OBJECTIVE

To examine the utility of repeated computed tomography (CT) coronary artery calcium (CAC) testing, we assessed risks of detectable CAC and its cardiovascular consequences in individuals with and without type 2 diabetes ages 45–85 years.

RESEARCH DESIGN AND METHODS

We included 5,836 individuals (618 with type 2 diabetes, 2,972 without baseline CAC) from the Multi-Ethnic Study of Atherosclerosis. With logistic and Cox regression we evaluated the impact of type 2 diabetes, diabetes treatment duration, and other predictors on prevalent and incident CAC. We used time-dependent Cox modeling of follow-up data (median 15.9 years) for two repeat CT exams and cardiovascular events to assess the association of CAC at follow-up CT with cardiovascular events.

RESULTS

For 45 year olds with type 2 diabetes, the likelihood of CAC at baseline was 23% vs. 17% for those without. Median age at incident CAC was 52.2 vs. 62.3 years for those with and without diabetes, respectively. Each 5 years of diabetes treatment increased the odds and hazard rate of CAC by 19% (95% CI 8–33) and 22% (95% CI 6–41). Male sex, White ethnicity/race, hypertension, hypercholesterolemia, obesity, and low serum creatinine also increased CAC. CAC at follow-up CT independently increased coronary heart disease rates.

CONCLUSIONS

We estimated cumulative CAC incidence to age 85 years. Patients with type 2 diabetes develop CAC at a younger age than those without diabetes. Because incident CAC is associated with increased coronary heart disease risk, the value of periodic CAC-based risk assessment in type 2 diabetes should be evaluated.

Numerous studies have documented the value of computed tomography (CT) for coronary artery calcium (CAC) in improving risk prediction for atherosclerotic cardiovascular disease (ASCVD). The incremental predictive value of CAC generalizes across several subpopulations with different risk profiles (1), including those with type 2 diabetes (2). The 2018 ACC/AHA guidelines on cholesterol management (3) recommend use of CT-CAC for refining risk assessment and shared decision-making for statin therapy in individuals with intermediate 10-year ASCVD risk (≥7.5% to <20%) and for select adults with borderline risk (5% to <7.5%). When the CAC score equals 0 Agatston units, the guidelines suggest that statin therapy may be less beneficial and can be deferred due to low expected ASCVD rates and absolute risk reduction. With nonzero scores, initiation of statin therapy should be considered.

For most patients with type 2 diabetes, initiation of statins is recommended regardless of estimated 10-year ASCVD risk, although periodic risk assessment and CT-CAC may inform decisions regarding statin dosing or addition of ezetimibe and targets for antihypertensive therapy. However, guidelines do not currently include clear recommendations for repeat CAC measurement, and more research is needed for evaluation of its clinical value in those with an initial CAC score of 0—a group with overall excellent prognosis (4).

It has been hypothesized that individuals who develop incident CAC may have a higher lifetime risk of ASCVD, as they are likely to be on a different long-term risk trajectory than those who remain at 0 (4). Therefore, the clinical utility of repeating CAC measurement depends on 1) the cumulative risk of newly detectable CAC and 2) its impact on long-term cardiovascular event rates that can potentially be modified by preventive interventions.

In previous analyses of Multi-Ethnic Study of Atherosclerosis (MESA) data, the relative risk was reported of several factors including type 2 diabetes for incident CAC (5,6). We extend these analyses by assessing absolute risks of developing CAC in subjects with and without type 2 diabetes across individuals’ life spans. In addition, we evaluate the independent association of CAC detected at follow-up CT with various cardiovascular events.

Study Design and Population

MESA was designed for evaluation of risk factors for ASCVD in a multiethnic cohort without a history of clinical ASCVD at baseline. A total of 6,814 U.S. individuals enrolled in MESA, ages 45–84 years, who self-identified as White, Black, Hispanic, or Asian. Study participants were recruited from six U.S. communities: Forsyth County, NC, Northern Manhattan and the Bronx, NY, Baltimore and Baltimore County, MD, St. Paul, MN, Chicago, IL, and Los Angeles County, CA, and had baseline visits between 2000 and 2002. For this study we used the National Heart, Lung, and Blood Institute (NHLBI) Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC) MESA data sets and supplemented with follow-up through 31 December 2017. CT-CAC scanning was completed in all participants at baseline exam 1. Based on a random selection procedure, follow-up CT-CAC scans were performed for 2,953 participants at exam 2 (September 2002–January 2004) and 2,805 participants at exam 3 (March 2004–July 2005). CT-CAC scans were performed for a more limited group of participants at exam 4 (September 2005–May 2007) and exam 5 (April 2010–December 2011): 1,406 and 3,305 participants, respectively. For the current study, from the 6,814 MESA participants, we excluded those with missing diabetes status (N = 24), missing follow-up (N = 20), and self-reported physician diagnosis of type 1 diabetes (N = 10) (Supplementary Fig. 1).

Figure 1

Likelihood of incident CAC at various ages by baseline age (years), sex, and type 2 diabetes status.

Figure 1

Likelihood of incident CAC at various ages by baseline age (years), sex, and type 2 diabetes status.

Predictor and Outcome Variables

Demographics, smoking status, medical history, and medication use were obtained by questionnaires. Physical exams included measurement of height and weight and three measurements of resting blood pressure. Blood pressure levels were calculated from the average of the last two measures. Serum laboratory tests included lipids, fasting glucose, and creatinine. Type 2 diabetes was defined according to fasting glucose ≥126 mg/dL, use of oral hypoglycemic medication and/or insulin, or self-reported physician diagnosis. CAC was measured twice for each participant either with electron-beam CT or multidetector row helical CT. Scans were read independently at a centralized reading center. The results of the two scans were averaged, and the amount of calcium was quantified with the Agatston scoring method. Inter- and intraobserver agreement was very high (0.93 and 0.90, respectively) (7).

