Impact of statins on progression of coronary artery calcium composition and density as assessed by noncontrast CT

Abstract

Purpose:

The effect of statins on progression of coronary artery calcium (CAC) density as measured by noncontrast computed tomography (CT) remains unknown. We examined the association of statin therapy with temporal changes in compositional calcium density using quantitative assessment of CAC scoring CT.

Methods:

This was a retrospective, single-center study of asymptomatic individuals undergoing serial CAC scoring CT. Scans were analyzed using a fully automated deep learning model, with quantification of total CAC volume and volumes of calcium compositional subtypes stratified by Hounsfield unit (HU) density: 130–199; 200–299; 300–399; ≥400 HU.

Results:

Of 316 patients (58.4±10.1 years; 49.1% male) with CAC present at baseline and rescanned at a mean interval of 3.8±1.7 years, 175 (55.4%) patients were statin-treated and 141 (44.6%) patients were statin-naive. In patients exhibiting only low-density calcium (130–199 HU) at baseline, statin therapy was associated with a temporal decrease in CAC volume (β −0.05[−0.09 to −0.02]; p<0.05). Among patients with ≥2 calcium compositional subtypes at baseline, statin therapy was associated with a greater temporal increase in volumes of each density stratum (130–199 HU: β 0.05[0.00–0.11]; 200–299 HU: β 0.11[0.07–0.14]; 300–399 HU: β 0.11[0.07–0.16]; ≥400 HU: β 0.11[0.06–0.16]; all p<0.05). Similar results were observed for changes in relative proportions of each density stratum (all p<0.001).

Conclusion:

In asymptomatic individuals undergoing serial CAC scoring CT, statin therapy was associated with a shift toward denser calcium, considered a more stable phenotype. Assessment of CAC density may capture more unique aspects of plaque progression and stability beyond traditional CAC scoring.

Graphical Abstract.

Changes in compositional CAC density with statin therapy

In asymptomatic individuals undergoing serial coronary artery calcium (CAC) scoring computed tomography (CT), a fully automated deep learning model was used to quantify CAC composition according to Hounsfield unit (HU) density strata: 130–199; 200–299; 300–399; ≥400 HU. Over a median of 3.8 years, statin therapy was associated with a temporal decrease in the proportion of low-density-only calcium, and an increase in the proportions of each density stratum within mixed-density calcium; i.e. a shift toward denser calcium.

INTRODUCTION

Coronary artery calcium (CAC) on noncontrast computed tomography (CT), as quantified by the Agatston score, is a well-established marker of calcified subclinical atherosclerotic burden and strong predictor of cardiovascular events [1]. Progression of the Agatston CAC score on serial CT provides independent and incremental prognostic value beyond the baseline score [2, 3]. Major global guidelines support repeat CAC scanning for patients in whom Agatston CAC score progression would inform the intensification of preventative management [4]. Clinically, the Agatston CAC score is expressed as the product of total calcified plaque area in cubic millimeters and a quantized peak calcium density weighting factor ranging from 1 to 4 [5]. CAC density reflects the concentration of calcium within an atherosclerotic plaque and is measured as the plaque attenuation on CT in Hounsfield units (HU). However, while the Agatston CAC score is upweighted for higher peak calcium density, recent evidence demonstrates that CAC density is inversely associated with lesion vulnerability and risk of major adverse cardiovascular events (MACE) [6, 7]. Thus, CAC density derived from routine CAC scoring CT has emerged as an important area of research, with rapid and fully automated deep learning-based methods for volumetric quantification now available [8–12]. Statin therapy lowers MACE risk, yet paradoxically accelerates progression of the Agatston CAC score [13, 14]. Studies applying serial coronary CT angiography (CCTA) in patients with stable coronary artery disease (CAD) have shown statin use to associate with calcified plaque volume progression and transformation toward a denser phenotype [15, 16]. This may be the mechanism by which these agents promote plaque stability. However, the long-term effect of statins on CAC density progression as measured by noncontrast CT has not been studied. We sought to examine the association of statin therapy with temporal changes in compositional calcium density on serial CAC scoring CT in asymptomatic individuals.

