PURPOSE: The ratio of cancer mortality and cancer incidence rates in a population has conventionally been used as an indicator of the completeness of cancer registration. More recently, the complement of the mortality-to-incidence ratio (1-M/I) has increasingly been presented as a surrogate for cancer survival. We discuss why this is mistaken in principle and misleading in practice. METHODS: We provide an empirical assessment of the extent to which trends in the 1-M/I ratio reflect trends in cancer survival. We used national cancer incidence, mortality and survival data in England to compare trends in both the 1-M/I ratio and net survival at 1, 5, and 10 years for 19 cancers in men and 20 cancers in women over the 29-year period from 1981 to 2009. RESULTS: The absolute difference between the 1-M/I ratio and 5-year net survival for 2009 was less than 5% for only 12 of the 39 cancer/sex combinations examined. For an additional 12, the 1-M/I ratio differed from 5-year net survival by at least 15%. The comparison is also unstable over time; thus, even when differences were small for 2009, the difference between 5-year net survival and the 1-M/I ratio had changed dramatically for most cancers between 1981 and 2009. CONCLUSION: The 1-M/I ratio lacks any theoretical basis as a proxy for cancer survival. It is not a valid proxy for cancer survival in practice, either, whether at 5 years or at any other time interval since diagnosis. It has none of the useful properties of a population-based survival estimate. It should not be used as a surrogate for cancer survival.