Pancreatic cancer is the third leading cause of cancer-related death in the USA(1). In 2019, approximately 56 770 subjects developed pancreatic cancer and an estimated 45 750 subjects died from the disease(1). Pancreatic cancer has been projected to become the second leading cause of cancer-related death by 2030 for the US population(Reference Rahib, Smith and Aizenberg2) despite improvements in 5-year cancer survival in the past decades(Reference Lepage, Capocaccia and Hackl3). Early detection is an effective approach to reduce cancer mortality, but an accurate screening test is not yet available for pancreatic cancer. The aetiology of pancreatic cancer remains elusive as cigarette smoking, family history, chronic pancreatitis and obesity are the only well-established risk factors(Reference Yadav and Lowenfels4). Therefore, it is important to identify modifiable risk factors for the primary prevention of pancreatic cancer.
Ca, Mg and phosphorus are essential minerals that are metabolically correlated and are crucial for many biologic and cellular functions(Reference Blaine5), including bone turnover, energy metabolism and inflammation(Reference Bendsen, Hother and Jensen6–Reference Zemel and Miller10). A growing body of evidence from experimental and human studies suggests that these minerals, particularly Ca, play a pivotal role in pancreatic carcinogenesis. Randomised trials showed that an increased intake of Ca significantly promoted faecal fat excretion and reduced levels of total cholesterol and LDL-cholesterol due to Ca soap formation and binding of bile acids in the intestine(Reference Bendsen, Hother and Jensen6,Reference Soerensen, Thorning and Astrup8) . Animal studies revealed that high-Ca diets induced weight loss through inhibiting lipogenesis, accelerating lipolysis and enhancing thermogenesis(Reference Zemel and Miller11). These findings offer a firm biological basis for the observation that dietary intake of Ca and the ratio of dietary Ca to phosphorus (Ca:P ratio) were inversely associated with obesity risk(Reference Pereira Dde, Lima and de Lima12). Obesity and diabetes have been linked to an increased risk of pancreatic cancer in many epidemiological studies(Reference Berrington de Gonzalez13,Reference Elena, Steplowski and Yu14) .
The essential roles of these elements in cellular functions suggest potential mechanisms related to carcinogenesis. Intracellular Ca+2 concentrations modulate the proliferation and apoptosis of immune cells and cancer cells, and the rise of cytosolic Ca+2 is necessary for efficient targeting and killing of tumour cells by cytotoxic T lymphocytes and natural killer cells(Reference Schwarz, Qu and Hoth15). Mg deficiency is commonly found in patients with diabetes(Reference Fatemi, Ryzen and Flores7). In a human experimental study, mild Mg depletion significantly lowered serum levels of Ca and 1,25-(OH)2D(Reference Fatemi, Ryzen and Flores7). Despite sound biological plausibility, the associations between intake of Ca, Mg and phosphorus and risk of pancreatic cancer have been inconsistent across previous epidemiological studies, with reports of a significant inverse association with intake of Ca(Reference Farrow and Davis16) and Mg(Reference Dibaba, Xun and Yokota17), a significant positive association with Ca intake(Reference Zablotska, Gong and Wang18) and a non-significant association with intake of Ca(Reference Genkinger, Wang and Li19), Mg(Reference Molina-Montes, Wark and Sánchez20) and phosphorus(Reference Bravi, Polesel and Bosetti21). Therefore, the present study sought to investigate these associations in a population-based, case-control study in Minnesota.
Materials and methods
Study population
The design and methodology of the case-control study of pancreatic cancer conducted from April 1994 to September 1998 in Minnesota have been described in detail elsewhere(Reference Anderson, Sinha and Kulldorff22,Reference Gross, Kruisselbrink and Anderson23) . Briefly, cases were patients recently diagnosed with pancreatic ductal adenocarcinoma (International Classification of Disease for Oncology, 3rd ed., code C25) and were 20 years or older, English-speaking and mentally competent. The source cohort was residents of the Upper Midwest and cases were recruited from all hospitals in the seven-county metropolitan area of Minneapolis and Saint Paul, Minnesota and the Mayo Clinic. Given the high fatality of pancreatic cancer, a rapid case-ascertainment system was adopted for case enrollment. The median number of days between diagnosis and first contact for the study was only 13 d for the cases recruited to the study.