METHODS

Study population

This was a retrospective, single-center study of consecutive asymptomatic outpatients free of known CAD (prior myocardial infarction, coronary revascularization, or any documented atherosclerosis on anatomical imaging or ischemia on functional testing) at baseline who underwent serial noncontrast CAC scoring CT ≥12 months apart at Monash Heart (Monash Health, Victoria, Australia) between January 2011 and January 2023. These patients were identified from a large institutional registry of 5,564 patients who underwent CAC scoring CT alone, without CCTA. Patients were referred by their primary care physician or cardiologist for assessment of baseline cardiovascular risk and subsequent change in risk over time; symptom status was ascertained from referral forms. We excluded patients with: i) Agatston CAC score of 0 reported on their baseline scan; ii) incomplete clinical data or missing information on statin therapy; iii) initiation of statin therapy >3 months after the baseline scan; iv) discontinuation of statin therapy prior to the follow-up scan; and v) inadequate CT image quality (Figure 1). Patients were classified as statin-treated if they were taking a statin prior to the baseline scan or were commenced on a statin ≤3 months after the baseline scan, and statin-naive if they were not taking a statin prior to both scans. For patients with 3 or more CAC scans, the first and last scans were analyzed. This study received institutional ethics approval.

Figure 1. CONSORT Diagram.

Among 746 consecutive asymptomatic outpatients who underwent serial noncontrast coronary artery calcium (CAC) scoring computed tomography (CT), we excluded those with an Agatston CAC score of 0, those with incomplete clinical or statin data, those who commenced statin therapy >3 months after the baseline scan or discontinued statin therapy prior to the follow-up scan; and those with inadequate CT image quality. Finally, 316 patients were eligible for the current analysis.

Definition of risk factors

Baseline cardiovascular risk factors were ascertained by patient questionnaires completed at the time of CAC scanning. Statin use prior to baseline and follow-up scans, and statin initiation ≤3 months after the baseline scan, were determined using patient questionnaires and review of electronic medical records, cross-referenced with Australian Pharmaceutical Benefits Scheme data which includes pharmacy dispensing records. High-intensity statin therapy was defined as atorvastatin 40 mg or 80 mg or rosuvastatin 20 mg or 40 mg [17]. Fasting lipid profiles ≤3 months prior to baseline and follow-up scans were obtained from online pathology databases. Diabetes mellitus was defined by a haemoglobin A1c ≥6.5% and/or use of diabetic medication. Hypertension was defined by a documented diagnosis or treatment with anti-hypertensive medication. Family history of CAD was defined as ≥1 first-degree family member with CAD before age 60 years. Smokers were defined as both active or former smokers.

Image acquisition

Noncontrast CAC scoring cardiac CT scans were performed on a 320-detector-row CT scanner (Aquilion ONE ViSION, Canon Medical Systems, Japan). During a single breath hold, 40–50 electrocardiogram (ECG)-gated slices were acquired from the carina to below the apex of the heart; tube voltage was 120 kVp and reconstruction slice thickness was 3.0 mm.

CAC quantification

All CAC scoring, volume, and density measurements from the CT images were quantified using a validated, fully automated deep learning model [8, 9]. This cascaded convLSTM (convolutional long-short term memory) system was previously trained on 3,000 ECG-gated CT scans and tested on 2,094 ECG-gated CT scans and 5,969 non-gated attenuation correction CT scans, demonstrating excellent agreement with expert manual measurements for CAC scoring in all vessels [9]. In the present analysis, calcified plaque was defined as ≥3 contiguous voxels with a density of ≥130 Hounsfield units (HU). All CAC metrics were automatically computed for the 3 major epicardial coronary vessels: left anterior descending artery, left circumflex artery, and right coronary artery. The left main coronary artery and ramus intermedius artery, when present, was considered part of the left anterior descending artery. Results were then summarized on a per-patient level. CAC scoring was performed by the deep learning model according to the Agatston method [5]. Absolute CAC volume (mm³) was quantified by summing all voxels contributing to calcified plaque, with calculation of the total CAC volume and volumes of calcium compositional subtypes based on predefined attenuation density strata: 130–199 HU (low); 200–299 HU (intermediate); 300–399 HU (high); and ≥400 (very high). We applied the same HU thresholds used for density weighting in the Agatston score formula [5]. The relative proportions (ranging from 0–1) of each calcium compositional subtype were also calculated. CAC at baseline was classified as “low-density-only calcium” (purely containing the 130–199 HU density strata) or “mixed-density calcium” (containing ≥2 density strata). Additionally, the deep learning model automatically measured the mean and peak calcium density in continuous HU, as well as the number of calcified lesions.

