Abstract:Abstract Background: The definition of metabolically healthy and unhealthy obesity for cancer risk remains uncertain and controversial. This study proposed a novel classifier based on biomarkers selected using machine learning (ML) and compared its risk stratification with the conventional definition. Methods: A prospective cohort study was conducted of 317,569 UK Biobank participants who were free of cancer and with body mass index (BMI) ≥18.5 kg/m2 at baseline. Individuals were classified into metabolically healthy and non-obese (MHNO), metabolically unhealthy and non-obese (MUNO), metabolically healthy and obese (MHO), and metabolically unhealthy and obese (MUO), according to body mass index (BMI) and six metabolic criteria. For the ML approach, LASSO regularization was used to select a subset from seventeen metabolic biomarkers to optimize C index. Clinical cut-off value was applied to this subset to define MHO. Multivariable Cox proportional hazards models were used to estimate hazard ratios (HRs) of total, obesity-related, type two diabetes (T2D)-related, and 23 site-specific cancers according to the conventional and ML definitions. Findings: Of 21 biomarkers, 17 were selected using LASSO. Compared with MHNO, individuals with MHO had higher risk of obesity-related cancer (HRconventional 1.22, 95% CI 1.15-1.29; HRML 1.23, 95% CI 1.16-1.30) and T2D-related cancer (HRconventional 1.25, 95% CI 1.19-1.39; HRML 1.27, 95% CI 1.20-1.34) after adjusting for sociodemographic and lifestyle factors. ML-defined metabolic status better-stratified individuals with normal weight for risk of total cancer (HRconventional1.02, 95% CI 0.99 -1.06; HRML 1.06, 95% CI 1.02 - 1.10) and some site-specific cancers, e.g., hepatocellular carcinoma (HRconventional 0.89, 95% CI, 0.53-1.48; HRML 2.52, 95% CI, 1.85-3.45). Interpretation: Compared with the conventional definition of metabolic health, a broader array of metabolic markers may help better stratify individuals for cancer risk. Keywords: Obesity, cancer, metabolically healthy obesity, hepatocellular carcinoma. Citation Format: Ziyi Zhou, Solange Parra-Soto, Yujia Lu, Zhe Fang, Kai Wang, Alaina Bever, Liyuan Tao, Fanny Petermann-Rocha, Jirapitcha Boonpor, Naveed Sattar, Carlos Celis-Morales, Jill P. Pell, Frederick K. Ho, Mingyang Song. Defining metabolically healthy and unhealthy obesity in relation to cancer risk: A prospective cohort study by using a machine learning approach in comparison with conventional definitions in the UK Biobank [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4870.

Machine learning‐based clustering identifies obesity subgroups with differential multi‐omics profiles and metabolic patterns

Machine Learning to Identify Metabolic Subtypes of Obesity: A Multi-Center Study

Subclassification of obesity for precision prediction of cardiometabolic diseases

244-LB: Data-Driven Metabolic Subgrouping of Obesity by Machine Learning (ML)

Cardiometabolic profiles and proteomics associated with obesity phenotypes in a longitudinal cohort of young adults

Multiomics Picture of Obesity in Young Adults

Identifying subgroups of childhood obesity by using multiplatform metabotyping

Multiomic signatures of body mass index identify heterogeneous health phenotypes and responses to a lifestyle intervention

Novel subgroups of obesity and their association with outcomes: a data-driven cluster analysis

Association between obesity phenotypes and dietary patterns: A two-step cluster analysis based on the China multi-ethnic cohort study

Multiomic investigations of Body Mass Index reveal heterogeneous trajectories in response to a lifestyle intervention

Abstract 4870: Defining metabolically healthy and unhealthy obesity in relation to cancer risk: A prospective cohort study by using a machine learning approach in comparison with conventional definitions in the UK Biobank

Using Machine Learning to Predict Obesity Based on Genome-Wide and Epigenome-Wide Gene–Gene and Gene–Diet Interactions

1252-P: Differences in Cardiometabolic and Immune Biomarkers between Novel Subphenotypes of Individuals at Elevated Risk for Type 2 Diabetes

Metabolomic Profiles of Body Mass Index in the Framingham Heart Study Reveal Distinct Cardiometabolic Phenotypes

Abstract EP02: Identification Of Genetic Signals For “Diabesity” --- Type 2 Diabetes And Obesity -- Among African American And European American Participants In Four Cohorts Of The TOPMed Consortium

WHOLE GENOME SEQUENCING ANALYSIS OF BODY MASS INDEX IDENTIFIES NOVEL AFRICAN ANCESTRY-SPECIFIC RISK ALLELE

Multi-omics landscape of childhood simple obesity: novel insights into pathogenesis and biomarkers discovery

Identification of Novel Metabolic Subtypes Using Multi-Trait Limited Mixed Regression in the Chinese Population

Lifestyle‐based subtypes of childhood obesity and their association with cardiometabolic factors: A cluster analysis

Integrative Analysis of Multi-Omics Data to Detect the Underlying Molecular Mechanisms for Obesity in Vivo in Humans