طريقة جديدة للنظر داخل قلبك بدون جراحة — هل يمكن لفحص التصوير المقطعي الشرياني التاجي المتقدم تحسين التنبؤ بخطر الإصابة بالنوبات القلبية؟

نماذج المخاطر التقليدية،, تصوير الأوعية المقطعي المحوسب للشرايين التاجية, ، والكمية المدعومة بالذكاء الاصطناعي لوحة تحليل: ما يظهره الدليل وما لا يظهره

تحليل سكابيس (SCAPIS) والتحول نحو تقييم المخاطر القائم على الأمراض

تحليل بيرغستروم وآخرون لدراسة الصور الحيوية القلبية الرئوية السويدية (سكابيس)، نُشر على الإنترنت في مجلة الجمعية الطبية الأمريكية في 9 نوفمبر 2025 وفي النسخة المطبوعة باسم جاما. 2026;335(3):245–254، وهو أكبر اختبار قائم على السكان حتى الآن لسؤال بسيط: هل النظر مباشرة إلى الشرايين التاجية أخبرك بشيء أن عوامل الخطر و نقاط الكالسيوم لا تفعل؟

تابعت الدراسة 24,791 فرداً تتراوح أعمارهم بين 50 و64 عاماً، تم اختيارهم عشوائياً من عامة السكان في ستة مستشفيات جامعية سويدية ولم تكن لديهم إصابات مثبتة أمراض القلب والأوعية الدموية في الأساس، ولمتوسط مدة 7.8 سنوات.¹ وكانت النتيجة حدوث أول حالة غير مميتة احشاء عضلة القلب أو الموت من مرض الشريان التاجي; وقعت 304 أحداث من هذا القبيل. وكان السؤال هو ما إذا كان إضافة الشريان التاجي التصوير المقطعي المحوسب نتائج تصوير الأوعية التاجية المقطعي المحوسب (CCTA) إلى نموذج يحتوي على معادلة الأمهات المُجمَّعة (PCE) و نقاط كالسيوم الشريان التاجي (CACS) أسست التنبؤ.¹

مدى تصلب الشرايين أهم من أي شيء فردي تضيق. . نقاط مشاركة القطاعات, ، والذي يحسب عدد القطع التاجية الـ 18 التي تحتوي على أي لويحة، وارتبط بقوة بالأحداث: درجة SIS من 3 إلى 4 حملت نسبة الخطر (الموارد البشرية) لـ 2.71 (95% CI، 1.34–5.44) ومؤشر SIS أكبر من 4، فإن معدل الخطر (HR) يبلغ 5.27 (95% CI، 2.50–11.07)، مقارنةً بالدرجات الأقل. وكان وجود تصلب الشرايين غير المتكلس مصحوبًا بمعدل خطر يبلغ 1.66 (95% CI, 1.23–2.22).¹

Predictive variable	Hazard ratio	95% CI	Model
SIS 1–2	1.00	Reference	Fully adjusted
SIS 3–4	2.71	1.34–5.44	Fully adjusted
SIS >4	5.27	2.50–11.07	Fully adjusted
Noncalcified atherosclerosis	1.66	1.23–2.22	Fully adjusted
Stenosis ≥50%	1.22	0.89–1.69	Fully adjusted

Table 1. Association of CCTA findings with first coronary events in SCAPIS.¹ All values are from the fully adjusted model. Note that once plaque extent is accounted for, obstructive stenosis (≥50%) is ليس independently associated with first events in this primary-prevention cohort — the confidence interval crosses 1.0. Unadjusted and partially adjusted models give larger stenosis estimates and should not be quoted as if they were adjusted results.

Adding the CCTA variables to a model containing PCE and CACS improved discrimination modestly: the C-statistic rose from 0.764 to 0.779 (P = .004).¹ Reclassification improved as well, with a net reclassification improvement of 0.133 (95% CI, 0.031–0.165). Among participants who went on to have an event, 14.2% were correctly moved into a higher risk category; among those who did not have an event, 1.6% were incorrectly moved upward.¹ Because the overall event rate was low, most of this reclassification occurred among people the PCE had labeled low risk (<5%) — which is precisely the group in which a missed diagnosis is most consequential, and also the group in which unnecessary treatment is hardest to justify.

