Anthropometric metabolic subtypes and health outcomes: A data-driven cluster analysis.

Abstract

Aims: The aims of the study were to develop and validate WHOLISTIIC, a data-driven cluster analysis for identifying anthropometric metabolic subtypes.

Materials and methods: K-means cluster analysis was performed in 397 424 UK Biobank participants based on five domains, that is, central obesity (waist-to-height ratio), general obesity (body mass index [BMI]), limb strength (handgrip strength), insulin resistance (triglyceride to high-density lipoprotein cholesterol [HDLc] ratio) and inflammatory condition (neutrophil-to-lymphocyte ratio). Replication was done in the NHANES. Cox proportional hazards regression models were used to estimate the associations of clusters with incident adverse health outcomes.

Results: Six replicable clusters were identified. Compared with individuals in cluster 1 (lowest BMI with preserved handgrip strength), individuals in cluster 2 (highest handgrip strength) were not at increased risk of all-cause mortality despite higher BMI, but had small yet significant increased risks of cardiovascular mortality, incident major adverse cardiovascular events (MACE), chronic renal failure and decreased risks of mortality due to respiratory disease, as well as incident dementia; individuals in cluster 3 (lowest handgrip strength and borderline elevated BMI), cluster 4 (highest triglyceride-to-HDLc ratio and moderately elevated BMI), cluster 5 (highest neutrophil-to-lymphocyte ratio and borderline elevated BMI) and cluster 6 (highest BMI) had substantially increased risks of all-cause, cardiovascular, and cancer mortality, incident MACE and chronic renal failure. The associations of anthropometric clusters with the risk of mortality were replicated in the NHANES cohort.

Conclusions: Anthropometric metabolic subtypes identified with easily accessible parameters reflecting multifaceted pathology of overweight and obesity were associated with distinct risks of long-term adverse health outcomes.