Prospective evaluation of a self-report-guided strategy for targeted coronary artery calcium imaging

Technical Summary: Prospective Evaluation of a Self-Report–Guided Strategy for Targeted Coronary Artery Calcium Imaging

Problem Statement
Coronary artery calcium (CAC) imaging via non-contrast computed tomography (CT) is a validated method for directly assessing subclinical calcified coronary atherosclerosis and refining cardiovascular risk stratification beyond traditional risk factors. However, population-wide CAC screening is not currently recommended due to cost, resource constraints, and accessibility issues. While current guidelines suggest selecting individuals for CAC imaging based on traditional risk scores (typically those at intermediate risk), these approaches may fail to identify all individuals with prognostically relevant disease. There is a need for efficient pre-screening strategies to target CAC imaging toward individuals most likely to benefit, thereby reducing unnecessary imaging while identifying those with elevated CAC who may require preventive intervention.

Methodology
The study, titled PRIO-Now (PRedict and Identify cOronary atherosclerosis–Now), was a prospective, observational cohort study conducted in two steps within the Gothenburg catchment area in Sweden.

1. Study Population and Recruitment:

   • 8,000 individuals aged 59–60 years (50% male) were randomly selected from the Swedish population register and invited via postal mail.
   • The narrow age range was chosen to minimize age-related variation in CAC prevalence and to evaluate the strategy in a clinically relevant age group for preventive intervention.

2. Step 1: Web-Based Self-Report Screening:

   • Invited individuals completed a 14-item web-based questionnaire covering demographics, anthropometrics (self-reported weight, height, waist, and hip circumference), smoking history, medical history (medication use, diagnosed diabetes/hypertension), and family history of myocardial infarction.
   • A pre-existing prediction model (developed in the SCAPIS cohort) was applied to estimate the probability of having CAC ≥100.
   • Selection Threshold: A predicted probability of ≥15.3% (representing the top 30% of risk in the development cohort) defined "high risk." Individuals meeting this threshold were invited to Step 2.
   • Exclusions: Participants with self-reported history of myocardial infarction, CABG, PCI, or angina were excluded.

3. Step 2: Clinical Assessment and Imaging:

   • High-risk participants underwent clinical assessment (anthropometrics, blood pressure, ECG, fasting blood samples) and non-contrast ECG-gated coronary CT imaging.
   • CAC scores were calculated in Agatston units.
   • Clinical Management: Participants received written reports. Those with CAC ≥100 and CAC ≥300 were provided with study-defined LDL-C targets (≤1.8 mmol/L and ≤1.4 mmol/L, respectively) and, for CAC ≥300, direct contact with a cardiologist.

4. Statistical Analysis:

   • The primary outcome was the agreement between the model-predicted probability of CAC ≥100 and the observed prevalence among CT completers.
   • Metrics included the expected/observed ratio, calibration intercept/slope, Brier score, and Spiegelhalter's Z-test.
   • Secondary outcomes included the prevalence of lipid-lowering therapy use and LDL-C levels relative to CAC categories.

Key Results

• Participation: Of 8,000 invited, 2,588 (32%) completed the questionnaire. After exclusions, 2,375 eligible respondents remained.
• Risk Classification: 814 participants (34% of eligible respondents) were classified as high risk (predicted probability ≥15.3%).
• CT Completion: Of the 814 high-risk individuals, 643 attended clinical assessment, and 563 completed CT imaging.
• Model Calibration (Primary Outcome):
  • Among the 563 CT completers, the mean predicted probability of CAC ≥100 was 28.3% (95% CI 27.2–29.3).
  • The observed prevalence of CAC ≥100 was 28.4% (95% CI 24.8–32.4).
  • The expected/observed ratio was 0.99.
  • The Brier score was 0.19, and Spiegelhalter's Z-test indicated no evidence of miscalibration (p=0.92).
  • Calibration plots showed agreement across risk strata (tertiles and quintiles).
• Treatment Gaps:
  • Among participants with CAC ≥100 (n=160), 64% were not receiving lipid-lowering therapy.
  • Among treated individuals with CAC ≥100, approximately 75% had LDL-C >1.8 mmol/L; only 8% were both treated and at the study-defined target.
  • Among participants with CAC ≥300 (n=71), 61% were not receiving lipid-lowering therapy, and only 1.4% were both treated and at the target (≤1.4 mmol/L).
• Incidental Findings: 38.2% of participants had incidental non-coronary findings. 76% of those with lung parenchymal abnormalities were referred for follow-up CT.
• Self-Report vs. Clinical Data: Predicted probabilities based on self-reported waist/hip circumference were highly correlated (R=0.96) with those recalculated using clinical measurements, suggesting self-report data did not significantly bias risk estimates.

Significance and Claims
The paper claims that a self-report-guided strategy successfully enabled targeted CAC imaging in a model-selected cohort. The core finding is that the observed prevalence of CAC ≥100 among participants undergoing CT was comparable to the mean model-predicted probability, demonstrating that the self-report model is well-calibrated for prospective selection.

The authors suggest that this approach supports the use of simple self-report data for pre-selection, potentially making CAC imaging more resource-efficient by focusing on high-probability individuals. Furthermore, the study highlights a significant clinical opportunity: a large proportion of individuals identified with substantial subclinical atherosclerosis (CAC ≥100 and ≥300) were not receiving lipid-lowering therapy or were not at guideline-recommended LDL-C targets. The authors posit that this screening strategy could help identify these treatment gaps and guide more targeted preventive interventions.

The study acknowledges limitations, including a 32% response rate (with socioeconomic and sex disparities), potential attrition bias, and the narrow age range (59–60 years), which may limit generalizability to other age groups or populations. The authors note that the high-risk group was predominantly male, reflecting the higher prevalence of CAC in men at this age, and suggest future implementations might need to consider older age groups for women.