The investigators themselves characterize the gain as modest.¹ That word is worth keeping. SCAPIS shows that direct imaging of plaque adds real, statistically robust information beyond risk factors and calcium scoring in a middle-aged primary-prevention population. It does not show that imaging transforms individual risk prediction, and it was not designed to show that acting on the imaging improves outcomes.

Two Different Questions: Detecting Stenosis vs Predicting Events

Much of the confusion in the popular coverage of AI-assisted CCTA comes from treating two very different performance metrics as interchangeable. They are not, and conflating them inflates the apparent capability of the technology.

• Diagnostic accuracy asks: how well does the test agree with a reference standard — typically invasive quantitative coronary angiography (QCA) — about whether a stenosis is present الآن?
• Prognostic discrimination asks: how well does the test rank people by their likelihood of having a cardiovascular event في المستقبل?

An area under the curve (AUC) of 0.91 for the first question is not evidence of an AUC of 0.91 for the second. Diagnostic AUCs against an anatomic reference standard are systematically higher than event-prediction AUCs, because future events depend on plaque biology, hemodynamics, thrombotic propensity, treatment, and chance — not only on anatomy. The two tables below are therefore presented separately and should not be merged into a single ranking.

Diagnostic accuracy: agreement with invasive angiography

In a post hoc analysis of the PACIFIC-1 cohort (208 patients with new-onset stable chest pain, all of whom underwent both CCTA and invasive QCA), AI-guided quantitative CT (AI-QCT) was compared directly with human readers of differing experience for the detection of ≥50% stenosis on a per-patient basis.³

Reader	AUC (95% CI)	Task
AI-QCT	0.91 (0.87–0.95)	≥50% stenosis vs QCA
Level-3 expert reader	0.77 (0.70–0.83)	≥50% stenosis vs QCA
Level-2 reader A	0.79	≥50% stenosis vs QCA
Level-2 reader B	0.76	≥50% stenosis vs QCA

Table 2. Per-patient diagnostic accuracy for obstructive stenosis, PACIFIC-1 post hoc analysis, invasive QCA as reference standard.³ These are diagnostic AUCs. They describe agreement with invasive angiography about present anatomy. They do not describe prediction of myocardial infarction.

The correct reading of this result is that automated quantitative analysis agreed with the invasive reference standard more closely than expert visual assessment did, and did so reproducibly. That is a meaningful finding about interpretive consistency — visual CCTA reading is known to be experience-dependent and to overestimate stenosis — but it is a statement about diagnosis, not about prognosis.

Prognostic discrimination: predicting future events

In the CONFIRM2 registry, AI-QCT was tested for incremental prognostic value over the qualitative and semiquantitative measures currently recommended in practice. Adding AI-QCT improved discrimination for major adverse cardiovascular events over CAD-RADS 2.0, CACS, and the modified Duke Index.⁴

Comparator (alone)	AUC alone	AUC + AI-QCT
CAD-RADS 2.0	0.79	0.81
Coronary شريان نقاط الكالسيوم	0.70	0.79
Modified Duke Index	0.76	0.81
PCE + CACS (SCAPIS, + CCTA)¹	0.764	0.779

Table 3. Prognostic discrimination for future events.¹˒⁴ The CONFIRM2 rows describe specific models within one registry of symptomatic patients referred for CCTA; the SCAPIS row describes an asymptomatic general-population cohort and non-AI CCTA variables. These are ليس interchangeable, and none of them is a universal performance characteristic of AI-QCT.

Taken together, the honest summary is this: quantitative, AI-assisted CCTA measurements can provide incremental prognostic information beyond conventional clinical risk assessment, calcium scoring, and qualitative CCTA reads — but the magnitude of that improvement varies substantially with the population studied, the endpoint chosen, and the model specification. In the contemporary prognostic studies discussed here, reported event-prediction AUCs for AI-QCT are approximately 0.75 to 0.81 rather than 0.90.¹˒⁴

What Quantitative Plaque Analysis Measures, and Why It Matters

Traditional risk equations such as the Framingham Risk Score و Pooled Cohort Equations infer vascular risk from surrogate variables — age, sex, ضغط الدم, lipids, تدخين, مرض السكري — rather than from the artery itself.⁹˒¹⁰ They perform reasonably at the population level and imperfectly at the individual level, because the relationship between risk factors and plaque development is heterogeneous. Two people with identical risk-factor profiles can have very different coronary arteries.

