Care needed in interpretation of cancer survival measures
2014; Elsevier BV; Volume: 385; Issue: 9974 Linguagem: Inglês
10.1016/s0140-6736(14)62292-3
ISSN1474-547X
Autores Tópico(s)Multiple and Secondary Primary Cancers
ResumoCancer patient survival obtained from population-based cancer studies is the optimum method to monitor and assess the effectiveness of patient care.1Dickman PW Adami HO Interpreting trends in cancer patient survival.J Intern Med. 2006; 260: 103-117Crossref PubMed Scopus (247) Google Scholar Consideration of these estimates in conjunction with estimates of cancer incidence and mortality is still important.2Ellis L Woods LM Estève J Eloranta S Coleman MP Rachet B Cancer incidence, survival and mortality: explaining the concepts.Int J Cancer. 2014; 135: 1774-1782Crossref PubMed Scopus (107) Google Scholar As a result, there is much interest in the assessment of progress in terms of cancer survival.3Coleman M Forman D Bryant H et al.Cancer survival in Australia, Canada, Denmark, Norway, Sweden, and the UK, 1995–2007 (the International Cancer Benchmarking Partnership): an analysis of population-based cancer registry data.Lancet. 2011; 277: 127-138Summary Full Text Full Text PDF Scopus (970) Google Scholar, 4De Angelis R Sant M Coleman MP et al.the EUROCARE-5 Working GroupCancer survival in Europe 1999–2007 by country and age: results of EUROCARE-5—a population-based study.Lancet Oncol. 2014; 15: 23-34Summary Full Text Full Text PDF PubMed Scopus (1494) Google Scholar Quantification of the improvements in cancer patient survival because of successes in some areas, such as treatment, diagnostic techniques, and awareness or screening campaigns, is of paramount importance to health-care officials, health policy makers, and charities supporting these developments in cancer control. However, to make a fair comparison—to compare like with like—is essential to assess progress accurately. In The Lancet, Manuela Quaresma and colleagues5Quaresma M Coleman MP Rachet B 40-year trends in an index of survival for all cancers combined and survival adjusted for age and sex for cancers in England and Wales, 1971–2011: a population-based study.Lancet. 2014; (published online Dec 3.)http://dx.doi.org/10.1016/S0140-6736(14)61396-9PubMed Google Scholar report the creation of a survival index that is useful to compare improvements in cancer patient survival over time. The authors show that the index of net survival increased substantially in cancer patients in England and Wales over the period 1971–2011. To make comparisons across time for all cancer sites combined is difficult because of changing age distributions for patients with cancer alongside a changing distribution of sites. For instance, if less fatal cancers become relatively more common, an improvement would be seen in all-cancer survival, but not necessarily because of any improvements for any individual cancer site. One further issue with comparisons of survival after cancer is that the chance of dying from another cause also changes over time. Quaresma and colleagues5Quaresma M Coleman MP Rachet B 40-year trends in an index of survival for all cancers combined and survival adjusted for age and sex for cancers in England and Wales, 1971–2011: a population-based study.Lancet. 2014; (published online Dec 3.)http://dx.doi.org/10.1016/S0140-6736(14)61396-9PubMed Google Scholar overcome these difficulties by defining a survival index that estimates average relative survival in different time periods, but the index is standardised to the distribution of both age and cancer site in 1996–99. Relative survival compares the survival in patients with cancer to a relevant cohort without cancer, and is important to account for the changing risk of dying of something else. Under specific assumptions,6Perme MP Stare J Estève J On estimation in relative survival.Biometrics. 2012; 68: 113-120Crossref PubMed Scopus (446) Google Scholar average relative survival estimates marginal net survival, which is completely independent of other-cause mortality. The standardisation to 1996–99 neatly overcomes the other issues of comparability (age and cancer distributions), but comes at the cost of the inability to interpret the measures. Measures that are optimum for comparability cannot be easily interpreted in terms of the survival experience of individuals, and vice versa. The authors note that the figures they report are readily interpreted incorrectly.7Borland S Half of cancer patients are living for ten years or more: number classed as having beaten the disease doubles since the 1970s.http://www.dailymail.co.uk/health/article-2615416/half-cancer-patients-living-ten-years-number-classed-having-beaten-disease-doubles-1970sDate: 2014Google Scholar, 8Boseley S Cancer survival rates: half of new UK patients 'can expect to live for another decade'.http://www.theguardian.com/society/2014/apr/29/cancer-patients-uk-survival-ratesDate: April 29, 2014Google Scholar The survival index for all cancers combined is something of a holy grail for headline writers, because it gives one summary figure to show progress in cancer as a whole over time. However, similarly to all averages and summaries, important detail lies behind the number. The survival index (and the site-specific estimates) measure net survival, which adds substantially to the difficulty in interpretation, but that subtlety is often missed. The all-cancer survival index of 50% at 10 years might be interpreted as meaning that 50% of cancer patients will survive for 10 years or more, which is wrong—as the authors discuss5Quaresma M Coleman MP Rachet B 40-year trends in an index of survival for all cancers combined and survival adjusted for age and sex for cancers in England and Wales, 1971–2011: a population-based study.Lancet. 