Related Experiment Video
Updated: May 27, 2026
Competing-Risk Nomogram for Predicting Cancer-Specific Survival in Multiple Primary Colorectal Cancer Patients after Surgery
Published on: September 27, 2024
Prognostic models for outcome following liver resection for colorectal cancer metastases: A systematic review
L Spelt1, B Andersson, J Nilsson
1Department of Surgery, Clinical Sciences Lund, Lund University, Lund, Sweden. lidewij.spelt@med.lu.se
Background:
Liver resection provides the best chance for cure in colorectal cancer (CRC) liver metastases. A variety of factors that might influence survival and recurrence have been identified. Predictive models can help in risk stratification, to determine multidisciplinary treatment and follow-up for individual patients.
Aims:
To systematically review available prognostic models described for outcome following resection of CRC liver metastases and to assess their differences and applicability.
Methods:
The Pubmed, Embase and Cochrane Library databases were searched for articles proposing a prognostic model or risk stratification system for resection of CRC liver metastases. Search terms included 'colorectal', 'liver', 'metastasis', 'resection', 'prognosis' and 'prediction'. The articles were systematically reviewed.
Results:
Fifteen prognostic systems were identified, published between 1996 and 2009. The median study population was 305 patients and the median follow-up was 32 months. All studies used Cox proportional hazards for multi-variable analysis. No prognostic factor was common in all models, though there was a tendency towards the number of metastases, CRC spread to lymph nodes, maximum size of metastases, preoperative CEA level and extrahepatic spread as representing independent risk factors. Seven models assigned more weight to selected factors considered of higher predictive value.
Conclusion:
The existing predictive models are diverse and their prognostic factors are often not weighed according to their impact. For the development of future predictive models, the complex relations within datasets and differences in relevance of individual factors should be taken into account, for example by using artificial neural networks.
