of the pancreas. overestimation of risk, since patients suffering from a specific Clin. Genet., 9:463-469, 1976. cancer are more sensitized than noncancer patients toward 4. Reimer, R. R., Fraumeni, J. R. Jr., Ozols, R. F., and Bender, R. Pancreatic

recalling episodes of cancer in their relatives. Information cancer in father and son. Lancet, 1 : 91 1 , 1977.

5. Katkhouda, N., and Mouiel, J. Pancreatic cancer in mother and daughter. 16. La Vecchia, C., Negri, E., D’Avanzo, B., Ferraroni, M., Grarnenzi, A., ILetterl. Lancet, 2: 747, 1986. Savoldelli, R., Franceschi, S. Medical history, diet and pancreatic cancer. 6. Dat, N. M., and Sontag, S. J. Pancreatic carcinoma in brothers. ]Letterl. Oncology, 47: 463-466, 1990. Ann. Intern. Med., 97:282, 1982. 1 7. La Vecchia, C., Negri, E., D’Avanzo, B., Boyle, P., and Franceschi, S. 7. Chen, C. J., Liang, K. Y., Chang, A. S., Lu, S. N., Liaw, Y. F., Chang, Medical historyand pnimarylivercancer. Cancer Res., 50:6274-6277, 1990. W. Y., Sheen, M. C., and Lin, T. M. Effects of hepatitis B virus, alcohol drink- ing, cigarette smoking and familial tendency of hepatocellular carcinoma. 1 8. Breslow, N. E., and Day, N. E. Statistical Methods in Cancer Research: Hepatology, 13: 398-406, 1991. The Analyses of Case-Control Studies, IARC Sci. PubI. 32. Lyon, France: 8. Shen, F. M., Lee, M. K., Gong, H. M., Cai, X. Q., and King, M. C. Complex International Agency for Research on Cancer, 1980.

segregation analysis of primary hepatocellular carcinoma in Chinese families: 19. Bruzzi, P., Green, S. B., Byar, D. P., Brinton, L. A., and Schairer, C. Interaction of inherited susceptibility and hepatitis B viral infection. Am. J. Estimating the population attributable risk for multiple risk factors using case- Hum. Genet., 49: 88-93, 1991. control data. Am. J. Epiderniol., 122: 904-91 4, 1985. 9. Alberts, S. R., Lanier, A. P., McMahon, B. J., Harpster, A., Bulkow, L. R., Heyward, W. 1., and Murray, C. Clustering of hepatocellular carcinoma in 20. Kelly, J. P., Rosenberg, L., Kaufman, D. W., and Shapiro, S. Reliability of Alaska native families. Genet. Epiderniol., 8: 127-139, 1991. personal interview data in a hospital-based case-control study. Am. J. Epi- derniol., 131:79-90, 1990. 10. Liang, K. Y., and Beaty, H. M. Measuring familial aggregation by using odds-ratio regression models. Genet. Epiderniol., 8: 361-370, 1991. 21 . Trichopoulos, D., Day, N. E., Kaklamani, E., Tzonou, A., Munoz, N.,

1 1 . N’Dri, N., Hadengue, A., and Benhamou, J. P. Familial hepatocellular Zavitsanos, X., Koumantaki, Y., and Trichopoulou, A. Hepatitis B virus, to- carcinoma in B or non-A non-B cirrhosis. Report of 4 families. Presse Med., bacco smoking and ethanol consumption in the etiology of hepatocellular 20:207-209, 1991. carcinoma. Int. J. Cancer, 39: 45-49, 1987. 12. Garber, J. E., and Shipley, W. Carcinoma of the gall bladder in three 22. Yu, M. C., Mark, T., Hanisch, R., Peters, R. L., Henderson, B. E., and Pike, members of a family. ]Letterl. Cancer Genet. Cytogenet., 39: 1 41-1 42, 1989. M. C. Hepatitis, alcohol consumption, cigarette smoking, and hepatocellular 1 3. Devor, E. J., and Buechley, R. W. Gallbladder cancer in Hispanic New carcinoma in Los Angeles. Cancer Res., 43: 6077-6079, 1983. Mexicans. I. General population, 1957-1977. Cancer (Phila.), 45: 1705- 1712, 1980. 23. Beasley, P., Lin, C., Swang, L. Y., and Chien, C. S. Hepatocellular car- A prospective study of 32,707 men in Taiwan. 14. Byers, T. Diet and cancer: any progress in the interim? Cancer (Phila.), cinoma and hepatitis B virus. Lancet, 2:1129-1133, 1981. 62: 1713-1724, 1988. 15. Lynch, H. 1., Albano, W. A., Ruma, T. A., Schrnitz, G. D., Costello, 24. Boyle, P., Hsieh, C. C., Maisonneuve, P., La Vecchia, C., Macfarlane, K. A., and Lynch, J. F. Surveillance/management of an obligate gene carrier: G. J., Walker, A. M., and Trichopoulos, D. Epidemiology of pancreas cancer the cancer family syndrome. Gastroenterology, 84: 404-408, 1983. (1988). Int. J. Pancreatol., 5: 327-346, 1989.

E Fernandez, C La Vecchia, B D'Avanzo, et al.

Cancer Epidemiol Biomarkers Prev 1994;3:209-212.

Updated version Access the most recent version of this article at: http://cebp.aacrjournals.org/content/3/3/209

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cebp.aacrjournals.org/content/3/3/209. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.