Precision cardiovascular risk prediction in type 1 diabetes: An IMI2 SOPHIA analysis

Precision cardiovascular risk prediction in type 1 diabetes: An IMI2 SOPHIA analysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Brief Communication Precision cardiovascular risk prediction in type 1 diabetes: An IMI2 SOPHIA analysis Bart Van der Schueren, Sofia Pazmino, Stefanie Schmid, Jordi Blanch, and 18 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6429567/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Cardiovascular disease (CVD) is a major long-term complication and the leading cause of morbidity and mortality among individuals with type 1 diabetes (T1D), with a substantially higher prevalence compared to the general population and driven by multiple interrelated risk factors—underscoring the urgent need for accurate risk assessment. To support more tailored approaches to CVD prevention, a recent article in Nature Medicine showed five discordant phenotypic risk profiles for CVD. These were identified in the general population in Europe with diverse relationship between body mass index and cardiometabolic biomarkers. Here, we explore their applicability in 24,323 people with T1D. Improved glycemic control was linked to a decrease in CVD risk as people with T1D with lower glycated hemoglobin belonged to the baseline concordant cluster. This supports the contention that glycemic control in people with T1D is an integral part of lowering CVD risk. Health sciences/Endocrinology/Endocrine system and metabolic diseases/Diabetes/Type 1 diabetes Health sciences/Risk factors Figures Figure 1 Figure 2 Main People living with type 1 diabetes (T1D) have a higher risk for developing cardiovascular disease (CVD), resulting in a decreased life-expectancy of about 10 years when compared to the general population. 1 , 2 CVD assessment in T1D is challenging—not only do traditional risk factors like hypertension and dyslipidemia apply, but chronic hyperglycemia also plays a role. However, the extent of its detrimental effect is still unclear, as CVD risk remains elevated even when good glycemic control is achieved. 2 , 3 Therefore, the observation that weight gain is on the raise in people living with T1D is of great concern, 4 , 5 as it may further increase their CVD risk. In order to facilitate precision prevention of CVD, Coral et al 6 identified five discordant profiles in the general population, where cardiometabolic risk differs from what is typically expected based on body mass index (BMI). Here, we aim to briefly put the study of Coral et al 6 into the perspective of people living with T1D to explore how overweight and obesity worsens their already elevated CVD risk. International/multi-center cross-sectional data from adults (> 35 years) living with T1D, followed between 2010 and 2022 were used for this replication study. Data sources were the University Hospital Leuven (KUL) in Belgium 4 , the German/Austrian/Luxembourgian/Swiss diabetes prospective (DPV) follow-up registry 7 and the Jordi Gol I Gurina Foundation (SIDIAP) 8 in Spain. The same biomarkers traditionally associated with CVD risk were included: age, sex, smoking, BMI, fasting glucose, lipids (HDL, LDL, triglycerides), blood pressure, serum creatinine, ALT, CRP and waist/hip ratio. The first visit with the most information available for all biomarkers was used. Missing data was imputed using information from similar patients with k-nearest neighbor algorithm. KUL included a population of 1,663 individuals (51% female, 49% male), DPV included 9,490 individuals (53% female, 47% male) and SIDIAP had a population of 13,170 (58% female and 42% male) used in this analysis. An overview of the analysis pipeline has been previously published. 6 Briefly, to replicate the identified subgroups 6 , we calculated discordant scores based on the expected population-level associations between BMI and each biomarker, adjusted for age and smoking, generated from the UK Biobank (UKB). We then computed allocation probabilities into the concordant and discordant profiles described by Coral et al. 6 , and visualized these allocations in relation to the concordant and discordant profiles using the Uniform Manifold Approximation and Projection (UMAP) embedding. Three risk profiles were mostly represented in the population with T1D: concordant, hyperglycemic and inflammatory (Fig. 1 ). The majority of individuals (55–76%) belong to the discordant hyperglycemic profile (Fig. 1 d., e. and f.), while this cluster is the smallest (~ 2.5%) in the general population analysis done by Coral et al. 6 Furthermore, in the population with T1D, mean HbA1c% was lower in the concordant cluster (mean ± standard deviation 6.9 ± 0.9/7.1 ± 1.0 in KUL and 7.1 ± 1.6/7.3 ± 1.5 in DPV for females/males, respectively) than in the hyperglycemic (7.8 ± 1.0/7.8 ± 1.3 in UZ Leuven and 7.9 ± 1.5/8.0 ± 1.5 in DPV for females/males respectively) and inflammatory (7.4 ± 1.1/7.7 ± 1.0 in KUL for females/males respectively) profiles. Same as Coral et al. 6 , we compared two types of sex-specific survival prediction models: one based solely on the previously mentioned biomarkers with variables and interactions used in the current CVD risk stratification tool endorsed by the European Society of Cardiology (SCORE2) 9 , 10 and another that also included profile allocation probabilities, to assess whether adding profile information would improve the prediction of major adverse cardiac events (MACE). Through comparison of nested models, we showed that adding profile allocation probabilities improved the predictive capability of these models, reflected by significant likelihood ratio tests and difference in C-statistics (Supplemental Table 1). This improvement was also reported by Coral et al 1 , particularly for men in UKB. Other risk engines specifically for diabetes, like the United Kingdom Prospective Diabetes Study (UKPDS) 11 and even T1D-specific engines such as STENO-T1D 12 have been shown to underestimate the risk of CVD in T1D 11 or do not include BMI. 12 Nevertheless, the newer LIFE-T1D 13 risk score that contains similar variables to SCORE2 with more focus on kidney health (albuminuria), concomitant complications (retinopathy) and age of T1D diagnosis has considered BMI as an important predictor. Many medical decisions involve balancing trade-offs, such as treating patients to prevent disease-related harm while minimizing the risk of treatment side effects, or diagnosing patients with a disease while avoiding unnecessary testing for healthy individuals. Net benefit is a decision-analytic measure that is increasingly used to quantify both benefits and harms on a common scale. To assess the additional net benefit of discordant profiles, decision curves were used to evaluate whether implementing interventions to prevent CVD would be beneficial (Fig. 2 ). Coral et al 6 showed that models with and without discordant profile data, generally performed better than no intervention or universal intervention across various MACE probability thresholds up to 15%. It also showed that adding profile information at a 10% 10-year MACE risk threshold resulted in an average net benefit of 4 additional correct interventions and 37 additional unnecessary interventions correctly avoided per 10,000 individuals tested. 