Technical notes
How prevalence has been estimated
State and territory cancer registries provide cancer registration data to the AIHW, which then compiles the data into the Australian Cancer Database (ACD).
The ACD contains de-identified details of every cancer diagnosis in Australia since 1982. The ACD has very high-quality national data on all diagnosed cases of breast cancer, but (except for Victoria and Tasmania) it has no information on stage at primary diagnosis. Stage refers to how advanced the cancer was when first diagnosed.
While all cases of cancer are routinely notified to and recorded by cancer registries and compiled nationally into the Australian Cancer Database (ACD), the same is not true for progression of cancers to metastatic stage. Therefore, it is not possible to directly count cases that become metastatic. This inability goes to the purpose of the ACDA and the development work they are supporting.
Recently, the ACD, MBS and PBS data were linked to form the first version of the national de-identified cancer and treatment analysis asset (CaT-Link). As well as having many other uses, CaT-Link provides the means to count people diagnosed with breast cancer who then commence MBS or PBS subsidised treatment specific to metastatic disease.
Establishment of CaT-Link coincided with calls by BCNA to count the number of women living with MBC, the formation of the ACDA to develop national cancer data capacity, and CINSW development of methods to count cases of MBC using linked cancer registry and treatment data.
Having access to their own state-specific linked data (e.g. state hospital inpatient data and radiotherapy data), CINSW, CCV and CAQ were then able to use these methods to count MBC cases in their own states.
As hospitals, radiotherapy and other data are not included in CaT-Link, AIHW counts of MBC cases treated with MBS or PBS were adjusted upwards based on these NSW, Victorian and Queensland counts. This process required the development of new methods and the use of a number of assumptions (see Box 2 below).
These national estimates were then compared against estimates calculated using another method (MIAMOD), which uses breast cancer incidence and mortality data to estimate prevalence.
For a detailed explanation of the methods involved, and the specific datasets used, as well as the assumptions and limitations, see the methods paper.
Box 2: Summary of assumptions and limitations
The method requires scaling up of national counts of cases that had received MBS and PBS subsidised treatments, using details of the state-specific counts produced by CINSW, CCV and CAQ that are more complete, so as to estimate the number of Australian women and men living with MBC.
Methods assume that treatment of metastatic breast cancer is broadly similar across Australia.
A number of cases will be missed because some treatments for MBC can also be used for less advanced stages of breast cancer – i.e. they are not specific only to MBC and therefore cannot be used to count cases. This approach was also used by CINSW, CCV and CAQ; it aims to minimise over counting.
Data available to CINSW, CCV, CAQ and AIHW differed, affecting completeness of counts:
- CINSW, CCV, CAQ mostly had access to state-specific collections covering different time periods. Consequently, methods and counts differed between the three states.
- CCV and CAQ had, respectively, limited and no access to MBS and PBS data in their linkages, with missing counts estimated by AIHW using CaT-Link data.
- Because AIHW does not have hospital or radiotherapy data in CaT-Link, it used the aggregated counts by CINSW, CCV and CAQ to scale up the national estimates and take account of this missing data.
AIHW estimates for NSW, Victoria and Queensland will differ from the counts previously produced by CINSW, CCV and CAQ, reflecting differences in the available data and completeness of counts, the assumptions used to model national estimates, or interstate differences in treatment (if any).
Until more complete national linked cancer and treatment data become available, these estimates should be considered preliminary experimental estimates and interpreted cautiously.
For a detailed explanation of the methods involved, and the assumptions and limitations, see the methods paper.