Partial nephrectomy (PN) represents the gold-standard management of T1a and select T1b renal cell carcinoma (RCC) as it has been noted to maximize postoperative renal function when compared to radical nephrectomy (RN) with similar oncologic outcomes.¹ Historically, local recurrence (LR) following PN has been variably defined using broad definitions that include metachronous de novo tumors and true LR within the PN bed.^2-4 Previously, a group from MD Anderson used a narrow definition to include new, enhancing tumor growth within the prior PN defect.⁵ Our group sought to identify predictors of LR, oncologic outcomes, and the associations of LR with cancer-specific survival and distant metastasis using the narrow definition provided by Wood et al.⁵ Additionally, we created a risk score using readily available preoperative variables which can be used to counsel patients on risk of LR.

Using a prospectively maintained institutional nephrectomy registry, we identified patients undergoing PN for unilateral, sporadic, localized (cN0M0) RCC between 2000 and 2019 with documented LR. Patients with syndromes predisposing to renal malignancy were excluded from our analysis. Cross-sectional imaging (CT and/or MRI) was reviewed, and patients with new, enhancing tumor growth within the PN defect were identified. Separate multivariable models were created for pre- and postoperative factors associated with LR. A scoring algorithm using the HRs from the preoperative model was developed (see Table). Cox proportional HRs were used to test the association of LR with distant metastasis and cancer-specific death, and in these models LR was defined as a time-dependent covariate to evaluate whether the presence of LR was associated with time from PN to distant metastasis and RCC-related death.

Table 1. Multivariable Analysis of Preoperative Features With True Local Recurrence Following Partial Nephrectomy

Figure. Predicted local recurrence (LR)–free survival based on a multivariable scoring model utilizing preoperative factors, including age at surgery >60 years, symptoms at presentation, solitary kidney, open surgical approach, and clinical T staging. Scores 8, 9, and 10 were combined into 1 group due to limited patient numbers. CI indicates confidence interval; PN, partial nephrectomy.

Ultimately, 2,164 adults treated with PN for unilateral, sporadic, localized RCC were identified, and 106 true LRs (4.8%) were diagnosed at a median of 3.3 years (IQR 1.7-5.1) after surgery. Estimated LR rates at 5 and 10 years were 4.1% and 6.2%, respectively. Preoperative factors associated with LR included age ≥60 years, symptoms at presentation, solitary kidney, open surgical approach, and increasing cT stage (see Table). The HRs from the preoperative model were used to create a scoring algorithm to predict the risk of LR following PN (see Figure). Postoperative factors associated with true LR included clear cell histology, pT stage, microscopic positive margins, and Fuhrman grade (see Table). The majority of patients were managed with percutaneous ablation (n=52, 49%) or completion RC (n=32, 30%). We typically reserve repeat PN for syndromic patients; thus, no patients ultimately underwent repeat PN. Multivariable analysis was performed on the subset of patients with clear cell RCC to determine if there was an association between LR and the development of distant metastasis. After adjusting for tumor size, pT classification, and grade, we found that LR was associated with distant metastasis (HR 6.25; 95% CI 3.54-11.0; P < .001). After further adjustment for the presence of distant metastases, LR was significantly associated with death from RCC (HR 1.93; 95% CI 1.08-3.50; P = .03).

Prior studies have suggested LR is rare following PN, with rates typically ranging from 1%-7%.^2-5 However, these reports are limited by heterogeneous, broad definitions of recurrence, and capture a significant number of metachronous de novo tumors which do not represent failure of the index operation. To date, few institutional series have examined LR in the PN bed. The largest such series by MD Anderson reported an LR rate of 2%, including only patients with new, enhancing tumor growth within the PN bed.⁵ Interestingly, our series suggests that with long-term follow-up, about 6% of patients can expect to develop LR after PN. Many of these occur >5 years after the initial operation, highlighting the importance of continued long-term monitoring in this population.

Preoperative factors associated with LR in our series included age ≥60 years, symptoms at presentation, solitary kidney, open surgical approach, and increasing cT stage. Using preoperative factors, we created a risk calculator that will aid clinicians when counseling patients prior to PN. To our knowledge, this represents the first preoperative risk calculator excluding metachronous, ipsilateral recurrences. Postoperative predictors of recurrence were similar to those identified by Wood et al,⁵ including positive microscopic margins, higher pT stage, and solitary kidney. Additional factors associated with LR included clear cell histology and open approach. Typically, open surgical approach is reserved for complex renal masses at our institution, and this likely reflects increasing tumor complexity.

A central goal of our study was to describe the natural history of true LR and associations with metastasis and death from RCC. Ultimately, 23 patients presented with metastasis before or concurrently with LR, likely reflecting metastases from the original tumor that were not identified prior to PN. Of the 83 patients who presented with an isolated LR, 26 (30%) subsequently developed metastatic progression, and half of these developed within 1.5 years of identification of the LR. The remaining half occurred 1.5 years or more following diagnosis of LR, possibly indicating these arose from the LR. When LR was analyzed as a time-dependent covariate on multivariable analysis, we found LR was associated with subsequent metastasis and death from RCC. Given the data showing half of metastases after LR occurred more than 1.5 years following diagnosis of LR, these data support the treatment of LR in the absence of metastatic disease.

This study is not without limitations, most importantly the retrospective, single-institution design. Additionally, our institutional registry does not record tumor complexity data, which have been shown to be associated with LR in the series by MD Anderson.⁵ However, data are conflicting regarding this finding, as other studies have shown nephrometry score is not associated with ipsilateral LR when controlling for tumor size.⁶ Given the limited cohort size, we were unable to compare LR treatment approaches. This will likely require additional multi-institutional series to provide sufficient sample size. Despite the noted limitations, to our knowledge this represents the largest study of patients experiencing LR in the PN bed. These findings provide insight into the natural history of LR following PN, establishing a rationale for treatment of LR to potentially prevent metastasis and death from RCC.