Article Text

Socioeconomic inequalities in interval colorectal cancer are explained by differences in faecal haemoglobin concentration and age: a register-based cohort study
  1. Ulrik Deding1,2,
  2. Morten Kobaek-Larsen1,2,
  3. Henrik Bøggild3,
  4. Lasse Kaalby1,2,
  5. Marianne Kirstine Thygesen1,2,
  6. Gunnar Baatrup1,2
  1. 1Department of Surgery, Odense University Hospital, Svendborg, Denmark
  2. 2Department of Clinical Research, University of Southern Denmark, Odense, Denmark
  3. 3Department of Health Science and Technology, Public Health and Epidemiology Group, Aalborg University, Aalborg, Denmark
  1. Correspondence to Ulrik Deding; Ulrik.Deding{at}


Objective To estimate the risk of interval colorectal cancer (CRC) in faecal immunochemical test (FIT) negative screening participants according to socioeconomic status.

Design In this register-based study, first round FIT negative (<20 µg hb/g faeces) screening participants (biennial FIT, citizens aged 50–74) were followed to estimate interval CRC risk. Multivariate Cox proportional hazard regression models estimated HRs based on socioeconomic status defined by educational level and income. Models were adjusted for age, sex and FIT concentration.

Results We identified 829 (0.7‰) interval CRC in 1 160 902 individuals. Interval CRC was more common in lower socioeconomic strata with 0.7‰ for medium-long higher education compared with 1.0‰ for elementary school and 0.4‰ in the highest income quartile compared with 1.2‰ in the lowest. These differences did not translate into significant differences in HR in the multivariate analysis, as they were explained by FIT concentration and age. HR for interval CRC was 7.09 (95% CI) for FIT concentrations 11.9–19.8 µg hb/g faeces, and 3.37 (95% CI) for FIT between 7.2 and 11.8 compared with those <7.2. The HR rose with increasing age ranging from 2.06 (95% CI 1.45 to 2.93) to 7.60 (95% CI 5.63 to 10.25) compared with those under 55 years.

Conclusion Interval CRC risk increased with decreasing income, heavily influenced by lower income individuals more often being older and having increased FIT concentrations. Individualising screening interval based on age and FIT result, may decrease interval CRC rates, reduce the social gradient and thereby increase the screening efficiency.


Data availability statement

No data are available. Data were stored on governmental secure logged servers and were pseudoanonymised before the authors were granted access. These data can only be accessed after individual approval has been granted and the data contains personal individual information. the data can therefore not be shared.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Population-based colorectal cancer screening has proven capable of decreasing cause-specific and all-cause mortality.

  • Guidelines recommend that screening programmes should aim to limit inequalities.

  • Social inequalities in colorectal cancer screening are evident in several stages of the process; faecal immunochemical test (FIT) participation, correct FIT sample collection, colonoscopy participation and risk of incomplete colonoscopy.


  • Social inequalities in interval colorectal cancer are evident in a National population-based screening programme using FIT, followed by optical colonoscopy in FIT positive individuals aged 50–74 years.

  • Individuals of lower income are at higher risk of interval colorectal cancer compared with their peers of higher-income groups.

  • Differences in risk between socioeconomic status subgroups are explained by differences in FIT concentration and age.


  • Population-based colorectal cancer screening programmes invite citizen based on their age and most use FIT to determine who is a candidate for optical colonoscopy.

  • Therefore, age and FIT results are easy available information in all FIT negative screening participants.

  • Screening programmes could increase efficiency, decrease the number of interval colorectal cancers and potentially limit the social gradient by differentiating the screening interval based on age and FIT result.


