Age-Specific Spatio-Temporal Patterns of Female Breast Cancer Mortality in Spain (1975–2005)

Age-Specific Spatio-Temporal Patterns of Female Breast Cancer Mortality in Spain (1975–2005)
GOICOA, PHD, JAIONE ETXEBERRIA, MSC, MAR
IA D. UGARTE, PHD, TOMAS
MD, PHD ANA F. MILITINO, PHD, AND MARINA POLLAN,

PURPOSE: In recent decades, a decline in breast cancer mortality has been observed across Europe, and also in Spain. Our objective is to assess the spatio-temporal pattern during the period 1975–2005 by specific age groups (!45, 45–64, >65) in the Spanish provinces. METHODS: For each age group, a spatio-temporal P-spline model with a B-spline basis is used to smooth the mortality risks. Smoothing is carried out in three dimensions: longitude, latitude, and time, allowing for a different time evolution of both spatial components. The age-specific decline is calculated as the maximum of the estimated curve in each province. A confidence band for each curve is also provided. RESULTS: For the first age group (!45), the decline in the different provinces is observed between 1986 and 1991. For women aged between 45 to 64 years, the change occurs between 1990 and 1993. For the third age group (>65), change points range from 1992 to 2000, unlike Malaga and Cadiz where the change has not been observed in the studied period. Northern and some Mediterranean provinces are the areas with higher mortality risks for all the age groups. CONCLUSIONS: A different behavior for breast cancer mortality risks is observed for different provinces among the age specific groups. The decline of mortality is delayed for the oldest age group. Province differences in the implementation of screening programs could explain some of the observed differences. Ann Epidemiol 2010;20:906–916. Ó 2010 Elsevier Inc. All rights reserved. KEY WORDS:

Breast Cancer Mortality, Spatio-Temporal Trends, P-spline Models.

INTRODUCTION Breast cancer (BC) is the most common tumor in European women and is the first cause of death by cancer in females (1). In 1995, mortality ranged from 14.5 death/100,000 women in Greece to 27 deaths/100,000 women in Denmark (age-adjusted to the world population) (2). In 2004, breast cancer female mortality figures in Europe varied between 12 deaths/100,000 in Spain and 19.3 deaths/100,000 women in Hungary (2). Since the 1990s, age-adjusted mortality rates for breast cancer have declined in most of the developed world, particularly in the young and middle-aged groups (35–64 years). For example, in the United Kingdom and Switzerland, breast cancer mortality decreased about 30% between 1990 and 2002, whereas in most Southern, Northern, and Western European countries the decline was between 15% and 25%, and in the Eastern Europe,

From the Department of Statistics and O.R., Public University of Navarre, Pamplona, Spain (M.D.U., T.G., J.E., A.F.M.); the Cancer Epidemiology Unit, National Centre for Epidemiology, Carlos III Institute of Health, Madrid, Spain (M.P.); Institute of Public Health of Navarre, Pamplona, Spain (J.E.); and the CIBER in Epidemiology and Public Health, Spain (J.E., M.P.). Address correspondence to: Mar
ia D. Ugarte, PhD, Departamento de Estad
istica e Investigaci
on Operativa, Universidad P
ublica de Navarra, Campus de Arrosad
ia, 31006 Pamplona, Spain. Tel.: þ0034-948-169202. E-mail: [email protected]. Received March 18, 2010; accepted July 19, 2010. Ó 2010 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010

