ENHANCING LONG TERM PREDICTION OF GLOMERULAR FILTRATION RATE IN LIVING KIDNEY DONORS AND RECIPIENTS WITH COMPUTED TOMOGRAPHY-BASED KIDNEY VOLUMES

Kidneys possess an inherent compensatory capacity that tends to preserve glomerular filtration rate (GFR) despite nephron loss, complicating prediction of expected GFR in donors and recipients. Because numbers of nephrons vary between individuals and cannot be directly assessed, non-invasive proxies are needed. Given nephrons are concentrated in the kidney cortex, CT-derived cortical volume is a plausible, non-invasive surrogate of nephron number. In this study, we evaluated whether CT-derived cortical kidney volume enhances prediction of GFR in living donors and their recipients beyond standard clinical factors. We also compared the performance of cortical and total kidney volume.

Contrast-enhanced CT kidney scans were segmented to quantify cortical and total kidney volumes. Living donor I125-iothalamate-based mGFR was evaluated at 50 days and 5 years. Recipient eGFR (CKD-EPI 2021) was recorded at 3 and 6 months and at 1, 3, 5, and 10 years post-transplant. In donors, cross-sectional multivariate linear regressions were conducted, modelling mGFR as the outcome at both timepoints, with age, sex, and pre-donation mGFR as clinical factors, with or without cortical or total kidney volume as predictors. In recipients, repeated measurements of eGFR were analysed by means of linear mixed effects models (N=173, complete cases; REML) with random slopes and intercepts. Models with clinical factors (donor age and pre-donation eGFR, recipient age) were compared with models with clinical factors plus z-scored cortical or total volume, leveraging all available observations with non-missing variables. Model parsimony versus clinical factors was compared using information criteria and potential cortical or total volume×time since transplantation interactions were examined in sensitivity analyses. Recipient multivariate linear regressions were also conducted at each timepoint, modelling eGFR in the recipient as the outcome with donor age, pre-donation eGFR and recipient age as clinical factors, with or without cortical or total kidney volume as predictors.

In donors, cortical volume improved mGFR model fits over clinical factors alone and total volume at ~50 days and 5 years post-transplant (Table 1).
In recipients, adding cortical volume to the linear mixed‑effects model improved fit over clinical factors alone (st. β = 2.73, p<0.001; ΔAIC −8.8; ΔBIC −6.2) and yielded larger information-criterion gains than adding total kidney volume (st. β = 1.99, p<0.05; ΔAIC −5.9; ΔBIC −3.3). These results were robust to the addition of volume×time interaction terms (interaction terms p>0.05). In recipient linear regression models, cortical volume increased adjusted R² at every timepoint. Models with total volume lost significance after 1 year and yielded smaller average adjusted R² improvements (Table 2).

CT-derived kidney volumes add meaningful, structure-sensitive information that improves the prediction of long-term kidney function in both living kidney donors and their recipients, beyond clinical factors alone. Cortical kidney volume outperformed total kidney volumes. These data support cortical volume as a non-invasive marker of nephron numbers.