Abstract
BACKGROUND: Response to induction chemotherapy has been shown to predict outcome in patients with high-risk neuroblastoma (HR-NB), with those achieving a complete response (CR) having superior outcomes. METHODS: We evaluated whether conventional prognostic factors remain prognostic in subsets of patients defined by response to induction. 1244 Patients from four COG high-risk trials were included. End-induction response was coded as CR, partial response (PR) or better, less than PR without progressive disease (PD), and PD. Cox regression models were performed to calculate event-free and overall survival (EFS, OS) hazard ratios, including interaction terms between induction response and prognostic factors including sex, age, stage, primary tumor location, LDH, ferritin, ploidy, MYCN status, ALK status, histology, MKI, grade, and study era. RESULTS: Among patients who achieved a CR after induction, INSS stage 4 disease and trial era were the only factors that remained significantly associated with inferior OS. For those who achieved less than a PR, adrenal primary site, MYCN amplification, and 1p LOH were associated with inferior outcomes. Multivariable models showed that end-induction response remained prognostic of EFS and OS even after controlling for other factors. Multiple significant statistical interactions were observed between end-induction response and other prognostic factors. CONCLUSION: The impact of conventional prognostic factors is not static in patients with HR-NB. Instead, response to induction chemotherapy modifies the effect of conventional prognostic factors. These data can help to further refine prognosis for patients with variable responses to induction and help to identify candidates who might benefit from treatment other than standard post-induction therapy.