An estimated 79,000 new cases of renal cancers will be diagnosed in the United States this year.¹ Most of these patients present with nonmetastatic disease, and many will undergo surgery. While pathological staging is closely associated with the risk of cancer progression and death following surgery, it can often paint an incomplete picture of a patient’s true individual risk when used in isolation. Several other clinicopathological factors have been consistently shown to be associated with oncologic outcomes, including tumor size, histology, grade, tumor necrosis, nodal involvement, rhabdoid or sarcomatoid features and symptoms at time of diagnosis, among others.

A variety of cancer nomograms and statistical models have been developed and utilized over the years to aid in predicting oncologic outcomes of patients following surgery for renal cell carcinoma (RCC). These have largely been based on, and validated using, retrospective clinical data collected prior to the current era of systemic therapy options. In 2019, Correa et al published a study testing the performance of 8 commonly used RCC risk models (UISS, SSIGN, Leibovich, Kattan, MSKCC, Yaycioglu, Karakiewicz and Cindolo) using data from the ASSURE prospective adjuvant phase III clinical trial (ECOG-ACRIN E2805). The authors observed that in this population all these models tended to underperform in terms of their predictive ability compared to their originally reported retrospective validations, with c-index ranging from 0.556 for UISS to 0.688 for SSIGN.²

Given these limitations, the same authors subsequently used the data from the ASSURE trial to develop a new prognostic model for disease-free survival (DFS), overall survival (OS) and early disease progression. They grouped patients into low-, intermediate- or high-risk categories based on a scoring system. The variables included were histology, tumor size, grade, necrosis and nodal involvement. Their global c-indices for DFS, OS and early disease progression were 0.680, 0.686 and 0.694, respectively. The patients at highest risk were those patients with clear cell, papillary type II or variant histology. In addition, patients with vascular invasion, tumor size >10 cm, Fuhrman grade 4, necrosis and positive regional lymph nodes were also at very high risk. Patients in the highest risk group had 33.1% DFS at 5 years and 49% OS at 5 years.³

Khene et al attempted to externally validate the new ASSURE prognostic model in a “real life” setting outside of a clinical trial in a study that included 1,372 patients from 10 centers and concluded that this model had moderate c-index (0.65 for DFS and 0.63 for OS) and performed similarly to existing nomograms in this setting.⁴

A recently published systematic review compared the performance of many of these prognostic models to each other and concluded that while some models may perform better for certain outcomes (recurrence-free survival, OS or cancer specific survival), there is no single best model. Utility and practicality in daily clinical practice may be based on what clinical and pathological data the individual clinicians have available to them.⁵ European kidney cancer guidelines from the European Society for Medical Oncology and European Association of Urology are consistent with these findings and, while recommending the use of prognostic models, do not specify which one is best used.

Figure. Validation data on PFS by AUA risk category. Figure used with permission of the Society of Urologic Oncology.⁷

In 2021, the AUA published updated guidelines on the evaluation, management and followup of patients with localized renal cancer.⁶ The guidelines recommend that clinicians stratify patients into one of 4 newly developed risk categories following surgery based on pathological stage, grade and margin status to help guide the surveillance schedule following partial and radical nephrectomy. These categories were meant to be both parsimonious and simple to use in daily clinical practice, while providing enough risk discrimination to be clinically relevant. The guidelines cite data listing approximate rates of disease recurrence by risk category.⁶ The table summarizes the findings.

These new risk categories were subsequently validated by Zganjar et al using data from Mayo Clinic’s prospective nephrectomy registry. The authors limited the analysis to clear cell and papillary histology. The new AUA risk stratification system performed well and demonstrated a robust c-index for progression-free survival (PFS; 0.775–0.780) and cancer specific survival (0.811–0.830) among patients with clear cell RCC and papillary RCC after surgery (see figure). Patients in the very high risk category had a 5-year PFS of 19%–40%.⁷ The authors also compared the performance of their institutional risk model (Leibovich score) to the AUA risk stratification system and observed that among patients with clear cell RCC, the 2018 Mayo model performed better (although additional variables are needed for the Leibovich score).

As the list of systemic therapy options for patients with RCC continues to grow and the multidisciplinary approach for managing these patients becomes increasingly individualized, accurate risk stratification is paramount for patient counseling, recommendations for adjuvant therapies and inclusion in clinical trials. Identifying those patients at highest risk of progression and death helps us decide the most appropriate candidate to consider for adjuvant therapy. Accurately defining which patients are at highest risk will help us design the next generation of clinical trials that maximize potential benefit to these patients while avoiding adverse events for patients who have low risk of recurrence.

There are common themes that unite most of these nomograms and push patients into the lowest risk categories such as low stage, low grade and negative margins, and into the highest risk categories such as high stage, sarcomatoid features, positive nodes and positive margins. While there are many risk stratification tools to predict outcomes for our patients with kidney cancer, using ones similar to the AUA risk groups that are both accurate and simple enough to use in a busy clinical practice is paramount.