Exploring lipoprotein(a) and incident heart failure in cancer patients with anthracycline-associated neoplasms

Abstract

Anthracyclines remain cornerstone therapy for ~2.7 million cancer patients annually, conferring up to 30% survival benefits. However, dose-dependent cardiotoxicity affects ~9% of patients, manifesting as heart failure (HF). Lipoprotein(a) [Lp(a)], a genetically determined biomarker, promotes oxidative stress and inflammation – key CTRCD pathways. Traditional categorisation of Lp(a) using arbitrary cut-offs risks information loss, particularly where baseline risks, treatment toxicity, and competing events interact complexly. Although elevated Lp(a) is associated with cardiovascular events overall, its role in oncology and interaction with cardiotoxicity remains unknown, supporting evaluation as a CTRCD marker. To explore baseline Lp(a) as a candidate biomarker of HF susceptibility following chemotherapy by (i) characterizing its distribution in anthracycline-eligible cancer patients and (ii) evaluating its association with HF across the full concentration range, rather than relying on cut-offs. Observational cohort study using UK Biobank data including 3,775 participants with anthracycline-associated neoplasms. Baseline Lp(a) was measured pre-chemotherapy; prevalent HF cases were excluded. Primary outcome was incident HF, defined as composite of HF diagnosis, hospitalisation, or HF-death. Lp(a) distribution and HF incidence were evaluated by concentration ranges. Associations were examined using logistic regression with Lp(a) modelled continuously (per 1 standard deviation [SD]) and restricted cubic splines (RCS) to assess nonlinearity. Sensitivity analyses evaluated clinical thresholds (75, 100, 125 nmol/L). Results reported as odds ratios (ORs) with 95% confidence intervals (CI). In 3,775 patients, 121 incident HF events occurred (3.2%) over median follow-up of 10.2 years (interquartile range [IQR] 5.4–14.8). Lp(a) was right-skewed (median 21.4 [IQR 10.1-59.3] nmol/L). Highest absolute HF incidence occurred at lowest Lp(a) range (0–10 nmol/L: 4.15%), without monotonic increase across distribution. Each 1-SD increase in Lp(a) was associated with non-significant lower odds of HF (OR 0.85 [95% CI 0.69–1.04]; p=0.13), consistent with HF events clustering at lower Lp(a) ranges and risk plateau at higher levels rather than direct dose-response. RCS showed no nonlinearity (p=0.23). Sensitivity analyses using clinical cut-offs (75, 100, 125 nmol/L) produced inconsistent estimates with wide CI and poor discrimination. In this exploratory analysis, baseline Lp(a) showed no linear or threshold association with incident HF post-chemotherapy. Higher HF incidence at very low Lp(a) levels is hypothesis-generating and may reflect Lp(a) as marker of underlying systemic vulnerability rather than protection. Findings support continuous Lp(a) modeling in cardio-oncology to capture risk heterogeneity and inform future, adequately powered studies for precision risk stratification amid growing global burden of CTRCD.