Estimating the lifetime risks of cancer: the best measure depends on your purpose

Estimates adjusted for competing risks of mortality can be more meaningful, but make some comparisons more difficult

The Australian Institute of Health and Welfare began collecting cancer incidence statistics at the population level during the mid‐1980s, based on state and territory registry data that are now collated by the Australian Cancer Database, supplemented by mortality data from the Australian Bureau of Statistics.1 These data inform Australian policy makers, researchers, and the general public about cancer profiles and trends.1

Population cancer statistics are important for broad population health surveillance and for providing key information about historical and emerging trends in socio‐demographic differences in cancer incidence and mortality at the global, national, and local levels. Comparisons of cancer incidence and outcomes in Australia with overseas data are enabled by reports of the International Association of Cancer Registries (www.iacr.com.fr) and the International Agency for Research on Cancer (https://www.iarc.fr). Australian registries are ranked by these agencies in the highest category for quality of data, but work is still required in many countries to improve population‐based cancer incidence and mortality data, especially in countries of low economic development.2,3

Accounting for competing risks in estimates of cancer incidence and mortality was introduced in the research literature in the 1980s and is now facilitated by widely available statistical software, such as the DevCan package of the United States National Cancer Institute (https://surveillance.cancer.gov/devcan). Competing risk analysis accounts for the risk of mortality from causes other than the index cancer, reflecting the reality that the risk of developing and dying from a cancer is affected by mortality from other causes. Competing risk estimates are particularly relevant for older populations, in which the risks of death from competing causes are high.4

Conventional cumulative lifetime risk analyses estimate incidence and mortality risk with the unlikely assumption that only death attributable to the index disease is possible. They do not estimate cancer risks in real world circumstances, under which risk can be reduced by intervening deaths from other causes. However, this approach has the advantage that it detects changes and other differences in cancer risk that result from cancer‐specific interventions or other causes, as the estimates are not influenced by variations in competing mortality.

In real world settings, competing mortality reduces the specific risks of cancer and cancer‐related death. In this issue of the MJA, Bach, Lo and colleagues5 report expectedly lower risk estimates for cancer and cancer‐related death in Australia after adjusting for intervening deaths from other causes. Comparisons are reported for the most frequent cancers: those of the lung, bowel, breast, prostate, and skin (melanoma). Differences between conventional and competing risk estimates were found to be larger for women than men, and were smallest for melanoma of the skin. The general trend to smaller differences between conventional and competing risk‐adjusted estimates reported by Bach, Lo and their co‐authors reflects the declining risks over time of death from competing causes, particularly from cardiovascular disease.

When citing statistics for health promotion or public awareness purposes, care should be taken to ensure that the information used is fit for purpose and that messaging is precise. For informing the public about the risks of developing and dying from cancer, lifetime risks adjusted for competing mortality are generally more relevant, although the differences in estimates may not be large. Competing risk estimates, like conventional cumulative risk, are based on cross‐sectional data that do not reflect future developments, so both types of estimates should be used with caution in risk messaging.6

Both conventional cumulative risks and rates, and those adjusted for competing mortality risks, have value. Each estimate can be appropriate, depending on the purpose for which it is used, but the distinction between them needs careful explanation.

For global comparisons, and to retain an international context and point of comparison for local socio‐demographic and temporal trends, it is important, on the other hand, to retain the methods employed by the International Association of Cancer Registries and the International Agency for Research on Cancer.