Comment on 'Estimating average annual per cent change in trend analysis' Reply

Office of Inspector General, U.S. Department of Veterans Affairs, Washington, DC, USA.
Statistics in Medicine (Impact Factor: 1.83). 12/2009; 28(29):3670-82. DOI: 10.1002/sim.3733
Source: PubMed


Trends in incidence or mortality rates over a specified time interval are usually described by the conventional annual per cent change (cAPC), under the assumption of a constant rate of change. When this assumption does not hold over the entire time interval, the trend may be characterized using the annual per cent changes from segmented analysis (sAPCs). This approach assumes that the change in rates is constant over each time partition defined by the transition points, but varies among different time partitions. Different groups (e.g. racial subgroups), however, may have different transition points and thus different time partitions over which they have constant rates of change, making comparison of sAPCs problematic across groups over a common time interval of interest (e.g. the past 10 years). We propose a new measure, the average annual per cent change (AAPC), which uses sAPCs to summarize and compare trends for a specific time period. The advantage of the proposed AAPC is that it takes into account the trend transitions, whereas cAPC does not and can lead to erroneous conclusions. In addition, when the trend is constant over the entire time interval of interest, the AAPC has the advantage of reducing to both cAPC and sAPC. Moreover, because the estimated AAPC is based on the segmented analysis over the entire data series, any selected subinterval within a single time partition will yield the same AAPC estimate--that is it will be equal to the estimated sAPC for that time partition. The cAPC, however, is re-estimated using data only from that selected subinterval; thus, its estimate may be sensitive to the subinterval selected. The AAPC estimation has been incorporated into the segmented regression (free) software Joinpoint, which is used by many registries throughout the world for characterizing trends in cancer rates.

Download full-text


Available from: Limin X Clegg,
47 Reads
  • Source
    • "" (Clegg, Hankey et al. 2009)used segmented regression while analyzing the trend using the annual average percent change in rat population and denoted it as (sAPCs). (Clegg, Hankey et al. 2009) assumed that the change in the population of rats was constant within each time partition but varied between (among) different time partitions. Also different groups (racial subgroups) might have different transition points and thus different constant rates in different partitions at a common time interval (past 10 years) were assumed to be held. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The modeling non-linear behavior between independent variables and response variable remained a challenging task for the researchers. We consider precipitation data set for two monitoring stations for twenty seven years period during monsoon. Generalized linear models, Generalized additive model and piecewise regression models are used to find appropriate model to describe the characteristics of dependent variable related to independent. The results of these models are compared by means of cross validation and coefficient of determination. It is observed that generalized linear model perform poorly then generalized additive model and piecewise regression model. Since generalized additive model is non-parametric and is more flexible to model non-linear behavior therefore it also explains more variation of dependent variable then piecewise regression models. However,the difference between generalized additive model and piecewise regression model becomes negligible if segmented points are observed from generalized additive model and then those segmented points are used to estimate piecewise regression.
  • Source
    • "We calculated the age-standardised incidence and mortality rates in HK according to the World Standard Population in 2000. Non-linearities in trends of varying time periods were characterised using segmented annual percent change from segmented regression (Clegg et al, 2009) or join-point regression analysis (Kim et al, 2000). We modelled breast cancer incidence and mortality using APC models (details in Supplementary Information 1) (Clayton and Schifflers, 1987; Holford, 1991, 1992). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background: Projections of future trends in cancer incidence and mortality are important for public health planning. Methods: By using 1976-2010 data in Hong Kong, we fitted Poisson age-period-cohort models and made projections for future breast cancer incidence and mortality to 2025. Results: Age-standardised breast cancer incidence (/mortality) is projected to increase (/decline) from 56.7 (/9.3) in 2011-2015 to 62.5 (/8.6) per 100,000 women in 2021-2025. Conclusions: The incidence pattern may relate to Hong Kong's socio-economic developmental history, while falling mortality trends are, most likely, due to improvements in survival from treatment advancement and improved health service delivery.
    British Journal of Cancer 10/2014; 112(1). DOI:10.1038/bjc.2014.532 · 4.84 Impact Factor
  • Source
    • "To assess mortality trends by educational level, we estimated the annual percent change in mortality (APC) based on a Poisson model that incorporated an interaction between educational level and year. The APC measures the average rate of change in the mortality rate per year (Clegg et al., 2009). At a second stage, we estimated rate ratios (RR) of mortality by educational level. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Non-communicable diseases have become the leading cause of death in middle-income countries, but mortality from injuries and infections remains high. We examined the contribution of specific causes to disparities in adult premature mortality (ages 25-64) by educational level from 1998 to 2007 in Colombia. Data from mortality registries were linked to population censuses to obtain mortality rates by educational attainment. We used Poisson regression to model trends in mortality by educational attainment and estimated the contribution of specific causes to the Slope Index of Inequality. Men and women with only primary education had higher premature mortality than men and women with post-secondary education (RRmen=2·60, 95% confidence interval [CI]:2·56, 2·64; RRwomen=2·36, CI:2·31, 2·42). Mortality declined in all educational groups, but declines were significantly larger for higher-educated men and women. Homicide explained 55·1% of male inequalities while non-communicable diseases explained 62·5% of female inequalities and 27·1% of male inequalities. Infections explained a small proportion of inequalities in mortality. Injuries and non-communicable diseases contribute considerably to disparities in premature mortality in Colombia. Multi-sector policies to reduce both interpersonal violence and non-communicable disease risk factors are required to curb mortality disparities.
    Preventive Medicine 03/2014; 64. DOI:10.1016/j.ypmed.2014.03.018 · 3.09 Impact Factor
Show more