When BMI looks fine – but disease risk is already building

Two people can have the same BMI – and radically different metabolic health: how their bodies process sugar, fat and energy in ways that shape disease risk. One may already be on a path toward diabetes and cardiovascular disease, whereas the other is not.

“We see this mismatch all the time in the clinic. That discrepancy is exactly what motivated us,” says Rima M. Chakaroun, a physician-scientist at the Wallenberg Laboratory, Sahlgrenska University Hospital in Gothenburg, Sweden. “BMI is easy to measure, but it compresses enormous biological complexity into a single number. It works at the population level but was never designed to explain individual biology.”

In the clinic, Chakaroun saw patients who appeared healthy yet had severe insulin resistance, abnormal lipid profiles or hypertension – whereas others with obesity showed few metabolic warning signs. The problem, she realised, was not weight itself but what weight fails to reveal biologically.

To expose that blind spot, Chakaroun analysed data from more than 1,400 adults during her second postdoctoral work, systematically comparing multiple biological layers. Genetics, lifestyle measures and several omics datasets were examined, but the strongest and most consistent signal came from the blood metabolome alone.

The result was a new, data-driven measure of metabolic health – metabolic BMI, or metBMI – derived from circulating metabolites and designed to capture biological risk that the scale alone cannot detect.

“This is not about redefining obesity cosmetically,” Chakaroun stresses. “It is about identifying biological risk earlier – before disease develops.” Individuals whose metBMI exceeded their measured BMI had two- to fivefold higher odds of developing type 2 diabetes, metabolic syndrome and fatty liver disease – despite often being labelled low risk by conventional standards.

Together, the findings point toward a future in which metabolic health is assessed not just by scales and tape measures but by molecular signals that reflect how the body actually functions – opening the door to more precise prevention and treatment.

Why BMI fails individual biology

For decades, BMI has been the dominant tool for defining obesity in both research and clinical practice. It is simple, cheap and easy to apply at scale. But its limitations are equally well known.

“BMI worked for what it was meant to do,” Chakaroun says. “But biology does not know where we draw those lines.”

For Chakaroun, this gap was not an abstract methodological concern – she encountered it repeatedly in the clinic. She describes seeing “extremely fit-looking people” arrive with horrendous blood sugar control and lipid profiles alongside patients with obesity who showed few metabolic complications – experiences that shifted her perspective.

This clinical disconnect has long been recognised in research. Concepts such as “metabolically healthy obesity” and “metabolically unhealthy normal weight” were introduced to capture it, but their definitions have varied widely and often rely on a narrow set of clinical markers. Chakaroun argues that these categories remain limited because they fail to reflect how metabolism operates across tissues, organs and biological systems.

The patients who do not fit the BMI rule

What changed in recent years is the ability to look deeper. Large-scale omics technologies – including metabolomics, proteomics and microbiome profiling – now enable thousands of molecular features to be measured simultaneously. Rather than asking whether someone crosses a single diagnostic threshold, Chakaroun says, the question becomes how the metabolic system as a whole is behaving.

“Clinical cutoffs are useful administratively,” says Fredrik Bäckhed, senior author and Professor of Molecular Medicine at the University of Gothenburg, Sweden, “but biology does not suddenly change because you cross a line. Individual molecular layers may correlate with BMI, but none of them alone captures metabolic health in full.”

The stakes are high. Obesity-related diseases account for a growing share of global illness – and prevention depends on identifying risk before weight or symptoms force action.

Reading the body’s chemistry – not the scale

To move beyond BMI as a blunt proxy for metabolic health, a framework was established integrating clinical measurements with multiple layers of molecular data. Rather than discarding established clinical tools, the goal was to place them in a broader biological setting.

“The key idea was not to replace clinical information,” says Rima M. Chakaroun, “but to embed it within a much richer biological context.”

The core analysis drew on the Impaired Glucose Tolerance and Microbiota Study cohort from a community-based study of adults aged 50–65 years from the Gothenburg area, established by Fredrik Bäckhed together with Göran Bergström.

More than 1,400 participants with no diagnosed cardiovascular disease underwent extensive phenotyping, including body-composition imaging and blood-based clinical chemistry. In parallel, the researchers collected plasma metabolomics, plasma proteomics, gut-microbiome sequencing and genetic data.

“The cohort was designed to capture metabolism as an interacting biological system,” Bäckhed says, “not as a checklist of isolated risk markers.”

