Age–period–cohort analysis of breast cancer mortality rates in Andalucia (Spain)

A. Cayuela1,*, S. Rodríguez-Domínguez2, M. Ruiz-Borrego3 and M. Gili4

1 Supportive Research Unit and 3 Oncology Department, Virgen del Rocío Hospital, Seville; 2 Benacazón Health Center, Seville; 4 Department of Social Sciences, University of Seville, Andalucia, Spain

Received 13 May 2003; revised 31 October 2003; accepted 2 January 2004


    ABSTRACT
 Top
 ABSTRACT
 Introduction
 Patients and methods
 Results
 Discussion
 REFERENCES
 
Background:

The aim of this paper is to analyse breast cancer mortality in Andalucia (Spain) between 1975 and 1999 based on age–period–cohort.

Patients and methods:

Mortality data were obtained from the Mortality Registry of Andalucia. Deaths and population were divided into 13 age groups and five 5-year periods. From this, age-specific mortality rates for 17 birth cohorts were computed. These were plotted and fitted to Poisson regression models to assess age, period and cohort effects.

Results:

The best fit was found for the complete model, which simultaneously considered the effects of age–period–cohort. Cohort effects were found to be more important than period effects in terms of model fit.

Conclusion:

These effects were manifest as a seemingly consistent increase in the relative risk of breast cancer mortality with a three-fold increase in women born in the 1950s relative to those born in the 1890s.

Key words: Andalucia, breast cancer, epidemiology, mortality


    Introduction
 Top
 ABSTRACT
 Introduction
 Patients and methods
 Results
 Discussion
 REFERENCES
 
Although breast cancer mortality rates in women have been increasing in most developed countries since 1950, recently mortality rates seem to have levelled off or started to decline in many of these countries [1].

Although overall age-standardized breast cancer mortality rates declined by 7% in Europe between 1988 and 1996 [2], breast cancer is the leading cause of death from cancer in women (17%) [3].

In Spain there is no evidence of a decline or levelling off of mortality in recent birth cohorts or in recent years [1, 4]. In a previous report we analysed trends in mortality from breast cancer in Andalucia (Spain) during the period 1975–1992, showing that overall age-standardized breast cancer mortality rates increased (2% per year) [5]. In this study the trend of breast cancer mortality could be explained by changes in factors that act around the time of death (period effect) and by risk factors that are present in early life (cohort effect). To provide updated information about breast cancer mortality in Andalucia, we have now considered death rates between 1975 and 1999 and applied an age–period–cohort model in order to separate the effects of age, period at death and cohort of birth on breast cancer mortality.


    Patients and methods
 Top
 ABSTRACT
 Introduction
 Patients and methods
 Results
 Discussion
 REFERENCES
 
Annual data on breast cancer deaths (code 174 in the eighth and ninth revisions of the International Classification of Diseases) were taken from the official vital statistics published by Andalucia’s Institute for Statistics for the years 1975–1999, which was the last year for which data was available at the time of the study. The data were arranged in five 5-year periods from 1975–1979 to 1995–1999 and thirteen 5-year age groups from 25–29 years to ≥85 years. The 5-year age and period grouping produced 17 partially overlapping 10-year birth cohorts (calculated as period minus age and defined by their respective mid-years, starting with 1890).

Parallel population groups were constructed using mid-year official population estimates. From these data, age-specific mortality rates were calculated for each birth cohort and plotted for the purposes of graphical presentation. Age-adjusted breast cancer mortality rates were calculated by the direct method, using the world standard population as the standard [6].

Evaluation of the effects of age, birth cohort and period of death was performed by means of a log-linear Poisson model, fitted using the General Linear Interactive Modelling (GLIM) software with appropriate macros. The predictive variables (age, cohort and period) were sequentially included in the model. Age was considered first since breast cancer mortality increases markedly with age. In the next step, three two-factor ‘age + period’, ‘age + cohort’ and ‘cohort + period’ models were considered. In the third model, values for cohort and period were derived after estimating the age effect alone. Age effects are expressed in terms of rate per 100 000 population. Cohort of birth and period of death are expressed as relative values compared to their weighted average (i.e. as estimators of relative risk for each cohort and period). The final model was defined by minimizing the sum of the Euclidean distances between the three two-factor models [79]. The resulting adjusted values for this model were then plotted for presentation. Goodness-of-fit was evaluated by comparing the deviance for each model with that for the ‘age’ model.


    Results
 Top
 ABSTRACT
 Introduction
 Patients and methods
 Results
 Discussion
 REFERENCES
 
Age-adjusted breast cancer mortality rates (Figure 1) increased in Andalucia during the period 1975–1993 (average annual increase, 2.7%). From that time until 1999, rates fell from 16.8 to 15.3 per 100 000, the lowest since 1986 (–1.6% a year).



