Why are most colorectal cancers diagnosed outside of screening? A retrospective analysis of data from the English bowel screening programme

Objective Despite several interventions to increase participation in England, most colorectal cancers (CRCs) are diagnosed outside of the screening programme. The aims of this study were to better understand why most CRCs are diagnosed externally, the extent to which this is due to suboptimal uptake of screening, and the extent to which it is due to other factors, such as false-negative test results. Setting / Methods We performed a clinical audit of 1011 patients diagnosed with CRC at St Mark's Hospital (Harrow, UK) between January 2017 and December 2020. Data on the diagnostic pathway and screening history of individuals were extracted from the bowel cancer screening system and assessed using descriptive statistics. Results 446/1011 (44.1%) patients diagnosed with CRC were eligible for screening at the time of diagnosis. Of these, only 115/446 (25.8%) were diagnosed through screening. Among those diagnosed via non-screening pathways, 210/331 (63.4%) had never taken part in screening, 31/331 (9.4%) had taken part but were not up to date, and 89/331 (26.9%) had taken part and were up-to-date (of these, 82/89 [92.2%] had received a normal or weak positive test result, and 5/89 [5.6%] had received a positive result and declined colonoscopy). Conclusion Nearly two-thirds of screening eligible patients diagnosed through a non-screening pathway had never taken part in screening. This represents the single largest source of inefficiency within the screening programme, followed by missed findings and inconsistent participation. Given the improved outcomes associated with screen-detected cancers, there is a strong public health mandate to encourage participation.


Introduction
Colorectal cancer (CRC, also referred to as 'bowel cancer') is the second leading cause of death from cancer in Europe. 1 Several large randomised controlled trials (RCTs) have shown that regular faecal immunochemical test (FIT) screening, between the ages of 45 and 80, can significantly reduce the mortality of the disease among people who complete the test. 2 As a result, many European countries have implemented FIT-based screening programmes for the early detection of CRC. 3 England introduced its national bowel cancer screening programme in 2006. For the first four years, guaiac faecal occult blood test screening (gFOBt) was offered to men and women aged 60-69 years. In 2010, the programme was extended to include men and women up to the age of 74. Then, in 2019, the programme switched to biennial FIT-based screening for CRC, following the success of a national pilot, which demonstrated significant improvements in uptake (a one-off flexible sigmoidoscopy was also available for adults aged 55-59 from 2013 onwards; however, this was decommissioned in January 2021). 4 Despite the availability of a national screening programme, which is free at the point of delivery, most CRCs in England are diagnosed outside of the screening programme. Indeed, a recent review of sex-related differences in routes to diagnosis found that only 8.1% and 5.1% of CRCs diagnosed in men and women, respectively, are diagnosed through screening. 5 Why so many people are diagnosed with CRC outside of screening is not entirely clear. Indeed, ∼40% of people who are diagnosed with CRC are diagnosed at an age when they are eligible for screening (i.e. 60-74 years of age), 6 so one might expect the proportion diagnosed through screening to be much greater. Possible explanations for the low diagnostic rate of screening include suboptimal use of screening among the eligible population (one study found that, while 70% of invitees participate at least once over three screening rounds, only 44% participate in all three rounds), 7 false negative test results (at present, the programme uses a threshold of 120 μg/ g to select individuals for colonoscopy; however, the cancer detection rate is almost doubled when a threshold of 20 μg/g is used) 4 and non-attendance at follow-up colonoscopy (approximately 15% of adults who receive an abnormal FIT screening result do not attend a diagnostic investigation). 4 To date, no studies have quantified the reason why so many individuals are diagnosed outside of screening, nor how route to diagnosis influences CRC outcomes in screening-eligible adults. The aim of this study, therefore, was to determine the extent to which diagnoses made outside the screening programme are due to non-participation in screening, false negative test results, and non-attendance at diagnostic investigation (i.e. colonoscopy), and how, if at all, these pathways influence stage at diagnosis.

Study design
This study used a retrospective design to analyse data on 1011 adults diagnosed with CRC.

Population
The population of interest was all men and women who were diagnosed with CRC at St Mark's Hospital (Harrow, UK), during a four year period (1st January 2017 -31st December 2020).

