Collection of population-based cancer staging information in Western Australia--a feasibility study.

Abstract

BackgroundRoutine data from cancer registries often lack information on stage of cancer, limiting their use. This study aimed to determine whether or not it is feasible to add cancer staging data to the routine data collections of a population-based Western Australian Cancer Registry (WACR).MethodsFor each of the five most common cancer types (prostate, colorectal, melanoma, breast and lung cancers), 60 cases were selected for staging. For the 15 next most common cancer types, 20 cases were selected. Four sources for collecting staging data were used in the following order: the WACR, the hospital based cancer registries (HBCRs), hospital medical records, and letters to treating doctors. If the case was unable to be fully staged, due to lack of information on regional lymph node invasion or distant metastases, we made the following assumptions. Cases which had data available for tumour (T) and regional lymph nodes (N), but no assessment of distant metastasis (MX) were assumed to have no distant metastases (M0). Cases which had data for T and M, but no assessment of regional nodal involvement (NX) were assumed to have no regional nodal involvement (N0).ResultsThe main focus of this project was the process of collecting staging data, and not the outcomes. For ovary, cervix and uterus cancers the existence of a HBCR increased the stageable proportion of cases so that staging data for these cancers could be incorporated into the WACR immediately. Breast and colorectal cancer could also be staged with adequate completeness if it were assumed that MX = M0. Similarly, melanoma and prostate cancer could be staged adequately if it were assumed that NX = N0 and MX = M0. Some cases of stomach, lung, pancreas, thyroid, testis and kidney cancers could be staged, but additional clinical input – on pathology request forms, for example – would be required to achieve useable levels of completeness. For the remaining cancer types either staging is widely regarded as not relevant, and no generally-accepted system exists, or an acceptable level of completeness is not achievable.ConclusionAdding stage to routinely collected information in a cancer registry is possible for many cancer types, particularly if the assumptions regarding missing data are found to be acceptable or if the guidelines for MX = M0 asumptions are clarified. These findings should be generalizable to most cancer registries in developed countries, if hospital-based cancer registries or other specialized databases are accessible.