br We hypothesized increasing age e
We hypothesized increasing age eﬀects on CRC incidence across all observed countries, VH-298 eﬀects that correspond to changes of the macro-environment of the populations, and increasing cohort eﬀects for the generations who were increasingly exposed to more westernized living environments.
2. Materials and methods
2.1. Study design and data sources
We conducted a population-based study using APC modeling, and included all men and women in the population over the study period for each country. Age- and sex-specific annual number of new CRC cases and annual population data are used to estimate the incidence. Age- and sex-specific annual population and incidence data for each country were obtained from the Cancer Incidence in Five Continents Time Trends (CI5plus) databases of the World Health Organization (WHO) International Agency for Research on Cancer (IARC) . CI5plus provides access to various databases containing information on the occurrence of cancer worldwide held and managed by the Section of Cancer Surveillance of IARC. While no large-scale database can be perfect, quality control procedures are instituted to ensure the quality of the data, identify the areas and degree of imperfection, and assist in data interpretation . The CI5plus system also ensures that the data items recommended for registration to be kept minimal, with emphasis on the quality rather than the volume of information . Moreover, intensive and extensive use of the registry's data tends to maintain and improve their quality .
From the data source, we chose representative Asian countries/re-gions including Japan, China (Hong Kong, Shanghai), Singapore and India and compared them against representative western countries in-cluding the UK, the US and Australia. We selected these countries for their representativeness of the diﬀerent models of the epidemiologic transition . While the UK, the US and Australia represent well-es-tablished Western populations, the Asian populations represent the accelerated epidemiologic transition model. Shanghai is a representa-tion of the Chinese populations that are recently undergoing ac-celerated socio-economic development; Hong Kong and Singapore re-present mature Asian economies; and Japan has the most established economy with high incidence of the disease. On the other hand, India with a low CRC incidence is included for comparative purposes. Due to data availability and completeness, selected regions of some countries were included, while population-wide data were used for all the others. We included Merseyside and Cheshire, North Western, South and Western Regions, Birmingham and West Midlands Region, Yorkshire of England and Scotland for the UK; the Surveillance, Epidemiology, and
End Results (SEER) Regions for the US; New South Wales, Queensland, South Australia, Tasmania, Victoria and Western Australia for Australia; Miyagi and Osaka Prefectures for Japan; and Mumbay and Chennai for India. Details of the registries included in this study are also shown in Supplementary Table 1.
2.2. Statistical analysis
CRC incidence rates were presented per 100,000 people, and di-rectly age-standardized to the WHO World Standard Population  for international comparison. We grouped the incidence data into five-year age groups in order to control for potential inconsistency and fluctuating data noise. We maximized the length of the study period; i.e., to maximize the number of five-year age groups and the number of five-year periods according to each country’s data availability. We used data of 1983–2007 for the UK, 1978–2007 for the USA, 1983–2007 for Australia, 1978–2007 for Japan, 1983–2007 for Hong Kong, 1988–2007 for Shanghai, 1968–2007 for Singapore and 1983–2007 for India. Cases below the age of 20 were excluded due to low-to- zero incidence in younger ages. Number of CRC cases over the last four five-year periods in each country is shown in Supplementary Table 2. For the observed periods, we modeled CRC incidence using Poisson APC regression models to decompose rates over time by age, period and cohort, with 95% confidence intervals.