Moving beyond BMI: Exploring the Heterogeneity of Obesity

“Moving Beyond BMI: Exploring the Heterogeneity of Obesity” was a virtual symposium sponsored by the National Institutes of Health (NIH) Obesity Research Task Force (ORTF), which convened on May 5, 2022. The seminar was opened to the public, and about 1,353 people attended via Zoom and NIH VideoCast.

Obesity is a chronic disease with deleterious effects on health and well-being. Body mass index (BMI ) >/= 30 is used in routine assessment to indicate obesity, but it is an imperfect proxy for health risk. The purpose of the ORTF  seminar was to convene extramural scientists with diverse expertise to discuss BMI and adiposity and their use to explore the heterogeneity of obesity, including risk for obesity development and complications, pathophysiology, and response to treatment, as well as implications for implementing obesity prevention and treatment. Presenters addressed the role of adipocytes and adipose tissue metabolism in the heterogeneity of obesity, metabolically healthy (MHO) and unhealthy obesity (MUO) and their relationship to cardiovascular disease (CVD) risk, genetic subclassification of obesity and its role in precision health, and use of machine learning to inform intervention targets that address the heterogeneity of obesity in children from diverse households.

The Division of Cardiovascular Sciences in the National Heart, Lung, and Blood Institute (NHLBI) hosted the “Moving Beyond BMI” seminar. The seminar co‑chairs, representing the ORTF, moderated the sessions.

Watch on Demand the Virtual Meeting

Moving Beyond BMI: Exploring the Heterogeneity of Obesity Workshop (NIH VideoCast)

Background

Obesity, a major contributor to serious health conditions such as diabetes and cardiovascular disease (CVD), has increased substantially in prevalence over the past 30 years in both the United States and globally. However, not all individuals with obesity are equally susceptible to the adverse health problems typically associated with obesity. Obesity disparities exist in the United States, with African Americans, Hispanics and Native Americans, and populations with low socio-economic status having a disproportionate prevalence of obesity. A phenotype of apparently “metabolically healthy” obesity (MHO) has been identified, in which individuals have increased body fat, but do not have negative health outcomes. Understanding not only which individuals are most at risk for (or protected from) obesity-related health conditions, but also the mechanisms by which increased body fatness or body fat distribution does or does not contribute to disease in specific individuals or subgroups can help target prevention and treatment strategies.

Presentations

After a brief introduction by Gary Gibbons, M.D., NHLBI Director and ORTF Co-Chair, the symposium consisted of the following presentations:

The Role of Adipocytes and Adipose Tissue Metabolism in the Heterogeneity of Obesity

Philipp E. Scherer, Ph.D., University of Texas Southwestern Medical Center’s Touchstone Diabetes Center, highlighted the heterogeneity of adipocytes and adipose tissue. He noted that in addition to metabolic complications commonly associated with obesity, elevated BMI is also associated with an increased incidence of cancerous tumor lesions. Further, dysfunctional adipocytes and adipose tissue are the target for several parasites and viruses, including SARS-CoV-2. The adipocyte is the most critical cell driving the obesity epidemic; adipocytes may be white, brown, or beige. Fat cells can de-differentiate, thereby reversing negative health states. Individual fat pads differ in their basic physiological makeup and location. How each stores excess calories under different conditions contributes to the overall metabolic function of the system. Dr. Scherer described mouse models for metabolically healthy and unhealthy obese (MHO or MUO) and lean (MHL or MUL) conditions. He provided examples of mouse studies in which fat pads have been experimentally manipulated, including those in which leptin levels have been altered and others in which iron levels in the matrix of the mitochondria of the fat cells have been manipulated. Dr. Scherer emphasized the role of adiponectin and ceramides as potential biomarkers of obesity.

Metabolically Healthy and Unhealthy Obesity

Samuel Klein, M.D., Washington University School of Medicine, described research indicating that a robust MHO phenotype exists and remains stable over time. He cited findings from a long-term study comparing individuals who met rigorous MHO criteria with those with a metabolically healthy lean (MHL) phenotype and those with MUO. People with the MHO phenotype were more insulin-resistant than MHL individuals. Compared with those with the MUO phenotype, those with the MHO phenotype had greater oxygen tension in adipose tissue, less collagen and collagen gene expression, and fewer inflammatory markers. However, there was considerable heterogeneity among people, with some overlap in both collagen and inflammation between individuals with the MHL phenotype and those with the MUO phenotype. Dr. Klein identified possible sources of metabolic dysfunction in the MUO phenotype, including the role of adipokines, particularly plasminogen activator inhibitor 1 (PAI-1); plasma exosomes that secrete microRNA, lipids, and proteins into the bloodstream; concentrations of glucose and insulin; and possibly blood concentrations of fatty acids.

