Prognostic models for predicting postoperative recurrence in Crohn's disease: a systematic review and critical appraisal

Background and Aims: Prophylaxis of postoperative recurrence is an intractable problem for clinicians and patients with Crohn's disease. Prognostic models are effective tools for patient stratification and personalised management. This systematic review aimed to provide an overview and critically appraise the existing models for predicting postoperative recurrence of Crohn's disease.Methods: Systematic retrieval was performed using PubMed and Web of Science in January 2022. Original articles on prognostic models for predicting postoperative recurrence of Crohn's disease were included in the analysis. The risk of bias was assessed using the Prediction Model Risk of Bias Assessment (PROBAST) tool. This study was registered with the International Prospective Register of Systematic Reviews (PROSPERO; number CRD42022311737).Results: In total, 1948 articles were screened, of which 15 were ultimately considered. Twelve studies developed 15 new prognostic models for Crohn's disease and the other three validated the performance of three existing models. Seven models utilised regression algorithms, six utilised scoring indices, and five utilised machine learning. The area under the receiver operating characteristic curve of the models ranged from 0.51 to 0.97. Six models showed good discrimination, with an area under the receiver operating characteristic curve of >0.80. All models were determined to have a high risk of bias in modelling or analysis, while they were at low risk of applicability concerns.Conclusions: Prognostic models have great potential for facilitating the assessment of postoperative recurrence risk in patients with Crohn's disease. Existing prognostic models require further validation regarding their reliability and applicability.

Automated summary

This systematic review evaluates existing models for predicting postoperative recurrence of Crohn's disease and finds that prognostic models have great potential for assessing the risk of recurrence. However, further validation is needed for the reliability and applicability of existing models.