View larger version (10K):
[in this window]
[in a new window]
 
Figure 1. Age-adjusted breast cancer mortality rates: Andalucia, 1975–1999.

 
Age-specific breast cancer mortality rates plotted against the central year of birth cohort are depicted on a semi-logarithmic scale in Figure 2. As described in the figure keys, rates for all age groups have been joined by lines. This presentation enables graphical assessment of the age, period and cohort effects on breast cancer mortality rates. It should be noted that cohort values relating to earlier and recent periods are based on fewer age-specific rates, and hence are less reliable than the central period. Furthermore, recent values are based on fewer deaths. As indicated, mortality increased markedly with age in all cohorts.



View larger version (13K):
[in this window]
[in a new window]
 
Figure 2. Age-specific 5-year mortality rates from breast cancer per 100 000 woman-years by middle year of birth for Andalucia (1975–1999). Closed diamond, 85+ years; closed square, 80–84 years; closed triangle, 75–79 years; cross, 70–74 years; open circle, 65–69 years; black line, 60–64 years; closed square on dotted line, 55–59 years; star, 50–54 years; open diamond, 45–49 years; open square on dotted line, 40–44 years; open triangle, 35–39 years; cross, 30–34 years; closed circle, 25–29 years.

 
A significant reduction in deviance (a gauge of the efficacy of models describing breast cancer mortality) was achieved when shifting from the age model [deviance (D): 285.8; degrees of freedom (df): 52] to models including age + period (D: 105.7; df: 48) or age + cohort (D: 76.3; df: 36). This indicates that both cohort and period effects should be taken into account, even though cohort effects were found to be more important than period effects in terms of model fit. The best fit was found for the full model simultaneously considering the effects of age, period and cohort (D: 55.2; df: 33).

Figure 3 depicts the graphs for the age, period and cohort effects. The age effect grows exponentially, but then dips around the age of 50 years before continuing to rise once more. Risk by birth cohort shows an increase from the 1890 generation up to the 1960 cohort and then a decline in women born after 1960. Period of death values also tended to decline.



View larger version (12K):
[in this window]
[in a new window]
 
Figure 3. Breast cancer mortality in Andalucia. Age, period and cohort effects.

 

    Discussion
 Top
 ABSTRACT
 Introduction
 Patients and methods
 Results
 Discussion
 REFERENCES
 
In Andalucia, breast cancer is still the most frequent cause of death from cancer with 890 deaths annually. After a continuous increase in breast cancer mortality for decades, age-standardized mortality rates have declined slightly since 1993 (–1.6% annual). This decline, while not dramatic, is statistically significant (P <0.05) and is similar to declines in other areas [8]. This trend can be attributed to changes in the risk of dying from breast cancer as people get older (age effects), changes due to methods of classifying cause of death, short-term effects of early detection and the effects of new medical procedures, which can reduce breast cancer mortality near the time of death (period effects), and changes that are associated with the generation to which people belong (cohort effects). These might include reproductive, hormonal and perhaps dietary factor exposures [9].

The age effect grows exponentially, but then dips around the age of 50 years before continuing to rise once more. This phenomenon, known as the ‘Clemmensen hook’, has been observed in different countries with reference to both incidence and mortality, and is interpreted as the overlapping of two curves corresponding to pre- and post-menopausal tumours, respectively.

The period effect shows a levelling off and decline in breast cancer mortality in recent years. This suggests that the decline might be due to mass phenomena that affected the Andalucian population and exerted their effects with a short latency time. Most prominent among the possible candidates for these phenomena are changes in the health-care system and quality of mortality statistics. Changes in how cause of death is registered and coded can influence the cause-specific mortality rates that are reported. The quality of the Spanish cause of death statistics is similar to that of the USA and some European countries [10].

Health-care services reform in Andalucia has been accompanied by major improvements in the health-care system. There has been an increase in the number of physicians per 100 000 population and in the number of primary health-care centres. The use of mammography increased sharply in the mid-1980s, and a breast cancer screening programme was introduced in 1995. All women aged 50–65 years are invited for screening every 2 years. Evidence from the programme itself indicates that the increase in activity was gradual (the screening programme covered 58.7% of the target population in 1999) and that the ‘prevalent’ round of screening will not be completed until the year 2002.

Breast cancer mortality began falling in Andalucia around 1993, before the breast screening programme was introduced. In Andalucian women in the target age group (50–65 years), breast cancer age-adjusted mortality rates fell by 4.3% during the periods 1990–1994 and 1995–1999. This may be the result of screening, although greater falls in mortality also took place in younger women (mortality in 25–49 year olds fell by 14.3% between the periods 1990–1994 and 1995–1999). It is suggested that primary medicine could play a fundamental role in making women more aware of breast cancer, regardless of early detection campaigns [11]. The impact of mammographic screening during the present study is thus likely to be small.