Information on the occurrence of clinical events was obtained at intervals of 9–12 months by telephone interview. Events were adjudicated based on abstracted medical records and independent assessment by two physicians using predefined criteria. We defined ASCVD as myocardial infarction, cardiac death, or fatal or nonfatal stroke based on the definition used in the Pooled Cohort Equations (8). Coronary heart disease was defined as myocardial infarction or cardiac death. Stroke was defined as a documented focal neurologic deficit lasting 24 h or until death or, if <24 h, a clinically relevant lesion on brain imaging. Heart failure required a combination of heart failure symptoms, physician diagnosis, and medical treatment. We further specified heart failure by ejection fraction <50% (heart failure with reduced ejection fraction) vs. ≥50% (heart failure with preserved ejection fraction) based on echocardiogram or other imaging studies assessed during the hospitalization when available in medical records. We defined major adverse cardiovascular events (MACE) as myocardial infarction, percutaneous coronary intervention, coronary artery bypass surgery, angina, peripheral artery disease, stroke, or transient ischemic attack.

Statistical Analysis

To increase the robustness of prediction modeling of CAC at baseline and follow-up CT, we included those with a follow-up CT-CAC at exam 2 or 3 and complete predictor variables (N = 5,836) (Supplementary Fig. 1). Within this study population, 2,972 had a 0 CAC score and 618 were classified as having type 2 diabetes at baseline. We used logistic regression to model CAC present at baseline.

To model incident CAC at follow-up CT, we restricted the study population to those with a CAC score of 0 at baseline and performed cause-specific Cox regression with attained age as time scale and delayed entry. We defined delayed entry on the basis of age at exam 2, when the follow-up CT was scheduled for exam 3, to ensure that these participants were considered at risk for developing CAC in a fashion similar to that for participants who had the follow-up CT scheduled for exam 2. We used death due to other causes and MACE as competing events. We considered MACE a competing event, arguing that a history of MACE would nullify the clinical relevance of CAC for initiation of preventive medication. Cumulative incidence was calculated as the sum of chances for new detection of CAC at each possible age given participants remaining CAC and MACE free until that age.

Predictors were selected from full models including all candidate predictors (5,9) in a backward stepwise manner with Akaike information criterion as selection criterion. We included serum creatinine, as high creatinine has, surprisingly, been associated with lower risk of incident CAC (5). We tested for interaction between included predictors and type 2 diabetes and explored nonlinear associations of continuous variables by including quadratic terms. We selected interaction and nonlinear terms based on a likelihood ratio test with P < 0.01 as criterion. For Cox regression models, we tested for violation of the proportional hazards assumption (10) with P < 0.05 as an indication for violation and left out predictors when indicated. In addition, we developed a color chart for cumulative incidences of CAC through age 85 years based on models including sex and baseline type 2 diabetes status.

Because neighborhood socioeconomic condition levels affect the performance of cardiovascular risk predictions (11), we performed an internal-external cross-validation procedure (12) to assess predictive performance of the multivariable models across diverse populations according to a neighborhood social cohesion score. This score is a sum score of responses on a 5-point Likert scale to questions relating to trust in neighbors, shared values with neighbors, willingness to help neighbors, and extent to which neighbors got along, with neighborhood defined as an area ∼1 mile (1.6 km) around the participant’s home. Lower social cohesion correlates with cardiovascular risk factors and event rates in MESA (1315). In each validation step, we refitted logistic regression and cumulative incidence functions combining data from three approximately equally sized groups defined by quartiles of social cohesion scores, and validity was tested in the one subset left out. For assessing predictive performance, we used the C-statistic for discrimination and the observed–to–expected risk ratio for calibration. For assessing recalibration, we estimated the calibration-in-the-large intercept (0 = optimal) and calibration intercept plus slope (1 = optimal) for logistic regression (16). For cause-specific Cox regression equations, only the calibration slope should be estimated (1 = optimal on the log hazard scale) (17).

Finally, we evaluated the independent association of CAC detected at follow-up CT with ASCVD events and heart failure. To increase the robustness of epidemiological modeling, we used time-dependent Cox regression of the larger data set of MESA participants with complete data on baseline type 2 diabetes and CAC status (N = 6,760) and follow-up for two repeat CT scans (first at exam 2 or 3 and second at exam 5), cardiovascular events, and death through 31 December 2017. We included participants with CAC at baseline to better define the potential value of acquiring additional information about CAC status at a follow-up CT versus just relying on CAC status assessed only once. To account for missing values in follow-up CT-CAC and covariables, we performed multiple imputation 20 times with a flexible additive model including covariables; status variables for ASCVD, heart failure, and death; and the Nelson-Aalen estimators of the cumulative hazard for these events. We included CAC status at exam 4 to improve imputations. Rubin’s rules were used for summarizing the effect estimates and variances of the 20 different analyses across multiple imputed data sets.

