Use of the Steno T1 Risk Engine Identifies Preclinical Atherosclerosis Better Than Use of ESC/EASD-2019 in Adult Subjects With Type 1 Diabetes at High Risk

Cardiovascular disease (CVD) is the leading cause of death in the general population (1). Similarly, it is also the leading cause of death among patients with type 1 diabetes (T1D) (2). Large studies have shown a reduction in cardiovascular morbidity and mortality either under intensive insulin therapy (3) or with improved adherence to lipid-lowering drugs (4). Intensive treatment of these and other cardiovascular risk factors (CVRF), optimization of therapeutic education, and advances in the management of chronic disease have led to a significant trend toward a lower rate of cardiovascular events in this population (5,6). Despite the aforementioned, the long-term risk of CVD among T1D patients is four- to eightfold higher than in the general European population (7,8).

Given the importance of CVD in persons with T1D, tools to predict cardiovascular events are essential to tailor cardiovascular risk management according to each individual’s risk. Classically, cardiovascular risk scales, which estimate the risk of events according to clinical and laboratory variables, have been used for this purpose (e.g., Systematic Coronary Risk Evaluation [SCORE] or the Framingham-REGICOR equation) (9). These equations are intended to be applied to the general population, although they have been extrapolated to the T1D group. These approaches usually show little predictive value in the case of T1D (10), given the different pathophysiology and the fact that CVRF specific to this entity, such as albuminuria or duration of diabetes, are not taken into account. Subsequently, specific T1D scales with a higher performance have emerged, such as the Steno T1 Risk Engine (Steno-Risk) (11), use of which has been shown to have an independent association with identification of several markers of subclinical atherosclerosis (arterial stiffness [12] or carotid plaques visualized by ultrasonography [13]) and which may also be useful for prediction of CVD events in other T1D populations (14). Similarly, the recently published 2019 ESC Guidelines on diabetes, pre-diabetes, and cardiovascular diseases developed in collaboration with the EASD (ESC/EASD-2019) specifically defined differentiated subgroups of patients with T1D at high/very high cardiovascular risk (15). However, agreement between these two strategies has been scarcely studied (16), and in no previous study has their utility been assessed in identifying individuals more prone to presenting CVD.

With this background, in this study we aimed to assess the concordance between the cardiovascular risk estimation scales ESC/EASD-2019 and Steno-Risk and analyze the relationship of their use with identification of preclinical atherosclerosis in subjects with T1D without CVD but at high risk.

This was a cross-sectional investigation conducted in participants with T1D without CVD (coronary artery disease, ischemic stroke, peripheral artery disease, or heart failure) enrolled from a specialized Diabetes Unit of the Hospital Clinic in Barcelona, Spain. Hospital Clinic is a tertiary university hospital, with a reference population of 540,000 individuals, and actively follows 2,000–2,500 patients with T1D. All subjects met the subsequent inclusion criteria: age ≥40 years; the presence of diabetic nephropathy, regardless of age or diabetes duration; and/or duration of T1D ≥10 years with one or more CVRF.

We considered the following as additional CVRF: active smoking habit (excluding former smokers), diabetic retinopathy, hypertension (defined as systolic blood pressure ≥140 mmHg and diastolic blood pressure ≥90 mmHg or treatment with antihypertensive drugs), triglycerides >150 mg/dL, low HDL cholesterol (<40 mg/dL in men, <45 mg/dL in women), family history of premature CVD in first-degree relatives (defined as CVD occurring at <55 years of age in men and <65 years in women), severe hypoglycemia (defined as an episode of confirmed hypoglycemia requiring external assistance for recovery) or hypoglycemia unawareness (defined as a score >3 on the Clarke test, with the validated Spanish version [17]), and prior preeclampsia/eclampsia in women.

The study protocol was conducted according to the Declaration of Helsinki. All patients provided informed consent, and the study was approved by the local ethics committee (Barcelona, Spain).

Both demographic and clinical data including duration of T1D, family history of premature CVD in first-degree relatives, history of microvascular diabetes complications, and medical treatment (multiple-dose insulin, continuous subcutaneous insulin infusion, lipid-lowering agents, and antihypertensive and antiplatelet drugs) were obtained from medical records. The term “high-intensity statin therapy” was defined as statin doses that reduced LDL cholesterol by ≥50%.