Outcomes

Analysis was performed on the per-patient level. We examined the association of statin therapy with temporal change in compositional CAC volumes and proportions by density strata. We also assessed the change in mean and peak calcium HU density with statin therapy.

Statistical analysis

Continuous variables are presented as mean ± standard deviation, regardless of distribution, for uniformity of presentation. Categorical variables are expressed as frequencies and percentages. Comparisons of continuous variables were performed using linear regression, with a log(x+1) transformation for variables with substantial positive skewness. Comparisons of binary variables were performed using logistic regression. Temporal change in compositional CAC volumes was modeled using random-slopes linear mixed-effects regression, with a log(x+1) transformation and a separate model for each density stratum. Temporal change in compositional CAC proportions was modeled using fractional multinomial logistic regression, with low-density-only calcium used as the reference category. This method is well suited for compositional outcomes, where several proportions across categories must always sum to 1. Modeling all strata together in a single fractional multinomial logistic model ensures that increases in one stratum are balanced by decreases in the others. Temporal change in mean and peak calcium HU density was modeled using random-slopes linear mixed-effects regression [18]. All models were adjusted for statin use, time in years, an interaction term between statin use and years, baseline patient characteristics (age, sex, hypertension, diabetes, smoking history), baseline low-density lipoprotein cholesterol level, and baseline total CAC volume. For visual interpretation, the predictive margins from the statistical models were plotted over time. In the statin-treated group, we examined CAC volume change in patients taking a statin prior to the baseline scan versus those commenced on a statin following the baseline scan; interaction terms between statin treatment status (statin-continued or statin-commenced) and years were added to the base models and their beta-coefficients compared. To account for the potential effect of new calcified lesions on the results, we performed a sensitivity analysis including only patients with identical numbers of lesions at baseline and follow-up. Finally, given that an Agatston CAC score of 100 Agatston units (AU) is a widely used clinical threshold for risk stratification and treatment [4], we performed a subgroup analysis examining compositional CAC volume change in patients with Agatston CAC score 1–99 and ≥100 AU; interaction terms between statin therapy, Agatston CAC score ≥100 AU, and years were added to the base models. All statistical analysis was performed using Stata version 16.1 (StataCorp, College Station, TX, USA) and R version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). A p-value <0.05 indicated statistical significance.

RESULTS

Study population and baseline characteristics

The final study population consisted of 316 patients (58.9 ± 10.1 years of age; 49.1% men), of whom 175 (55.4%) were statin-treated and 141 (44.6%) were statin-naive (Figure 1). The mean interscan interval was 3.8 ± 1.7 years (median 104.6 months; interquartile range [IQR] 67.3 to 146.1 months). There was no significant difference in the interscan interval between statin-treated and statin-naive patients (median 104.6 [IQR 67.6 to 148.3] vs. 109.0 [64.9 to 152.1] months. p=0.11). Among statin-treated patients, 98 (56.0%) were taking a statin prior to their baseline scan and 77 (44.0%) were commenced on a statin ≤3 months after the baseline scan. A total of 68 (38.9%) patients were on a high-intensity statin. At baseline, statin-treated patients were older (60.5 ± 10.1 vs. 56.9 ± 9.7 years) and had a higher prevalence of hypertension (56.0% vs. 29.8%), diabetes (18.3% vs 9.2%), and family history of CAD (69.1% vs. 51.8%) than statin-naive patients (p<0.001 for all; Table 1). More statin-treated patients were also taking aspirin (22.9% vs 14.2%), angiotensin-converting enzyme inhibitors or angiotensin receptor blockers (40.0% vs. 24.8%), and beta-blockers (19.4% vs 9.9%) at baseline compared with statin-naive patients. In statin-treated patients, total cholesterol (201.1 ± 50.3 vs. 220.4 ± 46.4 mg/dL, p=0.001) and low-density lipoprotein cholesterol (116.0 ± 46.4 vs. 139.2 ± 42.5 mg/dL, p<0.001) were lower at baseline and exhibited an interval reduction between baseline and follow-up (total cholesterol: −9.3 ± 20.9 vs. 0.39 ± 14.7 mg/dL; low-density lipoprotein cholesterol −8.1 ± 18.6 vs −0.39 ± 12.4 mg/dL; both p<0.05) compared with statin-naive patients.

