- Journal List
- J Natl Cancer Inst Monogr
- PMC4829054
As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsem*nt of, or agreement with, the contents by NLM or the National Institutes of Health.
Learn more: PMC Disclaimer | PMC Copyright Notice
J Natl Cancer Inst Monogr. 2014 Nov; 2014(49): 145–186.
Published online 2014 Nov 19. doi:10.1093/jncimonographs/lgu024
PMCID: PMC4829054
PMID: 25417231
Angela B. Mariotto,
Anne-Michelle Noone, Nadia Howlader, Hyunsoon Cho, Gretchen E. Keel, Jessica Garshell, Steven Woloshin, and Lisa M. Schwartz
Author information Copyright and License information PMC Disclaimer
This article has been corrected. See J Natl Cancer Inst Monogr. 2015 May; 2015(51): 97.
Abstract
Survival statistics are of great interest to patients, clinicians, researchers, and policy makers. Although seemingly simple, survival can be confusing: there are many different survival measures with a plethora of names and statistical methods developed to answer different questions. This paper aims to describe and disseminate different survival measures and their interpretation in less technical language. In addition, we introduce templates to summarize cancer survival statistic organized by their specific purpose: research and policy versus prognosis and clinical decision making.
Survival statistics are the most used measures to estimate cancer patients’ prognosis and the likely course of their disease and are of great interest to patients, clinicians, researchers, and policy makers. Although a seemingly simple concept, survival can be confusing: there are many different survival measures with a plethora of names and statistical methods developed to answer different questions. Because most of the work has been published in technical journals, clinicians and members of the public may not appreciate the many cancer survival statistics available and how to interpret them. For example, relative survival is often used to estimate a cancer patient’s survival. However, relative survival—also called net survival—represents the net effect of a cancer diagnosis, that is, the chances of surviving assuming that cancer is the only possible cause of death. Because cancer patients, of course, can also die from competing causes, the patients’ chance of dying from the cancer, dying from other “competing” causes, or surviving—also called crude survival measures—are more relevant survival statistics for cancer patients and the clinicians treating them.
This paper has two main objectives. The first is to describe the different survival measures, the methods and assumptions behind them and their respective interpretation in less technical language. The second is to provide a presentation template for summarizing cancer survival statistics for major cancer sites, organized by measures that answer policy and research questions and measures most useful for individual cancer patients in clinical decision making.
Cancer Survival Versus Mortality Statistics: Two Sides of Different Coins
In common usage, survival and mortality are two sides of the same coin: a person is either alive or dead. But in cancer statistics, survival and mortality are two sides of different coins. Mortality measures the number of cancer deaths among the entire population (ie, people with and without cancer). It is the chance that a person in the population will die of a cancer over a period of time, usually a year. Survival is the number alive among people with cancer. It is the chance that a cancer patient will be alive some years (typically five or 10 years) after diagnosis (Table 1). For clarity, the table refers to “population mortality” and “survival for cancer patients”. The key difference between population mortality and cancer survival statistics is the denominator. For mortality, the denominator is the whole population, but for survival, the denominator only includes people diagnosed with cancer (in both cases, the denominator is typically measured as person-years at risk).
Table 1.
Comparison of mortality rate and survival statistics. Understanding progress against cancer requires examination of mortality, survival, and incidence
Open in a separate window
Open in a separate window
In the cancer registry setting, survival is sometimes called “population-based survival”. This term erroneously sounds like it refers to survival for the entire population, with and without cancer. Instead, population-based survival refers to survival of all cancer patients diagnosed in a defined population area as opposed to survival of the usually highly selected (and often unrepresentative) cancer patients who participated in randomized trial.
Survival is sometimes used as a policy measure of cancer burden and is often used to compare cancer outcomes between different populations and time periods. However, it is well known that survival is more sensitive to biases (eg, lead time and length biases) than population mortality. For example, longer survival may reflect later deaths—but it can also reflect earlier diagnosis or over diagnosis (detecting cancer cases that progress so slowly that the person dies of other causes) with no change in death. Consequently, mortality is the preferred statistic for comparisons of cancer burden between different populations and across time. Nevertheless, mortality statistics alone cannot distinguish between the effects of primary prevention, earlier detection or better treatment. A paper in this monograph (1) discusses the use of cancer survival as a cancer burden measure, its biases and highlights the importance of interpreting survival trends in the context of incidence and mortality.
