Assessing Long-Term Stability of Lipoprotein(a) in Frozen Blood Plasma Samples: A 14-Year Analysis from The Irish Longitudinal Study on Ageing (TILDA)

<ns7:p> Background Degradation of lipoprotein(a) [Lp(a)] during storage may compromise the accuracy of measurements derived from historic or biobanked samples, with implications for retrospective research and cardiovascular risk estimation. Understanding if frozen Lp(a) values reliably reflect concentrations measured in fresh plasma is essential. This study aims to assess the stability of Lp(a) by comparing fresh plasma concentrations with short- and long-term frozen samples from matched participants in a longitudinal cohort. Method Using data from the Irish Longitudinal study on Ageing, paired fresh and short-term frozen (4 days at -80 <ns7:sup>o</ns7:sup> C), or long-term frozen (>10years at -80 <ns7:sup>o</ns7:sup> C) were analysed using an isoform-insensitive immunoassay. Bland-Altman analyses assessed absolute and percentage differences between fresh and frozen measurements and evaluated whether degradation varied across Lp(a) concentration. Cardiovascular risk classification based on Lp(a) thresholds was compared between fresh and frozen samples. Results Short-term frozen samples demonstrated modest Lp(a) reduction (mean -5.8 nmol/L; 12.1% decrease), whereas long-term storage resulted in substantial loss (mean -41.1 nmol/L; 61% decrease). Percentage differences showed no proportional bias, indicating uniform degradation across concentration ranges. Storage effects altered cardiovascular risk classification in 10% of short-term samples and 49% of long-term samples. Notably, 67% of individuals with fresh Lp(a) ≥105 nmol/L were no longer above this threshold after long-term storage, with 18% misclassified as <62 nmol/L. Conclusions Lp(a) undergoes significant degradation during frozen storage, particularly after long-term biobanking. These changes can substantially underestimate elevated Lp(a) prevalence and cardiovascular risk. Retrospective studies relying on long-stored samples should therefore interpret Lp(a) measurements with caution. </ns7:p>