Metabolomic profiling refines cardiovascular risk stratification beyond SCORE2-diabetes in patients with concurrent type 2 diabetes and pulmonary dysfunction.
Li M, Shen Y, et al. • Frontiers in endocrinology • 2026
Adding a 12-metabolite signature to SCORE2-Diabetes produced a statistically significant but moderate improvement in 10-year MACE risk discrimination in patients with concurrent type 2 diabetes and lung function impairment (C-index increase from 0.670 to 0.722), though the authors caution these internally validated findings do not establish immediate clinical usefulness.
Key Findings
Results
SCORE2-Diabetes showed lower discrimination for MACE in patients with lung function impairment compared to those with preserved lung function.
C-index was 0.670 in the lung function impairment (LFI) cohort versus 0.707 in the reference cohort with preserved lung function.
The reference cohort comprised 12,130 UK Biobank participants with T2D and preserved lung function.
The primary LFI cohort comprised 3,706 UK Biobank participants with T2D and lung function impairment.
Median follow-up was 12 years.
845 of 3,706 LFI participants developed MACE during follow-up.
Results
Adding a 12-metabolite machine learning-derived signature to SCORE2-Diabetes significantly improved MACE discrimination in the LFI cohort.
C-index increased from 0.670 to 0.722 with addition of the 12-metabolite signature (absolute difference 0.052; P < 0.001).
The improvement was described as 'statistically significant but moderate.'
249 plasma metabolites were profiled using nuclear magnetic resonance (NMR) spectroscopy.
The machine-learning framework combined LASSO Cox regression, random forest survival analysis, and extreme gradient boosting (XGBoost) to derive a consensus signature.
Internal validation used 1,000 bootstrap resamples with the complete modeling pipeline repeated within each resample.
Results
Reduced metabolite models also showed improved discrimination over SCORE2-Diabetes alone, though with progressively lower C-indices.
A five-metabolite model achieved a C-index of 0.708.
A two-metabolite sensitivity model yielded a C-index of 0.699 (95% CI, 0.683–0.716).
All reduced models exceeded the base SCORE2-Diabetes C-index of 0.670 in the LFI cohort.
Results
Adding the metabolomic signature to SCORE2-Diabetes produced meaningful risk reclassification using 2023 ESC SCORE2-Diabetes categories.
Categorical net reclassification improvement (NRI) was 14.8% (95% CI, 10.8%–18.8%).
Reclassification was assessed using the 2023 European Society of Cardiology SCORE2-Diabetes risk categories.
Internal calibration and decision curve analysis (DCA) suggested improved agreement and net benefit.
Conclusions
The study was conducted in UK Biobank participants with T2D and no cardiovascular disease at baseline, and findings require external validation before clinical application.
The authors explicitly state that internal calibration and DCA results 'require external validation.'
The authors conclude these findings 'do not establish immediate clinical usefulness.'
The study design was prospective observational using UK Biobank data.
The authors call for 'further evaluation in independent cohorts and prospective impact studies.'
What This Means
This research suggests that a widely used cardiovascular risk calculator called SCORE2-Diabetes is less accurate at predicting heart attacks and strokes over 10 years in people who have both type 2 diabetes and impaired lung function, compared to people with diabetes alone. In a large study of UK Biobank participants followed for about 12 years, roughly 1 in 4 people with both conditions experienced a major cardiovascular event, highlighting how vulnerable this group is. The standard risk score had a meaningfully lower accuracy (C-index 0.670) in the lung-impaired group than in those with normal lung function (C-index 0.707).
To try to improve prediction, the researchers used advanced machine learning techniques to identify patterns among 249 blood metabolites — small molecules detectable by a specialized blood test — that were linked to future cardiovascular events. They identified a 12-metabolite 'signature' that, when added to the standard risk score, improved prediction accuracy (C-index rising to 0.722) and correctly reclassified nearly 15% of patients into more appropriate risk categories. Simpler versions using only 5 or 2 metabolites also outperformed the standard score alone, suggesting some flexibility in potential implementation.
This research suggests that routine metabolite blood testing could one day help doctors better identify which patients with diabetes and lung problems are at highest cardiovascular risk. However, the authors are careful to note that these findings come from a single dataset with internal validation only, and the results must be confirmed in separate, independent patient populations before any clinical use could be considered. The study does not provide evidence that acting on this improved risk classification would change patient outcomes.
Li M, Shen Y, Huang Y, Liu L, Bin W. (2026). Metabolomic profiling refines cardiovascular risk stratification beyond SCORE2-diabetes in patients with concurrent type 2 diabetes and pulmonary dysfunction.. Frontiers in endocrinology. https://doi.org/10.3389/fendo.2026.1933396