Patients with diabetes mellitus (DM) were generally found to have two to four times of the risk to develop cardiovascular events compared to those without DM. Accurate cardiovascular disease (CVD) risk assessment is critical for patients with DM to guide the preventive therapy. Approaches of evaluating the CVD risk for those with DM included the historical “CVD risk equivalent” approach and the current risk score approach. Some risk reclassification tools are recommended when the treatment decisions are not clear based on risk score assessment. The current study investigated all three approaches in CVD risk assessment among patients with DM.
We identified the predictors of CVD risk equivalent and redefined the CVD risk equivalent conditions in DM patients in a pooled cohort of four large US community-based cohorts. We examined the relative CVD risk comparing those with DM but no CVD history (DM+/CVD-) vs. those with no DM but a CVD history (DM-/CVD+) at baseline. Overall DM+/CVD- had 17% lower CVD risk than those with DM-/CVD+. DM+/CVD- participants with HbA1c≥7%, DM duration over 10 years, or DM medication use had similar CVD risk as those with DM-/CVD+ while those without these factors had lower CVD risk. Subgroup analysis comparing the hazard ratios (HR) of DM+/CVD- vs. DM-/CVD+ was done by conventional CVD risk factors. DM+/CVD- were found to have similar CVD risk as those DM-/CVD+ among women, those age <55 years, White race, or with high triglycerides groups. One with DM+/CVD- was defined to have CVD risk equivalent DM if his/her relative CVD risk was as high as or higher than that if he/she had DM-/CVD+. The CVD risk profile and CVD risk were compared between the CVD risk equivalent subgroups in DM+/CVD-. Among those with DM+/CVD-, 17.5% were found to have CVD risk equivalent DM, who had lower mean 10-year ASCVD risk score compared to those with non-CVD-risk equivalent DM (14.8% vs. 22.7%, p<0.0001) however had much higher observed CVD risk, with adjusted hazard ratios (HRs) compared to those with DM-/CVD- being 2.65 (95% CI: 2.37-2.97) vs. 1.40 (95% CI: 1.31-1.49) , respectively.
We developed and validated a set of new risk scores for DM macrovascular complications from a pooled cohort of the US population. We pooled 4,183 CVD-free adults with DM (aged 30-86 years, 45% male and 45% Black) from five US population-based cohorts. We developed 10-year Diabetes Mellitus Risk Scores (DMRS) for total CVD [myocardial infarction, cardiac revascularization, stroke, heart failure (HF) and CVD death], atherosclerotic CVD (ASCVD), and separately for coronary heart disease (CHD), stroke and HF. Age, sex, hemoglobin A1c (HbA1c), serum creatinine, systolic blood pressure and current smoking were the most important predictors of all endpoints. DMRS had good internal discrimination and calibration (c-statistics: 0.70-0.76; calibration slopes: 1.03-1.16 comparing observed vs. predicted risk). Scores were externally validated in 6642 CVD-free subjects from the Action to Control Cardiovascular Risk in Diabetes trial Follow-on (ACCORDION) cohort and were compared with Framingham Risk Scores (FRS), UK Prospective Diabetes Study (UKPDS) risk engines and 2013 Pooled Cohort Equation (PCE) for each endpoint. In the ACCORDION cohort, DMRS showed superior performance over FRS, UKPDS and PCE (c-statistics 0.62-0.71 vs. 0.55-0.60, p <0.05 for CVD comparing DMRS vs. FRS and PCE and CHD comparing DMRS vs. FRS).
In addition, we comprehensively evaluated the incremented prediction from three subclinical atherosclerosis (SA) measures, namely coronary artery calcium (CAC), carotid intima media thickness (CIMT) and ankle brachial index (ABI) beyond the DMRS in 931 CVD-free subjects with DM (mean age of 62.3 years, with 43.8% males) in the MESA cohort. CAC was found to be associated with CVD, ASCVD, CHD, HF and stroke after adjustment of DMRS (HR ranged 1.11-1.28, all p <0.05). We calculated the Harrell’s c-statistics and net reclassification index (NRI) in the following model comparisons for each event: (1) single SA measures + DMRS vs. DMRS; (2) pairwise comparison of three models with single SA measure + DMRS; (3) CAC+CIMT (or ABI, or CIMT+ABI)+DMRS vs. CIMT(or ABI, or CIMT+ABI)+DMRS. The Harrell’s c-statistics of DMRS were 0.65, 0.66, 0.66, 0.68 and 0.65 for CVD, ASCVD, CHD, HF and stroke, respectively. CAC+DMRS increased the C-statistics to 0.70, 0.68, 0.74, 0.68 and 0.62 (p value <0.05 for CVD and CHD) while the change was minimal with the addition of CIMT or ABI to DMRS. CAC showed superiority in c-statistics and NRI to CIMT and ABI as well as beyond CIMT, ABI or both for CVD and CHD events. The results demonstrated that CAC remained the strongest CVD risk reclassifier among CAC, CIMT and ABI for patients with DM.
The new definition of CVD risk equivalent DM and its algorithm has the potential to help pick those whose DM is more severe than other DM patients regarding the CVD risk. More importantly, the high DM-conferred CVD risk in those with CVD risk equivalent DM were not captured by the current CVD risk assessment tools like PCE which only includes DM as binary predictors and neglects all the heterogenous CVD risk associated with DM. As to the estimation of global CVD risk, our new DMRS were demonstrated to have better prediction performance than existing risk scores including PCE, FRS and UKPDS in the DM population. Given that our DMRS were not yet perfect CVD risk estimation tools, we can further use cardiac CT scanning to get CAC score, which were found to have superior reclassification and discrimination ability to CIMT and ABI, to assist the CVD risk assessment for patients with DM.