Bayesian analysis of glomerular filtration rate trajectories in kidney transplant recipients: a pilot study

Abstract

BACKGROUND: Detailed modeling and analysis of renal (dys)function trajectories has not been undertaken in kidney transplant recipients. Although previous studies have assumed linear trajectories, this likely represents an oversimplification. METHODS: In this study, a Bayesian smoothing technique was undertaken to create 10,000 Monte Carlo samples for each of 158 patients over a median of 88 months. Specific parameters investigated were the prevalence of nonlinear trajectories, periods of nonprogression, and of rapid progression. RESULTS: Forty-five (28%) patients displayed high probability (>80%) for a nonlinear trajectory. Periods of nonprogression were also common, present in 110 (70%) patients. A substantial proportion of patients showed deviation from the classic paradigm of progressive linear loss of graft function with 137 (87%) patients displaying nonlinearity or nonprogression. Only nine (6%) patients demonstrated at least one episode of nonprogression after an episode of progression, that is, once progression occurred, a subsequent period of nonprogression was uncommon. Episodes of nonprogression were less common (P < 0.001) in patients whose grafts subsequently failed, whereas episodes of rapid progression were more common (P = 0.04). CONCLUSION: This study highlights the often nonlinear and nonprogressive nature of renal function decline after transplantation. Heightened understanding of the factors influencing these trajectories should help inform patients and clinicians alike.