Abstract Background Donor-derived cell-free DNA (dd-cfDNA) is reportedly a valuable tool for graft surveillance following kidney transplantation (KTx). Possible changes in dd-cfDNA(%) reference values over time have not been evaluated. For long-term monitoring after KTx, changes in host cfDNA might represent a biasing factor in dd-cfDNA(%) determinations. Methods Plasma samples were obtained (n = 929) 12–60 months after engraftment in a cross-sectional cohort of 303 clinically stable KTx recipients. Total cfDNA(copies/mL), dd-cfDNA(%), and dd-cfDNA(copies/mL) were determined using droplet-digital PCR. Stability of threshold values in these stable KTx recipients over time was assessed by 80th, 85th, and 90th quantile regression. Results Upper percentiles of total cfDNA showed a significant decline of −1902, −3589, and −4753 cp/mL/log(month) (P = 0.014, <0.001, and 0.017, respectively), resulting in increasing dd-cfDNA(%) percentiles by 0.25, 0.46, and 0.72%/log(month) (P = 0.04, 0.001, and 0.002, respectively), with doubling of the 85th percentile value by 5 years. In contrast, dd-cfDNA(cp/mL) was stable during the observation period (P = 0.52, 0.29, and 0.39). In parallel increasing white blood cell counts and decreasing tacrolimus concentrations over time were observed. After 5 years, the median total cfDNA was still 1.6-fold (P < 0.001) higher in KTx recipients than in healthy controls (n = 135) and 1.4-fold (P < 0.001) higher than patients with other medical conditions (n = 364). Conclusions The time-dependent decrease of host cfDNA resulted in an apparent increase of dd-cfDNA fraction in stable KTx patients. For long-term surveillance, measurement of absolute dd-cfDNA concentrations appears to be superior to percentages to minimize false positive results.