Can a Calcium Score Be Wrong?

A coronary artery calcium (CAC) score can absolutely be wrong, and it happens more often than most people realize. The score can overstate risk, understate it, or shift between risk categories depending on which scanner was used, whether your heart was perfectly still during the scan, and even which software processed the images. A zero score, widely considered reassuring, misses noncalcified plaque in roughly one out of every ten symptomatic patients. And a rising score sometimes reflects medications doing exactly what they should. None of this means the test is useless, but treating any single number as a definitive verdict on your heart health is a mistake.

When a Zero Score Does Not Mean Zero Risk

The most consequential way a calcium score can mislead you is the false sense of security that comes with a score of zero. The Agatston scoring method detects calcified deposits in coronary arteries, but atherosclerotic plaque exists in forms that contain little or no calcium. These noncalcified or “soft” plaques are invisible to the standard calcium scan. In a study of 447 symptomatic patients who scored zero, about 10.5% turned out to have noncalcified coronary plaques on CT angiography, and four of those patients had blockages greater than 50%.¹ A separate analysis of 367 zero-score patients found a similar rate of hidden plaque at 9.3%, and among those patients, nearly half had significant coronary obstructions.²

Young adults face a particular version of this problem. A recent study found that among young adults with chest pain and a zero calcium score, about 2.2% still had stenosis of 70% or greater.³ That percentage sounds small in isolation, but it represents real people with significant blockages who might be sent home with a clean bill of health. Among symptomatic patients more broadly, a large study of over 1,100 people found that a zero calcium score still carried a negative predictive value of about 99% for significant stenosis, which sounds excellent until you consider that the remaining 1-2% translates into missed disease in thousands of patients when applied across a population.⁴

Different Scanners, Different Scores

Your calcium score can change meaningfully depending on which CT machine produces it. A study comparing four different scanner brands found statistically significant differences in Agatston scores across all of them. The Toshiba scanner produced the highest scores, followed by Philips, GE, and Siemens.⁵ When comparing a 64-slice scanner to a dual-source scanner, researchers found correlation was high on paper but the actual limits of agreement were wide, with volume score variability between scanners reaching about 45%. Part of the problem turned out to be one vendor’s scoring software systematically overestimating volume scores, a bias confirmed by phantom measurements.⁶

Even within the same machine, the reconstruction algorithm matters. On Siemens scanners, switching from one type of image reconstruction to another increased the Agatston score by a measurable margin. GE and Siemens scanners both showed significant volume score increases with iterative reconstruction compared to traditional methods.⁵ Tube voltage settings also play a role: simply lowering the voltage from the standard 120 kVp to 90 kVp increased score variability by about 67%. Thinner slice thickness helped reduce that variability, but the combination of low voltage and thick slices produced the least reliable numbers.⁷

For you as a patient, the practical takeaway is straightforward: if you are getting repeat calcium scans to track progression over time, doing them on the same scanner with the same settings matters. A score that jumps from 150 to 220 might reflect genuine plaque growth, or it might reflect a switch from one machine to another.

Motion Artifacts and Heart Rate

Your heart does not hold still during a CT scan. That movement creates blurring and artifacts that inflate the apparent size of calcified deposits. Research on motion correction has found that the effect is not trivial. In one study applying a motion correction algorithm, the average Agatston score dropped from about 494 to 410 after correcting for cardiac movement. That shift was enough to change the risk category for about 13% of patients.⁸

Scans performed without ECG synchronization, where the machine does not time its images to the heartbeat, are especially vulnerable. A study on automatically detecting motion-affected calcium deposits found that scan-to-scan reproducibility improved substantially when the analysis excluded calcifications severely affected by cardiac motion.⁹ If you have an irregular or fast heart rate, the scan is more likely to blur individual plaques, pushing the score up. This is one reason that dedicated cardiac CT protocols with ECG gating produce more reliable scores than opportunistic reads from chest CTs ordered for other reasons.

