Fax: (507) 284-1803
Risk of malignancy in first-degree relatives of patients with pancreatic carcinoma
Article first published online: 23 MAY 2005
Copyright © 2005 American Cancer Society
Volume 104, Issue 2, pages 388–394, 15 July 2005
How to Cite
McWilliams, R. R., Rabe, K. G., Olswold, C., De Andrade, M. and Petersen, G. M. (2005), Risk of malignancy in first-degree relatives of patients with pancreatic carcinoma. Cancer, 104: 388–394. doi: 10.1002/cncr.21166
- Issue published online: 29 JUN 2005
- Article first published online: 23 MAY 2005
- Manuscript Accepted: 9 MAR 2005
- Manuscript Revised: 2 FEB 2005
- Manuscript Received: 27 DEC 2004
- Lustgarten Foundation for Pancreas Cancer Research
- National Cancer Institute. Grant Number: R01 CA97075
- Mayo Clinic SPORE in Pancreatic Cancer. Grant Numbers: P20 CA 10270, R25T CA 92049
- pancreatic carcinoma;
- first-degree relative;
Approximately 5–10% of pancreatic carcinoma (PC) patients report a family history of the disease. In some families, mutations of tumor suppressor genes have been elucidated, but for most the causative gene remains unidentified. Counseling the families of PC patients regarding their risk of cancer remains problematic because little information is available.
The authors analyzed family history questionnaires completed by 426 unselected, sequential Mayo Clinic patients with PC. The prevalence of malignancy reported among 3355 of their first-degree relatives was compared with the Surveillance, Epidemiology, and End Results Project (SEER) 9 (2000) registry. Age-adjusted and gender-adjusted standardized incidence ratios (SIRs) were generated.
Greater than 130,000 person-years at risk for cancer among the first-degree relatives were analyzed. The risk of PC was found to be increased among the first-degree relatives of patients with PC (SIR of 1.88; 95% confidence interval [95% CI], 1.27–2.68), as was the risk of liver carcinoma (SIR of 2.70; 95% CI, 1.51–4.46). Lymphoma (SIR of 0.28; 95% CI, 0.12–0.55), bladder carcinoma (SIR of 0.55; 95% CI, 0.31–0.89), breast carcinoma (SIR of 0.73; 95% CI, 0.57–0.92), lung carcinoma (SIR of 0.62; 95% CI, 0.47–0.80), and prostate carcinoma (SIR of 0.71; 95% CI, 0.54–0.92) were found to be underrepresented. When the proband was age < 60 years, the risk of PC to first-degree relatives was found to be increased further (SIR of 2.86; 95% CI, 1.15–5.89). In this subgroup, no other malignancies were found to be significantly increased, although the risks of melanoma (SIR of 1.73; 95% CI, 0.70–3.57), ovarian carcinoma (SIR of 2.20; 95% CI, 0.72–5.12), and colon carcinoma (SIR of 1.37; 95% CI, 0.80–2.19) were suggestive.
There was a nearly twofold increased risk of PC in the first-degree relatives of PC probands. This risk was found to increase nearly threefold when patients were diagnosed before age 60 years. At the current time, in the absence of a pedigree suggestive of known familial cancer syndromes, the current study results do not support targeted screening for other malignancies in the first-degree relatives of patients with sporadic PC. Cancer 2005. © 2005 American Cancer Society.
Pancreatic carcinoma (PC) is a highly lethal and common malignancy, accounting for an annual incidence of 1–10/100,0001 and a mortality rate of > 95% by 5 years.2 Of the approximately 30,000 incident cases diagnosed each year in the U.S., approximately 5–10% of patients have a family history of PC.3 Genetic studies of these kindreds have led to the identification of mutations in the BRCA2 gene in approximately 12–20% of families, often in the absence of cases of breast or ovarian carcinoma.4, 5 In addition, mutations in the CDKN2A gene have been identified in familial PC kindreds, especially when melanoma is also observed in the families.6 Segregation analysis has alluded to a possible major gene conferring risk for PC, which to our knowledge has yet to be identified.7 As with most malignancies, somatic and environmental factors such as tobacco smoking8 and pancreatitis9 also appear to influence risk.
