AJCN. First published ahead of print August 1, 2012 as doi: 10.3945/ajcn.111.030924.

Magnesium intake and colorectal tumor risk: a case-control study and meta-analysis14
Petra A Wark, Rosa Lau, Teresa Norat, and Ellen Kampman
ABSTRACT Background: Dietary magnesium might be related to colorectal tumor risk through the pivotal roles of magnesium in cellular metabolism, insulin resistance, and systemic inammation. Objective: We evaluated the hypothesis of whether higher dietary magnesium intake is associated with reduced colorectal tumor risk. Design: A case-control study on colorectal adenomas (768 cases; 709 polyp-free control subjects) and a meta-analysis of colorectal adenomas (3 case-control studies) and carcinomas (6 prospective cohort studies) were conducted. Dietary magnesium was estimated from food-frequency questionnaires in the case-control study and most studies in the meta-analyses. Data analysis comprised multiple logistic regression analysis (case-control study) and xed- and random-effects meta-analyses. Results: The case-control study showed a nonsignicant inverse association between dietary magnesium intake and risk of colorectal adenomas (OR for every 100-mg/d increase: 0.81; 95% CI: 0.62, 1.06). However, inverse associations were observed only in subjects with BMI (in kg/m2) $25, in subjects aged $55 y, and for advanced adenomas. Associations did not vary by the calcium-to-magnesium intake ratio. In the meta-analysis, every 100-mg/d increase in magnesium intake was associated with 13% lower risk of colorectal adenomas (OR: 0.87; 95% CI: 0.75, 1.00) and 12% lower risk of colorectal cancer (RR: 0.88; 95% CI: 0.81, 0.97). Conclusions: Our ndings support the hypothesis that higher intakes of dietary magnesium are associated with lower risk of colorectal tumors. The consumption of magnesium-rich foods may be a new avenue to explore further in the search for cancer-prevention strategies. Am J Clin Nutr doi: 10.3945/ajcn.111.030924.

growth in later stages (3). Observational studies also suggested possible benecial effects. The rate of colorectal cancer for people in the highest fth of dietary magnesium intake was 41% lower than the rate of colorectal cancer for people in the lowest fth of intake in the Swedish Mammography Cohort (10). Most subsequent prospective cohort studies also reported inverse associations (1114), but they tended to be weaker, were not always signicant, or were visible only in analyses stratied by site, sex, or BMI. Dietary magnesium was not associated with colorectal cancer risk in a prospective study from Germany (15). Most (1620), but not all (21, 22), case-control studies observed lower risks of colorectal cancer in subjects with high magnesium intakes, which was also seen in 3 case-control studies in colorectal adenomas (2325). In one of these studies (25), magnesium intake was most markedly associated with reduced risk of colorectal adenomas in people whose calcium intakes, compared with magnesium intakes, were relatively small. Calcium competes with magnesium for intestinal absorption and transport. We aimed to shed additional light on the role of magnesium intake in colorectal carcinogenesis by studying advanced, nonadvanced, multiple, and single adenomas separately in a casecontrol study in which we also evaluated whether the hypothesized inverse association between dietary magnesium and adenoma risk was stronger in people with low calcium-to-magnesium intake ratios. In addition, we conducted meta-analyses of the association of magnesium intake with risks of colorectal adenomas and carcinomas. Where possible, we stratied analyses by tumor location. In addition, we explored whether the associations depended on sex and BMI.

INTRODUCTION

Magnesium is an essential mineral, which is most notably present in foods rich in dietary ber, nonstarchy vegetables, fruit, nuts, and dairy products. A high consumption of these foods is thought to reduce risk of colorectal cancer (1). Magnesium is required for many physiologic processes that affect colorectal carcinogenesis, including DNA synthesis and repair, glucose metabolism, the regulation of cell proliferation and apoptosis, and defense against oxidative stress (2, 3). Furthermore, magnesium may affect insulin sensitivity and inammatory responses (46), which also inuence colorectal carcinogenesis (7). Indeed, magnesium hydroxide reduced the incidence of chemically induced colorectal cancer in rats (8, 9). Its protective effect, however, might be limited to early stages in tumor development, with some evidence pointing toward the promotion of tumor

1 From the Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom (PAW, RL, and TN), and the Division of Human Nutrition, Wageningen University, Wageningen, Netherlands (EK). 2 The funders had no role in the data analysis, data interpretation, or writing of the article. 3 Supported by the Dutch Cancer Society (grant LUW 96-1374), the Netherlands Organization for Health Research and Development (grant 980-10020), the Netherlands Foundation for Digestive Diseases (grant WS99-72), and the World Cancer Research Fund. 4 Address correspondence to PA Wark, Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, St Marys Campus, Norfolk Place, Paddington, London W2 1PG, United Kingdom. E-mail: p.wark@imperial.ac.uk. Received November 17, 2011. Accepted for publication June 5, 2012. doi: 10.3945/ajcn.111.030924.

Am J Clin Nutr doi: 10.3945/ajcn.111.030924. Printed in USA. 2012 American Society for Nutrition

