A large European study reveals hidden cardiovascular risk patterns in type 1 diabetes and shows how smarter profiling can help doctors detect complications earlier and tailor prevention.
Study: Accurate prediction of cardiovascular risk in type 1 diabetes: IMI2 SOPHIA analysis. Image credit: Chinnapong/Shutterstock.com
Cardiovascular disease (CVD) is the leading cause of death in type 1 diabetes (T1D). The presence of chronic hyperglycemia in addition to lipid abnormalities and hypertension complicates risk assessment. The following research was conducted to fill this gap. nature communications applied existing phenotype-based risk prediction tools to T1D patients and refined CVD risk stratification based on the discordance between body mass index (BMI) and cardiometabolic biomarkers.
Existing literature shows that despite good glycemic control, CVD risk remains high in T1D patients, making weight gain a major concern. In a previous study in which one of these authors participated, researchers distinguished five discordant profiles in the general population to improve detection of CVD risk.
Mismatch profile assignment in T1D
In the current study, we sought to assess whether these profiles apply to T1D by replicating this framework in a T1D population. The authors expanded their analysis to include patients with T1D. They analyzed cross-sectional data from approximately 44,000 T1D patients in three cohorts (KUL, DPV, and SIDIAP) across multiple centers in Europe. Traditional CVD markers were analyzed, including demographic, anthropometric, lifestyle, blood biomarkers, and blood pressure data.
They calculated a discrepancy score for each individual based on how closely their biomarkers matched their BMI. Each subject was then assigned a probability for each phenotype, emphasizing the continuous rather than categorical nature of this exercise. Finally, we plotted all this data using the uniform manifold approximation and projection (UMAP) method and compared the results with those of the original study.
Discordant hyperglycemic profiles common in T1D
This showed that the three profiles of concordance, hyperglycemia, and inflammation are mainly expressed in T1D, while other profiles are present at a lower frequency.
The discordant hyperglycemic phenotype accounted for 2.5% of people in the original study, compared to 55% to 76% in the T1D population. Compared to matched profiles, glycated hemoglobin (HbA1c) was higher in the hyperglycemic group, and lower HbA1c levels were associated with matched profiles. (lower risk, more similar to the general population) profile.
Current model shows selective improvement compared to traditional scoring
We then compared two sex-specific survival prediction models. One is based on SCORE2, a CVD risk stratification tool recommended by the European Society of Cardiology, which incorporates biomarkers and other CVD risk-related variables, and the other adds the probability of an assigned profile. The objective was to identify which performs better in predicting major adverse cardiac events (MACE).
We found that adding profile assignment probabilities improved predictions for specific models, outcomes, and cohorts, but not universally.
Significant likelihood ratio tests (a typical approach commonly used to compare nested models) demonstrated improved prediction of macrovascular complications in the KUL cohort, but this was limited to men.
Similarly, men in the SIDIAP cohort had improved MACE predictions, and women in the same cohort had improved extended MACE predictions. In addition, the prediction of retinopathy was enhanced for men in the KUL cohort and women in the DPV cohort.
These findings are consistent with the improvements in MACE prediction reported by the original study authors, particularly in UK men.
Comparison with other tools
By comparison, other risk prediction tools designed for diabetes, such as the UK Prospective Diabetes Study (UKPDS) (mainly developed for type 2 diabetes) and T1D-specific tools such as STENO-T1D, have been shown to underestimate CVD risk in T1D or do not include BMI, unlike the new LIFE-T1D, which also considers renal and retinal complications.
Current model is advantageous in net profit analysis
Considering the benefit to the population of using these tools, beyond just improved predictive performance, the original study showed a net benefit of using any model, including models with discordant profile data, up to a range of MACE probabilities up to 15% compared to treating no one or treating everyone (no intervention or universal intervention, respectively).
If the 10-year MACE risk was 10%, the model identified 4 more people who received appropriate treatment while avoiding 37 unnecessary interventions per 10,000 people tested.
Decision curve analysis estimates that if this model were used to test only T1D patients, this threshold would correctly perform two additional interventions (for men in the SIDIAP cohort), avoiding 5,746 unnecessary interventions per 10,000 people tested.
The specific mechanisms underlying these distinctions appear to date back to differences in fasting blood sugar levels in men and women, differences in systolic blood pressure in women, and low-density lipoproteins (LDL, “bad” cholesterol) in men. These represent potential preventive targets to reduce CVD risk in this population.
Chronic hyperglycemia in T1D may mask other associated cardiovascular risk factors and profiles, making risk stratification more difficult. Furthermore, the CVD pathway in people with good glycemic control may be different from that in people with hyperglycemia.
Advantages of this approach
Of note, this approach relies on routinely collected clinical biomarkers, does not require additional specialized testing beyond the easily obtained metric waist-hip ratio, and does not impose any additional burden on the healthcare system. However, if integrated into clinical routines, such as through publicly available digital tools, they could help clinicians decide when and how to perform CVD prevention testing in at-risk T1D patients. (e.g. https://shiny.gbiomed.kuleuven.be/UMAP_app/), especially if: Linked to electronic medical records.
Furthermore, any small improvement in the ability to accurately identify high-risk and low-risk patients is important for early identification and prevention of complications.
Potential limitations
Although cross-sectional data were used in this study, real-world data change over time and risk profiles may shift toward other phenotypes. Longitudinal studies are needed to track changes in these profiles in the general population and T1D populations.
However, emerging evidence suggests that cross-sectional data can predict outcomes as well or better than longitudinal data and are more accessible in daily clinical practice.
This study used European data, which limited generalizability.
Meaning and future direction
This result supports the validity of the original model when extended to T1D patients. We also demonstrate an association between better glycemic control and a lower risk profile, rather than establishing a direct causal reduction in CVD risk, contrary to previous reports suggesting that glycemic control alone may not influence CVD risk.
Future risk models should include glycemic control in risk allocation. Longitudinal studies could help validate risk categories based on BMI and track the evolution of risk profiles. This will help develop prevention and treatment strategies tailored to type 1 diabetes.
Click here to download your PDF copy.
Reference magazines:
- Pazmino, S., Schmid, S., Blanch, J., et al. (2026). Accurate prediction of cardiovascular risk in type 1 diabetes: IMI2 SOPHIA analysis. Nature Communications. Toi: https://doi.org/10.1038/s41467-026-72029-z. https://www.nature.com/articles/s41467-026-72029-z
