Factors associated with advanced colorectal cancer differ between young and older adults in England: a population‐based cohort study

Advanced stage presentation of colorectal cancer is associated with poorer survival outcomes, particularly among young adults. This study aimed to determine whether demographic risk factors for advanced stage presentation differed between young and older adults.


Introduction
In England, colorectal cancer (CRC) is the third most common cancer in men and women and accounts for 35 000 new diagnoses each year [1]. While overall agestandardized incidence rates of CRC have remained stable in the UK [2,3], this trend masks recent changes in age-specific incidence rates. There is a substantial body of evidence to suggest that the incidence of CRC in adults under 50 years is rapidly increasing in England and other nations with a high human development index (HDI) [4][5][6][7][8][9], although the cause of this increase is not well understood and is thought to be associated with a rising prevalence of obesity, adoption of sedentary lifestyles and changes to the gut microbiome [10][11][12][13]. Conversely, recent studies from the United States have shown a decline in incidence rates in adults aged over 50 years that has largely been attributed to the introduction of screening [5,14,15].There is emerging evidence that young-onset and older-onset CRC may in fact be biologically distinct entities. Despite a higher prevalence of hereditary syndromes in young-onset CRC, most tumours are sporadic in nature and typically located distally [16], display microsatellite stability and are CpG island methylator phenotype-low [17][18][19].
Advanced stage presentation is associated with worse survival outcomes, with 5-year net survival in England ranging from 10.3% in Stage IV disease to 91.7% in Stage I disease [20]. Population-based studies in other nations with a high HDI have shown that younger age, social deprivation and Black ethnicity are associated with advanced stage CRC [21][22][23][24][25], but demographic risk factors have not been explored in detail and it is unknown whether these risk factors differ between young and older adults. In the UK, there is scant evidence regarding the association between demographic factors, such as socioeconomic status (SES) and ethnicity, and advanced stage presentation due to a lack of complete data in nationally curated cancer registries until recently [26]. Take-up of the Bowel Cancer Screening Programme (BCSP) in England is known to be reduced in men, more deprived socioeconomic groups and areas of increased ethnic diversity [27], but how these trends are reflected in the stage at presentation of the screening age population has yet to be demonstrated. Identifying the demographic factors associated with advanced stage presentation is of critical importance as it would help guide targeted interventions at a population level.
Therefore, the aim of this study was to determine whether demographic risk factors for advanced stage presentation of CRC differed between young and older adults in the English population using data from the National Cancer Registration and Analysis Service (NCRAS) from 2012 to 2015 inclusive.

Method
This study was reported according to the STROBE guidelines for epidemiological studies (Table S1 in the online Supporting Information) [28]. Data were obtained on all patients aged 20 years and above from 2012 to 2015 using data from the National Cancer Registration and Analysis Service (NCRAS) (request ID ODR1718_067). NCRAS is the population-based cancer registry for England, operated by Public Health England, which collects, quality assures and analyses data on all patients with cancer in the English population. NCRAS receives data from multiple sources across the National Health Service including multidisciplinary team (MDT) meetings, pathology reports, molecular testing results, treatment records, hospital activity records, hospital patient administration systems and operational standards, such as cancer waiting times, to ensure complete coverage [26].

Procedures
All new diagnoses of CRC were identified using the ICD10 codes for colonic (C18.1, C18.2-C18.9), rectosigmoid (C19) and rectal (C20) tumours. Appendiceal tumours (C18.1) were excluded due to their differing aetiology. Patients diagnosed prior to 2012 were not included because of the high level of incompleteness for stage and ethnicity data.