Can molecular signals reveal hidden risk?

With these data in hand, the team asked a counterintuitive question: could BMI itself be predicted purely based on molecular information? Using machine-learning models, Chakaroun tested whether BMI could be predicted from various data sources. Models based on dietary information performed poorly, whereas metabolite-based models consistently outperformed all others.

“We trained the models to ask a very specific question,” Chakaroun says. “Based on someone’s metabolites and microbiome, what BMI would you expect them to have?”

The resulting estimate – metBMI – was then compared with each participant’s actual BMI. The most informative signal lay not in the prediction itself but in the mismatch between the two. It was like reading the engine rather than the car’s paintwork – a view that reveals problems long before they show on the surface.

“If the BMI we calculate from metabolites is higher,” Chakaroun explains, “it tells us that this person already has metabolic changes we would normally expect to see from someone with a higher body weight.”

These residuals identified individuals whose metabolic state diverged sharply from what their body size alone would suggest – people who would otherwise pass unnoticed in routine care.

“You do not just add weight and get disease,” Chakaroun says. “Metabolism does not work in a straight line.”

Does it hold up in real people?

To ensure that the approach was not cohort-specific, the framework was validated in the independent Swedish Cardiopulmonary Bioimage Study (SCAPIS) cohort, which represents a broader population with a higher burden of cardiometabolic disease. The researchers also tested the model in a geographically independent bariatric-surgery cohort in Germany to examine whether presurgery metabolic profiles could help predict postsurgical outcomes.

“Validation across independent cohorts was essential,” says Fredrik Bäckhed.

Finally, the team assessed how well metBMI aligned with real-world disease biology by comparing it with insulin resistance, lipid profiles, inflammatory markers and other clinical risk factors.

“What mattered to us,” Chakaroun says, “was whether this framework aligned with real metabolic risk – not just with numbers on a scale.”

Together, these methods establish metBMI as a biologically grounded, data-driven measure that exposes obesity’s hidden metabolic diversity.

“Otherwise, we are still guessing risk from the outside,” Chakaroun says.

The real insight emerged when the researchers compared biology with body size directly.

When body size and disease risk diverge

When the researchers compared metBMI with measured BMI, a clear pattern emerged: BMI-aligned averages concealed profound individual differences. metBMI and BMI were strongly correlated overall, but many participants deviated sharply from the line.

“That spread is where risk is hiding,” Bäckhed says. “Averages can look reassuring, but individual biology tells a very different story.”

Individuals whose metBMI exceeded their actual BMI displayed a constellation of unfavourable metabolic features.

Importantly, metBMI correlated more strongly with markers of visceral fat quality – including surrogates of whether fat tissue was inflamed and behaving more like a diseased organ than a passive energy store – than with total fat mass alone, suggesting that how fat tissue behaves biologically may matter more than how much fat is present.

Despite having a normal or only moderately elevated BMI, these individuals showed higher insulin resistance, impaired glucose handling, adverse lipid profiles and increased inflammatory markers.

“Metabolites change early,” Chakaroun explains. “They start moving before disease is clinically visible – yet metabolically, these individuals look much more like people with obesity.”

When “healthy-looking” hides metabolic danger

Conversely, some individuals with obesity had a lower-than-expected metBMI, indicating a biologically healthier profile than BMI alone would suggest. Chakaroun is careful to frame this finding precisely.

“This does not mean obesity is harmless,” she stresses. “But it shows that the biological pathways leading to disease are not identical in everyone.”

The metabolic divergence was not driven by a single factor. Lifestyle variables such as diet and physical activity contributed to metabolic risk, whereas smoking showed no significant association – likely reflecting the low prevalence of smokers in the cohorts studied – they did not fully explain metBMI. Instead, the signal reflected coordinated shifts across hundreds of metabolites, proteins and microbial features – capturing biology that conventional lifestyle measures cannot easily resolve.

The analysis also revealed sex-specific patterns. Among men, metBMI aligned more closely with lipid metabolism and physical activity, whereas among women it was more strongly associated with inflammatory and immune-related pathways – indicating that metabolic heterogeneity follows distinct biological routes.

Roughly one quarter to one third of the circulating metabolites contributing to metBMI were linked to the gut microbiome, underscoring its central role in shaping systemic metabolic risk.

A coordinated biological shift – not random noise

Notably, some of these microbial signatures could be traced not only to gut bacteria but also to oral-associated species. As Bäckhed explains, this reflects an increased presence of aerotolerant, oral-associated bacteria in the gut – pointing to altered gut conditions among individuals with higher metBMI.