Given the factors that influence the risk of PC, a systematic analysis of nonpancreatic malignancy risk among the relatives of patients diagnosed with PC would meet two objectives. First, it would provide data for cancer risk assessment for relatives that would be useful for counseling purposes. Second, the analysis could detect potential clustering of other malignancies in families of patients with apparently “sporadic” cases of PC, yielding clues to its etiology. For example, if a high proportion of colorectal carcinoma was noted, a greater role of abnormalities in DNA mismatch repair genes might then be implicated. Or, if primarily smoking-related malignancies were noted to be increased, the significance of the role of tobacco exposure might be elucidated further. Reports of cancer risks in families have been approached using epidemiologic designs, although with variable results.
To clarify the role of variations in methodology and recruitment rates in prior reports, we conducted a study using our Mayo Clinic Pancreatic Cancer Patient Registry. Our large practice serves a region of the upper midwestern U.S., a region of largely Northern European ancestry. Also unique to our study is the ultra-rapid recruitment of patients (at the time of their first visit to the Mayo Clinic), in which all family history and risk factor data were self-reported by the index patients. We hypothesized that the increased rates of malignancies in these relatives existed due to either shared risk factors (such as smoking) or inherited cancer susceptibility syndromes, resulting from either known or unknown genes.
MATERIALS AND METHODS
Since October 2000, we have used ultra-rapid case-finding to enroll patients with PC into a clinic-based biospecimen and information resource. This methodology recruits nearly 80% of patients with PC who visit the Mayo Clinic. Informed, written consent is documented for all patients enrolled in the registry, who also self-complete a detailed questionnaire regarding risk factors and family history. Those with family histories in which at least one blood relative was diagnosed with PC have been further characterized for family-based studies. This family study portion of the registry is part of a collaboration with several other institutions (the University of Texas M. D. Anderson Cancer Center, the Dana-Farber Cancer Center, the Johns Hopkins Medical Institutions, Creighton University, Karmanos Cancer Institute, and Mount Sinai Hospital [Toronto, Ontario, Canada) known as the Pancreatic Cancer Genetic Epidemiology Consortium (PACGENE; R01 CA97075)10.
We analyzed the family histories provided through the questionnaires completed by 426 consecutive patients at the Mayo Clinic who provided informed consent between October 2000 and February 2004 and who had documented PC. Of these probands, 98.4% were white, with the remainder comprised of African-American, Middle Eastern, Hispanic, and Native-American patients. Approximately 55.6% of the probands were male.
Although pathologic verification of the relatives in the current study was not possible, the reporting of cancer incidence in first-degree relatives has been shown consistently to be 60–80% sensitive for a history of cancer, compared with population-based databases.11, 12 Beyond first-degree relatives, family history and reporting are less accurate. Therefore, we limited our analysis to first-degree relatives. Data retrieved from the questionnaires included type of cancer, patient age at and date of the cancer diagnosis, date of birth/death, and ever/never/unknown smoking status. Rates of malignancy for 14 common malignancies were compared with population data from the Surveillance, Epidemiology, and End Results Project (SEER) 9 (1973–2000),13 using rates adjusted for age and gender, concerning the risk of cancer incidence per person-year. Based on our data and the expected population rates in the SEER database, standardized incidence ratios (SIRs) were generated.
Individuals age younger than 20 years were excluded from the analysis because the risk of developing a solid tumor malignancy before this age is considered to be small. The medical records of all patient-recorded diagnoses were analyzed by a single physician (R.M.) to clarify the site of origin of the tumor. All probands had their diagnoses confirmed pathologically as PC when tissue was available. For those patients in whom no biopsy was performed (2 of 426 patients, 0.47%), the clinical diagnostic criteria of weight loss, pancreatic mass on imaging studies, the presence or absence of obstructive jaundice, and an elevated CA 19-9 antigen level were considered sufficient for a diagnosis of PC in the absence of pathologic confirmation.
