Demographic trends in the incidence of young‐onset colorectal cancer: a population‐based study

Background Evidence is emerging that the incidence of colorectal cancer is increasing in young adults, but the descriptive epidemiology required to better understand these trends is currently lacking. Methods A population‐based cohort study was carried out including all adults aged 20–49 years diagnosed with colorectal cancer in England between 1974 and 2015. Data were extracted from the National Cancer Registration and Analysis Service database using ICD‐9/10 codes for colorectal cancer. Temporal trends in age‐specific incidence rates according to sex, anatomical subsite, index of multiple deprivation quintile and geographical region were analysed using Joinpoint regression. Results A total of 56 134 new diagnoses of colorectal cancer were analysed. The most sustained increase in incidence rate was in the group aged 20–29 years, which was mainly driven by a rise in distal tumours. The magnitude of incident rate increases was similar in both sexes and across Index of Multiple Deprivation quintiles, although the most pronounced increases in incidence occurred in the southern regions of England. Conclusion Colorectal cancer should no longer be considered a disease of older people. Changes in incidence rates should be used to inform future screening policy, preventative strategies and research agendas, as well as increasing public understanding that younger people need to be aware of the symptoms of colorectal cancer.


Introduction
Colorectal cancer is a major cause of cancer-related mortality and is the third most common cause of cancer death in the UK 1,2 . Advances in the surgical and oncological management of colorectal cancer are the most likely explanation for the UK age-standardized mortality rate decreasing from 49 to 27 per 100 000 person-years over the past 40 years 2 .
Despite age-standardized incidence rates remaining static in the UK, as well as in other nations with a high human development index 3 , there is increasing evidence that incidence rates are increasing in adults aged less than 50 years. A US study 4 , using Surveillance, Epidemiology, and End Results data, revealed a doubling in the incidence rate of both colonic and rectal cancers among patients aged between 20 and 54 years since 1974. Similar findings have been demonstrated in cohorts from Canada 5,6 , Australia 7 , New Zealand 8 and, most recently, Europe 9 , suggesting that the underlying risk of colorectal cancer is increasing in young people.
Although men are well recognized to have a higher incidence of colonic and rectal cancer in older age groups, there is little difference in the incidence rates between men and women aged less than 40 years 10,11 . UK data have shown that men have a higher proportion of rectal tumours, but that women have a higher proportion of right-sided tumours 12 . However, data on anatomical subsite have not been linked to age-specific incidence trends in the UK population. Data from North America 4,6 suggest that incidence rate increases have been driven by an increase in distal tumours, whereas European data 9 suggest that incidence rate increases have been more pronounced for colonic cancer.
Socioeconomic status (SES) is associated with several important colorectal cancer risk factors 13 -15 . In the UK, data from Northern Ireland 16 have shown no difference in age-standardized incidence between deprivation deciles, unlike in Scotland, where men from more deprived areas have an increased incidence of colorectal cancer with evidence of an increasing deprivation gap over time 17,18 . Previous studies have focused on SES as a risk factor for colorectal cancer incidence, but this has never been analysed in the context of recent changes in age-specific incidence trends in young adults. Significant variations in the burden of disease exist between the nine regions of England, including variation in the age-standardized rate of years of life lost to colorectal cancer 19,20 . Understanding whether there is a socioeconomic and regional variation in incidence rate trends in the young population could help elucidate potential aetiological factors.
Although data from the UK have been incorporated in recent Europe-wide population-based studies 9 , a more detailed description of the epidemiology underlying the recent increase in colorectal cancer incidence in young adults is required. This is vital, as young adults typically present with more advanced tumours that carry a poorer prognosis. A more thorough knowledge of the descriptive epidemiology would help inform future preventative strategies. Therefore, the aim of this study was to determine temporal trends in incidence of colorectal cancer stratified by sex, anatomical subsite in the colorectum, SES and geographical region of England.

Methods
This study is reported according to the STROBE guidelines for epidemiological studies 21 . Data were obtained on all patients diagnosed with colorectal cancer aged 20 years and above from 1974 to 2015 using data from the National Cancer Registration and Analysis Service (NCRAS) (request ID: ODR1718_067). NCRAS is a UK-wide partnership operated by Public Health England to collect data on all types of cancer, including colorectal cancer, occurring in the English population.
Mid-year population estimates were obtained from the Office for National Statistics (ONS) to provide population data stratified by age. Mid-year population estimates in conjunction with the number of new diagnoses were used to calculate age-specific incidence density rates per 100 000 person-years, referred to hereafter as the age-specific incidence rates, for each age group using the following.
Age-specific incidence rate

