The Association of Coronary Artery Calcium Score with Heart Failure – a Literature Review

The initial screening was based on article titles and abstracts, from which 32 studies were selected. After removing duplicates and excluding articles that did not meet the main inclusion criteria of the review, 23 studies were included. Of these, 12 investigated the association between CAC and the risk of HF or HF-related mortality. Eight studies examined the relationship between CAC and cardiovascular disease (CVD), cardiovascular disease events (CVDE), or major adverse cardiovascular events (MACE), which included HF. The remaining three studies assessed the prognostic value of the CAC score in patients with HF.

Of the 23 studies included in the review, 19 were cohort studies and 4 were cross-sectional. The mean age of the participants ranged from 40 to 74 years, and sample sizes ranged from 157 to 66,636. The studies were conducted in various regions; most in the United States, involving large cohorts (e.g., MESA and CARDIA), and in Europe (e.g., the Heinz Nixdorf Recall cohort study in Germany and the Rotterdam Study in the Netherlands). Other study locations included Canada, Japan, South Korea, Thailand, and Saudi Arabia (Table 1).

Four of the included studies used the Multi-Ethnic Study of Atherosclerosis (MESA) cohort as their study population. MESA was designed as a prospective, observational epidemiological cohort study aimed at characterizing the development and progression of subclinical cardiovascular disease in a large population without known cardiovascular disease. Between 2000 and 2002, the study enrolled 6,814 participants aged 45-84 years from four different ethnic groups across six regions in the United States. Participants with pacemakers or defibrillators, previous coronary bypass surgery, heart valve replacement, percutaneous revascularization, or other cardiac surgery were excluded.⁴

The included studies used different approaches to evaluate the relationship between the CAC score and HF. Some studies assessed the direct association between CAC and the risk of developing HF, whereas others analyzed outcomes separately for HF with preserved ejection fraction (HFpEF) and HF with reduced ejection fraction (HFrEF). Risk stratification for HF-related mortality was investigated in studies such as those from the Coronary Artery Consortium, while other studies focused on the association between higher CAC scores and the risk of HF development.

All included studies reported an association between CAC and HF, although results varied after adjustment for traditional risk factors. In the Rotterdam Study, a prospective, population-based cohort study initiated in 1990 that included subjects older than 55 years, 1,897 asymptomatic individuals underwent CAC scoring and were followed for a median of 6.8 years. The primary outcome was HF or coronary heart disease (CHD). After adjustment for other cardiovascular risk factors, higher CAC scores were associated with an increased risk of HF, with a hazard ratio (HR) of 4.1 for CAC scores > 400 compared with scores < 10.⁵

Another study by Ahmed Fathala et al. evaluated 204 patients with new-onset HF who underwent CCTA in the preceding 3 months, of whom 35 had obstructive CAD. In multivariable logistic regression analysis, the CAC score was the independent predictor of HF attributable to CAD. Notably, 63% of patients had a CAC score of 0, which correlated with normal CCTA findings and yielded a 100% negative predictive value.⁶

Regarding the different results after adjustment for traditional risk factors, a study including participants from the Heinz Nixdorf Recall Study reported attenuation of the association between CAC score and the presence of congestive heart failure (CHF). The Heinz Nix-dorf Recall Study is a prospective epidemiological cohort study conducted in Germany that enrolled 4,814 participants, aiming to improve cardiovascular risk prediction using imaging and non-imaging variables.⁷ In the study by Kälsch et al., 4,230 participants without known CAD but with a diagnosis of CHF were included. CHF was identified in 105 participants (2.5%), and CAC extent was significantly higher in subjects with CHF (median CAC score 64.7) than in those without CHF (median CAC score 11.6). To assess the association between CAC and CHF, a logarithmically transformed CAC burden was used. In univariate analysis, the CAC score was associated with CHF (odds ratio (OR) 1.16). However, in the fully adjusted multivariate model, which included age, sex, Framingham risk factors, and cardiovascular medication, the association became statistically non-significant (OR 1.07). The authors concluded that the observed association between CAC and CHF is partly explained by risk factors that are common in the pathophysiology of both CAC and CHF.⁸

