A Prospective Study of Magnesium and Iron Intake and Pancreatic Cancer in Men

Abstract

Many studies have investigated the relation between magnesium and iron intake and diabetes and, separately, between diabetes and pancreatic cancer. However, no known study has examined the direct association of magnesium and iron intake with pancreatic cancer risk. The authors obtained magnesium and iron intake data using food frequency questionnaires from the US male Health Professionals Follow-up Study, which began in 1986. During 851,476 person-years and 20 years of follow-up, 300 pancreatic cancer cases were documented. Cox proportional hazards models were used to estimate relative risks, adjusting for age, smoking, and body mass index. No associations were observed between magnesium or iron intake and pancreatic cancer (highest vs. lowest quintile: relative risk (RR) = 0.94, 95% confidence interval (CI): 0.66, 1.32 and RR = 0.93, 95% CI: 0.65, 1.34, respectively). Similarly, iron or magnesium supplement use was not related to pancreatic cancer. A statistically significant inverse relation was noted between magnesium and pancreatic cancer for subjects with a body mass index of ≥25 kg/m² (RR = 0.67, 95% CI: 0.46, 0.99; P-trend = 0.04). Although, overall, no relation between magnesium or iron intake and pancreatic cancer was observed in this cohort of men, an inverse association with magnesium was suggested among overweight individuals, which should be examined in other studies.

Pancreatic cancer is the fourth leading cause of cancer-related mortality in the United States, and it is estimated that approximately 35,240 Americans will die of the disease in 2009 (1). Pancreatic cancer is rapidly fatal, with little effective treatment and a 5-year survival rate of 5% (1). Although various diet and lifestyle factors have been studied in association with pancreatic cancer risk, few risk factors other than cigarette smoking, diabetes, and obesity have been consistently and significantly associated with the disease. Identification of modifiable risk factors for the disease may be crucial to reducing pancreatic cancer morbidity and mortality.

Conditions that may lead to high insulin levels in pancreatic cancer secretions have been shown to be related to pancreatic cancer risk (2). In observational studies, type 2 diabetes has been consistently associated with an elevated risk of pancreatic cancer, and current evidence supports a role for glucose intolerance, insulin resistance, and hyperinsulinemia in the etiology of pancreatic cancer. Previous cohort studies with long follow-up periods have consistently found an association between elevated postload or fasting glucose levels and higher pancreatic cancer risk (3–6). Given that recent studies on pancreatic cancer suggest that glucose intolerance and insulin resistance may play a role in carcinogenesis, dietary factors related to insulin resistance, such as those that may improve insulin sensitivity, may have an impact on pancreatic cancer risk.

Several studies have examined the relation between dietary micronutrient intake and the risk of type 2 diabetes. Current research indicates that magnesium is a critical cofactor for several enzymes involved in glucose metabolism and is hypothesized to be involved in glucose homeostasis, insulin action, and development of type 2 diabetes (7–9). Many patients with type 2 diabetes exhibit hypomagnesemia (10). Although it is established that diabetes can induce hypomagnesemia, several studies have shown that magnesium deficiency has also been found to be a risk factor for type 2 diabetes (11–14). Insulin secretion, binding, and action are among the mechanisms that have been suggested to explain the effect of intracellular magnesium on diabetes pathogenesis (12). In the same cohort of US male health professionals examined in the present analysis, an inverse relation between magnesium intake and diabetes was observed (12). A recent meta-analysis of cohort studies supports a statistically significant inverse association between magnesium intake and the risk of type 2 diabetes (15). The overall estimate indicated a 15% reduction in risk of type 2 diabetes for a 100-mg/day increase in magnesium intake (15). Although no known studies have examined the association between magnesium intake and pancreatic cancer risk, studies have shown an inverse relation between magnesium intake and other cancers, including a recent cohort study finding that magnesium intake was inversely associated with colorectal cancer risk (16).

