Consideration of Statin Therapy in the Interpretation of Artificial Intelligence-Derived Coronary Artery Calcium Density

I read with great interest the study by Kim et al. [1] demonstrating that coronary artery calcium (CAC) density derived from artificial intelligence (AI)-based analysis of non-electrocardiogram-gated chest CT was independently and inversely associated with major adverse cardiovascular events (MACE) in a health-screening cohort. The authors should be commended for advancing the opportunistic use of routine chest CT for cardiovascular risk stratification and for elegantly dissecting the joint prognostic value of CAC volume and density. I write to offer a cautionary note regarding the clinical interpretation of their central finding—namely, that higher CAC density predicts lower MACE risk (adjusted hazard ratio [HR] per standard deviation, 0.786; 95% confidence interval, 0.659–0.936)—in view of the potential influence of unmeasured lipid-lowering therapy.

The authors candidly acknowledge that “lipid-lowering therapy … could not be reliably assessed,” identifying it as a potential confounder [1]. I respectfully suggest that the direction of this confounding merits particular attention, as it may have significant implications for the clinical interpretation of AI-derived CAC density. Statins remain the cornerstone of atherosclerotic cardiovascular disease prevention, and large-scale meta-analyses have robustly shown that they substantially reduce MACE in at-risk individuals [2]. Simultaneously, mounting imaging evidence indicates that statins promote coronary plaque calcification and densification. In a serial coronary CT angiography study, van Rosendael et al. [3] demonstrated that statin therapy was associated with larger increases in high-density calcium and “1K plaque” volumes alongside reductions in low-attenuation plaque. Consequently, patients receiving statins may simultaneously exhibit higher CAC density and lower MACE risk—precisely the inverse association reported by Kim et al. [1]. This raises the possibility that the observed protective effect of CAC density may be partly influenced by concomitant statin use in a subset of the cohort.

In observational studies, this pattern represents a classic confounding structure: statin use is a common cause of both the exposure (increased CAC density) and the outcome (reduced MACE). Given the absence of data on statin therapy, it is difficult to exclude the possibility that the observed inverse association is partly driven by this pharmacologic confounding pathway. Thus, the adjusted HR of 0.786 may—at least in part—reflect the effects of statins on plaque composition and cardiovascular risk, in addition to—or even independent of—the intrinsic biologic stability of dense plaques. Prior work from the CAC Consortium has examined CAC prognostication among statin users versus non-users [4], and meta-analytic evidence confirms that the inverse association between density and events is most evident after adjustment for overall plaque burden [5,6]. These findings collectively underscore that the prognostic interpretation of CAC density may need to account for statin exposure status.

I emphasize that my intention is not to challenge the biological plausibility that denser calcification reflects more stable plaque architecture—a concept well supported by the literature [5,6]. Rather, I wish to highlight that in the absence of statin data, the extent to which the observed inverse association is attributable to plaque biology alone, as opposed to the concomitant pharmacologic effects of statins, remains uncertain. I believe the authors’ conclusion would be strengthened if the potential contribution of statin therapy could be addressed in future analyses.

To this end, I respectfully suggest that future studies leverage Korea’s National Health Insurance Service database, which comprehensively captures prescription records. A stratified analysis by statin use status—or the inclusion of statin exposure as a time-varying covariate—would help clarify whether the inverse association between CAC density and MACE persists independently of pharmacologic confounding. Such an approach would meaningfully advance the field by disentangling the prognostic signal of plaque biology from the effects of preventive therapy. I look forward to future investigations from this group that may provide further insights into this significant issue.