For cancer patients, the main statistic of interest is not population mortality, but individual survival. Survival, not mortality, answers the question that cancer patients want to know: what is my chance of staying alive given my diagnosis? Clearly, survival is an important statistics from a clinical perspective that can provide prognosis for particular cancer types and cancer patients.
Different Measures of Survival: Dealing With Competing Causes of Death
Different survival measures answer different questions. Table 2 classifies survival into three main groups: overall survival (includes all causes of death), cancer prognosis (net survival that removes competing causes of death), and actual prognosis (crude probabilities that consider competing causes of death). We have added the terms cancer prognosis and actual prognosis to use language that is more transparent than the technical statistical terms of net survival and crude survival, respectively. Both cancer prognosis and actual prognosis are calculated differently depending on whether cause of death information is available.
Table 2.
Definitions and interpretations of prognosis statistics (ie, case-based measures) using the example of prostate cancer
Open in a separate window
Open in a separate window
Overall Survival
Overall survival—also called all-cause, observed, and crude survival—is the most easily understood survival measure. It estimates the chance of remaining alive some time after diagnosis. Because it uses death from all causes as the endpoint (as opposed to death from a specific cause, which can be misattributed), overall survival is the most reliable and available survival measure. However, it is not specific enough to provide information on survival associated with a cancer diagnosis. Higher survival may reflect fewer deaths from other causes or fewer deaths from the specific cancer.
Cancer Prognosis (Net Survival): Survival Measures That Remove Competing Causes of Death
Researchers and clinicians have long been interested in measures that isolate the effect of a cancer diagnosis on survival: to estimate the chances of surviving a cancer while removing possible distortions from competing causes of death. Such cancer prognosis measures are associated with the cancer biology, that is, what happens with the cancer or natural history of the disease in the absence of other causes of death. It also answers questions about the efficacy (in clinical settings or randomized trials) or the effectiveness (in cancer registries) of cancer interventions. In these settings, differences in cancer survival will reflect differences in cancer rather than competing causes of death. We consider net survival, that is, survival measuring the net effect of a cancer diagnosis after removing the effects of competing causes of death as a cancer prognosis measure. The two commonly used methods to estimate cancer prognosis, relative survival (2,3) and cause-specific survival (4), are described here.
Relative Survival: Relying on Life Tables to Estimate Cancer Prognosis
Relative survival is the ratio of overall survival for cancer patients to the expected survival of a comparable group of cancer-free individuals. It provides a measure of excess mortality experienced by cancer patients without requiring cause of death information. Its initial motivation was closely related to the idea of “cure”. Researchers were interested in studying if and when overall survival for cancer patients returned to the same level as the general population, that is, when the excess deaths associated with a cancer diagnosis was zero, so patients no longer died from their cancer. For most cancer registries, cause of death information obtained from death certificate is either unavailable or unreliable due to misclassification errors or inherent ambiguities in determining the underlying cause of death. For example, a metastasis site might be reported as the cause of death rather than the true underlying cause, the original cancer site. Consequently, most registries have traditionally reported relative survival.
Since a comparable group of cancer-free individuals is difficult to obtain, expected survival is estimated using general population life tables. The underlying assumptions are that cancer deaths are a negligible proportion of all deaths in the general population and that cancer and noncancer are independent competing causes of death. Expected survival is calculated from the population life tables by matching an imaginary individual from the general population whose survival is represented by the respective life table. Cancer patients are matched on age, year, sex, race, and geographic area (eg, national, state, census, and so on) if available. Expected survival using life tables can be calculated in Surveillance, Epidemiology, and End Results (SEER)*Stat using any of the following four methods: Ederer I(2), Ederer II(3) (default), Hakulinen (5,6) and Pohar–Perme (7) (soon to be implemented in SEER*Stat). The methods differ in how long the matched individuals are considered to be at risk of death (8). For five-year survival, most of the methods provide very similar relative survival estimates. The new Pohar–Perme method provides the only unbiased estimate of “net survival” (7); however, it has larger variance compared with Ederer II, which may cause estimate instability especially for long-term survival and small data (9–11).