Studies have also suggested a possible role for iron in insulin resistance and diabetes (17). Type 2 diabetes is a common manifestation of hemochromatosis, a genetic disease leading to overabsorption of iron (18), which has generated the hypothesis that high iron stores may increase type 2 diabetes risk. Iron is a potential catalyst in several cellular reactions that lead to the production of reactive oxygen species. These reactions may cause tissue damage and increase oxidative stress, which may affect the risk of type 2 diabetes (19). A prospective study of dietary iron intake and type 2 diabetes found that heme iron intake was positively associated with type 2 diabetes risk, although no association was found for total iron or dietary iron intake (20).

An exploratory case-control study found that higher levels of serum iron were associated with pancreatic cancer risk (21); thus far, no study is known to have examined dietary intake of magnesium and iron in relation to pancreatic cancer risk. The primary aim of this study was to assess magnesium and iron intake and the risk of pancreatic cancer in a prospective study of male US health professionals with 20 years of follow-up.

MATERIALS AND METHODS

Study population

Detailed characteristics of the Health Professionals Follow-up Study have been described elsewhere (22). The study began in 1986, when 51,529 US health professionals, aged 40–75 years, provided detailed information on personal characteristics related to lifestyle and medical history, such as age, weight, height, smoking status, and physical activity. Information for this cohort was obtained at baseline and subsequently via biennial questionnaires. The response rate to these follow-up questionnaires has exceeded 90%.

Dietary assessment

To assess dietary intake, the 1986 baseline questionnaire and follow-up questionnaires included a semiquantitative food frequency questionnaire (23). On the 131-item questionnaire, participants responded to questions assessing their average frequency of intake of specific foods, based on serving size, over the previous year. Calculation of specific nutrients was determined by multiplying the frequency of each food reported by the participant by the nutrient content of the specific portion size consumed. The food composition values for magnesium and iron were based mainly on data from the US Department of Agriculture, supplemented by other publications and manufacturers’ data (24). To determine the total intake of magnesium and iron, intakes of those nutrients from all food sources were summed.

In addition to reporting intake of specific foods, participants also reported their intake of multivitamins and their use of magnesium and iron supplements. Total intake of magnesium and iron was calculated by summing the amount obtained from both diet and supplements. Nutrient intake was adjusted for total energy intake via the residual approach (25). For these analyses, categories of total magnesium and iron intake were determined by dividing each nutrient into quintiles of intake. In the primary analyses, we assessed total magnesium and iron intake at baseline. In secondary analyses, we updated nutrient intake every 2 years to assess the impact of cumulative intake of magnesium and iron over time, which reduces the measurement error that may result from a single baseline measure. We also assessed dietary intake of these nutrients (not including supplemental intake) and magnesium and iron supplement intake alone (ever vs. never users). In addition, we assessed heme iron intake because it has been shown to be associated with an elevated risk of insulin resistance (26).

The validity of the food frequency questionnaire in this cohort, compared with 2 one-week diet records, was evaluated in a random sample of 127 men from the Health Professionals Follow-up Study cohort and has been reported elsewhere (22, 23). The Pearson correlation coefficient between food frequency questionnaire and dietary record for magnesium intake was 0.66 after within-person variation was taken into account. The Pearson correlation coefficient between food frequency questionnaire and dietary record for iron intake was 0.54 after within-person variation was taken into account (22).

Pancreatic cancer ascertainment

Deaths in the Health Professionals Follow-up Study cohort were reported mainly by family members or the postal service via follow-up questionnaires. For nonrespondents, the National Death Index was searched, and this method has been previously shown to be highly sensitive (27).

In addition to reporting of medical history on the baseline questionnaires, follow-up questionnaires asked participants to report on specific medical conditions diagnosed within the 2-year period since the previous follow-up questionnaire. If participants or a family member reported a diagnosis of pancreatic cancer, we attempted to obtain medical records or pathology reports; when doing so was not possible, we attempted to follow up to confirm the self-reported cancer with the participant. We also attempted to confirm via medical records any diagnosis of pancreatic cancer reported as the primary cause of death on a death certificate. Approximately 95% of cases of pancreatic cancer were confirmed by medical records (all had complete information on histology), and the remaining cases were confirmed by death certificate, a physician, or a family member. All histologically confirmed cases were exocrine pancreatic cancer cases.