Socioeconomic and demographic differences in colorectal cancer screening participation are evident worldwide.1 Screening participation is associated with a reduction in both colorectal cancer mortality and all-cause mortality in long-term follow-up.2 The higher mortality in non-participants may be due to higher stages of cancer at diagnosis, but advanced staged cancers detected in screening also holds better prognosis than cancers detected outside screening.3 This probably entails an increased mortality risk in lower socioeconomic subgroups due to their lower participation rates. European guidelines state that participation should not be limited by financial barriers and that one aim of screening is to reduce health inequalities.4

Nevertheless, social inequalities are evident in participation at both faecal immunochemical test (FIT) and colonoscopy.5 6 The proportion of positive FIT samples and the risk of a following incomplete colonoscopy is higher in lower socioeconomic subgroups.7 8 Why this is, has yet to be determined, but it has been established that lifestyle and health behaviour affects the risk of colorectal cancer.9 In a North-American study, socioeconomic status has been identified as a predictor of the diversity in the colonic microbiota.10 Lifestyle affects the risk of developing inflammatory bowel disease (IBD) and, in IBD patients, lifestyle affects the severity of disease,11 12 while diet may also prevent the development of colorectal cancer.13 Further, health behaviour such as smoking, alcohol and diet varies between socioeconomic subgroups14 15 and may lead to increased risk of colorectal cancer. Whether this will be reflected in increased FIT concentrations is unknown. One possible explanation could be that lifestyle also affects the colonic microbiota and increases the risk of inflammation in the colon, entailing vasodilation and increased vascular permeability, swelling and redness,16 which could then increase the risk of haemoglobin in the lumen when faeces come into contact with the mucosal lining. Further, the cell turnover increases when inflammation occurs, and the risk of cancers developing increases with cell proliferation.17

We know that the risk of interval cancer correlates to FIT value, sex and age in FIT negative individuals,18–21 but whether a social gradient in the frequency of interval cancer exists is yet to be established. One study did not find statistically significant differences in interval colorectal cancer rates in FIT negative subjects by socioeconomic score, but was based on only 27 cases.22

We aimed to estimate the risk of interval colorectal cancer in FIT negative screening participant according to socioeconomic status.


In this register-based study, individuals with a negative FIT (below 100 ng hb/mL buffer equivalent to 20 µg hb/g faeces) in the first round of the Danish Colorectal Cancer Screening Programme were identified. They were followed to estimate the risk of interval colorectal cancer defined as a colorectal cancer diagnosis before the next screening round. They were followed from the date of their negative FIT result until next screening invitation, death, emigration, lost to follow-up or end of follow-up 24 September 2018, whichever came first. The first round of screening in Denmark was rolled out over 4 years (2014–2017), after which citizens are repeatedly invited every 2 years. Follow-up in our study was therefore limited to 2 years in individuals who were not reinvited within 2 years. Individuals lost to follow-up or with missing data on exposure, outcomes or covariates were excluded from the analyses.


All data between registers were linked using the personal civil registration number assigned to all individuals in Denmark at birth or when immigrating.

The Danish Colorectal Cancer Screening Database (DCCSD) provided data on sex, age, date of screening participation, date of invitations and FIT results. The DCCSD is a clinical database with high validity for programme monitoring and research purposes.23 It holds a high level of agreement (>90%) with hospital records, and receives data directly from the Invitation and Administration Module (IAM), which automatically plans and executes individuals’ screening invitation trajectory. IAM uses the civil registration number, which includes the date of birth and sex of the citizen. Besides the IAM, the DCCSD also comprises data from the Danish National Patient Registry (DNPR) and the National Pathology registry.

The DNPR provided data on colorectal cancer diagnoses during follow-up. The DNPR was established in 1977 and holds information on all patient contacts with Danish hospitals, including diagnoses, procedures, treatments and dates. All diagnoses in the DNPR are registered using the International classification of Diseases (ICD) coding. The DNPR is the most comprehensive register and has since the year 2000 been the basis of all funding for public hospitals. Therefore, the registrations from public hospitals since then is assumed to be complete.24 Colorectal cancer diagnoses were identified in the DNPR as ICD-10 codes ‘C18’ or ‘C20’, or ICD-8 codes ‘15300’, ‘15301’, ‘15302’, ‘15309’, ‘15319’, ‘15329’, ‘15339’, ‘15380’, ‘15389’, ‘15399’, ‘15409’, ‘15410’, ‘15411’, ‘15419’, or ‘15429’. ICD-8 codes were included in order to identify individuals with previous colorectal cancer as these were used up until the year 1993.

