Estimating relative survival for cancer: An analysis of bias introduced by outdated life tables

Relative survival analyses of cancer data often incorporate outdated information about expected survival when current information is not readily available. The assumption is that any bias introduced into the estimation of expected survival, and hence, into the estimate of relative survival, will be negligible. However, empirical studies of potential bias have yet to be published.

Data are from the Canadian Cancer Registry with mortality follow-up through record linkage to the Canadian Vital Statistics Death Database. Period method relative survival ratios (RSRs) for 2005-2007 were derived using life tables centred on the 2006 Census of Population to estimate expected survival. The analysis was repeated using life tables from 5 and 10 years earlier.

Deriving expected survival from life tables 5 years out of date resulted in increases in RSRs for all cancers. These increases became greater with lengthening survival duration. For example, increases in 1-, 5- and 10-year RSRs were 0.2, 0.8 and 1.7 percentage units, respectively, for all cancers combined. Increases in 5-year survival were highest for prostate (2.0) and bladder cancer (1.6); among males (1.2); and among people aged 75 to 99 at diagnosis (1.9). Differences were approximately double when life tables 10 years out of date were used.

The use of historical rather than current expected survival data in calculating RSRs for cancer may lead to consequential overestimation of survival.