Methods
Search Strategy
We conducted a systematic review of the published works without language restrictions and in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines. We searched Pubmed and Embase from their inception to March 2013 and systematically identified observational studies that evaluated the association between metabolic syndrome and incidence or prevalence of fractures. We used the following main search terms with no restrictions: "metabolic syndrome" or "insulin resistance syndrome" or "syndrome X" in combination with "bone" or "fracture" or "osteoporosis" or "metabolic bone diseases " or "osteopenia" or "bone mineral density" or "BMD". We also scanned the reference lists from published original articles and previous reviews for more relevant studies not identified in the databases search.
Study Selection
We included studies in the meta-analysis that met all of the following criteria: (1) the study had a population based observational design, (2) published original data relevant to a possible association between metabolic syndrome and fractures, (3) reported the odds ratio (OR) or relative risk (RR) and its 95% confidence interval (CI). In case of multiple publications had overlap their populations and reported with the same study design, the most recent publication was included in order to avoid duplicate observation, unless more inclusive and detailed data was found in other publications. To gather more relevant information, we consulted researchers with professional knowledge at this area for the presence of unpublished reports.
Data Extraction
Two of our reviewers independently evaluated all relevant articles and identified eligible studies from the databases. During data abstraction, differences and disagreements were resolved through discussion to come to an agreement. Following information was recorded by a standardized data extraction form: last name of the first author, publication year, name of the study, geographic region of original study, composition and age range of study population, type of fractures, definitions of metabolic syndrome, unadjusted and adjusted OR or RR with corresponding 95% CI and adjustment factors of interest. If possible, we also extracted the baseline data of cohort studies for the combined estimates. We contacted authors of the primary studies for additional information when necessary.
Statistical Analysis
The primary outcome of the pooled analysis was focused on a comparison of the summary effect of fractures risk in people with metabolic syndrome versus those without. Both unadjusted and adjusted values were extracted for the pooled analysis. When studies presented results from various covariates analyses, we used the one adjusted the most study-specific confounders. The combined estimates were calculated separately by averaging the natural logarithmic OR or RR weighted by their inverse of variance based on a fixed or random effects model within or between study variations, depending on the overall heterogeneity. Heterogeneity of effect size across studies was assessed by using Cochran's Q and the I statistic and p value < 0.10 or I value > 50% was considered to be heterogeneous. In the I statistic, values of 25, 50, and 75% are considered to represent low, medium, and high heterogeneity, respectively. If substantial heterogeneity was detected, pooled effect estimates were calculated using a random effects model by the method of DerSimonian and Laird. If not, the combined estimates were presented based on the fixed effects model by using the inverse variance method.
To assess the influence of individual studies on the pooled result, we conducted sensitivity analyses to investigate the influence of a single study on the overall risk estimate by omitting one study in each turn. We used the Begg' s adjusted rank correlation test and the Egger' s regression asymmetry test to detect publication bias and p > 0.05 for both tests was considered to be no significant publication bias. Subgroup analyses according to sex (male/female), types of fractures (non-vertebral/vertebral/any fractures), definitions of metabolic syndrome [strict National Cholesterol Education Program' s Adult Treatment Panel III (NCEP-ATP III) criteria-2001/strict NCEP-ATP III criteria-2005/other modified criteria] and geographical area (Europe/America/Asia) were used to assess the impacts of study characteristics on outcomes.
All statistical analyses were performed using STATA version 11.0 (Stata Corp, College Station, TX, USA).