The aim of the current study was to determine the association between ABO blood group and the risk and progression of pancreatic ductal adenocarcinoma (PDAC) in the Han Chinese ethnic group. During the period of 2000–2009, 1,431 patients with PDAC and 1,449 age- and sex-matched controls were recruited in two university-affiliated hospitals. An unconditional multivariable logistic regression analysis was used to estimate adjusted odds ratios (ORs). The relationship between patient ABO blood group and clinicopathologic features was also analyzed. Compared with subjects having blood group O, a modestly higher risk was observed among cases with blood group A or AB with adjusted ORs (95% confidence interval) of 1.368 (1.127–1.661) and 1.391 (1.053–1.838), respectively. The TNM stages of tumors in patients with non-O blood groups (A, B or AB) were more highly advanced than in patients with blood group O (p < 0.001). Among patients who underwent a potentially curative operation, the median survival time of patients with blood group O was significantly longer than that of patients with non-O blood groups (16.0 months vs. 11.0 months, p = 0.001, log-rank test). This study shows evidence of an association between blood group type and risk for development and progression of PDAC. These findings merit further confirmation in a large population-based prospective study in patients of the Han Chinese ethnic group.
Every year, about 30,000 new cases of pancreatic ductal adenocarcinoma (PDAC) are diagnosed in the United States. PDAC has the poorest prognosis of all gastrointestinal malignancies with a 5-year survival of less than 5%.1 Curative resection offers the only chance of cure, but only 9% of patients undergo this procedure.2 Typically, by the time of initial diagnosis, metastatic disease is present.3 Understanding the etiology and identifying high-risk individuals are essential for early diagnosis and successful treatment of this usually fatal disease.4 However, little is known about the risk factors for this disease.
Pancreatic cancer has been reported to be associated with several germ line mutations, including P16,5STK11,6CFTR7 and BRCA2.8 Furthermore, epidemiologic studies have suggested some environmental factors contribute to development of pancreatic cancer, including cigarette smoking,9 a medical history of type 2 diabetes and chronic pancreatitis. However, these genetic changes and environmental factors account only for a minority of pancreatic cancer cases. So, there is an urgent need to identify high- or low-risk population subsets.
Human blood group antigens are glycoproteins expressed on red blood cells and a variety of epithelial cells, including urothelium and gastrointestinal mucosa. The phenotypic A and B antigens are terminal carbohydrates synthesized by the addition of single sugars catalyzed by a series of specific glycosyltransferases. Phenotype O is characterized by the absence of A and B glycosyltransferases; only a protein backbone, the H antigen, is present. The glycosyltransferase is encoded by the ABO gene, which spans about 19 kb of genomic DNA on chromosome 9, band q34.10, 11 Since the initial report by Aird et al. indicating a correlation of gastric cancer with blood type A,12 the relationship between ABO blood group antigens and carcinogenesis or progression of human tumors, including PDAC, has been suggested by many investigations.12–18 For example, a variety of cancers were reported to be positively associated with blood group A, such as carcinomas of stomach,12 breast,13, 14 pancreas,15 and ovary.16 A positive association with blood group O was reported for several other tumor types: squamous cell carcinoma of the esophagus,17 skin melanoma18 and acute lymphoblastic leukemia.19 Specifically, two recent independent prospective cohort studies conducted by Wolpin et al.20 suggested that ABO blood type was associated in a statistically significant manner with the risk of developing pancreatic cancer. Compared with participants with blood group O, those with non-O blood groups (A, AB or B) were more likely to develop pancreatic cancer. Amundadottir et al.21 identified the contribution of genetic variation in the ABO locus of 9q34 to pancreatic carcinogenesis by conducting a two-stage genomewide association study. Furthermore, subjects with non-OO genotype(AO, AA, AB, BO or BB) were found to have ORs of 1.33–2.42 for developing pancreatic cancer compared with OO genotype.22 In addition, ABO blood groups may affect not only the development but also the subsequent risk of death of certain cancers.23–27 Examining the effect of ABO blood group on incidence, clinical characteristics and outcome within the same population is, therefore, important.
To understand more fully the relation between ABO blood type and the risk of developing pancreatic cancer, we completed a large case–control study of pancreatic cancer in the Han Chinese ethnic group and ascertained information on a variety of potential exposures related to pancreatic cancer risk, including a history of diabetes, alcohol consumption, smoking habits and a family history of pancreatic cancer. We evaluated whether ABO blood type was a risk factor for pancreatic cancer and whether this relationship was modified by other possible risk factors. In addition, correlations between ABO blood groups and clinicopathologic features and outcome in patients with pancreatic cancer were investigated.
