A rapid, genome-agnostic approach to personalized systemic therapy for patients with adrenocortical carcinoma

Background Adrenocortical carcinoma (ACC) is a rare, aggressive malignancy with 5-year survival rates less than 15% with standard approaches to systemic therapy. Personalized medicine using next generation sequencing has not been applicable to the vast majority of patients given limited actionable mutations present in this population. We therefore sought to evaluate a rapid, genome-agnostic approach to personalized systemic therapy for patients with ACC. Methods We retrospectively analyzed patients with metastatic ACC who underwent surgical intervention between June 2022 and August 2025. Tumor samples were evaluated ex vivo using cell mass-based response testing, and predicted treatment sensitivities were compared with clinical outcomes based on RECIST v1.1 criteria. Patients were stratified according to whether they received therapies predicted to elicit a response. A non-homogeneous Markov multistate model with transition-specific Cox proportional hazards modeling was used to assess the effects of receiving a therapy predicted to be efficacious and surgery across disease states. Results Fifteen patients, encompassing 188 within-patient transition intervals and 73 clinical events, were included. Patients treated with chemotherapy agents predicted to elicit a response had a longer median time to progression than those receiving therapies predicted to be ineffective (9.6 vs. 4.8 months; HR 0.437, 95% CI 0.157–1.212; p = 0.10), although the difference was not statistically significant. The effect of predicted-response therapy varied by disease state; however, the extreme estimate after progression (HR < 0.001; p < 0.001) should be considered exploratory. Conclusions In this small retrospective cohort of patients with metastatic ACC, receiving therapy predicted to elicit a response was associated with a longer time to progression, but the difference was not statistically significant. The limited number of patients, repeated observations, potential confounding, and sparse events make these findings hypothesis-generating. These findings support further prospective validation in larger cohorts.