In these analyses, participants were excluded if they were diagnosed with cancer (other than nonmelanoma skin cancer) before baseline. After exclusions, 47,893 participants followed over 20 years (1986–2006) were eligible for analysis, and 300 incident cases of pancreatic cancer were documented.

Measurement of nondietary covariates

Data on lifestyle factors were obtained at baseline and in all follow-up questionnaires. Participants reported current smoking status, intensity of smoking (average number of cigarettes/day), and history of smoking. For past smokers and those who quit during follow-up, time since quitting was calculated. Age and smoking variables were updated every 2 years during follow-up, and the following categories of smoking were created based on previous analyses in the Health Professionals Follow-up Study cohort (28): never smoker, quit 15 years ago or more, quit less than 15 years ago and smoked 25 pack-years or less in the past 15 years, quit less than 15 years ago and smoked more than 25 pack-years in the past 15 years, current smoker with 25 pack-years or less in the past 15 years, and current smoker with more than 25 pack-years in the past 15 years. Body mass index (BMI) was calculated from self-reported height and weight as weight in kilograms divided by height in meters squared. The correlation coefficient between self-reported weight and measured weight was 0.96 (29). Physical activity (in metabolic equivalents × hours/week) was estimated on the basis of the reported time spent engaged in various activities, weighted by intensity level. The validity of self-reported physical activity has been reported previously (30). Participants also reported medical history at baseline and during follow-up, including history of diabetes.

Statistical analyses

We computed person-time of follow-up for each participant from the return date of the baseline questionnaire in 1986 to the date of pancreatic cancer diagnosis, death, or end of follow-up (January 31, 2006), whichever occurred first. In the primary analysis, participants were divided into quintiles of total magnesium intake and quintiles of total iron intake. To calculate incidence rates of pancreatic cancer, we divided the number of incident cases in each quintile of intake by the number of person-years in that quintile. Relative risks were computed for each category of intake by dividing the incidence rates in each of the 4 upper quintiles by the rate in the lowest quintile (reference group).

In secondary analyses, for a more precise measure of long-term intake, we calculated cumulative average intake of magnesium from food frequency questionnaire data available up to the beginning of each 2-year follow-up period (31). For example, to model pancreatic cancer incidence in the 1988–1990 period, we used magnesium intake reported in 1986; for the 1990–1992 period, we used the average of the 1986 and 1990 intakes, and so forth. Evaluating cumulative average intake reduces within-person variation and characterizes effects of magnesium intake over time.

We used Cox proportional hazards models stratified by age and time period for all analyses to estimate relative risks. In multivariate analyses, we adjusted for potential confounders including age, BMI (quintiles at baseline), height, history of diabetes, physical activity (quintiles at baseline), smoking history, and total caloric intake. We also performed stratified analyses according to levels of BMI, age, and history of diabetes.

All P values were 2-sided. We performed tests for trend using the median value for each quintile of magnesium and iron intake and analyzing it as a continuous variable in the models. All statistical analyses were performed with SAS software (SAS Institute, Inc., Cary, North Carolina).

RESULTS

Table 1 shows the characteristics of the study population according to quintiles of magnesium and iron intake at baseline. Men in the highest quintile of total magnesium intake were more likely to have an increased intake of iron, vitamin D, and multivitamins and a lower intake of saturated fat. Participants with a higher intake of magnesium were also slightly older, more likely to have a history of diabetes and be more physically active than those in the lower quintiles, and they were less likely to smoke. Men in the highest quintile of total iron intake were slightly older and were less likely to smoke and more likely to exercise and take multivitamin supplements than were men in the lowest quintile. Higher total iron intake was associated with lower intake of saturated fat and higher glycemic load and was associated with higher intake of magnesium and vitamin D.

Table 1.

