European Journal of Clinical Nutrition (2012) 66, 1182–1186; doi:10.1038/ejcn.2012.135; published online 3 October 2012 Magnesium intake and risk of colorectal cancer: a meta-analysis of prospective studies G-C Chen1, Z Pang2 and Q-F Liu1 1Department of Epidemiology, School of Public Health, Soochow University, Suzhou, China 2Department of Gastroenterology, Suzhou Municipal Hospital (North Campus), Suzhou, China Correspondence: Professor Q-F Liu, Department of Epidemiology, School of Public Health, Soochow University, 199 Renai Road, Dushu Lake Higher Education Town, Suzhou 215123, China. E-mail: email@example.com Received 25 May 2012; Revised 7 August 2012; Accepted 3 September 2012 Advance online publication 3 October 2012 Top of page Abstract Epidemiologic studies have suggested that magnesium intake may be associated with a decreased risk of colorectal cancer (CRC), but the findings have been inconsistent. We aimed to assess this association by conducting a meta-analysis of prospective studies. We performed a literature search on PubMed database through July 2012 to identify prospective studies of magnesium intake in relation to CRC risk. Reference lists of the retrieved articles were also reviewed. A random-effects model was used to compute the summary risk estimates. Eight prospective studies containing 338 979 participants and 8000 CRC cases met the inclusion criteria. The summary relative risk (RR) for the highest vs lowest category of magnesium intake for CRC was 0.89 (95% CI, 0.79–1.00), with little evidence of heterogeneity. Restricting the analysis to six studies that have adjusted for calcium intake yielded a similar result. For colon and rectal cancer, the pooled RR was 0.81 (95% CI, 0.70–0.93) and 0.94 (95% CI, 0.72–1.24), respectively. In the dose–response analyses, the summary RRs for an increment of magnesium intake of 50 mg/day for colorectal, colon and rectal cancer were, respectively, 0.95 (95% CI, 0.89–1.00), 0.93 (95% CI, 0.88–0.99) and 0.93 (95% CI, 0.83–1.04), and there was some evidence of heterogeneity; omitting one study that substantially contributed to the heterogeneity yielded generally similar results, but with low heterogeneity. We detected no indication of publication bias. On the basis of the findings of this meta-analysis, a higher magnesium intake seems to be associated with a modest reduction in the risk of CRC, in particular, colon cancer. Keywords: prospective studies; colorectal cancer; magnesium; meta-analysis Top of page Introduction Colorectal cancer (CRC) comprises at least 13% of all cancer patients in Europe, and remains the third most common type of cancer in the United States.1, 2 Diet and lifestyle factors have an important role in the etiology of CRC, and a large proportion of CRC incidence might be prevented by a healthy lifestyle.3, 4 However, a limited number of dietary factors were identified to be associated with the risk of CRC. Magnesium is abundant in many foods and is involved in a wide variety of biochemical reactions that modulate key cell functions, such as proliferation, differentiation, migration and apoptosis.5, 6 It also has a crucial role in genetic stability and DNA synthesis.7 In animal experiments, supplemental magnesium has been demonstrated to reduce the incidence of colon cancer, possibly by means of inhibition of c-myc oncogene expression in the colon cancer cells.8, 9, 10, 11 Intake of magnesium has also been reported to reduce the risk of type 2 diabetes,12 a potential risk factor for CRC.13, 14, 15 In addition, a recently published randomized controlled study reported that magnesium treatment significantly decreased fasting C-peptide concentrations.16 Multiple lines of evidence have suggested that high circulating concentrations of C-peptide, a marker for insulin secretion, are associated with an increased risk of CRC in humans.17, 18 This raised the hypothesis that intake of magnesium may decrease the risk of CRC by improving insulin sensitivity and decreasing insulin levels. Evidence from epidemiologic studies has also suggested that intake of magnesium may be associated with a reduction in the risk of CRC, but the findings have been inconsistent.19, 20, 21, 22, 23, 24, 25, 26 The purpose of this study was to examine the relationship between magnesium intake and the risk of CRC by undertaking a meta-analysis of prospective studies. We attempted to plan, conduct and report this meta-analysis in adherence to the guidelines of the ‘Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group'.27 Materials and methods Search strategy We performed a literature