1 of 10

Copyright (C) 2012 by the American Society for Nutrition

2 of 10
SUBJECTS AND METHODS

WARK ET AL

Case-control study Study population Trained staff from 10 outpatient clinics in the Netherlands recruited cases and control subjects among all eligible people who underwent endoscopy of the large bowel between June 1997 and October 2002. The trained staff either handed eligible people the information package at the time of endoscopy or mailed the package #3 mo thereafter. Cases had to ever have had a histologically conrmed colorectal adenoma, whereas control subjects were never diagnosed with any type of polyp. Potential participants had to be between 18 and 75 y old, be white, be able to speak Dutch, be free of hereditary colorectal cancer syndromes and chronic inammatory bowel diseases, and lack histories of colorectal cancer and (partial) bowel resection. The overall response rate was 54% and ranged from 35% to 91% in the various outpatient clinics. A total of 1526 people gave written informed consent, completed food-frequency and general questionnaires, and agreed to provide a blood sample. Forty-nine people were excluded because their dietary data either were incomplete or resulted in implausible energy intakes. Medical records of the remaining 768 cases and 709 control subjects were reviewed to obtain information on the indication for endoscopy, polyp history, colonoscopy completeness, and adenoma location, size, and number. Ninety-two percent of cases and 85% of control subjects underwent full endoscopy (ie, full colonoscopy or sigmoidoscopy combined with an X-ray). The case-control study was approved by the Medical Review Boards of participating hospitals and Wageningen University. All participants provided written informed consent. Exposure assessment The 178-item food-frequency questionnaire referred to the year before study inclusion or symptom onset (26, 27). Data from the Dutch Food Composition Table (28) were used for conversion to daily energy and nutrient intakes. The questionnaire was suitable for ranking individuals according to most food groups and nutrients, with their 6- and 12-mo reproducibility of the assessment being moderate to high (26, 27). The general questionnaire inquired about demographic, socioeconomic, and lifestyle factors and medical history. It included a question on drugs used at least monthly, which was used to identify takers of magnesium-containing drugs (ie, magnesium hydroxide, carbonate, oxide, peroxide, and sulfate). Statistical analyses Intakes of all nutrients except alcohol were adjusted for total energy intake by obtaining the residuals of linear regression analyses of log-transformed nutrients on log-transformed total energy intake and subsequently adding the log of the nutrient intake of interest at the mean caloric intake (29). The resulting values were exponentiated and subdivided into quantiles according to the distribution in control subjects. Logistic regression was used to evaluate associations of energyadjusted magnesium intake and use of magnesium-containing drugs with colorectal adenoma risk. ORs for different strata were obtained by inclusion of crossproduct terms with quantiles of magnesium intake into regression models. Corresponding

likelihood ratio tests provided additional evidence for or against effect modication. All models on dietary magnesium included age and total energy intake as continuous variables and sex. The following factors were evaluated as potential confounders: BMI; smoking; alcohol consumption; physical activity; education level; family history of colorectal cancer; gastrointestinal complaints; use of nonsteroidal antiinammatory drugs and multivitamins; consumption of fruit, vegetables, and red meat; and dietary intakes of calcium, folate, zinc, ber, riboavin, and vitamins B-6 and B-12. Only dietary folate and vitamin B-6 altered the OR for dietary magnesium by .10% and were included in the nal regression models. To test for a linear trend, the median intake in control subjects within a given exposure category was assigned to the corresponding category and subsequently evaluated as a single ordinal term. Analyses of magnesium-containing drugs were adjusted for age, sex, and bowel complaints. We examined whether the associations depended on the histopathology, number, or location of adenomas by using multinomial logistic regression analyses. Participants with advanced adenomas had either a (tubulo)villous adenoma according to the pathology report or an adenoma $1 cm according to the endoscopy report. Multiple-adenoma patients were dened as patients diagnosed with one or more adenomas at full colonoscopy. Cases with a single adenoma who underwent a full colonoscopy were classied as subjects with distal (between the rectum and the splenic exure) and subjects with proximal lesions. In sensitivity analyses, we restricted the analyses to people who underwent their rst endoscopy; excluded users of the drug furosemide (which may increase the renal loss of magnesium) (30), multivitamin users, and subjects who indicated having made dietary changes because of experiencing bowel complaints; and repeated the analyses by using quintiles instead of tertiles. In additional sensitivity analyses, we explored whether the associations depended on heavy alcohol consumption ($40 g/d for men and $20 g/d for women) or having diabetes because these increase renal excretion of magnesium (30), on high physical activity because of possible losses via sweat (31), and on older age or the presence of bowel complaints because of possible magnesium malabsorption (30). Meta-analysis Study identication and selection We performed a systematic search for publications on magnesium intake from the diet or supplements and risk of colorectal adenomas and carcinomas in PubMed (http://www.ncbi.nlm.nih. gov/pubmed) without any language restriction from 1966 to 31 July 2011 by using the search strategy implemented for the World Cancer Research Fund/American Institute for Cancer Research report (1). Medical subject headings and text words covered a broad range of factors on foods and foods components, physical activity, and anthropometric measures. We also handsearched reference lists from retrieved articles, reviews, and meta-analyses. The complete protocol and full search strategy used is available at http://www.dietandcancerreport.org/cu/ (32). See Supplemental Text 1 under Supplemental data in the online issue for a description of the search terminology. The reporting on the systematic review and meta-analysis followed the preferred reporting items for systematic reviews and meta-analyses statement (33).

MAGNESIUM INTAKE AND COLORECTAL TUMOR RISK

3 of 10

The owchart shown in Figure 1 illustrates the study-selection process. An overview of the 20 original publications from which data were extracted is available online (see Supplemental Tables 13 under Supplemental data in the online issue) and is summarized in the Results. Publications that presented RR estimates and their variances or sufcient data to obtain these effect measures were eligible for inclusion in the meta-analysis. If multiple publications presented ndings on the same study population, only the most recent information was used. The meta-analysis eventually comprised 9 studies, which were inclusive of our case-control study. Statistical analyses We rst conducted a meta-analysis of the dose-response relation of magnesium from diet or supplements with risk of colorectal adenomas and carcinomas. Category-specic risk estimates were transformed into estimates of the RR by the use of an estimation of generalized least-squares for trends with assumption of a linear relation (34). Mean or median exposure values per category level were used directly in the trend estimation when provided; otherwise, the mean exposure level for each category was computed (35). We also performed a meta-analysis in which we compared the highest to lowest categories of intake (36). Fixed- and random-effects models were tted in both types of meta-analyses for which the most adjusted risk estimate from

each study was used. Heterogeneity between studies was quantied with the I2 statistic as a measure of the proportion of total variation in estimates because of heterogeneity. We excluded one study at a time in the sensitivity analyses and prepared funnel plots to assess for small study effects that may have been caused by publication bias. The Stata/SE statistical software package was used for all data analyses (version 9.2; Stata Corp).
RESULTS

Case-control study The study population according to case-control status is described in Table 1. Cases were, on average, older and more likely to be men, have a higher BMI, and have ever smoked and drank more alcohol than control subjects. Cases had higher energy intakes; had higher absolute intakes of fat, folate, and vitamin B-12; and tended to consume more red meat and vegetables than did control subjects. Fewer cases than control subjects underwent endoscopy because of lower gastrointestinal complaints and used magnesium-containing drugs. Lifestyle and medical characteristics of control subjects according to energy-adjusted dietary magnesium intake are described in Table 2. Control subjects in the highest one-third of energy-adjusted intake of dietary magnesium were, on average, older than control subjects with lower intakes. The top one-third

FIGURE 1. Flow diagram of the systematic literature search on magnesium and colorectal tumors. 1See Supplemental Tables 13 under Supplemental data in the online issue for additional description of these studies (and our case-control study). 2The meta-analysis also included the case-control study presented in the current article. Thus, the total number of studies included was 9, 3 studies of which were case-control studies on colorectal adenomas. WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research.