Metabolically Healthy Obesity and Cardiovascular Risk:  Fact or Fiction?

Yvonne Commodore-Mensah, Ph.D., RN, Johns Hopkins University School of Nursing and the Johns Hopkins Bloomberg School of Public Health, argued that the MHO phenotype’s characterization as “healthy” is “fiction,” based on evidence from several large-scale epidemiological studies and the Atherosclerosis Risk in Communities (ARIC) study. Among the health risks associated with MHO that she identified are an increased risk of heart failure and elevated troponin levels, which have been associated with coronary heart disease (CHD) incidence, CHD-related death, and all-cause mortality. Even if those with the MHO phenotype are at lower risk for CVD than those with the MUO phenotype, they remain at risk for cancer, arthritis, and adverse COVID-19 outcomes. Dr. Commodore-Mensah emphasized the need for longitudinal studies of individuals with obesity, noting that almost half of the subjects in the ARIC study with the MHO phenotype transitioned to the MUO phenotype over a 6-year period. She also noted a need for more studies of these phenotypes among underserved racial groups and of their association with different CVD subtypes.

Genetic Subclassification of Obesity and Its Role in Precision Health

Ruth Loos, Ph.D., Novo Nordisk Foundation Center for Basic Metabolic Research at the University of Copenhagen described the findings of genome-wide association studies (GWAS) of huge data sets from 23andMe and the Genetic Investigation of ANthropometric Traits (GIANT) consortium that identified nearly 1,700 genetic loci associated with adiposity outcomes. Sixty-two loci were identified that were significantly associated with both higher adiposity and lower cardiometabolic risk. The 62-loci cluster into three subtypes. Genetic risk scores showed a distinct association signature for each subtype. High genetic risk factors by subtype before onset of disease: Subtype 1: High adiposity, low waist-to-hip ratio, very low triglycerides, very high HDL cholesterol; Subtype 2: High adiposity, low triglycerides, high HDL; and Subtype 3: High adiposity, low glucose, low HbA1c. She reported that genetic association data alone will most likely not result in a precise subclassification of obesity, as heritability contributes approximately 40–70 percent and environment 30–60 percent. GWAS data explain less than 15 percent of the variation in BMI.

Heterogeneity of Obesity in Children From Diverse and Immigrant/Refugee Households:  Using Machine Learning to Inform Intervention Targets

Jerica M. Berge, Ph.D., University of Minnesota Medical School, and Allan Tate, Ph.D., University of Georgia College of Public Health, reported on the use of machine learning in the Family Matters cohort study to identify multilevel predictors of child obesogenic risk using high-dimensional risk and protective factors data from Black, Hispanic, Hmong, Native American, and White study participants. Ten of 57 possible predictors at the child, parent, dyad, household, and neighborhood levels of children’s weight trajectory toward lower BMI were identified in the full sample. The top three predictors in the full sample were conversations between parent and child about the need for weight loss (dyad); parent BMI (parent); and child satiety responsiveness (child). The study found some heterogeneity in predictors, notably among Hispanic populations, but less racial and ethnic heterogeneity than expected. The current Family Matters randomized clinical trial is a community-based intervention that uses multiple interventions to impact the proximal influences on children’s BMI. These interventions include mobile health technologies to deliver ecological momentary interventions focused on reducing parental negative mood/stress; community health workers (CHWs) engaged in a home visiting model focused on health behavior change during mealtimes, food preparation and demonstrations, and try-it-yourself activities; and video feedback on Zoom-recorded video of foods served, diet quality, eating behaviors, food parenting practices, and the interpersonal atmosphere during mealtimes. Using motivational interviewing, a CHW gives feedback and sets behavior change goals with the family around specific behaviors observed in the video footage.

Summary of Research Opportunities and Cross-Cutting Themes

Collectively, these seminars highlighted the need to utilize established and novel methodologies to further explore the heterogeneity of obesity across basic science, and epidemiologic and mechanistic/intervention studies. 

Basic Science

• Studies that manipulate fat pads and body fat distribution in animal models to understand adipocyte differentiation, fat storage, and biomarkers (e.g., leptin, adiponectin, and ceramides).
• Studies to determine whether alterations in adipose tissue biology and pathogenesis lead to metabolic dysfunction or protection from metabolic dysfunction in animal models and in individuals with obesity.
• Studies that determine the importance of the gut microbiome as well as leakage of bacteria and bacterial products across the intestine in animal models and in individuals with metabolically healthy obesity (MHO), metabolically unhealthy obesity (MUO), metabolically healthy leanness (MHL), and metabolically unhealthy leanness (MUL).