The Populations Education Registers (PER) provided data on the highest achieved level of education. The PER has a reported 0%–3% misclassifications and in 2008, 97% of the Danish population born between 1945 and 1990 have a registered level of education, but for immigrants this figure was lower (85%–90%).25

The Income Statistics Register (ISR) provided data on personal income. The ISR holds income information on anyone who has submitted a tax return form in Denmark and registers anyone economically active in Denmark. The ISR includes information on wages, entrepreneurial income, taxes, public transfer payments, public pensions, capital income, and private pension contributions and pay-outs. The ISR is complete and holds all registered income, however, any undeclared income cannot be covered by the register.26

Socioeconomic status

Socioeconomic status was assessed for each individual using educational level and income. These were treated as two separate exposures. Educational level was identified as highest level of completed education from the PER25 at baseline and grouped as elementary school, high school or vocational education, short higher education, and medium-long higher education. We considered income as the annual individual income in the year previous to the FIT submission as registered in the ISR.26 We then adjusted for inflation (year 2021) and categorised income by quartiles within the sample. The first income quartile was income below DKK208 897 (US$28 526), the second quartile was DKK208 897–DKK315 535 (US$28 526–US$43 088), the third quartile was DKK315 536–DKK447 865 (US$43 088–US$77 665), and the forth quartile was above DKK447 865 (US$77 665).


Age was included in the models as a categorical variable grouped as <55, 55–59, 60–64, 65–69 and 70 years or above. Sex was included as a binary variable. FIT concentration was included as a categorical variable grouped as <7.2 µg hb/g faeces, 7.2–11.8 µg hb/g faeces and 11.9–19.8 µg hb/g faeces. Concentrations were reported as numerical values between below 35–99 ng hb/mL buffer in the DCCDS. This was translated into μg hb/g faeces resulting in a minimum value reported as below 7.2 and a maximum value of 19.8 µg hb/g faeces. FIT concentrations were measured in the screening programme using the OC-Sensor Diana instruments (Eiken Chemical Co, Japan).


Baseline characteristics were compared using the χ2 test. Multivariate Cox proportional hazards regression models were conducted, estimating the hazard rate ratio as a proxy for the risk of interval colorectal cancer based on socioeconomic status in a simple model including both income and educational level and in a full multivariate model additionally adjusting for age, sex and faecal haemoglobin concentration. The authors discussed whether faecal haemoglobin concentration was to be considered a confounder or if it mediated the effect of socioeconomic status on the risk of interval colorectal cancer. This discussion arose since, as suggested in the introduction, lifestyle and health behaviour are associated with the risk of colorectal cancer and are also associated with socioeconomic status. We also know that lifestyle mediates some, but not all, inequalities in health related to socioeconomic status.27 FIT concentration may, therefore, simply be viewed as a marker of disease, but could also theoretically act as a proxy for lifestyle and behaviour in a mediation analysis. We, therefore, performed a series of regression models, as suggested by Baron and Kenny,28 confirming that faecal haemoglobin concentration was not a mediator. Schönfeld residuals were evaluated to confirm the proportional hazards assumption, and interactions between each covariate and educational level and income, respectively, were explored by including interaction terms in the models. Cross-tables of age group and FIT concentrations stratified by socioeconomic status were provided. Finally, cumulative incidence proportions curves stratified by the strongest predictors of interval colorectal cancer were created. Level of statistical significance was set at 5% and 95% CIs were calculated. Data management and statistical analysis were performed using SAS software V.9.4 (SAS Institute) and R statistical software package V.4.1.3 (R Core Team, Vienna, Austria). Sensitivity analyses were performed conducting the multivariate regression model first including the interaction term of educational level and FIT concentration, then, stratified by FIT concentration, and last, while stratifying the baseline hazard function by FIT concentration. Further, we conducted marginal effects analysis exploring the pattern of this interaction. Marginal effects analysis was conducted in STATA V.17 (StataCorp. 2021. Stata Statistical Software: Release V.17, StataCorp).


