Hidden Cardiometabolic Risk in Lean and Obese Adults: What Phenogroups Reveal

A high-performance sports car might display flawless exterior paint and pristine bodywork, yet internal diagnostics can reveal micro-friction inside the transmission or silent fuel-injector failure. For decades, routine medical assessments have relied on body mass index (BMI) and blood pressure as simple stand-ins for cardiovascular health. However, clinical reality is rarely binary. Standard labels such as lean or obese frequently obscure complex biological variations occurring beneath the surface.

A data-driven investigation published on the preprint server MedRxiv demonstrates that standard clinical labels fail to capture the true diversity of cardiometabolic disease. By analyzing multidimensional metabolic measurements with unsupervised machine learning, the researchers identified five distinct cardiometabolic phenogroups. The findings challenge the conventional assumption that a low BMI guarantees cardiovascular safety, or that obesity consistently follows a uniform path of metabolic breakdown.

Beyond BMI: The Blind Spots of Standard Cardiometabolic Risk

Traditional clinical practice categorizes patients using blunt thresholds: individuals are either lean or overweight, normotensive or hypertensive, diabetic or non-diabetic. While these categories provide practical shorthand in acute care, they overlook intermediate physiological states. Two individuals with an identical normal BMI of 22 can experience profoundly different internal biochemistry, with one displaying healthy tissue perfusion and the other accumulating silent metabolic stress.

To uncover these hidden patterns, the investigators applied latent class analysis (a statistical method that identifies unobserved subgroups within complex datasets) to the RESET discovery cohort consisting of 1,034 adults. The model analyzed 14 routine clinical and biochemical variables, including waist circumference, fasting plasma glucose, glycated hemoglobin (HbA1c), lipid fractions, and the Homeostatic Model Assessment of Insulin Resistance (HOMA-IR), which gauges how hard the pancreas must work to clear blood sugar.

Instead of forcing individuals into pre-existing clinical buckets, this unsupervised approach allowed underlying biological patterns to emerge naturally. The resulting data revealed that cardiovascular risk does not follow a simple linear path from lean and healthy to obese and diseased. Instead, distinct physiological subtypes emerge with divergent cellular signatures and unexpected patterns of cardiac strain, highlighting the need for synergistic metabolic calibration across diverse patient populations.

The Five Distinct Phenogroups: Hidden Risks in Discordant Profiles

The machine learning analysis uncovered five distinct metabolic subgroups across the cohort, revealing two discordant profiles where outward appearance diverged sharply from underlying metabolic status:

• Metabolically Preserved without Hypertension (23.6%, n=244): Characterized by minimal adiposity, optimal insulin sensitivity, low triglycerides, and preserved glycemic control.
• Metabolically Preserved with Hypertension (23.6%, n=244): Maintained favorable metabolic and lipid parameters, but presented with isolated systolic or diastolic blood pressure elevation.
• Lean-Insulin Resistant (13.5%, n=140): Displayed a normal or low BMI alongside marked insulin resistance, elevated fasting triglycerides, and impaired glucose tolerance.
• Obese-Insulin Sensitive (20.4%, n=211): Carried significant adipose tissue mass, yet retained preserved insulin sensitivity and normal fasting lipid levels.
• Obese-Insulin Resistant (18.9%, n=195): Exhibited high adiposity combined with severe insulin resistance, systemic dyslipidemia, and elevated blood pressure.

Phenogroup Subclinical Diastolic Dysfunction Rates

Metabolically Preserved:  [16.5%]
Lean-Insulin Resistant:   [36.7%]  <, Over 2x higher than Preserved
Obese-Insulin Sensitive:  [38.8%]  <, Over 2x higher than Preserved

The most striking finding appeared during comprehensive echocardiographic evaluation. Subclinical diastolic dysfunction (a subtle stiffening of the left ventricle that impairs how the heart relaxes between beats) was present in 36.7% of the Lean-Insulin Resistant group and 38.8% of the Obese-Insulin Sensitive group. Both rates were more than double the 16.5% prevalence observed in the Metabolically Preserved cohort.