Patients diagnosed when under 50 years of age comprised the young-onset colorectal cancer (YOCC) cohort and patients diagnosed when aged 50 years or above comprised the older-onset colorectal cancer (OOCC) cohort. Patients were further stratified according to gender, anatomical subsite (proximalcaecum to descending colon; distalsigmoid to rectum), geographical region, Index of Multiple Deprivation (IMD) quintile and ethnic group. The same anatomical subsite definition was used as in a previous study by the authors to ensure consistency [4]. The IMD is an area-based metric that combines weighted information on the following seven domains: income, employment, education, health, crime, barriers to housing and services and the living environment. England is divided into 32 844 lower-layer super output areas (LSOAs), each of which is ascribed a score based on these domains. The LSOAs are then ranked from least to most deprived and divided into five equal groups or quintiles (1, least deprived; 5, most deprived). Ethnicity was categorized according to the harmonized ethnic group categories for England used by the Office of National Statistics [29].
Tumours were staged as Union for International Cancer Control Stage I to IV based on the TNM classification system. Coding changes between subsequent editions of the TNM system affected sub-classifications within individual stages (e.g. Stage IIIA, IIIB), but not the broader stage categorization used in this study. Staging information inputted into the NCRAS registry was obtained from MDTs, pathology reports, imaging results and post mortems to determine a single anatomical stage at diagnosis [26]. The primary outcome of advanced stage presentation was defined as Stage III or IV disease at diagnosis.

Statistical analysis
Categorical data were presented as frequencies and percentages. variables were made using the chi-square test for independence and the chi-square test for trend as appropriate. Logistic regression models based on complete case analysis were constructed with stage as the dependent variable, dichotomized as early (Stage I/II) or advanced (Stage III/IV), for the YOCC and OOCC cohorts. Given the exploratory nature of the study in determining potential risk factors for advanced stage presentation, all predictor variables were included in the regression models. The potential modifying effect of SES on the association between ethnicity and stage at presentation was explored using the chi-square test for heterogeneity of ORs for unadjusted estimates and the likelihood ratio test for interaction between ethnic group and IMD quintile for adjusted estimates derived by logistic regression. The robustness of the findings from the main analysis was assessed using sensitivity analyses with different definitions of advanced stage disease, i.e. Stage I/II/III versus IV or Stage I versus Stage II/III/IV. Potential bias arising from missing stage information was assessed using multiple imputation under the assumption that stage data were missing at random. The same predictor variables were included in the imputation models as in the original regression analyses using 20 imputed data sets. All analyses were conducted using STATA 13.0 (StataCorp. 2014, Stata Statistical Software, Release 13, StataCorp LP, College Station, Texas, USA).

Risk factors for advanced stage presentation in the YOCC cohort
Among the YOCC cohort, other/not specified ethnicity (OR = 0.81, 95% CI 0.66-1.00, P = 0.045) compared with White ethnicity and residence in the East of England (OR = 0.76, 95% CI 0.62-0.94, P = 0.010) compared with London were independently associated with reduced likelihood of advanced stage presentation in the main analysis ( Table 2). The effect of ethnicity on advanced stage presentation was consistent across all IMD quintiles for each ethnic group (Table S2). Multiple imputation for missing stage data revealed minimal differences in effect estimates for any of the predictor variables compared with the main analysis (Table S3). However, none of the predictor variables were consistently associated with advanced stage presentation when using alternative cut-offs for this outcome in the sensitivity analysis (Tables S4 and S5).