Baseline Characteristics of Health Professionals Follow-up Study Participants According to Quintiles of Magnesium and Iron Intake, 1986

	Quintile of Magnesium Intake			Quintile of Iron Intake
	1 (Low)	3	5 (High)	1 (Low)	3	5 (High)
Mean intake (range), mg/day	257 (102–288)	343 (325–361)	479 (412–1,593)	10.5 (5.5–11.8)	14.4 (13.5–15.6)	42.4 (24–266.9)
Age, years	53.1	54.4	55.9	53.6	54.2	55.6
BMI, kg/m2	25.2	25.1	24.5	25.0	25.1	24.6
Physical activity, MET-hours/week	16.0	20.9	26.2	19.0	20.9	22.6
History of diabetes, %	1.8	3.1	4.6	2.1	3.5	3.3
History of cholecystectomy, %	3.2	3.0	2.9	2.6	2.9	3.3
Current smoker, %	12.6	9.3	6.9	13.8	8.8	7.9
Smoking history, %
Never	46.9	44.1	43.6	41.6	45.5	45.3
<10 pack-years	8.7	9.8	10.7	8.9	9.4	10.7
10–24 pack-years	16.2	18.1	19.3	18.2	18.6	17.9
25–44 pack-years	13.7	14.4	13.6	16.0	14.0	13.2
≥45 pack-years	8.8	7.6	7.1	9.4	7.2	7.5
Multivitamin use, %	27.7	38.3	64.1	26.1	31.5	79.1
Iron intake, mg	14	18	30
Magnesium intake, mg				306	346	417
NSAID use, %	31.8	34.8	39.2	33.0	34.2	41.1
Total calories, kcal	1,953	2,020	1,961	1,835	2,075	1,947
Saturated fat intake, g	27	25	21	26	24.4	23
Glycemic load	123	122	130	119	124	129
Vitamin D intake, mg	272	382	608	325	338	662

Abbreviations: BMI, body mass index; MET, metabolic equivalent task; NSAID, nonsteroidal antiinflammatory drug.

In this cohort, a modest, nonsignificant inverse relation was observed between total magnesium intake and pancreatic cancer risk when higher quintiles of intake were compared with the lowest quintile (Table 2). No dose-response relation was observed, and the test for trend was not statistically significant (highest quintile vs. lowest quintile: age-adjusted relative risk (RR) = 0.86, 95% confidence interval (CI): 0.61, 1.21; P-trend = 0.28). Results were similar, but slightly attenuated after adjustment for potential confounders including BMI, height, physical activity, history of diabetes, smoking, and total energy intake (multivariate RR = 0.94, 95% CI: 0.66, 1.32; P-trend = 0.52). Adding total iron intake to the model did not significantly alter the results. Given the unusual finding that individuals with the highest intake of magnesium are also more likely to have a history of diabetes, we conducted analyses excluding those with a history of diabetes at baseline and observed similar results for nondiabetics at baseline (RR = 0.93, 95% CI: 0.63, 1.38; P-trend = 0.51). We had limited power to observe an association among those subjects with a history of diabetes. In age-adjusted and multivariate analyses for dietary magnesium (excluding magnesium intake from supplemental sources), we found no significant association with pancreatic cancer risk (multivariate RR = 0.95, 95% CI: 0.68, 1.34; P-trend = 0.66). We also assessed the association between magnesium supplement use and risk of pancreatic cancer and found no significant association for ever versus never users (multivariate RR = 0.98, 95% CI: 0.54, 1.76).

Table 2.

Relative Risks of Pancreatic Cancer According to Quintiles of Magnesium and Iron Intake, Health Professionals Follow-up Study, 1986–2006