In Denmark, 1 190 540 individuals submitted a FIT sample and had a negative test result in the first round of screening. Among these, we excluded 4718 (0.40%) individuals that were registered with previous colorectal cancer and 88 (0.07‰) individuals that were emigrated or dead prior to participation. A further 24 832 (2.1%) were excluded due to missing information on socioeconomic status. This left 1 160 902 (97.5%) individuals for final analyses, of whom 829 (0.7‰) were diagnosed with colorectal cancer before their next screening invitation (ie, within a maximum of 2 years) (figure 1).

Figure 1

Flow of participants. FIT, faecal immunochemical test.

Significant differences in the proportions of interval colorectal cancer among participants were observed between groups of educational level, income quartiles, FIT concentrations, sex and age group. Participants with elementary school as their highest completed educational level had a higher proportion of interval colorectal cancer than participants with higher educational levels. Also, the proportion of interval colorectal cancer decreased with increasing income. The highest proportions of interval colorectal cancer in the subgroups were seen in the eldest (1.6‰) and those with high negative faecal haemoglobin concentrations (2.1‰ and 4.6‰) (table 1).

Table 1

Baseline characteristics stratified by interval colorectal cancer outcome

Compared with those with elementary school as their highest level of education, only those with high school/vocational education had a significant lower risk of interval colorectal cancer from the simple model with an HR 0.83 (95% CI 0.71 to 0.98). The risks of interval colorectal cancer were significantly lower for those in the second (HR 0.60, 95% CI 0.51 to 0.71), third (HR 0.44, 95% CI 0.36 to 0.54) and fourth (HR 0.36, 95% CI 0.29 to 0.45) income quartiles compared with those with the lowest income in the simple model (figure 2).

Figure 2

Risk of interval colorectal cancer based on socioeconomic characteristics from a simple cox proportional hazards regression model including educational level and income quartiles in the same model.

After adjusting for covariates, the only significant difference in risk for socioeconomic characteristics was a decreased risk for second income quartile (HR 0.76, 95% CI 0.64 to 0.91) compared with first quartile. The differences in risk were especially impacted by adjustments for FIT concentration and age. Sex also appeared to impact the results but to a lesser degree (figure 3). HR from the adjusted model were 3.37 (95% CI 2.81 to 4.04) for FIT concentrations 7.2–11.8 µg hb/g faeces, and 7.09 (95% CI 5.85 to 8.59) for FIT concentrations 11.9–19.8 µg hb/g faeces compared with those with <7.2 µg hb/g faeces. Compared with those under 55 years, the age groups were all at significantly increased risk of interval colorectal cancer, with HR 2.06 (95% CI 1.45 to 2.93), 3.24 (95% CI 2.34 to 4.48), 5.70 (95% CI 4.20 to 7.75) and 7.60 (95% CI 5.63 to 10.25), respectively. Males were at increased risk compared with females (HR 1.16, 95% CI 1.01 to 1.33) in the adjusted model (figure 3). The only statistically significant interaction found was between educational level and FIT concentration, leading to a series of sensitivity analyses. Results from the sensitivity analysis with the inclusion of the interaction term of educational level and FIT concentration did not differ significantly from results presented in figure 3 (online supplemental appendix A, figure A1). The multivariate regression model stratified by FIT concentration did not reveal any significant differences in HRs for educational levels across the strata (online supplemental appendix A, figure A2) and the model stratifying the baseline hazard function by FIT concentration resulted in HRs equal to those of figure 3 (first decimal). The marginal effects analysis did not reveal a pattern of differing effects on the predicted HRs across educational levels from FIT concentrations (online supplemental appendix A, figure A3).

Figure 3

Risk of interval colorectal cancer based on socioeconomic characteristics and covariates from an adjusted cox proportional hazards regression model. FIT, faecal immunochemical test; hb, haemoglobin.

