Measures of Cancer Survival

Three measures of cancer survival can be calculated in SEER*Stat software:

  • Observed all cause survival - Observed survival is an estimate of the probability of surviving all causes of death.
  • Net cancer-specific survival (policy-based statistic) - This is the probability of surviving cancer in the absence of other causes of death. It is a measure that is not influenced by changes in mortality from other causes and, therefore, provides a useful measure for tracking survival across time, and comparisons between racial/ethnic groups or between registries.
  • Crude probability of death (patient prognosis measure) - This is the probability of dying of cancer in the presence of other causes of death. It is a better measure to assess the impact of cancer diagnosis at an individual level since mortality from other causes play a key role. It measures mortality patterns actually experienced in a cohort of cancer patients on which many possible causes of death are acting simultaneously. The crude measure is reported as a cumulative probability of death from cancer rather than survival.

The SEER*Stat help system includes several frequently asked questions to clarify when net survival and crude probability of death would be used.

Approaches to Estimation of Cancer-Specific Survival

Relationship of the Survival Measure to the Estimation Method
Table shows the relationship of the survival measure (net or crude) to the estimation method (cause of death or expected survival)

Net cancer-specific survival and crude probability of death have two methods in which they can be estimated: using cause of death information or expected survival tables. When using cause of death information, there has been much debate over what is the right endpoint. If death certification were perfect, one would just use the specific form of cancer as the endpoint. However, if a cancer metastasizes, there are instances where the death certificate incorrectly lists the underlying cause of death as the metastatic site. In this instance, it may be best to use all cancers as the end point, especially if the patient only has one cancer. Work is ongoing to define more sophisticated algorithms for defining endpoints based on common sites of metastases for each cancer.

Regardless of whether one uses an approach which utilizes cause of death or expected lifetables, careful consideration should be given to exclusions from the analysis. A technical report from Boer et al., 2003 (PDF, 879 KB), summarizes various approaches to exclusions for survival analyses, as well as the choice of endpoints when death certificate information is utilized. The figure above illustrates the survival statistics that result from the combination of the two measures and twoestimation methods. A description of each is given below.

  • Relative survival - Cancer survival in the absence of other causes of death is calculated using survival life tables. Relative survival is defined as the ratio of the proportion of observed survivors (all causes of death) in a cohort of cancer patients to the proportion of expected survivors in a comparable cohort of cancer-free individuals. The formulation is based on the assumption of independent competing causes of death. Since a cohort of cancer-free individuals is difficult to obtain, we use expected life tables and assume that the cancer deaths are a negligible proportion of all deaths. (See the technical report from Cho et al., 2011 (PDF, 638 KB))
  • Cause-specific survival is a net survival measure representing survival of a specified cause of death in the absence of other causes of death. Estimates are calculated by specifying the cause of death. Individuals who die of causes other than those specified are considered to be censored. (See Marubini & Valsecchi, 1995)
  • Crude Probability of Death Using Expected Survival - This crude measure uses expected survival (obtained from the expected life tables) to estimate the probability of dying from other causes in each interval. Since a cohort of cancer-free individuals is difficult to obtain, we use expected life tables and assume that the cancer deaths are a negligible proportion of all deaths. (See Cronin & Feuer, 2000)
  • Crude Probability of Death Using Cause of Death Information - The probability of dying from cancer and dying from other causes in a cohort of cancer patients is calculated using cause of death information. (See Marubini & Valsecchi, 1995; Schairer et al., 2004)

Example: This figure shows crude and net probability of death from localized colorectal cancer for men and women diagnosed over the age of 70. Crude probability of death (cancer) is lower than net probability of death because localized colorectal cancer has good prognosis, and because mortality for other causes is high for that age group.

Cumulative Probability of Death in Men and Women Age 70+ Diagnosed with Localized Colorectal Cancer, 1996-2012, SEER 13 Registries
Cumulative Probability of Death in Men and Women Age 70+ Diagnosed with Localized Colorectal Cancer, 1996-2012, SEER 13 Registries Cumulative Probability of Death in Men and Women Age 70+ Diagnosed with Localized Colorectal Cancer, 1996-2012, SEER 13 Registries

The SEER*Stat matrix file used to obtain the percentages for the Cumulative Probability of Death figure shown above is available for download. You must have the SEER*Stat software in order to open this file - crude.vs.net.ssm.

SEER*Stat Features in Survival Analysis

Presentation from SEER*Stat Technical Webinar (July 14, 2011) (PDF, 909 KB)

These are slides from a webinar highlighting two features in the calculation of cancer survival added to SEER*Stat version 7.0:

  • Ederer II method to estimate expected survival in relative survival
  • Improved algorithm to specify the underlying cause of death: the SEER cause-specific death classification variable
Last Updated: 17 Mar, 2021