Influence of cancer susceptibility gene mutations and ABO blood group of pancreatic cancer probands on concomitant risk to first-degree relatives

Abstract

Background:

ABO blood group is associated with pancreatic cancer (PC) risk. Whether ABO blood group alone or when combined with inherited mutation status of index PC cases (probands) can enhance PC risk estimation in first-degree relatives (FDRs) is unclear. We examined FDRs’ risk for PC based on probands’ blood group and probands’ cancer susceptibility gene mutation status.

Methods:

Data on 23,739 FDRs, identified through 3,268 PC probands, were analyzed. Probands’ ABO blood groups were determined serologically or genetically, and 20 susceptibility genes were used to classify probands as “mutation-positive” or “mutation-negative.” SIRs and 95% CIs were calculated, comparing observed PC cases in the FDRs to the number expected in SEER-21 (reference population).

Results:

Overall, FDRs had two-fold risk of PC (SIR=2.00, 95%CI=1.79-2.22). PC risk was higher in FDRs of mutation-positive (SIR=3.80, 95%CI=2.81-5.02) than mutation-negative (SIR=1.79, 95%CI=1.57-2.04) probands (p<0.001). Risk magnitudes did not differ by ABO blood group alone (SIR_{blood-group-O}=1.57, 95%CI=1.20-2.03, SIR_non-O=1.83, 95%CI=1.53-2.17;p=0.33). Among FDRs of probands with non-O blood group, PC risk was higher in FDRs of mutation-positive (SIR=3.98, 95%CI=2.62-5.80) than mutation-negative (SIR=1.66, 95%CI=1.35-2.03) probands (p<.001), but risk magnitudes were statistically similar when probands had blood group O (SIR_{mutation-positive}=2.65, 95%CI=1.09-5.47, SIR_{mutation-negative}=1.48, 95%CI=1.06-5.47;p=0.16).

Conclusion:

There is a range of PC risk to FDRs according to probands’ germline mutation status and ABO blood group, ranging from 1.48 for FDRs of probands with blood group O and mutation-negative to 3.98 for FDRs of probands with non-O blood group and mutation-positive.

Introduction

Pancreatic cancer (PC), despite being the 10^th most common cancer among men and 8^th among women, is the 4^th leading cause of cancer-related death in both men and women (1). The major risk factors for PC include older age, male sex, diabetes mellitus, chronic pancreatitis, cigarette smoking, excessive alcohol intake, and obesity (2–4). Inherited (germline) genetic factors also play essential roles in PC development and can influence host physiological responses to modifiable risk factors, such as cigarette smoking and alcohol intake (2–5). Family history of PC and ABO blood group are also associated with PC risk, further demonstrating the importance of heritable factors in PC development (2,6–11).

Our group previously reported a higher risk of PC among first-degree relatives (FDRs) of index PC patients (probands), with substantially higher PC risk among FDRs of probands who carry one or more mutations in cancer susceptibility genes (7). ABO blood group is a genetically determined trait (12–14) and having a non-O ABO blood group has been consistently associated with higher risk for PC in both case-control and cohort studies (8–11,15,16). Given the consistent association between ABO blood group and PC risk, it is plausible that the familial aggregation of PC may in part be due to inheritance of ABO blood group alleles within families (12–14).

The precise mechanism by which ABO blood group predisposes to PC is unknown. The ABO gene encodes glycosyltransferase enzymes that catalyze the transfer of different sugar residues, N-acetyl-galactosamine or galactose, to a biosynthetic precursor (the H antigen) to form an A or B antigen, respectively, or AB antigen (13,17). Individuals who lack the glycosyltransferase enzyme activity have the O blood group and express the H antigen (12–14). These antigenic expressions on erythrocytes generally serve as the basis for ABO blood typing (13,17,18). The glycans that catalyze the biosynthesis of the ABO histo-blood groups play important physiological roles, including protein maturation and turnover, receptor binding and activation, and host immune responses (19). The ABO blood groups themselves have been implicated in modulation of systemic inflammation and membrane signaling (8,20–23), all of which could influence PC development (24,25). Specifically, ABO blood groups have been implicated in systemic inflammation measured by circulating levels of soluble intercellular adhesion molecule 1 (sICAM-1), tumor necrosis factor-alpha (TNF-α), C-reactive protein (CRP), interleukin 6 (IL6), or E-selectin (21–23). Studies have also shown that Helicobacter pylori (H pylori) colonization, a flagellated gram-negative bacillus infection, is associated with PC risk, but the association is modulated by ABO blood group, such that an association between CagA-negative H pylori seropositivity and PC risk is seen only in individuals who have a non-O blood group (26). Experimental studies have also shown that the expressions of ABO blood group antigens on the surfaces of tumor cells are associated with variation in in vitro cell motility, cellular resistance to apoptosis, and immune escape by tumor cells (27).