Cause-Specific Survival: Relying on Accurate Cause of Death Information
Cause-specific (4), also denoted cancer-specific, survival uses cancer death as the endpoint and censors people dying of other causes of death. Clinical studies have long used cause-specific survival because cause of death is typically available and accurately ascertained from detailed review of medical records and adjudication committees (12). The recent development of an algorithm, which more accurately attributes a cause of death to cancer (13), has made it possible for cancer registries to move to reporting cause-specific survival. This algorithm (described on the SEER Web site http://seer.cancer.gov/causespecific/) uses causes of death that are likely to be related to the particular cancer or as a consequence of a cancer diagnosis. In situations where relative survival may be considered the gold standard, validation studies demonstrated that the cause-specific survival using the new cause of death variable more closely resembled relative survival than cause-specific survival using the presumably less accurate, reported cause of death (13).
Cause-specific survival is considered a “net” measure because it removes competing causes of death: people dying of competing causes are censored (ie, they are not counted as “endpoints” but just removed from the “at risk” group in the same way that people who are lost to follow-up are removed). In effect, cause-specific survival may be interpreted as cancer survival in the hypothetical situation in which the cancer of interest is the only possible cause of death.
When to Use Relative Versus Cancer-Specific Survival
Relative survival is the preferred method to compare survival between different registries and across countries because cause of death may not be available or there may be variability in the accurate determination of cause of death across countries (14). However, relative survival can only be calculated when accurate life tables are available to represent expected survival of the cohort of cancer patients. When cancer patients differ considerably from the general population with respect to important personal factors, which may affect deaths from other causes (such as socio-economic status, health status, and health behaviors like smoking), relative survival can be biased.
Relative survival is overestimated when expected survival from life tables is too low. In fact, relative survival may even exceed 100%. This scenario is best illustrated for cancers found largely by screening, such as localized prostate and breast cancers. People who are screened have higher life expectancy than the general US population, perhaps because of better overall health, greater access to health care, or healthier lifestyles. This healthy screened effect was most recently demonstrated in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial, where participants in this screening trial had 30%–50% lower mortality rates for heart disease, injury, and kidney disease than expected (15). Because the expected survival of the general population is lower than for the screened population, relative survival is inflated by a denominator that is too small.
Relative survival is underestimated (ie, the denominator is falsely high) when expected survival from life tables is too high. Patients with smoking-related cancers (eg, lung cancer) typically have lower life expectancy than the general population because they face substantially higher risks of death from many cancers and from heart disease (16). Because the expected survival for the general population is higher than for smokers, relative survival is deflated by a denominator that is too big.
When life tables are not available or are unlikely to accurately estimate expected survival for a particular group of cancer patients, cause-specific survival may be more accurate than relative survival. Thus, SEER reports cause-specific instead of relative survival for Hispanics, Asians (eg, Chinese, Japanese, Filipino, Vietnamese), Native Americans, and Alaska Natives (17). The cause-specific method should also be used to estimate survival by factors that affect deaths from competing causes that life tables do not account for, such as chronic disease comorbidity and smoking status.
Actual Prognosis (Crude Probabilities): Survival Measures That Include Competing Causes of Death
Cancer prognosis communicates the net effect of a cancer diagnosis: the chance of surviving assuming the cancer was the only possible cause of death. But patients diagnosed with cancer may be far more interested in understanding what is likely to happen to them over time, specifically, their chance of dying from the cancer, versus dying from competing causes or surviving. Actual prognosis measures provide this information and have been developed using statistical competing risk methods. These statistics, also known as crude probability of death, crude survival, absolute risks, competing risks, cumulative incidence function, consider two (or more) endpoints: death due to cancer and death due to competing causes. As in reality, these events are considered mutually exclusive: a person can only die from one cause. Survival is calculated as one minus the probabilities of dying of cancer and dying of competing causes, and is exactly the same as overall survival.
