Combining Risk Factors and Faecal Immunochemical Testing in Colorectal Cancer Screening: a Randomized Controlled Trial

INTRODUCTION AND RATIONALE Colorectal Cancer (CRC) is the third most frequent diagnosed cancer worldwide, with 1.4 million new cases every year. Mortality of CRC is estimated at around 40%. To reduce the incidence of CRC and to abate the negative consequences associated with CRC, methods for early detection of (pre)malignant colorectal lesions have extensively been researched and implemented. Colonoscopy is the reference standard for the detection of advanced neoplasia, but it requires trained endoscopists, carries a high burden, harbours a risk for complications in patients, and increases societal costs.

Most screening programs use methods to select screening participants with a high risk of advanced neoplasia for colonoscopy. One such method is the detection of blood in stool with faecal immunochemical tests (FITs).

Multiple nation-wide screening programs have now been implemented using FIT, including one in the Netherlands. In the Dutch CRC screening program, using a FIT threshold of 47 µg Hb/g faeces, the sensitivity of FIT in detecting CRC is 85.5%. The sensitivity of FIT in detecting advanced neoplasia (AN) is however substantially lower: 38%.

FIT-based screening also generates a high number of false positives, leading to unnecessary colonoscopies and distress for those patients. Approximately 1 in 2 FIT-positives has advanced neoplasia; the first round of the Dutch CRC screening program, using a FIT threshold of 47 µg Hb/g faeces, reached a positive predictive value of 54%.

Previous studies have demonstrated the existence of clinical risk factors for CRC, such as male gender, age, family history of CRC, and smoking. In a ZonMW sponsored project (ZonMW 50-50115-96-521) our group has shown that higher age, male sex, a family history with close relatives in whom CRC was diagnosed, and active smoking were all associated with the presence of advanced neoplasia (Stegeman et al, Gut, 2014).

Multiple CRC risk models have been developed using these risk factors, but most show only weak discriminatory power. Others have proposed risk models that additionally included molecular markers, increasing performance and discriminatory power, but these markers can be cumbersome to collect and can be costly in respect to the relative cheap FIT test. the investigators propose to combine CRC risk factors with the quantitative FIT result. FIT is a quantitative test with a variable cut-off level, and individuals with higher values of faecal haemoglobin are at higher risk of having advanced neoplasia at the cost of a lower specificity. Therefore, the quantitative result can be used as a weight-factor in a risk model.

Based on information on these risk factors for CRC and the quantitative FIT-result, our FIT-based risk model for CRC screening calculates the personal risk of having advanced neoplasia at colonoscopy. Inviting screening participants with elevated risk for colonoscopy, rather than those whose FIT result exceeds a pre-specified threshold, has the potential to increase the effectiveness and efficiency of FIT-based CRC screening. However, this requires participants to complete the questions about risk.

Our hypothesis is that integration of FIT score with known risk factors (age, gender, smoking, and family history of CRC) improves the number of participants in whom advanced neoplasia is detected, relative to the number of invitees. The investigators will compare a risk-based approach to screening, as described above, to the conventional FIT-only screening in a randomized controlled screening trial.

OBJECTIVES The study hypothesis is that risk-based CRC screening significantly outperforms FIT-only CRC screening in terms of the relative number of invitees in whom advanced neoplasia is detected at colonoscopy.

Primary Research Question: does risk-based CRC screening result in a higher relative number of invitees in whom advanced neoplasia is detected at colonoscopy compared to FIT-only CRC screening?

Secondary Research Questions:

1. Does risk-based CRC screening result in a higher standardized screening yield in whom advanced neoplasia is detected at colonoscopy compared to FIT-only CRC screening? ii) Is the participation rate of risk-based CRC screening non-inferior compared to FIT-only screening? iii) Does risk-based CRC screening result in a higher relative number of invitees in whom advanced neoplasia is detected at colonoscopy compared to FIT-only CRC screening at higher FIT-positivity thresholds (from 15 µg Hb /g faeces)? iv) Does risk-based CRC screening result in a higher relative number of invitees in whom proximally located advanced neoplasia is detected at colonoscopy compared to FIT-only CRC screening?

METHODS The investigators will perform a randomized controlled trial in 23,000 randomly selected men and women (aged 55 to 75 years) eligible for participation in the national CRC screening programme. All eligible participants will be second time invitees, living within a 25 km radius from the colonoscopy centre Bergman Clinics Amsterdam, which is accredited for the national population screening program.

Eligible individuals are randomly allocated on a 1:1 basis to an invitation for either risk-based CRC screening (intervention group) or FIT-only screening (control group). Randomization per household will be done before invitation. Study invitations are sent out by Foundation of Population Screening Mid-West (BoMW) early November 2019 before sending the FIT sample kit of the national screening program to the study cohort from January 2020 to April 2020. The study invitation is combined with a leaflet containing information on study participation. It will also contain an informed consent form, and a one-page questionnaire (if allocated to the intervention arm) with a self-addressed envelope to the research facility. The informed consent form will bear a unique study identification code.