Larger proportions of individuals of lower socioeconomic status were seen in groups of increased FIT concentrations and higher age groups. For example, the proportions of individuals with FIT concentration 11.9–19.8 µg hb/g faeces were 1.8% for highest income quartile, increasing to 2.1%, 2.8% and 3.1% with decreasing income quartiles. Similar, proportions of individuals in the oldest group were 6.4% for the highest income quartile, increasing to 10.5%, 23.9% and 38.1% with decreasing income quartiles (table 2).

Table 2

Socioeconomic characteristics stratified by faecal haemoglobin concentration and age groups

The cumulative incidence proportion curves illustrates how individuals aged 60 or older while also having an increased FIT concentration (7.2–19.8 µg hb/g faeces) were at significantly increased risk of interval colorectal cancer even before 100 days had passed since FIT result, compared with individuals that are either below the age of 60 years or has an FIT concentration below 7.2, or both. After 2 years of follow-up, the cumulative incidence proportion was more than fivefold higher in the high risk group compared with low risk (figure 4). Cumulative incidence proportion curves illustrating the differences between age groups and FIT concentrations explicitly are included in online supplemental appendix A, figure A4,A5.

Figure 4

Cumulative incidence proportion curves of interval colorectal cancer stratified by risk predictors identified in multivariate cox proportional hazard regression model. Definitions: Increased FIT concentration, 7.2–19.8 μg hb/g faeces; increased age, 60 years or older. FIT, faecal immunochemical test; hb, haemoglobin.


Social inequalities in interval colorectal cancer incidence were evident in the Danish Colorectal Cancer Screening programme, where interval cancers occurred more often in those with lower income. However, this association seems to be carried by an overrepresentation of older individuals and individuals with detectable faecal haemoglobin concentrations of 7.2 µg hb/g faeces or more in the lower income quartiles, as shown from the full multivariate regression analysis. No significant inequality was seen based on educational level from the adjusted regression model, whereas high school/vocational education held a significantly lower risk of interval cancer compared with elementary school in the simple model. In both models, a u-shaped pattern of the HRs was evident for educational level. Even though lifestyle and socioeconomic status in general holds a relationship of the higher the status the healthier the lifestyle, for some specific lifestyle choices, the pattern may be more u-shaped. As an example, alcohol intake is higher in the lowest and highest socioeconomic subgroups of the Danish population, while at the same time being a risk factor for colorectal cancer development.29 30

Interaction, mediation and marginal effects analyses did not indicate that the FIT concentration mediated the association between educational level and interval colorectal cancer. Even though it may seem plausible that income and age interacts, as individuals with increasing age have a higher likelihood of withdrawal from the work force, we did not find other significant interactions in the model. As FIT concentration and age is not used to determine screening interval, this means that individuals of lower income are by proxy at higher risk of interval colorectal cancer. We found that interval cancers were more common in men than women, contradicting previous finding from a Scottish sample,31 although they used the guaiac-based faecal occult blood test (gFOBT) rather than the FIT as in the current study. This may affect the results if the FIT is more sensitive (compared with gFOBT) to left-sided cancers as opposed to right sided, as right-sided cancers were more common in women.31 In the Dutch screening programme where the FIT is also employed, a higher incidence rate of interval cancers was seen in males compared with females32 comparable to what we found. Regardless of which FIT positivity threshold is applied in screening, some cancers and precursor lesion will be missed. The threshold used in National screening programmes varies from 15 to 180 µg hb/g faeces (75–900 ng hb/mL buffer) between European countries with FIT-based screening.33 A negative FIT result carries the risk of false reassurance34 and neglect of symptoms. This could lead to a later stage cancer at time of diagnosis. It has been shown that interval cancers have a higher stage than screening detected cancers,22 and stage at diagnosis is strongly associated with mortality.35 To have the risk of false reassurance and risk of higher cancer stage at diagnosis differentiate between socioeconomic subgroups seems counterproductive in screening programmes that by European guidelines should aim to decrease inequalities in health.4 Only one other study investigating the association between socioeconomic status and interval cancers was identified. The authors did not report any significant differences in interval cancer rate between socioeconomic subgroups in the FIT negative arm of their trial.22 This may be either due to their low number of cases (n=27), due to their estimation of socioeconomic status relying on postal code means rather than individual assessments, or because no socioeconomic differences in interval colorectal cancer exists in the Netherlands. They did not adjust for age and FIT concentration, although with the power available it probably would not change the significance of their findings.