The Statin Paradox

Here is a scenario that confuses a lot of patients: you start taking a statin, your doctor orders a follow-up calcium scan, and the score has gone up. That sounds alarming, but it may be exactly what a protective medication does. The PARADIGM study confirmed that statins increase coronary calcium scores even as they reduce the overall volume of dangerous plaque and lower the risk of heart attacks.¹⁰ A similar finding came from the long-running Heinz Nixdorf Recall study, which found that statin use was associated with calcium score progression despite simultaneously lowering cardiovascular event risk. The researchers concluded that this pattern likely reflects plaque stabilization.¹¹

The biology behind this involves how plaques mature. Dangerous plaques tend to have a lipid-rich, necrotic core covered by a thin cap. When they calcify more densely, that process often represents the plaque becoming more stable and less likely to rupture. Statins appear to accelerate this conversion, shrinking the necrotic core while increasing calcification.¹² So a rising score on statins does not mean the same thing as a rising score in someone not taking medication. This is one of the most important ways a calcium score can be “wrong” in a functional sense: the number is technically accurate, but interpreting it at face value leads to the wrong clinical conclusion.

Not All Calcium Is Equally Dangerous

The standard Agatston scoring method assigns higher weight to denser calcium, which turns out to be somewhat backwards from a risk perspective. In the MESA cohort, which followed over 3,000 people with detectable calcium for a median of about seven and a half years, calcium density was inversely associated with heart attacks and cardiovascular events. At any given volume of plaque, denser calcium meant fewer events, not more.¹³ This makes pathological sense: dense, solid calcification signals a stable, healed plaque, while spotty, fragmented, or “microcalcification” patterns are associated with unstable lesions more prone to rupture.¹⁴

This creates a real limitation in the traditional score. Two people with the same Agatston number can have very different risk profiles depending on whether their calcium is concentrated in a few dense, stable plaques or scattered as tiny fragments across multiple unstable lesions. Researchers have proposed alternative scoring methods that separate volume from density to better capture this distinction, but these newer approaches have not yet replaced the Agatston method in routine practice.¹⁵

Age, Sex, and Ethnicity Shape What a Score Means

A calcium score of 100 does not mean the same thing in a 45-year-old woman as it does in a 70-year-old man. Calcium accumulates with age, and the rate differs substantially between sexes and ethnic groups. The MESA study, which established the most widely used reference percentiles, found that white men consistently had the highest calcium levels across all age groups. Among women, whites also had the highest values, while Hispanic women generally had the lowest. Black and Chinese participants fell in between, with their relative ranking shifting depending on age.¹⁶

This means your calcium score should ideally be interpreted against age-, sex-, and ethnicity-matched percentiles rather than raw thresholds. A score of 50 in a 40-year-old might place them well above the 90th percentile for their demographic, while the same score in a 75-year-old could be unremarkable. In young adults specifically, about one in ten in the general population and one in three of those with traditional risk factors have some detectable calcium. While calcium scoring is not recommended for universal screening in young adults, its ability to identify early disease and reclassify risk has drawn increasing interest.¹⁷¹⁸

Repeat Scans and the Noise Floor

If you get two calcium scans back-to-back on the same machine, the scores will not be identical. Every measurement has built-in variability, and for calcium scoring that variability is large enough to matter clinically. Work from the MESA cohort quantified the repeatability limits and found that the expected difference between two scans depends on the score itself, scaling up with higher scores, and also on body mass index. Larger patients have more noise in the images, which widens the range of plausible scores.¹⁹

The clinical consequence is that a modest change in score between two visits does not necessarily indicate real disease progression. A jump from 80 to 110 might sit entirely within measurement noise. Doctors tracking progression over time generally look for changes that exceed these repeatability thresholds before concluding that plaque has genuinely worsened. Without that discipline, patients can end up with unnecessary anxiety or escalation in treatment based on statistical noise rather than biological change.

Opportunistic Scoring From Chest CT Scans

Calcium scores are sometimes extracted “for free” from chest CTs ordered for other reasons, such as lung cancer screening. These scans use lower radiation doses and often lack ECG gating, which means the images are noisier and the heart may be blurred by motion. How reliable are these opportunistic scores? One study found that a low-dose chest CT protocol at 2.5 mm slice thickness and 130 Hounsfield units correlated reasonably well with reference standard results, achieving about 90% overall accuracy.²⁰

Ultra-low-dose scans with tin filtration, a newer technique that dramatically reduces radiation, have shown high correlation with standard ECG-gated calcium scoring but only moderate agreement in risk category assignment. Researchers developed adjusted thresholds for these scans that improved consistency from moderate to good.²¹ Deep learning denoising algorithms have further improved ultra-low-dose scan accuracy, pushing the ability to distinguish calcium-present from calcium-absent cases close to 100%.²² These scans are getting better quickly, but if you received a calcium score from a lung screening CT, it is worth knowing that the number may be less precise than one from a dedicated cardiac scan.

