Official Statistics

Conditional crude probabilities of death, split by cancer and non-cancer causes, for adult cancer patients in England followed up to 2019

Published 27 November 2020

Applies to England

Context

Cancer survival statistics are typically reported using measures, such as net survival, that do not take into account of non-cancer-cause mortality (or other-cause mortality). Net survival provides a measure of cancer survival that is particularly useful for comparisons across population groups where the other-cause mortality may differ. For example, the proportions of deaths due to obesity or cardiovascular diseases may be different across countries. As such, this can help aid policy makers in driving change for improvement in cancer healthcare.

In reality, however, cancer patients are always at risk of dying from other causes. Crude probabilities are a statistical measure that provide the all-cause probability of death (or mortality), partitioned into the probability of death due to cancer and the probability of death due to other causes. These statistical measures are commonly presented from diagnosis of cancer up until death, otherwise known as unconditional estimates.

It is also possible to present conditional crude probability estimates for patients based on the condition that they have survived for some time following their diagnosis, say 1 or 3 years post-diagnosis. As an example, the ‘5-year crude probability of death conditional on surviving 3 years’ refers to the mortality of patients at 5-years post-cancer diagnosis but have survived for at least 3 years. These conditional measures are much more informative for patients by providing updated estimates of their prognosis.

In general, it is useful to have both net survival and crude probability measures since they are informative for different purposes. This bulletin focuses on crude probabilities, showing the cancer and other-cause probability of death experienced by patients diagnosed with cancer between 2014 and 2018.

Variation in crude probabilities of death by cancer site

Figure 1 shows that 5-year unconditional all-cause probability of death for all ages combined varies between different cancer sites. The cancer site with the highest 5-year unconditional all-cause probability of death is mesothelioma at 96.0%. The cancer site with the lowest 5-year unconditional all-cause probability of death is thyroid at 13.1%.

Most cancer sites have a higher unconditional crude probability of death due to cancer compared to other causes. This shows that for most cancer sites the probability of death is more likely to be cancer. However, for melanoma and prostate cancer, the probability of death due to other causes is higher than cancer, which means for these patients, they are more likely to die of other causes than their cancer.

Figure 1: Site-specific crude probabilities of death split into cancer and other-causes for all ages at 5-years post-diagnosis, in patients diagnosed with cancer between 2014 to 2018.

Notes:

1. Crude probability of death is presented for persons for all sites except for:
   a. sex-specific cancers (cervix, ovary, uterus, and vulva for females and prostate for males); b. crude probability of death for cancer of the larynx is presented only for males, and for breast cancer is presented only for females since these cancers are relatively uncommon in the opposite sex.

5-year unconditional crude probabilities of death diminish in older patients

In younger age groups, cancer accounts for most of the deaths that these patients experience. For lung cancer patients aged 15 to 44 at diagnosis, 63.2% died within 5-years of diagnosis; 62.9% was due to their cancer diagnosis and the remaining 0.3% from other causes. Even for cancers with a generally good prognosis, a similar pattern occurs. Figure 2 shows that for female breast cancer patients aged 15 to 44 at diagnosis, 11.8% died within 5-years of diagnosis; 11.3% was due to their cancer diagnosis and 0.5% from other causes.

In older age groups, cancer still makes up most of the death experienced by patients but there is much wider variation by cancer site. Older patients diagnosed with cancers that typically have a poor prognosis also die mainly from their cancer. For pancreatic cancer patients aged 75 to 99 at diagnosis, 97.7% of patients died within 5 years of their diagnosis; 93.7% was due to their cancer diagnosis and 4.0% from other causes. For patients diagnosed with cancers that typically have a good prognosis, older patients are more likely to die from causes other than their cancer. Figure 2 shows that for female breast cancer patients aged 75 to 99 at diagnosis, 49.4% died within 5 years of diagnosis; 21.7% was due to their cancer diagnosis and 27.7% from other causes.

Figure 2: Cumulative crude probabilities of death due to cancer and other causes for breast cancer.