Quantitative CCTA measures the disease directly. Rather than reporting only the tightest narrowing, it characterizes plaque throughout the coronary tree by volume, composition, and vessel remodeling. Several of these features carry prognostic information:

• Total plaque volume (TPV). Higher total atherosclerotic burden is independently associated with future events across multiple cohorts. Effect sizes are study-specific and depend on the population, the units in which volume is expressed, and the covariates adjusted for; a risk estimate from one cohort should not be quoted as a general property of حجم اللويحة.⁴˒⁵
• Noncalcified plaque (NCP) volume. In CONFIRM2, each 50-mm³ increase in noncalcified plaque was associated with a 1% relative increase in MACE risk in women and 11.6% in men.⁵
• Low-attenuation plaque (≤30 HU). A CT marker associated with lipid-rich, necrotic-core plaque characteristics — an imaging correlate rather than a direct measurement of necrotic core. In SCOT-HEART, low-attenuation plaque burden was the strongest predictor of fatal or nonfatal myocardial infarction (adjusted HR 60 per doubling; 95% CI, 1.10–2.34), and a burden above 4% was associated with an HR of 4.65 (95% CI, 2.06–10.5).²
• Positive (outward) remodeling. Outward expansion of the vessel wall may allow substantial plaque to accumulate before marked luminal narrowing becomes apparent, which is one reason lumen-focused assessment can underestimate disease burden.

The biology underlying these measurements is more nuanced than a simple “calcium is safe, لويحة لينة is dangerous” dichotomy. Dense calcification is generally associated with more stable lesion behavior, and calcification does often follow inflammatory injury as part of a healing response — but لويحة متكلسة is not inert, and calcium burden remains a strong marker of overall atherosclerotic disease. Conversely, noncalcified plaque is not synonymous with vulnerable plaque; it is a heterogeneous category, only part of which has the lipid-rich, thin-capped phenotype associated with rupture. The useful statement is narrower and better supported: plaque composition adds prognostic information that calcium score alone does not capture.

A related point is frequently overstated in consumer coverage. It is well documented that many acute coronary events arise from lesions that were not severely obstructive on prior imaging, which is a central reason plaque burden and composition can carry information beyond stenosis severity. It does not follow that “most النوبات القلبية are caused by non-obstructive soft plaque” — that phrasing compresses several distinct issues (pre-event stenosis severity, plaque phenotype, rupture versus erosion, and the limitations of retrospective angiographic comparison) into a single claim that the evidence does not cleanly support.

Scanner Technology: What Wide-Detector CT Actually Delivers

Contemporary wide-detector CT systems can acquire the entire heart within a single cardiac cycle in appropriately selected patients. This reduces misregistration (“stitching”) artifacts that arise when a volume is assembled from multiple heartbeats, and it permits high-quality coronary imaging at relatively low radiation doses when protocols are optimized.

Several caveats belong alongside that statement, because they are routinely omitted:

• Marketing terms such as “640-slice” describe a reconstruction characteristic of a wide-area detector system rather than an independent measure of image quality, and they do not by themselves establish superior outcome prediction. Detector coverage (for example, 16 cm) and acquisition architecture are the more meaningful descriptors.
• Radiation dose cannot be stated as a single number. It varies with scanner generation, prospective versus retrospective gating, tube voltage and current, معدل ضربات القلب and rhythm, body habitus, scan length, and reconstruction algorithm.¹¹ Sub-millisievert coronary CTA is achievable in favorable patients with optimized protocols; it is not a general expectation.
• Spatial resolution figures are scanner- and protocol-specific and should be verified against the manufacturer’s technical specification for the exact system and acquisition mode in question before being cited.
• Single-beat acquisition reduces — but does not eliminate — the need for heart-rate control, and does not eliminate motion artifact.

The practical significance of wide-detector systems for this discussion is that reproducible, lower-dose acquisition makes serial imaging more feasible, which in turn makes longitudinal plaque tracking a technically feasible clinical and research proposition. Routine serial CCTA performed specifically to monitor plaque remains an evolving strategy rather than established standard care for most patients.