4 of 10

WARK ET AL
TABLE 1 Characteristics of the study population in our case-control study Characteristics Demographics and lifestyle factors Age (y) Men [n (%)] BMI (kg/m2) Low educational level [n (%)]3 Ever smokers [n (%)] High physical activity [n (% in top quintile)]4 Use of multivitamin supplements [n (%)] Medical history [n (%)] Positive family history of colorectal cancer Regular use of NSAIDs5 Bowel complaints or defecation problems as indication of endoscopy Users of drugs containing magnesium6 Dietary factors7 Total energy intake (kJ/d) Protein (g/d) Carbohydrates (g/d) Fat (g/d) Fruit (g/d) Vegetables (g/d) Red meat (g/d) Alcohol (g/d) Dietary calcium (mg/d) Dietary magnesium (mg/d) Dietary ber (g/d) Dietary folate (mg/d) Dietary riboavin (mg/d) Dietary vitamin B-6 (mg/d) Dietary vitamin B-12 (mg/d)
1 2

Adenoma cases (n = 768) 59.1 6 10.12 411 (53.5) 26.1 6 3.8 249 (35.7) 511 (67.0) 132 (18.4) 134 (17.5) 173 (23.5) 204 (26.6) 366 (47.7) 9 (1.17) 8703 6 2484 79.0 6 22.5 226.5 6 69.9 79.3 (62.6, 95.9)8 127.2 (73.7, 247.9) 113.2 (89.6, 142.7) 60.3 (33.4, 82.0) 9.6 (1.0, 24) 1055 (817.2, 1299) 337.7 (288.4, 395.0) 23.4 (19.2, 28.1) 192.8 (161.6, 231.2) 1.53 (1.25, 1.92) 1.59 (1.34, 1.90) 4.43 (3.40, 5.72)

Endoscopy control subjects (n = 709) 51.5 6 13.6 272 (38.4) 25.5 6 4.1 215 (33.0) 389 (55.2) 137 (20.1) 127 (17.9) 136 (20.1) 207 (29.2) 544 (76.7) 26 (3.67) 8434 6 2498 76.9 6 21.6 229.0 6 71.0 75.1 (58.4, 97.2) 124.7 (72.5, 243.9) 109.2 (83.3, 139.1) 53.5 (30.2, 79.0) 4.2 (0.33, 15) 1040 (795.9, 1323) 332.7 (277.5, 391.5) 22.6 (18.8, 27.2) 184.1 (155.6, 219.7) 1.54 (1.21, 1.90) 1.55 (1.30, 1.86) 4.18 (3.16, 5.52)

P1 ,0.001 ,0.001 0.0019 0.30 ,0.001 0.21 0.82 0.12 0.26 ,0.001 0.0016 0.038 0.065 0.50 0.034 0.35 0.025 0.017 ,0.001 0.91 0.15 0.076 0.0010 0.43 0.057 0.0059

Unadjusted P values were based on chi-square, t, or Mann-Whitney U test statistics as appropriate. Mean 6 SD (all such values). 3 Primary school or lower vocational training only. 4 On the basis of a continuous activity score. 5 NSAIDs, nonsteroidal antiinammatory drugs. Use $12 times/y. 6 Magnesium hydroxide, magnesium carbonate, magnesium oxide, magnesium peroxide, or magnesium sulfate. 7 Assessed values (not energy-adjusted or log-transformed). 8 Median; 25th, 75th percentiles in parentheses (all such values).

intake band contained more control subjects with a family history of colorectal cancer than other intake bands. The majority of control subjects underwent (full) endoscopy for the rst time. Only a few control subjects reported the use of furosemide, taking magnesium-containing drugs, or having diabetes, the latter of which was more common in the top one-third intake band. A larger proportion of control subjects in the bottom one-third intake band underwent endoscopy because of bowel complaints. The higher the dietary magnesium intake was, the lower was the risk of colorectal adenomas (Table 3). This particularly applied to risk of advanced and multiple adenomas (Table 3) but did not depend on their location [OR (95% CI) for highest compared with lowest tertile: 0.47 (0.28, 0.79) for proximal and 0.65 (0.46, 0.92) for distal adenomas]. The association between dietary magnesium and adenoma risk did not depend on the calcium-to-magnesium ratio (P-interaction = 0.86) or sex (P-interaction = 0.43) but depended on BMI and age (Table 3). Inverse associations between dietary magnesium intake and adenoma risk were detected only in overweight subjects and in

subjects aged $55 y. A comparable pattern was observed in analyses according to adenoma subtype (results not shown). Sensitivity analyses supported our ndings; the exclusion of subjects who underwent a repeated endoscopy (OR for highest compared with lowest tertile of dietary magnesium intake: 0.68, 95% CI: 0.47, 1.00), takers of magnesium-containing drugs (OR: 0.73; 95% CI: 0.53, 1.00), furosemide users (OR: 0.68; 95% CI: 0.49, 0.93), multivitamin users (OR: 0.70; 95% CI: 0.49, 0.98), and subjects who indicated having changed their diet (OR: 0.61; 95% CI: 0.40, 0.93) did not affect the conclusions. Besides, an inverse, albeit not signicant, association was also shown when quintiles of dietary magnesium intake were studied (OR for highest compared with lowest quintile: 0.72, 95% CI: 0.47, 1.10; P-linear trend = 0.07). These associations did not depend on heavy alcohol consumption, diabetes, high physical activity, or having bowel complaints (P-interaction $ 0.33). The data supported a linear dose-response relation (see Supplemental Figure 1 under Supplemental data in the online issue).