3 In a population with T1D, at the same threshold of 10%, the net benefit for MACE was of 16 additional correct interventions while correctly avoiding 1,274 unnecessary interventions for every 10,000 individuals tested (average of DPV and SIDIAP-Supplemental Table 2). Especially for T1D, this can become a way forward to aid in decision-making for CVD prevention, since the risk profile information was proven to be more impactful in T1D than in the general population. Finally, to explore how the relationship of BMI–biomarker discordance in this population with T1D differs from the general population, sex-specific linear models for the residuals were fitted while adjusting for age, smoking status and disease duration before clustering. In terms of starting points (intercept), the main differences were present in fasting glucose for both males and females. Additionally in females, differences in systolic blood pressure and ALT were seen, while in males, also LDL was found altered (Supplemental Table 3). These biomarkers could prove interesting targets for reducing CVD risk in people living with T1D with potentially greater effects than in the general population. While the risk profiles identified by Coral et al. 6 are well-suited for the general population, as expected, there is an overexpression of the hyperglycemic profile in individuals with T1D. Furthermore, better glycemic control (HbA1c) was associated with a lower CVD risk profile, such as the baseline concordant cluster, which more closely aligns with the general population. This highlights the critical importance of maintaining glycemic control in T1D management 14 as a key factor in effective CVD risk control. Also, it is in contradiction with previous findings stating that better glycemic control does not lower CVD. 15 On the other hand, our results point to chronic hyperglycemia masking other relevant risk profiles or biomarkers in people living with T1D, further hinting at different pathways to CVD depending on glycemic control 3 and mediated by traditional risk factors over time -disease duration-. 16 , 17 Future research in T1D would benefit from considering models based not only on typical CVD risk factors but also approaches that account for the impact of glucose control before allocating/clustering into risk profiles. It could also explore the use of longitudinal data to further validate BMI- subclassifications and their evolution in diverse cohorts. Understanding the heterogeneity of diabetes will be a key step towards precision medicine 18 , particularly for CVD prevention in T1D. Methods Study Cohorts University Hospital Leuven (KUL) Patient data were obtained from the type 1 diabetes (T1D) registry of University Hospital Leuven, 4 a tertiary care center located in the Flemish Brabant region -northern Belgium. All individuals with a clinical diagnosis of T1D who received care at this hospital were eligible for inclusion in the registry. A confirmed diagnosis by a healthcare provider was required. German/Austrian/Luxembourgian/Swiss diabetes prospective (DPV) follow-up registry This is a multicenter database initiated in Germany in 1995 to standardize the documentation of demographic and diabetes-related clinical data from specialized care centers and has now expanded to include centers in Austria, Luxembourg, and Switzerland. 7 Clinical and demographic data are entered locally using the standardized DPV electronic health record system. Participating centers submit anonymized data biannually to Ulm University, Germany, where they undergo plausibility checks. Any discrepancies identified are returned to the clinics for clarification or correction prior to data consolidation. The anonymized dataset is utilized for clinical research and quality monitoring purposes ( www.d-p-v.eu ). The DPV initiative has received ethical approval from the ethics committee of Ulm University, as well as the respective review boards of all participating institutions. 7 For this study, data from adult (> 35 years) individuals diagnosed with type 1 diabetes were analyzed. Jordi Gol I Gurina Foundation (SIDIAP) The Information System for Research in Primary Care (SIDIAP) 8 is a comprehensive database of electronic health records from primary care, designed to support healthcare research. It compiles data from 328 primary care centers operated by the Catalan Health Institute across Catalonia, Spain. Since 2006, SIDIAP has collected pseudo-anonymized health records for over 8 million individuals, covering approximately 75% of the Catalan population. Statistical analysis Data Preparation Same as Coral et al, 6 we included 13 biomarkers across all cohorts: fasting glucose (FG) in mmol/l; lipid fractions (HDL, LDL, triglycerides) in mmol/l; systolic and diastolic blood pressure in mmHg; serum creatinine in µmol/l; alanine aminotransferase (ALT) in U/l; C-reactive protein (CRP) in mg/l; waist-to-hip ratio in cm; age in years; smoking status (1 = current smoker, 0 = non-smoker); and sex (male/female). Units were harmonized across datasets. No BMI thresholds were applied. Outliers exceeding 5 standard deviations from the mean were excluded prior to analysis. All analyses were stratified by sex to account for known gender-related differences in BMI, biomarkers, and cardiometabolic risk. UMAP Projection and Profile Identification This cluster analysis follows a multi-step analytical pipeline 6 that combines linear modeling, residual extraction, dimensionality reduction, and unsupervised clustering to examine biomarker data. The main goals are to evaluate the relationship between a set of biomarkers and BMI (adjusting for additional covariates), and to identify latent subgroups based on the residual patterns from these models. The analysis is conducted separately for males and females. Biomarker residuals were projected into two dimensions using the umap function from the R package uwot (v0.1.16), with nearest neighbor count set as a function of cohort size. The parameter binary_edge_weights was set to TRUE, implementing PacMAP, a UMAP variant that better preserves both global and local data structures. Additionally, we used dens_scale = 1 to enable densMAP, which enhances density preservation. Clusters were identified by constructing a proximity network and applying the Leiden community detection algorithm after seeding with the leading eigenvector method (igraph, v2.0.2). The algorithm, run over 500 iterations, optimized modularity through node movement, partition refinement, and network aggregation. Eigen centrality scores were calculated per individual and used to determine cluster centres and covariance matrices within a Gaussian mixture model. To account for instability of central clusters, a concordant distribution (zero mean, identity covariance) was added. Individuals with high overlap between discordant and concordant profiles were assigned to the latter, improving cluster specificity. The final model provided both hard cluster assignments and profile probability scores for each individual. Model Comparisons We used likelihood ratio tests and changes in C-statistics to compare nested models. Variance explained by discordant profiles was quantified using likelihood-based methods, unaffected by threshold choices. We also performed decision curve analyses to assess clinical utility, quantifying net benefit (true positives) and net interventions avoided (true negatives) across varying probability thresholds, both per cohort and pooled across studies 19 . Predictive Value of Profiles To assess added predictive value, we compared nested regression models with and without discordant profile information. Using profile allocation probabilities in regression models is problematic due to their sum-to-one constraint. Coral et al 6 addressed this by using the log-contrast approach from compositional data analysis, using the concordant profile as reference. For each individual, discordant profile probabilities were divided by the concordant probability, and their natural logarithms were used as predictors. Nested Cox Regression Models We evaluated the association between profile allocations and risk of MACE using nested Cox proportional hazards models over 10-year follow-up periods. The predictors included variables from the SCORE2 algorithm (validated in diabetic populations), (see