Similar to net survival, crude probabilities can be calculated using either cause of death information (18,19) or expected survival using population life tables (20). However, to be useful as prognosis measures, crude probabilities need to be tailored to individual cancer patients and their level of comorbidity. As general life tables will not represent expected survival for different levels of comorbidity, the cause of death method is the better method to estimate individualized actual prognosis measures.
In this monograph, Howlader et al. (21) compares actual and cancer prognosis and provides actual prognosis estimates for major cancer sites by age categories and comorbidity. Because actual prognosis measures are more valuable when tailored to the individual, they are better reported in web applications, allowing the user to enter specific demographics, tumor characteristics, and comorbidity profile to obtain the respective estimated survival. The National Cancer Institute is developing the SEER Cancer Survival Calculator (22), a tool that will provide individualized actual prognosis for patients diagnosed with breast, prostate, colorectal, and head and neck cancers, accounting for many personal factors, such as stage, grade, age, sex, race, year of diagnosis, comorbidity, marital status, and socio-economic status (22). Also in this monograph, Feuer et al. (23) report the external validation of this tool using a group of patients diagnosed with colorectal and prostate cancers in a health maintenance organization.
When to Use Cancer or Actual Prognosis Measures?
Because cancer prognosis measures reflect the hypothetical situation where competing causes of death are removed, they are the best measures to represent trends, comparisons between different groups of cancer patients and the impact of cancer biology and other factors on cancer survival. The general idea is that changes in competing causes of death should not obscure cancer survival comparisons. As such, cancer prognosis measures are best suited to answer questions related to health policy, research, and biology.
Actual prognosis, on the other hand, better describes an individual’s chance of survival because it accounts for both the chance of dying from cancer and from competing causes. Because actual prognosis most closely reflects reality, these measures are most valuable in predictive tools, clinical decision making, and cost-effectiveness analyses. For example, older patients with coexisting comorbidity may have a higher probability of dying from competing causes than of dying from their cancer; in fact, the chance of dying from competing causes may preclude the benefit of cancer treatment.
Presentation Templates for Summarizing Cancer and Actual Prognosis Measures
We developed a presentation template to summarize measures of cancer prognosis and actual prognosis. The template is designed to more efficiently and clearly present: survival trends, the effect of prognostic and demographic characteristics on cancer prognosis, and actual prognosis measures for cancer patients and clinicians. We present the templates for eight major cancer sites: prostate, female breast, lung and bronchus, colon and rectum, urinary bladder, pancreas, corpus uteri cancers, and leukemia. To represent cancer prognosis we used five-year relative survival or five-year cause-specific survival, depending on which is more appropriate. We include 95% confidence intervals whenever feasible.
Cancer Prognosis Templates
Survival trends To illustrate trends in cancer survival, we show five-year relative survival by year at diagnosis in a table format. A figure with age-adjusted incidence and mortality trends is also presented to provide interpretation of changes in survival in terms of cancer progress or burden [see Cho et al. (1) for more details on trends interpretation].
Cancer prognosis by prognostic and demographic characteristics The marginal effect of age, race and clinical characteristics on 5 year relative survival is displayed as a bar chart with 95% confidence intervals. This template can be useful to inform researchers regarding which characteristics have a bigger effect on five-year cancer survival. Five-year relative survival is also displayed as a table by age groups and clinical characteristics, which may provide information on the effect of clinical factors on cancer prognosis for different age groups.
Cancer prognosis by race/ethnicity Because life tables are not available by race and ethnicity, we used five-year cause-specific survival stratified by race/ethnicity and stage together in a table format, to represent cancer prognosis.
Actual Prognosis Templates
Actual prognosis by age group is shown in horizontal bar charts for different stage and levels of comorbidity. Each bar chart displays the percentage of patients dying of cancer (black area), dying of competing causes (dark grey area) and surviving (light gray area) five years after diagnosis. The percent at the end of each bar represents the percentage of patients surviving. The first column of bar charts represent actual prognosis by stage and age for all patients with the specific cancer type, irrespective of their comorbidity status. It represents survival for a patient with the average comorbidity in the cancer population. The second and third columns of graphs show these data for patients with no comorbidities and those with severe comorbidities, respectively [see Howlader et al. (21) in this monograph for more details on interpretation].
