Prediction of measured GFR after living kidney donation from pre-donation parameters

Background One of the challenges in living kidney donor screening is to estimate remaining kidney function after donation. Here we developed a new model to predict post-donation measured glomerular filtration rate (mGFR) from pre-donation serum creatinine, age and sex. Methods In the prospective development cohort (TransplantLines, n = 511), several prediction models were constructed and tested for accuracy, precision and predictive capacity for short- and long-term post-donation I-125-iothalamate mGFR. The model with optimal performance was further tested in specific high-risk subgroups (pre-donation eGFR <90 mL/min/1.73 m(2), a declining 5-year post-donation mGFR slope or age >65 years) and validated in internal (n = 509) and external (Mayo Clinic, n = 1087) cohorts. Results In the development cohort, pre-donation estimated GFR (eGFR) was 86 +/- 14 mL/min/1.73 m(2) and post-donation mGFR was 64 +/- 11 mL/min/1.73 m(2). Donors with a pre-donation eGFR >= 90 mL/min/1.73 m(2) (present in 43%) had a mean post-donation mGFR of 69 +/- 10 mL/min/1.73 m(2) and 5% of these donors reached an mGFR <55 mL/min/1.73 m(2). A model using pre-donation serum creatinine, age and sex performed optimally, predicting mGFR with good accuracy (mean bias 2.56 mL/min/1.73 m(2), R-2 = 0.29, root mean square error = 11.61) and precision [bias interquartile range (IQR) 14 mL/min/1.73 m(2)] in the external validation cohort. This model also performed well in donors with pre-donation eGFR <90 mL/min/1.73 m(2) [bias 0.35 mL/min/1.73 m(2) (IQR 10)], in donors with a negative post-donation mGFR slope [bias 4.75 mL/min/1.73 m(2) (IQR 13)] and in donors >65 years of age [bias 0.003 mL/min/1.73 m(2) (IQR 9)]. Conclusions We developed a novel post-donation mGFR prediction model based on pre-donation serum creatinine, age and sex.

Automated summary

A new model was developed to predict post-donation measured glomerular filtration rate (mGFR) based on pre-donation serum creatinine, age, and sex. The model showed good accuracy and precision in predicting mGFR, and performed well in high-risk subgroups.