Epidemiologic Studies

• Studies that use large-scale epidemiologic data to understand the prevalence of different phenotypes of obesity and leanness, including MHO, MUO, MHL, and MUL, among diverse populations, with a goal of developing universal and harmonized definitions of these phenotypes.
• Studies that identify more refined subtypes in the population with obesity, for example, using precision tools to diagnose individuals with different phenotypes of obesity.
• Research to understand the factors underlying the heterogeneity in metabolic risk among people with similar BMI categories, including studies that elucidate the genetic differences among MHO and MUO to provide new insights into the mechanisms responsible for obesity-induced metabolic disease.
• Comprehensive approaches to assessing obesity in large cohort studies, including genetics, omics, environmental (e.g., geolocation technologies), and lifestyle factors to better diagnose individuals with obesity and provide the basis for targeted intervention, prevention, improved prediction, and prognosis.
• Studies that explore heterogeneity of obesity using longitudinal data, machine learning, and artificial intelligence techniques to understand heterogeneity of obesity in diverse populations.
• Studies that examine predictors of healthy weight and weight-related behaviors and trajectories in pre-adolescents, adolescents, and adults across race/ethnicity and sex. 

Mechanistic and Intervention Studies

• Studies that investigate the mechanistic effects of pharmacological interventions on adiponectin, leptin, and other hormones in adipocyte metabolism.
• Studies that examine the mechanisms by which MHO transitions to MUO and the metabolic risk and CVD subtypes associated with this transition.
• Mechanistic studies that use ecological momentary assessment (EMA) tools to effectively identify predictors of health behaviors in the moment using machine learning techniques.
• Mechanistic studies that use geolocation technologies to explain how location and place (e.g., home or work environments) influence health behaviors.
• Mechanistic studies using EMA to understand the influence of social and structural determinants of healthy weight and weight-related behaviors (e.g., weight stigma, harassment, discrimination) by race/ethnicity and sex.
• Mechanistic trials of multilevel interventions to understand how each component affects the determinants of obesity and promotes weight loss and cardiometabolic changes.

Cross-cutting themes included a focus on adipose tissue biology to understand adipocyte differentiation, fat storage and biomarkers (e.g., leptin, adiponectin, ceramides), as well as to determine whether alterations in adipose tissue biology and pathogenesis lead to metabolic dysfunction or protection from metabolic dysfunction in individuals with obesity.  Other themes included the need for a rigorous and universal definition of metabolic health and risk in obesity and to understand transitions in metabolic risk over time; the need for improved, more precise diagnosis of obesity, involving a comprehensive approach combining genetics, other omics, environmental, and lifestyle information; and the need for intervention studies to identify pharmacological agents to alter ceramide, adiponectin, and/or leptin production, as well as studies to test multilevel and targeted interventions to address multiple causes of obesity.  Finally, use of research methods such as ecological momentary assessment (EMA), machine learning, and geolocation technologies may be useful in exploring and understanding heterogeneity of obesity in diverse populations and locations.

Seminar Sponsors:

ORTF Co-Chairs

• Griffin P. Rodgers, M.D., National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)
• Gary H. Gibbons, M.D., National Heart, Lung, and Blood Institute (NHLBI)
• Diana W. Bianchi, M.D., Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD)

Symposium Moderators

• Laurie Friedman Donze, Ph.D., NHLBI
• Charlotte A. Pratt, Ph.D., RD, NHLBI

Symposium Speakers

• Jerica M. Berge, Ph.D., University of Minnesota Medical School
• Yvonne Commodore-Mensah, Ph.D., RN, Johns Hopkins University School of Nursing and the Johns Hopkins Bloomberg School of Public Health
• Gary H. Gibbons, M.D., Director, NHLBI
• Samuel Klein, M.D., Washington University School of Medicine
• Ruth Loos, Ph.D., Novo Nordisk Foundation Center for Basic Metabolic Research at the University of Copenhagen
• Philipp E. Scherer, Ph.D., University of Texas Southwestern Medical Center’s Touchstone Diabetes Center
• Allan Tate, Ph.D., University of Georgia College of Public Health

NIH ORTF Senior Leadership Group

• Sonia Arteaga, Ph.D., Office of the NIH Director (OD)
• Lisa Gansheroff, Ph.D., NIDDK
• Andrew Bremer, M.D., Ph.D., NICHD
• Laurie Friedman Donze, Ph.D., NHLBI
• Layla Esposito, Ph.D., NICHD
• Christopher Lynch, Ph.D., OD
• Linda Nebeling, Ph.D., RD, FAND, National Cancer Institute (NCI)
• Charlotte Pratt, Ph.D., RD, NHLBI
• Jaime Smith, Ph.D., NIDDK
• Karen Teff, Ph.D., NIDDK
• Niteace Whittington, Ph.D., NICHD
• Susan Yanovski, M.D., NIDDK