Data and Methods
Incidence and survival were calculated from the NCI SEER Program data. Registries joined the SEER program in different years. The SEER 9 registries were used for the calculation of time trends in incidence and survival from 1975 to 2010 and cover approximately 9% of the US population. SEER 18 was used for the remaining survival calculations, which include patients with a cancer diagnosis between 2004 and 2009 and study cutoff date December 31, 2010. These registries cover approximately 28% of the US population and have expanded reporting on race and ethnicity (eg, white, black, Asian/Pacific Islander, American Indian/Alaskan Native, and Hispanic). We used the derived American Joint Committee on Cancer, sixth edition, stage variable based on information collected for cancer cases diagnosed in 2004 and after under Collaborative Stage. Detailed information on staging may be found here: http://seer.cancer.gov/seerstat/variables/seer/ajcc-stage/6th/. SEER site recode variable based on the World Health Organization Inter national Classification of Diseases for Oncology, 3rd Edition (ICD-O-3) was used. Detailed information on the SEER site recode variable maybe found here: http://seer.cancer.gov/siterecode/index.html. Detailed information on tumor grade or differentiation may be obtained from SEER program coding and staging manual 2013 http://seer.cancer.gov/manuals/2013/SPCSM_2013_maindoc.pdf.
Mortality trends for the whole United States were estimated from deaths and causes of death data provided by the National Center for Health Statistics (www.cdc.gov/nchs) and retrieved using SEER*Stat software. The cancer sites used in these analyses were categorized according to the SEER cause of death recode. The associated ICD codes can be viewed on the SEER Web site (http://seer.cancer.gov/codrecode/1969+_d09172004/). Population estimates to calculate incidence and mortality rates were obtained from the US Census Bureau and available at SEER*Stat.
Incidence and mortalitytrends are included to provide interpretation of the changes in cancer survival trends (1). Mortality represents the whole US population, whereas incidence rates represent SEER 9 population and are adjusted to account for delay in the reporting of cases in the most recent years (delay-adjusted rates) (http://surveillance.cancer.gov/delay/). The line in Design 1 represents estimates from the Joinpoint model (24), which involves fitting a series of joined straight lines on a logarithm scale to the age-adjusted rates by calendar year.
Cancer prognosis was estimated as either relative survival or cause-specific survival. Relative survival was calculated by dividing all-cause (observed) survival to the expected survival. Expected survival is estimated using the Ederer II method (3) by linking cancer patients by sex, age, race, and year to the US (1970–2007) life tables available at (http://seer.cancer.gov/expsurvival/). Cancer survival trends is displayed as five-year relative survival by year at diagnosis estimates from a Joinpoint model on survival (25). The model fits linear segments to the hazard of dying as a function of calendar time and estimates points in which changes occurred. For patients diagnosed 2006 or later, the survival estimates reflect projected results from the model, because five-year observed is not available. Cause-specific survival was calculated to show cancer prognosis for specific race/ethnic groups because race/ethnicity life tables are not currently available. The SEER cause-specific death classification variable was used to identify deaths attributed to the specific cancer (13).
Actual prognosis is estimated as the five-year chance of dying from cancer, chance of dying of other causes and survival using the SEER cause-specific death classification to determine cause of death (13). Level of comorbidity before a cancer diagnosis for patients age 66 or older in the linked SEER-Medicare dataset were identified in previous analyses (26). Sixteen comorbid conditions identified by Charlson et al. (27) were identified from Medicare claims and summarized in an index, which was used to classify patients into comorbidity severity groups. For more information on the comorbidity index, refer to Howlader et al. (21) and Mariotto et al. (26).
All survival calculations used the complete method.
Discussion
Survival statistics are of great interest to clinicians, researchers, patients, and policy makers. Numerous methods and measures of cancer survival for cancer registry data have been developed, but not all are well known or in common use. This paper is an attempt to introduce the main cancer registry survival measures to a broad audience. To make the measures more accessible, we minimize technical language and provide explanations, suggest when to use them, and provide caveats for their interpretation. We introduce templates to summarize cancer survival statistics organized by their specific purpose: research-policy versus prognosis-clinical decision making. Although we report templates for eight major cancer sites, we plan to utilize these templates in annual reports and expand on the number of cancer types.