There has been considerable recent debate over whether the reduction in mortality from breast cancer can be attributed to improvements in treatment or screening. Our research indicates that for the period 1993–1999, improvements in treatment have probably played a major role in reducing breast cancer mortality in Andalucia. The fact that the decline in mortality occurred (in the 1990s) later than the increase in survival (reported in the 1980s) [12] reflects the relatively long life expectancy of women dying of breast cancer, who are known to survive on average 7–8 years after diagnosis [13].

Risk by birth cohort shows an increase from the 1890 generation to the 1960 cohort and then declines in women born after 1960. This finding must be interpreted carefully. However, cohort values for the most recent periods are based on fewer age-specific rates and fewer deaths (this is based solely on two points) and are therefore less reliable than central cohort values. Trends in birth cohort effects usually reflect risk factor trends, so this decline is surprising because trends in most known or suspected risk factors would seem to predict an increasing breast cancer risk in these cohorts.


    FOOTNOTES
 
* Correspondence to: Dr A. Cayuela, Hospitales Universitarios Virgen del Rocío, Edificio de Laboratorios, Unidad de Apoyo a la Investigación, Avenida de Manuel Siurot S/N, 41013-Sevilla, Andalucia, Spain. Tel: +34-955013293; Fax: +34-955013292; E-mail: aurelio.cayuela.sspa@juntadeandalucia.es Back


    REFERENCES
 Top
 ABSTRACT
 Introduction
 Patients and methods
 Results
 Discussion
 REFERENCES
 
1. Hermon C, Beral B. Breast cancer mortality rates are leveling off or beginning to decline in many western countries: analysis of time trends, age–cohort and age–period models of breast cancer mortality in 20 countries. Br J Cancer 1996; 73: 955–960.[ISI][Medline]

2. Levi F, Lucchini F, Negri E, La Vecchia C. The decline in cancer mortality in the European Union, 1988–1996. Eur J Cancer 2000; 36: 1965–1968.[CrossRef][ISI][Medline]

3. Bray F, Sankila R, Ferlay J, Parkin DM. Estimates of cancer incidence and mortality in Europe in 1995. Eur J Cancer 2002; 38: 99–166.[CrossRef][ISI][Medline]

4. La Vecchia C, Negri E, Levi F et al. Cancer mortality in Europe: effects of age, cohort of birth and period of death. Eur J Cancer 1998; 34: 118–141.[CrossRef][ISI][Medline]

5. Cayuela A, Rodríguez S. Análisis de la mortalidad en Andalucía 1975–1992. Sevilla: Instituto de Estadística de Andalucía 2000.

6. Waterhouse J, Muir CS, Shanmugaratnam K, Powell J. Cancer Incidence in Five Continents, Vol. IV. IARC Scientific Publication no. 42. Lyon: IARC 1982.

7. Decarli A, La Vecchia C. Age, period and cohort models: review of knowledge and implementation in GLIM. Rev Stat App 1987; 20: 397–409.

8. Osmond C, Gardner MJ. Age, period and cohort models applied to cancer mortality rates. Stat Med 1982; 1: 245–259.[Medline]

9. Gardner MJ, Osmond C. Interpretation of time trends in disease rates in the presence of generation effects. Stat Med 1984; 3: 113–130.[ISI][Medline]

10. Fernandez E, Gonzalez JR, Borras JM et al. Recent decline in cancer mortality in Catalonia (Spain). A joinpoint regression analysis. Eur J Cancer 2001; 37: 2222–2228.[CrossRef][ISI][Medline]

11. Prieto F, Serra Ll, La Vecchia C et al. Mortality trends and past and current dietary factors of breast cancer in Spain. Eur J Epidemiol 1996; 12: 141–148.[ISI][Medline]

12. García F, Bolumar F, Peris R. Quality of death certificates in Valencia, Spain. Am J Public Health 1989; 79: 1352–1354.[Abstract]

13. Villar A, Candel MT, Delgado F et al. Extensión del cáncer de mama en el momento del diagnostico inicial. Med Clin (Barc) 1992; 98: 285–289.[ISI][Medline]

14. Sant M, Capocaccia R, Verdecchia A et al. Survival of women with breast cancer in Europe: variation with age, year of diagnosis and country. Int J Cancer 1998; 77: 679–683.[CrossRef][ISI][Medline]

15. Sant M, Capocaccia R, Coleman MP et al. Cancer survival increases in Europe, but international differences remain wide. Eur J Cancer 2001; 37: 1659–1667.[CrossRef][ISI][Medline]





This Article
Abstract
Full Text (PDF)
E-letters: Submit a response
Alert me when this article is cited
Alert me when E-letters are posted
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Disclaimer
Request Permissions
Google Scholar
Articles by Cayuela, A.
Articles by Gili, M.
PubMed
PubMed Citation
Articles by Cayuela, A.
Articles by Gili, M.