Background Abdominal obesity, hyperglycemia, hypertriglyceridemia, low HDL cholesterol levels and hypertension frequently co-occur within individuals. The cluster of these features is referred to as the metabolic syndrome (MetS). The aim of this thesis was to investigate which metabolic endpoints should be studied in order to explain the clustering of MetS features best. Furthermore, genetic association studies were conducted to get more insight into the pathophysiology underlying the clustering of MetS features. Methods We conducted two studies to investigate which metabolic endpoints should be studied in order to best explain the clustering of MetS features. In the EPIC-NL case-cohort study we studied the model fit and predictive ability for type 2 diabetes (T2D) and cardiovascular diseases (CVD) of several MetS models, including traditional and novel MetS features. Model fit was analysed with confirmatory factor analysis. Furthermore, we reviewed published twin and family studies, which presented genetic correlation coefficients between different traditional and novel MetS features and between MetS and novel MetS features. We conducted four studies investigating which single nucleotide polymorphisms (SNPs) were associated with clustering of MetS features. First, we systematically reviewed published candidate gene studies on MetS. Second, we analysed whether SNPs associated with inflammatory biomarkers, waist circumference, insulin resistance, HDL cholesterol or triglycerides in genome wide association studies (GWAS) were also associated with MetS and MetS-score in a random sample of the EPIC-NL study. Third, in the Doetinchem cohort, we determined if SNPs of genes located in transcriptional pathways of glucose and lipid metabolism were associated with multiple MetS features simultaneously. Fourth, we evaluated the interaction between these SNPs and BMI in relation to glucose levels. Results A MetS model composed of the traditional MetS features and high sensitive C-reactive protein (hsCRP) optimally predicted T2D and CVD, while still representing a single entity. Our review of 9 twin and 19 family studies showed that genetic correlations were strongest, i.e. genetic pleiotropy was highest, between waist circumference and HOMA-IR, HDL cholesterol and triglycerides, and between adiponectin and MetS. After having systematically reviewed 25 genes in 88 candidate gene studies, we found evidence for an association of FTO rs9939609, TCF7L2 rs7903146, APOA5 C56G (rs3135506), APOA5 T1131C (rs662799), APOC3 C482T (rs2854117), Il6 174G>C (rs1800795) and CETP Taq-1B (rs708272) with MetS. SNPs associated with waist circumference in GWAS were on a group level significantly associated with MetS in a random sample of EPIC-NL, whereas a group of SNPs associated with insulin resistance was significantly associated with MetS-score. On the individual level MC4R rs17782312 and IRS1 rs2943634 were associated with MetS. In the Doetinchem cohort CETP Ile405Val (rs5882) and APOE Cys112Arg (rs429358) were associated with both the prevalence of low HDL cholesterol levels and with abdominal obesity. In this cohort, two highly correlated SNPs in the PPARGC1A gene, Gly482Ser (rs8192678) and Thr528Thr (rs3755863), showed a significant interaction with BMI on glucose levels. Conclusion Our results show that one MetS factor with or without hsCRP, can be used to study the clustering of MetS and MetS related features, because this factor can be represented as one statistical entity. However, in order to fully explain the clustering of MetS features, specific combinations of MetS features should be studied. Our results indicate that genetic pleiotropy is highest for the combination of HOMA-IR and waist circumference and the combination of HDL cholesterol and triglycerides. Therefore these combinations are good candidate endpoints for studies on genetic variants pleiotropic to several MetS and MetS related features. SNPs associated with the clustering of MetS features are involved in mechanisms traditionally believed to underlie MetS development, i.e. glucose metabolism and weight regulation, but also in other mechanisms, i.e. lipid metabolism and inflammation. This suggests that, although the MetS features may represent a statistical entity, there are multiple, related mechanisms explaining the clustering of MetS features.