The Colorectal cancer RISk Prediction (CRISP) trial: a randomised controlled trial of a decision support tool for risk-stratified colorectal cancer screening

Abstract

Background: A risk-stratified approach to colorectal cancer (CRC) screening could result in a more acceptable balance of benefits and harms, and be more cost-effective.

Aim: To determine the effect of a consultation in general practice using a computerised risk assessment and decision support tool (Colorectal cancer RISk Prediction, CRISP) on risk-appropriate CRC screening.

Design and setting: Randomised controlled trial in 10 general practices in Melbourne, Australia, from May 2017 to May 2018.

Method: Participants were recruited from a consecutive sample of patients aged 50-74 years attending their GP. Intervention consultations included CRC risk assessment using the CRISP tool and discussion of CRC screening recommendations. Control group consultations focused on lifestyle CRC risk factors. The primary outcome was risk-appropriate CRC screening at 12 months.

Results: A total of 734 participants (65.1% of eligible patients) were randomised (369 intervention, 365 control); the primary outcome was determined for 722 (362 intervention, 360 control). There was a 6.5% absolute increase (95% confidence interval [CI] = -0.28 to 13.2) in risk-appropriate screening in the intervention compared with the control group (71.5% versus 65.0%; odds ratio [OR] 1.36, 95% CI = 0.99 to 1.86, P = 0.057). In those due CRC screening during follow-up, there was a 20.3% (95% CI = 10.3 to 30.4) increase (intervention 59.8% versus control 38.9%; OR 2.31, 95% CI = 1.51 to 3.53, P<0.001) principally by increasing faecal occult blood testing in those at average risk.

Conclusion: A risk assessment and decision support tool increases risk-appropriate CRC screening in those due screening. The CRISP intervention could commence in people in their fifth decade to ensure people start CRC screening at the optimal age with the most cost-effective test.

Keywords: clinical decision support; colorectal neoplasms; early detection of cancer; general practice; primary care; referral and consultation.

Figure 1.

Consort diagram showing number of participants assessed for eligibility, randomised, and available for analysis of primary and secondary outcomes. ^a Denominator for percentages is those who received the allocated intervention in each group (366 for intervention group, 364 for control group). ^bParticipants who did not meet the inclusion criteria were excluded from all analyses. ^cParticipants who withdrew all data or who died before the primary endpoint were excluded from all analyses (four in each group). ^dThe number of participants remaining for analysis of the primary outcome at each timepoint. ^eThe number of participants who returned questionnaires at each timepoint contributing data to the analysis of secondary outcomes.

Figure 2.

Forest plot showing estimated effect sizes including planned sensitivity and subgroup analyses. ^a Difference in the percentage with appropriate CRC screening at 12-month follow-up between the intervention and control groups and respective 95% CI estimated using generalised linear model with the identity link function and binomial family adjusted for the general practice. ^bAccounting for data on CRC screening at baseline available to the GP to determine the type of CRC screening that was due. ^cExcluding three participants in control group with self-reported outcomes only. ^dEffect modification by whether participants were due CRC screening during 12-month follow-up (n = 184 in the intervention group and n = 180 in the control group) or not (n = 178 in the intervention group and n = 180 in the control group); interaction term estimated for between-group difference on the percentage scale 27.6% (95% CI = 15.7 to 39.5), P-value for effect modification <0.001. CRC = colorectal cancer.