=
No. of new cases in age group Mid-year population estimate of age group The European Standard Population 2013 (ESP 2013) was then used to derive age-standardized incidence rates for colonic and rectal cancer for the overall data set (20-49 years), in accordance with the methodology for direct standardization of the ONS 22 : Age-standardized incidence rate = ∑ (ESP of age group × age-specific rate) ∑ ESP of age group Colorectal cancer cases were further stratified by sex (using sex-specific population estimates from the ONS as above), anatomical subsite (proximal -caecum to descending colon; distal -sigmoid to rectum), geographical region (using region-based population estimates from the ONS from 1981 onwards) and Index of Multiple Deprivation (IMD) quintile (from 2001 onwards). IMD is an area-based metric that combines weighted information from seven domains: income (weighting 22⋅5 per cent), employment (22⋅5 per cent), education (13⋅5 per cent), health (13⋅5 per cent), crime (9⋅3 per cent), barriers to housing and services (9⋅3 per cent) and living environment (9⋅3 per cent). Lower-layer super output areas (LSOAs; 32 844 in England) are given a value based on these domains. IMD quintiles were calculated by ranking all LSOAs from most to least deprived and then splitting this ranking into five equal groups; each quintile has 20 per cent of the ranked areas. Age-specific incidence rates for men and women: a 20-29 years, b 30-39 years and c 40-49 years. Incidence rates per 100 000 are shown for each year; the range of values plotted on the y-axis varies according to magnitude of incidence rate, to highlight trends. Plotted lines indicate annual percentage changes (APCs). *Significant change in APC versus 0 (P < 0⋅050) using the permutation model of logarithmically transformed data. Age-specific incidence rates for proximal and distal tumours: a 20-29 years, b 30-39 years and c 40-49 years. Incidence rates per 100 000 are shown for each year; the range of values plotted on the y-axis varies according to magnitude of incidence rate, to highlight trends. Plotted lines indicate annual percentage changes (APCs). *Significant change in APC versus 0 (P < 0⋅050) using the permutation model of logarithmically transformed data. Data from 1974 to 1978 were excluded as it was not possible to map ICD-8 codes completely to later versions. of temporal trends in age-specific incidence rates according to sex, anatomical site, IMD quintile and geographical region. Permutation analysis of the log-transformed incidence rates was used to fit a series of joined lines with a minimum of 0 and a maximum of 5 join points. A series of comparisons among fitted models ranging from 0 to 5 join points was then undertaken to select the best-fit model. This procedure allowed estimation of the annual percentage change (APC) in incidence. The squared correlation coefficient (R 2 ) was used to estimate the goodness of fit of the Joinpoint regression models to provide an indication of the extent of agreement between modelled and observed values. Inspection of residuals under the models presented here did not give cause for concern; standard errors appeared homoscedastic, free from serial correlation and without any unduly influential observations.
Age-period-cohort modelling (National Cancer Institute; Age Period Cohort web tool; https://analysistools.nci .nih.gov/apc) was used to assess the independent effects of age, time period and cohort on colorectal cancer incidence rates 24 . This was performed for all adults aged over 20 years. Three 10-year age groups (20-29, 30-39 and 40-49 years for Joinpoint regression modelling) and four 10-year periods (1976-1985, 1986-1995, 1996-2005, 2006-2015) were used for age-period-cohort modelling as it was necessary to have age and time-period groups covering an equal timespan. Therefore, there were 11 birth cohorts starting in 1886 through to 1986 in 10-year bands. Reference values for the age-period-cohort model were chosen arbitrarily from the first cohort analysed (1976)(1977)(1978)(1979)(1980)(1981)(1982)(1983)(1984)(1985). Data presented from this model were shown as incidence rate ratios (IRRs) with 95 per cent confidence intervals to assess cohort effects. Local drift was estimated by examining age-specific net APC in incidence rates.

Age-standardized trends according to socioeconomic status
The age-standardized incidence rates for distal cancers increased more rapidly than those for proximal cancers in all IMD quintiles, except quintile 2 ( Fig. S3a-e, supporting information). There was no significant difference in the magnitude of incidence rate increases across the quintiles for either proximal (P = 0⋅110) or distal (P = 0⋅230) cancers.

Age-standardized trends according to geographical region
In 1985, age-standardized incidence rates of proximal cancers among 20-49 year olds were decreasing across all Annual percentage change (APC) in proximal colorectal cancer incidence rates calculated for each region of England using Joinpoint regression analysis: a 1985, b 1995, c 2005 and d 2015. *Significant change in APC versus 0 (P < 0⋅050) using the permutation model of logarithmically transformed data.
Annual percentage change (APC) in distal colorectal cancer incidence rates calculated for each region of England using Joinpoint regression analysis: a 1985, b 1995, c 2005 and d 2015. *Significant change in APC versus 0 (P < 0⋅050) using the permutation model of logarithmically transformed data.