For these time-dependent Cox regression analyses, we again used age as the time scale and stratified by age periods preceding and following the first follow-up CT-CAC exam at exam round 2 or 3. Sex and race/ethnicity were included as fixed covariates. Time-dependent covariates included type 2 diabetes status, statin therapy, antihypertensive therapy, and the traditional cardiovascular risk factors as included in the Pooled Cohort Equations (8). The value of these time-dependent covariates was assessed at baseline and at follow-up CT-CAC exams 2 or 3 and 5. We additionally censored follow-up at the CT-CAC exam date if a MACE event (i.e., excluding the ASCVD event of interest) occurred prior to the follow-up scan. We first tested whether the hazard ratio of CAC would change when its value is additionally based on follow-up CT and subsequently assessed whether this hazard ratio would change with adjustment for baseline CAC status. We tested for interaction of CAC with baseline type 2 diabetes status. We also assessed whether associations with heart failure were altered by sex and preceding coronary heart disease events (18,19). We used a likelihood ratio test modified for multiple imputations, with α = 0.05 as the criterion for statistical significance.

All analyses were performed with R, version 4.0.3 (R Foundation for Statistical Computing [https://www.r-project.org/]).

Institutional Review Board Approval

All MESA participants gave informed consent, and the study protocol was approved by the institutional review board (IRB) at each site. The Mount Sinai Hospital IRB deemed the study protocol as exempt from IRB review.

Study Population

Individuals with CAC present at baseline generally were older, more often were male and non-Hispanic White, and more frequently were hypertensive and taking statins. If they had type 2 diabetes, they had longer treatment duration (Table 1). Of the 2,972 MESA participants with CAC score 0 at baseline (median follow-up until follow-up CT-CAC exam 2 or 3: 2.3 years), 456 subjects had CAC newly detected at follow-up without experiencing MACE first and 60 subjects experienced a MACE or died prior and were censored for incident CAC.

Table 1

Study population (N = 5,836) baseline characteristics

No CAC at baseline (N = 2,972)CAC at baseline (N = 2,864)
Diabetes (N = 242)No diabetes (N = 2,730)Diabetes (N = 376)No diabetes (N = 2,488)
Age, years 59 (53–67) 56 (50–65) 67 (60–73) 67 (59–74) 
Male 96 (40) 1,009 (37) 233 (62) 1,433 (58) 
Race/ethnicity     
 Non-Hispanic White 27 (11) 980 (36) 99 (26) 1,175 (47) 
 Non-Hispanic Asian 20 (8) 327 (12) 38 (10) 296 (12) 
 Non-Hispanic Black 120 (50) 805 (29) 126 (34) 546 (22) 
 Hispanic 75 (31) 618 (23) 113 (30) 453 (18) 
Social cohesion score 17 (15–19) 18 (16–19) 18 (16–20) 18 (16–20) 
Education     
 Less than high school 62 (26) 420 (15) 108 (29) 384 (15) 
 High school 52 (22) 441 (16) 79 (21) 479 (19) 
 Some college/technical 69 (29) 813 (30) 105 (28) 685 (28) 
 College or graduate 59 (24) 1056 (39) 84 (22) 940 (38) 
Current smoker 32 (13) 340 (12) 48 (13) 318 (13) 
BMI, kg/m2 30.6 (26.8–35.0) 27.2 (24.2–30.9) 29.8 (26.4–33.6) 27.3 (24.4–30.6) 
Systolic blood pressure, mmHg 128 (115–143) 118 (107–133) 133 (118–148) 128 (115–142) 
Diastolic blood pressure, mmHg 72 (65–80) 71 (65–78) 72 (66–79) 73 (66–79) 
Antihypertensive treatment 147 (61) 687 (25) 238 (63) 1037 (42) 
Total cholesterol, mg/dL 186 (162–208) 192 (171–214) 185 (164–208) 193 (171–216) 
HDL cholesterol, mg/dL 46 (40–54) 51 (42–62) 43 (37–53) 48 (40–58) 
LDL cholesterol, mg/dL 110 (89–135) 116 (96–135) 110 (90–129) 118 (98–138) 
Triglycerides, mg/dL 113 (85–176) 104 (73–150) 138 (88–197) 111 (79–159) 
Statins 49 (20) 234 (9) 112 (30) 463 (19) 
Serum creatinine, mg/dL 0.9 (0.7–1.0) 0.9 (0.8–1.0) 1.0 (0.8–1.1) 1.0 (0.8–1.1) 
Diabetes treatment duration, years 3.0 (0.0–8.0) 4.0 (0.0–13.0) 
No CAC at baseline (N = 2,972)CAC at baseline (N = 2,864)
Diabetes (N = 242)No diabetes (N = 2,730)Diabetes (N = 376)No diabetes (N = 2,488)
Age, years 59 (53–67) 56 (50–65) 67 (60–73) 67 (59–74) 
Male 96 (40) 1,009 (37) 233 (62) 1,433 (58) 
Race/ethnicity     
 Non-Hispanic White 27 (11) 980 (36) 99 (26) 1,175 (47) 
 Non-Hispanic Asian 20 (8) 327 (12) 38 (10) 296 (12) 
 Non-Hispanic Black 120 (50) 805 (29) 126 (34) 546 (22) 
 Hispanic 75 (31) 618 (23) 113 (30) 453 (18) 
Social cohesion score 17 (15–19) 18 (16–19) 18 (16–20) 18 (16–20) 
Education     
 Less than high school 62 (26) 420 (15) 108 (29) 384 (15) 
 High school 52 (22) 441 (16) 79 (21) 479 (19) 
 Some college/technical 69 (29) 813 (30) 105 (28) 685 (28) 
 College or graduate 59 (24) 1056 (39) 84 (22) 940 (38) 
Current smoker 32 (13) 340 (12) 48 (13) 318 (13) 
BMI, kg/m2 30.6 (26.8–35.0) 27.2 (24.2–30.9) 29.8 (26.4–33.6) 27.3 (24.4–30.6) 
Systolic blood pressure, mmHg 128 (115–143) 118 (107–133) 133 (118–148) 128 (115–142) 
Diastolic blood pressure, mmHg 72 (65–80) 71 (65–78) 72 (66–79) 73 (66–79) 
Antihypertensive treatment 147 (61) 687 (25) 238 (63) 1037 (42) 
Total cholesterol, mg/dL 186 (162–208) 192 (171–214) 185 (164–208) 193 (171–216) 
HDL cholesterol, mg/dL 46 (40–54) 51 (42–62) 43 (37–53) 48 (40–58) 
LDL cholesterol, mg/dL 110 (89–135) 116 (96–135) 110 (90–129) 118 (98–138) 
Triglycerides, mg/dL 113 (85–176) 104 (73–150) 138 (88–197) 111 (79–159) 
Statins 49 (20) 234 (9) 112 (30) 463 (19) 
Serum creatinine, mg/dL 0.9 (0.7–1.0) 0.9 (0.8–1.0) 1.0 (0.8–1.1) 1.0 (0.8–1.1) 
Diabetes treatment duration, years 3.0 (0.0–8.0) 4.0 (0.0–13.0) 