Few studies have inferred FDRs’ risk for PC based on probands’ germline mutation status under assumptions of Mendelian laws of inheritance (7,28–30). To our knowledge, no study has yet investigated the potential contribution of ABO blood groups to the familial aggregation of PC. Therefore, the aims of this study were to investigate FDRs’ risk of PC based on probands’ ABO blood group and investigate FDRs’ PC risk by combining probands’ ABO blood group and germline cancer susceptibility gene mutation status. We used standardized incidence ratios (SIRs) to estimate PC risk among FDRs of PC probands in comparison with data from the U.S. Surveillance, Epidemiology, and End Results (SEER) program (31). We assessed SIR by probands’ ABO blood group alone, probands’ cancer susceptibility gene mutation status alone, and a combination of probands’ blood group and cancer susceptibility gene mutation status to more robustly estimate familial risk of PC to inform strategies for cancer risk assessment and generate evidence to guide genetic counseling in PC families.

Materials and Methods

Study population

Following approval by the Mayo Clinic Institutional Review Board, data were obtained from the Mayo Clinic Biospecimen Resource for Pancreas Research, a patient registry supported by the Mayo Clinic Specialized Program of Research Excellence (SPORE) in pancreatic cancer (4,7,32,33). The registry utilizes an ultra-rapid case ascertainment process, which ensures that approximately 86% of PC cases diagnosed at Mayo Clinic are recruited within 30 days of diagnosis (4,33). The PC probands included in this study were recruited between October 2000 and March 2020, with participation rate of ~70% (4,7). The primary reasons for non-participation in the registry are the severe morbidity associated with PC and rapid death of patients following PC diagnosis. For this study, we analyzed data on 23,739 FDRs, identified through 3,268 consecutively enrolled probands with a primary diagnosis of ductal adenocarcinoma of the exocrine pancreas, determined pathologically (~95%) or radiologically. The sample for this study overlaps partially with samples used in a previous study which primarily included data from the years 2000-2016 (7). The current sample contains 963 (29%) newly added probands, 6,577 (28%) newly added FDRs, with an additional 271,406 (28%) person-years at risk.

Data collection

The probands completed structured questionnaires that included questions about probands’ demographics and comprehensive family health history, including history of PC in an FDR (parents, siblings, and offspring). Therefore, analyses were based on probands’ recall of cancer diagnosis in their FDRs. However, in our registry, we have found a 98% concordance between probands report of PC diagnosis in an FDR and the FDRs’ own report (7). We compared the reported number of PC cases in the FDRs to the expected number of cases based on SEER data (SEER 21 registries, 2000-2017), covering ~28% of the general U.S. population (34). Because PC is rarely diagnosed before age 20 years, we restricted analyses to FDRs who were ≥20 years at the time of questionnaire completion, and constructed pedigrees based on family information provided by the probands.

ABO blood groups

Data on ABO blood groups were available for 2,082 (64%) of the 3,268 PC probands. For the majority of the probands (n=1,147), ABO blood groups were determined through serological testing, retrieved from medical records. For the remaining 935 probands, ABO blood groups were genetically determined using two tagging single nucleotide polymorphisms (SNPs rs505922/rs687289 and rs8176746) in the ABO gene locus (9q34.1-q34.2) that we (11) and others (8,35) have shown to predict serological ABO blood groups. The SNPs were genotyped as part of the Pancreatic Cancer Cohort Consortium’s genome-wide association studies (GWAS)(15,36) or the Pancreatic Cancer Case Control Consortium’s GWAS (37). We primarily used rs505922 and rs8176746 for imputation of the blood groups. Where rs505922 was not available for a particular proband, we used rs687289 instead, because rs505922 and rs687289 are in complete linkage disequilibrium (r² = 1.0 in HapMap CEU subjects)(38) and are perfect surrogates for each other (8) (see online Supplementary Methods for additional details). Studies have shown that these SNPs predict serological blood groups with ≥90% accuracy when blood groups are imputed genetically as A, B, AB, and O, and ≥99% accuracy when imputed as O versus non-O (8–10). However, we evaluated concordance between serologically determined and genotype determined ABO blood groups among probands who had data on both serotype- and genotype-derived ABO blood groups to assess the accuracy of our imputations (Supplementary Tables 1 and 2).

Cancer susceptibility genes

Genetic sequencing data were available for 2,799 (86%) of the probands. We selected previously established candidate cancer susceptibility genes for assessment of FDRs’ risk of PC based on whether their respective proband carried a pathogenic or likely pathogenic variant (mutation) in one or more cancer susceptibility genes (39). The genes were sequenced in two genetic studies among the probands. The first sequencing project was performed with a custom capture multiplex polymerase chain reaction using a QIAseq assay (Qiagen Inc.), as described (39). The second sequencing project was performed by Color Genomics™ (Burlingame, CA) using a clinical-grade assay (40). A detailed description of the sequencing methods, bioinformatics pipeline, and quality control checks used by Color Genomics™ has been published (40,41). For analyses, we selected 20 cancer susceptibility genes that overlapped the QIAseq and Color Genomics™ sequencing projects: APC, ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CDKN2A, CHEK2, EPCAM, MLH1, MSH2, MSH6, NBN, PALB2, PMS2, PTEN, RAD51C, RAD51D, and TP53. We classified probands as “mutation-positive” if they carried at least one pathogenic or likely pathogenic variant, as determined previously (39), in any of the 20 cancer genes. Probands who did not carry pathogenic or likely pathogenic variants in any of the 20 genes were classified as “mutation-negative”. FDRs’ risk of PC was based on the mutation status of their respective proband.