The other papers in this monograph complement this paper by providing applications, or describing methods and measures in more detail. Cho et al. (1) illustrate how trends in incidence and mortality can be used to interpret changes in survival. This paper also provides an explanation of the various biases that can affect survival, mostly caused by the introduction of screening or more advanced diagnostic techniques. Weir et al. (28) and Pinheiro et al. (29) study various issues with follow-up for ascertainment of vital status and how they impact survival estimates. Because life tables are an important component of relative survival estimation, Stroup et al. (30) studied the impact of national life tables versus state life tables on relative survival. Their study suggests the need to develop more appropriate life tables that better represent the varying mortality patterns in different populations for reporting of regional survival estimates. Lewis et al. (31) and Kish et al. (32) looked at cancer prognosis disparities in different populations. Lewis et al. (31) reports relative survival for adolescents and young adults diagnosed with cancer and compared with patients diagnosed at older ages. Kish et al. (32) investigates difference in five-year cause-specific survival among groups with different race, ethnicity and socioeconomic status (SES). Because life tables are not available by race/ethnicity or SES, they used cancer-specific survival. Finally, Stedman et al. (33) reports current estimates of cure fraction (the proportion of individuals that will not die of their diagnosed cancer) for selected cancers, based on SEER data, and investigates the effect of long versus short follow-up time on different types of models for estimating the cure fraction. Three monograph papers use actual prognosis measures: Howlader et al. (21), Feuer et al. (23), and Rabin et al. (34). Howlader et al. (21) compares cancer and actual prognosis measures and reports actual prognosis estimates for four leading cancers by age, comorbidity, and cancer stage. Feuer et al. (23) and Rabin et al. (34) are companion papers. They use the SEER Cancer Survival Calculator, which is being developed to be used as a web tool to provide individualized actual prognosis for prostate, female breast, colorectal and oral cancer patients. Feuer et al. (23) reports on the external validation of the tool using prostate and colorectal cancer patients’ data from a health maintenance organization. Rabin et al. (34) uses health maintenance organization data to describe service utilization patterns of subgroups of prostate cancer and colorectal patients who have different relative probabilities of dying of their cancer or other conditions as estimated by the tool. We mainly focused on methods and measures implemented in SEER*Stat that could be readily used with cancer registry data. However, there are other more technical population-based cancer survival topics that have not been covered. Some examples are age-standardized survival (http://seer.cancer.gov/stdpopulations/survival.html) (35), inclusion of multiple tumors (36), cohort definition and period survival (37) (http://surveillance.cancer.gov/survival/cohort.html), and projections of cancer survival (38).
Because different survival statistics answer different questions, both the producers and the end-users of cancer survival measures need to understand how to select and interpret the most appropriate statistic to answer the question of interest.
References
1. Cho H, Mariotto AB, Schwartz LM, Luo J, Woloshin S.When do changes in cancer survival mean progress? The insight from population incidence and mortality.J Natl Cancer Inst Monogr.2014;49:187–197. [PMC free article] [PubMed] [Google Scholar]
2. Ederer F, Axtell LM, Cutler SJ.The relative survival rate: a statistical methodology.Natl Cancer Inst Monogr.1961;6:101–121. [PubMed] [Google Scholar]
3. Ederer F, Heise H.Instructions to IBM 650 Programmers in Processing Survival Computations. Methodological Note no. 10, End Results Evaluation Section. Technical Report.Bethesda, MD: National Cancer Insitute; 1959. [Google Scholar]
4. Marubini E, Morabito A, Valsecchi MG.Prognostic factors and risk groups: some results given by using an algorithm suitable for censored survival data.Stat Med.1983;2(2):295–303. [PubMed] [Google Scholar]
5. Hakulinen T.Cancer survival corrected for heterogeneity in patient withdrawal.Biometrics.1982;38(4):933–942. [PubMed] [Google Scholar]
6. Hakulinen T.On long-term relative survival rates.J Chronic Dis.1977;30(7):431–443. [PubMed] [Google Scholar]
7. Perme MP, Stare J, Estève J.On estimation in relative survival.Biometrics.2012;68(1):113–120. [PubMed] [Google Scholar]
8. Cho H, Howlader N, Mariotto AB, Cronin KA.Estimating relative survival for cancer patients from the SEER Program using expected rates based on Ederer I versus Ederer II method. Surveillance Research Program, NCI, Technical Report #2011-01http://surveillance.cancer.gov/reports/tech2011.01.pdf. Published 2011. Accessed January 23, 2014.