Data are median (interquartile range) or count (%). To convert total, HDL, or LDL cholesterol to mmol/L, multiply by 0.0259; to convert triglycerides to mmol/L, multiply by 0.0113; and to convert creatinine to μmol/L, multiply by 88.4.

Long-term Predictions of CAC to 85 Years of Age

The estimated likelihood of having CAC at baseline was 23% for a 45-year-old individual with type 2 diabetes vs. 17% without. For each 5 years of diabetes treatment duration, the odds of CAC at baseline independently increased by 19% (95% CI 8–33) (Table 2).

Table 2

Multivariable-adjusted models for predicting CAC at baseline and follow-up CT exam

CAC at baseline (2,864 of 5,836)Incident CAC (456 of 2,972)Competing MACE or death (60 of 2,972)
Age per 10 years 2.47 (2.30–2.65) N/A N/A 
Male 2.73 (2.39–3.13) 1.66 (1.31–2.10)  
Race/ethnicity vs. Non-Hispanic White    
 Non-Hispanic Asian 0.83 (0.68–1.02) 0.58 (0.40–0.83)  
 Non-Hispanic Black 0.44 (0.38–0.52) 0.75 (0.59–0.94)  
 Hispanic 0.56 (0.47–0.66) 0.78 (0.60–1.00)  
Current smoker 1.71 (1.43–2.05)  2.17 (1.13–4.17) 
BMI per 5 kg/m2 1.10 (1.04–1.18) *  
Systolic blood pressure per 10 mmHg 1.06 (1.03–1.10)   
Diastolic blood pressure per 10 mmHg — 1.08 (0.98–1.19)  
Antihypertensive treatment 1.42 (1.24–1.62) 1.44 (1.17–1.77)  
HDL cholesterol per 10 mg/dL 0.95 (0.90–1.00) *  
LDL cholesterol per 10 mg/dL 1.07 (1.05–1.09)  0.94 (0.86–1.02) 
Triglycerides per 10 mg/dL 1.01 (1.00–1.02)   
Statins 1.81 (1.52–2.16) 1.32 (1.00–1.74)  
Serum creatinine, mg/dL  0.55 (0.31–0.96) 1.85 (0.86–3.99) 
Diabetes treatment duration per 5 years 1.19 (1.08–1.33) 1.22 (1.06–1.41) 1.34 (1.00–1.79) 
CAC at baseline (2,864 of 5,836)Incident CAC (456 of 2,972)Competing MACE or death (60 of 2,972)
Age per 10 years 2.47 (2.30–2.65) N/A N/A 
Male 2.73 (2.39–3.13) 1.66 (1.31–2.10)  
Race/ethnicity vs. Non-Hispanic White    
 Non-Hispanic Asian 0.83 (0.68–1.02) 0.58 (0.40–0.83)  
 Non-Hispanic Black 0.44 (0.38–0.52) 0.75 (0.59–0.94)  
 Hispanic 0.56 (0.47–0.66) 0.78 (0.60–1.00)  
Current smoker 1.71 (1.43–2.05)  2.17 (1.13–4.17) 
BMI per 5 kg/m2 1.10 (1.04–1.18) *  
Systolic blood pressure per 10 mmHg 1.06 (1.03–1.10)   
Diastolic blood pressure per 10 mmHg — 1.08 (0.98–1.19)  
Antihypertensive treatment 1.42 (1.24–1.62) 1.44 (1.17–1.77)  
HDL cholesterol per 10 mg/dL 0.95 (0.90–1.00) *  
LDL cholesterol per 10 mg/dL 1.07 (1.05–1.09)  0.94 (0.86–1.02) 
Triglycerides per 10 mg/dL 1.01 (1.00–1.02)   
Statins 1.81 (1.52–2.16) 1.32 (1.00–1.74)  
Serum creatinine, mg/dL  0.55 (0.31–0.96) 1.85 (0.86–3.99) 
Diabetes treatment duration per 5 years 1.19 (1.08–1.33) 1.22 (1.06–1.41) 1.34 (1.00–1.79) 

Shown are odds ratios (95% CI) for predicting CAC at baseline and hazard ratios (95% CI) for predicting incident CAC (censoring for MACE or death) and competing MACE or death (censoring for incident CAC).

*

Left out due to P value <0.05 for violation of proportional hazard assumption N/A, not applicable (age was used as time scale).