Eligibility criteria for controls were the same as those for cases, disallowing a diagnosis of pancreatic cancer. Controls were randomly recruited from the source population of cases. Controls aged 20–64 years were identified from drivers’ licences and state identity card database, while those aged 65 years or older were obtained from US Health Care Financing Administration (now Centers for Medicare and Medicaid Services) records. Controls were frequency matched to cases by age (within 5 years), sex and race.
A total of 460 cases were identified and met the eligibility criteria. Of these, 202 were excluded due to death prior to being contacted or interviewed (n 85), refusal to participate (n 79), disallowance by their physician (n 31) and inability to be reached or contacted (n 7). After these exclusions, 258 participated in the study, yielding a response rate of 56·1 %. A total of 1141 eligible controls were ascertained and 676 of them agreed to participate in the study, giving a response rate of 59·2 %. Dietary and alcohol intake data were not collected from 108 cases and 217 controls largely because cases were too frail to endure the interview process or because controls declined to respond to the FFQ. Finally, data from 150 cases and 459 controls were available for the present analysis.
Data collection
The present study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the institutional review boards of the University of Minnesota and the Mayo Clinic. Written informed consent was obtained from all subjects prior to the interview. A general questionnaire was used to solicit information regarding demographic characteristics (e.g. age, sex and race), socio-economic measures (e.g. education), family history of cancer, physical activity and cigarette smoking. A slightly modified version of the Willett FFQ was employed to assess the usual diet of the subjects. The Willett FFQ has been validated against dietary records, and validation studies showed that it had a reasonable level of reproducibility and validity for assessing individual nutrients and foods(Reference Feskanich, Rimm and Giovannucci24–Reference Smith-Warner, Elmer and Fosdick26). Specifically, the average de-attenuated correlation coefficient between the energy-adjusted nutrient intakes measured by the FFQ and diet records among 127 men was 0·65, with de-attenuated correlation coefficients of 0·61, 0·71 and 0·63 for Ca, Mg and phosphorus, respectively(Reference Rimm, Giovannucci and Stampfer25). In the present study, we used a 131-item Willett FFQ (HarvardSSFQ.5/93) that had been modified for Minnesota Cancer Prevention Research Unit studies to include additional vegetables, fruits and low-fat foods(Reference Smith-Warner, Elmer and Fosdick26). These modifications might have somewhat changed the reproducibility and validity of the Willet FFQ. The FFQ used in this case-control study has 153 individual foods or food groups (including alcohol consumption) commonly consumed in the USA and questions on use of nutrient supplements.
The general questionnaire and the FFQ were administered to study subjects by trained interviewers during face-to-face interviews. During the dietary survey, subjects were asked to recall how frequently they consumed each of the food items included in the FFQ in the year preceding pancreatic cancer diagnosis for cases or the referent date for controls. Dietary intake of total energy and nutrients was estimated by multiplying the portion size amount in each food item by the recalled frequency of consumption and summed over all food items. The amounts of energy and nutrients contained in portion sizes of all food items listed in the FFQ were derived from the Minnesota Colon Cancer Prevention Research Unit Studies database. Supplemental intake of Ca, Mg and phosphorus was also obtained from the responses of study subjects to the FFQ. Therefore, data on both total and dietary intakes of all three nutrients were available for the present analysis.
Statistical analysis
All dietary variables were adjusted for energy intake using the residual method prior to data analysis(Reference Willett and Stampfer27). Differences in categorical and continuous variables were examined with χ 2 test and t test, respectively. Pancreatic cancer risk, in relation to total and dietary intake of Ca, Mg and phosphorus, was estimated by performing unconditional logistic regression. OR and 95 % CI were calculated by comparing the second, third and fourth with the first quartile of total and dietary intakes of Ca, Mg and phosphorus. Cut-off points for creating the quartiles of each of the selected nutrients were based on their distributions among controls. Three regression models were constructed for each dietary variable. The first model estimated the effects of nutrients on pancreatic cancer risk without considering confounders. The second model was adjusted for age, sex, race, education (three levels), physical activity (h/week), cigarette smoking (never, former and current) and alcohol consumption (serving/week). The third model was additionally adjusted for intake of total energy, total fat, fibre, fruits and vegetables. The aforementioned covariates were adjusted as they are suspected or established confounders for the associations between dietary factors of interest and the risk of pancreatic cancer(Reference Yadav and Lowenfels4). The statistical significance of the linear trend across quartiles of each of the nutrients examined was tested by assigning a median intake value to each quartile and then treating these as values of a continuous variable.