	Quintile of Intake					P Trend
	1 (Low)	2	3	4	5
Total magnesium
Median intake, mg/day	263	307	343	384	457
No. of cases	66	58	58	47	71
No. of person-years	88,821	89,423	88,920	88,352	88,029
Age adjusted
RR	1.0	0.84	0.79	0.60	0.86	0.28
95% CI		0.59, 1.19	0.55, 1.13	0.41, 0.88	0.61, 1.21
Multivariate
RRa	1.0	0.87	0.84	0.64	0.94	0.52
95% CI		0.61, 1.25	0.59, 1.20	0.43, 0.93	0.66, 1.32
Dietary magnesium
Median intake, mg/day	260	302	335	371	432
No. of cases	68	53	62	42	75
No. of person-years	88,675	89,029	89,042	88,501	88,298
Age adjusted
RR	1.0	0.73	0.81	0.53	0.98	0.37
95% CI		0.51, 1.05	0.57, 1.14	0.36, 0.78	0.63, 1.23
Multivariate
RRa	1.0	0.75	0.86	0.56	0.95	0.66
95% CI		0.52, 1.08	0.60, 1.21	0.37, 0.82	0.68, 1.34
Total iron
Median intake, mg/day	10.8	12.6	14.4	17.8	35.9
No. of cases	63	52	63	62	60
No. of person-years	89,451	88,896	88,828	88,622	87,748
Age adjusted
RR	1.0	0.79	0.98	0.96	0.86	0.64
95% CI		0.55, 1.14	0.69, 1.39	0.67, 1.36	0.60, 1.23
Multivariate
RRa	1.0	0.84	1.08	1.08	0.93	0.82
95% CI		0.58, 1.21	0.75, 1.54	0.75, 1.55	0.65, 1.34
Dietary iron
Median intake, mg/day	10.6	12.2	13.5	15.2	19.9
No. of cases	58	66	50	61	65
No. of person-years	88,959	89,238	88,766	88,487	88,095
Age adjusted
RR	1.0	1.08	0.80	0.99	1.02	0.87
95% CI		0.76, 1.54	0.54, 1.17	0.69, 1.43	0.72, 1.47
Multivariate
RRa	1.0	1.14	0.84	1.08	1.13	0.52
95% CI		0.80, 1.63	0.57, 1.24	0.75, 1.55	0.78, 1.62
Heme iron
Median intake, mg/day	0.70	1.00	1.30	1.50	2.00
No. of cases	54	68	52	62	64
No. of person-years	88,512	89,109	89,057	89,002	87,865
Age adjusted
RR	1.0	1.30	1.00	1.18	1.29	0.33
95% CI		0.91, 1.87	0.68, 1.47	0.81, 1.71	0.89, 1.86
Multivariate
RRa	1.0	1.28	0.97	1.14	1.18	0.63
95% CI		0.89, 1.84	0.66, 1.43	0.78, 1.66	0.81, 1.72

Abbreviations: CI, confidence interval; RR, relative risk.

^a Adjusted for age, body mass index, height, history of diabetes (yes/no), physical activity (quintiles of metabolic equivalent task-hours/week), smoking history (categories), total caloric intake (quintiles).

Since the effect of dietary factors on insulin response may differ by BMI, we stratified the analyses by this variable to examine whether total magnesium intake may have a differential effect based on BMI (Table 3). We would expect diet to have a greater effect on insulin response in overweight subjects compared with those of a lower BMI because they will have lower insulin sensitivity. When we compared the highest with the lowest tertile of intake in multivariate models, we found a significant inverse relation between total magnesium intake and pancreatic cancer risk for subjects with a BMI of ≥25 kg/m² (RR = 0.67, 95% CI: 0.46, 0.99; P-trend = 0.04). We also performed analyses stratified by age (Table 3). For subjects aged 65 years or older, we did not find a significant association between total magnesium intake and pancreatic cancer risk (RR = 0.77, 95% CI: 0.56, 1.06; P-trend = 0.16).

Table 3.

Multivariate Relative Risks of Pancreatic Cancer According to Tertiles of Total Magnesium and Iron Intakes Stratified by BMI and Age, Health Professionals Follow-up Study, 1986–2006^a