The cumulative incidence of CAC was calculated at ages 45.5–87.6. The median age at incident CAC was 52.2 years for type 2 diabetes vs. 62.3 years without type 2 diabetes (Supplementary Fig. 2). The cumulative incidence of CAC at age 70 years for men with type 2 diabetes was 85%, resembling risk at age 85 years for women without type 2 diabetes (Fig. 1). For each 5 years of diabetes treatment duration, the hazard rate of incident CAC increased by 22% (95% CI 6–41) within the multivariable analysis. Other variables increasing the likelihood of incident CAC included male sex, nonHispanic White race/ethnicity, hypertension, hypercholesterolemia, and low serum creatinine. An overall test of interaction between these predictors and type 2 diabetes showed no indication of significant modification of associations: P = 0.42. The direction of hazard ratios for incident CAC was in agreement with odds ratios for CAC at baseline (Table 2).

For prediction of CAC presence at baseline, the C-statistic was 0.79 (95% CI 0.78–0.80) and varied from 0.76 to 0.81 at cross validation. Calibration was excellent, with observed-to-expected ratios ranging between 0.96 and 1.11, intercept differences varying around zero (−0.09 to 0.15), and calibration slopes varying around 1 (0.85–1.09) (Supplementary Table 1).

The C-statistic of cumulative incidence predictions was 0.69 (95% CI 0.67–0.72) and varied from 0.65 to 0.69 at cross validation. The mismatch between observed and expected cumulative incidence was small, with observed-to-expected ratios of 0.87–1.18. Although calibration slopes of cause-specific Cox regression equations within the cumulative incidence function deviated from 1, 1 was included in all 95% CIs (Supplementary Table 1).

Association of CAC at Follow-up CT With Cardiovascular Events

Of the 6,760 MESA participants (median follow-up time 15.9 years [range 0.2–17.4]), 3,897 had a detectable CAC score at the follow-up CT scan. A total of 735 participants developed ASCVD (472 coronary and 304 stroke events) and 383 developed heart failure, of whom 219 were estimated to have reduced ejection fraction.

Detectable CAC at baseline independently increased rates for all outcomes, and for heart failure the increase was particularly driven by heart failure with reduced ejection fraction (Table 3). The association of CAC at the baseline CT exam and heart failure (hazard ratio 1.45 [95% CI 1.12–1.88]), however, remained similar after additional adjustment for preceding coronary heart disease as a time-dependent covariate: 1.38 (95% CI 1.07–1.78). In modeling heart failure event rates, interactions of CAC status with sex were not statistically significant.

Table 3

Time-dependent Cox modeling of ASCVD, coronary heart disease, stroke, and heart failure

Hazard ratios (95% CI)
ASCVD (735 of 6,760)CHD (472 of 6,760)Stroke (304 of 6,760)Heart failure
Any (383 of 6,760)rEF (219 of 6,760)pEF (164 of 6,760)
Model 1: fixed covariate for CAC       
CAC at baseline CT exam 2.03 (1.68–2.44) 2.66 (2.06–3.43) 1.38 (1.05–1.81) 1.45 (1.12–1.88) 1.55 (1.09–2.21) 1.33 (0.89–2.00) 
Model 2: time-dependent covariate for CAC       
CAC at baseline or follow-up CT exam 2.26 (1.78–2.86) 3.26 (2.27–4.69) 1.47 (1.06–2.05) 1.42 (1.03–1.95) 1.30 (0.87–1.95) 1.63 (0.96–2.76) 
P, model 2 vs. 1 0.17 0.23 0.70 <0.001 0.001 0.71 
Model 3: allowing hazard ratio to vary       
CAC at baseline 2.49 (1.37–4.53) 3.29 (1.43–7.60) 1.88 (0.81–4.40) 1.09 (0.59–1.99) 1.12 (0.50–2.53) 1.03 (0.35–3.00) 
CAC at follow-up CT exam 2.22 (1.72–2.86) 3.26 (2.18–4.86) 1.41 (0.98–2.01) 1.55 (1.07–2.24) 1.36 (0.86–2.17) 1.88 (1.02–3.45) 
P, model 3 vs. 2 0.52 0.96 0.80 <0.001 <0.001 <0.001 
Model 4: adjusting for CAC at baseline and allowing its hazard ratio to vary       
CAC at baseline CT exam - until follow-up CT exam 2.53 (1.39–4.61) 3.37 (1.46–7.77) 1.90 (0.81–4.43) 1.10 (0.60–2.02) 1.14 (0.50–2.58) 1.03 (0.35–3.02) 
CAC at baseline CT exam - after follow-up CT exam 1.57 (1.24–1.98) 1.84 (1.34–2.53) 1.19 (0.83–1.70) 1.39 (0.98–1.97) 1.75 (1.05–2.92) 1.11 (0.68–1.82) 
CAC at follow-up CT exam 1.62 (1.19–2.22) 2.12 (1.30–3.45) 1.25 (0.80–1.95) 1.23 (0.78–1.94) 0.91 (0.48–1.71) 1.75 (0.87–3.53) 
P, model 4 vs. 3 <0.001 <0.001 0.56 0.27 0.14 0.92 
Hazard ratios (95% CI)
ASCVD (735 of 6,760)CHD (472 of 6,760)Stroke (304 of 6,760)Heart failure
Any (383 of 6,760)rEF (219 of 6,760)pEF (164 of 6,760)
Model 1: fixed covariate for CAC       
CAC at baseline CT exam 2.03 (1.68–2.44) 2.66 (2.06–3.43) 1.38 (1.05–1.81) 1.45 (1.12–1.88) 1.55 (1.09–2.21) 1.33 (0.89–2.00) 
Model 2: time-dependent covariate for CAC       
CAC at baseline or follow-up CT exam 2.26 (1.78–2.86) 3.26 (2.27–4.69) 1.47 (1.06–2.05) 1.42 (1.03–1.95) 1.30 (0.87–1.95) 1.63 (0.96–2.76) 
P, model 2 vs. 1 0.17 0.23 0.70 <0.001 0.001 0.71 
Model 3: allowing hazard ratio to vary       
CAC at baseline 2.49 (1.37–4.53) 3.29 (1.43–7.60) 1.88 (0.81–4.40) 1.09 (0.59–1.99) 1.12 (0.50–2.53) 1.03 (0.35–3.00) 
CAC at follow-up CT exam 2.22 (1.72–2.86) 3.26 (2.18–4.86) 1.41 (0.98–2.01) 1.55 (1.07–2.24) 1.36 (0.86–2.17) 1.88 (1.02–3.45) 
P, model 3 vs. 2 0.52 0.96 0.80 <0.001 <0.001 <0.001 
Model 4: adjusting for CAC at baseline and allowing its hazard ratio to vary       
CAC at baseline CT exam - until follow-up CT exam 2.53 (1.39–4.61) 3.37 (1.46–7.77) 1.90 (0.81–4.43) 1.10 (0.60–2.02) 1.14 (0.50–2.58) 1.03 (0.35–3.02) 
CAC at baseline CT exam - after follow-up CT exam 1.57 (1.24–1.98) 1.84 (1.34–2.53) 1.19 (0.83–1.70) 1.39 (0.98–1.97) 1.75 (1.05–2.92) 1.11 (0.68–1.82) 
CAC at follow-up CT exam 1.62 (1.19–2.22) 2.12 (1.30–3.45) 1.25 (0.80–1.95) 1.23 (0.78–1.94) 0.91 (0.48–1.71) 1.75 (0.87–3.53) 
P, model 4 vs. 3 <0.001 <0.001 0.56 0.27 0.14 0.92 