For subanalyses, we also stratified relatives into children, parents, and siblings; smoker/nonsmoker/unknown; gender; and age of the proband at diagnosis (age < 60 or ≥ 60 years). The date of birth was imputed for 798 of 3355 relatives (23.8%); the date of death was imputed for 185 relatives (5.5%). The date of birth of relatives was imputed via the use of the date of birth of the proband, using estimates of age for appropriate relatives (+20 years for fathers, +20 years for mothers, −20 years for children, and similar age for siblings, offset by reported birth order). For those relatives who had died with a diagnosis of cancer, the date of death was imputed 1 year after the date of diagnosis (if known), and only if the actual date of death was unknown. For the age at diagnosis of each malignancy (when it was not reported), we imputed the mean age at diagnosis, gender-specific, for each malignancy using means from the Mayo Clinic Tumor Registry from January 1, 1993 to December 31, 2002. When the date of death was available but the age at the onset of cancer was not, the date of death minus 1 year was imputed as the age at diagnosis.
A substudy was performed to assess the validity of the proband report on cancer incidence among first-degree relatives. Among those relatives for whom proband questionnaire information was available, 90 either filled out questionnaires themselves (n = 72) or had medical records/death certificates available (n = 18) for review. Relative self-report or medical records were used as a “gold standard” to assess the validity of the proband report. Nonmelanoma skin cancers were excluded from the analysis.
A total of 3355 first-degree relatives from 426 pedigrees were analyzed, which included 1372 siblings, 851 parents, and 1132 children. Approximately 52.1% of the relatives were male and 1106 (33.0%) were self-reported smokers. The rates were then calculated for the most common tumor types reported in our questionnaires: PC, bladder carcinoma, colorectal carcinoma, head/neck carcinoma, lung carcinoma, melanoma, prostate carcinoma, gastric carcinoma, ovarian carcinoma, leukemia, lymphoma, brain carcinoma, liver carcinoma, and breast carcinoma. Carcinoma of an unknown primary tumor (n = 36 patients) was excluded from the analysis because of uncertainty regarding the primary site of disease among reporting relatives. Nonmelanoma skin cancers also were excluded from the analysis because of inconsistent reporting. Results are shown in Table 1. The only malignancies found to be overrepresented in the relatives of patients with PC were PC (SIR of 1.88; 95% confidence interval [95% CI], 1.27–2.68) and liver carcinoma (SIR of 2.70; 95% CI, 1.51–4.46). Bladder carcinoma, lymphoma, breast carcinoma, lung carcinoma, and prostate carcinoma were found to be underrepresented.
|Cancer type||No. of relatives with incident cases (No. with imputed age)||% smokersb||Males||Females||Overall SIR; (95% CI)|
|No.||SIR (95% CI)||No.||SIR (95% CI)|
|Pancreas||30 (1)||46.4||18||2.10 (1.24–3.32)||12||1.62 (0.84–2.83)||1.88 (1.27–2.68)c|