9. Seppä K, Hakulinen T, Pokhrel A.Choosing the net survival method for cancer survival estimation [published online ahead of print].Eur J Cancer.2013;pii S0959-8049(13)00894-0. 10-1016/j.ejca.2013-03-019.. [PubMed] [Google Scholar]
10. Roche L, Danieli C, Belot A, et al.Cancer net survival on registry data: use of the new unbiased Pohar-Perme estimator and magnitude of the bias with the classical methods.Int J Cancer.2013;132(10):2359–2369. [PubMed] [Google Scholar]
11. Dickman PW, Lambert PC, Coviello E, Rutherford MJ.Estimating net survival in population-based cancer studies.Int J Cancer.2013;133(2):519–521. [PubMed] [Google Scholar]
12. Marubini EV, Valsecchi MG.Analysing Survival Data from Clinical Trials and Observational Studies.1st ed. West Sussex, UK: John Wiley & Sons, Inc.2004. [Google Scholar]
13. Howlader N, Ries LA, Mariotto AB, Reichman ME, Ruhl J, Cronin KA.Improved estimates of cancer-specific survival rates from population-based data.J Natl Cancer Inst.2010;102(20):1584–1598. [PMC free article] [PubMed] [Google Scholar]
14. Coleman MP, Quaresma M, Berrino F, et al.Cancer survival in five continents: a worldwide population-based study (CONCORD).Lancet Oncol.2008;9(8):730–756. [PubMed] [Google Scholar]
15. Pinsky PF, Miller A, Kramer BS, et al.Evidence of a healthy volunteer effect in the prostate, lung, colorectal, and ovarian cancer screening trial.Am J Epidemiol.2007;165(8):874–881. [PubMed] [Google Scholar]
16. Hinchliffe SR, Rutherford MJ, Crowther MJ, Nelson CP, Lambert PC.Should relative survival be used with lung cancer data?Br J Cancer.2012;106(11):1854–1859. [PMC free article] [PubMed] [Google Scholar]
17. Howlader N, Noone AM, Krapcho M, et al., eds. SEER cancer statistics review, 1975-2010. SEER Web site: SEER Cancer Statistics Review, 1975-2010. Bethesda, MD: National Cancer Institute; 2013. http://seer.cancer.gov/archive/csr/1975_2010/ Published April 2013. Updated June 14, 2013. Accessed May 28, 2014. [Google Scholar]
18. Kalbfleisch JD, Prentice RL.The Statistical Analysis of Failure Time Data.2nd ed. Hoboken, NJ: John Wiley & Sons; 2002. [Google Scholar]
19. Klein JP, Moeschberger ML.Survival Analysis: Techniques for Censored and Truncated Data.2nd ed. Berlin, Germany: Springer; 2003. [Google Scholar]
20. Cronin 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(13):1729–1740. [PubMed] [Google Scholar]
21. Howlader NM, Mariotto AB, Woloshin S, Schwartz LM.Providing clinicians and patients with actual prognosis: cancer in the context of competing causes of death.J Natl Cancer Inst Monogr.2014;49:255–264. [PMC free article] [PubMed] [Google Scholar]
22. Feuer EJ, Lee M, Mariotto AB, et al.The Cancer Survival Query System: making survival estimates from the Surveillance, Epidemiology, and End Results program more timely and relevant for recently diagnosed patients.Cancer.2012;118(22):5652–62. [PubMed] [Google Scholar]
23. Feuer EJ, Rabin B, Zou Z, et al.The Surveillance, Epidemiology, and End Results Cancer Survival Calculator SEER*CSC: validation in a managed care setting..2014;49:265–274. [PMC free article] [PubMed] [Google Scholar]
24. Kim HJ, Yu B, Feuer EJ.Selecting the number of change-points in segmented line regression.Stat Sin.2009;19(2):597–609. [PMC free article] [PubMed] [Google Scholar]