Is Plaque Modifiable? EVAPORATE and the Limits of What It Shows

EVAPORATE, a randomized, double-blind, placebo-controlled trial of 80 patients with elevated triglycerides on statin therapy, used serial CCTA to assess the effect of icosapent ethyl on plaque progression over 18 months. It found reduction in low-attenuation plaque volume in the treatment arm alongside progression in the دواء موهم arm.⁷

EVAPORATE is a mechanistic imaging trial. It was not an AI-QCT study and it was not powered for clinical events. It supports the proposition that plaque composition is modifiable and measurable over time; it does not validate the prognostic performance of any particular AI analysis platform, and it should not be cited as though it did.

TRANSFORM: A Trial Designed to Test the Outcome Question

TRANSFORM (NCT06112418) is a prospective, randomized, open-label, blinded-endpoint trial enrolling patients at elevated cardiovascular risk without known symptomatic disease. Participants are randomized to guideline-directed risk-factor-based care or to a care strategy driven by an AI-based coronary plaque staging system, with repeat imaging at 24 months in the imaging arm.⁸

TRANSFORM is ongoing. Its results are not yet available, and it therefore cannot be cited as evidence supporting any conclusion about the effectiveness of an imaging-guided prevention strategy. Its importance lies in what it is designed to establish: whether identifying and staging plaque earlier, and treating accordingly, actually reduces cardiovascular events compared with current practice. That is the question the existing observational and diagnostic literature cannot answer.

Impact on Clinical Decision-Making

There is reasonable evidence that quantitative plaque analysis changes what clinicians do. In the CERTAIN study, a multicenter crossover study of 750 consecutive patients referred for CCTA at five expert sites, physicians recorded their diagnosis and management plan based on conventional site interpretation and then repeated the assessment after AI-QCT analysis.⁶

• Diagnosis or management changed in 1% of patients (P < .001).⁶
• Statin initiation or intensification increased in an additional 1% of patients, and aspirin initiation in an additional 23.0% (P < .001 for both).⁶
• The anticipated need for downstream noninvasive and invasive testing fell by 1% (P < .001).⁶

These are changes in physician intent measured within a study design in which the same physicians assessed the same patients twice, in a fixed order, at high-volume expert centers. That design cannot exclude ordering effects, and intent to prescribe is not the same as improved outcomes. The finding is genuinely encouraging about clinical utility and genuinely insufficient as evidence of clinical benefit.

Coverage for AI-based coronary plaque analysis has expanded among United States payers since 2024, but policies remain plan-, indication-, and region-specific. Patients and clinicians should verify coverage directly with the payer rather than relying on general statements of availability.

Sex-Specific Implications

One of the most clinically relevant findings in this literature concerns sex differences in how plaque burden translates into risk. In the CONFIRM2 registry (3,551 symptomatic patients, 49.5% women, mean follow-up 4.8 ± 2.2 years), women had roughly half the plaque burden of men and about half the MACE rate (3.2% vs 6.1%). But the الخطر النسبي conferred by each increment of plaque was consistently higher in women.⁵

Plaque measure (per 50 mm³ increase)	Increase in MACE risk, women	Increase in MACE risk, men
Total plaque volume	+17.7%	+5.3%
Noncalcified plaque	+27.1%	+11.6%
Calcified plaque	+22.9%	+5.4%

Table 4. Sex-specific relative risk per 50-mm³ increment, CONFIRM2 registry (P for interaction < .001).⁵ Note that total plaque volume, noncalcified plaque, and calcified plaque have distinct values and must not be quoted interchangeably.

The clinical implication is that a plaque volume that looks reassuringly “low” by absolute standards derived largely from men may carry meaningful prognostic weight in a woman. Quantitative measurement, which reports actual volumes rather than a qualitative impression, is well suited to detecting exactly this pattern — provided the interpretation is sex-aware.

Synthesis and Conclusion

The central argument holds. Coronary CT angiography identifies atherosclerotic disease that risk-factor equations and calcium scoring do not fully characterize, and quantitative, AI-assisted analysis of those images adds diagnostic reproducibility and incremental prognostic information. SCAPIS provides strong contemporary evidence that CCTA improves prediction of first coronary events beyond the PCE and CACS in a middle-aged general population, particularly among people those tools classify as low risk.¹ CONFIRM2 shows that quantitative plaque measurement improves discrimination over the qualitative and semiquantitative measures used in practice today, and that plaque burden carries greater relative risk in women.⁴˒⁵ CERTAIN shows that clinicians change their management when given this information.⁶

What the evidence does not yet establish is equally important:

• It does not establish that AI-assisted CCTA predicts cardiovascular events with an AUC near 0.90. That figure comes from diagnostic accuracy against invasive angiography, not from event prediction.³˒⁴
• It does not establish that imaging-guided prevention improves clinical outcomes. No completed randomized trial has demonstrated this. TRANSFORM is designed to test it and has not reported.⁸
• It does not establish that any single scanner configuration or analysis platform is superior in outcome terms.