	Tertile of Intake			P Trend
	1 (Low)	2	3
Total magnesium
Median intake, mg/day	281	343	423
BMI, kg/m2
<25
No. of cases	36	35	51
No. of person-years	62,775	65,068	73,732
Multivariate
RRb	1.0	0.88	1.09	0.60
95% CI		0.54, 1.43	0.70, 1.71
≥25
No. of cases	70	50	46
No. of person-years	82,565	78,785	69,779
Multivariate
RRb	1.0	0.74	0.67	0.04
95% CI		0.51, 1.08	0.46, 0.99
Age, years
<65
No. of cases	27	29	25
No. of person-years	96,946	86,888	81,571
Multivariate
RRb	1.0	1.17	1.12	0.70
95% CI		0.69, 2.00	0.64, 1.98
≥65
No. of cases	82	58	79
No. of person-years	52,075	58,489	65,779
Multivariate
RRb	1.0	0.67	0.77	0.16
95% CI		0.48, 0.94	0.56, 1.06
Total iron
Median intake, mg/day	11.5	14.4	27.1
BMI, kg/m2
<25
No. of cases	36	42	44
No. of person-years	65,965	63,194	72,416
Multivariate
RRb	1.0	1.26	1.16	0.82
95% CI		0.79, 2.01	0.73, 1.83
≥25
No. of cases	55	53	58
No. of person-years	81,903	77,815	71,411
Multivariate
RRb	1.0	1.04	1.21	0.29
95% CI		0.70, 1.53	0.83, 1.78
Age, years
<65
No. of cases	30	19	32
No. of person-years	94,808	88,184	84,213
Multivariate
RRb	1.0	0.75	1.37	0.08
95% CI		0.42, 1.35	0.82, 2.28
≥65
No. of cases	65	77	77
No. of person-years	56,737	56,309	63,294
Multivariate
RRb	1.0	1.25	1.15	0.78
95% CI		0.89, 1.76	0.82, 1.61

Abbreviations: BMI, body mass index; CI, confidence interval; RR, relative risk.

^aP values for test for interactions: P_{magnesium,BMI} = 0.14; P_{magnesium,age} = 0.25; P_iron,BMI = 0.66; P_iron,age = 0.70.

^bAdjusted for age, BMI, height, history of diabetes (yes/no), physical activity (quintiles of metabolic equivalent task-hours/week), smoking history (categories), total caloric intake (quintiles).

We observed no significant association between total iron intake and pancreatic cancer when comparing the upper quintiles of intake with the lowest quintile (Table 2). No dose-response relation was observed, and the test for trend was not statistically significant (highest quintile vs. lowest quintile: RR = 0.86, 95% CI: 0.60, 1.23; P-trend = 0.64). Results were similar after adjustment for potential confounding variables (RR = 0.93, 95% CI: 0.65, 1.34; P-trend = 0.82). We also observed similar results for nondiabetics at baseline (RR = 0.88, 95% CI: 0.58, 1.33; P-trend = 0.72). In age-adjusted and multivariate analyses for dietary iron intake (excluding iron intake from supplemental sources), we found a nonsignificant increase in risk of pancreatic cancer when comparing the highest quintiles of intake with the lowest quintile (multivariate RR = 1.13, 95% CI: 0.78, 1.62; P-trend = 0.52).

We also assessed the association with heme iron intake because previous research has shown a significant association between heme iron intake and diabetes risk (3). We found a nonsignificant increase in risk across higher quintiles of intakes compared with the lowest quintile, and relative risks were attenuated after adjustment for potential confounding variables (highest vs. lowest quintile: multivariate RR = 1.18, 95% CI: 0.81, 1.72; P-trend = 0.63). We also assessed the association between iron supplement use and risk of pancreatic cancer and found no significant association for ever versus never users (multivariate RR = 1.10, 95% CI: 0.77, 1.57).

We stratified the analyses by age and BMI to examine whether total iron intake may have a differential effect based on these risk factors. A borderline significant trend was observed for total iron intake in the younger age group (RR = 1.37, 95% CI: 0.82, 2.28; P-trend = 0.08). No significant trends were identified when stratifying by BMI.

We conducted a secondary analysis using cumulative updating of dietary exposure from follow-up questionnaires. Results were similar to those obtained by using baseline data (data not shown).

DISCUSSION

In this prospective cohort study, we did not observe any overall associations between higher intakes of magnesium or iron and the risk of pancreatic cancer. However, we observed an inverse association for magnesium intake and pancreatic cancer risk that was statistically significant for overweight men. This finding was similar to one from a previous study on colorectal cancer that reported a stronger association between magnesium intake and colorectal cancer among those who were overweight (16). Nonetheless, this finding may be due to chance and will need to be examined in other cohort studies.

