Intra-individual variability in lipoprotein(a): insights from 922,304 measurements

Background Lipoprotein(a) is considered largely stable, and current guidelines often recommend a single lifetime measurement. However, emerging evidence suggests that intra-individual variation occurs. Purpose To quantify intra-individual Lp(a) variability and correlations with other biochemical parameters and investigate if variability reflects biological fluctuations or assay-dependent analytical noise. Methods We analyzed a dataset from a medical laboratory comprising 922,304 Lp(a) tests and standard clinical chemistries from 504,845 individuals. Two immunoturbidimetric assays calibrated in mg/dL were used: Roche Tina-quant, Gen. 1 (no longer in use), on MODULAR P (2000–2009) or Randox/Siemens reagents on Siemens ADVIA 2009–2025; ADVIA XPT from 2018. Longitudinal analyses included individuals with ≥2 measurements within same assay period (Roche n=32,089; Siemens n=29,376). Variability was assessed using correlation analyses, change distributions, transitions between Lp(a) categories (<30, 30–50, 50–70, 70–90, 90–110, >110 mg/dL), and mixed-effects models to estimate within- and between-person variance and predictors of log-Lp(a). Results Lp(a) was highly stable across repeated measurements, with high correlation between the first and last test (Spearman r=0.94 for Roche; r=0.96 for Siemens). Most individuals with <30 or >110 mg/dL remained in their category, while intermediate ranges frequently reclassified (Figure 1). For Siemens, 21% in the 30–50 mg/dL range exceeded 50 mg/dL at follow-up, while 15% fell below 30 mg/dL. Among those with 50–70 mg/dL, 14% fell below 50 mg/dL and 29% exceeded 70 mg/dL. Modelling indicated a baseline of around 67 mg/dL was required for ≥95% probability of remaining ≥50 mg/dL on repeat testing. Assay performance differed: Siemens had fewer deviations >25% and lower intra-individual CV (13,2%) than Roche (18,2%), with Roche showing greater variability at low and high Lp(a) (figure 2), suggesting largely analytical rather than biological variation. The expected 95% relative uncertainty was reduced by increasing number of independent measurements for both assays. Lp(a) increased modestly with age (+6.6% per decade). We found a few associations with other biomarkers that were statistically significant but small (LDL-C +0.93% per 1 mmol/L; creatinine +1.02% per 10 µmol/L; TSH +0,14% per 1 mIU/L; HDL-C -3.67% per 1 mmol/L; ALAT -0.36% per 10 U/L. Limitations Some variation may be attributed to changes in the use of medications, lifestyle or other factors that we lack information about. As the variation seems to be assay specific, our results are not directly transferable to other assays. Conclusion Lp(a) is highly stable at the population level with minimal covariation with other biochemical parameters. Yet, intra-individual variability remains, likely reflecting analytical rather than biological factors. There may be value in retesting some patients if clinically indicated in the future.