Diabetic nephropathy was assessed according to the albumin-to-creatinine ratio, with <30 mg/g considered as normal and ≥30 mg/g as diabetic nephropathy (confirmed on at least two of three consecutive determinations). The use of ACE inhibitors (ACEi) or angiotensin II receptor blockers (ARB), without a history of hypertension or CVD, was also considered as diabetic nephropathy. Fundus oculi was used for diagnosis of diabetic retinopathy, which was always confirmed by an ophthalmologist.

Anthropometric measurements (weight, height, and waist circumference) were also obtained. Patients were weighed wearing light clothing and barefoot, with use of a calibrated electronic scale. BMI was calculated as weight in kilograms divided by the square of height in meters. The midpoint between the lowest rib and the iliac crest was used to measure waist circumference.

Laboratory parameters were measured in fasting blood and first-morning urine spot samples. Lipid profile (including total cholesterol, triglycerides, and HDL cholesterol), glucose, creatinine, and albumin-to-creatinine ratio were assessed with standardized assays. LDL cholesterol was determined with the Friedewald formula. Non-HDL cholesterol was calculated as the deduction of HDL cholesterol from total cholesterol. Estimated glomerular filtration rate (eGFR) was assessed with the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. Glycated hemoglobin (HbA_1c) values (Tosoh G8 Automated HPLC Analyzer; Tosoh Bioscience, South San Francisco, CA) (Diabetes Control and Complications Trial [DCCT] aligned, normal range 4–6% [20–42 mmol/mol]) were also recorded.

Cardiovascular risk was assessed with two different strategies, specific to the T1D population: classification according to Steno-Risk (11) and the cardiovascular risk groups proposed by ESC/EASD-2019 (15). Briefly, Steno-Risk estimates the 10-year risk of fatal or nonfatal CVD (ischemic heart disease, ischemic stroke, heart failure, and peripheral artery disease) based on 10 variables (age, sex, diabetes duration, HbA_1c, systolic blood pressure, LDL cholesterol, albuminuria, eGFR, smoking habit, and regular exercise [≥3.5 h/week]). Individuals were classified as having low (<10%), moderate (10–19.9%), or high (≥20%) risk, accordingly (11). ESC/EASD-2019 defined the following cardiovascular risk groups: moderate risk, those with T1D duration <10 years and without any additional CVRF or chronic diabetes complications; high risk, those with T1D duration ≥10 years and any additional CVRF (but without target organ damage [TOD]); and very high risk, those with TOD, three or more major CVRF (hypertension, dyslipidemia, smoking habit, obesity), or early T1D onset (<10 years old) and T1D duration >20 years.

An exploratory analysis was also performed with use of the non-diabetes-specific, broadly used, pooled cohort equation risk estimation of the American College of Cardiology/American Heart Association (2013 ACC/AHA Guideline on the Assessment of Cardiovascular Risk [ACC/AHA-2013]) (18), as recommended by the American Diabetes Association (19). The equation estimates the 10-year risk of atherosclerotic CVD (coronary death or nonfatal myocardial infarction or fatal or nonfatal stroke) from several variables (age, sex, race, systolic and diastolic blood pressure, total cholesterol, HDL cholesterol, history of diabetes, hypertension treatment, and smoking status). Subjects were classified according to the categories low-borderline (<7.4%), intermediate (7.5–19.9%), and high (≥20%) risk, as appropriate.

Carotid plaques were evaluated with high-resolution B-mode ultrasound (ACUSON X700 [Siemens Healthineers] or Aplio a450 [Canon]) with the same electric linear array 5–10 MHz transducer. Predefined and standardized imaging protocols to measure carotid intima-media thickness (IMT) were used, and the presence of plaque was determined as previously described (20,21). Carotid images were visualized with B-mode and color Doppler in longitudinal and transverse planes to appraise circumferential asymmetry. Carotid plaques were defined according to focal echo structures trespassing into the arterial lumen by at least 50% of the surrounding IMT value or when IMT was at least 1.5 mm as measured from the media-adventitia interface to the intima-lumen surface (22). Two experienced endocrinologists performed all the procedures, and IMT and plaque ascertainment and measurements were made by the same researcher (A.J.A.) using semiautomatic software. The mean and mean maximum IMT of all carotid segments (common carotid, bulb, and internal carotid) were documented, in addition to the maximum height of carotid plaque. If carotid plaques were present, the maximum IMT was equal to the highest carotid plaque height. Peak systolic and end-diastolic velocities were used to determine carotid stenosis, and when significant plaques (IMT >2.5 mm) were found, the planimetric area was measured in a transversal view.