This register-based study enabled the inclusion of a large sample, needed for analyses of this rare outcome (0.71‰ in our sample with 2 years follow-up, and 1.72‰ in the Dutch sample with up to 2.3 years of follow-up).22 The Danish National registers enable us to identify individual socioeconomic characteristics rather than relying on means of geographical area, which is often a limitation. The average age of retirement from the work force in Denmark is approximately 65 years old. As our measure of socioeconomic status relies on income the year before participation, there is a risk of pensioners with a low income but great wealth being categorised as lower socioeconomic status. As we included age groups in the full multivariate regression model, this possible bias will be limited to the results of the simple model. The results from the current study are limited to the first round of screening, and future research would need to confirm whether they can be generalised to consecutive rounds of colorectal cancer screening.

Most colorectal cancer screening programmes are targeting individuals based on their age, and usually within the range of 45–80 years. In programmes using the FIT as a first screening tool, a threshold must be determined. Therefore, all FIT-based screening programmes will have information on individual age and FIT concentration at participation. This information could possibly be used to individualise screening intervals according to risk of interval cancer. By decreasing the interval in high-risk individuals (eg, individuals above the age of 60 with an increased FIT concentration), we may be able to limit the number of interval cancers and decrease the social gradient. If at the same time the interval is increased in low-risk individuals, this can be done without adding additional FIT and colonoscopies to the screening programme, which would affect the cost of the programmes. The latter part would go against the guidelines, recommending that the interval should not exceed 2 years, but this is based on a low level of evidence,4 which does not take individual risk assessment into account. Alternatively, the FIT concentration threshold for positivity could be differentiated by age. A recently published Dutch protocol describes how they aim to tailor personalised screening intervals based on prior faecal haemoglobin concentrations in the PERFECT-FIT trial.36 The results from that trial will hopefully shed light on the effects of such a strategy, both in terms of clinical outcomes, social inequalities and cost-effectiveness. By employing a ‘one screening interval fits all’ strategy, we are leaving valuable and easy accessible information unused.


In a large sample of FIT negative screening participants, the proportion of interval colorectal cancer increased as income decreased. This was explained by the association of lower income individuals more often being of older age and having higher, yet negative, FIT concentrations than those with higher income. We propose individualising the screening interval in FIT negative participants, based on individual risk assessments from age and FIT concentration in order to decrease the total number of interval cancers while limiting the social gradient while likely increasing the overall efficiency of the programme.

Data availability statement

No data are available. Data were stored on governmental secure logged servers and were pseudoanonymised before the authors were granted access. These data can only be accessed after individual approval has been granted and the data contains personal individual information. the data can therefore not be shared.

Ethics statements

Patient consent for publication

Ethics approval

Data were stored on governmental secure logged servers and were pseudoanonymised before the authors were granted access. The study was approved by the Danish Data Protection Agency (journals 19/32137). No ethical approval or consent forms were needed according to Danish law, as a register-based approach was employed. This work was supported by Odense University Hospital grant number A3610.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter @UlrikDeding

  • Contributors UD and MK-L conceived the idea for the study. UD and HB designed the statistical analysis plan. UD completed the data management and statistical analysis under the supervision of HB and GB. UD completed the first draft of the paper. UD, MK-L, HB, LK, MKT and GB revised and contributed to the manuscript. UD is the guarantor of the study.

  • Funding This work was supported by Odense University Hospital grant number A3610.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.