Renal medullary carcinoma (RMC) is a rare and highly aggressive kidney malignancy that predominantly affects adolescents and young adults with sickle hemoglobinopathies, particularly sickle cell trait. Characterized by loss of SMARCB1, the disease often presents at an advanced stage and is associated with poor survival outcomes.
Although understanding of RMC biology and treatment has advanced in recent years, predicting survival for individual patients remains challenging. Established prognostic systems, including the International Metastatic Renal Cell Carcinoma Database Consortium (IMDC) score, were developed in metastatic clear-cell renal cell carcinoma and have not been validated specifically for SMARCB1-deficient RMC.
To address this gap, investigators developed RMC-COMPASS, a disease-specific prognostic model designed to estimate overall survival (OS) using clinical and laboratory parameters available at diagnosis.
The study, titled “Development and Internal Validation of the RMC-COMPASS Clinical Prognostic Model for SMARCB1-deficient Renal Medullary Carcinoma,” was published in European Urology Oncology on November 6, 2026.
The research was conducted by Stepan M. Esagian, Peng Yang, Alexander D. Sherry, Rahul A. Sheth, Jose A. Karam, Najat C. Daw, Chad Tang, Nizar M. Tannir, and Pavlos Msaouel.
Study Design and Patient Characteristics
This retrospective, single-institution study included 180 patients with SMARCB1-deficient RMC diagnosed or treated at the University of Texas MD Anderson Cancer Center between January 2003 and October 2025.
The investigators aimed to develop and internally validate a prognostic model for OS, defined as the time from diagnosis to death from any cause. A total of 16 prespecified candidate predictors were evaluated, including clinical characteristics, laboratory measurements, and disease-related factors assessed at diagnosis before systemic treatment.
To develop the model and assess its performance, the investigators used multiple imputation to address missing data and repeated the model-development process across 1,000 bootstrap resamples within 50 imputed datasets. The final model used a log-normal accelerated failure time approach, with correction for estimated overfitting.
The median age at diagnosis was 29 years (interquartile range [IQR] 22–39), and 66% of patients were male. Most patients had sickle cell trait (88%), while 78% presented with stage IV disease. Platinum-based chemotherapy was administered as first-line treatment in 80% of patients. At a median follow-up of 53.4 months, median OS was 15.2 months (95% CI 13.2–18.9), with 142 deaths recorded during the study.
What Did RMC-COMPASS Show?
The final model, named Renal Medullary Carcinoma Clinical and Outcomes Model for Prognostic Assessment and Survival Stratification (RMC-COMPASS), included four routinely available parameters: Eastern Cooperative Oncology Group (ECOG) performance status, number of involved anatomic disease sites beyond the primary tumor, neutrophil-to-lymphocyte ratio (NLR), and corrected serum calcium.
Among the candidate predictors, ECOG performance status and the number of involved disease sites showed the highest selection frequencies during internal validation, appearing in 90% and 70% of bootstrap models, respectively.
After correction for optimism, RMC-COMPASS achieved a concordance index (C-index) of 0.64 (95% CI 0.61–0.71), indicating modest discrimination between patients with different survival outcomes. The optimism-corrected calibration slope was 0.895 (95% CI 0.589–1.20).
When compared with the IMDC score in the same study population, RMC-COMPASS demonstrated higher discrimination, with a C-index of 0.64 versus 0.58 (95% CI 0.54–0.62) for the IMDC score modeled as a linear term.
Calibration was assessed at 12, 18, and 24 months. At 24 months, the apparent integrated calibration index was 0.036, indicating relatively small absolute differences between predicted and observed survival probabilities in the development cohort.
At 24 months, decision curve analysis showed greater apparent net benefit for RMC-COMPASS than treat-all and treat-none strategies across threshold probabilities of death ranging from 40% to 80%. This analysis considered hypothetical decisions about intensified or investigational systemic therapy within a clinical trial and was not corrected for optimism.
Risk Stratification and Additional Analyses
The investigators also evaluated whether RMC-COMPASS could stratify patients into prognostic groups for use in future clinical trials. Based on predicted OS at 24 months, patients were classified into three categories: high risk, with predicted survival below 25% (60 patients); intermediate risk, with predicted survival between 25% and 50% (93 patients); and low risk, with predicted survival above 50% (27 patients).
Observed survival patterns were generally consistent with the model’s predictions across these groups. The investigators proposed that such stratification could help balance prognostic risk between treatment arms in prospective studies.
Additional analyses examined whether incorporating TP53 mutation status could improve the model. Among 85 patients with available TP53 data, adding this variable produced similar results for discrimination, calibration, and risk stratification.
The investigators also evaluated model performance across different diagnosis periods. For internal-external temporal validation, the four-predictor model was refitted in patients diagnosed before 2020 (n = 93) and evaluated without further tuning in those diagnosed from 2020 onward (n = 87).
In the later cohort, the C-index was 0.64 (95% CI 0.54–0.73), with a calibration slope of 0.839 (95% CI 0.351–1.30).
Although these findings suggested reasonably consistent discrimination across the two periods, the uncertainty around calibration remained substantial.
Clinical Implications and Limitations
RMC-COMPASS is the first internally validated prognostic model developed specifically for SMARCB1-deficient RMC. The model uses four routinely available clinical and laboratory parameters measured at diagnosis and does not incorporate treatment-related variables or additional molecular predictors.
The model may be particularly relevant in a disease where most patients present with advanced cancer and conventional staging provides limited information about individual survival.
By generating individualized survival estimates, RMC-COMPASS could support discussions between clinicians and patients about prognosis. Its risk categories may also be useful when designing prospective clinical trials, particularly for eligibility assessment and stratified randomization.
To facilitate its application, the investigators developed a freely accessible RMC-COMPASS web-based calculator, which provides individual survival estimates based on the four model parameters. However, the findings should be interpreted in the context of several limitations.
The retrospective, single-institution design introduces the possibility of referral and selection biases, while the cohort of 180 patients remains relatively small for prognostic model development. Missing clinical and laboratory data required multiple imputation, and some optimism in the reported performance may remain despite internal validation procedures.
Importantly, RMC-COMPASS has not yet been externally validated. Its performance in independent patient populations, including those treated at different institutions and in more recent clinical settings, remains to be established. The model has also not been validated for treatment selection. Therefore, although its estimates may inform prognostic discussions and future research, they should not independently guide therapeutic decisions.
Conclusion
RMC-COMPASS provides a disease-specific approach to estimating survival in patients with SMARCB1-deficient renal medullary carcinoma.
Using four routinely available parameters, the model demonstrated modest discrimination, showed higher discrimination than the IMDC score within the study cohort, and classified patients into three prognostic groups based on predicted 24-month survival. These findings suggest that RMC-COMPASS may provide a useful framework for individualized prognostic counseling and risk stratification in future clinical trials.
Nevertheless, external validation and potential recalibration in independent, multi-institutional cohorts are essential before the model can be considered for routine clinical use.
The full article is available in European Urology Oncology.
Find more videos on OncoDaily YouTube TV.
