Competing risks to breast cancer mortality in Catalonia
© Vilaprinyo et al; licensee BioMed Central Ltd. 2008
Received: 16 January 2008
Accepted: 12 November 2008
Published: 12 November 2008
Breast cancer mortality has experienced important changes over the last century. Breast cancer occurs in the presence of other competing risks which can influence breast cancer incidence and mortality trends. The aim of the present work is: 1) to assess the impact of breast cancer deaths among mortality from all causes in Catalonia (Spain), by age and birth cohort and 2) to estimate the risk of death from other causes than breast cancer, one of the inputs needed to model breast cancer mortality reduction due to screening or therapeutic interventions.
The multi-decrement life table methodology was used. First, all-cause mortality probabilities were obtained by age and cohort. Then mortality probability for breast cancer was subtracted from the all-cause mortality probabilities to obtain cohort life tables for causes other than breast cancer. These life tables, on one hand, provide an estimate of the risk of dying from competing risks, and on the other hand, permit to assess the impact of breast cancer deaths on all-cause mortality using the ratio of the probability of death for causes other than breast cancer by the all-cause probability of death.
There was an increasing impact of breast cancer on mortality in the first part of the 20th century, with a peak for cohorts born in 1945–54 in the 40–49 age groups (for which approximately 24% of mortality was due to breast cancer). Even though for cohorts born after 1955 there was only information for women under 50, it is also important to note that the impact of breast cancer on all-cause mortality decreased for those cohorts.
We have quantified the effect of removing breast cancer mortality in different age groups and birth cohorts. Our results are consistent with US findings. We also have obtained an estimate of the risk of dying from competing-causes mortality, which will be used in the assessment of the effect of mammography screening on breast cancer mortality in Catalonia.
Breast cancer is the leading cause of mortality among middle-aged women in many developed countries, including Catalonia, a region in the northeast of Spain, where it accounts for a fifth of all female cancer deaths and, on the average, fourteen years of potential life lost per death from this cause .
Causes of death other than breast cancer may influence mortality trends in two ways: 1) by changing the number of women at risk of having breast cancer and, 2) by changing the risk of dying of breast cancer once it has developed. For example, infectious diseases at the beginning of the 20th century reduced the number of women at risk for breast cancer and also competed with breast cancer as a cause of death in women with this disease. Thus, the effect of other causes of death in breast cancer incidence and mortality has changed over time depending on trends in other competing risks.
Rosenberg  used the multi-decrement life table methodology to partition overall mortality into mortality due to breast cancer and mortality due to other causes. In addition, multi-decrement life tables permit assessment of the impact that breast cancer mortality has on overall mortality by birth cohort and age. For instance, in the US, the reduction in overall mortality when removing breast cancer as a cause of death could be as high as 15% at some ages.
Evaluation of the impact of mammography and adjuvant therapy on breast cancer mortality reduction can be done using statistical models, for example those found in several studies sponsored by CISNET . The modeling process requires information such as the dissemination of mammography and adjuvant treatment programs, diagnostic characteristics of mammography, breast cancer incidence and mortality. Another required input is the competing-cause mortality, taking into account that deaths due to breast cancer occur in the presence of other causes of death, the above-mentioned competing risks.
The aims of the present work are: 1) to assess the impact of breast cancer mortality on overall mortality by birth cohort and age in Catalonia in the 20th century and 2) to assess the risk of death from other causes than breast cancer in cohorts born from 1900 to 2004. Our analysis is based on Rosenberg's work for the US and is part of a wider project that aims to model the impact of mammography in Catalonia (Spain) based on the methodology developed by Lee and Zelen for the CISNET project .
Catalonia is an autonomous region of Spain which has had authority over health-care planning, administration and provision since 1985. It has approximately one sixth of the Spanish population. The Catalan Health Department has implemented an independent information system and preventive programs in the region. Population breast cancer screening programs were initiated in the early nineties .
By the year 2007, the Catalan Health Service was providing services to 7 million inhabitants, including 3.5 million women. The female population grew from 800.000 women in 1900 to more than 3 million in 2004. The number of breast cancer deaths increased from an average of 612 per year in the period 1975–79 to 1060 per year in the period 1990–94 and then decreased to an average of 1000 per year in the period 2000–04. For the same periods, the percentage of breast cancer deaths among overall mortality in women increased from 3.5% to 5.2% and then decreased to 3.6%.