Statistical analysis

SIRs and 95% confidence intervals (CIs) were calculated to assess risk of PC among the FDRs by dividing the observed number of cases of PC by the expected number of cases calculated from age-specific (five-year intervals) and sex-specific incidence rates in the SEER-21 data (31) then multiplied by person-years at risk. The observed number of cases was based on probands’ report of PC diagnosis in an FDR. SIRs were calculated for the overall sample, and by FDRs’ smoking status and kinship to the proband based on age- and sex-specific incidence rates in SEER, and for male and female FDRs’ separately based on age-specific rates. We also calculated age- and sex-standardized incidence ratios for PC risk among the FDRs based on: (1) their respective proband’s ABO blood group (blood group A, B or AB, or O; and O vs. non-O), (2) the proband’s susceptibility gene mutation status (mutation-positive vs. mutation-negative), and (3) a combination of proband’s ABO blood group and mutation status (blood group O and mutation-positive, blood group O and mutation-negative, non-O blood group and mutation-positive, and non-O blood group and mutation-negative). We further calculated p-values to test for statistical difference in SIRs between groups assuming binomial distribution (42). These analyses were repeated in subgroups defined by FDRs’ sex, FDRs’ smoking status, and kinship to the proband. In a subset of probands who had data on both serological blood group and genetically determined blood group (n=393), we assessed concordance between the serotype- and genotype-derived ABO blood groups. All statistical tests were two-sided, and a p-value < 0.05 was considered statistically significant. Analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and R version 3.6.2.

Results

Descriptive characteristics of the 3,268 PC probands are reported in Table 1. The probands were predominantly white (98%), and a majority were male (55%). Mean age of PC diagnosis among the probands was 66 years. Data on ABO blood group were available for 2,082 (64%) probands, determined serologically (55%) or genetically. Most of the probands had blood group A (49%) or O (34%), fewer had B or AB (17%). Among probands with data on both serotype and genotype determined ABO blood groups, we found 99% (391/393) concordance between serotype- and genotype-derived blood groups when imputed as A, B or AB, and O, and remained 99% when compared as O versus non-O (Supplementary Table 2). Data on germline mutation status were available for 2,799 (86%) probands, among whom 266 (9.5%) tested positive for one or more mutations in the 20 cancer susceptibility genes tested (Table 1). The top five genes with the most pathogenic variants in the probands were ATM (n=72, 2.6% of probands), BRCA2 (n=66, 2.3%), CHECK2 (n=29, 1.0%), CDKN2A (n=23, 0.8%), and BRCA1 (n=18, 0.6%).

Table 1.

Descriptive characteristics of the pancreatic cancer probands

Patient characteristics	N = 3,268
Age, years a
Mean (SD)	65.7 (10.3)
Median	66.0
Range	(20.0-92.0)
Sex
Male	1,799 (55.0%)
Female	1,469 (45.0%)
Race
White	3,188 (97.7%)
Black/African American	33 (1.0%)
Asian/Asian-American	18 (0.6%)
American Indian/Alaskan Native	10 (0.3%)
Native Hawaiian/Other Pacific Islander	2 (0.1%)
Multiracial	13 (0.4%)
Unknown	4
ABO blood group b
A	1,019 (48.9%)
B	289 (13.9%)
AB	58 (2.8%)
O	716 (34.4%)
Missing	1,186
O vs. non-OABO blood group b,c
O	716 (34.4%)
Non-O	1,366 (65.6%)
Missing	1,186
ABO data source
Serology	1,147 (55.1%)
Genotype-derived	935 (44.9%)
Missing	1,186
Germline mutation status d
Mutation positive	266 (9.5%)
Mutation negative	2,533 (90.5%)
Missing	469

^aAge at PDAC diagnosis

^bBlood group was determined serologically or genetically imputed based on the following polymorphisms in the ABO locus that predict ABO blood groups (rs505922/rs687289 and rs8176749). Those with missing information did not have serological or genetic data available and were excluded from the ABO blood group analyses.

^cNon-O blood group was categorized by combining blood groups A, B, and AB into one group.

^dProbands were categorized as mutation-positive if they tested positive for one or mutations in any of the following cancer susceptibility genes: APC, ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CDKN2A, CHEK2, EPCAM, MLH1, MSH2, MSH6, NBN, PALB2, PMS2, PTEN, RAD51C, RAD51D, and TP53. Probands were classified as mutation-negative if they had no mutation in all 20 genes tested.