Supplementary Table 1). Individuals with pre-existing CVD were excluded. Phenotypic Discordance with BMI in T1D We examined the association between BMI and each biomarker, adjusting for age and smoking status and disease duration, using linear regression models (see Supplementary Table 2). Declarations Data availability KUL. The data that support the findings of this study are available on request from the corresponding author. DPV. The datasets generated and analysed during the current study are not publicly available. Due to data protection reasons, data on individual level cannot be provided. However, we can provide remote access to aggregated data if requested . SIDIAP. Any researcher is able to request SIDIAP data to conduct a study. A five-step procedure takes place before data access is granted: (i) the researcher(s) must send an application (standardized form available at www.sidiap.org and study protocol) to the SIDIAP team; (ii) the application is approved by SIDIAP’s Scientific Committee which evaluates the scientific quality and feasibility of the proposal; (iii) the study protocol is approved by the Clinical Research Ethics Committee of IDIAPJGol; (iv) the principal investigator of the study must sign a Good Practice form and, in some cases, an agreement between parties is needed; and (v) a meeting between the research team and the SIDIAP team is arranged to discuss the procedures and set the data extraction. Further information is available online (https://www.sidiap.org/index.php/menu-solicitudesen/application-proccedure) or by contacting Anna Moleras ( [email protected] ). Data access is limited to researchers from public organizations and collaboration with private institutions is possible when a study is required by a regulatory agency or for non-commercial studies within a European project financed by the European Commission. In accordance with current European and national law, the data used in this study are only available for the researchers participating in this study. Thus, we are not allowed to distribute or make publicly available the data to other parties. Code availability All analyses were performed in programming language R v.4.4 (https://www.r-project.org/). The scripts used can be found at https://github.com/danielcoral/SOPHIA_Cross_Sectional. ACKNOWLEDGEMENTS We also thank Carlos Marin P. for providing medical writing support and editorial assistance. FUNDING This work is part of the Stratification of Obesity Phenotypes to Optimize Future Obesity Therapy (SOPHIA) project (www.imisophia.eu). SOPHIA has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No. 875534, supported by the European Union’s Horizon 2020 research and innovation program and EFPIA, with additional support from T1D Exchange, Breakthrough T1D (former JDRF), and Obesity Action Coalition. COMPETING INTERESTS CM serves or has served on the advisory panel for Novo Nordisk, Sanofi, Eli Lilly and Company, Novartis, Boehringer Ingelheim, Bayer, Roche, Medtronic, Imcyse, Insulet, Biomea Fusion and Vertex. Financial compensation for these activities has been received by KU Leuven; KU Leuven has received research support for CM from Medtronic, Imcyse, Novo Nordisk, Sanofi and ActoBio Therapeutics; CM serves or has served on the Speaker’s bureau for Novo Nordisk, Sanofi, Eli Lilly and Company, Medtronic and Boehringer Ingelheim. Financial compensation for these activities has been received by KU Leuven. CM is president of EASD. All external support of EASD is to be found on www.easd.org. References Ahuja A, Agrawal S, Acharya S, Reddy V, Batra N. Strategies for Cardiovascular Disease Prevention in Type 1 Diabetes: A Comprehensive Review. Cureus 2024; 16 (8): e66420. Vergès B. Cardiovascular disease in type 1 diabetes, an underestimated danger: Epidemiological and pathophysiological data. Atherosclerosis 2024; 394 : 117158. Miller RG, Orchard TJ, Costacou T. 30-Year Cardiovascular Disease in Type 1 Diabetes: Risk and Risk Factors Differ by Long-term Patterns of Glycemic Control. Diabetes Care 2022; 45 (1): 142-50. Al Ozairi E, Steenackers N, Pazmino S, et al. Prevalence of obesity in people with and without type 1 diabetes across Belgium, Kuwait, and Mexico: an IMI2 SOPHIA study. eClinicalMedicine 2024; 77 : 102869. Conway B, Miller RG, Costacou T, Fried L, Kelsey S, Evans RW, Orchard TJ. Temporal patterns in overweight and obesity in Type 1 diabetes. Diabet Med 2010; 27 (4): 398-404. Coral DE, Smit F, Farzaneh A, et al. Subclassification of obesity for precision prediction of cardiometabolic diseases. Nature Medicine 2024. Grimsmann JM, Tittel SR, Bramlage P, et al. Disease heterogeneity of adult diabetes based on routine clinical variables at diagnosis: Results from the German/Austrian Diabetes Follow-up Registry. Diabetes Obes Metab 2022; 24 (11): 2253-62. Recalde M, Rodríguez C, Burn E, et al. Data Resource Profile: The Information System for Research in Primary Care (SIDIAP). Int J Epidemiol 2022; 51 (6): e324-e36. SCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe. Eur Heart J 2021; 42 (25): 2439-54. SCORE2-Diabetes: 10-year cardiovascular risk estimation in type 2 diabetes in Europe. Eur Heart J 2023; 44 (28): 2544-56. Zgibor JC, Ruppert K, Orchard TJ, Soedamah-Muthu SS, Fuller J, Chaturvedi N, Roberts MS. Development of a coronary heart disease risk prediction model for type 1 diabetes: the Pittsburgh CHD in Type 1 Diabetes Risk Model. Diabetes Res Clin Pract 2010; 88 (3): 314-21. Vistisen D, Andersen GS, Hansen CS, Hulman A, Henriksen JE, Bech-Nielsen H, Jørgensen ME. Prediction of First Cardiovascular Disease Event in Type 1 Diabetes Mellitus: The Steno Type 1 Risk Engine. Circulation 2016; 133 (11): 1058-66. Helmink MAG, Hageman SHJ, Eliasson B, et al. Lifetime and 10-year cardiovascular risk prediction in individuals with type 1 diabetes: The LIFE-T1D model. Diabetes Obes Metab 2024; 26 (6): 2229-38. Risk Factors for Cardiovascular Disease in Type 1 Diabetes. Diabetes 2016; 65 (5): 1370-9. Orchard TJ, Olson JC, Erbey JR, et al. Insulin resistance-related factors, but not glycemia, predict coronary artery disease in type 1 diabetes: 10-year follow-up data from the Pittsburgh Epidemiology of Diabetes Complications Study. Diabetes Care 2003; 26 (5): 1374-9. Miller RG, Costacou T, Orchard TJ. Risk Factor Modeling for Cardiovascular Disease in Type 1 Diabetes in the Pittsburgh Epidemiology of Diabetes Complications (EDC) Study: A Comparison With the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study (DCCT/EDIC). Diabetes 2019; 68 (2): 409-19. Bebu I, Braffett BH, Pop-Busui R, Orchard TJ, Nathan DM, Lachin JM. The relationship of blood glucose with cardiovascular disease is mediated over time by traditional risk factors in type 1 diabetes: the DCCT/EDIC study. Diabetologia 2017; 60 (10): 2084-91. Cefalu WT, Franks PW, Rosenblum ND, et al. A global initiative to deliver precision health in diabetes. Nat Med 2024; 30 (7): 1819-22. Vickers AJ, Cronin AM, Elkin EB, Gonen M. Extensions to decision curve analysis, a novel method for evaluating diagnostic tests, prediction models and molecular markers. BMC Medical Informatics and Decision Making 2008; 8 (1): 53. Additional Declarations Yes there is potential Competing Interest. CM serves or has served on the advisory panel for Novo Nordisk, Sanofi, Eli Lilly and Company, Novartis, Boehringer Ingelheim, Bayer, Roche, Medtronic, Imcyse, Insulet, Biomea Fusion and Vertex. Financial compensation for these activities has been received by KU Leuven; KU Leuven has received research support for CM from Medtronic, Imcyse, Novo Nordisk, Sanofi and ActoBio Therapeutics; CM serves or has served on the Speaker’s bureau for Novo Nordisk, Sanofi, Eli Lilly and Company, Medtronic and Boehringer Ingelheim. Financial compensation for these activities has been received