Risk factors for advanced stage presentation in the OOCC cohort
Advanced stage presentation was associated with Black/Black British (OR = 1.25, 95% CI 1.11-1.40, p < 0.001) or other/not specified ethnicity (OR = 1.07, 95% CI 1.01-1.14, P = 0.015) compared with White ethnicity, increasing levels of social deprivation (OR (IMD quintile 5 vs 1) = 1.11, 95% CI 1.07-1.16, P < 0.001] and residence in the East Midlands (OR = 1.11, 95% CI 1.04-1.17, P = 0.001) compared with London in the main analysis (Table 3). Compared with patients aged 50-59 years, patients in each of the age categories between 60 and 89 years were less likely to present with advanced stage disease (all P < 0.001). The likelihood of advanced stage presentation was also reduced among patients living in Yorkshire and The Humber (OR = 0.94, 95% CI 0.89-1.00, P = 0.042), the South West (OR = 0.88, 95% CI 0.84-0.93, P < 0.001) and the East of England (OR = 0.91, 95% CI 0.87-0.96, P = 0.001) compared with London, and among patients with distal tumours (OR = 0.89, 95% CI 0.87-0.92, P < 0.001). The effect of ethnicity on advanced stage presentation was not modified by IMD quintile for Black/Black British, Asian/Asian British and Mixed ethnicity compared with White ethnicity (Table S6). There was evidence of a deprivation gradient for other/not specified ethnicity (P = 0.051) in the unadjusted analysis, but this did not reach statistical significance and interaction between ethnicity and IMD quintile was not demonstrated in the adjusted analysis. Effect estimates  (Table S7), except for a reduced likelihood of advanced stage presentation in patients aged ≥ 90 years compared with 50-59 years (OR = 0.84, 95% CI 0.78-0.90, P < 0.001) and residence in the South East (OR = 0.95, 95% CI 0.90-0.99, P = 0.030) compared with London. The strength and direction of the associations between advanced stage presentation and age, level of social deprivation, geographical region and tumour location remained consistent in both the regression models in the sensitivity analysis (Tables S6 and S7). The association between Black/Black British or other/not specified ethnicity and advanced stage presentation persisted in the regression model with Stages I/II/III versus Stage IV as the outcome (Table S8), but not Stage I versus Stages II/III/IV (Table S9).

Discussion
This is the largest study based on English cancer registry data to explore risk factors associated with advanced stage presentation of any cancer type. It is unique in being the first study from the UK to demonstrate evidence of socioeconomic and ethnic inequalities in the stage at presentation of CRC. Young adults comprised only a small proportion of new diagnoses of CRC but were more likely to present with advanced disease than older adults. Furthermore, they were more likely to present with distal tumours, supporting the findings of other smaller cohort studies that there may in fact be biological differences between young and older onset disease [17][18][19].Of note, demographic risk factors associated with advanced stage presentation differed between the YOCC and OOCC cohorts. Among the OOCC cohort, increasing social deprivation, Black/ Black British ethnicity and residence in the East Midlands were strongly associated with advanced stage presentation, and these findings remained robust to the alternative definitions of advanced stage in the sensitivity analysis. Risk factors for advanced stage presentation were less consistent across the main and sensitivity analyses in the YOCC cohort, with no difference in risk associated with SES, ethnicity and geographical region.
The negative association between age and stage at presentation is well described, and was again demonstrated in this study for both proximal and distal tumours [21,25]. Reasons for advanced stage presentation in young adults remain unclear, despite the increased prevalence of hereditary tumours in this group [30]. Recent cohort studies suggest that while young adults experience a longer time to diagnosis than older adults this delay does not entirely explain the increased risk of advanced stage presentation and that biological factors may play a role [31]. Among the OOCC cohort, adults of screening age were the least likely to present with advanced disease, which may indicate the effect of screening on this population. This would support the findings from the initial randomized controlled trials of faecal occult blood testing that demonstrated a shift to earlier stage disease at presentation among participants in the screening arms [32,33].
The deprivation gradient associated with advanced stage presentation has been reported in other HDI nations [22,23], although this relationship was not demonstrated by previous studies conducted in the UK [34,35]. In this study, there was clear evidence of a deprivation gradient among the OOCC cohort but not the YOCC cohort, implying that age is an effect modifier in the relationship between SES and stage. Studies where analyses have been stratified by age report conflicting results. A population-based study from Florida demonstrated a deprivation gradient in both young and older adult cohorts [36]; whereas a New Hampshire study identified college-level education to be a risk factor for advanced stage presentation in young adults, but protective in older adults [37]. In the current study, the deprivation gradient among the OOCC cohort mirrors the deprivation gradient observed in screening uptake in England [27]. This may explain the lack of association between SES and stage in previous studies from the UK, as these were based on data collected prior to, or during the initial phase of, the screening programmes in England and Scotland [34,35].Despite only small differences in the absolute risk of advanced stage presentation across IMD quintiles, elimination of the deprivation gradient would have resulted in the diagnosis of about 320 fewer cases of advanced stage disease in the OOCC cohort annually. The absence of a deprivation gradient in the YOCC cohort is difficult to explain, although it does reflect the findings from a recent study of demographic trends in the incidence of young-onset CRC in England that showed increases in the incidence rate in CRC to be similar across all IMD quintiles [4].