|Bladder||16 (7)||76.9||12||0.54 (0.28–0.94)||4||0.57 (0.16–1.47)||0.55 (0.31–0.89)|
|Colorectal||67 (30)||41.4||41||0.96 (0.68–1.30)||26||0.69 (0.45–1.01)||0.83 (0.64–1.05)|
|Head/neck||14 (8)||75.0||12||0.86 (0.44–1.50)||2||0.34 (0.04–1.24)||0.71 (0.39–1.19)|
|Lung||59 (31)||82.5||39||0.62 (0.44–0.85)||20||0.60 (0.37–0.93)||0.62 (0.47–0.80)|
|Melanoma||14 (4)||35.7||6||0.50 (0.18–1.09)||8||0.84 (0.36–1.65)||0.65 (0.36–1.10)|
|Gastric||16 (3)||46.2||11||1.21 (0.61–2.17)||5||1.03 (0.34–2.40)||1.15 (0.66–1.87)|
|Leukemia||10 (4)||62.5||7||0.70 (0.28–1.44)||3||0.45 (0.09–1.30)||0.60 (0.29–1.10)|
|Lymphoma||8 (3)||0||1||0.06 (0.00–0.34)||7||0.57 (0.23–1.18)||0.28 (0.12–0.55)|
|Liver||15 (8)||53.8||7||1.80 (0.72–3.69)||8||4.84 (2.09–9.52)||2.70 (1.51–4.46)|
|Brain||10 (5)||50.0||5||0.95 (0.31–2.22)||5||1.37 (0.44–3.19)||1.12 (0.54–2.07)|
|Prostate||59 (17)||52.9||59||0.71 (0.54–0.92)||0||–||0.71 (0.54–0.92)|
|Breast||70 (23)||28.4||1||1.44 (0.04–8.02)||69||0.72 (0.56–0.92)||0.73 (0.57–0.92)|
|Ovarian||12 (5)||22.2||0||–||12||0.90 (0.47–1.58)||0.90 (0.47–1.58)|
The risk of PC was found to be increased among nearly all subgroups, though some subanalyses did not achieve statistical significance. When only parents and siblings were analyzed (Table 2), lymphoma, bladder carcinoma, breast carcinoma, prostate carcinoma, and lung carcinoma were found to be underrepresented, and only PC and liver carcinoma were increased. Among relatives reported as smokers, the PC risk was high (SIR of 2.52; 95% CI, 1.34–4.32), as were the risks for lung and liver carcinoma. Nonsmoking relatives (n = 1868) also demonstrated a trend toward an increased risk of PC, but this risk was found to be lower than that for smokers (SIR of 1.77; 95% CI, 0.99–2.92), in concordance with smoking being a known risk factor for PC. Among nonsmokers, as expected, lung carcinoma was found to be underrepresented (SIR of 0.22; 95% CI, 0.10–0.40), as was bladder carcinoma (SIR of 0.22; 95% CI, 0.05–0.65).
|Primary tumor||Siblings only (n = 1372)||Parents and siblings (n = 2223)||Nonsmoking relatives (n = 1868)||Smoking relatives (n = 1106)||Proband age < 60 yrs (n = 830)||Proband age ≥ 60 yrs (n = 2525)||Fathers only (n = 425)||Mothers only (n = 426)|
|Pancreas||1.61 (0.81–2.89)||1.92 (1.29–2.74)a||1.77 (0.99–2.92)||2.52 (1.34–4.32)||2.86 (1.15–5.89)||1.70 (1.08–2.55)||2.90 (1.54–4.96)||1.38 (0.51–3.01)|
|Bladder||0.47 (0.17–1.03)||0.56 (0.32–0.91)||0.22 (0.05–0.65)||0.91 (0.44–1.68)||0.86 (0.24–2.22)||0.49 (0.25–0.85)||0.76 (0.35–1.45)||0.25 (0.01–1.38)|
|Colorectal||0.73 (0.47–1.08)||0.84 (0.65–1.06)||0.79 (0.55–1.10)||0.93 (0.59–1.38)||1.37 (0.80–2.19)||0.73 (0.54–0.96)||1.10 (0.71–1.62)||0.73 (0.42–1.18)|
|Head/neck||0.64 (0.24–1.39)||0.75 (0.41–1.25)||0.32 (0.07–0.94)||1.22 (0.56–2.32)||0.29 (0.01–1.63)||0.80 (0.42–1.36)||1.08 (0.44–2.23)||0.34 (0.01–1.89)|
|Lung||0.73 (0.50–1.03)||0.63 (0.48–0.81)||0.22 (0.10–0.40)||1.39 (1.02–1.85)||0.58 (0.26–1.10)||0.62 (0.46–0.82)||0.56 (0.33–0.89)||0.50 (0.23–0.96)|
|Melanoma||0.60 (0.22–1.31)||0.58 (0.29–1.04)||0.82 (0.37–1.55)||0.68 (0.22–1.58)||1.73 (0.70–3.57)||0.40 (0.16–0.83)||0.20 (0.00–1.09)||1.00 (0.27–2.57)|