Description and sources of data.
Official census data (every 10 years)
National Institute of Statistics (INE)
Annual population estimated by the components method*
Annual population estimated by the components method*
Catalan Institute of Statistics (IDESCAT)
Catalan Mortality Registry
Breast cancer mortality
Catalan Mortality Registry
We grouped cohorts and ages into five-year categories in order to smooth data and minimize fluctuations present in some census and mortality data prior to 1960 , and to get more reliable estimates for young age groups. We used linear interpolation to estimate missing data or to split categories that were larger than five years in the original data source.
Deaths by calendar year were converted into deaths by year of birth (cohorts) using the relation year of birth = calendar year of death – age at death.
Actuarial life table methods were used to obtain the life table functions: the age specific death rate n M x,BY , the conditional probabilities of death n q x,BY , and the proportion of expected survivors by age l x,BY . These functions were estimated for a) the overall population, and b) for a population in which breast cancer had been eliminated as a cause of death, in the following steps:
where n D x,BY is the observed number of deaths and n K x,BY the midyear population in age group x to x+n. In our analysis n is equal to 5 years.
where n f x is the estimated fraction of years lived in the interval. Assuming a uniform distribution of deaths over time we have taken n f x = 0.5 for all age groups except the 0–4 age group, where n f 0 = 0.4. All-cause probabilities of death are labeled as .
3) We estimated probabilities of death from breast cancer, , in the interval [x, x+n) as described in Step 2, using only breast cancer deaths in the numerator.
4) Probabilities of death in the last age group, 85+, were estimated as = 1 for overall mortality, and for breast cancer mortality.
7) We obtained the life table function l x,BY , which indicates the proportion of survivors at age x from birth cohort BY, using the expression:
l x+n,BY = l x,BY (1 - n q x,BY ), starting from l 0,BY = 1
Following the same nomenclature used in probabilities of death, the functions and were computed. The indicates the proportion of survivors from birth cohort BY, at the beginning of each age interval, after removing breast cancer as a cause of death.
8) We computed the ratio of , which measures the impact in all-cause mortality of removing deaths from breast cancer at different birth cohorts and ages. A ratio equal to 1, in a specific group, would indicate that there is no breast cancer mortality in that group. A ratio equal to 0 would indicate that all mortality is due to breast cancer.
Breast cancer impact on overall mortality
Cohort multi-decrement life table functions and were obtained for a population of women born between 1900 and 2004, with breast cancer removed as a cause of death.
Risk of death from other causes than breast cancer
Proportion of survivors, , and proportion of survivors after removing breast cancer as a cause of death, , by age x and cohort of birth.
Year of Birth
The column Dif. in Table 2 contains the difference per 10,000 women. It can be interpreted as the number of women saved up to age x from an initial cohort of 10,000 women, when breast cancer is removed as a cause of death. If there were no breast cancer deaths, for each 10,000 women born in 1920–24 approximately 93 would have been saved before arriving at age 70, and 118 women for the cohort born in 1930–34.
Figure 2 and Table 2 show two different approaches to evaluating trends in breast cancer mortality in the presence of other causes of death by age and cohort. The estimates competing-cause mortality, the ratio shows the impact of breast cancer over global mortality, and indicates the number of women saved up to age x if breast cancer was removed as a cause of death.
Discussion and conclusion
Our study shows that breast cancer mortality in Catalonia has experienced important changes over the last century. There was an increasing impact of breast cancer on overall mortality in the first part of the century, with a peak for cohorts born in 1945–54 in the 40–49 age groups (approximately 24% of mortality was due to breast cancer in these age groups and cohorts). This increasing impact could be explained by a decrease in mortality from other causes of death and increased breast cancer incidence and mortality due to changes in lifestyle and reproductive patterns . In Catalonia, mortality from all causes in women has been decreasing an average of 1.6% per year since 1978 , whereas breast cancer incidence increased 2.2% per year between 1980–97 .