by KU Leuven. CM is president of EASD. All external support of EASD is to be found on www.easd.org . Supplementary Files SupplementalTables.docx Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6429567","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Brief Communication","associatedPublications":[],"authors":[{"id":449827138,"identity":"e8d25b3c-9c78-4e81-88a0-11f0fdf05043","order_by":0,"name":"Bart Van der Schueren","email":"data:image/png;base64,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","orcid":"","institution":"University of Leuven","correspondingAuthor":true,"prefix":"","firstName":"Bart","middleName":"Van der","lastName":"Schueren","suffix":""},{"id":449827139,"identity":"240f6b3b-c845-4406-92e7-83eec8dd993f","order_by":1,"name":"Sofia Pazmino","email":"","orcid":"","institution":"Department of Chronic Diseases and Metabolism, Clinical and Experimental Endocrinology, KU Leuven, Leuven, Belgium.","correspondingAuthor":false,"prefix":"","firstName":"Sofia","middleName":"","lastName":"Pazmino","suffix":""},{"id":449827140,"identity":"67e42cf0-6d6a-4b0f-a447-a24159dc2c9f","order_by":2,"name":"Stefanie Schmid","email":"","orcid":"","institution":"University of Ulm","correspondingAuthor":false,"prefix":"","firstName":"Stefanie","middleName":"","lastName":"Schmid","suffix":""},{"id":449827141,"identity":"f0e3b1bf-1a8e-4294-9fc4-e279664ba3d0","order_by":3,"name":"Jordi Blanch","email":"","orcid":"","institution":"Institut d’Investigació Biomèdica de Girona Dr. Josep Trueta","correspondingAuthor":false,"prefix":"","firstName":"Jordi","middleName":"","lastName":"Blanch","suffix":""},{"id":449827142,"identity":"a018f294-d5fd-4a16-bc5b-52258e1fad46","order_by":4,"name":"Daniel Coral","email":"","orcid":"https://orcid.org/0000-0002-8809-4544","institution":"Lund University","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"","lastName":"Coral","suffix":""},{"id":449827143,"identity":"1ef6571b-c34a-4a00-b8db-a25f72725889","order_by":5,"name":"Nele Steenackers","email":"","orcid":"","institution":"Maastricht University","correspondingAuthor":false,"prefix":"","firstName":"Nele","middleName":"","lastName":"Steenackers","suffix":""},{"id":449827144,"identity":"eff9603b-d6bf-4977-938b-e117bb7d30a3","order_by":6,"name":"Nicole Forestier","email":"","orcid":"","institution":"Diabetes centre Bad Aibling","correspondingAuthor":false,"prefix":"","firstName":"Nicole","middleName":"","lastName":"Forestier","suffix":""},{"id":449827145,"identity":"c8278daa-cb65-405a-b53e-34b14fe73242","order_by":7,"name":"Laura Gallardo Nuel","email":"","orcid":"","institution":"Institut d’Investigació Biomèdica de Girona (IDIBGI)","correspondingAuthor":false,"prefix":"","firstName":"Laura","middleName":"Gallardo","lastName":"Nuel","suffix":""},{"id":449827146,"identity":"4a9d5c25-82f0-4a2a-9e9d-e6421832df98","order_by":8,"name":"Naichhuan Zhang","email":"","orcid":"","institution":"University of Leuven","correspondingAuthor":false,"prefix":"","firstName":"Naichhuan","middleName":"","lastName":"Zhang","suffix":""},{"id":449827147,"identity":"7949608f-331f-48f5-bc09-ba96ef9d57e2","order_by":9,"name":"Birgit Reinhart-Steininger","email":"","orcid":"","institution":"Medical University of Vienna","correspondingAuthor":false,"prefix":"","firstName":"Birgit","middleName":"","lastName":"Reinhart-Steininger","suffix":""},{"id":449827148,"identity":"8850a16e-df18-4953-b5e7-cf6b1427dbc3","order_by":10,"name":"Yenny Leal","email":"","orcid":"","institution":"IDIBGI-CERCA, University of Girona","correspondingAuthor":false,"prefix":"","firstName":"Yenny","middleName":"","lastName":"Leal","suffix":""},{"id":449827149,"identity":"4bbb3a2d-0562-4874-a761-4a2e0e18e5d1","order_by":11,"name":"Amar van Laar","email":"","orcid":"","institution":"University of Leuven","correspondingAuthor":false,"prefix":"","firstName":"Amar","middleName":"van","lastName":"Laar","suffix":""},{"id":449827150,"identity":"37a3d2f4-dc93-4024-b05e-265b08a7586a","order_by":12,"name":"Rafael Ramos","email":"","orcid":"","institution":"Girona Biomedical Research Institute (IdiBGi)","correspondingAuthor":false,"prefix":"","firstName":"Rafael","middleName":"","lastName":"Ramos","suffix":""},{"id":449827151,"identity":"8b69d43b-858f-45a4-8958-0c572155d741","order_by":13,"name":"Shreekar Bharadwaj Araveti","email":"","orcid":"","institution":"University of Leuven","correspondingAuthor":false,"prefix":"","firstName":"Shreekar","middleName":"Bharadwaj","lastName":"Araveti","suffix":""},{"id":449827152,"identity":"d39d5483-83e7-4281-955a-418bf8390146","order_by":14,"name":"Kinga Nowak","email":"","orcid":"","institution":"University of Leuven","correspondingAuthor":false,"prefix":"","firstName":"Kinga","middleName":"","lastName":"Nowak","suffix":""},{"id":449827153,"identity":"3a8ebc8e-38dc-44ff-b4a2-acdd5346c6b8","order_by":15,"name":"Jonathan Rosen","email":"","orcid":"","institution":"Breakthrough T1D","correspondingAuthor":false,"prefix":"","firstName":"Jonathan","middleName":"","lastName":"Rosen","suffix":""},{"id":449827154,"identity":"76301d50-ef26-4e4b-983f-5b030947699a","order_by":16,"name":"Carmen Hurtado del Pozo","email":"","orcid":"","institution":"Breakthrough T1D","correspondingAuthor":false,"prefix":"","firstName":"Carmen","middleName":"Hurtado del","lastName":"Pozo","suffix":""},{"id":449827155,"identity":"f85773ab-06de-4daa-b230-0f751d897d3b","order_by":17,"name":"Christian-Dominik Möller","email":"","orcid":"","institution":"Bürgerhospital Frankfurt am Main","correspondingAuthor":false,"prefix":"","firstName":"Christian-Dominik","middleName":"","lastName":"Möller","suffix":""},{"id":449827156,"identity":"1201de56-86be-4919-8657-e78d8d7b3f65","order_by":18,"name":"Chantal Mathieu","email":"","orcid":"","institution":"Katholieke Universiteit Leuven","correspondingAuthor":false,"prefix":"","firstName":"Chantal","middleName":"","lastName":"Mathieu","suffix":""},{"id":449827157,"identity":"254c393c-965b-4e2e-af34-0d381578c66f","order_by":19,"name":"Paul Franks","email":"","orcid":"https://orcid.org/0000-0002-0520-7604","institution":"Lund University","correspondingAuthor":false,"prefix":"","firstName":"Paul","middleName":"","lastName":"Franks","suffix":""},{"id":449827158,"identity":"79bcb4e2-bad1-4d3a-baf1-d58a8eaa7c47","order_by":20,"name":"Jose-Manuel Fernandez-Real","email":"","orcid":"https://orcid.org/0000-0002-7442-9323","institution":"Department of Endocrinology, Diabetes and Nutrition, Hospital of Girona “Dr Josep Trueta”, Departament de Ciències Mèdiques, Universitat of Girona and CIBERobn Pathophysi","correspondingAuthor":false,"prefix":"","firstName":"Jose-Manuel","middleName":"","lastName":"Fernandez-Real","suffix":""},{"id":449827159,"identity":"21e383d9-5dc0-4207-be2e-76d9843917b5","order_by":21,"name":"Stefanie Lanzinger","email":"","orcid":"","institution":"University of Ulm","correspondingAuthor":false,"prefix":"","firstName":"Stefanie","middleName":"","lastName":"Lanzinger","suffix":""}],"badges":[],"createdAt":"2025-04-11 15:20:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6429567/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6429567/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":83649124,"identity":"978e0ea0-4340-4c00-a280-fab64af9b00f","added_by":"auto","created_at":"2025-05-30 06:37:22","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":220782,"visible":true,"origin":"","legend":"\u003cp\u003eUniform Manifold Approximation and Projection for Dimension Reduction (UMAP) 2D projections: the population with T1D from a) KU Leuven, b) DPV and c) SIDIAP are in colours projected on the general population in black. Percentage of population per cluster in d) KU Leuven, e) DPV and f) SIDIAP. The colours denote profile allocations: baseline concordant -BC-, discordant hypertensive -DHT-, discordant adverse lipid profile -DAL-, discordant liver transaminase -DLT-, discordant inflammatory -DIS- and discordant hyperglycemic -DHG- profiles\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6429567/v1/a3c41e9302eda916987bfd0e.png"},{"id":83649780,"identity":"6964a604-ad74-4daa-8802-1506bcc27395","added_by":"auto","created_at":"2025-05-30 06:45:22","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":104818,"visible":true,"origin":"","legend":"\u003cp\u003eDecision curves comparing the net benefit of using various strategies at different thresholds of disease probability up to 25%.