|Gastric||1.18 (0.47–2.43)||1.18 (0.67–1.92)||1.01 (0.41–2.08)||1.23 (0.45–2.68)||0.46 (0.01–2.55)||1.28 (0.72–2.10)||1.46 (0.59–3.01)||0.71 (0.08–2.54)|
|Leukemia||0.84 (0.31–1.82)||0.63 (0.30–1.16)||0.35 (0.07–1.02)||0.88 (0.29–2.06)||0.73 (0.09–2.64)||0.57 (0.25–1.13)||0.60 (0.12–1.74)||0.27 (0.01–1.51)|
|Lymphoma||0.16 (0.02–0.56)||0.30 (0.13–0.60)||0.54 (0.23–1.06)||0.00 (0.00–0.38)||0.39 (0.05–1.42)||0.25 (0.09–0.55)||0.14 (0.00–0.77)||0.81 (0.26–1.89)|
|Liver||2.41 (0.88–5.24)||2.42 (1.29–4.15)||2.27 (0.83–4.93)||3.44 (1.38–7.07)||3.32 (0.69–9.68)||2.58 (1.33–4.50)||1.03 (0.12–3.70)||5.37 (1.74–12.5)|
|Brain||1.21 (0.39–2.81)||1.11 (0.51–2.12)||1.11 (0.36–2.58)||1.29 (0.35–3.30)||1.85 (0.38–5.40)||0.96 (0.39–1.98)||0.90 (0.11–3.24)||1.17 (0.14–4.22)|
|Prostate||0.82 (0.55–1.17, n = 728)||0.71 (0.54–0.92,n = 1153)||0.75 (0.48–1.12, n = 817)||0.75 (0.50–1.10, n = 706)||1.10 (0.60–1.85, n = 431)||0.64 (0.47–0.85,n = 1317)||0.63 (0.42–0.90)||–|
|Breast (female only)||0.75 (0.51–1.06, n = 644)||0.66 (0.50–0.85,n = 1070)||0.75 (0.56–1.00, n = 1051)||0.87 (0.52–1.37, n = 400)||0.86 (0.47–1.44, n = 399)||0.70 (0.52–0.91,n = 1208)||–||0.58 (0.38–0.84)|
|Ovarian||1.01 (0.37–2.20, n = 644)||0.89 (0.44–1.59, n = 1070)||0.79 (0.32–1.62, n = 1051)||0.69 (0.08–2.48, n = 400)||2.20 (0.72–5.12, n = 399)||0.64 (0.26–1.31, n = 1208)||–||0.78 (0.25–1.81)|
If the proband was diagnosed with PC before the age of 60 years (n = 830 relatives), the risk for PC was found to be further increased among relatives, to nearly threefold (SIR of 2.86; 95% CI, 1.15–5.89). No other cancer rate differences were found to reach statistical significance, although the risks of colorectal carcinoma (SIR of 1.37; 95% CI, 0.80–2.19) and melanoma (SIR of 1.73; 95% CI, 0.70–3.57) appeared to trend upward in this subgroup. One possible explanation for these results would be the predisposition of patients with familial atypical mole and melanoma syndrome (FAMMM) and hereditary nonpolyposis colorectal carcinoma (HNPCC) to develop PC at a younger age, as has been suggested in some reports.14, 15
Given the large number of comparisons, a more conservative statistical methodology was used to assess the strength of our significant findings. We also analyzed the data using a significance of P ≤ 0.01. The risk for PC (SIR of 1.88; 95% CI, 1.11–2.96) and liver carcinoma (SIR of 2.70; 95% CI, 1.24–5.07) remained significant. The subgroup of probands age < 60 years was found to lose statistical significance (SIR of 2.86; 95% CI, 0.83–7.01), whereas the other PC subgroups remained significant. Using an even more conservative Bonferroni-corrected alpha16 (P = 0.0003968), only four associations remained among the comparisons in Tables 1 and 2. These were all decreases in the risk for lymphoma-overall (SIR of 0.28; 95% CI, 0.05–0.82), lung carcinoma-overall (SIR of 0.62; 95% CI, 0.37–0.96), lung carcinoma-parents/siblings (SIR of 0.63; 95% CI, 0.38–0.98), and lymphoma-parents/siblings (SIR of 0.30; 95% CI, 0.06–0.90).