Comprehensive family histories were provided by the probands and were used to derive the analytic sample of 23,739 FDRs, with up to 956,491 person-years at risk. The FDRs comprised 6,536 parents, 9,619 siblings, and 7,584 offspring. The average age of the FDRs was 60 years, with nearly equal proportion of men and women (49.2% vs. 50.8%), and most FDRs were non-smokers (Supplementary Table 3). Age- and sex-standardized incidence ratio showed a two-fold higher-than-expected risk of PC among the FDRs (SIR=2.00, 95%CI=1.79-2.22), compared with the SEER population (Table 2). We also found a higher-than-expected age-standardized PC risk among male (SIR=1.71, 95%CI=1.45-1.99) and female (SIR=2.30, 95%CI=1.97-2.67) FDRs. Additional analysis showed that the magnitude of risk for female FDRs is significantly higher than the magnitude of risk for male FDRs (between-group comparison, p=0.006). Further, PC risk was substantially higher among FDRs who ever smoked (SIR=2.45, 95%CI=2.09-2.85) than FDRs who never smoked (SIR=1.63, 95%CI=1.36-1.92) (between-group comparison, p<0.001). SIRs (95% CIs) for PC risk among mothers, fathers, and siblings of the probands were 3.50 (2.86-4.23), 2.94 (2.40-3.56), and 1.61 (1.34-1.93), respectively. For mutation status only, we found a substantially higher PC risk among FDRs of mutation-positive probands (SIR=3.80, 95%CI=2.81-5.02) than among FDRs of mutation-negative probands (SIR=1.79, 95%CI=1.57-2.04) (between-group comparison, p<0.001).

Table 2.

Risk of pancreatic cancer among first-degree relatives (FDRs) of pancreatic cancer probands

Study sample	Probands (N)	FDRs (N)	Person-Years at Risk	PC observed in FDR (N)	PC expected in FDR (N)	SIR (95% CI)a	p-valueb
Overall	3,268	23,739	956,491	336	168.3	2.00 (1.79-2.22)
Male FDRs	1,799	12,052	474,629	162	94.9	1.71 (1.45-1.99)	0.006
Female FDRs	1,469	11,687	481,862	174	75.7	2.30 (1.97-2.67)
FDRs’ smoking status
Never smokers	3,066	12,446	478,791	137	84.3	1.63 (1.36-1.92)	<0.001
Ever smokers	2,887	9,167	387,355	167	68.2	2.45 (2.09-2.85)
Kinship to proband
Mother	3,268	3,268	191,228	105	33.0	3.50 (2.86-4.23)
Father	3,268	3,268	175,391	103	35.1	2.94 (2.40-3.56)
Siblings	2,987	9,619	422,492	120	74.4	1.61 (1.34-1.93)	<0.001
Parentsc	3,268	6,536	366,618	208	64.5	3.22 (2.80-3.69)
Mutation positived	266	1,908	73,328	49	12.9	3.80 (2.81-5.02)	<0.001
Mutation negative	2,533	18,505	745,578	235	131.2	1.79 (1.57-2.04)

^aAge and sex standardized incidence ratio, except for sex-specific analyses which were standardzed by age alone.

^bP-values comparing male vs. female first-degree relatives; first-degree relatives who ever smoked vs. never smoked; siblings vs. parents; first-degree relatives of mutation-positive vs. mutation-negative probands.

^cMother and father combined into one group

^dProbands were categorized as mutation positive if they tested positive for one or mutations in the following genes: APC, ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CDKN2A, CHEK2, EPCAM, MLH1, MSH2, MSH6, NBN, PALB2, PMS2, PTEN, RAD51C, RAD51D, and TP53, otherwise the probands were categorized as mutation negative if they tested negative for all the genes.

Abbreviations: FDR, first-degree relative; PC, pancreatic cancer; SIR, standardized incidence rate

Table 3 presents age- and sex-standardized incidence ratios for PC among FDRs based on probands’ ABO blood group, and both the probands’ blood group and probands’ germline mutation status, compared with the SEER population. For ABO blood group only, we found a 1.75 (95%CI: 1.41-2.14) times higher-than-expected risk of PC for FDRs of probands with blood group A, a 2.06 (95%CI: 1.46-2.83) times higher-than-expected risk for FDRs of probands with blood group B or AB, and a 1.57 (95%CI: 1.20-2.03) times higher-than-expected risk for FDRs of probands with blood group O. Collectively, PC risk for FDRs of probands with a non-O blood group was 1.83 (95%CI=1.53-2.17) times higher than expected but this did not differ significantly from the risk among FDRs of probands with blood group O (p=0.33).

Table 3.

Risk of pancreatic cancer among first-degree relatives of pancreatic cancer probands by probands ABO blood group and proband’s germline susceptibility gene mutation status

Proband characteristica	Probands (N)	FDRs (N)	Person-Years at Risk	PC observed in FDR (N)	PC expected in FDR (N)	SIR (95% CI)	p-valuee
ABO blood group A	1,019	7,501	298,962	92	52.6	1.75 (1.41-2.14)
ABO blood group B/ABb	347	2,620	104,699	38	18.4	2.06 (1.46-2.83)
ABO blood group non-Oc	1,366	10,121	403,662	130	71.0	1.83 (1.53-2.17)
ABO blood group O	716	5,300	213,410	59	37.6	1.57 (1.20-2.03)	0.33
ABO blood group non-O
Mutation positive	142	998	38,508	27	6.8	3.98 (2.62-5.80)	<0.001
Mutation negative	1,123	8,404	334,684	98	58.9	1.66 (1.35-2.03)
ABO blood group O
Mutation positive	52	405	14,991	7	2.6	2.65 (1.06-5.47)	0.16
Mutation negative	608	4,472	180,546	47	31.8	1.48 (1.09-1.97)

^aABO blood group was determined by combining serological data with data on polymorphisms in the ABO locus that are known to encode ABO blood group.