To our knowledge, no previous study has examined the relation between dietary intake of iron or magnesium and pancreatic cancer risk. However, several studies have examined the association of these micronutrients with risk of other cancers, including 3 cohort studies on colorectal cancer that reported inverse associations with magnesium intake (16, 32, 33). The association in one study was stronger among overweight subjects (RR = 0.67, 95% CI: 0.41, 1.08 for colon cancer) (16), suggesting that the effect of magnesium in overweight subjects may be mediated through improved insulin sensitivity. In the 2 other cohort studies, one observed an inverse association between magnesium intake and colorectal cancer (RR = 0.59, 95% CI: 0.40, 0.87 for the highest vs. the lowest quintile) (32), and the other reported an inverse relation for colon cancer (RR = 0.77, 95% CI: 0.58, 1.03 for the highest vs. the lowest quintile) (33). Additional studies on micronutrient intake and cancer risk include those on magnesium in drinking water. One study found that a high magnesium content in drinking water was inversely related to liver cancer (34), and other studies found an inverse relation for breast, prostate, and ovarian cancers, but no association for other cancers (35–38).

Evidence supporting a role for insulin and insulin resistance in the etiology of pancreatic cancer is strong, and several previous epidemiologic studies have examined intake of magnesium in relation to type 2 diabetes (15). Animal studies have shown that a magnesium-deficient diet may cause impaired insulin secretion and action (39) and that supplementing with magnesium lowers the incidence of type 2 diabetes (40). In humans, several studies have suggested that supplemental magnesium intake may have beneficial effects on glucose metabolism and/or insulin sensitivity (41–45). Findings from a recent meta-analysis support a statistically significant inverse association between magnesium intake and risk of type 2 diabetes (15).

There is also evidence for the role of iron in insulin resistance. Iron is involved in the generation of hydroxyl radicals, and the resulting increase in oxidative stress may increase diabetes risk (46). In addition, the increase in oxidative stress may increase oxidative activation of precarcinogens and support tumor cell growth (47). Research has also suggested that high iron levels may obstruct insulin extraction in the liver, leading to peripheral hyperinsulinemia (48, 49). Iron may also be related to insulin resistance since direct iron deposition in β-cells in the pancreas can impair insulin secretion (50).

Intake of micronutrients may also be related to pancreatic carcinogenesis via other mechanisms, such as the role they play in inflammation. It appears that inflammation is involved in pancreatic cancer pathogenesis (51). However, the inflammatory mediators involved in development of pancreatic cancer are not well defined. Recent studies have reported that low serum levels of magnesium are independently related to elevated C-reactive protein concentration (52) and have shown that magnesium intake is inversely associated with systemic inflammation (53, 54). Another study found that magnesium intake from diet is modestly and inversely associated with some, but not all markers of systematic inflammation and endothelial dysfunction in healthy women (55).

Strengths of this study include the prospective design, which precludes recall bias, and detailed collection of dietary data as well as potential risk factors for pancreatic cancer, allowing for adjustment of several potential confounding variables. Furthermore, we updated reports of dietary intake every 4 years in a secondary analysis to minimize misclassification of exposure. Taking into account repeated dietary measurements creates less measurement error and allows the opportunity to account for changes in diet over time (30).

Error due to self-reported dietary intake is possible, and we would expect any measurement error to be random with respect to the outcome. Because measurement error in this case would result in attenuation of the effect estimates, it is possible that null findings may have resulted because of attenuation of the true estimate. It may be warranted for future studies to consider serum measures, as did Friedman and van den Eeden (21), who showed an association between serum iron and pancreatic cancer risk, because such biomarkers may provide more biologically relevant measures of intake than self-reported dietary measures. Although we cannot exclude residual confounding by other factors associated with intake of specific micronutrients, such as dietary variables correlated with magnesium intake, we did not observe a difference in risk in multivariate models adjusting for several dietary variables, including other micronutrients (results not shown).

In summary, although we did not find a significant relation between intake of magnesium or iron and risk of pancreatic cancer, our data suggest a modestly lower risk associated with total magnesium intake among overweight individuals. Further studies are warranted to understand the role of these micronutrients in pancreatic cancer pathogenesis.

Glossary

Abbreviations

BMI

body mass index

CI

confidence interval

RR

relative risk