The data were subjected to descriptive analysis. The Kolmogorov-Smirnov test was used to evaluate the normal distribution of continuous variables. Results for qualitative variables are stated as frequency and percentage, quantitative normal variables as mean ± SD, and nonnormal quantitative variables as median (first quartile–third quartile). Comparisons between binomial qualitative or quantitative variables were performed with parametric or nonparametric tests as appropriate. Cohen κ coefficient was used to assess concordance between scales (Steno-Risk vs. ESC/EASD-2019), either in the whole cohort or after division by sex. Furthermore, the characteristics of the individuals discordant with both strategies were also assessed.

The prevalence of preclinical atherosclerosis (no plaque, one plaque, or two or two or more plaques) according to risk strata with both strategies was assessed by means of the χ² test. The sensitivity and specificity of the scales for identifying individuals harboring carotid plaques (receiver operating characteristic curves) were further assessed, providing the Youden J statistic, accordingly. Finally, polynomial regression models were constructed for assessing the independent associations of the discordance in estimated risk (independent variable) with preclinical atherosclerosis (both presence of plaques or two or more 2 plaques [dependent variables]), with adjustment for variables not included in these strategies (HDL cholesterol, triglycerides, and cardioprotective treatment [i.e., statins, ACEi/ARB, and antiplatelet drugs].

As an exploratory analysis, in participants between 40 and 79 years old (n = 419), all of the previous comparisons between T1D-specific strategies were additionally compared with the ACC/AHA-2013 risk equation. Polynomial regression models were also constructed for the discordance in the estimated risk between Steno-Risk and ACC/AHA-2013 (with adjustment for triglyceride levels, statins, and antiplatelet drugs).

All of the statistical analyses were performed with IBM SPSS Statistics, version 26 (IBM Corporation, Armonk, NY), statistical software. The significance level was set as a P value <0.05.

A total of 501 subjects were included (53% men, mean ± SD), age 48.8 ± 10.25 years, median duration of diabetes 26.5 years (interquartile range 20.9–34.0). Overall, 29.5% were smokers, 25% had hypertension, 11% had nephropathy, and 38.3% had retinopathy (34.9% proliferative). There were 42.9% on statin treatment and 9.4% taking antiplatelet drugs. In the whole cohort, 41.3% had at least one carotid plaque (23.6% with two or more plaques), with no between-sex differences (P > 0.350 for both). As the number of plaques increased, there was a stepped increase in age, smokers, the presence of hypertension, diabetic nephropathy, and triglyceride values and a decrease in eGFR (P < 0.05 for all). The remaining characteristics are shown in Table 1.

With use of the Steno-Risk strategy, we classified one of four as having low cardiovascular risk and 43.9% and 31.1% as moderate or high risk, respectively. In contrast, in using the ESC/EASD-2019, we classified almost all individuals as at high or very high risk (51.3% and 48.1%), with virtually no individuals classified as moderate risk (Fig. 1). No differences in sex were found for any of the scales (P = 0.638 and P = 0.574) (Supplementary Fig. 1).

Clinical and laboratory characteristics according to estimated risk were further assessed. Regarding the Steno-Risk, a progressive increase in most of the CVRF was found as the estimated risk increased (namely, age, smoking status, hypertension, duration of diabetes, nephropathy, HbA_1c, waist circumference, retinopathy, and triglyceride levels), as well as in the use of the cardioprotective treatments (P < 0.05 for all comparisons) (Supplementary Table 1). Differences according to ESC/EASD-2019 categories (very high vs. moderate/high risk) were also assessed. Although some CVRF were increased for individuals at higher risk (i.e., BMI, hypertension, chronic complications, and HbA_1c; P < 0.05 for all), neither age, smoking habit, nor most of the cardioprotective drugs were statistically different (Supplementary Table 2).