AI Scoring Versus Human Reading

Automated calcium scoring powered by artificial intelligence is increasingly used to speed up reporting, especially for opportunistic scoring from non-cardiac CTs. Overall, AI and human readers agree closely: one study found a correlation coefficient of 0.996 between AI and expert readings. But despite that impressive headline correlation, 14% of patients were reclassified into a different risk category when the AI score was compared to the human score. The biggest source of disagreement was at the low end, particularly the distinction between a true zero and a very low score. AI tended to undercount calcium in the right coronary artery and sometimes misidentified dense tissue in the right ventricle as calcium.²³

That 14% reclassification rate matters because clinical decisions often hinge on which category a patient falls into. The boundary between a zero score and any calcium at all is a particularly consequential one, since a zero score typically leads to reassurance and no further testing. If the AI calls it zero when a careful human reader would have spotted a tiny deposit, that patient may miss an early intervention window.

Kidney Disease Complicates Everything

People with chronic kidney disease develop arterial calcification through a different pathway than typical atherosclerosis. In addition to the usual intimal calcification within plaque, kidney disease promotes medial calcification, where the middle layer of the artery wall stiffens and calcifies independently of plaque. The standard calcium scan cannot tell these two processes apart. As the MESA investigators noted, no current imaging method can reliably distinguish intimal calcification from medial calcification.²⁴ This means a high calcium score in someone with advanced kidney disease may overestimate the amount of actual atherosclerotic plaque, since a portion of the detected calcium reflects a fundamentally different disease process. People with diabetes face a similar issue, as diabetes also drives medial calcification.

How Scores Change Real Clinical Decisions

Despite all these limitations, calcium scoring has a measurable impact on what doctors and patients actually do. In a randomized trial comparing calcium score-guided statin decisions to traditional risk-calculator-based recommendations, over a third of patients in the calcium score group were reclassified from “start a statin” to “no statin needed,” while about 6% were reclassified in the other direction. Doctors were more likely to follow the study recommendation when it was based on a calcium score rather than a risk calculator, and patients who received calcium-guided advice were significantly more likely to actually take their prescribed statin three months later.²⁵

At the same time, newer risk calculators do not always agree with calcium scores. A recent analysis comparing calcium burden to the PREVENT risk estimation tool found that over half of patients were “missed” by the calculator, and nearly 80% of patients with high calcium burdens had their risk underestimated by the tool alone.²⁶ This kind of disagreement is actually an argument for the test’s value: the calcium scan adds information that clinical risk factors alone cannot provide. The score can be imprecise, subject to scanner variability, and blind to soft plaque, yet it still reshapes decisions in ways that improve both physician confidence and patient follow-through.

What Else the Scan Reveals

A calcium scoring CT images more than just your coronary arteries. Because the scan captures a slice of your chest, it can incidentally reveal findings in the lungs, liver, and surrounding structures. Pulmonary nodules are the most common surprise finding, particularly in older adults and smokers. Liver cysts, pleural fluid, chest wall abnormalities, breast calcifications, and enlarged lymph nodes in the mediastinum have all been reported as incidental discoveries.²⁷ Most of these findings are benign and require nothing more than noting them in the report, but some, particularly lung nodules in high-risk patients, have established follow-up protocols. An incidental finding is not a “wrong” calcium score in any strict sense, but it can generate anxiety and additional testing that the patient did not anticipate when agreeing to the scan.

Research has also explored whether the tissue surrounding the coronary arteries on a calcium scan, particularly fat around the vessels, can add predictive information. One study found that analyzing features of pericoronary fat on calcium scoring images produced better discrimination for predicting disease than the Agatston score alone, suggesting that the standard number we extract from these scans may be leaving useful information on the table.²⁸ Future scoring methods may integrate these features alongside calcium volume and density, narrowing the gap between what the scan captures and what it currently reports.