Diabetes mellitus is associated with more complex and progressive coronary atherosclerosis, leading to the development of cardiovascular disease (CVD) [1, 2]. However, cardiovascular risk in diabetics is heterogeneous [3] and clinical risk scores are limited in defining individual risk and thus intensity of preventive measures required. Non-invasive imaging modalities, particularly non-contrast enhanced computed tomography (CT) coronary artery calcium scoring (CACS), may improve the ability to risk stratify diabetics [4]. Past studies examined the additive prognostic value of CACS in diabetics in comparison to older risk score models [5, 6]. In recent years, the Pooled Cohort Equation (PCE) and the Multi-Ethnic Study of Atherosclerosis (MESA) risk scores, which both include diabetes, are more widely used in clinical practice for assessing future risk of cardiovascular events [7, 8]. However, limited data exist in diabetics regarding the additional predictive value of combining these risk models with CACS. Moreover, although it is widely recognized that CACS can improve clinical risk assessment in asymptomatic individuals, its routine use in diabetics is debated [9]. In the current study, we aimed to investigate the performance of the PCE and MESA risk scores with or without CACS in predicting long-term cardiovascular outcomes of diabetic patients without coronary artery disease (CAD).

In long-term follow-up of diabetic patients with additional risk factors but without known CAD, PCE and MESA risk scores predicted 10-year MI and cardiovascular death but not stroke. CACS was an independent predictor of MACE and each of its individual endpoints, improving discrimination of event risk prediction models. Absence of coronary artery calcification at study entry was evident in up to a quarter of the study population, and was associated with very-low event rates even in patients with high estimated clinical risk scores.

Diabetes mellitus was previously considered a CHD risk equivalent [16]. However, more recent data have shown that the CVD risk of patients with diabetes is markedly variable [17]. The present study population was comprised of middle-aged diabetic patients who did not have known CAD but had at least one additional cardiovascular risk factor. Half of the study cohort had an estimated PCE 10-year ASCVD risk of > 20%, considered a high-risk cut-off, as might be expected in diabetics. Notwithstanding, almost a quarter of the study population did not have any coronary artery calcification. CACS has significant prognostic value across a wide spectrum of risk factor profiles [4]. In particular, individuals who do not have any coronary artery calcification have low risk of cardiovascular events, and in this population CACS can serve as a “negative risk marker” [17, 18]. In the current study, among individuals classified at lower risk based on estimated PCE risk score < 20%, the presence of any CACS (> 0) was associated with four times higher event rates than in those with zero CACS. In contrast, among individuals traditionally identified as high risk based on estimated PCE risk score > 20%, CACS of zero was associated with remarkably low event rates. Of note, the PCE was previously reported to overestimate risk compared to the MESA multiethnic cohort of patients without baseline clinical CVD [19]. In the present analysis, the ability of low CACS to reclassify risk was further demonstrated by the MESA risk score estimation incorporating CACS. Malik and colleagues have analyzed diabetic patients participating in the MESA study, reporting that 38% of the patients with diabetes had a CACS of zero with annual CHD event rate similar to those without diabetes [20]. In a previous meta-analysis of 8 studies, CACS ≥ 10 AU in people with type 2 diabetes predicted all-cause mortality and cardiovascular events, suggesting that a finding of CACS < 10 may be clinically used to identify diabetic patients at low risk within this high-risk population [21]. Overall, these findings imply that CACS testing in diabetic patients could help reclassify risk of a significant proportion of patients to a lower risk group in the absence of coronary artery calcification.

Coronary artery calcification is often detected in asymptomatic individuals without known CVD. In a large cohort of asymptomatic individuals referred for CACS (only 7% diabetics), the addition of CACS to the PCE and to MESA_{(without CACS)} improved risk discrimination, particularly in the borderline and intermediate risk group [22]. Demonstration of subclinical atherosclerosis by CACS testing may benefit individuals who are at intermediate risk by traditional risk scores and have the greatest potential for risk reclassification and modification of outcome by intensification of preventive measures [23]. This may aid clinicians when evaluating the risk/benefit of preventive medications in diabetics, including the addition of aspirin which may be associated with adverse gastrointestinal effects, or the cost-effectiveness of PCSK9 monoclonal antibodies beyond statins [24]. In these cases, the use of CACS may improve the allocation of newer therapies and identify patients at higher risk who will have greater benefit in risk reduction. However, although diabetic patients with zero calcium have good long-term prognosis, and recent findings suggest that among individuals at intermediate risk with risk-enhancing factors, cardiovascular event rates are generally lower than the recommended threshold to initiate statin therapy when the CACS is zero [25], statins should be given to diabetic patients according to current clinical guidelines until prospective studies are performed. The presence and severity of CACS was repeatedly shown to improve discrimination and reclassification of future cardiovascular risk when compared to traditional risk factors, including in patients with type 2 diabetes [20, 26, 27]. Our findings further display the additive discriminatory capacity for MACE when adding CACS to clinical risk scores in diabetics. The discriminatory ability was even stronger when evaluating MI or cardiovascular death, excluding stroke as an outcome. CACS was shown in previous studies to serve as an independent predictor also for cerebrovascular events, improving discriminatory capacity [28, 29]. However, in a recent analysis of asymptomatic individuals, CACS was consistently a better predictor of CHD than stroke [30]. Of note, the MESA risk score was designed as predictor of CHD events and not stroke. Our findings are in line with these results, as both clinical risk calculators were predictors of stroke only if data on CACS was added, with a discriminatory ability that was far lower for stroke than that for MI or cardiovascular death.

Interestingly, in a restricted spline model, the adjusted HR for 10-year adverse events increased rapidly with the initial rise in CACS levels but became more moderate when CACS was > 400 AU for MACE and > 600 AU for MI or cardiovascular death. Although a wide confidence interval limits the interpretation of this analysis, it is possible that additional calcification of older plaques leads to their stabilization, while plaques with mild calcification are more active and prone to rupture or erosion, resulting in acute cardiovascular events [9, 31]. In this context, it should be noted that a significant proportion of the study population was treated with statins, which may have influenced the progression of CACS [32]. Statins were reported to increase coronary artery calcium density without increasing total calcium volume, which may stabilize existing plaque [33]. In addition, it was shown that CACS retains its predictive value among patients already on statin therapy, and therefore the utility of CACS testing should be maintained in diabetics, a population in whom the role of statins is unequivocal [34].

Strengths of this study include the prospective design and the recruitment of diabetic patients in a community-based setting under routine medical care. In addition, the availability of data on diabetes duration, baseline severity and complications, as well as long-term follow-up adds to the robustness of the results. Several limitations should be noted. First, patients with significant chronic kidney disease were excluded due to the use of CT angiography and the risk of contrast induced nephropathy. Moreover, most of our patients were under preventive medications at recruitment. These factors may impact on the generalizability of our findings. Second, the composite outcome event in the current study was not identical to that estimated by the PCE or MESA risk scores. Therefore, we did not investigate measures of calibration between models. Finally, the study was not designed to define treatment or to show benefit of routine screening on clinical outcomes. Treating physicians and patients received an assessment of risk as low, intermediate or high based on CACS. It is unknown whether this may have influenced clinician-patient behavior, and therefore impacted on clinical management.

Although clinical risk scores may provide important prognostic information, CACS is an independent predictor of cardiovascular events that provides additive discriminatory capacity and an ability to reclassify risk of diabetic patients without known CAD. This enhances risk assessment and may help personalize the use of preventive medications and lifestyle interventions with a potential for improved outcomes.