The subsequent decrease in impact was due, primarily, to the observed reduction in mortality from breast cancer during the 1990s, which affected all age groups (see  and official reports by the Catalan Department of Health ). Furthermore, the significant reduction in recent decades implies important gains in life expectancy for middle-aged women . This reduction has been attributed to the use of mammography and adjuvant treatments in the US . The phenomenon is now being studied in Catalonia.
In Catalonia, as in the US, breast cancer mortality had the greatest impact on global mortality in the 40–54 age groups during the 20th century . On one hand, breast cancer incidence and mortality are low before the age of 40 . Although breast cancer incidence and mortality increase with age, after age 50 there are other causes of death (competing risks) that are acting simultaneously, and therefore breast cancer may have less impact on overall mortality.
It is worthwhile to note that the values of the ratio were higher in the US, which means that the impact of breast cancer mortality was lower. The ratio depends on two risks, the risk of dying of causes other than breast cancer (numerator) and the risk of dying of any cause (denominator). For cohorts of women born in the middle of the 20th century, overall mortality rates at ages 40 to 54 years were 40% to 60% higher in the US than in Catalonia . On the other hand, breast cancer mortality rates for these ages and cohorts were similar or slightly higher in Spain than the US. Since breast cancer mortality rates represent about one fifth of the mortality in these age groups, differences in the overall mortality risk in the US and Catalonia may explain the differences seen in the ratio. But, other factors like reproductive patterns, hormonal replacement therapy use and fat intake could explain differences in the impact of breast cancer mortality in both countries.
Results from Rosenberg show that the lowest ratio (the highest impact of breast cancer mortality) for women born in 1930 occurs at older ages than for those born in 1950. This pattern is also shown in our data before any extrapolation (Figure 2a). In Catalonia this is related with a clear period effect during the 1990s. As we show in Figures 3 and 4, trends for breast cancer death probabilities by age increased from the 1970s to the 1990s and started a decreasing trend after that for all age groups.
The declines in mortality from breast cancer, which began in Europe in the late 1980s may be attributed in part to earlier detection by screening programs. But, since the declining trends started before screening was introduced and occurred also in non-screened age groups, improved cancer treatments such as adjuvant chemotherapy and tamoxifen may have been important determinants of breast cancer mortality reduction . In Catalonia, breast cancer mortality started to decrease at the beginning of the 1990s, concurrent with the dissemination of mammography. Since the effect of screening would be seen some years later, treatments probably had an important role during the 1990s. The contribution of each of these factors still needs to be determined in Catalonia.
Our work was undertaken in the Catalan region and not in the whole country of Spain. This is due to the fact that there is no national cancer registry in Spain. Instead, there are 12 local cancer registries, which show differences in the incidence of breast cancer by region (probably due to different reproductive patterns and degrees of economic development). The only statistics that are collected in a standardized manner at the national level are mortality data. Also, this study is part of a project that aims to assessing the cost-effectiveness of different early detection strategies on the reduction of breast cancer mortality. The cost-effectiveness analysis needs information on outcomes and costs that would be difficult to obtain at the national level. On the other hand, since Catalonia has approximately one sixth of the Spanish population, we believe that some of the results of our studies will be relevant for all of Spain.
This study has several limitations. First, the multi-decrement method that we used assumes independence of causes of death, which means that when breast cancer mortality is removed, the risk of dying due to the remaining causes of death is not affected. This assumption may not be true, since different causes of death may share the same risk factors. For example, body weight, smoking, and diet are associated not only with breast cancer but also with other health problems, like cardiovascular diseases. Therefore, the probability of dying of other causes could change when breast cancer death is removed. In our study, if mortality from other causes had decreased when eliminating breast cancer mortality, the impact of breast cancer mortality would be higher than the reported values. There are methods for competing risks analysis, such as cumulative incidence functions, that do not make any assumptions about independence of risks, but need information on the relationships among them [22, 23]. These methods are more complex than multi-decrement life tables and have been used to evaluate the effects of explanatory variables, such as assessing the effects of therapy in different groups with multiple endpoints. On the other hand, Chin Long Chiang proposed a method  that takes into account different conditional probabilities of death (crude, net and partial crude) which reflect the relationships among the different risks of death acting simultaneously. Nevertheless, although these methods could be more accurate than the one used in our work, they make assumptions about the risks' associations requiring information that was not available to us such as causes of death for women with breast cancer. Therefore, we chose the multi-decrement method, that Rosenberg also used , because it is simple and adequate to fulfill our objectives.