^bBlood group B and AB were combined into one group because the polymorphisms used to infer blood group were able to detect the presence of blood group B, but could distinguish blood group AB from blood group B.

^cNon-O blood group categorized by combining A, B, and AB into one group.

^dProbands were categorized as mutation positive if they tested positive for one or mutations in the following genes: APC, ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CDKN2A, CHEK2, EPCAM, MLH1, MSH2, MSH6, NBN, PALB2, PMS2, PTEN, RAD51C, RAD51D, and TP53, otherwise the probands were categorized as mutation negative if they tested negative for all of the genes.

^eP-value comparing first-degree relatives of probands with O vs. non-O blood groups, first-degree relatives of probands who are mutation-positive vs. mutation negative, first-degree relatives of probands with blood group O and mutation positive vs. those with blood group O and mutation negative, or first-degree relatives of probands with non-O blood group and mutation positive vs. those with non-O blood group and mutation negative.

Abbreviations: PC, pancreatic cancer; FDR, first-degree relative; SIR, standardized incidence ratio

In the combined analysis of probands’ ABO blood group and mutation status, we found that among FDRs of probands with blood group O, those FDRs of mutation-positive probands had a higher risk for PC (SIR=2.65, 95%CI=1.06-5.47), as did FDRs of mutation-negative probands (SIR=1.48, 95%CI=1.09-1.97) (Table 3). However, the risk estimates did not differ significantly by probands’ mutation status among FDRs of probands with ABO blood group O (p=0.16). By contrast, among FDRs of probands with non-O blood group, the FDRs of mutation-positive probands had a 3.98 times higher PC risk (SIR=3.98, 95%CI=2.62-5.80), while FDRs of mutation-negative probands had a 1.66 times higher risk (SIR=1.66, 95%CI=1.35-2.03), and the risk estimates differed significantly by probands’ mutation status (p<0.001). These differences are shown graphically in Figure 1.

Figure 1.

Standardized Incidence Ratios (SIRs) and 95% confidence intervals for associations of probands’ ABO blood group and mutation status on first-degree relatives’ risk for pancreatic cancer. P-values were computed for comparisons within the same blood group, but different mutation status.

Abbreviations: ABO-O, ABO blood group O; ABO-non-O, ABO blood group non-O; mut-pos, mutation-positive; mut-neg, mutation-negative.

We also performed subgroup analyses sex and smoking status of the FDRs according to the respective proband’s blood group and mutation status (Table 4). The major findings include a higher PC risk among females FDRs of probands with blood group O (SIR=2.13, 95%CI=1.49-2.95), whereas no association was found among male FDRs of probands with blood group O (SIR=1.09, 95%CI=0.69-1.63). Both male and female FDRs of probands with non-O blood group and mutation-positive had higher PC risk (male, SIR=3.99, 95%CI=2.23-6.58; female, SIR=3.88, 95%CI=2.00-6.77). However, while no association was found among male FDRs of probands with blood group O and mutation-negative (SIR=1.06, 95%CI=0.64-1.65), a significant association was found for female FDRs of probands with blood group O and mutation-negative (SIR=1.97, 95%CI=1.31-2.84). FDRs who ever smoked generally had higher PC risk than FDRs who never smoked irrespective of probands’ ABO blood group and mutation status. For example, a significant association was observed among FDRs who were smokers and are related to probands with blood group O (SIR=1.93, 95%CI=1.29-2.77), but no association was found among FDRs who never smoked and are related to probands with blood group O (SIR=1.43, 95%CI=0.97-2.08). FDRs who ever smoked and are related to mutation-positive probands with non-O blood group also had higher PC risk (SIR=4.92, 95%CI=2.69-8.26) than FDRs who never smoked and are related to mutation-positive probands with non-O blood group (SIR=2.69, 95%CI=1.23-5.11) (Table 4).

Table 4.

Standardized incidence ratios for pancreatic cancer (PC) risk among FDRs of pancreatic cancer probands according to probands’ ABO blood group and probands’ germline susceptibility gene mutation status, stratified by FDRs’ sex and FDRs’ smoking status