In the substudy validation of proband reporting of cancer, 19 malignancies were reported by the proband, whereas 22 cancers were identified through self-report or records, yielding a sensitivity of 86.4% for the diagnosis of cancer. Three relatives reported carcinoma diagnoses (of the breast, colon, and prostate, respectively) that were not reported in the probands' questionnaires. The primary tumor site was misreported for two relatives, whereas one relative had no primary tumor site listed by the proband. Specificity was 100% because no relatives were falsely reported as having cancer by the proband. Although not a random sample of all patients (because these relatives' records were procured during the workup for familial PC), they do provide insight with regard to the magnitude of the reliability of family history questionnaires targeting cancer. The sensitivity and specificity rates in the current study were comparable to those from prior reports for cancer family history.11, 12
The 30 cases of PC diagnosed among first-degree relatives was distributed among 29 families; therefore, the frequency of a reported PC family history was 29 of 426 patients (6.8%) in the current study, which is consistent with prior reports.17
To our knowledge, two cohort studies to date have reported the risk of PC to relatives. Hemminki et al.1 examined the Swedish Family-Cancer Database (which contains information regarding 10.2 million individuals), with 21,000 incident cases of PC reported over 37 years. In this database, the records of parents and offspring are linked so that familial studies can be performed. SIRs were generated for offspring risk for PC, sorted by parental malignancies. Only parental lung carcinoma (SIR of 1.52; 95% CI, 1.17–1.95) and PC (SIR of 1.68; 95% CI, 1.16–2.35) were found to confer an increased risk for PC in offspring. The increased risk for PC was found to be magnified if the parent was diagnosed with lung carcinoma before age 60 years (SIR of 3.14; 95% CI, 1.86–4.97). This could potentially be attributable to either cigarette smoking or secondhand smoke because tobacco exposure is a known risk factor for both PC18 and lung carcinoma.19 The risk for nonpancreatic malignancies in the siblings and offspring of patients with PC was not assessed.
Coughlin et al.20 reported data from the American Cancer Society Cancer Prevention Study II in 2000. This study examined data from 483,109 men and 619,199 women followed prospectively since 1982, using questionnaires completed at that time. Both men (relative risk [RR] of 1.5; 95% CI, 1.1–2.1) and women (RR of 1.7; 95% CI, 1.3–2.3) who reported a family history of PC at the time of study entry were found to have a higher risk of dying from PC, even after adjusting for age, race, smoking, and other lifestyle factors. To our knowledge, the risk for other malignancies was not reported.
Case–control studies also have examined familial risk in PC. Ghadirian et al.21 examined the family histories of 174 patients with incident PC who were seen at 1 hospital, comparing cancer rates among their first-degree relatives with those of 136 healthy controls (spouses of the probands). Probands were interviewed in person (36%) or by telephone (64%), with 174 of 208 potentially eligible patients (84%) participating. The rate of PC among the first-degree relatives of cases was found to be 5 times higher than among the first-degree relatives of controls (RR of 5.0; P = 0.01). No other malignancies demonstrated statistically significant differences between groups, although trends toward significance (RR > 2.0 in cases compared with controls) were observed among patients with melanoma, bladder carcinoma, cervical carcinoma, and esophageal carcinoma. The current study was not powered to detect subtle differences in cancer risk.
Silverman et al.22 conducted a population-based case–control study comparing risk factors in 484 PC patients versus 2099 controls. PC was reported more frequently among the first-degree relatives of cases than among those of controls (odds ratio [OR] of 3.2; 95% CI, 1.8–5.6). Ovarian carcinoma was found to be increased as well (OR of 5.3; 95% CI, 1.4–20.2), although the numbers of tumors were small. No increased rates of esophageal, gastric, colon, liver, lung, breast, endometrial, or prostate carcinoma were reported. Probands were interviewed in person, although only 484 of 1153 patients initially identified for the study could be interviewed. This was due for the most part to subject mortality; 471 patients of the potential subjects (41%) died before the interview could be conducted.
Schenk et al.23 reported a case–control study assessing the risk for PC among 1816 first-degree relatives of 247 patients with PC versus 420 controls. The relatives of PC patients had a higher risk for PC (RR of 2.49; 95% CI, 1.32–4.69). Smoking and age of the proband (age < 60 years) was found to further increase the risk (RR of 8.23; 95% CI, 2.18–31.07). The risk for other malignancies was not reported. All probands and controls were interviewed in person for family history information; however, only 247 of 358 eligible patients (69%) were interviewed. Due to the rapidly fatal nature of the disease, coupled with a median time between diagnosis and interview of 48 days among the probands, many patients died before an interview could be performed, even with a rapid-ascertainment case finding system. No proxy interviews were performed.