25. Yu B, Huang L, Tiwari RC, Feuer EJ, Johnson KA.Modelling population-based cancer survival trends using join point models for grouped survival data.J R Stat Soc Ser A Stat Soc.2009;172(2):405–425. [PMC free article] [PubMed] [Google Scholar]
26. Mariotto AB, Wang Z, Klabunde CN, Cho H, Das B, Feuer EJ.Life tables adjusted for comorbidity more accurately estimate noncancer survival for recently diagnosed cancer patients.J Clin Epidemiol.2013;66(12):1376–1385. [PMC free article] [PubMed] [Google Scholar]
27. Charlson ME, Pompei P, Ales KL, MacKenzie CR.A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.J Chronic Dis.1987;40(5):373–383. [PubMed] [Google Scholar]
28. Weir HK, Johnson CJ, Mariotto AB, et al.Evaluation of North American Association of Central Cancer Registries’ (NAACCR) data for use in population-based cancer survival studies.J Natl Cancer Inst Monogr.2014;49:198–209. [PMC free article] [PubMed] [Google Scholar]
29. Pinheiro PS, Morris CR, Liu L, Bungum TJ, Altekruse SF.The impact of follow-up type and missed deaths on population-based cancer survival studies for Hispanics and Asians.J Natl Cancer Inst Monogr.2014;49:210–217. [PMC free article] [PubMed] [Google Scholar]
30. Stroup AM, Cho H, Scoppa SM, Weir HK, Mariotto AB.The impact of state-specific life tables on relative survival.J Natl Cancer Inst Monogr.2014;49:218–227. [PMC free article] [PubMed] [Google Scholar]
31. Lewis DR, Seibel NL, Smith AW, Stedman MR.Adolescent and young adult cancer survival.J Natl Cancer Inst Monogr.2014;49:228–235. [PMC free article] [PubMed] [Google Scholar]
32. Kish JK, Yu M, Percy-Laurry A, Altekruse SF.Racial and ethnic disparities in cancer survival by neighborhood socioeconomic status in Surveillance, Epidemiology, and End Results (SEER) registries.J Natl Cancer Inst Monogr.2014;49:236–243. [PMC free article] [PubMed] [Google Scholar]
33. Stedman MR, Feuer EJ, Mariotto AB.Current estimates of the cure fraction: a seasibility study of statistical cure for breast and colorectal cancer.J Natl Cancer Inst Monogr.2014;49:244–254. [PMC free article] [PubMed] [Google Scholar]
34. Rabin BA, Ellis JL, Steiner JF, et al.Health-care utilization by prognosis profile in a managed care setting: using the Surveillance, Epidemiology, and End Results Cancer Survival Calculator SEER*CSC.J Natl Cancer Inst Monogr.2014;49:275–281. [PMC free article] [PubMed] [Google Scholar]
35. Corazziari I, Quinn M, Capocaccia R.Standard cancer patient population for age standardising survival ratios.Eur J Cancer.2004;40(15):2307–2316. [PubMed] [Google Scholar]
36. Weir HK, Johnson CJ, Thompson TD.The effect of multiple primary rules on population-based cancer survival.Cancer Causes Control.2013;24(6):1231–1242. [PMC free article] [PubMed] [Google Scholar]
37. Brenner H, Hakulinen T.Up-to-date and precise estimates of cancer patient survival: model-based period analysis.Am J Epidemiol.2006;164(7):689–696. [PubMed] [Google Scholar]
38. Mariotto AB, Wesley MN, Cronin KA, Johnson KA, Feuer EJ.Estimates of long-term survival for newly diagnosed cancer patients - a projection approach.Cancer.2006;106(9):2039–2050. [PubMed] [Google Scholar]
Articles from Journal of the National Cancer Institute. Monographs are provided here courtesy of Oxford University Press