All models had attained age as time scale and were adjusted for baseline sex and race/ethnicity and in addition for time-dependent type 2 diabetes status, current smoking status, systolic blood pressure, antihypertensive treatment status, total cholesterol, HDL cholesterol, and statin treatment status. CHD, coronary heart disease; pEF, preserved ejection fraction; rEF, reduced ejection fraction.

After controlling for time-dependent variables (model 3 [Table 3]), CAC at follow-up CT scan was associated with higher rates of subsequent ASCVD, coronary heart disease, and heart failure, with hazard ratios of 2.22 (95% CI 1.72–2.86), 3.26 (95% CI 2.18–4.86), and 1.55 (1.07–2.24), respectively. Conditional on baseline CAC status (model 4), the hazard ratio of CAC at follow-up CT scan remained statistically significant for ASCVD (1.62 [95% CI 1.19–2.22]), and coronary heart disease: 2.12 (95% CI 1.30–3.45). The hazard ratio of detectable CAC at follow-up CT scan was no longer statistically significant for stroke and heart failure after adjustment for baseline CAC status (Table 3). For hazard ratio results of covariables in the full models see Supplementary Tables 25.

Type 2 diabetes did not significantly alter the associations of CAC with ASCVD, coronary heart disease, stroke, or heart failure in any of the models (P values >0.05) (see Supplementary Table 6]).

Individuals with type 2 diabetes who were CAC free at the baseline CT developed detectable CAC at a younger age (∼10 years earlier) than those without type 2 diabetes. Factors that were associated with detecting CAC at a younger age included longer type 2 diabetes treatment duration, male sex, non-Hispanic White race/ethnicity, hypertension, hypercholesterolemia, and low serum creatinine. Detecting CAC at baseline was independently associated with higher total ASCVD, coronary heart disease, stroke, and heart failure rates, and newly detected CAC at follow-up scan significantly increased the risk for ASCVD and coronary heart disease.

In those with CAC score 0 at baseline, type 2 diabetes independently increased the likelihood of incident CAC and estimates of hazard ratios agreed with relative risks reported previously for MESA data (5,6). It has been well understood that high glucose levels and insulin resistance accelerate the progression of atherosclerosis and vascular calcification (1,20). Risk factors that correlate with type 2 diabetes such as leptin and hyperlipidemia have been shown to promote vascular calcification as well (1). However, thus far, the usefulness of 10-year ASCVD risk assessment in patients with type 2 diabetes has been controversial because type 2 diabetes has traditionally been considered a high ASCVD risk equivalent. Yet, multiple studies show that subgroups of patients with type 2 diabetes carry lower cardiovascular risk, especially those with a relatively brief duration of type 2 diabetes. Evidence that CT-CAC scores can potentially refine cardiovascular risk assessment in patients with type 2 diabetes has been established (2). Our results show that the absolute risk of developing newly detectable CAC is much higher for patients with type 2 diabetes, especially for those with longer diabetes treatment duration.

After adjustment for prior CAC status measured at the MESA baseline exam, detectable CAC at follow-up CT was associated with predominantly increased coronary heart disease rates. Our findings are substantially consistent with those of an earlier study with evaluation of progression of CAC and incident coronary heart disease using MESA data (21), although in the current analysis we use data on participants with much longer follow-up and use additional follow-up CT scans from MESA exam 5. Predictions of the likelihood of detecting new CAC at a follow-up CT may thus be particularly useful to guide decision-making for risk assessment intervals and incorporation of CT-CAC algorithms for those with type 2 diabetes in the prevention of ASCVD. Our findings are also consistent with results from the Heinz Nixdorff Recall Study (9), in that CAC score 0 on serial scans portended low risk, although CAC progression added only weakly to risk progression in that analysis.