MAGNESIUM INTAKE AND COLORECTAL TUMOR RISK

5 of 10

TABLE 2 Distribution of lifestyle and medical factors according to tertiles of energy-adjusted magnesium intake in endoscopy control subjects in our case-control study Energy-adjusted intake of dietary magnesium in endoscopy control subjects ,315.9 mg/d (n = 236) 49.7 6 14.82 88 (37.3) 25.1 6 3.98 77 (35.5) 146 (62.1) 49 (22.1) 4.08 (0.145, 18.1)5 35 (14.8) 39 (16.5) 32 65 3 208 233 198 (14.6) (27.5) (1.38) (88.1) (98.7) (83.9) 315.9 to ,358.85 mg/d (n = 236) 51.9 6 13.5 98 (41.5) 25.8 6 4.31 76 (34.4) 122 (52.1) 41 (18.0) 4.34 (0.725, 12.5) 22 (9.32) 43 (18.2) 49 71 8 195 231 172 (21.6) (30.1) (3.49) (82.6) (97.9) (72.9) $358.85 mg/d (n = 237) 53.0 6 12.4 86 (36.3) 25.6 6 3.93 62 (29.0) 121 (51.3) 47 (20.4) 4.17 (0.179, 12.9) 21 (8.86) 45 (19.0) 55 71 17 199 233 174 (23.8) (30.0) (7.56) (84.3) (98.3) (73.4)

Characteristics Demographics and lifestyle factors Age (y) Men [n (%)] BMI (kg/m2) Low educational level [n (%)]3 Ever smokers [n (%)] High physical activity [n (% in top quintile)]4 Alcohol consumption (g/d) Heavy alcohol consumption [n (%)]6 Use of multivitamin supplements [n (%)] Medical factors and history [n (%)] Positive family history of colorectal cancer Regular use of NSAIDs7 Self-reported diabetes Full endoscopy8 First endoscopy of the large bowel Bowel complaints or defecation problems as indication of endoscopy Made dietary changes because of bowel complaints9 Takers of drugs containing magnesium10 Takers of furosemide11
1 2

P1 0.027 0.46 0.14 0.31 0.031 0.42 0.80 0.070 0.78 0.038 0.79 0.0040 0.23 0.78 0.0060 0.062 0.59 0.19

104 (45.6) 11 (4.66) 5 (2.12)

111 (47.8) 8 (3.39) 2 (0.85)

130 (56.0) 7 (2.95) 1 (0.42)

Unadjusted P values were based on ANOVA, Kruskal-Wallis, or chi-square test statistics as appropriate. Mean 6 SD (all such values). 3 Primary school or lower vocational training only. 4 On the basis of a continuous activity score. 5 Median; 25th, 75th percentiles in parentheses (all such values). 6 A total of $40 g/d for men and $20 g/d for women. 7 NSAIDs, nonsteroidal antiinammatory drugs. Use $12 times/y. 8 Full colonoscopy or sigmoidoscopy combined with X-ray. 9 Not all changes are likely to have affected dietary magnesium intake. 10 Magnesium hydroxide, magnesium carbonate, magnesium oxide, magnesium peroxide, or magnesium sulfate. 11 This drug may increase renal loss of magnesium (30).

Cases were less likely to have used magnesium-containing drugs than were control subjects (OR: 0.52; 95% CI: 0.23, 1.20), but the point estimate was imprecise and not signicant. Meta-analysis Magnesium and colorectal adenomas The literature search revealed 3 case-control studies on magnesium and risk of colorectal adenomas (2325) but no prospective studies (see Supplemental Table 1 under Supplemental data in the online issue). The only study that made use of population control subjects did not provide cell counts or CIs and could not be included in the meta-analysis (OR for highest compared with lowest quartile: 0.43; P-linear trend = 0.030) (24). The meta-analysis, which, thus, was based on 2 published studies along with our case-control study and included a total of 1703 cases and 2253 control subjects, showed that the higher the intake of magnesium, the lower the risk of colorectal adenomas (OR for every 100-mg/d increase in dietary magnesium intake: 0.87; 95% CI: 0.75, 1.00; Figure 2). A linear doseresponse relation could be assumed (see Supplemental Figure 2 under Supplemental data in the online issue).

One of the included studies observed an inverse (but not signicant) association with dietary magnesium, whereas a signicant and stronger inverse association was seen when magnesium from supplements and magnesium from the diet were studied jointly (25). The use of combined rather than dietary estimates for this study in the meta-analysis increased the strength of the association (OR for every 100-mg/d increase in magnesium intake: 0.80; 95% CI: 0.71, 0.90). The highest compared with lowest comparison also showed that subjects with higher magnesium intakes had lower colorectal adenoma risks (OR: 0.76; 95% CI: 0.60, 0.96; see Supplemental Figure 1 under Supplemental data in the online issue). Magnesium and colorectal cancer Eleven publications on 7 case-control studies (1622, 3740) on magnesium intake and colorectal cancer were identied, and all of them were based on dietary estimates (see Supplemental Table 2 under Supplemental data in the online issue). A metaanalysis was not conducted because only 2 of the 11 publications reported sufcient data to undertake a meta-analysis (16, 22). Three (1620) of the 7 (1620) case-control studies showed inverse associations but only in the smallest of these three

