Age, tumor grade, and mitosis-karyorrhexis index (MKI) are independently predictive of outcome in neuroblastoma (NB)

Abstract

9558 Background: NB is a rare pediatric malignancy (USA incidence 1:100,000), and treatment is tailored according to risk. NB patients (pts) in the Children's Oncology Group are stratified to low, intermediate or high risk according to age at diagnosis, stage of disease, MYCN status, histopathology, and tumor cell ploidy. The International NB Pathologic Classification (INPC) uses age at diagnosis to classify tumors as Favorable or Unfavorable histology. This results in duplication of the prognostic contribution of age when both age and INPC, with other factors, are used to assign risk groups or build multivariable models of prognostic factors. To eliminate the confounding contribution of age and INPC, and determine if tumor pathology is predictive of outcome independent of age, we performed multivariable modeling using the underlying pathologic components of INPC. Methods: Using the largest cohort of NB pts ever assembled (n=11,054; 1980–2002; international), 1,860 pts with known age, INPC, diagnosis, tumor grade of differentiation, MKI, and outcome were identified. Half were selected at random (reserving the other half for validation) for analysis. A Cox multivariable model was used to perform survival tree regression to identify factors statistically significantly (p<0.05) prognostic of event-free (relapse, progression, secondary malignancy, death) survival (EFS). Factors tested were age, diagnosis, grade, and MKI. Results: Age (<547 v. ≥547 days) was the most significant factor (p<0.0001). Within pts <547 days, no factors were significantly prognostic. Within pts ≥547 days, stroma-poor NB and nodular (composite) ganglioNB was associated with significantly lower EFS than intermixed ganglioNB and ganglioneuroma, maturing (p<0.0001). Pts were further stratified within the latter diagnoses by MKI (Low/Intermediate v. High) (p=0.0016). These results were validated in the other half of the cohort. Conclusions: Histologic features of NB tumors are predictive of outcome. To remove confounding of the prognostic contribution of age, the underlying histologic features of the tumor (diagnosis, MKI, grade) should be used instead of INPC class to assign pts to risk groups or identify prognostic factors with multivariable models. No significant financial relationships to disclose.