The reasonable conclusion is a strong one without being an overclaim: direct visualization and quantification of coronary plaque represents a substantive advance over inference from risk factors alone, and it identifies disease in people whom conventional tools reassure. Whether acting on that information reduces heart attacks is a question currently under randomized investigation, and the answer will come from TRANSFORM and trials like it rather than from registries, diagnostic accuracy studies, or extrapolation.

For an individual patient, the practical takeaway is unchanged by any of these caveats: finding plaque may identify an opportunity to intensify evidence-based preventive treatment, and serial CCTA studies demonstrate that plaque characteristics are measurably modifiable.⁷ The appropriate posture toward the technology is confident engagement paired with accurate expectations about what the evidence currently supports.

References

1. Bergström G, Engström G, Björnson E, et al. Coronary computed tomography angiography in prediction of first coronary events. JAMA. 2026;335(3):245-254. doi:10.1001/jama.2025.21077. (Published online November 9, 2025.)
2. Williams MC, Kwiecinski J, Doris M, et al. Low-attenuation noncalcified plaque on coronary computed tomography angiography predicts myocardial infarction: results from the multicenter SCOT-HEART trial (Scottish Computed Tomography of the HEART). Circulation. 2020;141(18):1452-1462. doi:10.1161/CIRCULATIONAHA.119.044720
3. Bernardo R, Nurmohamed NS, Bom MJ, et al. Diagnostic accuracy in coronary CT angiography analysis: artificial intelligence versus human assessment. Open Heart. 2025;12(1):e003115. doi:10.1136/openhrt-2024-003115
4. van Rosendael A, Nakanishi R, Bax JJ, et al. Prognostic value of AI-based quantitative coronary CTA vs human reader-based visual assessment: results from the CONFIRM2 registry. JACC Cardiovasc Imaging. 2026;19(3):345-359. doi:10.1016/j.jcmg.2025.09.021
5. Feuchtner GM, Lacaita PG, Bax JJ, et al. AI-quantitative CT coronary plaque features associate with a higher relative risk in women: CONFIRM2 registry. Circ Cardiovasc Imaging. 2025;18(6):e018235. doi:10.1161/CIRCIMAGING.125.018235
6. Nurmohamed NS, Cole JH, Budoff MJ, et al. Impact of atherosclerosis imaging-quantitative computed tomography on diagnostic certainty, downstream testing, coronary revascularization, and medical therapy: the CERTAIN study. Eur Heart J Cardiovasc Imaging. 2024;25(6):857-866. doi:10.1093/ehjci/jeae029
7. Budoff MJ, Bhatt DL, Kinninger A, et al. Effect of icosapent ethyl on progression of coronary atherosclerosis in patients with elevated triglycerides on statin therapy: final results of the EVAPORATE trial. Eur Heart J. 2020;41(40):3925-3932. doi:10.1093/eurheartj/ehaa652
8. [Trial registry record] TRANSFORM — A randomized comparison of Cleerly coronary artery disease stage-based care versus risk factor-based care for primary prevention of cardiovascular events. ClinicalTrials.gov identifier NCT06112418. US National Library of Medicine. Accessed August 2026. Cited for trial design and status only; no results have been reported.
9. Goff DC Jr, Lloyd-Jones DM, Bennett G, et al. 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(25 suppl 2):S49-S73. doi:10.1161/01.cir.0000437741.48606.98
10. D’Agostino RB Sr, Vasan RS, Pencina MJ, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008;117(6):743-753. doi:10.1161/CIRCULATIONAHA.107.699579
11. Hausleiter J, Meyer T, Hermann F, et al. Estimated radiation dose associated with cardiac CT angiography. JAMA. 2009;301(5):500-507. doi:10.1001/jama.2009.54