As high Ca intake promotes faecal fat excretion and lowers levels of total cholesterol and LDL-cholesterol(Reference Bendsen, Hother and Jensen6,Reference Soerensen, Thorning and Astrup8) , the potential interactions on pancreatic cancer risk between each selected nutrient and total fat were evaluated with the likelihood ratio test. All statistical analyses were performed using SAS (version 9.4; SAS Institute Inc.), and a P-value of <0·05 was considered statistically significant.
Results
The mean ages of cases and controls were 65·8 and 66·3 years, respectively. Approximately 59·3 % of cases and 56·9 % of controls were male. Study subjects were predominately white (91·3 % for cases and 98·0 % for controls). Compared with controls, cases had a lower level of education and physical activity. Cases were more likely than controls to be former or current smokers and to report a history of diabetes (Table 1). Total intake of Ca (936 v. 1026 mg/d) and dietary intake of Mg (315 v. 331 mg/d) and phosphorus (1350 v. 1402 mg/d) were significantly lower in cases than in controls (Table 2).
Table 1. Characteristics of cases and controls in a population-based, case-control study of pancreatic cancer in Minnesota, 1994–1998*
(Numbers and percentages)
* Some variables have missing data.
† Values shown are mean (sd) for continuous variables and number (%) for categorical variables. t test and χ 2 test were used to compare differences in continuous and categorical variables between cases and controls, respectively.
‡ Data were missing from twenty-six cases and one control.
Table 2. Difference in intake of Ca, Mg and phosphorus (mean and sd) between cases and controls in a population-based, case-control study of pancreatic cancer in Minnesota, 1994–1998*
(Mean values and standard deviations)
* Some variables have missing data.
† t test was used to compare differences in intake of nutrients examined between cases and controls.
After adjustment for confounders, total and dietary intakes of Ca, Mg and phosphorus were not statistically significantly associated with the risk of pancreatic cancer (Table 3). In addition, no significant associations were observed for the ratio of total intake of Ca to phosphorus or the ratio of dietary intakes of Ca to phosphorus (data not shown). There were no significant interactions of total and dietary intakes of Ca, Mg and phosphorus with total fat intake in relation to pancreatic cancer risk (all P-interaction values >0·05). The analyses stratified by the median of total fat intake (72·2 g/d) did not reveal any clear patterns of differences in the associations between total and dietary intakes of nutrients considered and the risk of pancreatic cancer (Table A2 in the Appendix).
Table 3. Risk of pancreatic cancer in relation to intake of nutritional factors in a population-based, case-control study of pancreatic cancer in Minnesota, 1994–1998
(Odds ratio and 95 % confidence intervals)
* Adjusted for age, sex, race, education, physical activity, cigarette smoking and alcohol consumption per week.
† Additionally adjusted for intake of energy, total fat, fibre, fruits and vegetables.
Discussion
The present study found no evidence that there were statistically significant associations between total and dietary intakes of Ca, Mg and phosphorus and the risk of pancreatic cancer after adjustment for suspected and established confounders.
A number of cellular, animal and human studies have suggested that low levels of Ca are involved in pancreatic carcinogenesis. Experimental studies have shown that intracellular Ca+2 concentrations play a crucial role in the regulation of proliferation and apoptosis of immune and tumour cells and in the elimination of tumour cells by the innate immune system(Reference Schwarz, Qu and Hoth15). In addition, physiological intranuclear concentrations of Ca regulate DNA conformation and replication(Reference Dobi and Agoston28). Animal studies have consistently demonstrated the anti-obesity effect of dietary Ca. Transgenic mice fed on high-Ca diets exhibited an accelerated loss of fat and weight(Reference Zemel and Miller11). The results of animal studies have been partially replicated in human intervention trials where high Ca intake promoted faecal fat excretion and favourably influenced insulin resistance biomarkers(Reference Bendsen, Hother and Jensen6,Reference Soerensen, Thorning and Astrup8) . Although it is biologically plausible that Ca intake protects against pancreatic cancer, epidemiological studies evaluating the association between Ca intake and pancreatic cancer risk have yielded conflicting results.