breast cancer mortality only decreased moderately or remained stable during the same period (3). In Spain, the economic growth and social transformation of recent decades have allowed the improvement of treatments and advances in screening and early diagnosis. Spain is divided into autonomous regions, each one including one or more provinces (Figure 1). All the autonomous regions implemented breast cancer screening programs during the 1990s (some of them before others) and, although they mainly targeted women in the age range of 50–64 years, some included the group of 45–49 years (4). Until 1992, BC mortality increased (2.9% per year [95% confidence interval (CI): 2.5, 3.3]) in the whole country. After 1992, a downturn in BC mortality was observed in women of all ages although for women older than 64 the decrease was substantially lower (5). More precisely, the analysis per age group showed a downturn in mortality around 1991– 1993 in all groups, but the decline was more pronounced in younger women, particularly in those aged 25–44 years (annual percentage change [APC] 25–44 Z
4.0% and 95% CI:
4.4 to
3.5; APC 45–64 Z
3.1% and 95% CI:
3.4 to
2.9), whereas in the oldest group the decrease was substantially lower (APC O65 Z
1.3% and 95% CI:
1.7 to
0.9) (5). Spain is a heterogeneous country with geographical differences regarding lifestyles and socioeconomic factors. The development of the urbanization and industrialization has not been the same for all Spanish provinces and this 1047-2797/$ - see front matter doi:10.1016/j.annepidem.2010.07.102

AEP Vol. 20, No. 12 December 2010: 906–916

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

uit Zlogrit Zfðx1i ; x2i ; tÞ;

Selected Abbreviations and Acronyms BC Z breast cancer CI Z confidence interval APC Z annual percentage change

leads to have different geographical mortality patterns for each cancer typology. Regarding breast cancer, it is well known that reproductive and lifestyle factors are closely related to socioeconomic development (6, 7). This study is aimed at assessing possible differences between Spanish provinces regarding the decline in breast cancer mortality. Given the above-mentioned differences in evolution among age groups, age-specific spatio-temporal patterns are presented using data registered between 1975 and 2005.

METHODS Data Source To account for possible differences in age-specific spatiotemporal mortality patterns, breast cancer deaths for 50 Spanish provinces (excluding Ceuta and Melilla, two autonomous cities) and three age groups (!45, 45–64, >65) in the period 1975–2005 are considered. Data on population and deaths were obtained from records of the Spanish Statistical Institute. Statistical Analysis The statistical modeling is carried out using spatio-temporal P-spline models for each age group. These models are different to the commonly used conditional autoregressive (CAR) models in disease mapping. They smooth global spatio-temporal trends, whereas traditional CAR models account for local effects. Here, P-spline models with Bspline bases are considered for modeling spatio-temporal interactions (8). The model is nonseparable and anisotropic as it might provide a different amount of smoothing in each direction. Specifically, if we label the Spanish provinces as i Z 1,.,50, and the time period as t Z 1975,.,2005, conditional on the random region effects rit, the number of deaths in each area and time period (for a given age group), Cit is assumed to be Poisson distributed with mean mit Z eitrit, where rit represents the unknown relative risk of mortality, and eit is the expected number of deaths. Namely Cit jrit wPoissonðmit Zeit rit Þ; logmit Zlogeit þ logrit :

[1]

The specification of log rit gives rise to different models in disease mapping. In the P-spline approach the log-risk, log rit, is modeled as a smooth function of the covariates. Then

907

[2]

where x1i and x2i are the coordinates of the centroid of the ith small area (longitude and latitude respectively), t is the time, and f is a smooth function to be estimated using P-splines with B-spline bases (8). This P-spline model provides smoothed risks along space and time for a given age group. Smoothing is carried out in three dimensions: longitude, latitude, and time, allowing for a different time evolution of both spatial components (9). The model is estimated from an empirical Bayes approach using penalized quasi likelihood (PQL). Second-order correct mean-square error estimators of the log-risk predictor (10) are built. CIs for the risks are obtained by applying the exponential function to both ends of the CIs for the log-risk and and then, areas with extreme risks can be detected. The decline in mortality risks over the studied period is evaluated by considering the maximum value of the smoothed mortality curve in each region. Standardized mortality ratios (SMR) have been also calculated as an initial guess of the whole picture of the disease. All analysis and graphs were carried out in R 2.11.1 (11). RESULTS Table 1 shows the years in which the maximum value of breast cancer mortality risk is reached for each province and age group (i.e., the years when the risks start to decrease). In general, it is observed that the change point occurs first in the youngest age group (!45), then in the second group (45–64), and it is delayed for the third group (O65). For the sake of clarity, the provinces have been grouped according to the geographical location (see Figure 1) and the autonomous region to which they belong. Five groups are considered: the northern provinces, the north-central provinces, the Mediterranean area, the central region, and finally the southern provinces. In the North, the following provinces are considered: Coru~ na, Lugo, Ourense, Pontevedra (the Galician provinces); Asturias, Cantabria, Rioja; the Basque provinces: Alava, Guipuzcoa, and Vizcaya; Navarra, Huesca, Zaragoza, and Teruel. Then, in the north-central group, Burgos, Palencia, Leon, Zamora, Valladolid, Soria, Salamanca, Avila, and Segovia (region of Castilla Leon). The Mediterranean area is composed of Barcelona, Lleida, Girona, and Tarragona (region of Catalu~ na), Castellon, Valencia, and Alicante (region of Valencia), Baleares (Islands), and Murcia. The provinces located in the central part (from west to east) of Spain are Caceres, Badajoz (region of Extremadura); Madrid; Guadalajara, Toledo, Cuenca, Ciudad Real, and Albacete (region of Castilla la Mancha). Finally the Southern provinces are (from west to east) Huelva, Sevilla, Cadiz, Cordoba, Malaga, Jaen,

