In this paper, we sought to create a prediction tool to estimate long-term renal function outcomes after both radical and partial nephrectomy. We created separate tools for each, given that individual factors may be more or less important in each instance. Additionally, we used a hierarchical mixed effects modeling approach to account for the potential irregular timing of postoperative renal function measurements and interaction terms with time from surgery so that the effect and importance of these variables could change over time.
After radical nephrectomy, we identified age, diabetes, preoperative eGFR, preoperative proteinuria, and tumor size as important variables (given the modeling approach, time from surgery is a necessary input in the model, but practically the eventual online prediction tool would provide estimates at fixed time points). There was an interaction between age and time from surgery, representing a small age-related decline in renal function over time. After partial nephrectomy, age, the presence of a solitary kidney, diabetes, hypertension, preoperative eGFR, preoperative proteinuria, and open versus minimally-invasive surgical approach were important predictors. Furthermore, there were interactions between time from surgery and age, diabetes, preoperative eGFR, and preoperative proteinuria, representing greater progressive declines in renal function overtime-related to these factors. As detailed in our paper’s discussion, several facets of these models correspond well to our understanding of the underlying biology of renal function outcomes.
As with any prediction tool, validation in an external cohort would be helpful. Pending such validation, this renal function prediction tool may be helpful for clinicians as a component of a multifaceted discussion about the pros and cons of partial and radical nephrectomy. Whereas until now clinical gestalt is most commonly used to estimate renal function impacts, these models will allow for the clinicians to quantify the magnitude of difference in renal function outcomes. This can then be considered alongside the perceived perioperative risks based on tumor complexity and oncologic factors including technical feasibility. Ultimately, shared decision-making is ideal; these models simply provide objective numbers to facilitate discussion.
Written by: Bimal Bhindi, MD, MSc, FRCSC, Department of Urology, Mayo Clinic, Rochester, MN, USA; Southern Alberta Institute of Urology, Calgary, Alberta, Canada, and R. Houston Thompson, MD, Department of Urology, Mayo Clinic, Rochester, MN, USA
Read the Abstract