In 1990, Farrow et al. reported a reduced risk of pancreatic cancer associated with Ca intake in a small case-control study conducted in Western Washington State(Reference Farrow and Davis16). However, this potential beneficial effect was not replicated in a large case-control study performed in the San Francisco Bay area. In the latter study, dietary intake of Ca was associated with an elevated risk of pancreatic cancer among men (OR (95 % CI) for ≥ 1200 mg/d v. < 500 mg/d: 2·8 (1·2, 6·4))(Reference Zablotska, Gong and Wang18). A pooled analysis of fourteen prospective cohort studies in Western countries showed inverse but non-significant associations of total and dietary intakes of Ca with the risk of pancreatic cancer (OR (95 % CI) for dietary Ca intake of ≥ 1100 mg/d v. < 500 mg/d: 0·86 (0·69, 1·07))(Reference Genkinger, Wang and Li19). Likewise, the present study found an inverse, but not statistically significant, association between total and dietary intakes of Ca and pancreatic cancer risk (OR (0·72 (95 % CI 0·38, 1·37) when comparing subjects with a median dietary intake of 1300 mg/d with those with a median dietary intake of 541 mg/d).
It is possible that the discrepant findings in the studies discussed above are due to differences in methods of case ascertainment, quality of Ca intake data, between-person variation in Ca intake and control of confounding in those studies. Of note, scarce data on the association between Ca intake and pancreatic cancer are available from Asian populations which have relatively low intake of Ca (e.g. the median intake of Ca was only 328 mg/d for 11 937 Chinese adults residents)(Reference Huang, Wang and Zhang29). Therefore, epidemiological analyses in Asian countries may help us better understand the association between a wide range of Ca intake and the occurrence of pancreatic cancer.
Mg is involved in inflammatory cytokine excretion, immune response, DNA replication and cell cycle regulation(Reference Dai, Sandler and Barry30–Reference Vernon33). Randomised trials have revealed that Mg supplementation optimised circulating vitamin D levels(Reference Dai, Zhu and Manson34). In addition, low intake of Mg has been associated with an elevated risk of diabetes and the metabolic syndrome, which are both risk factors for pancreatic cancer(Reference Long and Romani35). Although the findings of these studies suggest that low Mg intake may also play a role in pancreatic carcinogenesis, few epidemiological studies have investigated this hypothesis. In the Vitamins and Lifestyle Study (VITAL), subjects whose Mg intake was < 75 % of the RDA (420 mg/d for men and 320 mg/d for women) had a significantly increased risk of pancreatic cancer, compared with those who met the RDA for Mg intake(Reference Dibaba, Xun and Yokota17,36) . However, this potential protective effect of Mg on pancreatic cancer was not found in the European Prospective Investigation into Cancer and Nutrition study(Reference Molina-Montes, Wark and Sánchez20) and the US male Health Professionals Follow-up Study(Reference Kesavan, Giovannucci and Fuchs37). The present study showed inverse but non-significant associations between total and dietary intakes of Mg with pancreatic cancer risk after adjustment for confounders, which is largely consistent with findings from the European Prospective Investigation into Cancer and Nutrition and the Health Professionals Follow-up Study.
It has been reported that mice fed a diet high in phosphorus exhibited an increased number and size of carcinogen-induced lung epithelial tumours(Reference Anderson38). The underlying mechanisms for this promoting effect could be that elevated intracellular phosphorus modulates active phosphorylated protein kinase B that stimulates cell cycle progression and other cellular events(Reference Anderson38). Although cigarette smoking that emitted diverse carcinogens is a risk factor shared by lung cancer and pancreatic cancer(Reference Meyer, Rohrmann and Bopp39), it remains unclear whether an increased risk of chemically induced lung tumour associated with high phosphorus intake in animals can be observed for pancreatic cancer in humans. To our knowledge, only one epidemiological study has investigated the association between phosphorus intake and pancreatic cancer risk(Reference Bravi, Polesel and Bosetti21). In that Italian population-based case-control study, phosphorus intake was not associated with an altered risk of pancreatic cancer(Reference Bravi, Polesel and Bosetti21), which is in agreement with the results of the present study.
A major advantage of the present study is that a rapid case-ascertainment system was used to recruit all cases, which was necessary to maximise case enrollment due to the rapidly fatal nature of pancreatic cancer. As a result, all cases were interviewed in person and no proxy interviews, which are prone to recall bias, were used. To enhance the validity of dietary intake data collected from the FFQ, food models were provided to participants to help them estimate serving sizes for foods they consumed(Reference Marley, Fan and Hoyt40).