In contrast, the study with the largest sample size among the included studies, conducted by Mhaimeed et al., demonstrated a strong correlation between CAC score and HF-related mortality. The study included 66,636 participants from the CAC Consortium, a multicenter, retrospective, observational real-world cohort study of patients older than 18 years who underwent CAC scoring.⁹ The primary outcome was HF-related mortality, with a median follow-up of 12.5 years. A total of 260 HF-related deaths were recorded, of which 75.3% occurred in patients with a CAC score ≥ 100, and more than half (53.1%) occurred in participants with a CAC score ≥ 400. Participants were categorized into four groups based on CAC score: 0 (reference), 1–99, 100–399, and ≥ 400. After multivariable adjustment for demographic factors, traditional cardiovascular risk factors, and competing risks of non-cardiovascular death, subdistribution hazard ratios (SHR) for HF-related mortality compared with the reference group were 2.27 for CAC 1–99, 3.68 for CAC 100–399, and 7.05 for CAC ≥ 400. Overall, the study demonstrated that higher CAC scores are strongly associated with an increased risk of HF-related mortality.¹⁰

Although most studies demonstrate an association between CAC score and HF, the degree of independent predictive value after adjustment for traditional risk factors remains inconsistent. This is mainly attributable to differences in study design, patient populations, and definitions of HF. One important methodological factor is the nature of the analysis (cross-sectional or prospective). For example, the Heinz Nixdorf Recall Study assessed the prevalence of self-reported CHF, whereas the Rotterdam Study evaluated the incidence of HF, and the CAC consortium focused on HF-related mortality. Additional variation arises from differences in the age and health status of the studied populations. The mean age in the Heinz Nixdorf Recall Study was 59 years, compared with 69.9 years in the Rotterdam Study and 54.4 years in the CAC Consortium cohort.

Some studies stratified HF outcomes into HFpEF and HFrEF. Three of these used data from the MESA cohort and demonstrated the predictive value of CAC for HF risk assessment using different methodological approaches. De la Rosa et al. used log-transformed CAC together with other CCTA biomarkers, such as left ventricular size index, aortic valve, mitral annulus, and thoracic aorta measurements, to predict new-onset HF among 6,667 patients from the MESA cohort. During a median follow-up of 17.7 years, new-onset HF was diagnosed in 426 patients: 173 (40.6%) with HFrEF, 193 (45.3%) with HFpEF, and 60 (14.1%) without a recorded ejection fraction (EF). Cox regression analysis showed that log-transformed CAC was an independent predictor of both new-onset HFrEF (HR 1.15) and new-onset HFpEF (HR 1.09).¹¹

In contrast, Sharma et al. examined the predictive value of CAC specifically for HFpEF and reported sex-specific differences. A total of 6,809 participants underwent CAC scoring and were followed for a median time of 11.2 years, during which 127 incident HFpEF hospitalizations were recorded. The overall incidence rate of HFpEF was 1.82 per 1,000 person-years, with a slightly higher rate in men than women (2.00 vs. 1.66 events per 1,000 personyears). Across CAC categories, HFpEF incidence increased progressively, from 0.99 events per 1,000 person-years in participants with a CAC score of 0 to 1.48 for CAC 1-100, 2.95 for CAC 101–300, and 5.39 for CAC > 300. In adjusted analyses, a CAC score of >300 was associated with an increased risk of HFpEF (HR 1.68); however, this association was significant only in women (HR 2.82) and not in men (HR 0.91). The authors concluded that, beyond traditional risk factors, CAC may be useful in HFpEF risk stratification in women but not in men, highlighting the need for further research to clarify sex-specific pathophysiological mechanisms in HFpEF.¹²

The third study examining HFpEF and HFrEF included, in addition to the CAC score, the pro-inflammatory cytokine interleukin-6 (IL-6). The rationale for including IL-6 was that chronic inflammation is a key contributor to both clinical and subclinical atherosclerosis and HF.^13,14 Ahmad et al. the association between HF and combined IL-6 and CAC burden in 6,592 participants from the MESA cohort. Participants were categorized into four groups: low CAC/low IL-6, low CAC/high IL-6, high CAC/low IL-6, and high CAC/high IL-6. A total of 422 HF events were reported, with incidence rates increasing across categories: 1.58 per 1,000 person-years in the low IL-6/low CAC group, 3.63 in the low IL-6/high CAC group, 5.04 in the high IL-6/low CAC group, and 9.14 in the high IL-6/high CAC group. In adjusted analyses, high IL-6 alone (HR 1.61) and high CAC score alone (HR 1.56) were each associated with increased HF risk compared with the reference group. For HFrEF, neither high IL-6 nor high CAC alone showed a statistically significant association after multivariable adjustment; however, their combination was associated with a higher risk of HFrEF (HR 2.34). In contrast, for HFpEF, both high IL-6 alone (HR 2.43) and high CAC score alone (HR 2.12) were associated with increased risk, whereas the combination of high IL-6 and high CAC did not confer additional risk.¹⁵