Strengths and Limitations

Our study has several strengths. We used data from a large community-based and diverse cohort study, MESA, and considered predictor variables that are routinely available. We used a cumulative incidence function with attained age as the time scale that provided us the opportunity to estimate and individualize CAC incidence and competing MACE and death rates over an extensive time horizon. We assessed our model’s performance, and adequate discrimination and calibration were demonstrated across different socioeconomic groups.

Our study must be viewed in the light of potential limitations. In particular, follow-up CAC scans were not performed for all participants and, as shown by others, MESA participants without a follow-up CAC tended to have somewhat worse risk factor distributions (5). Second, in agreement with the 2018 ACC/AHA guidelines on cholesterol management (3), we evaluated the association of any nonzero CAC scores with subsequent cardiovascular event rates. In this group, some individuals develop high CAC scores, and for them associations are generally stronger (21). Third, we included variables on use of statin and antihypertensive therapy, and their associations with incident CAC were likely affected by confounding by indication and should thus be interpreted with caution. Fourth, we omitted potentially important predictors such as hemoglobin A1c (HbA1c) and a family history of ASCVD or coronary heart disease, as HbA1c was not assessed at the MESA baseline exam and the family history of heart attack question had a high number of missing values for our study population (N = 290). Lastly, we did not validate our predictions in a separate, independent data set, and the extent of generalizability to other populations should thus be assessed in further research.

The 2018 ACC/AHA guidelines recommend considering CT-CAC as a risk-enhancing factor for guiding risk discussions regarding statin therapy in individuals aged 40–75 without diabetes but with an LDL cholesterol level of 70–189 mg/dL and a borderline or intermediate 10-year ASCVD risk (3). If the CAC score is 0 for these individuals, it is considered reasonable to withhold or postpone statin therapy and reassess in 5–10 years. However, data on the clinical utility of repeated CT-CAC testing are limited, and therefore no clear recommendations are included about repeat CT-CAC testing. We have shown that repeating CT-CAC could result in identifying more individuals with high cardiovascular risk (incident CAC) as well as individuals with low risk (two measures with CAC score 0), and our predictions of CAC incidence may help guide the interval for repeat testing.

In the 2018 ACC/AHA guidelines, adults aged 40–75 years with type 2 diabetes are stratified into a high-risk category, and moderate-intensity statin therapy is recommended regardless of estimated 10-year ASCVD risk. Moreover, withholding or postponing statin therapy is not currently recommended when a CAC score of 0 has been found for patients with diabetes. Thus, additional evaluation of strategies including CAC in this population is needed for determination of the utility of CAC to inform decisions about initiating or intensifying cholesterol-lowering therapy.

Yet, for patients with type 2 diabetes, newer drugs have now become available that also reduce ASCVD and heart failure event rates (22,23), although it remains to be seen for whom these benefits outweigh the side effects and high drug prices of these novel agents. As such, improving stratification of cardiovascular risk in patients with type 2 diabetes will likely become increasingly important to ensure cost-effective preventive care. The combination of our lifetime predictions and impact of newly detectable CAC on prediction of cardiovascular disease rates can be useful for decision-making on the scheduling of repeat CT scans. The algorithms for designing periodic risk assessment programs should be individualized and optimized using the expected benefits and harms of cardiovascular treatment and screening tests.

In conclusion, we estimated cumulative CAC incidence to age 85 years. Patients with type 2 diabetes develop CAC at a younger age than those without diabetes. The risk of incident CAC further increases with duration of diabetes treatment. Because incident CAC is associated with increased coronary heart disease risk, the value of periodic CAC-based risk assessment in individuals with type 2 diabetes should be evaluated.

This article contains supplementary material online at https://doi.org/10.2337/figshare.14365538.

Acknowledgments. This manuscript was prepared using MESA Research Materials obtained from the NHLBI and NHLBI Biologic Specimen and Data Repository Information Coordinating Center and does not necessarily reflect the opinions or views of the MESA or the NHLBI.

Funding. This study was supported by American Diabetes Association grant 1-18-ICTS-041.

The American Diabetes Association had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.

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

Author Contributions. All authors contributed to study concept and design. All authors contributed to acquisition, analysis, or interpretation of data. All authors contributed to critical revision of the manuscript for important intellectual content. B.S.F. and M.G.M.H. contributed to statistical analysis. K.E.F. obtained funding. All authors provided administrative, technical, or material support. B.S.F. was responsible for the decision to submit and publish the manuscript. B.S.F. 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 American College of Cardiology’s 68th Annual ScientificSession, New Orleans, LA, 16–18 March 2019.

