NEW research suggests that a predictive model combining clinical and endoscopic features can improve differentiation between colorectal sessile serrated lesions (SSLs) and hyperplastic polyps, a distinction that is critical for early colorectal cancer prevention.
Sessile serrated lesions are increasingly recognised as important precursors in the serrated pathway to colorectal cancer, yet they can be difficult to distinguish from benign hyperplastic polyps during routine endoscopy and pathological assessment. Misclassification may lead to missed opportunities for appropriate surveillance and intervention.
Building a Predictive Model
In this study, researchers analysed data from 1,628 patients to identify factors associated with SSLs and develop a nomogram-based prediction model. Patients were divided into training and validation cohorts, and multivariable logistic regression was used to determine independent predictors.
Eight variables emerged as significant predictors of SSLs: age, hyperlipidaemia, diarrhoea, Helicobacter pylori infection, gastric neoplasms, polyp size, lesion location, and laterally spreading morphology. These factors were incorporated into a predictive model designed to support clinical decision-making.
Strong Diagnostic Performance
The model demonstrated high discriminatory ability, with area under the curve values of 0.876 in the training dataset and 0.880 in the validation cohort. At the optimal threshold, the model achieved high specificity of over 90%, indicating strong ability to correctly identify SSLs, although sensitivity was more moderate.
Calibration analysis showed good agreement between predicted and observed outcomes, while decision curve analysis suggested meaningful clinical benefit across a range of diagnostic thresholds.
Implications for Clinical Practice
Accurate differentiation between SSLs and hyperplastic polyps remains a key challenge in colorectal cancer prevention. The authors suggest that incorporating structured prediction tools into clinical workflows could support more precise identification of high-risk lesions, particularly in cases where visual or histological features are ambiguous.
Improved identification of SSLs may also help guide appropriate endoscopic resection strategies and inform surveillance intervals, ultimately reducing the risk of progression to malignancy.
Looking Ahead
While the model shows strong internal validation, further external validation in diverse populations will be necessary before widespread adoption. Nonetheless, the findings highlight the growing role of data-driven tools in enhancing diagnostic accuracy and supporting personalised approaches to colorectal cancer prevention.
