Pancreatic cancer is 1 of the leading causes of death from cancer in Canada and is among the cancer types with the highest case-fatality rate. Pancreatic cancer is diagnosed in roughly equal numbers in males and females. There are few known risk factors for pancreatic cancer and few preventive or screening options are available for those considered to be at high risk.1
Smoking is the most important and most consistent environmental risk factor for pancreatic cancer; the risk is increased roughly 3-fold for men and women with heavy cigarette consumption.2, 3 Several dietary factors have been reported to be protective, including fresh fruits and vegetables, dietary fibre and vitamin C.4, 5 Other foods have been reported to increase the cancer risk, including salted and smoked foods, fried foods and refined sugar.5
It is clear that familial factors are also involved in susceptibility to pancreatic cancer.6 Several case-control studies have reported that patients with pancreatic cancer are more likely to have a relative affected with pancreatic cancer than are healthy controls. Falk et al.7 estimated the relative risk of pancreatic cancer to be 5.3 for first-degree relatives of patients with pancreatic cancer. Ghadirian et al.8 reported that 7.8% of French-Canadian patients with pancreatic cancer had an affected relative, compared to 0.6% of the controls. In an Italian study, the estimated relative risk for pancreatic cancer was 2.8 for those with an affected first-degree relative, after adjustment for smoking, alcohol, diet and selected medical factors.9
Many families with multiple individuals affected with pancreatic cancer have been described.8, 10, 11 In many families it appears that pancreatic cancer is the only site in excess. There are several familial cancer syndromes, however, that include pancreatic cancer and cancers of other types. Individuals who carry mutations in BRCA1 or BRCA2 are primarily at risk for breast and ovarian cancer, but appear to be at elevated risk for pancreatic cancer as well. The Breast Cancer Linkage Consortium estimate the relative risk to be 5.5 for pancreatic cancer to age 65 in BRCA2 carriers.12 In a study of Jewish men and women with pancreatic cancer, a BRCA2 founder mutation was seen in 4 of 39 patients (10.2%) and the lifetime pancreatic cancer risk was estimated to be 7%.13 The absolute risk for pancreatic cancer in carriers of BRCA1 mutations has not been estimated. Members of families with multiple cases of malignant melanoma, and who carry germline mutations in p16, are also at increased risk of pancreatic carcinoma. In some families the risk excess appears to be 13-fold.14
It is not clear how often pancreatic cancer is familial, nor what is the absolute lifetime risk of pancreatic cancer in first-degree relatives of probands. The range of cancer types seen in excess in relatives of pancreatic cases has not been clearly defined. The purpose of the present study is to estimate the risk of cancer of the pancreas and of other types among first-degree relatives of unselected pancreatic cancer patients, identified at teaching hospitals in Eastern Canada.
MATERIAL AND METHODS
Study subjects participated in a hospital-based case-control study, involving 9 hospitals in Ontario and Quebec. Potentially eligible patients with histologically-confirmed adenocarcinoma of the pancreas were approached in the physician's office or in the outpatient clinics (oncology, surgery and radiotherapy) of 4 tertiary care hospitals in Toronto (Princess Margaret Hospital, Mount Sinai Hospital, Toronto General Hospital, Wellesley Central Hospital) and in 5 hospitals in Montreal (Montreal General Hospital, Royal Victoria Hospital, Jewish General Hospital, Notre-Dame Hospital, Hotel-Dieu Hospital). A total of 174 cases were enrolled. There were 109 cases enrolled at the 4 University of Toronto hospitals and 65 cases from the McGill and University of Montreal teaching hospitals between 1995–99.
A total of 208 cases of pancreatic cancer were identified during the course of the study, of whom 30 cases died before interview. Four patients refused to participate. The remaining 174 cases were interviewed (84% participation rate).
For 157 of the cases, pancreatic cancer was the first occurrence of cancer; however 17 subjects had a previous diagnosis of cancer; including colorectal (4), breast (4) prostate (2), skin (2) and 1 each of melanoma, head and neck, bladder, and uterus. One of the breast cancers was in a male patient.
Controls were excluded if they had a past history of cancer of any type, excluding non-melanoma skin cancer. In Toronto, the spouses of the cases were the preferred controls because the great majority of spouses were from the same ethnic group as the case. It was possible to identify a spouse control for 59 of the 109 Toronto cases. It was not possible to identify a spouse control for the remaining cases, either because they were never married, they were divorced or widowed, or because the spouse was otherwise unwilling to participate. In Montreal, population-based controls were available. These controls were age-, ethnically- and gender-matched and were selected through random digit dialing. It was possible to identify a control for each of the 65 Montreal-based cases. An additional 12 controls from Montreal were suitable matches for the Toronto cases. In total, 136 controls were available for the 174 cases. The cases and controls are compared in Table I.