908

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

AEP Vol. 20, No. 12 December 2010: 906–916

FIGURE 1. Administrative division of Spain into provinces. Provinces belonging to the same autonomous regions share the same color. Note that Canary Island have changed their exact location. They are shown at the bottom right corner.

Granada, and Almeria (region of Andalucia) and Canary Islands with provinces Las Palmas and Tenerife. Figure 2 displays the SMR (black lines) and the smooth mortality risks (red lines) with their 95% confidence bands (green lines) for each of the 50 Spanish provinces in the period 1975–2005 and the first age group (!45). The vertical line indicates the maximum value of the smoothed mortality risk, considered as the change point in each province over the period 1975–2005. The horizontal line at one is placed for interpretation purposes. The smooth mortality risk (red line) and the lower bound of the confidence band (green line) above the horizontal line mean that the risk of death from breast cancer in that province and time is significantly higher than the risk of the whole country (Spain) in the studied period. The smooth mortality risk (red line) and the upper bound of the confidence band (green line) below the horizontal line mean that the risk of death from breast cancer in that province and time is significantly lower than the risk of the whole country (Spain) in the studied period. Finally, if the horizontal line is between the lower and the upper bounds of the confidence band, the risk of that province is not statistically different than the risk of the whole country. It can be observed that the decrease occurs first in Las Palmas (Canary Islands), Baleares, and the provinces of Catalu~ na (Lleida,

Girona, Barcelona, and Tarragona). This happens because these provinces had the highest mortality rates (12), and so the decrease is easily detected. Figure 3 shows the spatio-temporal distribution of BC mortality risks in Spain during the study period for the first age group. Here, unlike Figure 2, the evolution of the geographical pattern is observed. In general, the geographical differences along time are not very pronounced. The provinces of Huelva, Sevilla, Cadiz (southwestern Spain), Las Palmas (Canary Islands) and Lleida, Girona, Barcelona, and Tarragona (Catalu~ na) exhibit an excess of risks till 1996–1997, but afterward they tend to behave as the rest of provinces. Figure 4 displays the SMRs and the smooth mortality risks with their 95% confidence bands for the second age group (45–64). In this case, the change point occurs around 1991 in most provinces. It is also interesting to observe that Jaen and Granada exhibit risks significantly lower than Spain during the whole period. The province of Las Palmas (Canary Islands) had the highest risk at the beginning of the period, but after a pronounced decrease, its risk is similar to the rest of the country at the end of the study. Figure 5 shows the evolution of the geographical pattern for the second age group. The provinces located in the Central North and the Mediterranean area exhibit the highest

AEP Vol. 20, No. 12 December 2010: 906–916

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

909

TABLE 1. Years in which the maximum of breast cancer mortality is reached in each region and age group !45 years 45–64 years O65 years !45 years 45–64 years O65 years