2014; (published online Dec 3.)http://dx.doi.org/10.1016/S0140-6736(14)61396-9PubMed Google Scholar—but it is worth stressing this point. The problem relates to the three main adjustments necessary for making the index possible to compare across time. The measure is independent of competing mortality because of other causes that also heavily affect patients with cancer; for most sites, the diagnosed patients are of an older age. I show an example in which the relative survival estimate at 10 years is 50% at every age. England and Wales population mortality data are used to convert these estimates9Cronin KA Feuer EJ Cumulative cause-specific mortality for cancer patients in the presence of other causes: a crude analogue of relative survival.Stat Med. 2000; 19: 1729-1740Crossref PubMed Scopus (100) Google Scholar, 10Lambert PC Dickman PW Nelson CP Royston P Estimating the crude probability of death due to cancer and other causes using relative survival models.Stat Med. 2010; 29: 885-895Crossref PubMed Scopus (92) Google Scholar into the 10 year all-cause probability of death (table). An 80-year-old has a 16% chance of being alive at 10 years, despite the net estimate of 50%. For younger patients, the differences are smaller, but still exist. Overall, fewer patients with cancer will be alive 10 years after diagnosis if we use all-cause death as the outcome, rather than a net measure. Difficulties are also introduced by standardisation; patients with cancer diagnosed now (and in the future) might be older on average, and have a higher proportion of cancers with a better or worse prognosis than in 1996–99. The survival index is being standardised to a population that might not be a good representation of the present population structure, making present and future interpretability of the measure more difficult. Additionally, any biases in site-specific estimates will also contribute to the survival index.TableSynthetic example of survival by patient age in which 10-year net survival is 50% for all ages10-year net probability of death (1–relative survival)10-year crude (actual) probability of death due to cancer10-year crude (actual) probability of death due to other causes10-year crude (actual) probability of death due to any cause40 years0·500·490·020·5160 years0·500·480·080·5780 years0·500·420·420·84Derived from an England and Wales life table. Columns 2 and 3 give the actual probabilities of death due to the disease and due to other causes, and sum to the value given in column 4 (the overall probability of death). The first column gives a net measure and has no interpretation as the actual probability of death due to cancer. Open table in a new tab Derived from an England and Wales life table. Columns 2 and 3 give the actual probabilities of death due to the disease and due to other causes, and sum to the value given in column 4 (the overall probability of death). The first column gives a net measure and has no interpretation as the actual probability of death due to cancer. Quaresma and colleagues mainly concentrate on age-standardised measures. However, the age-group specific estimates need to be examined across each site to understand fully whether improvements in outcomes occur across all ages simultaneously, which is unlikely to be the case. For instance, certain treatment improvements will be seen only for younger patients. Although the overall measures reported as headline figures are important, understanding of the reasons for the changes over time is even more important. Quaresma and colleagues fit very sophisticated models to arrive at the values that they present. A lot of extra information can be garnered from these models. Metrics that underline in which subgroups of patients and at what point in follow-up improvements have occurred should be further used.11Lambert PC Holmberg L Sandin F et al.Quantifying differences in breast cancer survival between England and Norway.Cancer Epidemiol. 2011; 35: 526-533Summary Full Text Full Text PDF PubMed Scopus (35) Google Scholar, 12Adolfsson J Lambert PC et al.ElorantaHow can we make cancer survival statistics more useful for patients and clinicians: an illustration using localized prostate cancer in Sweden.Cancer Causes Control. 2013; 24: 505-515Crossref PubMed Scopus (41) Google Scholar Overall, Quaresma and colleagues should be congratulated on a carefully constructed method for comparisons across time. This is a useful epidemiological measure for quantification of overall improvements in cancer survival. The site-specific indices are particularly useful to understand whether progress has been made and when in calendar time we observe the improvement. However, caution should be used when interpreting the figures themselves. I would encourage the continued development of statistical approaches that help to pinpoint the reasons for these overall improvements. I would also further stress that other statistical metrics, which are more readily interpretable for patients and health-care professionals, should be used.12Adolfsson J Lambert PC et al.ElorantaHow can we make cancer survival statistics more useful for patients and clinicians: an illustration using localized prostate cancer in Sweden.Cancer Causes Control. 2013; 24: 505-515Crossref PubMed Scopus (41) Google Scholar I declare no competing interests. 40-year trends in an index of survival for all cancers combined and survival adjusted for age and sex for each cancer in England and Wales, 1971–2011: a population-based studyThese findings support substantial increases in both short-term and long-term net survival from all cancers combined in both England and Wales. The net survival index provides a convenient, single number that summarises the overall patterns of cancer survival in any one population, in each calendar period, for young and old men and women and for a wide range of cancers with very disparate survival. The persistent sex difference is partly due to a more favourable cancer distribution in women than men. Full-Text PDF Open Access
Referência(s)