In the OOCC cohort, Black/Black British ethnicity was associated with a 25% increase in the odds of advanced stage presentation compared with White ethnicity. This may represent an underestimate as Black ethnicity is more frequently misclassified than White ethnicity and may explain why other/not specified ethnicity was also associated with an increase in the odds of advanced stage presentation [38]. In the USA, the association between Black ethnicity and advanced stage presentation is well described and does not appear to be modified by age [24,36]. Paradoxically, other/not specified ethnicity was shown to be protective for advanced stage presentation in the YOCC cohort. This finding cannot be readily explained and may be the result of type 1 error as the associated Pvalue was only just below the 5% threshold for statistical significance. In the UK, Black and minority ethnicity is linked to poorer health outcomes in older adults [39,40], although the effect of ethnicity on stage at presentation for any cancer type has never been studied due to historically low levels of completeness of ethnicity data in English cancer registries [26]. The increased risk of advanced stage presentation among adults of Black/Black British ethnicity in the OOCC cohort may be attributed to screening behaviour, as screening uptake is known to be lower in areas of increased ethnic diversity [27]. However, advanced stage presentation was not associated with Asian/Asian British or Mixed ethnicity, indicating that other factors, such as lifestyle or diet, may offset the risk attributed to low screening uptake in these ethnic groups.
There was no evidence that deprivation modified the effect of Black/Black British ethnicity on advanced stage presentation in either the YOCC or OOCC cohorts. In the OOCC cohort, this suggests that the increased risk of advanced stage presentation among adults of Black/Black British ethnicity could have been related to similarly low levels of screening uptake across all IMD quintiles, but also to potential differences in tumour biology compared with other ethnicities that predisposed to more rapid disease progression. Evidence for genetic differences conferring adverse outcomes in Black patients is limited to one small retrospective study and merits further investigation [41]. The geographical variation in advanced stage presentation was most pronounced in the OOCC cohort. Residence in the East Midlands was a risk factor for advanced stage presentation, whereas residence in the South West and East of England was shown to be protective. It is difficult to reconcile the increased risk associated with residence in the East Midlands with regional screening uptake rates, as first-time uptake is known to be higher in the East Midlands compared with the baseline reference region of London [27]. A recent report from the National Cancer Intelligence Network suggests that the most likely cause of geographical variation in stage at presentation for all cancers is the completeness of stage data [42]. Regions with the highest stage completeness reported the highest proportion of early stage disease, with Trent (now East Midlands) reporting the lowest proportions of stage completeness and early stage disease [42]. In the current study, it is noteworthy that the East Midlands also had the lowest stage completeness of all the English regions at 79.2%, compared with 85.8% for the entire OOCC cohort (data not shown), although stage completeness was similarly low in the South East (80.6%) with no associated increase in the risk of advanced stage presentation. Also, multiple imputation was used to minimize bias associated with missing data and the effect estimates from the multiple imputation models were essentially unchanged from those obtained in the main analysis. Thus, variation in stage completeness by region may not have had a significant effect on stage at presentation.