Table I. Some Characteristics of the Study Population
Cases (n = 174)
Controls (n = 136)
Age of diagnosis
Cases and controls in Toronto completed a self-administered questionnaire designed to identify all cancers in the first-degree relatives, including site of cancer and age of diagnosis. The Montreal cases were interviewed either in person (36%) or by telephone (74%). The current age or the age of death was recorded for the first-degree relatives without cancer. Details were missing for 27 first-degree relatives (24 relatives of cases and 3 relatives of controls) and these were excluded from the analysis. A total of 966 case relatives and 903 control relatives were available for study.
For each case and control, a questionnaire was administered either in person or by telephone. Subjects were asked about their age, smoking history, drinking history and past medical history. A detailed pedigree was constructed for each case and control. Information on all cancer diagnoses in first-degree relatives of cases and controls was recorded, including site and age of diagnosis. The current age or age of death was recorded for relatives unaffected with cancer.
The cumulative risk of cancer of all types, and of pancreatic cancer in particular, was estimated in the first-degree relatives of the pancreatic cancer patients and of the controls. The relative was considered to be at risk for cancer until the development of cancer, until death from another cause, until age 85 or until January 2000. The relative risk of cancer in case relatives compared to control relatives was then calculated using Cox proportional hazards models. Cumulative incidences of cancer were calculated using actuarial methods and the statistical significance for these differences were tested with the log-rank test. We excluded the children of the cases and controls because some of these children are common to both subgroups, and there were very few cases of cancer in children of either cases or controls.
Because it is possible that genetic factors are more pronounced in early-onset cases, and in cases with multiple primary cancers, comparisons were also made by dividing the patient sample into those diagnosed before and after age 60, and into those with and without a previous history of cancer.
We studied a total of 174 cases of pancreatic cancer. The cases were diagnosed at a mean age of 62.3 years (range: 30–88 years). The majority of the cases were male (60%) and most had a past history of smoking (58%).
We were able to identify a total of 136 eligible controls. There were 966 first-degree relatives of the cases and 903 first-degree relatives of the controls. Although the cases and controls differed with regards to gender (Table I) the relatives of the cases and controls were similar (Table II). The mean age of the case relatives was 61.8 years and the mean age of the control relatives was 61.7 years, and the gender ratio was approximately equal. Smoking information was available for 974 relatives (52%). Among this sample, 37.9% of the case relatives smoked, compared to 33.3% of the control relatives.
Table II. Comparison of First-Degree Relatives of Cases and First-Degree Relatives of Controls
Case relatives (n = 966)
Control relatives (n = 903)
A total of 150 cancers was reported in the 966 first-degree relatives of the cases (15.5%) and a total of 122 cases of cancer were reported in the 903 first-degree relatives of the controls (13.5%). We identified 16 pancreas cancer patients in the case relatives (1.7%) and only three pancreas cancer cases in the control relatives (0.3%). The other types of cancers observed in the relatives of the cases and controls are presented in Table III. The relative risks are based on a comparison of the cumulative risks of the specific cancers in the relatives of the cases, compared to the relatives of the controls, and the significance of the association is measured using the log-rank test.
Table III. Relative Risks for Cancers of Different Sites in First-Degree Relatives of Pancreas Cancer Cases Versus Controls
Number of cancers observed
Head and neck
Overall, the cumulative incidence of cancer of any type was 33.8% for the case relatives to age 85, compared to 32.0% for the control relatives (relative risk 1.15; p = 0.25). Only the risk of pancreatic cancer was significantly in excess (relative risk 5.0; p = 0.01) (Fig. 1). There were 16 cases with a relative with pancreatic cancer (9.2%) compared to 3 controls with an affected relative (2.2%). The absolute risk of pancreatic cancer to age 85 was 4.7% for the 966 first-degree relatives of the cases, compared to 1.4% for the first-degree relatives of the controls. Non-significant excesses were also seen for liver cancer, brain tumors, bladder cancers, cervical cancers and esophageal cancers (Table III).
An early age-of-onset is one of the characteristic features of a number of hereditary cancers. The mean age of onset of the 15 cases of pancreatic cancer with a relative affected with pancreatic cancer was 61.6 years, compared to a mean age of 62.4 years for the 159 non-familial cases (p = 0.92). The risk of pancreatic cancer, however, was higher for the relatives of cases diagnosed before age 60, than for the cases diagnosed after age 60 (relative risk 2.0; p = 0.18). Among first-degree relatives of cases diagnosed before age 60 the cumulative risk was 7.2%. Among first-degree relatives of patients diagnosed after age 60 the cumulative risk was 3.8%. The cumulative risk of any type of cancer in the first-degree relatives of the early-onset cases was 43.3%, compared to 30.4% for the relatives of the late-onset cases (relative risk = 1.45; p = 0.03).