Table 1.

Baseline clinical characteristics and lipid profile

	All Patients (n=316)	Statin-Treated Patients (n = 175)	Stain-Naive Patients (n = 141)	P-value
Age, years	58.9 ± 10.1	60.5 ± 10.1	56.9 ± 9.7	0.001
Male	155 (49.1)	85 (48.6)	70 (49.6)	0.85
Body mass index, kg/m2	27.7 ± 4.3	27.9 ± 4.3	27.3 ± 4.3	0.25
Systolic blood pressure, mmHg	125.2 ± 18.5	125.7 ± 19.2	124.5 ± 17.5	0.64
Diastolic blood pressure, mmHg	70.7 ± 11.2	71.1 ± 11.3	70.1 ± 11.0	0.51
Hypertension	140 (44.3)	98 (56.0)	42 (29.8)	<0.001
Diabetes mellitus	45 (14.2)	32 (18.3)	13 (9.2)	0.02
Family history of CAD	194 (61.4)	121 (69.1)	73 (51.8)	0.002
Smoking history	88 (27.8)	47 (26.9)	41 (29.1)	0.66
Aspirin	60 (19.0)	40 (22.9)	20 (14.2)	0.05
ACE inhibitor or ARB	105 (33.2)	70 (40.0)	35 (24.8)	0.01
Beta-blocker	48 (15.2)	34 (19.4)	14 (9.9)	0.02
Baseline lipid profile
Total cholesterol, mg/dL	208.8 ± 50.3	201.1 ± 50.3	220.4 ± 46.4	0.001
Low-density lipoprotein, mg/dL	127.6 ± 46.4	116.0 ± 46.4	139.2 ± 42.5	<0.001
High-density lipoprotein, mg/dL	58.0 ± 15.5	59.0 ± 14.2	58.3 ± 19.1	0.79
Triglycerides, mg/dL	132.9 ± 70.9	124.0 ± 71.2	132.9 ± 79.7	0.97
Change in lipid profile between baseline and follow-up CAC scans
Total cholesterol, mg/dL	−5.4 ± 19.0	−9.3 ± 20.9	0.39 ± 14.7	<0.001
Low-density lipoprotein, mg/dL	−5.0 ± 16.6	−8.1 ± 18.6	−0.39 ± 12.4	0.001
High-density lipoprotein, mg/dL	−0.39 ± 4.3	−0.38 ± 4.6	−0.40 ± 4.5	0.99
Triglycerides, mg/dL	−0.89 ± 23.0	−1.8 ± 21.3	2.7 ± 25.7	0.02

ACE = angiotensin converting enzyme; ARB = angiotensin receptor blocker; CAC = coronary artery calcium; CAD = coronary artery disease; CT = computed tomography.

Baseline CAC measurements

Mean baseline CAC measurements for all patients are presented in Table 2. Statin-treated patients had a greater number of calcified lesions (3.7 ± 2.7 vs. 2.1 ± 1.1, p<0.001) and higher Agatston CAC scores (174 ± 327 vs. 58 ± 86 AU; both p<0.001) compared with statin-naive patients. Per-patient total CAC volume was higher in statin-treated versus statin-naive patients (154.2 ± 216.2 vs. 50.4 ± 62.9 mm³, p<0.001), as were volumes of each density stratum (p<0.05 for all; Table 2). In addition, mean and peak calcium HU density (218.4 ± 61.6 vs. 207.7 ± 70.8 HU and 483.3 ± 298.5 vs. 383.4 ± 260.4 HU, respectively; both p<0.001) were higher among statin-treated versus statin-naive patients.

Table 2.

