Metabolic Syndrome and Risk of Cancer

A Systematic Review and Meta-analysis

Katherine Esposito, MD, PHD; Paolo Chiodini, PHD; Annamaria Colao, MD; Andrea Lenzi, MD; Dario Giugliano, MD, PHD


Diabetes Care. 2012;35(11):2402-2411. 

In This Article


We screened 2,628 potentially relevant, nonduplicate articles. The κ score for concordance between reviewers rating the articles was 0.62–0.77. The final number of articles[8–50] included in the meta-analysis was 43 (Supplementary Fig. 1), which reported on 116 datasets (Supplementary Table 2). All articles were published in English. The characteristics of included studies are summarized in Table 1 and Supplementary Table 3. The analysis included 38,940 cancer cases (18,180 men and 20,201 women, plus 559 cases for gallbladder cancer not divided by sex). The median follow-up per cohort studies and per cancer site varied from 3 years (endometrium) to 12.2 years (prostate). Notably, no North American population data contributed to the summaries for gallbladder, ovary, thyroid, and bladder cancers. The proportion of studies in which the definition of metabolic syndrome was traditional varied by cancer sites: higher for colorectal (14 vs. 8 nontraditional); approximately equal for breast, hepatobiliary and prostate; and lower for pancreas. No further differentiation was made when the number of datasets for cancer site was four or fewer. The number of potential confounding factors (cancer-site–specific risk factors) included in the adjusted analyses also varied (Supplementary Table 3).

Figure 1.

Summary risk estimates by cancer sites in men (A) and in women (B).

Fig. 1A and B shows the results of meta-analyses of RR (for presence of metabolic syndrome) in men and in women, respectively, for cohort studies only. Separate meta-analyses for some relevant sites and for sex are given in Supplementary Figs. 1–6. In men, the presence of metabolic syndrome was associated with liver (Fig. 2; RR 1.43, P < 0.0001) and colorectal (Fig. 2A; 1.25, P < 0.001) cancers and weakly associated with bladder cancer (1.10, P = 0.013). Between-study heterogeneity was low or moderate for liver, colorectal, and bladder cancer (I 2 = 0.0, 35, and 0.0%, respectively) (Fig. 1A). The quality of the evidence was high for the association with colorectal cancer (high number of datasets and events, narrow CIs, and low heterogeneity), moderate for liver cancer, and low for bladder cancer. The overall risk of bias was low, as all studies were prospective cohort studies and most adjusted for many confounders.

Figure 2.

Meta-analyses for some common cancer sites in both sexes: colorectal and liver cancer in men (A) and colorectal, breast postmenopausal, endometrial, and pancreatic cancer in women (B). ES, effect size; MS, metabolic syndrome.

In women, the presence of metabolic syndrome was associated with endometrial (Fig. 2B; RR 1.61, P = 0.001), pancreas (Fig. 2B; 1.58, P < 0.0001), breast postmenopausal (Fig. 2B; 1.56, P = 0.017), rectal (1.52, P = 0.005), and colorectal (Fig. 2B; 1.34, P = 0.006) cancers; the association with ovary cancer (1.26) was of borderline significance (P = 0.054). Between-study heterogeneity was high for endometrial, breast postmenopausal, and colorectal cancers and moderate or low for rectal (I 2 = 35%), pancreas (0.0%), and ovary cancers (8.8%) (Fig. 1B). The quality of the evidence was moderate for the association with colorectal and pancreas cancers and low for endometrium and breast postmenopausal cancers. The overall risk of bias was low, as all studies were prospective cohort studies and most adjusted for many confounders. Associations with metabolic syndrome were stronger in women than in men for pancreas (P = 0.01), rectal (P = 0.01), and bladder (P = 0.01) cancers.

We also examined whether estimates varied between populations in cancer sites for which we had at least two datasets from the main geographical regions (Table 2). For colorectal cancer, for example, we recorded a positive association in U.S. and Europe populations for men and in Europe populations for women; for postmenopausal breast cancer, the positive association was lost in Europe populations; for liver cancer, the association remained significant in Europe and Asia populations for men only; and for prostate cancer the association became negative in U.S. populations (almost exclusively whites, RR 0.79, P = 0.001, I 2 = 9%). This last figure was also significant if a mortality study (dataset n = 69) was excluded (RR 0.75 [95% CI 0.60–0.94], P = 0.011, I 2 = 0.0%).

We also examined mortality from cancer in the available studies for which we had at least two datasets (Table 2). There were three cohort studies from the U.S. (two for both sexes, one for men only) and a case-control study from China (both sexes) for colorectal cancer only.[8,9,16,20] Risk estimate for cancer mortality was 1.61 (P < 0.0001), with no heterogeneity (I 2 = 0.0%).

We also examined whether results for cancer association differed according to whether studies of different design (case-control and patient series) were included in the full analysis (Table 2). For breast cancer, the inclusion of two case-control studies[26,27] with 3,950 cases produced a significant overall association (11 datasets, 9,643 cases) of 1.23 (P = 0.009) with high heterogeneity (88%). For liver cancer, the inclusion of two large case-control studies,[29,30] with an additional 4,951 cases, produced a significant association for women (RR 1.62, P < 0.0001).

The definitions used for diagnosis of metabolic syndrome affected estimates of the association between metabolic syndrome and cancer risk (Table 2). For both sexes, the estimates remained similar for colorectal cancer (RR 1.33 and 1.22 for traditional versus nontraditional definitions); for liver cancer, both definitions achieved significant associations (1.88 and 1.51); for pancreas cancer, there was no association with traditional definitions (RR 1.13, P = 0.745); for prostate cancer, both definitions gave similar results; and for breast cancer, there was an association with traditional definitions only (1.45, P = 0.025).

To summarize the results for cohort studies with available data (32 cohorts), risk estimates for single factors were equal to metabolic syndrome in 15 cohorts, higher in 11 cohorts, and lower in 6 cohorts. For colorectal cancer, for example, all increased cancer risk was explained by diabetes alone,[9,10] diabetes and waist,[10] diabetes and BMI,[12] and triglycerides >150 mg/dL;[15] other single or combined factors explained part (from 30 to 50%) of the increased risk conveyed by metabolic syndrome: BMI,[11,21] waist,[16,19] BMI and lipid,[17] BMI and dysglycemia,[14] and hypertension.[21]

Influence analysis showed that no single study affected the sex-specific summary estimates for most sites. Moreover, we did not note funnel plot asymmetry for cancer sites where a sufficient number of datasets exited to run the Egger test (colorectal cohorts men, P = 0.912; colorectal cohorts women, P = 0.201; and prostate cancer cohorts, P = 0.085).