Discussion
This large study based on a single, national population registry describes detailed epidemiological changes in colorectal cancer incidence in a young adult population. The finding that the incidence of colorectal cancer is increasing rapidly in young adults supports recent findings from other nations with a high human development index 4 -9,25 . Rapid increases were observed in adults aged 20-39 years, which appear to be driven by increases in the rate of distal tumours. Incidence rate increases in the English population appear to be similar in both sexes and across all socioeconomic groups. Importantly, incidence rates have increased the fastest in the southern regions of England, particularly in the South West, where the incidence of distal cancers is now increasing by more than 10 per cent each year. A substantial birth cohort effect was observed, with dramatic increases in IRRs from the mid-1960s onwards. This is similar to the observations in North American studies 4,6 , although IRR increases in these studies appeared to have occurred in birth cohorts born approximately 15 years earlier. This suggests that any exposure to underlying risk factors may have occurred earlier in the North American population. Tumours in young adults are thought to be sporadic 26 , with environmental factors likely to play a significant causative role. The rising incidence of colorectal cancer in young adults coincides with several environmental changes, most notably increasing childhood and adult obesity rates 27 . It is recognized that obesity in early life leads to an increased risk of developing colorectal cancer 28,29 . Therefore, the increases in colorectal cancer incidence in young men and women may reflect the recent obesity trends in the UK, where prevalence rates among adults aged 35-54 years have increased from 15⋅4 to 26⋅3 per cent in men, and from 17⋅9 to 24⋅5 per cent in women, between 1993 and 2004 30 .
The more pronounced increase in the incidence rate of distal compared with proximal tumours contrasts with findings from recent European data 9 , but is similar to the results from several North American studies 4,6,31 . Although risk factors associated with an increased risk of colorectal cancer have been identified, the strength of their association with tumour development at individual sites in the colorectum remains unclear. Differences in the way environmental factors promote tumorigenesis at various sites in the colorectum suggest that proximal and distal tumours may be biologically distinct entities 32 ; this may explain why the incidence in distal tumours among this English cohort has increased more rapidly. The biological differences in earlyversus late-onset colorectal cancer have been explored in several studies. A recent large cohort study 33,34 characterizing the clinical and molecular features of early-onset colorectal cancer demonstrated enrichment of certain phenotypes, such as consensus molecular subtype 1 (CMS1) in distal tumours among adults aged less than 50 years. Other work 35,36 has shown low levels of microsatellite instability in colorectal cancer in young adults. Additionally, there is a prevalence of mutations in genes such as β-catenin 35,37 and KRAS 38 . Interestingly, the combination of altered environmental exposures and different tumour biology suggests that young-adult colorectal cancer may be a different disease from later-onset disease.
This study provided no evidence of an association between SES and the rate of increase in incidence of either proximal or distal tumours, in contrast to previous studies 16 -18 where higher incidence rates were observed in more deprived groups. Although factors associated with an increased risk of colorectal cancer, such as obesity, low-fibre diet and reduced physical activity, are known to be associated with lower SES 1 -15 , changes in obesity prevalence trends are actually similar between socioeconomic groups 39 ; this may partly explain the lack of association between SES and colorectal cancer incidence rate increases in the present study. Furthermore, obesity is one of many risk factors associated with the development of colorectal cancer, and is itself caused by several complex societal, genetic and environmental interactions. It is perhaps not surprising that understanding the causative effects of single environmental risk factors is challenging 40 .
Geographical inequalities in health have been well characterized in England. Incidence rates of all cancers are higher in the north of England than in the south, although there is minimal variation in colorectal cancer incidence by region 41 . Recent incidence rate increases in colorectal cancer were observed across all English regions in the present study, although the most marked increases occurred in the south. It is difficult to explain why incidence rates are increasing more rapidly in young adults in the south given that risk factors such as obesity are increasing faster in northern regions 39 . It is important to point out that the effect of regional variations in access to healthcare/endoscopy services on colorectal cancer incidence rates remains unknown, and it may be that the incidence rate increases seen in the more affluent, southern regions have been driven by increased awareness and access to medical care.
The main strengths of this study are the size and completeness of the data set. Data were obtained from NCRAS, a nationally curated cancer registry, with 100 per cent complete data for 1974-2012 and 98⋅4 per cent complete data for 2013-2015. Unfortunately, stage-specific data were not recorded routinely until 2012, so further analysis of incidence rate trends according to tumour stage was not possible. It is important to know whether the increase in young-onset colorectal cancer has been driven by an increase in the detection of early-stage disease, particularly in regions and socioeconomic groups that may have increased health awareness and access to endoscopy services. Data presented in this study are population-based and specific causal inferences cannot be made. In addition, IMD quintile and geographical region are group-level metrics and are unable to account for individual-level contextual effects that could have affected the association between these variables and colorectal cancer incidence rates. Finally, endoscopy is being used increasingly in England 42 , and it could be argued that this accounted for the rising incidence of colorectal cancer. However, detection bias is unlikely as incidence rates were decreasing until the 1990s, and the most rapid increases were noted in the youngest age groups which are the least likely to attend for endoscopic examination.
The incidence rate of young-onset colorectal cancer is increasing, particularly among adults aged 20-39 years. This trend appears to be predominantly driven by a rise in distal tumours. Incidence rate increases of a similar magnitude have been observed in both sexes and across IMD quintiles, but are most pronounced in the south of England. Importantly, there is a strong birth cohort effect and it is likely that the increased risk in the youngest cohorts will be carried forward as they age, which will place a significant burden on future healthcare resources. The role of environmental factors, such as diet, obesity, physical exercise and the gut microbiota, in the development of young-onset colorectal cancer is incompletely understood and requires further research. Reducing the screening age to below 50 years will have significant resource implications in the current economic climate 43 and, instead, there should be more focus on risk-stratifying symptomatic younger patients to further investigation using tests such as quantitative faecal immunohistochemical testing.