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6429567/v1/5c5a9a2ca3293e2eea5daf0d.png"},{"id":83649792,"identity":"e1bf0f0a-6546-465b-83cf-60a4ec2add4d","added_by":"auto","created_at":"2025-05-30 06:45:27","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":847369,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6429567/v1/ab76ab07-5ae2-4ec4-a390-87e68b6b7ffe.pdf"},{"id":83649120,"identity":"8a86f365-fc2d-4c73-881b-41f4970af9a6","added_by":"auto","created_at":"2025-05-30 06:37:22","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":34414,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTables.docx","url":"https://assets-eu.researchsquare.com/files/rs-6429567/v1/cced5f1e51ed1f724681b0d5.docx"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nCM serves or has served on the advisory panel for Novo Nordisk, Sanofi, Eli Lilly and Company, Novartis, Boehringer Ingelheim, Bayer, Roche, Medtronic, Imcyse, Insulet, Biomea Fusion and Vertex. Financial compensation for these activities has been received by KU Leuven; KU Leuven has received research support for CM from Medtronic, Imcyse, Novo Nordisk, Sanofi and ActoBio Therapeutics; CM serves or has served on the Speaker’s bureau for Novo Nordisk, Sanofi, Eli Lilly and Company, Medtronic and Boehringer Ingelheim. Financial compensation for these activities has been received by KU Leuven. CM is president of EASD. All external support of EASD is to be found on www.easd.org.","formattedTitle":"Precision cardiovascular risk prediction in type 1 diabetes: An IMI2 SOPHIA analysis","fulltext":[{"header":"Main","content":"\u003cp\u003ePeople living with type 1 diabetes (T1D) have a higher risk for developing cardiovascular disease (CVD), resulting in a decreased life-expectancy of about 10 years when compared to the general population.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e CVD assessment in T1D is challenging\u0026mdash;not only do traditional risk factors like hypertension and dyslipidemia apply, but chronic hyperglycemia also plays a role. However, the extent of its detrimental effect is still unclear, as CVD risk remains elevated even when good glycemic control is achieved.\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e Therefore, the observation that weight gain is on the raise in people living with T1D is of great concern,\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e as it may further increase their CVD risk. In order to facilitate precision prevention of CVD, Coral et al\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e identified five discordant profiles in the general population, where cardiometabolic risk differs from what is typically expected based on body mass index (BMI). Here, we aim to briefly put the study of Coral et al\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e into the perspective of people living with T1D to explore how overweight and obesity worsens their already elevated CVD risk.\u003c/p\u003e \u003cp\u003eInternational/multi-center cross-sectional data from adults (\u0026gt;\u0026thinsp;35 years) living with T1D, followed between 2010 and 2022 were used for this replication study. Data sources were the University Hospital Leuven (KUL) in Belgium\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, the German/Austrian/Luxembourgian/Swiss diabetes prospective (DPV) follow-up registry\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e and the Jordi Gol I Gurina Foundation (SIDIAP)\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e in Spain. The same biomarkers traditionally associated with CVD risk were included: age, sex, smoking, BMI, fasting glucose, lipids (HDL, LDL, triglycerides), blood pressure, serum creatinine, ALT, CRP and waist/hip ratio. The first visit with the most information available for all biomarkers was used. Missing data was imputed using information from similar patients with k-nearest neighbor algorithm. KUL included a population of 1,663 individuals (51% female, 49% male), DPV included 9,490 individuals (53% female, 47% male) and SIDIAP had a population of 13,170 (58% female and 42% male) used in this analysis.\u003c/p\u003e \u003cp\u003eAn overview of the analysis pipeline has been previously published.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e Briefly, to replicate the identified subgroups\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e, we calculated discordant scores based on the expected population-level associations between BMI and each biomarker, adjusted for age and smoking, generated from the UK Biobank (UKB). We then computed allocation probabilities into the concordant and discordant profiles described by Coral et al.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e, and visualized these allocations in relation to the concordant and discordant profiles using the Uniform Manifold Approximation and Projection (UMAP) embedding. Three risk profiles were mostly represented in the population with T1D: concordant, hyperglycemic and inflammatory (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The majority of individuals (55\u0026ndash;76%) belong to the discordant hyperglycemic profile (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ed., e. and f.), while this cluster is the smallest (~\u0026thinsp;2.5%) in the general population analysis done by Coral et al.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e Furthermore, in the population with T1D, mean HbA1c% was lower in the concordant cluster (mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation 6.9\u0026thinsp;\u0026plusmn;\u0026thinsp;0.9/7.1\u0026thinsp;\u0026plusmn;\u0026thinsp;1.0 in KUL and 7.1\u0026thinsp;\u0026plusmn;\u0026thinsp;1.6/7.3\u0026thinsp;\u0026plusmn;\u0026thinsp;1.5 in DPV for females/males, respectively) than in the hyperglycemic (7.8\u0026thinsp;\u0026plusmn;\u0026thinsp;1.0/7.8\u0026thinsp;\u0026plusmn;\u0026thinsp;1.3 in UZ Leuven and 7.9\u0026thinsp;\u0026plusmn;\u0026thinsp;1.5/8.0\u0026thinsp;\u0026plusmn;\u0026thinsp;1.5 in DPV for females/males respectively) and inflammatory (7.4\u0026thinsp;\u0026plusmn;\u0026thinsp;1.1/7.7\u0026thinsp;\u0026plusmn;\u0026thinsp;1.0 in KUL for females/males respectively) profiles.\u003c/p\u003e \u003cp\u003eSame as Coral et al.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e, we compared two types of sex-specific survival prediction models: one based solely on the previously mentioned biomarkers with variables and interactions used in the current CVD risk stratification tool endorsed by the European Society of Cardiology (SCORE2) \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e and another that also included profile allocation probabilities, to assess whether adding profile information would improve the prediction of major adverse cardiac events (MACE). Through comparison of nested models, we showed that adding profile allocation probabilities improved the predictive capability of these models, reflected by significant likelihood ratio tests and difference in C-statistics (Supplemental Table\u0026nbsp;1). This improvement was also reported by Coral et al\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e, particularly for men in UKB. Other risk engines specifically for diabetes, like the United Kingdom Prospective Diabetes Study (UKPDS)\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e and even T1D-specific engines such as STENO-T1D\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e have been shown to underestimate the risk of CVD in T1D\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e or do not include BMI.