The key strengths of this study are the size and completeness of the dataset, which has allowed the first national-level analysis of predictors for advanced stage presentation of CRC to be undertaken in England. However, there are some important limitations. This study was exploratory in nature and not powered to detect a prespecified difference in outcome for a given predictor variable. The YOCC cohort was large (n = 7075), but considerably smaller than the OOCC cohort (n = 128 345) and therefore more susceptible to type II error. Combining the data presented here with other international datasets would reduce the impact of type II error, but the benefits of doing so would have to be weighed against the potential heterogeneity introduced by including data from countries where data may be less complete, or less accurate, and healthcare systems do not provide universal coverage. Another limitation is that IMD, geographical region and ethnicity are all grouplevel metrics that do not fully account for individual-level contextual factors that may have influenced the association between these variables and stage at presentation. Furthermore, non-White ethnicity is more likely to be misclassified than White ethnicity and thus the effect estimates of ethnic group on stage at presentation should be viewed with caution [38].
In summary, young adults with CRC are more likely to present with advanced disease and distal tumours than older adults. Demographic factors associated with advanced stage presentation differ between young and older populations. There is clear evidence of a deprivation gradient and increased risk of advanced stage presentation associated with Black/Black British ethnicity in older adults that is not observed in the young, which reflect similar patterns in screening uptake. The rising incidence of CRC in young adults in England is characterized by a strong cohort effect with the increased risk observed in the youngest cohorts, i.e. those aged 20-29 and 30-39 years, likely to be carried forward as they age [4]. More than 60% of all tumours among adults aged 50-59 years were Stage III/IV at presentation and this age group is likely to face the dual burden of rising CRC incidence and advanced stage presentation unless the screening age is lowered. The Bowel Scope Screening Programme, which commenced in 2013, offers a one-off flexible sigmoidoscopy at the age of 55 years, but only covers half of all general practice populations in England and has the lowest uptake in the most deprived and ethnically diverse areas [43,44]. Therefore, strategies that are specifically tailored to groups at the highest risk of advanced stage presentation are warranted. These should involve a more nuanced approach to the use of quantitative faecal immunohistochemical testing (qFIT), which has already been shown to increase screening uptake among previous nonresponders [45]. Broadening access to diagnostic endoscopy has significant resource implications in the current healthcare climate and qFIT should be used as a risk stratification tool in symptomatic younger patients.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1. STROBE checklist. Table S2. Risk of advanced stage presentation, defined as Stage I/II versus Stage III/IV, by ethnic group and stratified by Index of Multiple Deprivation (IMD) quintile in the young-onset colorectal cancer cohort. a Adjusted for age, gender, English region and tumour location. b The chi-square test for heterogeneity of odds ratios was used to assess for interaction in the unadjusted analysis of ethnicity stratified by IMD quintile for each ethnic group. The likelihood ratio test was used to assess interaction in the adjusted analysis by comparing logistic regression models with and without an interaction term for ethnicity-IMD quintile and thus, only one P-value has been presented.  Table S3. Logistic regression model of demographic factors associated with advanced stage presentation, defined as Stages I/II versus Stages III/IV, in the young-onset colorectal cancer cohort after multiple imputation of stage. Table S4. Logistic regression model of demographic factors associated with advanced stage presentation, defined as Stages I/II/III versus Stage IV, in the young-onset colorectal cancer cohort. Table S5. Logistic regression model of demographic factors associated with advanced stage presentation, defined as Stage I versus Stages II/III/IV, in the young-onset colorectal cancer cohort. Table S6. Risk of advanced stage presentation, defined as Stages I/II versus Stages III/IV, by ethnic group and stratified by Index of Multiple Deprivation (IMD) quintile in the older-onset colorectal cancer cohort. a Adjusted for age, gender, English region and tumour location. b The chi-square test for heterogeneity of odds ratios was used to assess for interaction in the unadjusted analysis of ethnicity stratified by IMD quintile for each ethnic group. The likelihood ratio test was used to assess interaction in the adjusted analysis by comparing logistic regression models with and without an interaction term for ethnicity-IMD quintile and thus, only one P-value has been presented. Table S7. Logistic regression model of demographic factors associated with advanced stage presentation, defined as Stages I/II versus Stages III/IV, in the older-onset colorectal cancer cohort after multiple imputation of stage.