Identifying Patients With Hereditary and Familial Colorectal Cancer by Using an Online Risk Tool

Design: This is a multicenter prospective comparative cohort study, using a stepped-wedge design. All clusters (hospitals) start with the control strategy and switch, one by one, to the intervention strategy. Every 9 weeks a hospital will start implementing the questionnaire after a short training period of 1 week. As a referral to a clinical geneticist can take up to several months, we allow at least 6 months for the collection of the results of the genetic tests.

Control strategy: Each hospital starts with the control strategy. Just before the start of this study, all hospitals will be sent a list with the clinical genetics referral criteria that will be used for this study. Each hospital can decide on how to use these criteria.

Intervention strategy: Before the start of the study, all hospitals will be informed when the use of the online questionnaire will be implemented. They will also receive a manual on how to use the online CRF and questionnaire. The online CRF will be used to send out invitations for the questionnaire, to report baseline characteristics (age, sex, nationality, educational level, native language), reasons for non-participation, the number of changes that need to be made when verifying completed questionnaires and to calculate referral advice. In the training week, a researcher from the AMC will present the study to all involved persons and he will facilitate the implementation of the online CRF and questionnaire. All nurses and doctors participating in this study will be provided with their own login codes for the online CRF. Patients included in the training week will not be included in the analysis. All consecutive patients with CRC that have a first appointment at the outpatient clinic will be invited to complete the online questionnaire and they receive a record number. In case patients decline to complete the questionnaire, this will be reported by a nurse or gastroenterologist in the online CRF. If the patient consents, an email will be sent to the patient with a link to the questionnaire and the question to complete it before the next visit to the outpatient clinic. The online CRF will automatically result in a check for indications for referral to a clinical geneticist. The nurse or gastroenterologist can report this advice in the patient file and can decide on referring the patient.

Checking referrals to a clinical geneticist: Every two months, all patients included in the study will be tracked by searching for reports on clinical genetics consultation in their medical files and a check with local genetic centers will be done.

Statistical analysis of the primary outcome measure: The difference between the control group and intervention group in the proportion of patients who receive surveillance advice by a clinical geneticist will be calculated. A nonlinear mixed model will be used for testing the hypothesis of no difference. We will model the probability of receiving surveillance advice as a linear function, after a logit transformation, with as explanatory variables the hospital, the time interval, and the questionnaire effect: Logit(pijk)=μ+αi+βj+Χijθ, Where Logit(pijk) is the natural logarithm of the odds of a referral to a clinical geneticist for individuals (k) in cluster i at time j. μ is the overall mean, αi is a random effect for hospital i, βj is a fixed effect corresponding to time interval j, and Χij is an indicator variable for implementation of the questionnaire in hospital i at time j and θ is the questionnaire effect. Time effects will be analyzed in a discrete way (season, month) in case a primary effect is found. If no time effect is present, time function will be excluded from the model. For the primary outcome measures, we will use the intention-to-treat principle, including all eligible patients in the analysis.

Statistical analysis of the secondary outcome measures: for each secondary outcome measure, differences between the control and intervention group will be calculated using relative risks with corresponding 95% confidence intervals. Secondary outcome measures to be calculated in the intervention group only, will be reported using descriptive statistics.

Sample size calculation: For the sample size calculation, a two-group chi-squared test of equal proportions was used as a starting point for comparing proportions in referrals in the control group versus the intervention group. Significance level was set at 5% (two sided) with a power of 80%. Of all CRC cases, approximately 20% are related to familial or hereditary factors. In the literature the reported proportion of CRC patients with an increased familial risk of CRC that is appropriately referred to a clinical geneticist ranges from 8% to 67%, with most studies reporting proportions between 10% and 30%. We estimate that 20% of patients at risk are referred, resulting in a referral rate of 4% of all CRC patients. We anticipate that the implementation of the questionnaire can increase this proportion up to 15%.To detect this difference we would need a sample size of 111 patients per treatment group. For the stepped wedge design the total sample size (111x2) needs to be multiplied by a design effect, that can be calculated using the following formula, where k is the number of steps, b is the number of baseline measurements, n is the number of patients per cluster, t is the number of measurements after each step and ρ is the intracluster correlation (ICC), which compares the within-group variance with the between-group variance:

DEsw= ((1+ρ(ktn+bn-1))/(1+ρ (0.5ktn+bn-1))) . ((3(1-ρ))/(2t(k-1/k)))

The ICC was estimated at 0.10 as no previously reported ICC for this topic was available.

As there are five participating hospitals, a total number of five steps (k) will be possible. As the study duration will be one year, each step will take 2 months. After randomization, each hospital starts with a training period of one week that will be added to these two months. We expect two to eight eligible patients per week per hospital, which results in 16 to 64 patients per hospital per step. With this information, a DEsw can be calculated. Assuming 16 patients per step, DEsw is 0.45. Assuming 64 patients per step, DEsw is 0.47. This corresponds with a required total sample size of 100 and 104 patients, respectively. When divided by the total number of patients per cluster, 6.2 and 1.6 clusters are needed for the minimum and maximum number of included patients respectively. We therefore believe that our number of five clusters would be sufficient.

Institutional review board: This study was approved by the institutional review board (IRB) of the Academic Medical Center (AMC) in Amsterdam, the Netherlands. This IRB decided that the study did not meet the requirements of the Medical Research Involving Human Subjects Act (WMO). All local ethics committees confirmed the local practicability of this study.

Consent: According to the requirements of the AMC Research Code, digital approval will be asked from the patients for making use of personal data for this study in the intervention group. This means that questionnaires start with a question about the willingness to participate in this study. They can answer this question by indicating 'yes' or 'no'. If 'no' is answered, results will be removed from the database. Use of the digital questionnaire is considered safe as all answers are collected anonymously and could only be linked to a patient with a secured key document. This study will be carried out in accordance with the Helsinki Declaration. No approval for data use will be asked from patients in the control group, as this would increase awareness and would therefore bias results. Moreover, data from these patients will only be collected anonymously.

Monitoring and quality assurance: The AMC researcher will monitor all centers every month. Each site that is randomized will be visited for a training week.