The concordance between the two strategies was poor (κ = −0.079, P < 0.001) (Supplementary Table 3), even after we performed the analysis for the highest category of risk for both strategies (κ = 0.186 for all the cohort) (Supplementary Table 4), with no sex differences (κ = 0.173 and 0.195 for women and men; respectively) (Supplementary Table 4). In analyses of patients with discrepant results between the two strategies, we found that those with a higher estimated risk with Steno-Risk (12%) were older and had longer diabetes duration and worse glycemic control, higher prevalence of hypertension and higher triglyceride levels, worse renal function, and greater use of cardioprotective drugs. Conversely, those with discrepant lower risk in Steno-Risk (28%) were younger, with a better overall cardiovascular risk profile as well as lower use of cardioprotective drugs (Supplementary Table 5).

According to Steno-Risk, most of the patients at low risk were free of plaques (82%) and only 4% harbored two or more plaques. Conversely, two of three of the patients in the high-risk category showed plaques (46% with two or more plaques; P < 0.001 for linear trend) (Fig. 2). Regarding the ESC/EASD-2019 strategy, irrespective of the allocation of high or very high risk, most showed no plaques (63% and 55% for high and very high risk, respectively) (Fig. 2). Accordingly, no differences were found in preclinical atherosclerosis status, as the estimated risk increased with this strategy (P for trend = 0.127) (Fig. 2). Furthermore, no statistical differences were found in plaque presence in comparing some of the very-high-risk subgroups with the whole high-risk group according to ESC/EASD-2019 (T1D + TOD, n = 118; 45.4% of plaque presence; early-onset T1D + ≥20 years of diabetes duration, n = 89; 40.4% of plaques; P = 0.077 and P = 0.604).

Whereas the area under the curve (AUC) in receiver operating characteristic curves was statistically significant for Steno-Risk (AUC 0.691; Youden J statistic 0.236; P < 0.001), those harboring plaques were not identified with use of ESC/EASD-2019 (AUC 0.538; P = 0.149) (Supplementary Fig. 2). Accordingly, diagnostic performance regarding plaque identification was better with Steno-Risk, both for those at moderate-high risk (Steno-Risk ≥10%; sensitivity = 0.889 and specificity = 0.348; Youden J statistic = 0.236) and high risk (Steno-Risk ≥20%; sensitivity = 0.483 and specificity = 0.809; Youden J statistic = 0.292), vs. ESC/EASD-2019 for very high risk (sensitivity = 0.527 and specificity = 0.553; Youden J statistic = 0.080). Similar results were found for two or more plaques, with AUC 0.736 (P < 0.001) for Steno-Risk (sensitivity = 0.958 and specificity = 0.314 and sensitivity = 0.602 and specificity = 0.777 for moderate-high and high risk, respectively) and AUC 0.537 (P = 0.226) for ESC/EASD-2019 (sensitivity = 0.542 and specificity = 0.539 for very-high risk) (Supplementary Fig. 2). We also analyzed performance using the absolute risk values estimated with the Steno-Risk equation, showing a slight improvement compared with the classification into three groups (AUC 0.725 for the presence of plaque [Youden J statistic = 0.378] and AUC 0.772 for two or more plaques [Youden J statistic = 0.457]; P < 0.001 for both) (Supplementary Fig. 3).

Finally, concordance in cardiovascular risk assessment concerning preclinical atherosclerosis was also studied. Irrespective of the estimated risk with ESC/EASD-2019, individuals classified as high risk with Steno-Risk showed a significantly higher prevalence of plaques (P < 0.001) (Supplementary Table 5). Even after adjustment for several variables not included in any of the strategies (HDL cholesterol, triglycerides, and cardioprotective treatment), discordance and higher risk according to Steno-Risk showed a similar or even greater association with presence of preclinical atherosclerosis than concordance and high/very high risk according to both strategies (presence of plaques odds ratio [OR] 2.914 [95% CI 1.268–6.716] and 2.152 [0.991–4.673] and two or more plaques 4.938 [2.346–10.395] and 5.29 [2.746–10.189] for discordance with higher Steno-Risk and concordance with high/very high risk; respectively) (Fig. 3). Conversely, even in individuals classified as at very high risk with ESC/EASD-2019, if Steno-Risk was low/moderate, the presence of plaques was similar to that of those identified as non–high risk according to both strategies (P > 0.254) (Fig. 3).