Baseline CAC measurements and annualized change according to statin use

	All Patients	Statin-Treated Patients	Stain-Naive Patients	P-value
	(n = 316)	(n = 175)	(n = 141)
Baseline CAC measurements
Agatston score, AU	115.6 ± 260.2	174.0 ± 327.0	58.3 ± 86.2	< 0.001
Total volume, mm3 Volumes by density strata, mm3	107.9 ± 173.9	154.2 ± 216.2	50.4 ± 62.9	< 0.001
130–199 HU	48.1 ± 65.7	68.3 ± 80.5	23.1 ± 22.6	< 0.001
200–299 HU	29.7 ± 49.5	42.5 ± 61.5	13.8 ± 18.8	< 0.001
300–399 HU	13.7 ± 27.8	20.3 ± 35.1	5.6 ± 9.1	< 0.001
≥400 HU	16.3 ± 43.6	23.1 ± 55.4	8.0 ± 18.3	0.002
Mean density, HU	213.6 ± 66.0	218.4 ± 61.6	207.7 ± 70.8	0.15
Peak density, HU	438.7 ± 286.1	483.3 ± 298.5	383.4 ± 260.4	0.002
Number of lesions, n	7.9 ± 2.0	3.7 ± 2.7	2.1 ± 1.1	< 0.001
Annualized change in CAC measurements
Agatston score, AU	33.3 ± 56.2	52.0 ± 68.9	10.2 ± 15.2	< 0.001
Total volume, mm3/year	24.3 ± 40.0	37.2 ± 49.2	8.4 ± 10.9	< 0.001
Volumes by density strata, mm3/year
130–199 HU	8.5 ± 13.8	12.6 ± 16.9	3.4 ± 5.2	< 0.001
200–299 HU	6.9 ± 11.8	10.6 ± 14.2	2.4 ± 4.8	< 0.001
300–399 HU	4.4 ± 8.5	7.0 ± 10.6	1.2 ± 2.1	< 0.001
≥400 HU	4.4 ± 10.5	7.0 ± 13.4	1.1 ± 2.5	< 0.001
Mean density, HU/year	5.9 ± 10.5	8.1 ± 12.1	3.1 ± 7.4	< 0.001
Peak density, HU/year	22.7 ± 35.1	30.7 ± 38.7	12.8 ± 27.1	< 0.001
Number of lesions, n/year	0.2 ± 0.4	0.3 ± 0.4	0.1 ± 0.3	< 0.001

AU = Agatston units; CAC = coronary artery calcium; HU = Hounsfield units.

Association of statin therapy with compositional CAC changes by density strata

Statin-treated patients exhibited a higher annualized increase in total CAC volume and volumes of each density stratum compared with statin-naive patients (p<0.05 for all; Table 2). Table 3 shows the adjusted interaction terms between statin therapy and years for the change in volume of each density stratum. In patients with low-density-only calcium (130–199 HU) at baseline, statin therapy was associated with a decrease in CAC volume over time (β −0.05 [−0.09 to −0.02], p=0.006) compared with no statin therapy. Among patients with mixed-density calcium (≥2 compositional subtypes) at baseline, statin therapy was associated with a more rapid increase in the volumes of each density stratum over time (130–199 HU: β 0.05 [0.00 to 0.11]; 200–299 HU: β 0.11 [0.07 to 0.14]; 300–399 HU: β 0.11 [0.07 to 0.16]; ≥400 HU: β 0.11 [0.06 to 0.16]; p<0.05 for all) compared with no statin therapy. Temporal changes in compositional CAC volumes with and without statin treatment are shown in Online Resource 1. There was no significant difference in the rate of CAC volume change between patients who were taking a statin prior to the baseline scan and those who were commenced on a statin following the baseline scan (difference in beta-coefficients ranging from 0 to 0.06; all p≥0.05; Online Resource 1 and Online Resource 6).

Table 3.

