Age-Period-Cohort Models in Cancer Surveillance Research: Ready for Prime Time?

Authors' Affiliation: National Cancer Institute, Division of Cancer Epidemiology and Genetics, Bethesda, Maryland.
Cancer Epidemiology Biomarkers & Prevention (Impact Factor: 4.13). 06/2011; 20(7):1263-8. DOI: 10.1158/1055-9965.EPI-11-0421
Source: PubMed


Standard descriptive methods for the analysis of cancer surveillance data include canonical plots based on the lexis diagram, directly age-standardized rates (ASR), estimated annual percentage change (EAPC), and joinpoint regression. The age-period-cohort (APC) model has been used less often. Here, we argue that it merits much broader use. First, we describe close connections between estimable functions of the model parameters and standard quantities such as the ASR, EAPC, and joinpoints. Estimable functions have the added value of being fully adjusted for period and cohort effects, and generally more precise. Second, the APC model provides the descriptive epidemiologist with powerful new tools, including rigorous statistical methods for comparative analyses, and the ability to project the future burden of cancer. We illustrate these principles by using invasive female breast cancer incidence in the United States, but these concepts apply equally well to other cancer sites for incidence or mortality.

Download full-text


Available from: William F Anderson, Sep 09, 2014
  • Source
    • "The EAPC can be estimated through an age-drift model, and we can easily build a confidence interval for the EAPC using the standard error of the drift [13]. The age-drift model is a simplified version of the age-cohort model where the cohort effect is parameterized using only the linear component. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Tobacco is currently the largest risk factor for cancers of the lung, lip/oral cavity/pharynx (LOCP) and esophagus. Variations in tobacco consumption over time have led to changes in cancer incidence in the general population. Data on the incidence of cancers at these sites in adults aged 20-44 years old are scarce. Our objective was to provide estimates of incidence trends for these cancers in France among this age group over the last 30 years. Observed incidence data over the period 1982-2010 for the 20-44 age group were provided from six cancer registries (eight for esophagus) covering approximately 6% of the French population. Age-period-cohort models were used on the observed period, and estimates of cancer incidence for France in 2012 were provided on the basis of short-term predictions. In men, a sharp decline was observed over time for LOCP and esophageal cancers, while lung cancer saw only a slight decline. In women, a large increase was seen in lung cancer incidence, while LOCP cancer incidence did not vary significantly. Smoking behaviors among adults aged 20-44 impact incidence trends in cancers of the lung, LOCP and esophagus, although other factors are involved, particularly in LOCP and esophageal cancers. Our results highlight the importance of preventative efforts which particularly target women aged 20-44. Efforts to curb tobacco smoking in men should also be pursued. Copyright © 2015 Elsevier Ltd. All rights reserved.
    07/2015; 39(5). DOI:10.1016/j.canep.2015.07.001
  • Source
    • "The period when death occurs may also reflect important changes in factors that affect mortality (e.g., introduction of new treatments). The cohort effect may indicate changes in exposures that are particular to specific generations [15,16]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background Individuals who live in rural areas are at greater risk for brain cancer, and pesticide exposure may contribute to this increased risk. The aims of this research were to analyze the mortality trends and to estimate the age-period-cohort effects on mortality rates from brain cancer in two regions in Rio de Janeiro, Brazil. Methods This descriptive study examined brain cancer mortality patterns in individuals of both sexes, >19 years of age, who died between 1996 and 2010. They were residents of a rural (Serrana) or a non-rural (Metropolitan) area of Rio de Janeiro, Brazil. We estimated mortality trends using Joinpoint Regression analysis. Age-period-cohort models were estimated using Poisson regression analysis. Results The estimated annual percentage change in mortality caused by brain cancer was 3.8% in the Serrana Region (95% confidence interval (CI): 0.8–5.6) and -0.2% (95% CI: -1.2–0.7) in the Metropolitan Region. The results indicated that the relative risk was higher in the rural region for the more recent birth cohorts (1954 and later). Compared with the reference birth cohort (1945–49, Serrana Region), the relative risk was four times higher for individuals born between 1985 and 1989. Conclusions The results of this study indicate that there is an increasing trend in brain cancer mortality rates in the rural Serrana Region in Brazil. A cohort effect occurred in the birth cohorts born in this rural area after 1954. At the ecological level, different environmental factors, especially the use of pesticides, may explain regional disparities in the mortality patterns from brain cancers.
    BMC Cancer 05/2014; 14(1):320. DOI:10.1186/1471-2407-14-320 · 3.36 Impact Factor
  • Source
    • "Age-period-cohort (APC) models are a statistical approach to try and separate the three etiologically important, but interdependent, time variables to provide further insights into the underlying forces influencing the burden of disease. These methods have been successfully employed over many years [8]. The linear dependence of age, time period of diagnosis or death, and year of birth creates the identification problem in statistical APC models but this is equally present, although less easily seen, in the accepted classic graphical methods of cohort analysis [7]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background. Non-Hodgkin lymphoma (NHL) incidence rates have increased considerably in New Zealand. Methods. Incidence and mortality rates for NHL from 1981 to 2010 were calculated. Trends in age-specific rates were analysed and age-period-cohort models fitted to explore generation-specific changes in incidence and mortality. Results. NHL incidence increased by 67% for men and 74% for women between the 1981-1985 and 2006-2010 time periods in New Zealand. For women born about 1936 and men born about 1946, NHL incidence and mortality have diverged suggesting an improved prognosis for recent generations. Conclusion. The strong generation effects suggest that an exposure before 25 years of age is of major importance in determining the lifetime risk of NHL in New Zealand. NHL incidence rates in New Zealand will continue to increase in the future and probably more in females than males, as generations with increased risk age. Current hypotheses for the cause of NHL do not explain the trends observed. A decline in the prevalence of a protective factor may have also contributed to these trends. Examination of trends for subtypes of NHL and innovative testable hypotheses that may explain these trends are needed.
    Journal of Cancer Epidemiology 04/2014; 2014(6):315378. DOI:10.1155/2014/315378
Show more