This reveals that carrying covert insulin resistance while lean impairs heart muscle relaxation just as severely as carrying significant excess weight. It also demonstrates that excess adipose tissue can compromise myocardial relaxation even when circulating insulin and glucose metrics remain within normal limits. Understanding these early mechanical shifts connects closely to ongoing research into preventing heart stiffness and diastolic dysfunction.

Proteomic Fingerprints and Multi-Cohort Validation

To understand the molecular mechanisms driving these divergent phenotypes, the researchers mapped circulating plasma proteins across the study groups. The biological pathways separating the discordant groups proved entirely different, despite their nearly identical rates of subclinical heart strain.

The Lean-Insulin Resistant subgroup exhibited a pronounced systemic inflammatory proteomic signature, marked by pathways tied to vascular endothelial stress and cellular immune activation. In contrast, the Obese-Insulin Sensitive subgroup showed a strong signature of hepatic metabolic stress and liver injury pathways, operating alongside a shared core of endocrine signaling disruption.

Biological Divergence Leading to Convergent Cardiac Risk:

Lean-Insulin Resistant, >  Systemic Inflammation & Vascular Stress, >  Diastolic Stiffness
Obese-Insulin Sensitive, >  Hepatic Metabolic Stress & Lipid Overload, >  Diastolic Stiffness

The investigators constructed a practical decision tree using the most critical routine variables, allowing clinicians to map individuals into these five phenogroups without running complex algorithmic models. They then validated the classification system across three independent external populations totaling over 350,000 individuals:

1. The PICMAN Cohort (n = 120): Confirmed the distinct proteomic and inflammatory signatures identified in the discovery group.
2. The UK Biobank (n = 344,817): Verified the proteomic fingerprints and confirmed that both Lean-Insulin Resistant and Obese-Insulin Sensitive individuals face elevated long-term cardiovascular disease events compared to metabolically preserved individuals.
3. The CHARLS Cohort (n = 12,145): Validated the clinical outcomes and predictive accuracy of the decision tree model across an independent international demographic.

This extensive multi-cohort validation confirms that data-driven phenotyping captures genuine, reproducible biology rather than statistical noise.

Actionable Precision Longevity: Rethinking Subclinical Cardiac Health

These findings suggest that relying solely on BMI or standard lipid panels leaves major cardiovascular blind spots. An individual deemed healthy based on a normal body weight may harbor silent microvascular inflammation and diastolic stiffening driven by undetected insulin resistance. Conversely, assuming every individual with an elevated BMI has severe systemic insulin resistance overlooks the distinct hepatic pathways that can drive their risk.

Moving beyond blunt metrics requires assessing metabolic health with greater granularity. Clinicians and proactive individuals can evaluate underlying physiology by measuring fasting insulin alongside fasting glucose to calculate HOMA-IR, tracking advanced lipid markers such as triglyceride-to-HDL ratios, and considering resting echocardiographic strain metrics when risk factors appear discordant.

Understanding whether cardiovascular vulnerability stems primarily from immune-inflammatory pathways or hepatic metabolic stress allows preventive strategies to be tailored more effectively. Precision cardiometabolic care relies on identifying these distinct internal mechanisms before overt clinical disease develops.

Study Limitations and Research Context

Several considerations should be kept in mind when interpreting these findings:

• Preprint Status: This research was published as an early-stage study on MedRxiv and has not yet completed the formal peer-review process.
• Observational Design: While the study links specific proteomic pathways and phenogroups to subclinical heart dysfunction and cardiovascular outcomes across large biobanks, it does not demonstrate that altering these specific proteomic markers directly reverses diastolic stiffening.
• Intervention Studies Needed: The decision tree successfully categorizes individuals into risk phenogroups, but prospective randomized clinical trials are required to determine whether targeted lifestyle or pharmacological therapies tailored to specific phenogroups achieve better outcomes than standard care guidelines.

Clinical Conclusions

Just as high-end telemetry exposes mechanical friction beneath a pristine vehicle chassis, advanced metabolic profiling uncovers silent cardiovascular stress that simple scales and tape measures miss entirely. By recognizing the biological divergence between lean insulin resistance and obese insulin sensitivity, clinical practice can move closer to truly personalized cardiovascular prevention.