Similar to studies examining the overall association between CAC and HF, some analyses reported conflicting results, largely due to differences in methodology and variable definitions. For example, de la Rosa et al.¹¹ primarily used log-transformed CAC as a continuous variable, whereas Ahmad et al.¹⁵ categorized CAC into two groups, and Sharma et al.¹² used both categorical CAC and log-transformed CAC. Another important difference concerned the left ventricular ejection fraction (LVEF) thresholds used to define HF subtypes. In the study by de la Rosa et al.,¹¹ HFrEF was defined as EF < 45% and HFpEF as EF ≥ 45%, while Ahmad et al.¹⁵ defined HFrEF as EF ≤ 40% and HFpEF as EF ≥ 50%.

Diastolic dysfunction (DD) refers to a range of alterations in cardiac function, including increased myocardial stiffness, abnormal relaxation, and increased end-diastolic pressure, and represents a major contributor to HFpEF.¹⁶ Of the 23 included studies, three investigated the association between CAC score and DD. Although all three demonstrated an independent association between CAC and DD, they differed in study population, methodology, and endpoints. For example, Haddad et al. studied 2,082 participants from the Project Baseline Health Study who underwent CAC scoring and echocardiography. After multivariable adjustment, subclinical coronary atherosclerotic burden assessed by CAC score was independently associated with DD. Two continuous diastolic parameters, e’ velocity and the E/e’ ratio, showed an independent association with increasing CAC score categories.¹⁷

In contrast to the other studies, Nitta et al. assessed diastolic function using gated single-photon emission computed tomography (SPECT). The study tested the hypothesis that CAC extent is associated with left ventricle (LV) diastolic parameters in patients without evidence of myocardial ischemia, with peak filling rate (PFR) as a key parameter. The results showed that patients with CAC had lower PFR than those without CAC. After multivariate linear regression adjusting for sex, hypertension, LV end-diastolic volume (LVEDV), and LV mass index, ln(CAC score + 1) was identified as an independent predictor of LV diastolic parameters. The authors suggested that micro-vascular damage secondary to arterial stiffness and endothelial dysfunction caused by CAC may represent a pathophysiological mechanism underlying DD.¹⁸

While the studies by Haddad et al. and Nitta et al. were cross-sectional, the third study, by Choi et al., performed a retrospective longitudinal analysis examining the relationship between CAC score and the progression of DD in a cohort of 15,193 adults in Korea. The CAC scores were categorized into three groups: CAC 0, CAC 1–99, and CAC ≥ 100. DD was defined using echocardiographic parameters, including septal e’, septal E/e’, left atrial volume index (LAVI), and tricuspid regurgitation velocity. Definite DD was defined as ≥ 3 of 4 abnormal parameters, whereas definite or probable DD was defined as ≥ 2 of 4 abnormal DD parameters. The adjusted prevalence ratio for definite DD in patients with CAC ≥ 100 was 1.72 compared with those with CAC = 0, while the prevalence ratio for definite or probable DD was 1.83. Regarding mortality and the combined effect of CAC and DD, patients with CAC ≥ 100 and definite DD had the highest mortality risk with an HR for all-cause mortality of 3.91 compared with to the reference group (CAC < 100 without definite DD). In addition, CAC ≥ 100 was independently associated with progression to definite DD over time, with an adjusted HR of 1.95 compared with CAC = 0. Overall, the study demonstrated a strong association between CAC score and the prevalence of DD, an increased mortality risk when both CAC and DD were present, and an independent effect of significant CAC burden (≥ 100) on DD progression over time.¹⁹

Most studies included in this review focused specifically on the association between CAC and HF; however, some also evaluated HF as part of broader cardiovascular outcomes such as stroke, aortic disease, unstable angina, myocardial infarction, and coronary artery revascularization procedures. All studies demonstrated a strong predictive value of CAC for CVDE or MACE, although results varied owing to differences in study populations and methodologies. For example, Wada et al. focused on a Japanese population and identified sex-specific CAC thresholds for MACE prediction,²⁰ whereas a study based on the CARDIA cohort examined young adults and found that any CAC score above zero was associated with an increased risk for CVD.²¹ In contrast, Wakaki et al. demonstrated that high CAC scores predict CVDE using non-ECG-gated CT imaging.²²

