Individual risk modelling for esophagectomy: a systematic review

Area of Science:

• Surgical Oncology
• Medical Statistics
• Health Services Research

Background:

• Esophagectomy outcome prediction models are crucial for surgical decision-making.
• Existing models require evaluation for clinical credibility, methodological rigor, and predictive performance.

Purpose of the Study:

• To systematically review and compare existing models for predicting outcomes after esophagectomy.
• To assess the clinical credibility, methodological quality, and performance of these predictive models.

Main Methods:

• A systematic review of PubMed, EMBASE, and Cochrane databases was conducted.
• Model and study quality were appraised using the Minne et al. framework.
• Data from twenty retrospective studies published over several years were analyzed.

Main Results:

• All models utilized logistic regression; heterogeneity and retrospective designs were common.
• P-POSSUM demonstrated superior calibration and discrimination for mortality prediction.
• Few models, excluding Southampton and Amsterdam scores, effectively predicted surgical complications.

Conclusions:

• Current esophagectomy outcome models have significant limitations, necessitating larger, contemporary validation studies.
• Further research is required to refine existing models and explore novel techniques like artificial neural networks.