Dr. Winston (Zhenguo) Qiu, Surveillance & Reporting C-MORE
Canproj: the Mixture of Cancer Projection Methods Based on Age, Period, and/or Cohort Models
Cancer incidence/mortality can be projected by extrapolating past trends using statistical models. Trends in cancer incidence/mortality may be described as trends over age at diagnosis, year of diagnosis (period), and/or year of birth (cohort) (Osmond, 1985). Cancer rates usually increase with age. Period effects are related to events that quickly change incidence or mortality with the same order of magnitude regardless of the age group. Cohort effects involve risk factors that are shared by a specific generation as they age together. Based on the age-period-cohort model (Clayton & Schifflers, 1987), the “Nordpred” method (Møller, 2004) was developed for cancer projection. In 2010-2012, the Cancer Projection Analytical Network (C-Proj) Working Team, funded by Canadian Partnership Against Cancer (CPAC), developed the approach, a mixture of cancer projection methods used in recent three decades and the corresponding R-package “Canproj”, in which the best fitted model for a given data can be selected and used for projection. In Canproj, the candidate models are: age-only, age-period (including common trend and age-specific trend), age-cohort and age-period-cohort; negative-binomial distribution may replace the Poisson distribution when over-dispersion appears; and “log”, “sqrt” and “power 5” link functions can be applied. Validation analyses using Alberta and Nova Scotia Cancer Registry data have shown that the Canproj methods outperform other traditional used approaches, such as the five-year average method, Poisson regression method (Dyba T, Hakulinen & Päivärinta, 1997), joinpoint regression method (Kim, Fay & Feuer et al. 2000), the polynomial regression & natural spline methods (Carstensen, 2007) and the Bayesian Markov chain Monte Carlo methods with different prior settings (Bray, 2002; Clèries, Ribes & Esteban et al, 2006).
 Osmond C. Using age, period and cohort models to estimate future mortality rates. Inter J Epidem, 1985, 14: 124 – 9.
 Clayton D & Schifflers E. Models for temporal variation in cancer rates. I. Age-period and age cohort models. Stat Med, 1987a, 6: 449 – 467.
 Clayton D & Schifflers E. Models for temporal variation in cancer rates. II. Age-period-cohort models. Stat Med, 1987b, 6: 469-481.
 Møller B. Prediction of cancer incidence—Methodological considerations and trends in the Nordic countries 1958 – 2022. Dissertation, Faculty of Medicine, University of Oslo, Unipub AS, Oslo, 2004.
 Dyba T, Hakulinen T & Päivärinta L. A simple non-linear model in incidence prediction. Stat Med, 1997, 16: 2297 – 309.
 Kim HJ, Fay MP & Feuer EJ et al. Permutation tests for joinpoint regression with applications to cancer rates. Stat Med, 2000, 19: 335 – 51 (correction: 2001, 20:655).
 Carstensen B. Age-period-cohort models for the lexis diagram. Stat Med, 2007, 26(15): 3018 – 45.
 Bray I. Application of Markov chain Monte Carlo methods to projecting cancer incidence and mortality. J R Stat Soc Ser C, Appl Stat, 2002, 51: 151 – 64.
 Clèries R, Ribes J & Esteban L et al. Time trends of breast cancer mortality in Spain during the period 1977–2001 and Bayesian approach for projections during 2002–2016. Annals of Oncology, 2006, 17: 1783 – 91.
Pretest Time: Start: 12:45 End: 1:00 MST/MDT
*CAL – Holy Cross – Rm G020A/B
CAL – Richmond Road Diagnostic Treatment Centre – Rm 3005
EDM – Cross Cancer – Rm 4134
EDM – Sun Life – Brdrm 1538