Implications and issues related to familial pancreatic cancer: a cohort study of hospitalized patients

This study states the reasonable hypothesis that a significant proportion of patients presenting with pancreatic cancer have a familial incidence such which this retrospective study identified.

It has been reported that in a study involving 570 families and including 9204 relatives whose probands were 3- to 5-fold more often heavy smokers than the general population, and 9.3 % of them reported a positive family history of pancreatic cancer; in addition, in first-degree relatives, only mortality from pancreatic cancer was significantly increased and the lifetime risk of dying from pancreatic cancer was 4.1 % for the relatives of all probands, and was 7.2 % for the relatives of probands who developed disease prior to 60 years of age [12]. These data suggest that genetic susceptibility to pancreatic cancer may be attributable to moderate- to low-penetrance genes. Several guidelines have also been developed to manage these high-risk individuals [13, 14] and registries have been established [15, 16] in order to better investigate this topic. However, there are no data regarding the incidence of familial pancreatic cancer in a real world of hospitalized patients. Thus, our study retrospectively evaluated all patients suffering from pancreatic cancer treated in Sant’Orsola-Malpighi Hospital in a period of three years. Regarding the primary endpoint of the study, only 1.6 % of patients had at least one relative with pancreatic adenocarcinoma while 0.5 %, had a family history of familial adenocarcinoma. Pancreatic cancer has a high mortality rate but its low incidence, together with the inability to arrive at a diagnosis using low cost imaging methods, does not seem to justify the execution of screening the general population, and it is therefore important to try to intervene first on the risk factors, such as smoking, coffee, alcohol and diet [17]. Based on the observations made in this study with respect to cases of familial pancreatic cancer, the introduction of screening at least 10 years before the age of the youngest family member affected would be recommended with the use of imaging methods, such as MRI and/or CT so that early intervention can be carried out with resection surgery. It should also be noted that our percentage of patients who should be screened is similar to that of another Italian study [8] and it would be appropriate, given the low number of patients to be followed, to abandon the current lack of interest regarding pancreatic cancer, at least in the familial form [18]. Only one of our patients was diagnosed at an advanced age and had never undergone a radiological investigation for screening as recommended by the Italian guidelines for familial cancer [13]. In addition, 1.6 % of patients had a tumor associated with FAP; this issue merits better investigation in clinical practice because screening for colorectal carcinoma may add useful information in selecting patients with FAP, and these patients should also be investigated and followed for the development of pancreatic cancer [19]. It is possibly that the future development of diagnosing the predisposition of familial pancreatic cancer is to evaluate genomic modifications; in fact, it has been recently reported that prevalence of mutations among familial pancreatic cancer probands was 1.2 % for BRCA1, 3.7 % for BRCA2, 0.6 % for PALB2, and 2.5 % for CDKN2A [20]. In addition, the same authors reported that familial pancreatic cancer probands carry mutations in the four genes with a significant more high frequency (8.0 %) than nonfamilial pancreatic cancer probands (3.5 %) [20]. On the other hand, other authors have been also reported that eleven pathogenic mutations have identified (3 in ATM, 1 in BRCA1, 2 in BRCA2, 1 in MLH1, 2 in MSH2, 1 in MSH6, and 1 in TP53) and that the prevalence of mutations was 3.8 %; the carrier status was associated significantly with breast cancer in the proband or first-degree relative, and with colorectal cancer in the proband or first-degree relative, but not family history of pancreatic cancer, age at diagnosis, or stage at diagnosis [21]. Thus, we need of further studies on this topic but genetic testing of multiple relevant genes in probands with a positive family history of cancer is reasonable.

Other data from our study confirmed what is already known, i.e. that approximately half of the study population smoked [20], about 50 % habitually consumed coffee [20] and less than a quarter of the patients (24.1 %) drank alcohol [22]. It has been suggested that obesity is a risk factor for pancreatic cancer [23], but our data seemed to confirm that, in the majority of cases (50.0 %), Italian patients were of normal weight, 38.2 % were overweight and only 6.6 % were obese. Probably, the diet of the Italian population is different from the American diet so the risk related to obesity does not appear to be evident [24]. Our data were in line with those of the literature regarding the location of pancreatic cancer; in fact, this tumor is most frequently localized in the head of the pancreas (60 %) [25], and the size of the mass, greater than 3 cm, means that the diagnosis is unfortunately still too late [26]. An interesting datum was that IPMNs were found to be associated with a pancreatic mass in 8 % of the patients and was multifocal in 3 patients. This indicated that pancreatic adenocarcinoma probably arises from the degeneration of an unrecognized mucinous tumor [27]; hence, the importance of a proper diagnosis and follow-up of mucinous cystic lesions [28]. Another interesting fact is that of the 187 patients studied, approximately 6.0 % had a synchronous carcinoma and 13 % metachronous cancer; these patients require a different clinical approach which has not yet been defined. Even if our study suffers from the limitations of retrospective studies, this is the first study, however, which evaluated the incidence of familial pancreatic cancer in an unselected hospitalized population; the post-hoc analysis revealed that the population study was not unpowered. In addition, it should be pointed out that only 62 % of the patients studied (187/301) had a correct Classification ICD-9-CM; much effort should be made to improve the correct application of Classification ICD-9-CM as also suggested by the Italian Association of Medical Oncologists [29]. One problem of studies regarding familial pancreatic cancer is the method used to assess the family history of these subjects. PancPro is a Bayesian modeling framework used to assess the pancreatic cancer risk of patients with a family history of pancreatic cancer [30]. However, in this study, information was collected using a questionnaire and no natural language processing system techniques. Using PancPro, it has been reported that the sensitivity of finding a subject with a family history of pancreatic cancer is 93 %, the positive predictive value is 97 % when family history is extracted, but this method considers only family histories [31] and does not extract specific family members or classify family members as primary, secondary and unknown relatives [32]. An algorithm has recently been proposed based on the Unstructured Information Management Architecture (UIMA) framework consisting of section segmentation, relation discovery and negation detection [32]. This system was evaluated using data from two institutions, and the results showed that rule-based natural language processing approaches for specific information extraction tasks are portable across institutions; however, customization of the algorithm regarding the new dataset improves its performance and this may increase the possibility of detecting the familial origin of pancreatic cancer.