Purpose: Adrenocortical tumors, especially cancers, remain challenging both for their diagnosis and prognosis assessment. The aim of this article is to identify molecular predictors of malignancy and of survival.
Patients and methods: One hundred fifty-three unilateral adrenocortical tumors were studied by microarray (n = 92) or reverse transcription quantitative polymerase chain reaction (n = 148). A two-gene predictor of malignancy was built using the disease-free survival as the end point in a training cohort (n = 47), then validated in an independent validation cohort (n = 104). The best candidate genes were selected using Cox models, and the best gene combination was validated using the log-rank test. Similarly, for malignant tumors, a two-gene predictor of survival was built using the overall survival as the end point in a training cohort (n = 23), then tested in an independent validation cohort (n = 35).
Results: Unsupervised clustering analysis discriminated robustly the malignant and benign tumors, and identified two groups of malignant tumors with very different outcome. The combined expression of DLG7 and PINK1 was the best predictor of disease-free survival (log-rank P approximately 10(-12)), could overcome the uncertainties of intermediate pathological Weiss scores, and remained significant after adjustment to the Weiss score (P < 1.3 x 10(-2)). Among the malignant tumors, the combined expression of BUB1B and PINK1 was the best predictor of overall survival (P < 2 x 10(-6)), and remained significant after adjusting for MacFarlane staging (P < .005).
Conclusion: Gene expression analysis unravels two distinct groups of adrenocortical carcinomas. The molecular predictors of malignancy and of survival are reliable and provide valuable independent information in addition to pathology and tumor staging. These original tools should provide important improvements for adrenal tumors management.