Of the n = 419 individuals for whom the risk equation could be applied, almost two-thirds (62%) showed low or borderline risk, 27.7% intermediate risk, and only 10.3% high risk with ACC/AHA-2013, with a higher estimated risk in men than women (P < 0.001) (Supplementary Fig. 4). Regarding patient characteristics according to risk strata, a stepped increase in the proportion of men, age, and presence of active smoking habit, hypertension, and diabetic nephropathy was observed as the estimated risk increased (Supplementary Table 6). Furthermore, triglycerides, creatinine, and cardioprotective treatment increased and HDL cholesterol decreased as the ACC/AHA-2013 risk categories increased (Supplementary Table 6).

There was a gradual increase in the number of plaques as the estimated risk increased, with at least one plaque in 32.7%, 62.9%, and 79.1% of those at low-borderline, intermediate, and high risk, respectively (P for lineal trend <0.001) (Supplementary Fig. 5). However, most of the patients harboring carotid plaques were classified as non–high risk (82.3% of patients with plaques) and almost half (44.3%) fell into the low-borderline category. The discriminatory diagnostic capacity of the ACC/AHA-2013 equation was also compared with T1D-specific scores in these 419 individuals. The ability to detect preclinical atherosclerosis was similar to that shown by Steno-Risk (presence of plaque AUC 0.673 and 0.664, and two or more plaques AUC 0.688 and 0.693 for ACC/AHA-2013 and Steno-Risk; P < 0.001 for both) (Supplementary Fig. 6A–D) and higher than that by ESC/EASD-2019 (AUC 0.558 and 0.556 for the presence of plaque and two or more plaques) (Supplementary Fig. 6A and B). Furthermore, in evaluating only those in the highest risk category with the three scales, Steno-Risk better discriminated individuals with the most advanced atherosclerosis (two or more plaques AUC 0.669, P < 0.001, AUC 0.579, P = 0.012, and AUC 0.560, P = 0.058, for Steno-Risk, ACC/AHA-2013 and ESC/EASD-2019) (Supplementary Fig. 6E and F). The performance of the three strategies in identifying individuals with preclinical carotid atherosclerosis is summarized in Supplementary Table 7.

Finally, risk classification agreement between Steno-Risk and ACC/AHA-2013 was also analyzed. Individuals at higher risk by Steno-Risk showed a significantly higher prevalence of plaques, irrespective of ACC/AHA-2013 risk category allocation (P < 0.001) (Supplementary Table 8). And, after adjustment for possible confounders not included in both equations, participants with discordance in risk assessment with higher risk as estimated with Steno-Risk showed a greater atherosclerotic burden (OR 3.333 [95% CI 1.930–5.756], P < 0.001, for two or more plaques) (Supplementary Fig. 7).

In our sample of T1D adults at prespecified high risk undergoing primary prevention, there was poor concordance between cardiovascular risk scales, with an overall higher estimated risk with ESC/EASD-2019 compared with Steno-Risk. Furthermore, this discrepancy also translated into identification of individuals with more advanced atherosclerosis. Whereas ESC/EASD-2019 classification failed to detect individuals harboring more carotid plaques, the Steno-Risk was directly associated with the presence and number of plaques in a stepped manner, even after adjustment for other CVRF. Moreover, the performance of ESC/EASD-2019 guidelines in identifying individuals with higher plaque burden was worse than that of cardiovascular risk scores specifically designed for the general population. As far as we know, this is the first study assessing the differences between these strategies and the relationship of their use with identification of preclinical atherosclerosis as detected by ultrasound in an adult population with T1D at high cardiovascular risk.