Prognosis changes over time

Most crude probability statistics measure prognosis from the time of diagnosis, called unconditional crude probabilities. Conditional crude probabilities measure prognosis at some time after cancer diagnosis conditional on survival up to that time. In the cancers analysed, the 5-year crude probability of death, having survived 3 years after diagnosis, is lower than the 5-year crude probability of death for patients who have just been diagnosed with cancer.

Figure 3 illustrates unconditional and conditional crude probabilities of death by age group for prostate cancer. For patients aged 85 to 99, the all-cause unconditional 5-year crude probability of death is 80.4%, while the 5-year crude probability of death conditional on surviving 3 years is 49.5%. A similar improvement in prognosis can be seen for the other 4 age groups and for all ages combined.

For the same cohort of patients, the all-cause unconditional 5-year crude probability of death (80.4%) was partitioned into 38.7% for prostate cancer deaths and 41.7% for other-cause deaths. The all-cause 5-year crude probability of death conditional on surviving 3 years (49.5%) was partitioned into 19.1% for prostate cancer deaths and 30.4% for other-cause deaths. This means that the risk of death from prostate cancer decreases by 2-fold at 5-years after diagnosis for patients who have survived at least 3 years compared to those who have just been diagnosed.

Figure 3: Cumulative crude probabilities of death due to cancer deaths and other causes for prostate cancer in those aged 85 to 99.

Full sized charts of unconditional, conditional on surviving 1 year and conditional on surviving 3 years, crude probabilities of death by age group, for all 26 cancer sites, can be found in the appendix. The data demonstrates the varying levels of improvements in prognosis, by cancer site, as mortality decreases over time.

Interpretation of these statistics

Cancer in adults is defined using the International Statistical Classification of Diseases 10th Revision (ICD-10) and by morphology and behaviour codes in the International Classification of Diseases for Oncology, Second Edition (ICD-O-2).

Adult cancer mortality estimates are based on conditional crude probabilities, which are calculated using the methods presented by Cronin and Feuer, in the complete approach, and conditional on surviving 1 and 3 years from diagnosis.

Confidence intervals (at the 95% level) are included in the data tables, which are a measure of the statistical precision of a probability and show the range of uncertainty around the calculated estimate.

A robustness criterion is applied to the estimates to ensure high-quality results. The whole series of mortality estimates were suppressed for each cancer site, age group, sex-specific combination if the cumulative mortality estimates were not monotonically increasing at individual yearly time points.

Quality and Methodology

The Cancer survival Quality and Methodology Information (QMI) and Quality assurance of administrative data used in cancer registrations and cancer survival statistics reports contain important information on:

• the strengths and limitations of the data and how it compares with related data
• uses and users of the data
• how the output was created
• the quality of the output including the accuracy of the data

Cancer registrations in England can take up to 5 years after the end of a given calendar year to reach 100% completeness, due to the continuing accrual of late registrations. The back series is not revised to account for late registrations (except in childhood when previous years are estimated annually). Please see the Cancer Registrations Statistics bulletin for more information.

More detailed information on the methods used to estimate conditional crude probabilities in England can be found in Conditional crude probabilities of death for English cancer patients.

Authors and Collaborators

For queries relating to this bulletin, please contact us via email at [email protected].

This publication was produced by Public Health England using the methodology developed in a joint collaboration between Public Health England (PHE) and the University of Leicester (UoL).

Responsible Statistician: Kwok Wong (Public Health England)

Production Team and Collaborators: Kwok Wong (PHE), John Broggio (PHE); Marta Emmett (PHE); Sophie Finnigan (PHE); Thomas Higgins (PHE); Roger Hill (PHE); Dian Xu (PHE); Ann Saxton (PHE); Paul Lambert (UoL); Mark Rutherford (UoL).

Acknowledgements

Data for this work is based on patient-level information collected by the NHS, as part of the care and support of cancer patients. The data is collated, maintained and quality assured by the National Cancer Registration and Analysis Service, which is part of Public Health England (PHE).