Why BMI May Not Be an Independent Multimorbidity Driver

For preventing type 2 diabetes, coronary artery disease, and stroke from piling up in the same person, blood pressure, smoking, and blood sugar still deserve center stage. A 2025 genetic analysis adds two less familiar candidates: household income and lung function. It also complicates the usual "lose weight first" hierarchy. In the study's adjusted model, body mass index (BMI) no longer had an independent association with cardiometabolic multimorbidity, while systolic blood pressure, HbA1c, smoking, household income, and forced expiratory volume in one second (FEV1) did.

That does not make BMI irrelevant, nor does it mean a weight-loss program cannot improve health. It suggests that BMI's effect on this particular composite outcome may largely run through downstream factors such as blood pressure, glycemic control, and lipids. Those are the factors clinicians can measure repeatedly and treat directly.

A genetic map of clustered cardiometabolic disease

Dun Li and colleagues' 2025 study examined cardiometabolic multimorbidity: having at least two of type 2 diabetes, coronary artery disease, and stroke. The CMM genome-wide association study included 367,147 UK Biobank participants of European ancestry. The investigators also used coronary artery disease data from 1,165,690 participants and stroke data from 1,308,460 participants.

The team screened 35 modifiable factors with two-sample Mendelian randomization (MR), then used multivariable MR to account for correlated exposures within the same risk domains. Finally, they used multi-response MR to examine whether factors appeared to affect diabetes, coronary disease, and stroke similarly or differently.

MR uses genetic variants associated with an exposure as instruments. Because people inherit those variants before adult lifestyle and disease develop, the approach can reduce some confounding and reverse causation that distort ordinary observational studies. It cannot prove that changing a trait today will produce the same result as a lifetime genetic tendency. It also depends on assumptions that can fail, especially when genetic variants influence several traits at once.

BMI and screen time looked large before adjustment

In the initial, unadjusted MR screen, 13 of 35 factors passed Bonferroni correction. Leisure screen time had the largest risk-increasing estimate for CMM, with an odds ratio of 1.255. BMI followed at 1.243, and systolic blood pressure at 1.125. Higher household income had the largest protective estimate, with an odds ratio of 0.794; educational attainment was also associated with lower risk, at 0.858. FEV1 had an odds ratio of 0.953 per standard-deviation increase.

These figures identify candidates, not independent effects. Screen time travels with movement, diet, sleep, adiposity, socioeconomic circumstances, and other traits. BMI also shares biological pathways with blood pressure, glucose regulation, and lipids. Treating each unadjusted odds ratio as a separate causal contribution would double-count a crowded metabolic pathway.

That is why the multivariable analysis matters. After adjustment for domain-level collinearity, Li et al. reported six independently significant factors, including household income, smoking, HbA1c, systolic blood pressure, and FEV1. BMI and leisure screen time did not retain independent status.

The most plausible reading is mediation: excess adiposity and sedentary behavior may raise CMM risk partly by worsening metabolic risk factors that the adjusted model already includes. A 2025 two-sample MR study by Wei et al. supports that interpretation for sedentary behavior and stroke. Its mediation analysis identified BMI, lipid markers, and glycemic indices as pathways linking sedentary behavior to stroke and its subtypes. It does not show that screen time has no effect. It shows why its apparent direct effect can shrink after a model controls for the metabolic consequences.

What the income and FEV1 findings add

Smoking cessation, blood pressure control, and glycemic management already anchor cardiometabolic prevention. The study reinforces those priorities rather than overturning them. Its more novel signals are household income and FEV1.

Household income remained independently associated with lower CMM risk even after the model accounted for smoking and systolic blood pressure. The study cannot identify the pathway. Income could affect food security, health care access, housing, working conditions, chronic stress, or several of these at once. A 2026 MR study by Kong et al. adds relevant life-course context: it examined birthweight and healthy aging, identifying socioeconomic and behavioral mediators. It does not establish income as an independent primary exposure, so it cannot confirm Li et al.'s result. It does show that socioeconomic pathways can appear in genetic analyses of long-term health.

FEV1, a standard spirometry measure of how much air a person can forcefully exhale in one second, also retained an independent protective association. That finding is emerging, not settled. A 2025 prospective UK Biobank cohort study by Xiao et al. found that sarcopenia tracked with faster progression from one cardiometabolic disease to multimorbidity. The cohort result is observational, so it does not establish a causal link between muscle loss and CMM. Still, it fits the possibility that physiological reserve captures risk beyond conventional metabolic markers.

Neither study establishes why FEV1 might matter. Systemic inflammation, hypoxic endothelial stress, and shared cardiopulmonary biology remain hypotheses, not mechanisms demonstrated here.

Why the results should not become a new risk score tomorrow

The CMM endpoint combines several disease pairings: diabetes plus coronary disease, diabetes plus stroke, coronary disease plus stroke, and all three together. Those combinations do not share identical biology. Insulin resistance may dominate one pairing, while atherothrombotic and cerebrovascular processes matter more in another. A composite outcome can hide factor-specific effects.

More important, MR estimates lifetime genetically predicted exposure. An odds ratio of 1.243 for BMI does not tell you how much a person's CMM risk will fall after losing a specified amount of weight over six months or five years. Likewise, the household-income estimate does not predict the effect of a raise, a cash transfer, financial counseling, or a less precarious job.

A 2025 review by Kjeldsen et al. describes this gap between MR and intervention evidence for cardiovascular risk factors. For blood pressure, MR estimates of benefit have sometimes exceeded effects seen in randomized trials. Clinical decisions should therefore continue to rely on trial evidence for treatments such as antihypertensive drugs, smoking cessation therapies, and glucose-lowering medications.

The same review also flags horizontal pleiotropy: genetic variants used as instruments for BMI, blood pressure, and educational attainment can influence many other traits. In Li et al.'s multivariable model, BMI's attenuation could reflect genuine mediation, but it could also partly reflect pleiotropic genetic effects being redistributed across correlated exposures. The analysis cannot fully separate those explanations.

There are other constraints. All contributing datasets were restricted to people of European ancestry, so the estimates cannot be assumed to apply to South Asian, East Asian, African, or Latin American populations. UK Biobank participants also tend to be healthier, more educated, and more affluent than the wider UK population. Sample overlap between UK Biobank exposure and outcome data can create correlated estimation errors despite statistical corrections. And the CMM outcome has not yet been independently replicated outside UK Biobank.

The practical interpretation

Use BMI as a useful upstream marker, not as the only scoreboard. Weight reduction often improves blood pressure and glucose control, and those improvements matter. But when weight, systolic blood pressure, HbA1c, and smoking status point in different directions, this study supports paying close attention to the latter three rather than assuming the scale settles the question.

• Stop smoking and use evidence-based treatment to make that realistic.
• Measure and control blood pressure.
• Track glucose status when risk warrants it, including HbA1c.
• Treat a decline in lung function as a reason to investigate health more broadly, not as proof of a specific cardiometabolic mechanism.
• Take financial instability seriously as a health exposure, while recognizing that no trial-tested CMM intervention follows directly from this MR result.

Li et al.'s study refines prevention priorities. It does not license a declaration that BMI no longer matters, nor does it turn income or spirometry into proven CMM treatments. Its strongest contribution is more modest and more useful: cardiometabolic risk may travel through measurable metabolic mediators, while socioeconomic conditions and physiological reserve deserve more attention than standard risk algorithms usually give them.