Rubin et al1 make an interesting contribution to the complex issue of the role of primary care in improving cancer outcomes. However, they refer to survival rates from diagnosis as the benchmark of improvements in care. Unfortunately, survival rates from diagnosis are a relatively poor indicator of the efficacy of treatment as they obscure two major biases: (1) the lead-time bias; and (2) the over-diagnosis bias. Lead-time bias results in an apparent improvement in survival rates by diagnosing disease earlier but without affecting mortality. The over-diagnosis bias is the discovery of non-progressive disease, for example, many cases of prostate and breast cancer. Identification of non-progressive disease is highly likely to improve apparent outcomes as it means the disease that never would have caused death is included in outcome data and, therefore, results in a falsely favourable impression of the effect of intervention.
Mortality rates are a far better indicator of treatment effectiveness for cancer.2 It is generally not understood that there is a lack of correlation between 5-year survival rates and mortality rates due to the operation of the biases mentioned above.3 If we are going to compare outcomes of cancer treatment it is essential that we use measures that are replicable between healthcare systems: mortality rates achieve this, survival rates do not.
- © British Journal of General Practice, January 2011