The article by Lobo et al. in the last issue of the Argentine Journal of Cardiology (1) presents an uncomfortable truth: in patients with type 2 diabetes, cardiovascular risk scores are not always consistent and do not provide equivalent information. In this cohort of patients in primary prevention, the authors observed different results after using various calculators. Depending on the model, between 10% and 70% of patients were considered to be at high risk. What do clinicians do with that information?
The discordance between tools is more than a statistical problem. This is a practical issue: which patients should receive more intensive treatment? In which patients should I be more aggressive in managing their lipids or in choosing antidiabetic drugs with cardiovascular benefits? This study shows that, although there is good correlation between scores, the actual concordance (defined as the agreement in the final classification) is low. This leaves us, once again, relying on clinical judgment as our compass.
Diabetes is not a uniform condition, but rather one with different phenotypes and clinical courses. Most importantly, the cardiovascular risk associated with diabetes varies from person to person. For years, it was assumed that type 2 diabetes mellitus was a heart disease equivalent. Today we know that this is a dangerous simplification. However, we have moved to the opposite extreme, where scores have become increasingly complex, incorporating dozens of variables... and yet there is still a lack of consensus. (3)
In this context, the search for subclinical disease, such as carotid atheromatosis, which was evaluated in this study, makes sense again. Finding a plaque in an "intermediate" risk patient may justify more aggressive interventions. (4) Conversely, if a patient has no other risk factors and shows no signs of vascular damage, we can be more cautious.
The message is clear: scores help, but they do not decide for us. In the meantime, while we await a simple, locally calibrated, and pragmatic model, we go on doing our best: listening to patients, reviewing their medical records, interpreting their exams, and making decisions collaboratively. Sometimes, this can be more valuable than any algorithm. (5)
Ethical considerations
Not applicable.
Conflicts of interest
None declared.
(See authors' conflict of interests forms on the web).
AUTHORS’ REPLY
We greatly appreciate your thorough review and valuable feedback on our work. We fully agree that the heterogeneity of cardiovascular risk scores in patients with type 2 diabetes poses a challenge not only in terms of methodology, but also for decision-making in clinical practice. It is noteworthy to mention that risk scores depend on the population in which they were developed, which may affect their applicability and accuracy in different contexts. The variability of the same risk score across different populations is not a new issue. Brindle et al. addressed this problem by comparing the Framingham risk score among 71 727 patients in 27 studies. This reflects the difficulty of applying a risk score to a population other than the one in which it was developed. (Brindle P., Beswick A., Fahey T., Ebrahim S. Heart 2006;92:1752-1759).
As previously mentioned, the differences between the various tools require physicians to integrate evidence with clinical judgment, the patient's individual history, and, to a growing extent, the search for markers of subclinical damage. We believe this is a key point: cardiovascular risk scores are a useful tool for coordinating the use of resources, but they do not replace comprehensive assessment or shared decision-making. Rather, they complement these processes.
We also agree with the observation on how the paradigm has evolved, starting from the idea that all type 2 diabetes cases are a heart disease equivalent (NCEP ATP III, JAMA, 2001; 285: 2486-2497; SAC Consensus on Cardiovascular Prevention, Rev. Argent. Cardiol. 2020;88:9-3), to the current situation in which the complexity of the models does not always translate into clear utility in clinical practice. In this scenario, the identification of carotid plaques or other indicators of subclinical vascular disease may provide additional clinically relevant criteria for personalizing treatment intensity.
As the recent consensus statements of the Argentine Society of Cardiology have recommended, risk stratification should consider not only available scores but also the assessment of markers of subclinical vascular damage, particularly in patients with type 2 diabetes in primary prevention. (SAC Consensus Statement on Cardiovascular Prevention. Rev Argent Cardiol. 2020;88:9-3; Rev Argent Cardiol. 2024;92:F-19). Both documents underscore the need to move toward simpler models that are calibrated to our population and useful for daily clinical practice, in line with the considerations outlined in the letter.
In short, the discussion raised in your letter enriches the debate and reinforces the need to move towards simpler prediction models that are better calibrated to local populations and, most importantly, can be effectively implemented in daily practice. Meanwhile, and as is well emphasized, the combination of science, clinical experience, and dialog with the patient remains our most robust tool.
Yours sincerely,
The authors
