Testing
Adding Lp(a) improves pre-test probability calibration for obstructive CAD, derivation cohort of 4,262 (Eur Heart J Cardiovasc Imaging 2026)
Original title: Incorporation of lipoprotein(a) levels improves calibration of pre-test likelihood estimates of obstructive coronary artery disease
Derivation cohort of 4,262 patients (54% male, mean age 58) without known coronary artery disease (CAD), testing whether adding Lp(a) to the risk-factor-weighted clinical likelihood (RF-CL) model improves prediction of obstructive CAD; validated in an external cohort of 1,595 patients (49% male, mean age 60). An Lp(a)-adjusted model (RF-CL-Lp(a)) multiplied RF-CL by 1.5 in patients with Lp(a) 125 nmol/L or more. Obstructive CAD was present in 11.1% (derivation) and 12.9% (validation, significant stenosis); relative risk with elevated Lp(a) was 1.51 (95% CI 1.23-1.86) in derivation and 1.19 (95% CI 0.88-1.60) in validation. RF-CL-Lp(a) modestly improved discrimination in derivation (AUC 0.743 vs 0.740) and calibration in patients with elevated Lp(a), and improved reclassification in derivation but not validation. The authors conclude adding elevated Lp(a) to RF-CL improves obstructive CAD prediction accuracy in high-Lp(a) patients.
Original abstract
Aims: Risk factor-weighted clinical likelihood (RF-CL) estimates the probability of obstructive coronary artery disease (CAD) in patients without known CAD. We examined whether adding lipoprotein(a) [Lp(a)] measurements to the RF-CL model improves predictions of obstructive CAD.
Methods And Results: In a derivation cohort (N = 4262; 54% male; mean age 58 years), the prevalence of obstructive CAD at invasive angiography with fractional flow reserve was assessed by Lp(a)-strata. On the basis of initial results, an Lp(a)-adjusted model (RF-CLLp(a)) was developed: RF-CL was multiplied by 1.5 in patients with elevated Lp(a) (≥125 nmol/L) and otherwise unchanged. Discrimination, calibration, and reclassification were compared. Findings were validated in an external validation cohort (N = 1595; 49% male; mean age 60 years) using a comparative endpoint; significant stenosis at invasive angiography or coronary computed tomography.In the derivation cohort, 473 patients (11.1%) had obstructive CAD; in the validation cohort, 206 patients (12.9%) had significant stenosis. The relative risk in patients with elevated Lp(a) was 1.51 [95% confidence interval (CI) 1.23-1.86] and 1.19 (95% CI 0.88-1.60) in the derivation and validation cohort, respectively. In the derivation cohort, the RF-CLLp(a) model showed a higher area under the receiver operating curve than the RF-CL model [0.743 (standard error 0.011) vs. 0.740 (0.013)] and better calibration in patients with elevated Lp(a). Reclassification from RF-CL to RF-CLLp(a) improved likelihood stratification in the derivation cohort but not in the validation cohort.
Conclusion: Adding elevated Lp(a) as a risk factor to the RF-CL model improves accuracy of obstructive CAD in patients with high Lp(a).
Summary written by lp-a.org from the published abstract; figures as published. Page updated 17 August 2026. Methods.