6 of 10

WARK ET AL
TABLE 3 Association between energy-adjusted intake of dietary magnesium (mg/d) and risk of colorectal adenomas in the case-control study1 Comparisons, subgroups, and parameters All adenomas compared with control subjects All2 Cases/control subjects (n) OR (95% CI) BMI (in kg/m2) ,253 Cases/control subjects (n) OR (95% CI) BMI $253 Cases/control subjects (n) OR (95% CI) Age ,55 y4 Cases/control subjects (n) OR (95% CI) Age $55 y4 Cases/control subjects (n) OR (95% CI) Nonadvanced adenomas compared with control subjects All5 Cases/control subjects (n) OR (95% CI) Advanced adenomas compared with control subjects All5 Cases/control subjects (n) OR (95% CI) Single adenoma compared with control subjects All6 Cases/control subjects (n) OR (95% CI) Multiple adenomas compared with control subjects All6 Cases/control subjects (n) OR (95% CI)
1 2

,315.9 mg/d

315.9 to ,358.85 mg/d

$358.85 mg/d

P-linear trend

270/236 Referent 104/134 Referent 161/100 Referent 72/141 Referent 198/95 Referent

267/236 0.90 (0.68, 1.19) 121/109 1.39 (0.93, 2.09) 143/127 0.62 (0.42, 0.90) 100/138 1.16 (0.77, 1.74) 167/98 0.72 (0.50, 1.05)

231/237 0.73 (0.53, 1.00) 89/112 0.87 (0.56, 1.35) 140/124 0.64 (0.43, 0.95) 86/125 1.01 (0.65, 1.56) 145/112 0.56 (0.37, 0.83)

0.050

0.58

0.025

0.98

0.004

89/236 Referent

92/236 0.93 (0.64, 1.35)

92/237 0.92 (0.61, 1.39)

0.69

163/236 Referent

157/236 0.86 (0.62, 1.19)

125/237 0.60 (0.42, 0.87)

0.007

104/208 Referent

111/195 0.96 (0.67, 1.38)

106/199 0.85 (0.57, 1.27)

0.43

144/208 Referent

132/195 0.89 (0.63, 1.26)

103/199 0.69 (0.46, 1.03)

0.072

ORs (95% CIs) were adjusted for age, sex, total energy intake, dietary folate, and vitamin B-6. OR for every 100-mg/d increase in magnesium intake: 0.81 (95% CI: 0.62, 1.06). 3 P-interaction = 0.011. 4 P-interaction = 0.081. 5 Despite not being signicant, the odds of having advanced rather than nonadvanced colorectal adenomas in people in the highest tertile of dietary magnesium intake seemed to be lower than the odds of having advanced rather than nonadvanced colorectal adenomas in people in the lowest tertile of dietary magnesium intake (OR: 0.66; 95% CI: 0.42, 1.02; Plinear trend = 0.064). 6 Applies to subjects who underwent a full endoscopy (700 cases and 602 control subjects). The difference between associations with multiple and single adenomas was not signicant (OR for having multiple rather than single adenomas for the highest compared with the lowest tertile of dietary magnesium: 0.81; 95% CI: 0.52, 1.26; P-linear trend = 0.36).

studies (399 cases and 399 control subjects) (16) was significance reached. The study in 1993 cases and 2410 control subjects showed no evidence for an association (39, 40) and neither did the 3 studies with ,300 cases and control subjects each (21, 22, 38). In addition, 6 prospective cohort studies were identied (1015), which included a total of 5834 cases. Follow-up time ranged from 7.9 to longer than 17 y, and 3 of these studies included only women (see Supplemental Table 3 under Supplemental data in the online issue). The highest compared with lowest category of intake (RR: 0.84; 95% CI: 0.73, 0.97; see Supplemental Figure 1 under Supplemental data in the online issue) as well as the dose-response meta-analysis (Figure 2) showed inverse associations of similar magnitude for colorectal,

colon, and rectum cancer risks. Between-study heterogeneity was observed, but this appeared to be largely due to the Swedish Mammography Cohort (10). Exclusion of the latter study only slightly attenuated the pooled RR [RR for colorectal cancer for every 100-mg/d increase in magnesium intake: 0.90; 95% CI: 0.83, 0.99; I2 = 5.4%, P for heterogeneity = 0.38; compared with RR of 0.88 (Figure 2) when all studies were included]. Doseresponse graphs suggested a linear relation (see Supplemental Figure 2 under Supplemental data in the online issue). Two studies presented analyses according to sex and obesity. The Netherlands Cohort Study on Diet and Cancer reported comparable results for men and women and observed an inverse dose-response relation for colon and proximal cancer in overweight individuals only (13). In contrast, the Japan Public Health

MAGNESIUM INTAKE AND COLORECTAL TUMOR RISK

7 of 10

FIGURE 2. Dose-response meta-analysis of magnesium intake and risk of colorectal adenomas and cancer. The center of the solid boxes represents the sizes of the RR for each study, with the area of these boxes representing the weight that each study contributes to the overall RR. Horizontal lines indicate the corresponding 95% CIs. The center of the diamonds indicates summary RRs, and the left and right extremes represent the corresponding CIs. *Case-control studies. D+L, DerSimonian and Laird (random-effects) model; IV, prospective cohort studies IV that represent the inverse-varianceweighted (xed-effects) model.

Centerbased Prospective Study only detected associations in men with BMI (in kg/m2) ,25 (14). The literature search did not identify any studies on magnesium-containing drugs.
DISCUSSION

In addition to conducting case-control analyses of colorectal adenomas, we provided a comprehensive overview of the epidemiologic evidence on magnesium and colorectal tumors available. Both our case-control study and the meta-analyses showed inverse associations between magnesium intake and risk of colorectal tumors. In the case-control study, associations with colorectal adenomas were restricted to overweight individuals and subjects aged $55 y and were more clearly visible for advanced and multiple adenomas. The associations did not depend on the calcium-to-magnesium intake ratio. Neither the case-control

study nor the meta-analysis suggested differential associations according to tumor location. The available data have limitations. Our case-control study included more cases than any other identied study on magnesium intake and colorectal adenomas, but overall, the available evidence was limited. Only 3 studies on adenomas and 6 studies on carcinomas have been published to our knowledge. In contrast to the cohort studies on which the meta-analysis of colorectal cancer was based, the meta-analysis of colorectal adenomas included only endoscopy-based case-control studies. Endoscopybased populations may include a proportion of people with chronic diarrhea or intestinal and biliary stulae in whom magnesium absorption is impaired (30, 41). Because the association between dietary magnesium intake and risk of colorectal adenomas was similar in subjects with and without bowel complaints, the presence of this subcategory does not compromise the validity in our