Estimating cardiovascular risk in people with T1D remains a challenge nowadays. Predictive tools such as the classification proposed by the ESC/EASD-2019 emphasize the increased risk associated with age, the presence of TOD, and the duration of diabetes (15), mainly based on the excess of mortality found in the Swedish National Diabetes Register among individuals with earlier disease onset (23). However, some criticism has been raised regarding classifying all the patients with early-onset and long duration of T1D as at very high risk (24). Similarly, in our cohort, among patients of this kind there was a prevalence of preclinical atherosclerosis similar to that of individuals with a lower estimated risk. This lack of effectiveness in identifying patients at risk implies that, with the exception of subjects <35 years of age, duration of diabetes <10 years, and no other CVRF, the majority of patients would be considered at high/very high risk and, therefore, with an indication for lipid-lowering and even antiplatelet therapy. The classification of cardiovascular risk proposed in the recently published 2021 ESC Guidelines on cardiovascular disease prevention in clinical practice (ESC-2021) (25) has not undergone substantial changes in this respect. In this regard, other guidelines with use of age and diabetes duration as the main drivers for the estimation of cardiovascular risk (such as the National Institute for Health and Care Excellence [NICE] guidelines [26]), have shown poor performance for risk stratification. For instance, if these guidelines were applied to the Swedish National Diabetes Register’s cohort, 81–90% and 100% of patients between 20–40 years and ≥40 years of age, respectively, would be on statins (27). Similarly, if we look at the regional and national databases of adult individuals with T1D in our country, the average age is >40 years and mean duration of diabetes between 10 and 20 years (28–30). Therefore, the vast majority would also be classified as at high or very high risk according to ESC/EASD-2019, and intensified treatment of CVRF would be warranted as well. One of the possible explanations of this poor performance of the European guidelines is that other known predictors of mortality and CVD, such as glycemic control, are not taken into account (31). Conversely, the Steno-Risk equation includes several pathology-specific variables, which could allow greater individualization of risk and, consequently, tailored treatment. In our cohort, the concordance between these two strategies was poor and up to 60% of those at very high risk according to ESC/EASD-2019 were at low/moderate risk according to Steno-Risk (Supplementary Table 4). Pointing in the same direction, in a study conducted in an unselected Italian cohort of T1D patients (16), only 12% aged >35 years were included in the highest risk categories according to both scales and ∼3% were for analyses of the entire cohort. In our study, the concordance across the highest risk categories was greater (20%), which is to be expected considering that among our cohort patients were older (mean ± SD age 48.8 ± 10.25 vs. 36 ± 12 years) and had a longer diabetes duration (median 26.5 years [interquartile range 20.9–34.0] vs. 19 ± 11 years), with higher prevalence of other CVRF (Table 1).

A relevant finding of our study was the stepped relationship between preclinical atherosclerosis and estimated cardiovascular risk according to Steno-Risk (Fig. 2), which was also independent of that yielded by ESC/EASD-2019 (Fig. 3). In no previous study have investigators specifically evaluated the presence of preclinical atherosclerosis by validated methods or the incidence of CVD according to different risk scales among the T1D population. In a sample of n = 302 individuals with T1D, our group recently showed that Steno-Risk was independently associated with the presence and the number of carotid plaques, even after adjustment for other CVRF (13). Similarly, Boscari et al. (14) found that a higher risk according to this score better discriminated between subjects with a higher prevalence of carotid plaques and those who presented a cardiovascular event. Although the follow-up time was almost 5 years, it should be noted that the sample was small (n = 223) and the number of CVD events very low (n = 3); therefore, further studies are needed to ascertain its usefulness. Pending results of prospective studies with hard CVD outcomes, the findings of the current study provide the first evidence that Steno-Risk could better identify individuals with T1D at the highest risk of CVD (vs. ESC/EASD-2019).