1.
Greenland
P
,
Blaha
MJ
,
Budoff
MJ
,
Erbel
R
,
Watson
KE
.
Coronary calcium score and cardiovascular risk
.
J Am Coll Cardiol
2018
;
72
:
434
447
2.
Malik
S
,
Zhao
Y
,
Budoff
M
, et al
.
Coronary artery calcium score for long-term risk classification in individuals with type 2 diabetes and metabolic syndrome from the Multi-Ethnic Study of Atherosclerosis
.
JAMA Cardiol
2017
;
2
:
1332
1340
3.
Lloyd-Jones
DM
,
Braun
LT
,
Ndumele
CE
, et al
.
Use of risk assessment tools to guide decision-making in the primary prevention of atherosclerotic cardiovascular disease: a special report From the American Heart Association and American College of Cardiology
.
Circulation
2019
;
139
:
e1162
e1177
4.
Khera
A
,
Greenland
P
.
Coronary artery calcium: if measuring once is good, is twice better?
Circulation
2018
;
137
:
680
683
5.
Kronmal
RA
,
McClelland
RL
,
Detrano
R
, et al
.
Risk factors for the progression of coronary artery calcification in asymptomatic subjects: results from the Multi-Ethnic Study of Atherosclerosis (MESA)
.
Circulation
2007
;
115
:
2722
2730
6.
Wong
ND
,
Nelson
JC
,
Granston
T
, et al
.
Metabolic syndrome, diabetes, and incidence and progression of coronary calcium: the Multiethnic Study of Atherosclerosis study
.
JACC Cardiovasc Imaging
2012
;
5
:
358
366
7.
McClelland
RL
,
Jorgensen
NW
,
Budoff
M
, et al
.
10-year coronary heart disease risk prediction using coronary artery calcium and traditional risk factors: derivation in the MESA (Multi-Ethnic Study of Atherosclerosis) with validation in the HNR (Heinz Nixdorf Recall) Study and the DHS (Dallas Heart Study)
.
J Am Coll Cardiol
2015
;
66
:
1643
1653
8.
Goff
DC
 Jr
,
Lloyd-Jones
DM
,
Bennett
G
, et al.;
American College of Cardiology/American Heart Association Task Force on Practice Guidelines
.
2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines
.
Circulation
2014
;
129
(
Suppl. 2
):
S49
S73
9.
Lehmann
N
,
Erbel
R
,
Mahabadi
AA
, et al.;
Heinz Nixdorf Recall Study Investigators
.
Value of progression of coronary artery calcification for risk prediction of coronary and cardiovascular events: result of the HNR Study (Heinz Nixdorf Recall)
.
Circulation
2018
;
137
:
665
679
10.
Grambsch
PM
.
Goodness-of-fit and diagnostics for proportional hazards regression models
.
Cancer Treat Res
1995
;
75
:
95
112
11.
Dalton
JE
,
Perzynski
AT
,
Zidar
DA
, et al
.
Accuracy of cardiovascular risk prediction varies by neighborhood socioeconomic position: a retrospective cohort study
.
Ann Intern Med
2017
;
167
:
456
464
12.
Steyerberg
EW
,
Harrell
FE
 Jr
.
Prediction models need appropriate internal, internal-external, and external validation
.
J Clin Epidemiol
2016
;
69
:
245
247
13.
Echeverría
S
,
Diez-Roux
AV
,
Shea
S
,
Borrell
LN
,
Jackson
S
.
Associations of neighborhood problems and neighborhood social cohesion with mental health and health behaviors: the Multi-Ethnic Study of Atherosclerosis
.
Health Place
2008
;
14
:
853
865
14.
Kaiser
P
,
Diez Roux
AV
,
Mujahid
M
, et al
.
Neighborhood environments and incident hypertension in the Multi-Ethnic Study of Atherosclerosis
.
Am J Epidemiol
2016
;
183
:
988
997
15.
Hussein
M
,
Diez Roux
AV
,
Mujahid
MS
, et al
.
Unequal exposure or unequal vulnerability? Contributions of neighborhood conditions and cardiovascular risk factors to socioeconomic inequality in incident cardiovascular disease in the Multi-Ethnic Study of Atherosclerosis
.
Am J Epidemiol
2018
;
187
:
1424
1437
16.
Steyerberg
EW
,
Vergouwe
Y
.
Towards better clinical prediction models: seven steps for development and an ABCD for validation
.
Eur Heart J
2014
;
35
:
1925
1931
17.
Royston
P
,
Altman
DG
.
External validation of a Cox prognostic model: principles and methods
.
BMC Med Res Methodol
2013
;
13
:
33
18.
Bakhshi
H
,
Ambale-Venkatesh
B
,
Yang
X
, et al
.
Progression of coronary artery calcium and incident heart failure: the Multi-Ethnic Study of Atherosclerosis
.
J Am Heart Assoc
2017
;
6
:
e005253
19.
Leening
MJ
,
Elias-Smale
SE
,
Kavousi
M
, et al
.
Coronary calcification and the risk of heart failure in the elderly: the Rotterdam Study
.
JACC Cardiovasc Imaging
2012
;
5
:
874
880
20.
Katakami
N
.
Mechanism of development of atherosclerosis and cardiovascular disease in diabetes mellitus
.
J Atheroscler Thromb
2018
;
25
:
27
39
21.
Budoff
MJ
,
Young
R
,
Lopez
VA
, et al
.
Progression of coronary calcium and incident coronary heart disease events: MESA (Multi-Ethnic Study of Atherosclerosis)
.
J Am Coll Cardiol
2013
;
61
:
1231
1239
22.
Cefalu
WT
,
Kaul
S
,
Gerstein
HC
, et al
.
Cardiovascular outcomes trials in type 2 diabetes: where do we go from here? Reflections from a Diabetes Care Editors’ Expert Forum
.
Diabetes Care
2018
;
41
:
14
31
23.
Das
SR
,
Everett
BM
,
Birtcher
KK
, et al
.
2018 ACC expert consensus decision pathway on novel therapies for cardiovascular risk reduction in patients with type 2 diabetes and atherosclerotic cardiovascular disease: a report of the American College of Cardiology Task Force on Expert Consensus Decision Pathways
.
J Am Coll Cardiol
2018
;
72
:
3200
3223
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/content/license.