8 of 10

WARK ET AL

case-control study. Recall bias, however, cannot be ruled out in any case-control study. Although we appropriately took into account potential confounders, the possibility of residual confounding can also never be ruled out. The same applies to the other studies included in the meta-analysis. Selection bias is always a potential issue in case-control studies. The overall response rate of our case-control study was 55% and varied between hospitals that had somewhat different recruitment strategies. The available data did not allow us to explore nonresponse bias in detail. However, we reduced the possibility of selection bias by recruiting cases and control subjects among people who underwent endoscopy of the large bowel by using similar recruitment strategies. We chose this group because adenomas are common in the general population and are often only detected by chance because they are mostly asymptomatic (42). Restriction of the analyses to participants who underwent their rst endoscopy, underwent a full endoscopy, or did not report to have made dietary changes because of bowel complaints did not affect our conclusions. Although our choice of study population may have affected external validity, our selection of control subjects from endoscopy patients can, thus, be regarded as an advantage. One of the other casecontrol studies included in the meta-analysis selected cases likewise (25). The majority of studies included in the meta-analysis, which were inclusive of our case-control, assessed magnesium intake through food-frequency questionnaires. Food-frequency questionnaires are suitable for ranking individuals according to magnesium intake, and the reproducibility of its assessment is high (43, 44). Thus, associations of high compared with low magnesium intakes can be investigated despite the likely underestimation of intake compared with that in food records (41). In contrast to the US studies (11, 12), we did not take magnesium intake from supplements into account, but this has unlikely affected our ndings. Supplement use is rare in the Netherlands (13, 45), and the exclusion of multivitamin users did not change the conclusions. However, nondifferential misclassication of magnesium intake might have occurred. Magnesium from drinking water has been associated with colorectal cancer risk (46). However, sourcespecic data on the magnesium content of drinking waters were not available in the case-control study or meta-analysis, and the loss of magnesium during food processing (31) could not be estimated. Because reporting errors also inevitably lead to some degree of measurement error, true associations between magnesium intake and risk of colorectal tumors may be stronger than observed. In the case-control study, we observed that inverse associations between dietary magnesium and colorectal adenomas were limited to people who were overweight, which was also observed in the Netherlands Cohort Study on Diet and Cancer for colon cancer (13) and in the Health Professionals Follow-Up Study for pancreatic cancer (47). This effect modication might have been a result of effects of magnesium on insulin resistance and responses, which are thought to play a role in carcinogenesis (7). Magnesium supplementation reduced insulin resistance and improved insulin sensitivity and b cell function even in patients without diabetes (46). The postulated role of magnesium in the causation of type 2 diabetes (48) and increased risk of colorectal cancer seen in type 2 diabetes patients (49) lend additional support for this mechanism. However, the Japan

Public Health Centerbased Prospective study only observed an inverse association in subjects with BMI ,25 (14). b Cell function is relatively low in lean Asian people (50), and dietary magnesium could, thus, be mostly benecial in this subgroup. Our case-control study and meta-analysis suggested a linear dose-response relation, which was contradicted by only one study in which an inverse association was seen when magnesium intakes from the diet and supplements were studied jointly but not when dietary intakes were considered alone (25). These observations may suggest that the achievement of magnesium homeostasis (30) does not sufce to explain the association between magnesium intake and colorectal tumor risk. However, different individuals may require different dietary intake levels to achieve homeostasis depending on their age, sex, and calcium intake. We did not nd differences according to sex in our studies and we observed an inverse association only in participants $55 y old. This observation requires conrmation because it is hard to explain and may have been by chance: In contrast to the Tennessee Colorectal Polyp Study (25), we did not nd evidence for effect modication by the calcium-to-magnesium intake ratio in the case-control study. Differences in genetic makeup between our study populations might potentially account for this discrepancy. For instance, the Tennessee Colorectal Polyp Study showed that carriers of the Thr1482Ile allele of the transient receptor potential melastatin 7 gene, which plays a role in magnesium homeostasis, were at greater risk of adenomas, particularly if carriers had relatively high calcium compared with magnesium intakes (25). We investigated associations for advanced, nonadvanced, multiple, and single adenomas and carcinomas separately because animal studies suggested that magnesium may be benecial only in the early stages of colorectal carcinogenesis (3). In line with this hypothesis, the meta-analysis showed strikingly similar risk estimates for adenoma and carcinomas. The case-control study, however, pointed toward stronger associations with advanced and multiple than with nonadvanced or single adenomas. This nding warrants replication because it may imply that magnesium might also affect intermediate stages of colorectal carcinogenesis. Whether magnesium-containing drugs, such as some laxatives and antacids, affect colorectal cancer tumor risk remains to be established. In line with a study in rats (8, 9), we observed a nonsignicant reduced risk of colorectal adenomas in takers of magnesium-containing drugs, which are prescribed for hyperacidity, constipation, or severe magnesium deciency (30). Nonetheless, the statistical power was limited, and only some of the control subjects who were prescribed magnesium-containing drugs were free of bowel complaints, which made it hard to separate effects of treatment and indication. Furthermore, the use of laxatives was not associated with colorectal tumor risk in US women (51, 52). In conclusion, we observed that magnesium intake was inversely associated with risk of colorectal adenomas and colorectal cancer. However, the number of studies on this topic is small. In particular, the observation that the association was limited to overweight individuals in the case-control study requires conrmation. Elucidation of pathways that involve dietary magnesium and insulin resistance in colorectal carcinogenesis is also warranted.