Primary outcome measures
Data on the screening history and results of each individual were extracted from the bowel cancer screening system: an electronic system that provides up-to-date information about a person's bowel cancer screening status. Specifically, data on whether individuals had 'never participated in bowel cancer screening' and 'participated in the most recent round of bowel cancer screening, prior to their diagnosis', were extracted from the bowel cancer screening system, as well as whether the result was 'positive' (3 or more positive gFOBt samples [of 6], a FIT of >120 µg/g, or detection of adenomas / cancer during a flexible sigmoidoscopy), 'weak positive' (1 or 2 positive gFOBt samples [of 6]), or 'normal' (0 positive gFOBt samples [of 6], a FIT of <120 μg/g, or no adenomas / cancer during a flexible sigmoidoscopy). Where participants had a 'positive' result, additional data were extracted regarding whether they attended, or did not attend, a diagnostic investigation, such as colonoscopy.

Secondary outcome measures
Data on several additional variables of interest were also extracted from the bowel cancer screening system, including sex ('male', 'female'), age at diagnosis (measured continuously, in years), Ethnicity ('White British', 'South Asian', 'Any other Asian ethnicity', 'Any Black Ethnicity', 'Any other White Ethnicity', 'Mixed/Other'), Region ('Brent', 'Harrow', 'North Ealing'), area-level deprivation (measured using the 'Index of Multiple Deprivation' [IMD], which is the government's official measure for area-level deprivation, and uses census data to create a scale ranging from 0 to 80), 9 Lymph node involvement at diagnosis (yes, no), Metastatic disease at diagnosis (yes, no) and the source by which individuals were diagnosed / referred for colonoscopy (screening, primary care, accident and emergency [A&E], other).

Analysis
Descriptive statistics (i.e. frequencies and percentages) were used to describe the proportion of individuals who were eligible for screening, the demographic characteristics of those individuals, their previous screening participation, results and colonoscopy attendance, and the route by which they were diagnosed with CRC. Descriptive statistics were also used to report the involvement of lymph nodes and other organs at diagnosis.
Univariate and multivariate logistic regression were used to identify predictors of being diagnosed through screening (within the screening eligible sample) and having lymph node involvement and metastasis at diagnosis.
The threshold for statistical significance was 0.05; the data were analysed using SPSS (version 27.0).

Missing data
Cases with missing data were excluded from the analyses.

Ethics
Ethical approval was not required for this study, as it was considered 'service evaluation', as opposed to 'research', by the National Health Service (NHS) Health Research Authority.

Predictors of route to diagnosis
In the multivariate analysis (

Predictors of lymph node involvement in the screening eligible population
In the multivariate analysis (Table 3), being referred for colonoscopy through general practice was associated with increased odds of lymph node involvement at diagnosis, compared with being referred for colonoscopy through screening (64.7% vs. 45.0%; aOR: 2.33, 95%CIs: 1.36, 4.01; p: 0.002). Similarly, having never participated in screening was associated with increased odds of lymph node involvement, compared with those who were diagnosed through screening (63.9% vs. 45.0%; aOR: 2.22, 95%CIs: 1.28, 3.86; p: 0.005, respectively).

Predictors of metastasis in the screening eligible population
In the multivariate analysis (

Summary of main findings
This study found that only one in four CRCs diagnosed in screening eligible adults are diagnosed through the screening programme. In addition, this study found that most CRCs diagnosed outside the screening programme (six in ten) are diagnosed in people who have never taken part in bowel cancer screening, and that a considerable proportion (three in ten) are diagnosed in people who were up-to-date with their bowel cancer screening, but had received a normal or weak positive result that did not require further investigation (a small proportion were also diagnosed in those who received an abnormal result, but did not attend further investigation when invited). This study also found that screening eligible adults who were diagnosed through the screening programme were less likely to have lymph node involvement at diagnosis, compared with those who were referred for colonoscopy through primary care (no differences were observed for those diagnosed through  Bold-type indicates significance: * = P < 0.05; ** = P < 0.01; *** = P < 0.001. A&E or other pathways, although this may have been due to lower numbers [and thereby lower statistical power]). Patients diagnosed through A&E or other pathways were, however, more likely to have metastasis than those diagnosed through screening, as were those diagnosed through primary care. Finally, this study found that individuals who were up-to-date with screening (but received a normal / weak positive result) or had never participated in screening were more likely to have lymph node involvement at diagnosis, compared with those who were diagnosed through screening, while those who had previously participated in screening but were not up-to-date were no more or less likely to have lymph node involvement (as above, this is likely due to small numbers and, thereby, reduced statistical power). Patients who had previously participated in screening but were not up-to-date were, however, more likely to have metastatic disease at diagnosis than those diagnosed through screening, as were those who had never taken part in screening, or had taken part and were up-to-date.