	Associations by FDRs’ sex
	Male FDRs (n=12,052)				Female FDRs (n=11,687)
Proband characteristicsa	PC observed in FDR (N)	PC expected in FDR (N)	Person-Years at Risk	SIR (95% CI)	PC observed in FDR (N)	PC expected in FDR (N)	Person-Years at Risk	SIR (95% CI)
ABO blood group A	45	29.3	146,674	1.53 (1.12-2.05)	47	23.9	152,288	1.97 (1.44-2.61)
ABO blood group B/ABb	21	10.3	51,446	2.04 (1.26-3.12)	17	8.4	53,253	2.03 (1.18-3.26)
ABO blood group non-Oc	66	39.6	198,121	1.67 (1.29-2.12)	64	32.3	205,541	1.98 (1.53-2.53)
ABO blood group O	23	21.2	105,812	1.09 (0.69-1.63)	36	16.9	107,599	2.13 (1.49-2.95)
ABO blood group non-O
Mutation positive	15	3.8	18,798	3.99 (2.23-6.58)	12	3.1	19,710	3.88 (2.00-6.77)
Mutation negative	50	32.9	164,700	1.52 (1.13-2.00)	48	26.7	169,984	1.80 (1.33-2.38)
ABO blood group O
Mutation positive	3	1.4	7,200	2.08 (0.42-6.09)	4	1.2	7,791	3.27 (0.88-8.37)
Mutation negative	19	18.0	89,921	1.06 (0.64-1.65)	28	14.2	90,625	1.97 (1.31-2.84)
	Associations by FDRs’ smoking statusa
	Ever smokers (n=9,167)				Never smokers (n=12,446)
Proband characteristicsa	PC observed in FDR (N)	PC expected in FDR (N)	Person-Years at Risk	SIR (95% CI)	PC observed in FDR (N)	PC expected in FDR (N)	Person-Years at Risk	SIR (95% CI)
ABO blood group A	41	20.9	118,857	1.96 (1.41-2.66)	41	26.6	151,205	1.54 (1.11-2.09)
ABO blood group B/ABb	20	7.3	41,601	2.73 (1.67-4.22)	12	9.2	52,313	1.30 (0.67-2.28)
ABO blood group non-Oc	61	28.2	160,458	2.16 (1.65-2.77)	53	35.8	203,518	1.48 (1.11-1.94)
ABO blood group O	29	15.0	85,400	1.93 (1.29-2.77)	27	18.9	107,244	1.43 (0.94-2.08)
ABO blood group non-O
Mutation positive	14	2.8	16,168	4.92 (2.69-8.26)	9	3.3	19,010	2.69 (1.23-5.11)
Mutation negative	46	23.3	132,612	1.97 (1.44-2.63)	40	29.5	167,739	1.35 (0.97-1.85)
ABO blood group O
Mutation positive	4	1.0	5,949	3.82 (1.03-9.78)	3	1.3	7,352	2.32 (0.47-6.77)
Mutation negative	25	13.0	73,975	1.92 (1.24-2.83)	20	15.9	90,607	1.25 (0.77-1.94)

^aABO blood group was determined by combining serological data with data on polymorphisms in the ABO locus known to encode ABO blood groups.

^bBlood group B and AB were combined into one group because the polymorphisms used to infer blood group were able to detect the presence of blood group B, but could distinguish blood group AB from blood group B.

^cNon-O blood group categorized by combining A, B, and AB into one group.

^dProbands were categorized as mutation positive if they tested positive for one or mutations in the following genes: APC, ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CDKN2A, CHEK2, EPCAM, MLH1, MSH2, MSH6, NBN, PALB2, PMS2, PTEN, RAD51C, RAD51D, and TP53, otherwise the probands were categorized as mutation negative if they tested negative for all of the genes.

^eP-value comparing first-degree relatives of probands with O vs. non-O blood groups, first-degree relatives of probands who are mutation-positive vs. mutation negative, first-degree relatives of probands with blood group O and mutation positive vs. those with blood group O and mutation negative, or first-degree relatives of probands with non-O blood group and mutation positive vs. those with non-O blood group and mutation negative.

Abbreviations: PC, pancreatic cancer; FDR, first-degree relative; SIR, standardized incidence ratio

Additional analyses were performed by probands’ blood group and combination of probands’ ABO blood group and mutation status according to FDRs’ kinship to the proband (Table 5). The results generally show higher PC risk among mothers of the FDRs, except that in probands with non-O blood group and were mutation-positive, fathers of the proband had substantially higher risk (SIR=8.05, 95%CI=4.15-14.06) than mothers (SIR=4.09, 95%CI=1.32-9.55) and siblings (SIR=2.90, 95%CI=1.32-5.50). Among FDRs of probands with blood group O and mutation-negative, a significant association was found among mothers of the proband (SIR=23, 95%CI=1.91-5.11) but no association was found among fathers or siblings (Table 5).

Table 5.

Risk for pancreatic cancer among FDRs of pancreatic cancer probands according to probands ABO blood group and probands germline susceptibility gene mutation status, stratified by kinship to the proband