MAGNESIUM INTAKE AND COLORECTAL TUMOR RISK

We thank the endoscopy staff and gastroenterologists of the following hospitals in the Netherlands: Slingeland Ziekenhuis (Doetinchem), Ziekenhuis Gelderse Vallei (Ede), Radboud University Nijmegen Medical Centre (Nijmegen), Antonius Ziekenhuis (Nieuwegein), Meander Medisch Centrum (Amersfoort), Ziekenhuis Rijnstate (Arnhem), Ziekenhuis Rivierenland (Tiel), Slotervaart Ziekenhuis (Amsterdam), Jeroen Bosch Ziekenhuis (Den Bosch), and Canisius Wilhelmina Ziekenhuis (Nijmegen). We are also grateful to from the National Institute of Public Health and the EnvironMarga Ocke ment, Bilthoven, Netherlands, for calculation of food and nutrient intakes and to Lorijn van Rooijen for contributing to the literature research. We also thank all of the Wageningen University staff who were involved in the recruitment, data collection, and data management of the case-control study and to all staff who contributed to the World Cancer Research Fund metaanalysis project on colorectal adenomas and carcinomas at Wageningen University and Imperial College London. The authors responsibilities were as followsTN and EK: designed the research and had primary responsibility for the nal content of the manuscript; PAW and RL: conducted the research and analyzed data; PAW: drafted the manuscript; and all authors: provided critical revisions for important intellectual content and read and approved the nal manuscript. None of the authors had a conict of interest.

9 of 10

16.

17.

18.

19.

20.

21.

22. 23.

REFERENCES
1. World Cancer Research Fund/American Institute for Cancer Research. Food, nutrition, physical activity, and the prevention of cancer: a global perspective. 1st ed. Washington DC: AIRC, 2007. 2. Wolf FI, Maier JA, Nasulewicz A, Feillet-Coudray C, Simonacci M, Mazur A, Cittadini A. Magnesium and neoplasia: from carcinogenesis to tumor growth and progression or treatment. Arch Biochem Biophys 2007;458:2432. 3. Hartwig A. Role of magnesium in genomic stability. Mutat Res 2001; 475:11321. 4. Guerrero-Romero F, Tamez-Perez HE, Gonzalez-Gonzalez G, Salinas-Martinez AM, Montes-Villarreal J, Trevino-Ortiz JH, Rodriguez-Moran M. Oral magnesium supplementation improves insulin sensitivity in non-diabetic subjects with insulin resistance. A double-blind placebo-controlled randomized trial. Diabetes Metab 2004;30:2538. 5. Mooren FC, Kruger K, Volker K, Golf SW, Wadepuhl M, Kraus A. Oral magnesium supplementation reduces insulin resistance in nondiabetic subjects - a double-blind, placebo-controlled, randomized trial. Diabetes Obes Metab 2011;13:2814. 6. Guerrero-Romero F, Rodriguez-Moran M. Magnesium improves the beta-cell function to compensate variation of insulin sensitivity: double-blind, randomized clinical trial. Eur J Clin Invest 2011;41: 40510. 7. Giovannucci E. Metabolic syndrome, hyperinsulinemia, and colon cancer: a review. Am J Clin Nutr 2007;86:s83642. 8. Mori H, Morishita Y, Shinoda T, Tanaka T. Preventive effect of magnesium hydroxide on carcinogen-induced large bowel carcinogenesis in rats. Basic Life Sci 1993;61:1118. 9. Tanaka T, Shinoda T, Yoshimi N, Niwa K, Iwata H, Mori H. Inhibitory effect of magnesium hydroxide on methylazoxymethanol acetateinduced large bowel carcinogenesis in male F344 rats. Carcinogenesis 1989;10:6136. 10. Larsson SC, Bergkvist L, Wolk A. Magnesium intake in relation to risk of colorectal cancer in women. JAMA 2005;293:869. 11. Folsom AR, Hong CP. Magnesium intake and reduced risk of colon cancer in a prospective study of women. Am J Epidemiol 2006;163: 2325. 12. Lin J, Cook NR, Lee IM, Manson JE, Buring JE, Zhang SM. Total magnesium intake and colorectal cancer incidence in women. Cancer Epidemiol Biomarkers Prev 2006;15:20069. 13. van den Brandt PA, Smits KM, Goldbohm RA, Weijenberg MP. Magnesium intake and colorectal cancer risk in the Netherlands Cohort Study. Br J Cancer 2007;96:5103. 14. Ma E, Sasazuki S, Inoue M, Iwasaki M, Sawada N, Takachi R, Tsugane S. High dietary intake of magnesium may decrease risk of colorectal cancer in Japanese men. J Nutr 2010;140:77985. 15. Li K, Kaaks R, Linseisen J, Rohrmann S. Dietary calcium and magnesium intake in relation to cancer incidence and mortality in a German

24. 25.

26.

27.

28.

29. 30. 31. 32.

33.

34.

35.

36.

37.

38.

39.

