Two-protein signature of novel serological markers apolipoprotein-A2 and serum amyloid alpha predicts prognosis in patients with metastatic renal cell cancer and improves the currently used prognostic survival models.

BACKGROUND

In metastatic renal cell cancer (mRCC), the Memorial Sloan-Kettering Cancer Center (MSKCC) risk model is widely used for clinical trial design and patient management. To improve prognostication, we applied proteomics to identify novel serological proteins associated with overall survival (OS).

PATIENTS AND METHODS

Sera from 114 mRCC patients were screened by surface-enhanced laser desorption ionization time-of-flight mass spectrometry (SELDI-TOF MS). Identified proteins were related to OS. Three proteins were subsequently validated with enzyme-linked immunosorbent assays and immunoturbidimetry. Prognostic models were statistically bootstrapped to correct for overestimation.

RESULTS

SELDI-TOF MS detected 10 proteins associated with OS. Of these, apolipoprotein A2 (ApoA2), serum amyloid alpha (SAA) and transthyretin were validated for their association with OS (P = 5.5 x 10(-9), P = 1.1 x 10(-7) and P = 0.0004, respectively). Combining ApoA2 and SAA yielded a prognostic two-protein signature [Akaike's Information Criteria (AIC) = 732, P = 5.2 x 10(-7)]. Including previously identified prognostic factors, multivariable Cox regression analysis revealed ApoA2, SAA, lactate dehydrogenase, performance status and number of metastasis sites as independent factors for survival. Using these five factors, categorization of patients into three risk groups generated a novel protein-based model predicting patient prognosis (AIC = 713, P = 4.3 x 10(-11)) more robustly than the MSKCC model (AIC = 729, P = 1.3 x 10(-7)). Applying this protein-based model instead of the MSKCC model would have changed the risk group in 38% of the patients.

CONCLUSIONS

Proteomics and subsequent validation yielded two novel prognostic markers and survival models which improved prediction of OS in mRCC patients over commonly used risk models. Implementation of these models has the potential to improve current risk stratification, although prospective validation will still be necessary.