\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e Nevertheless, the newer LIFE-T1D\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e risk score that contains similar variables to SCORE2 with more focus on kidney health (albuminuria), concomitant complications (retinopathy) and age of T1D diagnosis has considered BMI as an important predictor.\u003c/p\u003e \u003cp\u003eMany medical decisions involve balancing trade-offs, such as treating patients to prevent disease-related harm while minimizing the risk of treatment side effects, or diagnosing patients with a disease while avoiding unnecessary testing for healthy individuals. Net benefit is a decision-analytic measure that is increasingly used to quantify both benefits and harms on a common scale. To assess the additional net benefit of discordant profiles, decision curves were used to evaluate whether implementing interventions to prevent CVD would be beneficial (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Coral et al\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e showed that models with and without discordant profile data, generally performed better than no intervention or universal intervention across various MACE probability thresholds up to 15%. It also showed that adding profile information at a 10% 10-year MACE risk threshold resulted in an average net benefit of 4 additional correct interventions and 37 additional unnecessary interventions correctly avoided per 10,000 individuals tested.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e In a population with T1D, at the same threshold of 10%, the net benefit for MACE was of 16 additional correct interventions while correctly avoiding 1,274 unnecessary interventions for every 10,000 individuals tested (average of DPV and SIDIAP-Supplemental Table\u0026nbsp;2). Especially for T1D, this can become a way forward to aid in decision-making for CVD prevention, since the risk profile information was proven to be more impactful in T1D than in the general population.\u003c/p\u003e \u003cp\u003eFinally, to explore how the relationship of BMI\u0026ndash;biomarker discordance in this population with T1D differs from the general population, sex-specific linear models for the residuals were fitted while adjusting for age, smoking status and disease duration before clustering. In terms of starting points (intercept), the main differences were present in fasting glucose for both males and females. Additionally in females, differences in systolic blood pressure and ALT were seen, while in males, also LDL was found altered (Supplemental Table\u0026nbsp;3). These biomarkers could prove interesting targets for reducing CVD risk in people living with T1D with potentially greater effects than in the general population.\u003c/p\u003e \u003cp\u003eWhile the risk profiles identified by Coral et al.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e are well-suited for the general population, as expected, there is an overexpression of the hyperglycemic profile in individuals with T1D. Furthermore, better glycemic control (HbA1c) was associated with a lower CVD risk profile, such as the baseline concordant cluster, which more closely aligns with the general population. This highlights the critical importance of maintaining glycemic control in T1D management\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e as a key factor in effective CVD risk control. Also, it is in contradiction with previous findings stating that better glycemic control does not lower CVD.\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e On the other hand, our results point to chronic hyperglycemia masking other relevant risk profiles or biomarkers in people living with T1D, further hinting at different pathways to CVD depending on glycemic control\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e and mediated by traditional risk factors over time -disease duration-.\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e Future research in T1D would benefit from considering models based not only on typical CVD risk factors but also approaches that account for the impact of glucose control before allocating/clustering into risk profiles. It could also explore the use of longitudinal data to further validate BMI- subclassifications and their evolution in diverse cohorts. Understanding the heterogeneity of diabetes will be a key step towards precision medicine\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e, particularly for CVD prevention in T1D.\u003c/p\u003e "},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eStudy Cohorts\u003c/h2\u003e\n \u003cp\u003eUniversity Hospital Leuven (KUL)\u003c/p\u003e\n \u003cp\u003ePatient data were obtained from the type 1 diabetes (T1D) registry of University Hospital Leuven,\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e a tertiary care center located in the Flemish Brabant region -northern Belgium. All individuals with a clinical diagnosis of T1D who received care at this hospital were eligible for inclusion in the registry. A confirmed diagnosis by a healthcare provider was required.\u003c/p\u003e\n \u003cp\u003eGerman/Austrian/Luxembourgian/Swiss diabetes prospective (DPV) follow-up registry\u003c/p\u003e\n \u003cp\u003eThis is a multicenter database initiated in Germany in 1995 to standardize the documentation of demographic and diabetes-related clinical data from specialized care centers and has now expanded to include centers in Austria, Luxembourg, and Switzerland.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e Clinical and demographic data are entered locally using the standardized DPV electronic health record system. Participating centers submit anonymized data biannually to Ulm University, Germany, where they undergo plausibility checks. Any discrepancies identified are returned to the clinics for clarification or correction prior to data consolidation. The anonymized dataset is utilized for clinical research and quality monitoring purposes (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.d-p-v.eu\u003c/span\u003e\u003c/span\u003e). The DPV initiative has received ethical approval from the ethics committee of Ulm University, as well as the respective review boards of all participating institutions.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e For this study, data from adult (\u0026gt;\u0026thinsp;35 years) individuals diagnosed with type 1 diabetes were analyzed.\u003c/p\u003e\n \u003cp\u003eJordi Gol I Gurina Foundation (SIDIAP)\u003c/p\u003e\n \u003cp\u003eThe Information System for Research in Primary Care (SIDIAP)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e is a comprehensive database of electronic health records from primary care, designed to support healthcare research. It compiles data from 328 primary care centers operated by the Catalan Health Institute across Catalonia, Spain. Since 2006, SIDIAP has collected pseudo-anonymized health records for over 8\u0026nbsp;million individuals, covering approximately 75% of the Catalan population.