Association of statin use with change in compositional CAC volumes by density strata

	Low-density-only calcium		Mixed-density calcium (>2 compositional subtypes)
	Low: 130–199 HU		Low: 130–199 HU		Intermediate: 200–299 HU		High: 300–399 HU		Very high: >400 HU
Model*	Adjusted β coefficient (95% CI)†	P-value	Adjusted β coefficient (95% CI)†	P-value	Adjusted β coefficient (95% CI)†	P-value	Adjusted β coefficient (95% CI)†	P-value	Adjusted β coefficient (95% CI)†	P-value
Statin use	−0.24 (−0.43 to −0.06)	0.01	0.61 (0.39 to 0.83)	<0.001	0.51 (0.26 to 0.75)	<0.001	0.35 (0.10 to 0.60)	0.01	0.41 (0.06 to 0.76)	0.02
Years	0.01 (−0.02 to 0.04)	0.45	0.06 (0.03 to 0.10)	0.001	0.06 (0.03 to 0.09)	<0.001	0.04 (0.01 to 0.08)	0.01	0.04 (−0.00 to 0.07)	0.06
Statin x years	−0.05 (−0.09 to −0.02)	0.01	0.05 (0.00 to 0.11)	0.02	0.11 (0.07 to 0.14)	<0.001	0.11 (0.07 to 0.16)	<0.001	0.11 (0.06 to 0.16)	<0.001

^*β coefficients for each density strata derived from separate linear mixed models.

^†Adjusted for baseline age, sex, hypertension, diabetes, smoking, low-density lipoprotein cholesterol, and total calcified plaque volume.

CI = confidence interval; HU = Hounsfield units.

The proportions of each density stratum within mixed-density calcium increased over time with statin therapy (130–199 HU: β 1.09 [0.48 to 1.70]; 200–299 HU: β 1.14 [0.53 to 1.76]; 300–399 HU: β 1.18 [0.56 to 1.79]; ≥400 HU: β 1.11 [0.49 to 1.73]; p<0.001 for all; Table 4); while the proportion of low-density-only calcium decreased (Figure 2). Without statin therapy, there was no significant change in the proportions of any density stratum within mixed-density calcium, nor in the proportion of low-density-only calcium (Figure 2).

Table 4.

Association of statin use with change in compositional CAC proportions by density strata

	Mixed-density calcium (≥2 compositional subtypes)
	Low: 130–199 HU		Intermediate: 200–299 HU		High: 300–399 HU		Very high: ≥400 HU
	Adjusted β coefficient (95% CI)*	P-value	Adjusted β coefficient (95% CI)*	P-value	Adjusted β coefficient (95% CI)*	P-value	Adjusted β coefficient (95% CI)*	P-value
Statin use	0.56 (−0.24 to 1.37)	0.17	0.29 (−0.52 to 1.11)	0.48	0.31 (−0.54 to 1.15)	0.47	0.19 (−0.70 to 1.08)	0.67
Years	−0.01 (−0.12 to 0.12)	0.98	−0.05 (−0.17 to 0.06)	0.35	−0.07 (−0.19 to 0.05)	0.28	−0.05 (−0.18 to 0.08)	0.43
Statin x years	1.09 (0.48 to 1.70)	<0.001	1.14 (0.53 to 1.76)	<0.001	1.18 (0.56 to 1.79)	<0.001	1.11 (0.49 to 1.73)	<0.001

^*Adjusted for baseline age, sex, hypertension, diabetes, smoking, low-density lipoprotein cholesterol, and total calcified plaque volume. Low-density-only calcium used as reference category.

CI = confidence interval; HU = Hounsfield units.

Figure 2. Change in compositional CAC proportions by density strata according to statin use.

Estimated changes in the proportions of compositional coronary artery calcium (CAC) subtypes by Hounsfield unit (HU) density strata according to statin use are derived from fractional multinomial logistic regression models, with predictive margins from the models plotted over time. Statin therapy was associated with a temporal decrease in the proportion of low-density-only calcium, and an increase in the proportions of each density stratum within mixed-density calcium. Without statin therapy, there was no significant change in the proportions of any density subtype.

In sensitivity analysis restricted to patients (n = 162; 51.3%) with identical numbers of calcified lesions at baseline and follow-up, statin therapy remained associated with an increase in the volumes and proportions of all compositional subtypes within mixed-density calcium, and a decrease in the volume and proportion of low-density-only calcium (Online Resources 2–3 and Online Resources 7–8). In subgroup analysis of patients with Agatston CAC score 1–99 (n = 202; 63.9%) and ≥100 (n = 114; 36.1%), no significant interaction between statin therapy and Agatston CAC score was observed for changes in the volumes of low-density-only calcium or each density stratum within mixed-density calcium (all p>0.05 for interaction; Online Resource 4 and Online Resource 9).