search through July 2012 on PubMed database using the following search terms: magnesium; cancer, colorectal cancer, colorectal neoplasms, colon, rectum; and cohort study, prospective study, follow-up study. No language restrictions were imposed. In addition, we also comprehensively reviewed the reference lists of the retrieved articles to identify additional studies. Study selection Studies were included if they met the following criteria: (i) the study design was prospective cohort or case–cohort or nested case-control; (ii) the exposure of interest was intake of dietary magnesium or total magnesium (dietary and supplements combined); (iii) the outcome of interest was colorectal, colon or rectal cancer and (iv) the relative risk (RR) estimates (or odds ratios (OR) in nested case–control studies) with corresponding 95% confidence interval (CI) were provided, or could be calculated using the raw data presented in the studies. First, the titles and abstracts of the relevant studies were reviewed, and then the full papers of potentially eligible studies were obtained. Data extraction and quality assessment The following data were extracted from each included eligible study using a standardized report form: author name, publication year, length of follow-up, study location, the sex of participants, the age range of participants, number of cases and participants (cases and controls or cohort size), the measure of exposure, the range of exposure, exposure assessment method, variables adjusted for in the analysis, the RR or OR of CRC and the corresponding 95% CIs for each category of magnesium intake. We extracted the maximally adjusted RR or OR with the corresponding 95% CI for the highest vs lowest category of magnesium intake for use in the main analyses. We did not assess the study quality using a quality score. Instead, we reported some characteristics of the included studies that are the indicators of study quality,27 such as the numbers of cases and participants, duration of follow-up, assessments of exposure and adjustment for potential confounding factors. Data extraction was conducted independently by two authors (G-CC and ZP), with any disagreements resolved by consensus. Statistical analysis We used a DerSimonian and Laird random-effects model,28 which considers both within- and between-study variation to calculate the summary risk estimate. Because the outcomes were relatively rare, the ORs in a nested case–control study25 were considered to be approximations of RRs. One study21 presented results for both dietary and total magnesium intakes (dietary and supplements combined); we used the results for dietary magnesium in the primary analyses for better comparability between studies. We also attempted to investigate the association of magnesium with CRC risk by colorectal subsites, by sex, anatomical subsites of the colon and also by body mass index (BMI), if possible. We also conducted a dose-response analysis by using the method proposed by Greenland and Longnecker29 and Orsini et al.30 This method requires that the number of cases, person-years (or controls in a nested case–control study) and the risk estimates with their variance estimates for at least three quantitative exposure categories are known. For one study20 that did not present the number of cases and person-years in each exposure category, we used the variance-weighted least square regression model to estimate the slopes. For each study, the median or mean level of magnesium intake for each category was assigned to each corresponding RR estimate. When the median or mean intake per category was not provided, we assigned the midpoint of the upper and lower boundaries in each category as the average intake. If the highest or lowest category was open-ended, we assumed the width of the interval to be the same as in the closest category. We presented the results of linear dose–response for an increment of 50 mg/day. Heterogeneity test was performed by using Q and I2 statistics.31 For the Q statistic, a P-value of <0.1 was considered statistically significant heterogeneity. Potential publication bias was investigated by the use of Begg’s funnel plots and Egger’s regression asymmetry test.32 All statistical analyses were done using STATA software, version 11.0 (STATA, College Station, TX, USA). All P-values are two-sided and the level of significance was <0.05, unless explicitly stated. Top of page Results Study characteristics A flowchart showing the process of study selection is shown in Figure 1. Our literature search identified eight eligible prospective