Assessing the feasibility and validity of the Toronto Childhood Cancer Stage Guidelines: a population-based registry study

Background Cancer stage at diagnosis is crucial for assessing global efforts to increase awareness of childhood cancer and improve outcomes. However, consistent information on childhood cancer stage is absent from population cancer registries worldwide. The Toronto Childhood Cancer Stage Guidelines, compiled through an international consensus process, were designed to provide a standard framework for collection of information on stage at diagnosis of childhood cancers. We aimed to assess the feasibility of implementing the Toronto Guidelines within a national population cancer registry. Methods We did a population-based registry study using data from the Australian Childhood Cancer Registry and included data from children aged 0-14 years diagnosed between Jan 1, 2006, and Dec 31, 2010 with one of 16 childhood cancers listed in the Toronto Guidelines (acute lymphoblastic leukaemia, acute myeloid leukaemia, Hodgkin's lymphoma, non-Hodgkin lymphoma, neuroblastoma, Wilms' tumour, rhabdomyosarcoma, non-rhabdomyosarcoma soft tissue sarcoma, osteosarcoma, Ewing's sarcoma, retinoblastoma, hepatoblastoma, testicular cancer, ovarian cancer, medulloblastoma, and ependymoma). We extracted data from medical records, and assigned stage according to the Tier 1 criteria (basic) and Tier 2 criteria (more detailed, requiring data from cytology, imaging, and other diagnostic tests, where available) using computer algorithms derived from the Toronto Guidelines. Additionally, expert reviewers independently assigned Tier 2 stage to a random subsample of 160 cases (ten per malignancy type). Feasibility of the guidelines was assessed on the percentage of cases that could be staged, agreement between stage assigned by the algorithms and the expert reviewers, and the mean time (min) taken to collect the required data. Findings We obtained data for 1412 eligible children. Stage could be assigned according to Tier 2 criteria for 1318 (93%) cases, ranging from 48 (84%) of 57 cases of non-rhabdomyosarcoma soft tissue sarcoma to 46 (100%) cases of hepatoblastoma. According to Tier 1 criteria, stage could be assigned for 1329 (94%) cases, ranging from 131 (87%) of 151 cases of acute myeloid leukaemia to 46 (100%) cases of hepatoblastoma. By contrast, stage at diagnosis was recorded by the treating physician for 555 (39%) of the 1412 cases. The computer algorithm assigned the same stage as did one or more independent expert reviewers in 155 (97%) of the 160 cases assessed. The mean time taken to review medical records and extract the required data was 18.0 min (SD 9.5 per case). Interpretation The Toronto Guidelines provide a highly functional framework that can be used to assign cancer stage at diagnosis using data routinely available in medical records for most childhood cancers. Data on staging have the potential to inform interventions targeting improved diagnosis and survival.