!45 years 45–64 years O65 years !45 years 45–64 years O65 years

Coru~ na

Lugo

Ourense

Pontevedra

Asturias

Cantabria

Rioja

Alava

Guipuzcoa

Vizcaya

Navarra

1987 1991 1997

1988 1992 2000

1989 1992 2000

1988 1991 1996

1989 1991 1996

1990 1991 1996

1990 1991 1994

1990 1991 1993

1989 1991 1992

1990 1992 1993

1989 1990 1993

Huesca

Zaragoza

Teruel

Burgos

Palencia

Leon

Zamora

Valladolid

Soria

Salamanca

Avila

1989 1990 1998

1989 1990 1997

1989 1991 1996

1990 1992 1996

1990 1992 1997

1989 1991 1996

1990 1992 1998

1990 1992 1997

1990 1992 1995

1990 1993 1999

1991 1993 1997

Segovia

Lleida

Girona

Barcelona

Tarragona

Castellon

Valencia

Alicante

Murcia

Baleares

Madrid

1990 1993 1996

1988 1991 1997

1988 1991 1993

1988 1991 1993

1989 1991 1996

1989 1991 1995

1990 1991 1996

1989 1992 1996

1990 1992 1997

1988 1992 1993

1990 1993 1996

Caceres

Badajoz

Guadalajara

Toledo

Cuenca

Ciudad Real

Albacete

Huelva

Sevilla

Cadiz

Cordoba

1990 1993 1998

1991 1992 1997

1990 1993 1996

1991 1993 1996

1990 1992 1997

1991 1993 1997

1990 1992 1999

1990 1993 1999

1991 1993 1998

1991 1993 2005

1991 1993 1997

45 years 45-64 years O65 years

Malaga

Jaen

Granada

Almeria

Las Palmas

Tenerife

1991 1993 2005

1990 1993 1998

1990 1993 1998

1990 1993 1996

1986 1991 1993

1986 1991 1994

mortality risks together with the provinces of Las Palmas, Sevilla, and Cadiz around the beginning of the 1990s. From there on, the geographical pattern changes and the behavior of the provinces is rather similar with respect to the whole of Spain. Finally, results for the third age group (O65) are presented in Figure 6 and Figure 7. More differences among provinces are observed in this age group. In particular, Las Palmas, Baleares, and Barcelona (to a lesser extent) show an excess of mortality during the whole period. However, Jaen, Galicia, and Cuenca and to a lesser extent Leon, Zamora, and Salamanca show risks significantly lower than the whole of Spain. In this age group, the change point occurs before 1995 in Navarra, Basque country provinces (Alava, Guipuzcoa and Vizcaya), Rioja, Girona, Barcelona, Baleares, and Canary Islands (Figure 6). In Navarra and Basque Country this might be explained because they were the first provinces implementing screening programs whereas as Baleares and Canary Islands had the highest mortality risks, the decrease is easily observed. Finally, it is interesting to point out that in Cadiz and Malaga a change point has not been observed in the studied period. Figure 7 displays the spatio-temporal pattern of BC mortality of the older age group. The geographical pattern does not change that much along the years. The Central Northern provinces, the Mediterranean area, and Canary Islands exhibit an excess of risk. DISCUSSION The results presented in this study show that breast cancer mortality has different behavior by provinces and age

groups. The main conclusion obtained with respect to the age groups and the BC mortality behavior is that the decline of mortality is delayed when the age of the group is increased. Regarding the geographical patterns, differences have also been found in BC mortality. More precisely, for women younger than 45 years, the mortality pattern shows a clear change between the first and second half of the period in relation to the provinces with higher risk. Whereas in the first half of the period the northern, Mediterranean, and southwestern provinces are the areas with higher mortality risks, after 1992–1993 southwestern provinces are mainly those that have the highest risks. For women between 45 and 65 years, the northern, Mediterranean, and southwestern provinces are the areas with higher mortality risks until 1995. Since 1996, some central and Mediterranean provinces have higher risks. For women over 65 years, the northeastern, some central, southwest, and north-Mediterranean provinces are those with greatest risk in the whole period. The decrease in breast cancer mortality is a generalized phenomenon. The decline starts earlier in younger women probably reflecting the increasing survival of cases (13). The average number of years that a woman with breast cancer is expected to live has risen throughout the whole period. As a result, a continuously growing number of young women with breast cancer diagnosis will die at later ages or even will live a normal life. Two factors are responsible for the gain in life expectancy among cases, namely, advances in therapy, particularly in chemotherapy, and an earlier detection throughout the implementation of screening programs.