studies (seven cohort studies and one nested case–control study25), which contained a total of 338 979 participants and 8000 CRC cases. Characteristics of the included studies are shown in Table 1. These studies were published between 2005 and 2012. Three of the eight studies were conducted in the United States, four in Europe and one in Japan. Four studies included both men and women, and the other four studies consisted of women only. The follow-up durations of the prospective cohort studies ranged from 7.9 to 28 years. All studies provided multivariable risk estimates for CRC, six studies for both colon and rectal cancer. Figure 1. Flow chart showing the process of study selection. Full figure and legend (57K) Table1 - Characteristics of eight prospective studies on magnesium intake and colorectal cancer. Full table Most studies used a validated food-frequency questionnaire (FFQ) in exposures assessment; one study25 used food diaries. Age and total energy intake were adjusted for in all included studies and most studies additionally adjusted for BMI (N=7), physical activity (N=7), vitamin D or multivitamin use (N=7), calcium use (N=6), smoking status (N=6), alcohol consumption (N=6), dietary fiber intake (N=5) and folate use (N=5). Magnesium intake and CRC The pooled analysis of all studies showed that the summary RR of CRC for the highest vs lowest category of magnesium intake was 0.89 (95% CI, 0.79–1.00). The association was statistically significant (P-trend=0.045), with no heterogeneity (Figure 2). Figure 2. Pooled random-effects RR (95% CI) of colorectal cancer risk comparing the highest with the lowest category of magnesium intake. Full figure and legend (66K) On repeating the analysis by omitting the only nested case–control study,25 which adjusted for the smallest number of confounding factors, the summary was 0.87 (95% CI, 0.77–0.99) and the association was statistically significant (P-trend=0.031), with no heterogeneity (P=0.43, I2=0.0%). One study21 presented results for both dietary and total magnesium intakes. If the results for total magnesium were used in the pooled analysis, the summary RR was 0.89 (95% CI, 0.79–1.00) and the association was statistically significant (P-trend=0.048), with no heterogeneity (P=0.46, I2=0.0%). A high calcium intake has been demonstrated to reduce the absorption of both calcium and magnesium.33 Hence, we further examined the association of magnesium intake with the risk of CRC by removing two studies21, 25 that have not adjusted for calcium intake. After that, the summary RR was 0.87 (95% CI, 0.75–1.00) and the association was statistically significant (P-trend=0.046), with low heterogeneity (P=0.34, I2=12.1%). The summary RR of six19, 20, 21, 22, 23, 26 studies for women was 0.86 (95% CI: 0.75–0.99), with no heterogeneity (P=0.67; I2=0.0%); only two studies22, 23 provided results for men. Six19, 20, 21, 22, 23, 26 of the eight studies also conducted site-specific analyses. For these studies, the pooled RRs for colon and rectal cancer were, respectively, 0.81 (95% CI, 0.70–0.93) and 0.94 (95% CI, 0.72–1.24). Little evidence of heterogeneity was observed in these analyses. Three studies19, 22, 23 also presented results on anatomical subsites of colon cancer. The summary RRs for proximal and distal colon cancer were 0.73 (95% CI, 0.53–1.01) and 0.79 (95% CI, 0.44–1.43), respectively. Three studies22, 23, 26 also presented the risk estimates, stratifying by BMI. For these studies, the pooled RRs were 0.87 (95% CI, 0.54–1.41) for those with a BMI <25 kg/m2 and 1.08 (95% CI, 0.77–1.51) for those with a BMI 25 kg/m2. Dose–response analysis of the primary studies showed that the summary RRs for an increase in magnesium intake of 50 mg/day for colorectal, colon and rectal cancer were, respectively, 0.95 (95%CI, 0.89–1.00; N=8), 0.93 (95%CI, 0.88–0.99; N=6) and 0.93 (95%CI, 0.83–1.04; N=6). There was some evidence of heterogeneity (P=0.05, 0.17 and 0.04, respectively; I2=49.3%, 33.3% and 54.9%, respectively). In the sensitivity analyses, in which one study at a time was removed and the rest pooled, we found that the study by Larsson et al.19 substantially contributed to the heterogeneity among studies. After excluding this study, the summary RRs for colorectal, colon and rectal cancer were 0.96 (95% CI, 0.92–1.00), 0.94 (95% CI, 0.89–0.99) and 0.99 (95% CI, 0.92–1.06), respectively, with little or no evidence of heterogeneity observed (P=0.30, 0.22 and 0.46, respectively; I2=16.0, 28.6 and 0.0%, respectively). Publication bias Begg’s funnel plots suggested the absence of asymmetry, with