	Mother (n=3,268, from 3,268 pedigrees)				Father (n=3,268, from 3,268 pedigrees)				Siblings (n=9,619, from 2,987pedigrees)
Proband characteristicsa	PC observed in FDR (N)	PC expected in FDR (N)	Person-Years at Risk	SIR (95% CI)	PC observed in FDR (N)	PC expected in FDR (N)	Person-Year at Risk	SIR (95% CI)	PC observed in FDR (N)	PC expected in FDR (N)	Person-Years at Risk	SIR (95% CI)
ABO blood group A	28	9.4	59,928	2.98 (1.98-4.30)	30	10.9	54,255	2.76 (1.86-3.95)	29	23.2	131,949	1.25 (0.84-1.79)
ABO blood group B/ABb	10	3.1	19,865	3.21 (1.54-5.90)	16	3.6	18,240	4.39 (2.51-7.12)	11	8.2	46,629	1.34 (0.67-2.40)
ABO blood group non-Oc	38	12.5	79,793	3.03 (2.15-4.16)	46	14.5	72,495	3.17 (2.32-4.23)	40	31.4	178,578	1.27 (0.91-1.73)
ABO blood group O	20	6.5	41,619	3.06 (1.87-4.73)	12	7.7	38,573	1.56 (0.80-2.72)	25	16.6	94,520	1.50 (0.97-2.22)
ABO blood group non-O
Mutation positive	5	1.2	7,784	4.09 (1.32-9.55)	12	1.5	7,455	8.05 (4.15-14.06)	9	3.1	17,653	2.90 (1.32-5.50)
Mutation negative	31	10.4	66,019	2.99 (2.03-4.25)	33	11.9	59,589	2.77 (1.91-3.89)	29	26.0	147,482	1.12 (0.75-1.60)
ABO blood group O
Mutation positive	1	0.4	2,793	2.28 (0.03-12.69)	0	0.5	2,736	NE	5	1.1	6,515	4.36 (1.41-10.18)
Mutation negative	18	5.6	35,482	3.23 (1.91-5.11)	12	6.6	32,956	1.82 (0.94-3.18)	17	14	79,670	1.21 (0.71-1.94)

^aABO blood group was determined by combining serological data with data on polymorphisms in the ABO locus that are known to encode ABO blood group.

^bBlood group B and AB were combined into one group because the polymorphisms used to infer blood group were able to detect the presence of blood group B, but could distinguish blood group AB from blood group B.

^cNon-O blood group categorized by combining A, B, and AB into one group.

^dProbands were categorized as mutation positive if they tested positive for one or mutations in the following genes: APC, ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CDKN2A, CHEK2, EPCAM, MLH1, MSH2, MSH6, NBN, PALB2, PMS2, PTEN, RAD51C, RAD51D, and TP53, otherwise the probands were categorized as mutation negative if they tested negative for all of the genes.

^eP-value comparing first-degree relatives of probands with O vs. non-O blood groups, first-degree relatives of probands who are mutation-positive vs. mutation negative, first-degree relatives of probands with blood group O and mutation positive vs. those with blood group O and mutation negative, or first-degree relatives of probands with non-O blood group and mutation positive vs. those with non-O blood group and mutation negative.

Abbreviations: PC, pancreatic cancer; FDR, first-degree relative; NE, not estimable; SIR, standardized incidence ratio

Discussion

We investigated PC risk among 23,739 FDRs with 956,491 person-years at risk, identified through 3,268 PC probands who were recruited at Mayo Clinic. The results show a two-fold higher than expected risk of PC among FDRs. By estimating FDRs’ risk according to probands’ ABO blood group, we found that the FDRs had higher-than-expected risk of PC regardless of the probands’ blood group, and the magnitude of risk did not differ significantly among FDRs of probands with blood group O versus non-O. In separate analyses based on probands’ cancer susceptibility gene mutation status, we found higher PC risk in both FDRs of mutation-positive probands and FDRs of mutation-negative probands, but the risk magnitude was significantly higher in FDRs of mutation-positive probands. We also found that FDRs’ risk for PC was substantially and significantly higher when the related proband had a non-O blood group in addition to carrying at least one mutation in a cancer susceptibility gene than when the proband had a non-O blood group but did not carry a mutation in any of the genes tested. We also examined associations of two characteristics (sex and smoking status) of the FDRs in the context of the probands’ genetic status and found that PC risks were higher in both male and female FDRs, although female FDRs had significantly higher magnitude of risk than male FDRs. FDRs who smoked also had significantly higher PC risk compared to FDRs who never smoked. FDRs who ever smoked and are related to mutation-positive probands with non-O blood group had highest PC risk (SIR=4.92) compared to all subgroups of non-smokers. Collectively, the combination of a non-O blood group and presence of a cancer susceptibility gene mutation in a proband appears to contribute to PC risk among FDRs of the PC probands.

In a previous study involving fewer FDRs (n=17,162) and an earlier version of SEER data (SEER 9, 1973-2013), we estimated a two-fold higher risk of PC among FDRs of PC probands (7), which is consistent with the findings of the present study that is based on a much larger sample size and more recent SEER data with expanded number of cancer registries. Other studies have reported similarly higher risks of PC among FDRs of PC probands (43,44). In this study, we also found that female FDRs had significantly higher PC risk than male FDRs, as was observed previously (7), but the reasons are not clear and could be due to potential differences in the reporting of PC diagnosis by the probands for male versus female FDRs. We also previously estimated a 4.32 times higher PC risk among FDRs of probands who carry one or more mutations in cancer susceptibility genes (7), slightly higher than the present SIR estimate of 3.80 among FDRs of mutation-positive probands.