910

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

AEP Vol. 20, No. 12 December 2010: 906–916

FIGURE 2. Smoothed breast cancer mortality risks (in red) and confidence bands (in green) using a P-spline model for women aged under 45 years in the period (1975–2005). SMRs have also been included. The x-axis represents the year and y-axis represents the mortality risk.

AEP Vol. 20, No. 12 December 2010: 906–916

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

911

FIGURE 3. Breast cancer mortality risks spatio-temporal distribution between 1975–2005 for women under 45 years of age. Note that Canary Island have changed their exact location. They are shown inside the rectangle at the bottom right corner.

The different behavior by provinces is likely to reflect differences in breast cancer occurrence. This, in turn, is also influenced by the implementation of screening programs. In breast cancer incidence, a statistically significant interruption in the steady rise was detected before 2000 in Navarra and the three Basque Country provinces of Vizcaya, Guipuzcoa, and Alava. In all of these provinces, breast cancer screening programs were fully implemented over fairly short periods and required only 1–2 years to achieve full coverage of the respective target populations. In the remaining provinces of Spain, full implementation of screening programs took longer and was achieved only recently, between 1999 and 2006 depending on the region. As a result, screening saturation was not observed (14). This evidence means that, even though all Spanish provinces have population-based screening programs, many of them have reached full coverage very recently and the time elapsed is not enough to expect a reduction in breast cancer mortality attributable to this cause (7, 14). In European countries such as United Kingdom (England, Wales, and Scotland), Finland, Iceland,

Netherlands, and Sweden, the role of the screening in breast cancer mortality was evaluated during the period 1985– 1997 (15), all of them with national screening programs. A statistically significant decline in breast cancer mortality in all age groups was found in all these countries. In the same study, countries without national breast cancer screening, such as Czech Republic, Denmark, Estonia, France, Italy, Norway, Slovakia, Slovenia, and Switzerland, did not show a decrease in breast cancer mortality in women older than 65 years (15). Differences in incidence and hence in mortality can also be explained by a heterogeneous distribution of breast cancer risk factors. In our country, a small area study showed that breast cancer mortality is positively associated with socio-economic level and negatively associated with rurality and the presence of a higher proportion of people from older generations, even though these differences tend to disappear in younger women (16). This heterogeneity is explained by the distribution and evolution of reproductive factors associated to breast cancer. Geographical differences in age at menarche among women born in different regions have

912

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

AEP Vol. 20, No. 12 December 2010: 906–916

FIGURE 4. Smoothed breast cancer mortality risks (in red) and confidence bands (in green) using a P-spline model for women aged between 45 and 64 years in the period (1975–2005). SMRs have also been included. The x-axis represents the year and y-axis represents the mortality risk.

AEP Vol. 20, No. 12 December 2010: 906–916

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

913

FIGURE 5. Breast cancer mortality risks spatio-temporal distribution between 1975–2005 for women aged between 45–64 years. Note that Canary Island have changed their exact location. They are shown inside the rectangle at the bottom right corner.

been described (17), and there are also differences regarding the age at first birth and number of children (18). In 1975, the Spanish fecundity rate was over 2 in all Spanish regions, the average rate being 2.8 (18). A marked decline was observed afterward, and 10 years later, the areas registering the lowest fertility were the Basque Country (1.29), Asturias (1.31), Navarra (1.42), and Aragon (1.44), far below the overall Spanish rate (1.64) (18). On the other hand, the age at first birth increased from 25.2 years in 1975 to 29.1 in the year 2000, Navarra and the Basque Country being the regions where women delayed longer childbearing (18). Other important risk factors include obesity and alcohol consumption. Obesity increases the risk of breast cancer in postmenopausal women (19). A study estimated that approximately 80% of the Spanish population older than 60 years of age were overweight or obese (20). The distribution of obesity is also heterogeneous, with a higher prevalence in Andalusia, Canary Islands, Murcia, Aragon, and Extremadura (21). Regarding alcohol consumption, women in Valencia, Aragon, Navarra, and Galicia had a higher prevalence of alcohol intake (21).