Material and Methods
Patients eligible for the study were enrolled between October 1, 2000 and August 31, 2008 at Ruijin Hospital, Shanghai, People's Republic of China or between January 1, 2000 and August 31, 2009 at Changhai Hospital, Shanghai, People's Republic of China. There were 2,018 potential pancreatic cancer patients during the study period and 1,798 patients with histologically confirmed primary pancreatic adenocarcinoma. Of these, 367 cases were not included because 298 lacked ABO blood group data, 37 had a history of cancer and 32 missed recruitment (Fig. 1). The remaining 1,431 patients (589 from Ruijin Hospital and 842 from Changhai Hospital) were enrolled in this study. Eligible patients were Han Chinese and were able to donate a blood sample at the time of diagnosis. The stage of pancreatic cancer for all patients was determined based on the 2002 UICC on cancer staging system.28 Those with other types of pancreatic disease, such as neuroendocrine tumors, cysts, or patients unable or unwilling to give informed consent were excluded.
Subjects considered for inclusion as controls consisted of patients admitted to the same hospitals for any acute conditions based on discharge diagnoses. We sampled controls stratified by age (in 5-year age groups) and sex to match the distribution of the case series. There were a total of 1,449 controls. Of these, 35% were admitted to the hospital for trauma, 28% for acute surgical conditions, 24% for eye diseases and 13% for a variety of other acute illnesses. Patients with any malignant diseases were excluded. In addition, conditions related to alcohol and tobacco consumption (e.g., respiratory diseases, peptic ulcer and hepatic disease) or any chronic diseases that might have resulted from substantial lifestyle modifications (e.g., cardio- or cerebrovascular diseases) were also excluded based on discharge diagnoses.
Clinicopathological data or potential risk factors for pancreatic cancer were acquired from patients with approval from the ethics of the chamber of physicians of Ruijin Hospital and Changhai Hospital. Cases and controls were personally interviewed for demographic characteristics that contained information on sociodemographic indicators (including marriage status and birth place), personal habits (including smoking status and alcohol drinking), a history of diabetes mellitus and a family history of pancreatic cancer. Birth place was categorized into eastern China (including the city of Shanghai and provinces of Jiangsu, Zhejiang, Fujian, Jiangxi, Anhui and Shandong) or non-eastern China. Smoking status was categorized into two groups based on whether or not patients had smoked more than 100 cigarettes during their lifetime. Alcohol consumption was estimated based on answers to questions regarding drinking habits. These answers were integrated into a scoring system which was used to classify alcohol consumption as “low or moderate” (0–30 g/day) or “high” consumption (≥30 g/day).29, 30 No attempts to more accurately quantify smoking status or alcohol consumption were made. Diabetes was defined as present if there was a fasting serum glucose level greater than 7.0 mmol/L or a previous diagnosis of diabetes mellitus diagnosed based on the American Diabetes Association criteria.31 ABO blood type and Rh factor were obtained using routine clinical tests.
Patient follow-up was obtained through review of hospital records, contact with family members of the patients or review of the Cancer Registry of Shanghai. Ruijin Hospital patients were observed until December 31, 2008 and Changhai Hospital patients were observed until September 30, 2009. Overall survival (OS) was defined as the interval between the dates of diagnosis to the time of last follow-up or death due to pancreatic cancer. Censoring occurred if patients were still alive or were dead from other reasons at last follow-up.
All statistical analyses were conducted by using the SPSS 13.0 statistical software program (SPSS, Chicago, IL). Crude and adjusted odds ratios (OR) and 95% confidence intervals (CI) for each variable were calculated by using unconditional logistic regression. Potential confounders were included in the multivariate analysis performed in a stepwise manner at a significance level of p < 0.15. Equations included terms for a family history of pancreatic cancer in first-degree relatives, a history of diabetes, blood types, smoking status and alcohol drinking. Pearson's χ2 test, Fisher's exact probability and Student's t-test were used to compare the sociodemographic and clinicopathologic data. An analysis of survival was performed using the Kaplan–Meier method and differences between survival curves were performed by using the log-rank test. All tests were two-tailed and a p value < 0.05 was considered to be significant.
The distribution of demographic features of cases and controls are shown in Table 1. The median age was 61 (range 28–89) years for pancreatic cancer patients and 59 (range 25–85) years for controls. Most of the patients and controls were males (63.2% and 62.1%, respectively), were born in eastern China (94.9% and 95.3%, respectively), were married (93.2% and 94.0%, respectively) and were Rh factor positive (99.8% and 99.7%, respectively). No statistically significant differences were found between patients and controls in terms of these variables, suggesting that the frequency matching was adequate. The distribution of birth places of missing cases were comparable to studied cases (p = 0.541, data not shown). In addition, the frequency distribution of ABO blood groups of controls in the current study was similar to those for Han people in Shanghai, China as recorded in population registries32 (p = 0.468, data not shown).
