Stage at Diagnosis and International Survival Variation in Childhood Tumors in the BENCHISTA Study

Abstract

Importance: Understanding the reasons for variations in population-level survival differences in childhood cancer is important to guide improvement efforts. Collaboration between population-based cancer registries (CRs) to apply the international consensus Toronto guidelines to record tumor stage at diagnosis is a key first step.

Objective: To test whether survival probabilities by tumor stage vary internationally, using 6 childhood solid tumors as exemplars.

Design, setting, and participants: The International Benchmarking of Childhood Cancer Survival by Stage (BENCHISTA) population-based retrospective cohort study included all incident cases of neuroblastoma, Wilms tumor, medulloblastoma, osteosarcoma, Ewing sarcoma of bone, and rhabdomyosarcoma diagnosed between January 1, 2014, and December 31, 2017, with 3-year follow-up for survival. A total of 73 CRs from 27 countries (23 European as well as Australia, Brazil, Canada, and Japan) constituted the dataset. Analyses were conducted from June 2023 to December 2024.

Main outcomes and measures: Three-year overall survival (OS) by stage for each tumor type, with comparisons between countries grouped into 5 predefined European areas. Multivariable Cox and logistic models estimated each area's hazard or odds ratio of death compared with Central Europe (Austria, Belgium, France, Germany, Switzerland, and the Netherlands), adjusted by age group and stage.

Results: A total of 9883 cases were included; 4452 (45%) were girls and overall median (IQR) age was 54 (22-122) months; stage completeness was 93% (9199 of 9883). Three-year OS rates were as follows: Wilms tumor, 95% (95% CI, 94%-96%); neuroblastoma, 83% (95% CI, 81%-84%); medulloblastoma, 79% (95% CI, 77%-81%); Ewing sarcoma, 78% (95% CI, 75%-80%); rhabdomyosarcoma, 77% (95% CI, 74%-79%); and osteosarcoma, 75% (95% CI, 73%-77%). Geographical variations in age-adjusted OS were found for neuroblastoma, medulloblastoma, Ewing sarcoma, and rhabdomyosarcoma. Following additional adjustment for stage, differences were no longer significant for neuroblastoma (in the UK and Ireland) and rhabdomyosarcoma (in Eastern Europe) while becoming significant for neuroblastoma in Eastern Europe (hazard ratio, 1.36; 95% CI, 1.05-1.76) and medulloblastoma in Southern Europe (hazard ratio, 1.42; 95% CI, 1.03-1.94). However, no mitigation of survival variation was observed for Ewing sarcoma in the UK and Ireland (hazard ratio, 2.06; 95% CI, 1.39-3.04) and Eastern Europe (hazard ratio, 1.87; 95% CI, 1.22-2.86) as well as for medulloblastoma in Eastern Europe (hazard ratio, 1.68; 95% CI, 1.13-2.49).

Conclusions and relevance: In this BENCHISTA cohort study of 6 solid tumors, international variation in population-level OS was associated with differences in tumor stage distribution for some cancer types and regions. Additional factors are suggested for further investigation. The results have important implications for national health systems for monitoring early diagnosis efforts and supporting collaboration between CRs and clinicians to sustain standardized use of Toronto guidelines to improve understanding of survival variation in childhood cancer.

Figure 1.. Three-Year Overall Survival by Toronto Stage for Each Tumor Type and European Geographical Area

Stage definitions are according to the Toronto guidelines.^, Neuroblastoma was staged at tier 1; all other tumors were staged at tier 2. For full definitions of stages by tumor type, refer to eFigure 3 in Supplement 1. For neuroblastoma, L indicates localized; LR, locoregional; M, metastatic; MS, metastatic with specific pattern of metastases in children younger than 18 months. For Wilms tumor, y indicates that the tumor was staged after preoperative chemotherapy. For medulloblastoma, M0 indicates cerebrospinal fluid negative for metastases; M1, tumor cells in cerebrospinal fluid; M2, visible metastasis in brain; and M3, visible metastases in spine or in cervicomedullary junction. For osteosarcoma and Ewing sarcoma, L indicates localized; and M, metastatic.

Figure 2.. Tumor-Specific Hazard Ratios (HRs) or Odds Ratios (ORs) From Models With and Without Stage Adjustment, by Geographical Area

Central Europe was the reference category, and areas with significant differences from Central Europe are shaded. The vertical line at 1 represents the null value for the effect estimates, indicating no difference in the 3-year risk of death compared with the reference category, after adjustment for covariates. Points to the right of this line indicate a higher risk relative to Central Europe, while points to the left indicate a lower risk. HRs are reported for all cancers except osteosarcoma, where the proportional hazards assumption was violated; therefore, ORs estimated using an inverse probability censoring-weighted logistic model are shown. Horizontal bars represent the 95% CI for each estimate.