Finally, hormonal replacement therapy has proved to have an important influence in breast cancer occurrence among postmenopausal women (22). However, the use among Spanish women has been very limited (14, 23). There may be geographical differences in the prevalence of use, but no study is available on this issue. In conclusion, it is impossible to estimate the specific contribution of each of the agents influencing the spatio-temporal evolution of breast cancer mortality in our country. Geographical differences tend to disappear, probably reflecting a more uniformed distribution of the above mentioned factors, namely diffusion of screening, reproductive behavior, obesity and other life-style factors, in recent years. In geographical analysis, annual mortality risks in each geographical unit can be very unstable, even for very common diseases such as breast cancer. Therefore, in the last 20 years the use of sophisticated statistical models has greatly increased. For example, CAR models, incorporating spatial dependence are used widely. In this study, P-spline models with B-spline bases are used instead. These models are more appropriate than CAR models for situations where

914

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

AEP Vol. 20, No. 12 December 2010: 906–916

FIGURE 6. Smoothed breast cancer mortality risks (in red) and confidence bands (in green) using a P-spline model for women aged over 65 years in the period 1975–2005. SMRs have been also included. The x-axis represents the year and y-axis represents the mortality risk.

AEP Vol. 20, No. 12 December 2010: 906–916

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

915

FIGURE 7. Breast cancer mortality risks spatio-temporal distribution between 1975–2005 for women over 65 years of age. Note that Canary Island have changed their exact location. They are shown inside the rectangle at the bottom right corner.

the risks are supposed to change gradually over space and time (8). Besides, penalized splines offer a flexible methodology to build anisotropic and non-separable models that easily model the spatial trend using the centroids of the provinces. These models have the advantage of being represented as generalized linear mixed effects models (GLMM) and then, existing theory can be applied. For example, the GLMM representation facilitates the estimation of the smoothing parameters that become variance components in the GLMM framework. In summary, P-spline models are very appropriate to analyze spatio-temporal trends of mortality (or incidence) risks by age groups allowing for spatio-temporal interactions and a different amount of spatial smoothing in each direction (longitude and latitude). The final estimated curve (and its confidence bands) easily show the trend changes along the time. In this study it has been observed that the decline in breast cancer mortality is gradual in all age groups although it is delayed for the oldest age group. This result is explained by the increased survival in breast cancer cases,

which implies that cases diagnosed in younger groups will contribute to breast cancer mortality at older ages. This work has been supported by the Spanish Ministry of Science and Innovation (Project MTM2008-03085).

REFERENCES 1. Izquierdo A, Gispert R, Saladie F, Espinas JA. An
alisis de la incidencia, la supervivencia y la mortalidad seg
un las principales localizaciones tumorales, 19852019: C
ancer de mama. Med Clin (Barc). 2008;131(Suppl 1):50–52. 2. World Health Organization. Mortality database. WHO;2-6-2009. 3. Hery C, Ferlay JM, Boniol M, Autier P. Quantification of changes in breast cancer incidence and mortality since 1990 in 35 countries with Caucasian-majority populations. Ann Oncol. 2009;19:1187–1194. 4. Ascunce N, Ederra M, Barcos A, Zubizarreta R, Fern
andez AB, Casamitjana M. Situaci
on del cribado de c
ancer de mama en Espa~ na: caracter
isticas y principales resultados de programas existentes. In: Castells X, Sala MAN, et al., eds. Descripci
on del cribado del c
ancer en Espa~ na a. Proyecto DESCRIC. Madrid Barcelona: Ministerio de Sanidad y Consumo, Agencia d’Avaluaci
o de Tecnolog
ia i Recerca Mediques de Catalunya; 2007:31–73. 5. Cabanes A, Vidal E, Perez-Gomez B, Aragones N, Lopez-Abente G, Pollan M. Age-specific breast, uterine and ovarian cancer mortality trends in Spain: changes from 1980 to 2006. Cancer Epidemiol. 2009;33:169–175.