Association of statin therapy with changes in mean and peak calcium HU density

There was a higher annualized increase in the mean and peak continuous HU density of CAC among statin-treated versus statin-naive patients (Table 2). In linear mixed models adjusted for baseline age, sex, cardiovascular risk factors, low-density lipoprotein cholesterol level, and total CAC volume, statin therapy was associated a greater temporal increase in the mean (β 5.93 [4.33 to 7.54]) and peak: β 19.22 [12.04 to 26.40]; both p<0.001) CAC HU density compared with no statin therapy (Figure 3).

Figure 3. Change in mean and peak CAC HU density according to statin use.

Temporal changes in the mean and peak continuous Hounsfield unit (HU) density of coronary artery calcium (CAC) are derived from linear mixed models. Statin use was associated with a greater temporal increase in the mean and peak density compared with no statin use.

DISCUSSION

In this study of asymptomatic individuals undergoing serial CAC scoring CT, statin therapy was associated with progression of compositional CAC density; specifically: i) a decrease in the volume and proportion of low-density-only calcium; ii) an increase in the volumes and proportions of all compositional subtypes within mixed-density calcium; and iii) an increase in the mean and peak continuous HU density. These findings were independent of patient characteristics and total CAC volume at baseline, and robust to sensitivity analyses including only patients with identical numbers of calcified lesions at baseline and follow-up. Finally, the effect of statins on compositional CAC volume progression did not differ significantly between patients with baseline Agatston CAC score above or below 100 AU.

To our knowledge, this is the first study applying noncontrast CAC scoring CT to assess long-term changes in coronary calcification density with statin therapy. A sub-analysis of the PARADIGM (Progression of Atherosclerotic Plaque Determined by Computed Tomography Angiography Imaging) study previously showed statins to induce a transformation toward denser calcium on CCTA, as evidenced by greater progression of high density (>700 HU) calcium and less progression of low-density (351–700) calcium [16]. Extending this investigation to noncontrast CT, Vogel et al performed a pooled analysis of two randomized controlled trials of high- versus low-intensity statin therapy in patients undergoing serial CAC scoring scans [21]. They demonstrated that although the mean calcium HU density in the overall population increased at 1 year, the change in density did not differ between the 2 treatment groups. Notably, the effect of statin therapy compared with placebo on calcium density was not evaluated and the follow-up period was relatively short. Our current study, although non-randomized, provides insights into the magnitude and directionality of calcium density progression in statin-treated versus statin-naive patients on noncontrast CT over a mean of 3.8 years. We applied an automated deep learning model which quantified the volumetric change in calcium compositional subtypes by density strata, capturing the shift toward a greater volume and proportion of higher-density calcium [9]. This represents deeper phenotyping of calcified plaque on routine CAC scoring CT beyond measuring the raw mean or peak HU attenuation.

On the molecular level, statins are thought to accelerate the calcification cascade that accompanies atherosclerotic plaque progression. Coronary artery calcification pathologically begins as microcalcifications, which are commonly observed in the deeper areas of necrotic core as a result of dying smooth muscle cells and macrophages [22, 23]. Microcalcifications then coalesce into larger calcium fragments, which grow into sheet-like deposits of dense calcium. Histopathologic and intracoronary imaging studies suggest that spotty calcification is associated with unstable plaques, and extensive calcification with stable plaques [23, 24]. While microcalcifications cannot be reliably detected by CT scanners due to limited spatial resolution, larger calcifications such as fragments and sheets can be identified with increasing HU attenuation [25]. Our findings support the hypothesis that low-density calcium on noncontrast CT represents calcification early in the evolutionary process toward more stable dense calcific sheets. We observed the volume of low-density-only calcium to regress with statin therapy; this may represent small, predominantly noncalcified lesions in which statins reduced the surrounding necrotic core and promoted growth of calcium fragments into larger masses [16]. Meanwhile, the statin-induced increase in volumes of all density strata among patients with mixed-density calcium may reflect the presence of more non-calcified plaque at baseline; this is converted to low-density calcium, which in turn undergoes further densification along the evolutionary cascade [16]. Although the basic mechanism for enhanced vascular calcification with statins remains unclear, recent evidence from animal models suggests that this may be partly due to increased IL-1β secretion by macrophages in response to inflammasome stimulation, or increased differentiation of vascular smooth muscle cells into osteoblast-like cells [26, 27].

