A clinical predictive model can effectively stratify the risk of colorectal cancer among patients undergoing colonoscopy. A multicenter, prospective derivation and validation study

Aims The balance between prescription of colonoscopies and healthcare resources has been broken, leading to long waiting lists. Therefore, a triage system to define colonoscopy priority is needed. We recently demonstrated that Italian RAO criteria can effectively stratify CRC risk among patients undergoing colonoscopy. Aim of the present study was to derive and validate a predictive model for colorectal cancer (CRC) based on such criteria, aiming to risk-stratify patients undergoing colonoscopy outside organized screening programs.

Methods Multicenter prospective observational study involving 19 institutions in Italy. Consecutive adult patients undergoing colonoscopy for clinical indication outside CRC screening organized programs were eligible to be included. Colonoscopies with inadequate bowel cleansing and incomplete were excluded. RAO criteria comprise three groups of priority to colonoscopy based on different clinical indications, namely “B” (high priority), “D” (intermediate priority), and “P” (low priority). Predictive models for CRC were derived through multivariable logistic regression analysis, and their discriminant power assessed according to area under the ROC curve (AUROC). Three cohorts were considered, i.e. (i) the derivation cohort including patients enrolled in Emilia-Romagna Centers in three consecutive months for each Center between November 2022 and May 2023, (ii) the geographical validation cohort in Italy outside Emilia-Romagna, and (iii) the prospective validation cohort in Emilia-Romagna from June to September 2023.

Results 6,691 patients (mean age 61 years, female sex 50.2%) were included. CRC was found in 183 (2.7%) cases. In the derivation cohort, predictive models for CRC diagnosis included: age 60-69 (OR 3.54, CI 1.28-9.82) and age≥70 (OR 4.91, CI 1.87-12.89), no previous colonoscopy in 10 years (OR 2.94, CI 1.61-5.26), and indication to colonoscopy alternatively according to RAO “D” (OR 5.42, CI 2.25-13.08) and “B” (OR 24.80, CI 11.05-55.69) priority criteria, ASGE criteria (OR 2.54, CI 1.54-4.2) or presence of≥1 alarm feature (OR 2.52, CI 1.57-4.06). The model based on RAO criteria outperformed models based on ASGE criteria or alarm features, yielding significantly higher discriminant power (AUROC 0.85 vs. 0.75 and 0.74, p<0.01). The model was geographically and prospectively validated with AUROC 0.81 and 0.84, respectively. We defined three CRC risk categories, i.e. low-risk (CRC<1%), intermediate-risk (CRC 1-5%), and high-risk (CRC≥5%). When considering both the validation cohorts (n=4,158), if we gave low priority to low-risk patients, we would delay 2,412 (58%) colonoscopies with 20 (0.8%) CRCs, whereas giving high priority to high-risk patients would anticipate 636 (15.3%) colonoscopies with 69 (10.9%) CRCs.

Conclusions We derived a simple and accurate clinical prediction tool to stratify patient CRC risk before prescribing colonoscopy, with two-fold implication: on one hand, it might help clinicians in appropriate prescription of colonoscopy; on the other hand, it might help healthcare providers to define the priority of colonoscopies and optimize resource allocation. [1] [2] [3]

Publication History

Article published online:
15 April 2024

• References

• 1 Juillerat P, Peytremann-Bridevaux I, Vader J-P. et al. Appropriateness of colonoscopy in Europe (EPAGE II). Presentation of methodology, general results, and analysis of complications. Endoscopy 2009; 41: 240-6
• 2 ASGE Standards of Practice Committee. Appropriate use of GI endoscopy. Gastrointest Endosc 2012; 75 (06) 1128-1131
• 3 AGENAS. Manuale RAO (Raggruppamenti di Attesa Omogenei). PNGLA 2019-2021; AGENAS, Roma 2020