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003eStatistical analysis\u003c/h2\u003e\n \u003cp\u003eData Preparation\u003c/p\u003e\n \u003cp\u003eSame as Coral et al,\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e we included 13 biomarkers across all cohorts: fasting glucose (FG) in mmol/l; lipid fractions (HDL, LDL, triglycerides) in mmol/l; systolic and diastolic blood pressure in mmHg; serum creatinine in \u0026micro;mol/l; alanine aminotransferase (ALT) in U/l; C-reactive protein (CRP) in mg/l; waist-to-hip ratio in cm; age in years; smoking status (1\u0026thinsp;=\u0026thinsp;current smoker, 0\u0026thinsp;=\u0026thinsp;non-smoker); and sex (male/female). Units were harmonized across datasets. No BMI thresholds were applied. Outliers exceeding 5 standard deviations from the mean were excluded prior to analysis. All analyses were stratified by sex to account for known gender-related differences in BMI, biomarkers, and cardiometabolic risk.\u003c/p\u003e\n \u003cp\u003eUMAP Projection and Profile Identification\u003c/p\u003e\n \u003cp\u003eThis cluster analysis follows a multi-step analytical pipeline\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e that combines linear modeling, residual extraction, dimensionality reduction, and unsupervised clustering to examine biomarker data. The main goals are to evaluate the relationship between a set of biomarkers and BMI (adjusting for additional covariates), and to identify latent subgroups based on the residual patterns from these models. The analysis is conducted separately for males and females.\u003c/p\u003e\n \u003cp\u003eBiomarker residuals were projected into two dimensions using the umap function from the R package uwot (v0.1.16), with nearest neighbor count set as a function of cohort size. The parameter binary_edge_weights was set to TRUE, implementing PacMAP, a UMAP variant that better preserves both global and local data structures. Additionally, we used dens_scale\u0026thinsp;=\u0026thinsp;1 to enable densMAP, which enhances density preservation. Clusters were identified by constructing a proximity network and applying the Leiden community detection algorithm after seeding with the leading eigenvector method (igraph, v2.0.2). The algorithm, run over 500 iterations, optimized modularity through node movement, partition refinement, and network aggregation. Eigen centrality scores were calculated per individual and used to determine cluster centres and covariance matrices within a Gaussian mixture model. To account for instability of central clusters, a concordant distribution (zero mean, identity covariance) was added. Individuals with high overlap between discordant and concordant profiles were assigned to the latter, improving cluster specificity. The final model provided both hard cluster assignments and profile probability scores for each individual.\u003c/p\u003e\n \u003cp\u003eModel Comparisons\u003c/p\u003e\n \u003cp\u003eWe used likelihood ratio tests and changes in C-statistics to compare nested models. Variance explained by discordant profiles was quantified using likelihood-based methods, unaffected by threshold choices. We also performed decision curve analyses to assess clinical utility, quantifying net benefit (true positives) and net interventions avoided (true negatives) across varying probability thresholds, both per cohort and pooled across studies\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003ePredictive Value of Profiles\u003c/p\u003e\n \u003cp\u003eTo assess added predictive value, we compared nested regression models with and without discordant profile information. Using profile allocation probabilities in regression models is problematic due to their sum-to-one constraint. Coral et al\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e addressed this by using the log-contrast approach from compositional data analysis, using the concordant profile as reference. For each individual, discordant profile probabilities were divided by the concordant probability, and their natural logarithms were used as predictors.\u003c/p\u003e\n \u003cp\u003eNested Cox Regression Models\u003c/p\u003e\n \u003cp\u003eWe evaluated the association between profile allocations and risk of MACE using nested Cox proportional hazards models over 10-year follow-up periods. The predictors included variables from the SCORE2 algorithm (validated in diabetic populations), (see Supplementary Table\u0026nbsp;1). Individuals with pre-existing CVD were excluded.\u003c/p\u003e\n \u003cp\u003ePhenotypic Discordance with BMI in T1D\u003c/p\u003e\n \u003cp\u003eWe examined the association between BMI and each biomarker, adjusting for age and smoking status and disease duration, using linear regression models (see Supplementary Table\u0026nbsp;2).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eKUL.\u0026nbsp;\u003c/strong\u003eThe data that support the findings of this study are available on request from the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDPV.\u0026nbsp;\u003c/strong\u003eThe datasets generated and analysed during the current study are not publicly available. Due to data protection reasons, data on individual level cannot be provided. However, we can provide remote access to aggregated data if requested\u003cstrong\u003e.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSIDIAP.\u003c/strong\u003e Any researcher is able to request SIDIAP data to conduct a study. A five-step procedure takes place before data access is granted: (i) the researcher(s) must send an application (standardized form available at www.sidiap.org and study protocol) to the SIDIAP team; (ii) the application is approved by SIDIAP\u0026rsquo;s Scientific Committee which evaluates the scientific quality and feasibility of the proposal; (iii) the study protocol is approved by the Clinical Research Ethics Committee of IDIAPJGol; (iv) the principal investigator of the study must sign a Good Practice form and, in some cases, an agreement between parties is needed; and (v) a meeting between the research team and the SIDIAP team is arranged to discuss the procedures and set the data extraction. Further information is available online (https://www.sidiap.org/index.php/menu-solicitudesen/application-proccedure) or by contacting Anna Moleras ([email protected]). Data access is limited to researchers from public organizations and collaboration with private institutions is possible when a study is required by a regulatory agency or for non-commercial studies within a European project financed by the European Commission.\u003c/p\u003e\n\u003cp\u003eIn accordance with current European and national law, the data used in this study are only available for the researchers participating in this study. Thus, we are not allowed to distribute or make publicly available the data to other parties.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll analyses were performed in programming language R v.4.4 (https://www.r-project.org/). The scripts used can be found at https://github.com/danielcoral/SOPHIA_Cross_Sectional.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eACKNOWLEDGEMENTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe also thank Carlos Marin P. for providing medical writing support and editorial assistance.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eFUNDING\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work is part of the Stratification of Obesity Phenotypes to Optimize Future Obesity Therapy (SOPHIA) project (www.imisophia.eu). SOPHIA has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No. 875534, supported by the European Union\u0026rsquo;s Horizon 2020 research and innovation program and EFPIA, with additional support from T1D Exchange, Breakthrough T1D (former JDRF), and Obesity Action Coalition.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCOMPETING INTERESTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCM serves or has served on the advisory panel for Novo Nordisk, Sanofi, Eli Lilly and Company, Novartis, Boehringer Ingelheim, Bayer, Roche, Medtronic, Imcyse, Insulet, Biomea Fusion and Vertex. Financial compensation for these activities has been received by KU Leuven; KU Leuven has received research support for CM from Medtronic, Imcyse, Novo Nordisk, Sanofi and ActoBio Therapeutics; CM serves or has served on the Speaker\u0026rsquo;s bureau for Novo Nordisk, Sanofi, Eli Lilly and Company, Medtronic and Boehringer Ingelheim. Financial compensation for these activities has been received by KU Leuven. CM is president of EASD. All external support of EASD is to be found on www.easd.org.