As an exploratory analysis, we also assessed the performance of the ACC/AHA-2013 risk equation in our sample. Despite being a non-T1D-specific tool, it showed superior efficacy to detect atherosclerosis compared with the ESC/EASD-2019 classification and efficacy similar to that of Steno-Risk (Supplementary Fig. 6). However, some caveats should be noted. First, the ACC/AHA-2013 risk equation cannot be applied for individuals <40 years of age. Taking into account that T1D specially affects younger individuals (32), the use of this nonspecific score would leave out of the estimation a large number of patients who could benefit from early cardioprotective treatment. Second, in the context of a disease in which CVD is the leading cause of early mortality (2), it seems surprising that the overall estimated cardiovascular risk could be low in the vast majority of the individuals (Supplementary Fig. 4). In the same sense, the ACC/AHA-2013 risk equation also failed to identify the patients with preclinical atherosclerosis, since >82% of patients with at least one carotid plaque were allocated as non–high risk and, thus, not requiring cardioprotective treatment. This is in contrast to Steno-Risk, according to which only 11.1% of patients with preclinical carotid atherosclerosis were classified as low risk and almost half of those with plaques were correctly classified as high risk (Fig. 2 and Supplementary Fig. 5). Finally, even though ACC/AHA-2013 performed better than other T1D-specific strategies (i.e., ESC/EASD-2019), Steno-Risk seems to better identify individuals with more advanced atherosclerosis (two or more carotid plaques), even over the estimation of this general population risk scale (Supplementary Fig. 7). Altogether, this reinforces the statement of the recent American Diabetes Association/ European Association for the Study of Diabetes consensus, in which specific risk scores (i.e., Steno-Risk) are recommended for selecting individuals for whom cardioprotective treatment should be prescribed (33).

Classically, and after the irruption of the large Diabetes Control and Complications Trial (34) and the observational Epidemiology of Diabetes Interventions and Complications (EDIC) study (3), the management of T1D was and continues to be eminently glucocentric. In this regard, in the last years, a lot of technological advances have helped with improvement of glycemic control (35). However, despite good glycemic control, and even without the presence of other chronic complications, cardiovascular risk in T1D patients is never equal to that of the general population (36). Therefore, studies focused on other aspects of the pathophysiology of CVD in T1D are mandatory. Thus, both nonspecific (i.e., age, dyslipidemia, or obstetric complications) and T1D-specific (i.e., nephropathy and retinopathy) factors have been involved in cardiovascular risk in this population (37–40). Our findings are that 1) Steno-Risk better identifies individuals with a higher prevalence of CVRF vs. ESC/EASD-2019 (Supplementary Tables 1 and 2) and 2) this scale was strongly associated with identification of preclinical atherosclerosis, which could help with overcoming this glucocentric scenario. Taking into account that the 10 variables included in the score are easily available (11), Steno-Risk should be implemented in clinical practice.

Generic CVD events prediction tools are not easily extrapolated to the T1D population, and even in using those T1D-specific tools (i.e., Steno-Risk), >40% of the individuals are classified as at moderate risk, which is an indeterminate area in CVD prevention (Fig. 1). Against this background, the use of other potential markers that can increase predictive performance becomes paramount. Imaging biomarkers, such as the detection of carotid plaques by ultrasound, have demonstrated their usefulness as predictors of CVD (41) and could serve to reclassify risk in subjects in the general population according to the recently released ESC-2021 guidelines on CVD prevention (25). In addition, the visualization of plaques by the patients themselves can increase adherence to a healthy lifestyle and cardioprotective drugs and may even reduce the risk of cardiovascular events (42). Although the usefulness of this biomarker in the setting of the T1D population is less known, our study could help to fill the gap in this promising research area. Thus, carotid ultrasonography could be especially useful in the setting of moderate risk allocation with Steno-Risk, in which case there is uncertainty in identifying individuals with atherosclerosis (Fig. 2). Furthermore, there is also some preliminary evidence that other biomarkers aimed at detecting the quality of diet (43) or systemic inflammation (44) could be useful and open the door to different approaches and therapeutic targets. The implementation of biomarkers in daily practice could help to intensify cardiovascular disease prevention in patients with greater plaque burden, although it also allows de-escalation of unnecessary treatments for those with discordance between the estimated risk and the presence of preclinical atherosclerosis.