One method of studying the predictive value of the CAC score has been to assess whether adding it to traditional risk scores improves risk stratification. In this regard, Wada et al. conducted a multicenter, prospective cohort study in Japan that enrolled 982 patients to evaluate the contribution of CAC to the prediction of MACE. CAC scores were categorized as no CAC (0), mild (0 < score < 100), moderate (100 ≤ score < 300), and severe (above 300). MACE included hospitalization for unstable angina, HF, aortic disease, cardiovascular death, myocardial infarction, revascularization, and stroke. A key finding of the study was the presence of sex differences in CAC burden: CAC was detected in 77.5% of male participants compared with 55.3% of female participants, and severe CAC was detected in 25.5% of men but only 7.5% of women. The primary objective was to determine whether adding CAC to the Suita risk score improved the C-statistic for predicting MACE.²⁰ The Suita score, developed for the Japanese population, estimates cardiovascular risk based on age, sex, smoking, diabetes, blood pressure, low-density lipoprotein, high-density lipoprotein, and chronic kidney disease.²³ In this cohort, the C-statistic for the Suita score alone was 0.650. Adding CAC improved the C-statistic when thresholds of ≥ 100, ≥ 200, ≥ 300, or ≥ 400 were applied, with the greatest improvement observed at the ≥ 300 threshold, yielding a continuous net reclassification index (NRI) of 0.652. In men, the C-statistic improved with CAC thresholds of ≥100 and ≥200, with an NRI of 0.630 at the ≥ 200 threshold. In women, improvement was observed only when a CAC threshold of ≥400 was added, with an NRI of 0.698. Overall, the study demonstrated that the CAC score has a significant predictive value for MACE, with sex-specific differences in its impact on risk stratification.²⁰

One demographic that is often underrepresented in cardiovascular risk assessment studies is young adults. Carr et al. addressed this gap by evaluating the association between CAC, CHD, CVD (including HF), and all-cause mortality in adults aged 32–46 years. A total of 3,043 participants from the CARDIA study were included and followed for 12.5 years. The presence of CAC was associated with a higher risk of CVD events (HR 3.0) and an increased rate of CHD events (HR 5.0). The study also demonstrated that even low CAC scores (1–19) significantly increased the risk of CHD in young adults, suggesting that early screening should be considered.²¹

In a retrospective cohort study, Wakaki et al. examined the relationship between CAC score and future CVD and non-CVD events in patients with established CVD. Participants were categorized into three groups based on CAC score: low (median 0), intermediate (median 200), and high (median 1,601). The primary endpoint was CVD events, including cardiac death, nonfatal myocardial infarction, hospitalization for CHF, stroke, and unplanned cardiac surgery. Kaplan–Meier analysis confirmed that patients with a high CAC score had a higher risk of CVD events than those with a low score. In multivariable Cox analysis, CAC score, smoking, and dialysis were independently associated with CVD events.²²

An additional application of the CAC score has been explored in studies evaluating its prognostic value in patients already diagnosed with HF. Both Ota et al. and Hashimoto et al. reported that higher CAC scores are associated with worse clinical outcomes.²³,²⁵ Ota et al. investigated the prognostic value of CAC for all-cause mortality in patients with CHF. A total of 550 patients were enrolled and stratified into three groups based on CAC score: 0, 1—999, and ≥ 1,000. After adjustment for traditional risk factors, brain natriuretic peptide (BNP) and LVEF, patients with CAC ≥ 1,000 had a significantly higher risk of all-cause mortality (HR 1.564) compared with those with CAC 0.²⁵

Hashimoto et al. evaluated the role of CAC in predicting MACE in patients with non-ischemic HFpEF (NIHFpEF), comparing its prognostic value with that of the I-beta-methyl-piodophenyl-pentadecanoic acid (I-BMIPP) defect score (BDS). The study included 108 patients with NIFH-pEF who underwent non-contrast chest CT and I-BMIPP SPECT. Multivariable Cox regression analysis demonstrated that both CAC score >0 and high BDS (≥4) were independent predictors of MACE. When combined, CAC score and BDS provided improved prognostic stratification, with a 75% MACE rate in patients with CAC >0 and high BDS compared with 3% in those with CAC = 0 and low BDS.²⁴