regard to colorectal, colon or rectal cancer in relation to magnesium intake. This was corroborated by the P-value for Egger regression asymmetry test (all P-values >0.45). Top of page Discussion Findings from this meta-analysis of prospective studies, including more than 8000 CRC cases, indicate that a high magnesium intake may provide a modest protection against CRC risk. The reduction in risk is 11% (95% CI, 0–21%) for the highest vs lowest category of magnesium intake and, on average, 5% (95%CI, 0–11%) for each additional 50 mg/day increase in magnesium intake. The observed beneficial effect of magnesium intake against CRC appeared to be independent of calcium intake and restricted to colon cancer only. One case–control study from France showed that CRC cases had a lower intake of magnesium compared with the controls, but the difference was not significant after adjusting for other nutrients.34 Several subsequent case–control studies have yielded inconsistent results on the relationship between magnesium intake and the risk of CRC, with both inverse35 and null associations suggested.36, 37 Another large case–control found that total magnesium intake was linked to a significant lower risk of colorectal adenoma, particularly in those subjects with a low ratio of calcium:magnesium.38 Studies of a case–control design were subject to biases such as recall and selection, because lifestyles and diet habits in retrospective case–control studies are determined after the diagnosis of cancer. The major strengths of this meta-analysis were that all included studies were of a prospective design, which eliminates the possibility of recall and selection biases. Furthermore, most studies had a long duration of follow-up and a large number of CRC cases. These features of the study enhanced the statistical power to assess the long-term effects of magnesium intake on CRC risk. This meta-analysis also has several limitations. First, as a meta-analysis of epidemiological studies, it is not able to solve the problem of confounding that is universal in the included studies. We cannot entirely rule out the possibility of some confounding factors as a potential explanation for the observed findings. However, most of the included studies have adjusted for major potential confounders. Omitting the only nested case–control study adjusting for the smallest number of confounders or excluding the studies that have not adjusted for calcium intake did not materially alter the summary risk estimate. A second limitation was that most included studies assessed diet with a FFQ, and all but two studies25, 26 assessed magnesium intake only at baseline. This could introduce measurement error and lead to a misclassification of exposure. As an instrument to quantify nutrient intakes, FFQ is relatively rough, even though most FFQs have been validated before application. Another source of misclassification of exposure is that magnesium intake from water, which would be about 1–30 mg/day, as estimated by Larsson et al.,39 has not been taken as a source of magnesium into account. Although misclassification will not create false associations, it would tend to reduce possible real association toward the null. Therefore, the observed reduction in CRC risk is possibly a conservative estimate. Third, we were not able to assess the impact of magnesium from supplement use on CRC risk, because current data of this aspect are scanty. Among the participants in the included studies, supplement magnesium only accounts for a small proportion of total magnesium intake. For instance, magnesium from supplement contributes <8% of total magnesium in the Nurses’ Health Study, 3 and the corresponding figure in the Iowa Women’s Health Study was <5%;20 moreover, in the Netherlands cohort study, only 0.2% individuals used magnesium supplement. Fourth, subgroup analyses to test the possible impact of geographic areas and years of follow-up were not conducted due to the limited numbers of studies. Finally, given that our meta-analysis was based on published studies, publication bias that results from a tendency to publish only positive results also merits consideration. In this meta-analysis, little indication of such bias was detected. In summary, findings from this meta-analysis of prospective studies suggest that a higher or increased intake of magnesium may help to mildly reduce the risk of CRC, in particular colon cancer. Given the observed association of magnesium intake with CRC risk is of borderline significance, the findings are required to be confirmed by more large prospective studies.