Table 1. Demographics of the study population
Association between ABO blood type and pancreatic cancer risk
Unconditional logistic regression analysis was used to estimate associations between ABO blood type and risk of pancreatic cancer and other factors and pancreatic cancer risk (Table 2). The unadjusted ORs (95% CI) of risk of pancreatic cancer associated with the smoking status, alcohol consumption, family history of pancreatic cancer and diabetes were 1.492 (1.264–1.761), 1.184 (0.977–1.436), 1.474 (1.078–2.015) and 5.574 (4.432–7.010), respectively. Compared with subjects having blood group O, the unadjusted ORs (95% CI) of risk of pancreatic cancer associated with blood group A, B and AB were 1.391 (1.156–1.672), 1.030 (0.849–1.251) and 1.353 (1.038–1.765), respectively. Because demographic factors for cases and controls were comparable, we chose to present the ORs for blood groups after adjusting for smoking status, diabetes, alcohol consumption and family history of pancreatic cancer. After adjustment for the above variables, the significance was not changed.
Table 2. Pancreatic cancer risk factors and association with ABO blood type: Univariate and multivariate logistic regression analyses
The presence of at least one A or B blood group allele was reported to be associated with the risk of pancreatic cancer among Caucasian men and women in the United States.20–22 We further analyzed the risk of the disease for correlation with the presence of at least one A or B blood group allele in Han people. Compared with subjects having blood group O, those having any blood group A (blood groups A or AB) or having non-O blood groups (blood groups A, B or AB) had adjusted ORs of 1.365 (1.137–1.638) or 1.229 (1.040–1.453). By contrast, we found no association between any blood group B (blood groups B and AB) and pancreatic cancer risk (OR of 1.111, 0.929–1.328).
Clinicopathologic variables and outcome in patients with pancreatic cancer
Clinicopathologic variables for patients grouped by blood type are shown in Table 3. No significant difference was found regarding location of tumors among cases with various blood groups. However, the TNM stages of tumors among patients with various blood groups were found to be significantly different (p = 0.003). When the TNM stages of tumors in patients with blood group O and patients with non-O blood groups (A, B and AB) were evaluated, we found that the TNM stages of tumors in patients with non-O blood types were significantly more advanced (p < 0.001, data not shown).
Table 3. Clinical characteristics of pancreatic cancer cases sorted by ABO blood type
During the study period, survival data were available for 755 of the 1,431 patients. We were unable to obtain survival data on 676 cases due to incorrect contact information. The clinicopathologic variables of those patients whose OS data were unavailable were comparable to those whose survival data were available (Supporting Information Table S1). At the termination of the study, 130 patients were still alive and the OS rate was 43% at 1 year, 6% at 3 years and 2% at 5 years. The median survival time in subjects having the blood groups A, B, AB and O were 9.0, 9.0, 9.1 and 11.1 months, respectively; the differences were not significant as indicated by the log-rank test (p = 0.869, Fig. 2a). In addition, there was no significant difference in the survival time of patients with non-O blood groups (A, B and AB) and the O blood group (p = 0.657, Fig. 2b). However, among patients who underwent a potentially curative operation, of which there were 316, the median survival time in patients with the O blood group was significantly longer than that of patient with non-O blood groups (16.0 months vs. 11.0 months, p = 0.001, log-rank test) as shown in Figure 2c.
In this large case–control study of Han Chinese patients, patients with blood group O had lower incidence of pancreatic cancer compared with those with blood groups A and AB. The O blood type was also associated with the less advanced pancreatic cancer. To our knowledge, this is the first study showing a correlation between blood group O, and the development of pancreatic cancer in Han people.
Studies conducted several decades ago suggested a role for ABO blood group antigens in the development of cancers, including pancreatic cancer.15, 33, 34 For example, several retrospective case–control studies performed in the 1960s and 1970s indicated an association between blood type A and an increased risk of pancreatic cancer, compared with participants with blood group O or B.33–35 However, an increased risk of blood group B among pancreatic cancer cases was observed compared with a randomly selected group of patients admitted with nonmalignant diseases and a randomly selected group of blood donors in a more recent study.36 The lack of a consistent association between blood group and pancreatic cancer risk across these studies might stem from small sample sizes, retrospective data collection or the use of poorly matched control populations.20 Results from two large, independent prospective cohorts of Caucasians from the United States suggested that, compared with subjects with blood group O, those with non-O blood types (A, AB or B) were more likely to develop pancreatic cancer.20 In the current study, we retrospectively analyzed a large case–control group of patients and, similarly, found an association between blood type A or AB and an increased risk of pancreatic cancer compared with those with blood groups O [adjusted ORs 1.368 (1.127–1.661) and 1.391 (1.053–1.838), respectively]. However, compared with blood group O, no association between blood group B and pancreatic cancer risk was found in Han people. Our results are in contrast with a study conducted on subjects of European ancestry with blood group B20; in Europeans, adjusted ORs for patients with blood type B were 1.72 (1.25–2.38) compared with those with blood group O. The difference might be due to varying genetic backgrounds (populations with blood group B account for about 13% in European ancestry and about 28% in Han Chinese, respectively).