A second limitation concerns quality of information. The validity of population and mortality data has increased over time, but there may be errors in data from the earliest years of the study. We observed some fluctuations in the late 1930s and early 1940s, coinciding with the Spanish Civil War, a problem that could be addressed by smoothing the life table series' before using them in subsequent analyses. Also, to minimize variability we aggregated data in 5-year age groups and calendar years. The quality of mortality statistics has also changed over time, but different studies have shown that in Spain, deaths from cancer as a whole and leading cancer sites (lung, colon-rectum, prostate, stomach, pancreas, female breast, uterus, brain, leukemia, lymphomas and myeloma) were properly coded [25, 26]. Furthermore, breast cancer mortality data was provided by the Catalan Mortality Registry which, since the beginning of the 1980s, collects data and codes causes of death consistently and has adequate quality standards.
A third limitation arises from the assumptions made when completing information on breast cancer mortality probabilities for years without available data. Data from available years seemed adequate to us to fit an age-cohort model that predicted missing information. Other approximations such as the assumption that the probabilities of dying from breast cancer at a given age would be the same in neighboring years with data, as Rosenberg did, do not change the main results (data not shown). In any case, extrapolated data for the earlier cohorts may differ from the actual values and therefore results based on these data should be interpreted with caution.
In summary, we obtained cohort life tables for causes of death other than breast cancer in Catalonia (Spain). This makes it possible to quantify the impact of removing breast cancer on overall mortality in different age groups and birth cohorts (). As with the US results, our analysis found that the greatest impact of breast cancer mortality on overall mortality was for women aged 40–54 born in the middle of the 20th century. The multi-decrement life tables method also provided estimations of the risk of dying from competing risks that will be used in the assessment of the effect of mammography screening on breast cancer mortality.
This work has been funded by projects PI06/1649 and PI06/90355 from the Fondo de Investigación Sanitaria, Spanish Ministry of Health, and project 068/27/06 from the Agència per a l'Avaluació de la Tecnologia i la Recerca Mèdiques, Department of Health, Generalitat de Catalunya, Spain. We are indebted to the reviewers for their comments, which have improved the quality of the manuscript.
- García F, Puigdefàbregas A, Gispert R: Anàlisi de la mortalitat a Catalunya, 2005. 2007, Barcelona: Catalan Department of Health, Servei d'informació i estudis. Generalitat de Catalunya, [http://www.gencat.cat/salut/depsalut/html/ca/dir1932/index.html]Google Scholar
- Rosenberg MA: Competing risks to breast cancer mortality. J Natl Cancer Inst Monogr. 2006, 15-19. 36
- Habbema JD, Schechter CB, Cronin KA, Clarke LD, Feuer EJ: Modeling cancer natural history, epidemiology, and control: reflections on the CISNET breast group experience. J Natl Cancer Inst Monogr. 2006, 122-126. 36
- Lee S, Zelen M: A stochastic model for predicting the mortality of breast cancer. J Natl Cancer Inst Monogr. 2006, 79-86. 36
- Masuet C, Seculi E, Brugulat P, Tresserras R: The practice of preventive mammography in Catalonia [Spain]: a step forward. Gac Sanit. 2004, 18 (4): 321-325.View ArticlePubMedGoogle Scholar
- INEbase: Official statistics website of Spain. [http://www.ine.es/en/inebmenu/indice_en.htm]
- IDESCAT:Official statistics website of Catalonia. [http://www.idescat.net]