916

Ugarte et al. FEMALE BC MORTALITY IN SPAIN 1975-2005

AEP Vol. 20, No. 12 December 2010: 906–916

6. Robert SA, Strombom I, Trentham-Dietz A, Hampton JM, McElroy JA, Newcomb PA, et al. Socioeconomic risk factors for breast cancerddistinguishing individual- and community-level effects. Epidemiology. 2004;15:442–450.

15. Botha JL, Bray F, Sankila R, Parkin DM. Breast cancer incidence and mortality trends in 16 European countries. Eur J Cancer. 2003;39: 1718–1729.

7. dos Santos I, Beral V. Socioeconomic differences in reproductive behaviour. IARC Sci Publ. 1997;138:285–308.

16. Poll
an M, Ramis R, Aragon
es N, P
erez-G
omez B, G
omez-Barroso D, Lope V, et al. Municipal distribution of breast cancer mortality among women in Spain. BMC Cancer. 2007;7:78.

8. Ugarte MD, Goicoa T, Militino AF. Spatio-temporal modeling of mortality risks using penalized splines. Environmetrics. 2010;21:270–289. 9. Lee D-J, Durb
an M. Pspline ANOVA type interaction models for spatio temporal smoothing. Statistical Modelling. 2010 (in press). 10. Ugarte MD, Militino AF, Goicoa T. Prediction error estimators in empirical Bayes disease mapping. Environmetrics. 2008;19:287–300. 11. R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2010. Available at: http://www.R-project.org. 12. L
opez-Abente G, Ramis R, Poll
an M, Aragon
es N, P
erez-G
omez B, G
omez-Barroso D, et al. Atlas municipal de mortalidad por c
ancer en Espa~ na, 1989–1998. Madrid: Instituto de Salud Carlos III; 2006.

17. Cabanes A, Ascunce N, Vidal E, Ederra M, Barcos A, Erdozain N, et al. Decline in age at menarche among Spanish women born from 1925 to 1962. BMC Public Health. 2009:449. 18. Arroyo A. Fecundidad. In: Arroyo A, ed. Tendencias demogr
aficas cas durante el siglo XX en Espa~ na. Madrid, Spain: Instituto Nacional de Estad
istica; 2004:119–162. 19. Bergstrom A, Pisani P, Tenet V, Wolk A, Adami HO. Overweight as an avoidable cause of cancer in Europe. Int J Cancer. 2001;91:421–430. 20. Gutierrez-Fisac JL, Lopez E, Banegas JR, et al. Prevalence of overweight and obesity in elderly people in Spain. Obes Res. 2004;12:710–715. 21. Ministerio de Sanidad y Consumo. I.N.E. Encuesta Nacional de Salud de Espa~ na 2006. Madrid, Spain: Ministerio de Sanidad y Consumo; 2007.

13. Berrino F, De Angelis R, Sant M, Rosso S, Lasota MB, Coebergh JW, et al. Survival for eight major cancers and all cancers combined for European adults diagnosed in 1995-99: results of the EUROCARE-4 study. Lancet Oncol. 2007;8:773–783.

22. Kerlikowske K, Miglioretti DL, Buist DS, Walker R, Carney PA. Declines in invasive breast cancer and use of postmenopausal hormone therapy in a screening mammography population. J Natl Cancer Inst. 2007;99: 1335–1339.

14. Poll
an M, Pastor-Barriuso R, Ardanaz E, Arg€uelles M, Martos C, Galcer
an J, et al. Recent changes in breast cancer incidence in Spain, 1980–2004. J Natl Cancer Inst. 2009;101:1584–1591.

23. Schneider HP. Cross-national study of women’s use of hormone replacement therapy (HRT) in Europe. Int J Fertil Womens Med. 1997;42(Suppl 2):365–375.