Our study findings may have implications for serial CAC assessment. Major global guidelines advocate for repeat CAC scanning at a 3–5 year interval for patients with CAC >0 in whom progression of the Agatston CAC score would support intensification or modification of preventative management [4]. By contrast, some primary prevention guidelines posit that among statin users, the Agatston CAC score has limited clinical utility and should thus be interpreted with caution [17]. Large cohort studies have shown that despite lowering MACE risk, statins paradoxically increase the Agatston CAC score on serial scanning [13, 14]. The present analysis demonstrates that this phenomenon is partly driven by an increase in CAC density, as density is upweighted in the Agatston formula. While we observed an increase in the volumes of all density strata with statin therapy, there was an overall greater rate of increase in the mean and peak HU density. Cumulative evidence suggests that higher lesion density is associated with atherosclerotic plaque stability on the per-plaque level [28, 29]. Thus, in daily practice, the population of statin users with high Agatston CAC scores likely encompasses high-risk individuals who have an extensive burden of calcified plaque as well as individuals with high-density stable plaques who are at relatively lower risk of MACE. As such, the prognostic value of the Agatston CAC score may be attenuated among statin users [30]. Our findings suggest that progression of calcified plaque density follows a predictable pattern in statin-treated patients, regardless of whether their baseline Agatston CAC score is above or below 100 AU; the widely used clinical threshold for risk stratification and treatment [4]. Herein we decompartmentalize the progression of CAC density and volumes in examining the natural history of atherosclerosis development and its alteration by statin therapy. In future, CAC density (particularly volume of low-density CAC) could potentially be used to refine clinical CAC assessment as it captures more unique aspects of plaque progression and stability compared with the traditional Agatston CAC score. Certainly, there is currently no standard definition of CAC progression, with various methods of quantification based on the Agatston CAC score yielding only moderate concordance [31]. Further, CAC progression determined by these methods does not offer additional prognostic information beyond the follow-up Agatston CAC score [32]. Importantly, recent evidence demonstrates that progression of mean CAC density on serial CAC scans is inversely associated with future MACE, providing incremental predictive value when added to baseline CAC density [7]. Future prospective studies will need to examine whether statin use and statin intensity modulate MACE risk in patients who exhibit CAC density progression.

Limitations

This study has several important limitations. First, assignment of statin therapy was non-randomized, resulting in differences in clinical risk profile and baseline CAC volume between statin-treated and statin-naive patients. These factors were adjusted for in multivariable models, yet we cannot exclude confounding by indication or time-varying confounders. Second, data were not available on vascular comorbidities such as stroke and peripheral arterial disease, and these may also represent potential confounders for the outcome. Third, the study population included asymptomatic individuals who were referred at the discretion of their physician for repeat cardiovascular risk assessment ≥12 months apart, with each CAC scoring scan incurring a fee for the patient. This introduces selection bias, as patients who declined repeat assessment or who underwent rescanning <12 months were ineligible; the generalizability of our findings to these groups is unknown. Fourth, our analyses involved multiple hypothesis testing across several related outcomes (volumes of density strata, proportions of density strata, mean density, and peak density), resulting in family-wise type I error inflation. However, we consistently observed small p-values (many were <0.001) which would have remained statistically significant even with formal multiplicity adjustment. Finally, we examined quantitative calcified plaque changes on noncontrast CAC scoring CT, and thus were unable to assess for corresponding changes in noncalcified plaque. However, prior studies have demonstrated the differential impact of statins on calcified and noncalcified plaque volumes quantified from CCTA in a stable CAD cohort [15, 16].

CONCLUSIONS

In asymptomatic individuals undergoing serial CAC scoring CT, statin therapy was associated with a shift toward denser calcium, considered a more stable phenotype. Assessment of CAC density may capture more unique aspects of plaque progression and stability beyond traditional CAC scoring.

LIST OF ABBREVIATIONS

AU

Agatston units

CAC

coronary artery calcium

CAD

coronary artery disease

CCTA

coronary computed tomography angiography

CT

computed tomography

ECG

electrocardiogram

HU

Hounsfield units

MACE

major adverse cardiovascular events