Given the disparate results regarding the risk of ovarian carcinoma in the relatives of PC patients in the study by Silverman et al.22 compared with the others noted earlier, in addition to the poorly understood nature of genetic susceptibility to PC, we elected to perform the current study. These results support the concept of inherited susceptibility for PC, but provide no clear evidence of a major association with other malignancies in unselected patients. The risk of PC was found to be increased nearly twofold in the first-degree relatives of patients with PC, and nearly threefold in the relatives of those patients diagnosed before age 60 years. The increased risk of PC in first-degree relatives supports the concept of the autosomal dominant transmission of a susceptibility to PC. Certainly, familial syndromes exist that predispose to other malignancies in addition to PC, including BRCA1/2, FAMMM,6 and HNPCC.24 Germline mutations in BRCA2, p16/CDKN2A, and DNA mismatch repair genes are known to explain these subsets of “familial” PC, but do not explain all kindreds with familial pancreatic cancer. We suggest that the lack of overrepresentation of colorectal carcinoma, melanoma, and breast carcinoma, in light of the elevated risk of PC noted herein, provides further evidence for as-yet undiscovered PC susceptibility gene or genes, unrelated to currently known cancer family syndromes.
We found that the reported risk of liver carcinoma is increased (SIR of 2.70; 95% CI, 1.51–4.46). However, because the liver is a common site of metastatic disease, and a relatively rare primary tumor site in the U.S., it is likely that some patients reported the site of metastasis rather than the primary tumor site. In the absence of pathologic or medical record verification, this finding cannot be considered conclusive. This will need to be confirmed with further study with verification of pathologic and medical records.
The underrepresentation of other malignancies identified in the current analysis likely may be a reflection of the expected 20–40% rate of underreporting11, 12 of malignancies among first-degree family members. The relatively large number of relatives reported to have an unknown primary site of malignancy suggests that perhaps a diagnosis of cancer is reported adequately, but the index patient may not be aware of the site of origin, if it can be determined clinically.
Potential limitations of this analysis are the possibility of the selective recruitment of patients with a family history of PC, a heightened proband awareness of family members with PC compared with other malignancies, and the focused workup of families with multiple cases of PC reported (for our PACGENE study). However, the recruitment of a large percentage of all patients with PC seen at the study institution (> 75%), coupled with the relatively low incidence of familial PC (7%), are important considerations. Patient awareness of family members' diagnoses cannot be controlled, but in limiting our analysis to first-degree relatives, we have likely minimized reporting and recall bias. The impact of the focused workup of families is minimized because of the necessity for the index patient to self-report at least one family member diagnosed with PC.
The results of the current study have shown that first-degree family members of patients with PC are at increased risk of developing the disease themselves. Whether this can be translated to PC screening programs will require more research. Aggressive screening regimens with endoscopic ultrasound, computed tomography, and blood CA 19-9 testing for patients at risk have to our knowledge not yet been validated, although studies are reportedly ongoing in high-risk kindreds.25–27
The identification of risk factors for PC is a vital focus for research. A known risk factor is family history, and the current study findings support this concept. The data from this study do not demonstrate any increased risk for malignancies for first-degree relatives of unselected patients with PC, although liver carcinoma may be overrepresented. PC patients with a strong family history of cancer suggesting hereditary breast carcinoma, ovarian carcinoma, colorectal carcinoma, or melanoma syndromes should be managed accordingly, however, and are best referred to a genetic counselor or medical genetics consultation.
- 102003: 14., , , et al. Pancreatic cancer genetic epidemiology (PACGENE) consortium [abstract]. Proceedings of the Lustgarten Foundation for Pancreatic Cancer Research, Boston, MA,
- 13National Cancer Institute D, Surveillance Research Program, Cancer Statistics Branch. Surveillance, Epidemiology, and End Results (SEER) Program SEER*Stat Database: Incidence - SEER 9 Regs Public-Use. Available from URL: www.seer.cancer.gov [accessed January 2004].
- 14Phenotypic variation in eight extended CDKN2A germline mutation familial atypical multiple mole melanoma-pancreatic carcinoma-prone families: the familial atypical mole melanoma-pancreatic carcinoma syndrome. Cancer. 2002; 94: 84–96., , , et al.