There are some limitations in our study. The response rates were less than 60 % for both cases and controls. Although the case response rate is relatively high among population-based, case-control studies of pancreatic cancer that do not rely on proxy interviews(Reference Baghurst, McMichael and Slavotinek41,Reference Nkondjock, Krewski and Johnson42) , the generalisability of our results might have been limited as subjects who agreed to participate in the study may be different from those who declined with regard to demographic, socio-economic and lifestyle factors. In addition, lack of complete dietary and alcohol intake data from some individuals who participated in the study reduced the number of cases and controls included in the present analysis and thus the power. Of note, however, our analysis showed that there were no significant differences in age, sex, race, education, smoking status, alcohol intake and physical activity between all subjects considered in the present analysis and most subjects excluded from the analysis (69·4 % of excluded cases and 56·7 % of excluded controls) due to lack of data on dietary and alcohol intake. A sample size of 150 cases and 459 controls may not offer adequate power for us to detect the potential moderate associations between intakes of minerals examined and risk of pancreatic cancer. Dietary measurement error, arising from intentional or unintentional misreporting of individual food intake, might have led to misclassification of some subjects with regard to intake of the nutrients examined and consequently attenuated the risk estimates if such measurement error were non-differential and substantial.
Reverse causality should be considered in any case-control studies of diet and cancer as patients may change their dietary habits in response to clinical symptoms and medical treatments after diagnosis. Although we assessed diet history for the period prior to diagnosis to avoid this bias, the illness may affect recall as well. Overweight and obesity have been linked to pancreatic cancer(Reference Larsson43), but we were unable to adjust for BMI in our analysis because body height and weight were not measured due to an oversight. This limitation was in part overcome by adjustment for energy intake and physical activity, the two main determinants of BMI(Reference Marley, Fan and Hoyt40). Overweight and obesity are associated with an increased risk of diabetes(Reference DeFronzo, Ferrannini and Groop44). In a sensitivity analysis, we adjusted for diabetes in regression models, with the exclusion of nine patients with diabetes diagnosed within 2 years of cancer diagnosis to minimise the possibility of diabetes induced by subclinical pancreatic tumour. This sensitivity analysis revealed that an additional adjustment for diabetes did not materially alter our original results. Nevertheless, failure to adjust for BMI might have distorted our findings. This case-control study was conducted 22–24 years ago. As there is still no screening test available for pancreatic cancer, most cases continue to be diagnosed at late stages. Survival from pancreatic cancer has improved slightly over the period since the present study was conducted primarily due to improved treatment. For example, 1-year relative survival was 18·5 % in 1990–1994 and 37·3 % in 2016(Reference Howlader, Noone and Krapcho45). Since our study was focused on aetiology and the cases in our study were recruited very soon after diagnosis, the trends in survival are unlikely to have affected the relevance of our findings.
In conclusion, there were non-significant associations between total and dietary intakes of Ca, Mg and phosphorus, and the risk of pancreatic cancer in this Upper Midwestern population of the USA. More epidemiological studies are warranted to evaluate whether Ca, Mg and phosphorus confer an altered risk of pancreatic cancer in populations with a relatively low intake of these minerals (e.g. Eastern Asian populations). As dietary intake of energy and nutrients is subject to the measurement error derived from recall bias, urinary biomarkers of Ca, Mg and phosphorus should be considered in future studies(Reference Sun, Bertrand and Franke46). A clear understanding of the roles of these nutrients in pancreatic cancer aetiology may offer innovative practical avenues for its primary prevention.
Acknowledgements
Research reported in this publication was supported by the National Institutes of Health (K.E.A., grant number R01CA58697). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The National Institutes of Health had no role in the design, analysis or writing of this article.
K. E. A. and J. Z. designed the research; H. F., Y. Y., K. E. A. and J. Z. conducted the research; H. F., Y. Y. and J. Z. performed statistical analysis; all authors drafted and/or revised it critically for important intellectual content; J. Z. is responsible for final content of manuscript. All authors read and approved the final manuscript.
The authors declare that they have no conflict of interest.
The present study has been approved by the institutional review boards of the University of Minnesota and the Mayo Clinic and has therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. All study subjects provided written informed consents prior to their inclusion in the study.
Appendix
Table A1. Characteristics of subjects in a population-based, case–control study of pancreatic cancer in Minnesota, stratified by quartiles of total Ca intake, total Mg intake and total phosphorus intake, 1994–1998*
Table A2. Risk of pancreatic cancer in relation to total and dietary intakes of Ca, Mg and phosphorus, stratified by the median intake of total fat, in a population-based, case–control study of pancreatic cancer in Minnesota, 1994–1998