Our study has several strengths: 1) its novelty, as to the best of our knowledge, this is the first study with analysis of the relationships of different risk scales with identification of a proven proxy of cardiovascular events, such as the presence of carotid plaques (41); 2) the quality of the data, which were carefully collected, including for pharmacological treatments and other clinical and analytical variables, reduces the risk of bias in our results; 3) the large sample included, and the use of standardized procedures, reduces the variability and, therefore, leads to more accurate results. However, our study also has limitations. Firstly, the cross-sectional design does not allow assessment of the appearance of atherosclerotic disease or cardiovascular events. In this regard, although the detection of carotid plaques by ultrasound has demonstrated predictive capacity for future cardiovascular events, it is a surrogate variable. Further prospective studies are needed to ascertain the usefulness of CVD risk scores in identifying those more prone to hard clinical events. Secondly, given that the study was carried out in a single tertiary hospital and that only prespecified high-risk patients without CVD were included, extrapolation to other patients with a different setting or cardiovascular profile should be made with caution. However, although our sample of study was a bit older, with a longer diabetes duration, than those of other recent data sets representative of our geographical area (28–30), the other key characteristics of cardiovascular risk, such as mean BMI, prevalence of smoking habit, and hypertension, or microvascular complications, were very similar. Thirdly, due to our inclusion criteria, high/very high risk strata according to ESC/EASD-2019 classification strategy could be overrepresented. Thus, the better performance of Steno-Risk for those classified as moderate risk by this strategy may be unproven. However, in looking at other studies with recruitment of a representative sample of adults with T1D from our same area (28–30), mean age was >40 years and in the case of most patients mean diabetes duration was >10 years, which, according to ESC/EASD-2019 guidelines, would, in fact, classify most at high/very high risk. Furthermore, even in other studies, with inclusion of T1D individuals with a priori lesser risk compared with our sample (mean age 36 years and diabetes duration 19 years; <20% with diabetic retinopathy), only 4% were classified as at moderate risk according to the ESC/EASD-2019 strategy (16). Altogether, this suggests that moderate risk according to this classification strategy would hardly be seen among nonpediatric T1D patients, allowing our results to be extrapolated to other adult populations with T1D. Finally, although regular exercise (≥3.5 h/week) was a variable included in Steno-Risk, more specific information about the intensity and frequency of physical activity was lacking. Since investigators of some previous studies have found a close relationships between this variable and future CVD events among the T1D population (45), it would have been interesting to assess the relationships of physical activity, estimated cardiovascular risk, and preclinical atherosclerosis.

In summary, in our sample of patients with T1D at high cardiovascular risk, the ESC/EASD-2019 generic risk scale showed poor concordance with the more T1D-specific Steno-Risk scale. The latter presented a significant and independent association with the presence and number of carotid plaques, even after adjustment for other variables strongly associated with atherosclerosis. Furthermore, and independently of estimations according to the ESC/EASD-2019 strategy, Steno-Risk better identified individuals with more advanced atherosclerosis, even after comparisons with the ACC/AHA-2013 pooled cohort equation. The availability of more accurate tools for predicting CVD events, and the advances in the knowledge of different CVRF in T1D, allows moving in the direction of individualized medicine to identify individuals who can benefit most from a more aggressive and targeted management.

Acknowledgments. The authors are grateful to Donna Pringle (freelance native English translator) for her helping in writing and editing the manuscript.

Funding. A.J.A received a research grant from Associació Catalana de Diabetis, “Ajut per a la recerca en diabetis modalitat clínica 2018.”

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. All authors discussed the results and commented on the final version of the manuscript. T.S.-N., M.G., V.P., L.B., C.V., J.B., I.V., A.P., E.E., I.C., and A.J.A. acquired and processed all clinical data. A.J.A. performed the U.S. measurements. T.S.-N. and A.J.A. contributed to data analysis and interpretation and wrote, reviewed, and edited the manuscript. V.P. and A.J.A. contributed to the study concept and design. M.G., V.P., C.V., J.B., I.V., I.C., and A.J.A. supervised the study and participated in data analysis and interpretation. T.S.-N. and A.J.A. wrote the manuscript, designed the figures, and had final responsibility for the decision to submit for publication. A.J.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 62nd National Congress of “Sociedad Española de Endocrinología y Nutrición” (SEEN), Sevilla, Spain, 13–15 October 2021, and XXXII National Congress of “Sociedad Española de Diabetes” (SED), Granada, Spain, 16–18 June 2021.