br First the application of screening strategies and colonos
First, the application of screening strategies and colonoscopic polypectomy may reduce CRC in the West. Nevertheless, since CRC screening is usually recommended to those aged 50 years old or above [10,34], the impact of screening towards the current younger age groups (who by definition also belong to the younger cohorts) would be smaller than that towards the older age groups/generations. The al-ternative explanation is related to the inherent diﬀerence in terms of macro-environmental exposures between the Western and the Asian populations. Westernized lifestyle factors (including low CAY10683 and ve-getable intake, high-fat diet, overweight and obesity, sedentary
Fig. 2. Parameter estimates and 95% confidence intervals of age eﬀects for colorectal cancer incidence in (a) UK, (b) USA, (c) Australia, (d) Japan, (e) Hong Kong, (f) Shanghai, (g) Singapore, and (h) India.
Fig. 3. Parameter estimates and 95% confidence intervals of period (triangles) and cohort (circles) eﬀects for colorectal cancer incidence in (a) UK, (b) USA, (c) Australia, (d) Japan, (e) Hong Kong, (f) Shanghai, (g) Singapore, and (h) India.
lifestyle, tobacco use and alcohol consumption) are well-established risk factors of CRC . The diﬀerential observation between the Wes-tern and the Asian populations may imply that despite continuous socio-economic development to adopt more western-like lifestyles, there may still be deep-rooted traditions among these Asian populations that may protect them against CRC, especially for those generations (i.e., 1950s cohorts) who still grew up in a macro-environment that largely consumed traditional non-Western diet . Further analytical epidemiological studies are needed to clarify the specific dietary habits that contribute to this eﬀect.
However, an upturn of cohort eﬀects among the younger 1960s–1970s generations in these Asian populations (Japan, Hong Kong, Shanghai and Singapore) is also observed. In fact, the timing of the upturn for these younger cohorts coincide with their greater early life exposures to westernized diet and lifestyle that came with rapid socio-economic development. This upturn is especially obvious for Japan, which experienced socio-economic development earliest in Asia. Our findings are consistent with the life course framework to under-standing the risk factors of lifestyle-related chronic, non-communicable diseases, which postulates that exposures to living conditions and macro-environmental changes throughout the life course are important driver of changes in population health. Early life exposure is considered to be a critical period that impacts the risk of non-communicable dis-eases during later life . Despite wide confidence intervals toward the younger cohorts for these Asian populations (Hong Kong, Shanghai and Singapore), the upturn of cohort eﬀects implies that if the life course explanation holds true, there will be an increased CRC risk for these cohorts as they grow older in the future. A longer period of ob-servation will be needed to confirm this notion.
There are limitations to our study. First, routinely collected data has its limitation by nature – since CRC incidence increases with age, many people of younger age may not have lived up to the normal age range that the disease is typically developed yet. Since the cohort eﬀect of younger generations are based on incidence in the younger age groups, there are more uncertainties towards the more recent birth cohorts as shown by the overlapping confidence intervals. A longer period of data collection would help clarify the eﬀects on the more recent generations. Also, the observed age range and lengths of period varied across dif-ferent populations under study. Even so, the time periods are long en-ough to generate reliable and comparable cohort estimates that gen-erally went back to the turn of the 20th century. Second, the three components of age, period and cohort are linearly dependent on each other, making the full model with all three components non-identifiable without an additional reference constraint. Even with an additional constraint, only second-order changes in slope can be interpreted. To confirm our results, we also conducted APC modeling using the alter-native method of partial least squares regressions , and found the same inflection points for all three eﬀects (data not shown). Third, while we cannot rule out diﬀerences in reporting across years, hence a possible presence of the period eﬀect, there is no reason to believe that these diﬀerences would occur in a systematic manner. Moreover, the present study used the most up-to-date data leading up to year 2007 for analysis. Since the lengths of follow-up period are longer for the older birth cohorts, there is stronger power and confidence to the findings of these older cohorts, thus enhancing robustness of our findings for in-ternational comparison. Last, this study treated CRC as one disease cluster, while in reality risk factors may diﬀerentially aﬀect the various sub-sites (colon vs rectum, left colon vs right colon). Any future study would benefit from analysis by the diﬀerent sub-sites. Nevertheless, the present study is a necessary first step to delineate the complex etiolo-gical risk factors of CRC with a global perspective.