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

Research Design and Methods

Data Sources

We followed the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) checklist for reporting systematic reviews and meta-analyses.[6] We systematically searched Medline, Embase, CENTRAL, CINAHL, and Web of Science through October 2011 for studies in humans of the association between metabolic syndrome and cancer. Our core search consisted of the terms metabolic syndrome, insulin resistance syndrome, and syndrome X, combined with specific terms for each cancer site: colorectal (colon and rectum), gastric, esophageal, hepatobiliary (liver and gallbladder), pancreas, lung, bladder, thyroid, renal, leukemia, malignant melanoma, multiple myeloma, and non-Hodgkin lymphoma for both sexes and prostate, breast, ovary, and endometrium for single sex. Relevant journals, bibliographies, reviews, and personal files were hand searched for additional articles. The search had no language restriction. The last search was performed on 31 October 2011. The electronic database search strategy for Medline is available in Supplementary Table 1.

Study Selection

We included studies if 1) their aim was to assess the effect of metabolic syndrome on risk of cancer or association with cancer, 2) they reported the definition of metabolic syndrome according to criteria of national or international scientific associations, federations, or organizations (traditional definitions) or if they used proxy indicators in the absence of the original data (nontraditional definitions), and 3) they included at least three factors, even in the absence of others. We included cohort studies, nested case-control studies, control arms from clinical trials, case-control studies, patient series, and mortality studies. We specified that every study must either report risk estimates (relative risks [RRs], odds ratios, hazard ratios, and standardized incidence ratio) with 95% CIs separately for men, women, or both or must report sufficient data to estimate these. If a site-specific dataset had been published more than once, we used the most recent publication. We included a specific cancer site in the analysis if there were at least two cohort datasets. We excluded studies that were not published as full reports, such as conference abstracts and letters to editors, and studies of cancer precursors (e.g., colorectal adenoma).

Data Extraction

From each retrieved article, we extracted the following data: name of the first author, year of publication, country where the study was performed, specific outcomes, follow-up time, proportion of men and women, total number of individuals, number of cases, and risk estimates and their 95% CIs (presence versus absence of metabolic syndrome). We collected data for the most adjusted model. Populations were categorized into four groups: U.S., Europe, Asia, and other. Returned articles were reviewed against inclusion and exclusion criteria by three reviewers (D.G., K.E., and P.C.) until interrater reliability (κ ≥ 0.60) was established. Methodological quality of each study was assessed according to three study components that might affect the strength of the association between metabolic syndrome and cancer risk: length of follow-up for cohort studies, whether metabolic syndrome definition was traditional or nontraditional, and the extent of adjustments for potential confounding factors. We also collected, where available, risk estimates of the association with cancer for each single factor of the syndrome taken at its highest level.

Data Synthesis and Analysis

The primary end point was to assess the association between metabolic syndrome and cancer risk in cohort studies. For the main outcome at each cancer site, we graded the evidence for study quality and for the risk of bias: study quality was based on the number of datasets, number of events, width of CIs, and heterogeneity; risk of bias was mainly based on type of study and adjustment for confounders. Unless otherwise stated, we used the most adjusted risk estimate from each study. Heterogeneity of the effect across studies was assessed by Q2 statistics, which is distributed as χ2 statistics.[7] A value of P < 0.10 was used to indicate lack of homogeneity (heterogeneity) among effects. I 2 statistics were provided to quantify the percentage of total variation across studies that was attributable to heterogeneity rather than to chance. I 2 values of 25, 50, and 75% correspond to cutoff points for low, moderate, and high degrees of heterogeneity. We used a fixed-effects model if I 2 value significance was >0.1; otherwise, we used a random-effect model. We did subgroup analyses for each site to identify study-level factors that modify the association between the presence of metabolic syndrome and cancer risk: these factors include sex, subsite (e.g., colon and rectum), definition of metabolic syndrome (traditional versus nontraditional), and design; for incidence of cancer, we considered cohort studies, nested case-control studies, and control arms of clinical trials. Sensitivity analyses evaluated whether the results could have been affected markedly by a single study and were repeated using a fixed-effects model. Publication bias was examined in funnel plots and with a regression asymmetry test: the Egger test is best for cancer sites with 10 or more datasets. We used STATA, version 9.0 (STATA, College Station, TX), to analyze data.