- S Shryock H, Siegel JS, Associates: The Methods and Materials of Demography (Studies in Population). 1976, San Diego: Academic PressGoogle Scholar
- Servei d'informació i estudis: Anàlisi de la mortalitat a Catalunya, diversos anys. Barcelona: Catalan Department of Health, Generalitat de Catalunya, [http://www.gencat.cat/salut/depsalut/html/ca/dir1932/index.html]
- Glei DA, Redondo RG, Argüeso A, Canudas-Romo V: Description of Spanish data. Human Mortality Database. University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany), http://www.mortality.org or http://www.humanmortality.de
- Elandt-Johnson RC, Johnson NL: Survival Models and Data Analysis. 1980, John Wiley & Sons edn. New York: John Wiley & SonGoogle Scholar
- R Development Core Team: R: A language and environment for statistical computing. 2008, Vienna, Austria: R Foundation for Statistical Computing, [http://www.R-project.org]Google Scholar
- Kelsey JL, Gammon MD: The epidemiology of breast cancer. CA Cancer J Clin. 1991, 41 (3): 146-165. 10.3322/canjclin.41.3.146.View ArticlePubMedGoogle Scholar
- Puig X, López-Abente G, Gispert R, Freitas A, Puigdefàbregas A: Tendències de la mortalitat a Catalunya, 1978–2002. Models edat/període/cohort. 2005, Catalan Department of Health, Generalitat de Catalunya, [http://www.gencat.cat/salut/depsalut/pdf/t1978-2002.pdf]Google Scholar
- Borràs JMBJ, Bosch FX, Fernández E, Galceran J, Gispert R, González JR, Izquierdo A, Marcos R, Moreno V, Peris M, Puig X, Sánchez V, Viladiu P: Càncer Catalunya 2001. 2001, L'Hospitalet: Institut Català d'Oncologia, Departament de Sanitat i Seguretat SocialGoogle Scholar
- Puig X, Ginebra J, Gispert R: Analysis of mortality time trend using generalized linear models. Gac Sanit. 2005, 19 (6): 481-485.View ArticlePubMedGoogle Scholar
- Serra I, Gispert R, Puig X, Torné M, Puigdefàbregas A: Impacte de l'edat i les causes de mort en els canvis de l'esperança de vida. Catalunya, 1987–2002. 2006, Catalan Department of Health, Generalitat de Catalunya, [http://www.gencat.cat/salut/depsalut/pdf/imcauses2006.pdf]Google Scholar
- Cronin KA, Feuer EJ, Clarke LD, Plevritis SK: Impact of adjuvant therapy and mammography on U.S. mortality from 1975 to 2000: comparison of mortality results from the cisnet breast cancer base case analysis. J Natl Cancer Inst Monogr. 2006, 112-121. 36
- Parkin DM, Whelan SL, Ferlay J, Storm H: Cancer Incidence in Five Continents, Vol. I to VIII. IARC CancerBase No 7, Lyon. 2005, IARC CancerBase No. 7, LyonGoogle Scholar
- Human Mortality Database. [http://www.mortality.org/]
- Botha JL, Bray F, Sankila R, Parkin DM: Breast cancer incidence and mortality trends in 16 European countries. Eur J Cancer. 2003, 39 (12): 1718-1729. 10.1016/S0959-8049(03)00118-7.View ArticlePubMedGoogle Scholar
- Fine J, Gray R: A Proportional Hazards Model for the Subdistribution of a Competing Risk. J Am Statist Ass. 1999, 94 (446): 496-509. 10.2307/2670170.View ArticleGoogle Scholar
- Schairer C, Mink PJ, Carroll L, Devesa SS: Probabilities of death from breast cancer and other causes among female breast cancer patients. J Natl Cancer Inst. 2004, 96 (17): 1311-1321.View ArticlePubMedGoogle Scholar
- Chiang CL: Competing risks in mortality analysis. Annu Rev Public Health. 1991, 12: 281-307. 10.1146/annurev.pu.12.050191.001433.View ArticlePubMedGoogle Scholar
- Perez-Gomez B, Aragones N, Pollan M, Suarez B, Lope V, Llacer A, Lopez-Abente G: Accuracy of cancer death certificates in Spain: a summary of available information. Gac Sanit. 2006, 20 (Suppl 3): 42-51. 10.1157/13101089.View ArticlePubMedGoogle Scholar
- de Aran Bares M, Perez G, Rosell J, Molina P: Exactness of mortality statistics by external and natural causes of death with medico-legal intervention in Catalonia, 1996. Gac Sanit. 2000, 14 (5): 356-362.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2407/8/331/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.