\u003c/p\u003e\n"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAhuja A, Agrawal S, Acharya S, Reddy V, Batra N. Strategies for Cardiovascular Disease Prevention in Type 1 Diabetes: A Comprehensive Review. \u003cem\u003eCureus\u003c/em\u003e 2024; \u003cstrong\u003e16\u003c/strong\u003e(8): e66420.\u003c/li\u003e\n\u003cli\u003eVerg\u0026egrave;s B. Cardiovascular disease in type 1 diabetes, an underestimated danger: Epidemiological and pathophysiological data. \u003cem\u003eAtherosclerosis\u003c/em\u003e 2024; \u003cstrong\u003e394\u003c/strong\u003e: 117158.\u003c/li\u003e\n\u003cli\u003eMiller RG, Orchard TJ, Costacou T. 30-Year Cardiovascular Disease in Type 1 Diabetes: Risk and Risk Factors Differ by Long-term Patterns of Glycemic Control. \u003cem\u003eDiabetes Care\u003c/em\u003e 2022; \u003cstrong\u003e45\u003c/strong\u003e(1): 142-50.\u003c/li\u003e\n\u003cli\u003eAl Ozairi E, Steenackers N, Pazmino S, et al. Prevalence of obesity in people with and without type 1 diabetes across Belgium, Kuwait, and Mexico: an IMI2 SOPHIA study. \u003cem\u003eeClinicalMedicine\u003c/em\u003e 2024; \u003cstrong\u003e77\u003c/strong\u003e: 102869.\u003c/li\u003e\n\u003cli\u003eConway B, Miller RG, Costacou T, Fried L, Kelsey S, Evans RW, Orchard TJ. Temporal patterns in overweight and obesity in Type 1 diabetes. \u003cem\u003eDiabet Med\u003c/em\u003e 2010; \u003cstrong\u003e27\u003c/strong\u003e(4): 398-404.\u003c/li\u003e\n\u003cli\u003eCoral DE, Smit F, Farzaneh A, et al. Subclassification of obesity for precision prediction of cardiometabolic diseases. \u003cem\u003eNature Medicine\u003c/em\u003e 2024.\u003c/li\u003e\n\u003cli\u003eGrimsmann JM, Tittel SR, Bramlage P, et al. Disease heterogeneity of adult diabetes based on routine clinical variables at diagnosis: Results from the German/Austrian Diabetes Follow-up Registry. \u003cem\u003eDiabetes Obes Metab\u003c/em\u003e 2022; \u003cstrong\u003e24\u003c/strong\u003e(11): 2253-62.\u003c/li\u003e\n\u003cli\u003eRecalde M, Rodr\u0026iacute;guez C, Burn E, et al. Data Resource Profile: The Information System for Research in Primary Care (SIDIAP). \u003cem\u003eInt J Epidemiol\u003c/em\u003e 2022; \u003cstrong\u003e51\u003c/strong\u003e(6): e324-e36.\u003c/li\u003e\n\u003cli\u003eSCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe. \u003cem\u003eEur Heart J\u003c/em\u003e 2021; \u003cstrong\u003e42\u003c/strong\u003e(25): 2439-54.\u003c/li\u003e\n\u003cli\u003eSCORE2-Diabetes: 10-year cardiovascular risk estimation in type 2 diabetes in Europe. \u003cem\u003eEur Heart J\u003c/em\u003e 2023; \u003cstrong\u003e44\u003c/strong\u003e(28): 2544-56.\u003c/li\u003e\n\u003cli\u003eZgibor JC, Ruppert K, Orchard TJ, Soedamah-Muthu SS, Fuller J, Chaturvedi N, Roberts MS. Development of a coronary heart disease risk prediction model for type 1 diabetes: the Pittsburgh CHD in Type 1 Diabetes Risk Model. \u003cem\u003eDiabetes Res Clin Pract\u003c/em\u003e 2010; \u003cstrong\u003e88\u003c/strong\u003e(3): 314-21.\u003c/li\u003e\n\u003cli\u003eVistisen D, Andersen GS, Hansen CS, Hulman A, Henriksen JE, Bech-Nielsen H, J\u0026oslash;rgensen ME. Prediction of First Cardiovascular Disease Event in Type 1 Diabetes Mellitus: The Steno Type 1 Risk Engine. \u003cem\u003eCirculation\u003c/em\u003e 2016; \u003cstrong\u003e133\u003c/strong\u003e(11): 1058-66.\u003c/li\u003e\n\u003cli\u003eHelmink MAG, Hageman SHJ, Eliasson B, et al. Lifetime and 10-year cardiovascular risk prediction in individuals with type 1 diabetes: The LIFE-T1D model. \u003cem\u003eDiabetes Obes Metab\u003c/em\u003e 2024; \u003cstrong\u003e26\u003c/strong\u003e(6): 2229-38.\u003c/li\u003e\n\u003cli\u003eRisk Factors for Cardiovascular Disease in Type 1 Diabetes. \u003cem\u003eDiabetes\u003c/em\u003e 2016; \u003cstrong\u003e65\u003c/strong\u003e(5): 1370-9.\u003c/li\u003e\n\u003cli\u003eOrchard TJ, Olson JC, Erbey JR, et al. Insulin resistance-related factors, but not glycemia, predict coronary artery disease in type 1 diabetes: 10-year follow-up data from the Pittsburgh Epidemiology of Diabetes Complications Study. \u003cem\u003eDiabetes Care\u003c/em\u003e 2003; \u003cstrong\u003e26\u003c/strong\u003e(5): 1374-9.\u003c/li\u003e\n\u003cli\u003eMiller RG, Costacou T, Orchard TJ. Risk Factor Modeling for Cardiovascular Disease in Type 1 Diabetes in the Pittsburgh Epidemiology of Diabetes Complications (EDC) Study: A Comparison With the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study (DCCT/EDIC). \u003cem\u003eDiabetes\u003c/em\u003e 2019; \u003cstrong\u003e68\u003c/strong\u003e(2): 409-19.\u003c/li\u003e\n\u003cli\u003eBebu I, Braffett BH, Pop-Busui R, Orchard TJ, Nathan DM, Lachin JM. The relationship of blood glucose with cardiovascular disease is mediated over time by traditional risk factors in type 1 diabetes: the DCCT/EDIC study. \u003cem\u003eDiabetologia\u003c/em\u003e 2017; \u003cstrong\u003e60\u003c/strong\u003e(10): 2084-91.\u003c/li\u003e\n\u003cli\u003eCefalu WT, Franks PW, Rosenblum ND, et al. A global initiative to deliver precision health in diabetes. \u003cem\u003eNat Med\u003c/em\u003e 2024; \u003cstrong\u003e30\u003c/strong\u003e(7): 1819-22.\u003c/li\u003e\n\u003cli\u003eVickers AJ, Cronin AM, Elkin EB, Gonen M. Extensions to decision curve analysis, a novel method for evaluating diagnostic tests, prediction models and molecular markers. \u003cem\u003eBMC Medical Informatics and Decision Making\u003c/em\u003e 2008; \u003cstrong\u003e8\u003c/strong\u003e(1): 53.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6429567/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6429567/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eCardiovascular disease (CVD) is a major long-term complication and the leading cause of morbidity and mortality among individuals with type 1 diabetes (T1D), with a substantially higher prevalence compared to the general population and driven by multiple interrelated risk factors\u0026mdash;underscoring the urgent need for accurate risk assessment. To support more tailored approaches to CVD prevention, a recent article in Nature Medicine showed five discordant phenotypic risk profiles for CVD. These were identified in the general population in Europe with diverse relationship between body mass index and cardiometabolic biomarkers. Here, we explore their applicability in 24,323 people with T1D. Improved glycemic control was linked to a decrease in CVD risk as people with T1D with lower glycated hemoglobin belonged to the baseline concordant cluster. This supports the contention that glycemic control in people with T1D is an integral part of lowering CVD risk.\u003c/p\u003e","manuscriptTitle":"Precision cardiovascular risk prediction in type 1 diabetes: An IMI2 SOPHIA analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-30 06:37:17","doi":"10.21203/rs.3.rs-6429567/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"b18940ec-a092-40c4-a0bd-452f0912368e","owner":[],"postedDate":"May 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":47868169,"name":"Health sciences/Endocrinology/Endocrine system and metabolic diseases/Diabetes/Type 1 diabetes"},{"id":47868170,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2025-05-30T06:37:17+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-30 06:37:17","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6429567","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6429567","identity":"rs-6429567","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}