Predicting Renal Function Outcomes After Partial and Radical Nephrectomy

Background

Partial nephrectomy (PN) is generally favored for cT1 tumors over radical nephrectomy (RN) when technically feasible. However, it can be unclear whether the additional risks of PN are worth the magnitude of renal function benefit.

Objective

To develop preoperative tools to predict long-term estimated glomerular filtration rate (eGFR) beyond 30 d following PN and RN, separately.

Design, setting, and participants

In this retrospective cohort study, patients who underwent RN or PN for a single nonmetastatic renal tumor between 1997 and 2014 at our institution were identified. Exclusion criteria were venous tumor thrombus and preoperative eGFR <15 ml/min/1.73 m².

Intervention

RN and PN.

Outcome measurements and statistical analysis

Hierarchical generalized linear mixed-effect models with backward selection of candidate preoperative features were used to predict long-term eGFR following RN and PN, separately. Predictive ability was summarized using marginal $R_{\text{GLMM}}^2$, which ranges from 0 to 1, with higher values indicating increased predictive ability.

Results and limitations

The analysis included 1152 patients (13 206 eGFR observations) who underwent RN and 1920 patients (18 652 eGFR observations) who underwent PN, with mean preoperative eGFRs of 66 ml/min/1.73 m² (standard deviation SD] = 18) and 72 ml/min/1.73 m² (SD = 20), respectively. The model to predict eGFR after RN included age, diabetes, preoperative eGFR, preoperative proteinuria, tumor size, time from surgery, and an interaction between time from surgery and age (marginal $R_{\text{GLMM}}^2=0.41$). The model to predict eGFR after PN included age, presence of a solitary kidney, diabetes, hypertension, preoperative eGFR, preoperative proteinuria, surgical approach, time from surgery, and interaction terms between time from surgery and age, diabetes, preoperative eGFR, and preoperative proteinuria (marginal $R_{\text{GLMM}}^2$). Limitations include the lack of data on renal tumor complexity and the single-center design; generalizability needs to be confirmed in external cohorts.

Conclusions

We developed preoperative tools to predict renal function outcomes following RN and PN. Pending validation, these tools should be helpful for patient counseling and clinical decision-making.

Patient summary

We developed models to predict kidney function outcomes after partial and radical nephrectomy based on preoperative features. This should help clinicians during patient counseling and decision-making in the management of kidney tumors.