Comparisons with the previous literature
The results of this study contrast with those of a larger study, conducted by White et al. (2018), which found that 8.1% of all CRCs diagnosed in men and 5.1% of all CRCs diagnosed in women are diagnosed through screening (present study: 11.2% [65/580] in men and 13.2% [57/431] in women). The reason for this discrepancy is not entirely clear; however, one possible explanation is that White et al. analysed data on people diagnosed with CRC between 2006 and 2013, before the age extension to 74 year olds was fully rolled-out (2014), and before the implementation of once-only flexible sigmoidoscopy screening for 55 year olds (also referred to as 'bowel scope screening'), which started in 2013 and was decommissioned in 2021. 10 Another possible explanation, or joint explanation, for the higher proportion diagnosed through screening is that White et al.'s study took place before the implementation of FIT, which began in 2019, and has higher uptake and sensitivity than gFOBt. 4,11 The results of this study are, however, consistent with other studies examining disease progression by diagnostic pathway. For example, an analysis comparing the clinical outcomes of screen-detected cancers and stage-matched cancers found that more screen-detected cancers were diagnosed at Dukes' Stage A, compared with interval cancers (40% [125/316] vs. 19% [36/187]). 12 The finding that individuals from 'Any other Asian background' were more likely to participate in bowel cancer screening than their White British counterparts was unexpected. The authors believe these differences might be due to the Gurkha population, who are, anecdotally, highly compliant with screening (and form a large part of the local community). The finding that individuals from North Ealing were less likely to be diagnosed through screening than individuals from Brent, however, was expected and is likely due to lower screening participation in the region. 13

Strengths and limitations
This study has several strengths. First, it used objective measures of screening participation and history, as opposed to selfreported measures, improving the reliability of the findings. Second, it used data on an ethnically diverse sample, improving the generalisability of the results. Finally, there was a wide range of co-variates, including age, sex, area-level deprivation, screening history and route to diagnosis, reducing the effect of confounding data in the final models.
This study also has several limitations, First, it was restricted to a single centre in London, and so does not represent the general population of England. Second, the sample size was relatively small, meaning that the study may have been underpowered to detect modest differences between subgroups. Finally, as the study was restricted to data stored on the bowel cancer screening system, it was not possible to include other co-variates which may have been important / of interest (e.g. co-morbidities).

Implications for policy and future research
This study has several implications for policy and research. First, the findings of this study reiterate the importance of health promotion campaigns encouraging regular participation in screening (not only did this study find that people who were diagnosed outside of the screening programme were more likely to have metastatic disease at diagnosis, but that this was true even of those who had taken part in screening, but were not up-to-date). Second, as the findings of this study were limited to a single centre in London, there is now a need to verify the findings with larger, national, datasets. Third, this study used data that were predominantly collected when gFOBt was the primary screening test, as opposed to FIT, and so repeat studies will be needed to see how this affects the proportion detected through screening in the future (given the analytical superiority of FIT, it is possible that some issues, such as 'false negatives results' and 'the proportion of cancers missed among individuals who are up-to-date with screening', may be overrepresented).

Conclusions
This study demonstrates that the majority of screening eligible patients who are diagnosed with CRC are diagnosed outside of the screening programme. In addition, this study demonstrates that individuals diagnosed outside of the screening programme, including those who have previously taken part in screening but are not up-to-date, are more likely to be diagnosed with metastatic disease. The results of this study, therefore, reinforce the importance of regular participation in screening and the need for public health strategies to promote uptake among previous non-responders.