prospective cohort (EPIC-Heidelberg). Cancer Causes Control 2011;22: 137582. e J, Charnay B, Bertheze ` ne P, Macquart-Moulin G, Riboli E, Corne Day N. Case-control study on colorectal cancer and diet in Marseilles. Int J Cancer 1986;38:18391. Tuyns AJ, Haelterman M, Kaaks R. Colorectal cancer and the intake of nutrients: oligosaccharides are a risk factor, fats are not. A case-control study in Belgium. Nutr Cancer 1987;10:18196. Yang G, Butian J, Yutang G, Fan J, Shu Z, Runie G. [Relationship between colorectal cancer and ten inorganic elements.] Zhonghua Yu Fang Yi Xue Za Zhi 1993;27:2825 (in Chinese). Yang G, Gao Y, Ji B. [Comparison of risk factors between left and right-sided colon cancer.] Zhongguo Yi Xue Ke Xue Yuan Xue Bao 1994;16:638 (in Chinese). Yang G, Butian J, Yutang G, Runie G. [Nutritional epidemiology of rectal cancer in a population-based case-control study.] Ying Yang Xue Bao 1993;15:30917 (in Chinese). Benito E, Stiggelbout A, Bosch FX, Obrador A, Kaldor J, Mulet M, Munoz N. Nutritional factors in colorectal cancer risk: a case-control study in Majorca. Int J Cancer 1991;49:1617. Zaridze DG, Filipchenko VV. [The role of nutrition in the etiology of cancer of the large intestine.] Vopr Onkol 1991;37:1528 (in Russian). e J, Kaaks R, Bertheze ` ne P. Macquart-Moulin G, Riboli E, Corne Colorectal polyps and diet: a case-control study in Marseilles. Int J Cancer 1987;40:17988. Benito E, Cabeza E, Moreno V, Obrador A, Bosch FX. Diet and colorectal adenomas: a case-control study in Majorca. Int J Cancer 1993;55:2139. Dai Q, Shrubsole MJ, Ness RM, Schlundt D, Cai Q, Smalley WE, Li M, Shyr Y, Zheng W. The relation of magnesium and calcium intakes and a genetic polymorphism in the magnesium transporter to colorectal neoplasia risk. Am J Clin Nutr 2007;86:74351. MC, Bueno de Mesquita HB, Goddijn HE, Jansen A, Pols MA, Ocke Van Staveren WA, Kromhout D. The Dutch EPIC food frequency questionnaire. I. Description of the questionnaire, and relative validity and reproducibility for food groups. Int J Epidemiol 1997;26(suppl 1): S3748. MC, Bueno de Mesquita HB, Pols MA, Smit HA, Van Staveren WA, Ocke Kromhout D. The Dutch EPIC food frequency questionnaire. II. Relative validity and reproducibility for nutrients. Int J Epidemiol 1997;26 (suppl 1):S4958. Stichting Nederlands Voedingsstoffenbestand. Dutch Food Composition Table. Nevo tabel: Nederlands voedingsstoffenbestand. The Hague, Netherlands: Voorlichtingsbureau voor de Voeding, 1996. Willett W. Nutritional epidemiology. 2nd ed. New York, NY: Oxford University Press, 1998. Rude RK. Magnesium deciency: a cause of heterogeneous disease in humans. J Bone Miner Res 1998;13:74958. Vormann J. Magnesium: nutrition and metabolism. Mol Aspects Med 2003;24:2737. World Cancer Research Fund/American Institute for Cancer Research. Available from: http://www.dietandcancerreport.org/cu (cited 5 June 2012). Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009;6:e1000097. Orsini N, Bellocco R, Greenland S. Generalized least squares for trend estimation of summarized dose-response data. Stata Journal 2006;6: 4057. ne G, Thompson SG. Methods for summarizing the risk associaChe tions of quantitative variables in epidemiologic studies in a consistent form. Am J Epidemiol 1996;144:61021. Egger M, Davey Smith G, Altman DG. Systematic reviews in health care: meta-analysis in context. 2nd ed. London, United Kingdom: BMJ Books, 2001. e J, Bertheze ` ne P, Southgate DA. Macquart-Moulin G, Durbec JP, Corne [Diet and colorectal cancer.] Gastroenterol Clin Biol 1983;7:27786 (in French). Arbman G, Axelson O, Ericsson-Begodzki AB, Fredriksson M, Nilsson E, Sjodahl R. Cereal ber, calcium, and colorectal cancer. Cancer 1992;69:20428. Slattery ML, Potter JD, Coates A, Ma KN, Berry TD, Duncan DM, Caan BJ. Plant foods and colon cancer: an assessment of specic foods and their related nutrients (United States). Cancer Causes Control 1997;8:57590.

10 of 10

WARK ET AL
47. Kesavan Y, Giovannucci E, Fuchs CS, Michaud DS. A prospective study of magnesium and iron intake and pancreatic cancer in men. Am J Epidemiol 2010;171:23341. 48. Dong JY, Xun P, He K, Qin LQ. Magnesium intake and risk of type 2 diabetes: meta-analysis of prospective cohort studies. Diabetes Care 2011;34:211622. 49. Jiang Y, Ben Q, Shen H, Lu W, Zhang Y, Zhu J. Diabetes mellitus and incidence and mortality of colorectal cancer: a systematic review and meta-analysis of cohort studies. Eur J Epidemiol 2011;26:86376. 50. Sakurai M, Miura K, Takamura T, Ishizaki M, Morikawa Y, Nakamura K, Yoshita K, Kido T, Naruse Y, Kaneko S, et al. J-shaped relationship between waist circumference and subsequent risk for Type 2 diabetes: an 8-year follow-up of relatively lean Japanese individuals. Diabet Med 2009;26:7539. 51. Dukas L, Platz EA, Colditz GA, Willet WC, Giovannucci EL. Bowel movement, use of laxatives and risk of colorectal adenomatous polyps among women (United States). Cancer Causes Control 2000; 11:90714. 52. Dukas L, Willett WC, Colditz GA, Fuchs CS, Rosner B, Giovannucci EL. Prospective study of bowel movement, laxative use, and risk of colorectal cancer among women. Am J Epidemiol 2000;151:95864.

40. Kampman E, Slattery ML, Caan B, Potter JD. Calcium, vitamin D, sunshine exposure, dairy products and colon cancer risk (United States). Cancer Causes Control 2000;11:45966. 41. Scientic Committee on Food. Opinion of the Scientic Committee on Food on the tolerable upper intake level of magnesium. Report no. SCF/CS/NUT/UPPLEV/54 Final. Brussels, Belgium: European Commission, 2001. 42. Peipins LA, Sandler RS. Epidemiology of colorectal adenomas. Epidemiol Rev 1994;16:27397. 43. Patterson RE, Kristal AR, Tinker LF, Carter RA, Bolton MP, AgursCollins T. Measurement characteristics of the Womens Health Initiative food frequency questionnaire. Ann Epidemiol 1999;9:17887. 44. Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA, Litin LB, Willett WC. Reproducibility and validity of an expanded selfadministered semiquantitative food frequency questionnaire among male health professionals. Am J Epidemiol 1992;135:111426. MC, Buurma-Rethans EJM, Fransen HP. Dietary supplement 45. Ocke use in the Netherlands. Current data and recommendations for future assessment Bilthoven, Netherlands: RIVM, 2005. 46. World Health Organization. Calcium and magnesium in drinking-water: public health signicance. Geneva, Switzerland: WHO Press, 2009.