Abstract Background It is important to assess the temporal reproducibility of circulating cytokines for their utility in epidemiological studies. However, existing evidence is limited and inconsistent, especially for the elderly population. Methods Sixty-five elderly (mean age = 77.89 ± 6.14 years) subjects were randomly selected from an existing prospective cohort study. Levels of 41 cytokines in 195 serum samples, collected at three separate visits that were up to 15.26 years apart, were measured by the Luminex technology. The temporal reproducibility of cytokines was estimated by the intraclass correlation coefficient (ICC) calculated using a mixed-effects model. In addition, data analyses were stratified by the median (4.49 years) of time intervals across sample collection. Sensitivity analyses were performed when excluding subjects with undetectable samples. Results A total of 23 cytokines were detectable in more than 60% of samples. Fair to good (ICC = 0.40 to 0.75) and excellent (ICC > 0.75) reproducibility was found in 10 (Eotaxin, VEGF, FGF-2, G-CSF, MDC, GM-CSF, TGFα, IP-10, MIP-1β, IL-1RA) and 5 (GRO, IFNγ, IL-17, PDGF-AA, IL-4) cytokines, respectively. The results were not changed dramatically in the stratification and sensitivity analyses. Conclusions Serum levels of the selected 15 cytokines measured with Luminex technology displayed fair to excellent within-person temporal reproducibility among elderly population.