This is the first study to evaluate FDRs’ risk of PC based on probands’ ABO blood group and a combination of probands’ ABO blood group and probands’ cancer gene mutation status. While we did not find a significant difference in the magnitude of risk between FDRs of probands with O versus non-O blood group, both groups of FDRs had higher-than-expected risk when compared with the SEER population. Importantly, we found that FDRs of probands who had a non-O blood group and tested positive for cancer susceptibility gene mutation had a statistically significantly higher magnitude of PC risk (SIR=3.98) than FDRs of probands with non-O blood group and tested negative for cancer gene mutation (SIR=1.66). Among FDRs of probands with blood group O, risk estimates were generally lower and did not vary significantly by probands’ mutation status. This supports the hypothesis that PC risk is often less in families with predominance of blood group O, over other ABO blood groups.

Although it is not entirely clear at this point whether ABO blood group alone can be used as an independent indicator for predicting familial PC risk, we posit that a combination of familial aggregation of non-O blood groups with presence of germline mutation(s) in cancer susceptibility genes among family members could robustly predict familial PC risk. We previously demonstrated the role of germline mutations in PC susceptibility (39) and the utility of a proband’s susceptibility gene mutation status to infer FDRs’ risk for PC (7). Here we add evidence that the presence of a non-O blood group in a proband in addition to germline mutation(s) in cancer susceptibility gene in the proband is strongly associated with PC risk among FDRs of the proband. This study demonstrates the importance of two discrete heritable factors in the familial aggregation of PC, which if confirmed by others, could guide genetic counseling considerations.

We also emphasize that in addition to heritability factors, shared environmental risk factors could contribute to the familial aggregation of PC. While heritable genetic factors are estimated to account for up to 10% of PC cases, cigarette smoking alone currently accounts for about 12% of PC cases (2,45). We found that FDRs who smoked had substantially higher PC risk, particularly if they are related to a mutation-positive proband. Rulyak et al. have shown that smoking in an independent risk factor for PC in familial PC kindreds (46). In that study, smokers had a 3.7-fold higher risk of PC and were diagnosed with PC a decade earlier than non-smoking family members (46). Otten et al.(47) and Gilman et al.(48) have both shown that parental smoking strongly influences smoking initiation by their offspring, which may partly explain the aggregation of smoking-related cancers, such as PC, in families. To a lesser extent, similar patterns could exist for other shared lifestyle factors, such as poor dietary habits or physical inactivity, both of which are associated with a higher risk of PC (2). This suggests a potential research direction that when combined with the risk factors we have examined here, non-genetic exposures of the FDRs may further enhance risk assessment.

Our study has several strengths and limitations. It is limited to a predominantly white population, which hinders ability to generalize broadly to minority populations. Diagnosis of PC among probands was determined histologically or radiologically, but such level of detail was not possible for the FDRs, and we relied on probands’ report of PC diagnosis in an FDR. In our pancreas registry, we have found a 98% concordance between probands’ report and FDRs’ self-report of PC diagnosis (7), indicating that the probands’ reports are highly reliable. We also did not have ABO blood group information for the FDRs. Further, we genetically imputed ABO blood groups for some of the probands (45%) using tag SNPs in the ABO gene locus that have been shown to predict serological ABO blood groups. The use of SNPs to determine blood groups is generally less granular than serotype-derived blood groups because of the possibility of measurement error. While the SNPs were able to identify the presence of the B ABO allele, they could not distinguish the B blood group from the AB blood group; therefore, we combined the B with the AB blood group for analyses to assess risk of PC in families of probands who carry the B allele. Additionally, only 20 cancer susceptibility genes were assessed; hence, mutations in other cancer genes in the probands could have a modest effect due to misclassification. Some cells in the stratified analyses are small and those results need to be interpreted with caution. Strengths of the study include its uniquely large sample size and the use of a structured and comprehensive risk factor questionnaire that ensured uniform collection of family history data. Availability of data on probands’ susceptibility gene mutation status and probands’ ABO blood groups also add to the study strengths. Furthermore, this is the first study to estimate FDRs’ risk of PC based on probands’ ABO blood group (9,13), providing suggestive evidence of PC risk based on familial aggregation of ABO blood group alleles.

In conclusion, we report a two-fold higher risk of PC among FDRs of PC probands compared to the SEER population. We also found that FDRs risk for PC is enhanced by the combined presence of non-O blood group and germline mutation(s) in cancer susceptibility genes in their respective proband, which could be useful for familial risk estimation and genetic counseling among relatives of PC probands. SIR for PC risk in FDRs ranged from as low as 1.48 for FDRs of probands with blood group O and mutation-negative to 3.98 for FDRs of probands with non-O blood group and mutation-positive.

Impact:

Combined ABO blood group and germline mutation status of probands can inform PC risk estimation in FDRs.

Abbreviations:

CI

confidence interval

FDR

first-degree relatives

GWAS

genome-wide association study

H pylori

Helicobacter pylori

PC

pancreatic cancer

SEER

Surveillance, Epidemiology, and End Results

SIR

standardized incidence ratios

Data availability statement:

The data may be made available to researchers upon request to Dr. Gloria Petersen (Petersen.gloria@mayo.edu). Ethical and legal restrictions apply to these data.