The model that the investigators will use in this study for risk-based screening is derived from a model developed in a primary colonoscopy screening trial (Stegeman et al, Gut, 2014). The investigators have re-evaluated our multivariable risk model for CRC, with the aim to simplify it for daily practice. A number of risk factors (such as elements from diet) added little to the performance of the model, whereas collecting the necessary information proved rather tedious. The quantitative FIT result, age, sex, family history and smoking behaviour were the most decisive factors in the model.

This simplified FIT-based risk model had a performance like that of the extended model in the development dataset. In this study the investigators will calculate the risk of harbouring advanced neoplasia with our simplified FIT-based risk model, based on the quantitative FIT result, age, sex, family history and smoking behaviour (see below). The threshold and expected positivity rate of the risk-based model will be matched to the expected positivity rate of the FIT-only group.

Regression equation of the updated model:

ln(odds AN) = -4.955073942+0.9196052173*Smoking+0.3364761549* √FIT+0.370575346*family history of CRC-0.006934851*FIT+0.0228752117*age+0.070972656*gender

Probability of advanced neoplasia:

P (AN)=e^((odds AN))/(1+e^((odds AN))) If invitees that are allocated to the intervention group (i.e. questionnaire and FIT) only return the questionnaire, they do not qualify for a colonoscopy. When consenting invitees in the intervention group prefer not to complete the questionnaire, FIT tests will be analysed at a cut-off of 15 µg Hb/g faeces.

The type of FIT that will be used is the one used in the Dutch national screening programme: FOB-Gold (Sentinel, Italy). FIT-analyses and sending of FIT results will be performed according to the logistics as being used in the national screening. The FIT result and the result of the questionnaire will be entered in a database, to calculate the participant's risk. All FIT positives (≥15 µg Hb/g faeces) will be referred for consultation for colonoscopy at regional endoscopic centres. For the purpose of this study, FIT-negative but risk-positive participants (i.e. the combination of the results of the questionnaire and the quantitative FIT result) will also be invited for consultation for colonoscopy.

Colonoscopy will be performed by endoscopists accredited for our national screening programme and according to national and international quality guidelines. Quality indicators and endoscopic findings will be systematically recorded in a database, according to the standards of the national screening programme. Lesions will be immediately removed, if possible, and otherwise biopsies will be obtained. Histopathological assessment of these tissue samples will provide a definitive diagnosis. Data on location, size, macroscopic aspect, morphology and polypectomy will be recorded for all colonic lesions, which will be collected and evaluated by an accredited GI-pathologist according to the Vienna criteria. Advanced adenomas are defined as adenomas larger than 10 mm, adenomas with high-grade dysplasia or adenomas with a villous component of at least 25%. Advice regarding surveillance colonoscopy after removal of adenomatous polyps, large (≥10 mm) serrated lesions or cancer will be given to the clients according to the Dutch CBO consensus.

Screening invitees can leave the study at any time for any reason if they wish to do so without any consequences. The investigator can decide to withdraw a screening invitee from the study for urgent medical reasons.

STATISTICAL ANALYSIS

Primary outcome The difference in screening yield will be estimated and expressed as an absolute difference per 1,000 invitees, with 95% confidence intervals. The investigators will test the null hypothesis of no difference in yield between the intervention and control group using the chi-square test.

Secondary outcome The standardized screening yield will be calculated by ranking participants according to their calculated risk (in risk-based screening) and comparing the yield of risk-based screening for the number of positives that matches the number of FIT-positives with FIT-only screening with a chi-square test. Additionally, the investigators will use the McNemar test statistic to test the null hypothesis of no gain in screening yield in the intervention group only, where the investigators can compare referrals based on FIT only with referrals based on FIT and calculated risk.

The participation rate will be calculated through the number of participants relative to the number of invitees in each arm. Differences in participation rate will be assessed using the chi square test. The difference in screening yield of proximally located advanced neoplasia between both arms will be estimated and expressed as an absolute difference per 1,000 invitees (with 95% confidence intervals). The null hypothesis of no difference in yield will be tested using the chi square test.

Statistical significance level P-values of lower than 0.05 will be considered statistically significant. All analysis between both arms are based on the intention-to-screen principle, unless stated otherwise. Statistical analyses will be performed using R.

Additional outcomes Model performance in comparison to the FIT will be assessed with the net reclassification improvement (NRI). In addition, the yield of the FIT-based risk model and the FIT will be assessed at a broad range of FIT thresholds and will be estimated and expressed as the number of detected advanced neoplasia per 1,000 invitees (intention-to-screen) and per 1,000 colonoscopies (per protocol). Differences in the yield will be tested using a chi square test. Any missing data that was intended to be used as input for the risk-model, will be classified as '0'.

ETHICAL CONSIDERATIONS The study will be conducted according to the principles of the Declaration of Helsinki (Version October 2008) and the Population Screening Act (WBO). Permission to conduct this study has been granted by the Dutch Health Council and the Dutch Minister of Healthcare.