Our finding of an association between blood type and pancreatic cancer risk should not be surprising for the following reasons: First, several studies have indicated that blood type is correlated with the host immunological and inflammatory state, which may be related to carcinogenesis.37–41 Single nucleotide polymorphisms (SNPs) at the ABO locus were recently found to be associated with the modification of circulating levels of two inflammatory cytokines, tumor necrosis factor-alpha (TNFα)40 and soluble intracellular adhesion molecule-1 (sICAM-1).41 TNFα, which is secreted by cancerous cells in increased amounts compared with normal cells, is associated with the ability to induce apoptosis in cancer cells.42, 43 Second, blood group antigens might alter the host response to stimuli of apoptosis. There appears to be a link between inhibition of apoptosis and carcinogenesis.44, 45
A relationship between the expression of blood group antigens and tumor behavior and subsequent clinical outcome was suggested previously in patients with various malignances. It was hypothesized that, as malignant cells became more poorly differentiated, they lose their ability to synthesize normal ABH antigens. Therefore, the loss of ABH expression in tumor tissues might correlate with an adverse prognosis in cancers of the lung, urinary bladder, uterine cervix and head and neck.46–48 Although evaluation of blood groups as prognostic markers was performed in patients with breast cancer, malignant melanoma, oral squamous cell carcinoma and esophageal squamous cell carcinoma,17, 18, 23, 24 findings were not consistent. In our study, there was a significant correlation between blood group O and the less advanced TNM stages of tumors. The other important finding was that, for pancreatic cancer patients who had undergone curative operations, those with blood group O lived longer than patients with non-O blood groups. These data suggest that non-O blood group may be associated with poorer survival. Examination of the question in a complete series of all cases in a defined population, with full information on presentation and treatment, will be helpful to confirm the reported association.
The ABO gene, on chromosome 9q34, determines blood type by encoding three glycosyltransferases.49–52 Blood group antigens are present on key receptors controlling cell proliferation, adhesion and motility, such as receptors for epidermal growth factor, integrins, cadherins and CD44. The expression patterns of these various receptors differ according to the type of cancer and, therefore, the role of ABH antigens in the biology of human cancers may also vary.53 A higher incidence of various types of carcinomas has been reported in individuals with A/B blood groups. The reason for this propensity, however, is not clear. It has also been postulated that A-positive and A-negative individuals may have different susceptibilities to carcinoma development. A study by Ichikawa et al. suggested that loss of expression of AB glycosyltransferase in human colon carcinoma cell lines enhances malignancy.54 A later study by the research group showed a marked enhancement of motility in the A-negative population compared with the A-positive population among tumor cells.55 These results raised the possibility that blood group antigens might alter the systemic inflammatory response and, thus, suggested a possible mechanism to explain the association between blood type and the development of cancer.
In any case–control study, limitations and strengths should be recognized. Because pancreatic cancer is a rapidly fatal disease, the number of cases for which survival data could not be obtained is relatively large. However, the distribution of birth places, one of the most relevant factors related to ABO blood group status, was comparable between missing cases and studied cases. Similarly, survival data for many patients in the current study were also unavailable. However, the clinicopathological and treatment variables, which were reported to be the most relevant to the outcome of this disease, between patients whose survival data were available and those whose survival data were not available were comparable. Second, the possibility of selection bias due to the use of in-patients as controls cannot be ruled out. Nevertheless, because the distribution of blood groups in our controls was not different from that in the general population and because both cases and controls belonged to a relatively homogeneous base population, they were matched by sex, age, healthcare systems and sociodemographic variables. Bias should be minimized by the matching. The blood groups in this study were abstracted from medical charts based on routine laboratory tests. More than 80% of participants had confirmatory ABO blood group tests. Hence, measurement error due to inaccurate blood group self-reporting was avoided. Finally, pathology slides and reports were reviewed by the study pathologist to ensure that diagnostic inclusion criteria were met and that only patients with histologically confirmed adenocarcinoma of the exocrine pancreas were included. This increased the reliability of our findings.
In conclusion, these data support the recently reported association between non-O blood types and risk of developing pancreatic cancer. They also suggest that the association is similar in populations of Asian and European ancestry. These findings merit further confirmation in a larger population-based prospective study in the Han Chinese ethnic group.
We thank Ms. Fei Yuan (Department of Pathology, Ruijin Hospital, Shanghai, China) for her skillful technical assistance.