What stood out to the researchers was the consistency of these patterns across data layers.

“What surprised us was how consistent this was across omics layers,” Chakaroun says. “It really looks like a systemic shift – not just noise in the data.”

Crucially, metBMI captured risk associations that traditional measures could not. metBMI showed strong links to metabolic disease pathways but displayed no robust association with subclinical atherosclerosis, indicating specificity for metabolic dysfunction rather than cardiovascular pathology more broadly.

When metBMI was included in statistical models alongside BMI, many previously strong associations between BMI and cardiometabolic risk markers lost significance – while metBMI remained predictive.

“In a sense,” Chakaroun explains, “metBMI absorbs much of the biological information that BMI was only indirectly pointing to.”

Why the same treatment works differently

These findings held up in the independent SCAPIS validation cohort, which includes individuals with more advanced metabolic disease. Here again, metBMI separated participants into metabolically distinct subgroups with markedly different clinical risk profiles. In the bariatric-surgery cohort, presurgical metabolic signatures were linked to postsurgical outcomes, indicating potential predictive value for treatment response.

Specifically, individuals with a high metBMI lost around 30% less weight after bariatric surgery than expected based on BMI alone – highlighting that biological subtype can influence the response even to the most effective obesity treatments.

“Weight loss is not the same as metabolic improvement,” Chakaroun notes. “And different biological pathways may respond differently to the same intervention.”

Taken together, the results recast obesity not as a single condition but as a spectrum of metabolic states – some visible on the scale and others detectable only in biology.

Why obesity risk starts before weight changes

The discovery of metabolically discordant obesity states raises fundamental questions about how obesity is diagnosed, monitored and treated. For Chakaroun, the core message is not that obesity needs redefining – but that risk does.

“This is not about redefining obesity,” she says. “It is about redefining metabolic risk. Metabolic disease does not begin at a BMI cutoff,” she says. “It begins in biology.”

“The goal is to intervene before disease becomes obvious – and harder to reverse,” Chakaroun says.

Individuals with a normal BMI but elevated metabolic risk could potentially be identified years before traditional markers signal danger. Chakaroun emphasises that timing matters.

“Once diabetes or cardiovascular disease is established, intervention becomes much harder,” she explains. “That earlier window – before disease is clinically obvious – is when you can really change the trajectory.”

Nevertheless, the findings complicate the notion of “metabolically healthy obesity”. Although some individuals with obesity showed a lower metabolic burden, Chakaroun is cautious about treating this as a stable or protective state.

“You might not have metabolic complications right now,” she says, “but biology is dynamic. Trajectories can change.”

The study also raises practical and ethical questions. Deep multi-omics profiling remains costly and technically demanding, and the authors are explicit about what they are not proposing.

From complex biology to practical markers

Instead, Chakaroun sees the current work as a way to identify which signals matter most. In the study, a reduced panel of just 66 metabolites retained a substantial fraction of metBMI’s explanatory power – hinting at a more scalable path forward.

“These data help us to understand which biological signals are really important,” she explains, “and which simplified markers might eventually make sense to use in routine care.”

Another implication lies in treatment selection. If obesity reflects multiple biological pathways, a one-size-fits-all intervention may be inefficient – or even misleading.

“Different metabolic signatures may respond very differently to lifestyle changes, medication or surgery,” Chakaroun says. “Stratification could improve outcomes and also how we use healthcare resources.”

From body size to disease prediction

Beyond obesity, the framework may extend to other complex, multifactorial conditions. Chakaroun stresses that the logic is not weight-specific.

“This approach is not limited to weight,” she emphasises.

Consistent with this view, metBMI was only weakly explained by genetic risk scores, indicating that the signatures captured by the model largely reflect modifiable biology rather than genetic destiny.

“These patterns are not fixed,” says Fredrik Bäckhed. “They reflect biology we may be able to change – and a move from descriptive categories toward a more mechanistic understanding of disease.”

Nevertheless, the authors urge caution. Longitudinal studies will be needed to determine how stable metabolic subtypes are over time – and how clinicians should act on this information.

“The promise is real,” Chakaroun says, “but translation has to be evidence-driven.”

“And it has to be biologically grounded,” Bäckhed adds. “Otherwise we risk replacing one blunt tool with another – and missing the same people all over again.”