A second characteristic of hereditary cancer syndromes is individuals with multiple primary cancers. A past history of cancer was reported for 17 cases of pancreatic cancer. The cumulative risk of pancreatic cancer in the first-degree relatives of the patients with multiple primary cancers was 12.3%, compared to 3.6% for the relatives of single primary cases (relative risk 2.2; p = 0.22) (Fig. 2).
Overall, the lifetime risk of any cancer was 53.0% in the smoking relatives of the cases, compared to 22.8% in the non-smoking relatives of the cases (p = 0.0007) and 36.4% for the smoking relative of the controls. We did not see a significant excess in the risk of lung cancer in the relatives of the pancreatic cancer cases compared to control relatives (Table III). This suggests that the familial aggregation of pancreatic cancer cannot be explained by familial smoking patterns. In our study we found a significant interaction between the age-of-onset of pancreatic cancer and the risk of lung cancer in the relatives. Among relatives of pancreatic cancer cases diagnosed before age 60, the cumulative risk of lung cancer was 7.6% and among relatives of pancreatic cancer cases diagnosed after age 60 the risk of lung cancer was 3.5% (relative risk 3.6; p =0.004) (Fig. 3). This association suggests that there is a common hereditary factor associated with both lung cancer and with early-onset pancreatic cancer.
We estimate that approximately 9% of cases of pancreatic cancer are familial. This is based on the observation that 16 of 174 pancreatic cancer patients had a first-degree relative with pancreatic cancer. We believe that the majority of familial cases have a hereditary basis. All hereditary cases, however, may not be familial, particularly if the penetrance of the predisposing gene is low. In this case, the hereditary proportion would exceed 10%.
The 10% fraction of hereditary cases cannot be explained on the basis of known susceptibility genes. In a previous study, we identified a BRCA1 mutation or BRCA2 or p16 mutation in 5 of the 102 cases (5%) from Toronto.17 Only 1 of these cases was from a family with a close relative with pancreatic cancer. Therefore, it appears that the syndrome of site-specific pancreatic cancer (pancreatic cancer without associated familial malignancies) is a distinct genetic syndrome that is not accounted for by known susceptibility genes. We have also investigated a possible association between genes involved in carcinogen metabolism and pancreatic cancer risk, but no positive associations were seen.15
It appears from our present study that there is a syndrome of familial pancreatic cancer, however it is not yet clear if there are other cancer types that are a manifestation of this syndrome. Hruban et al.16 reported an excess of cancer of the lung, breast and colon in the relatives of cases of familial pancreatic cancer. The patients, however, were selected for their family histories and were reported to a familial pancreatic cancer registry, and a comparable control group was not included. We did not see an excess of breast cancer in our study, and the excesses of colon and lung were small and non-significant.
Our study is a combined analysis of 2 different case-control studies, in Toronto and in Montreal. Although the study populations are different, we chose to combine the studies to increase the power and to improve the precision of the risk estimates. If the genetic factors differ in the 2 communities, this approach may not be valid. Because there were only 3 pancreatic cancers reported in the control relatives it was not possible to compare the relative risks separately for the 2 cities.
The risk estimates are based on comparison of cancer rates in the relatives of the cases and controls. Because the 2 groups of relatives were similar with respect to age, gender and ethnic group, this approach is valid, even though there are important differences in the cases and controls themselves. Furthermore, the principal findings relating to the familial risk of pancreatic cancer based on age of onset or multiple primary cancers is not dependent on the presence of a control group.
We were only able to obtain smoking histories on 52% of the relatives. In this sample the case relatives were more likely to have smoked than the control relatives (38% vs. 33%) but a difference of this size is not sufficient to explain the large difference seen in the pancreatic cancer experience of the 2 groups.
The diagnosis of cancer in the relatives were not confirmed, and it is possible that a patient with pancreatic cancer is more likely to accurately recall an affected relative with pancreatic cancer than is a control. In this case, the relative risk may be an over-estimate. On the other hand, if there has been inaccurate reporting of pancreatic cancer cases in relatives by cases and by controls, then this will likely result in an under-estimate of the true relative risk.
Our results indicate that there likely to be one or more genes that predispose to site-specific pancreatic cancer, and it should be possible to use a linkage approach to identify these. Because of the high case fatality rate associated with pancreatic cancer, however, it is often difficult to obtain blood samples from multiple living cases. New techniques use small quantities of DNA and it may be possible to perform linkage searches using DNA extracted from archived tumour specimens.
The lifetime risk for pancreatic cancer is 4.7% for first-degree relatives of pancreatic cancer cases, but this risk is greater for individuals who have a first-degree relative with pancreatic cancer diagnosed before age 60 (7.2%), or who have a relative with pancreatic cancer after another cancer diagnosis (12.3%). These high-risk individuals may be subjects for future studies of surveillance or chemoprevention.