Prognostic models and predictors of recurrence in resectable colorectal liver metastases: A systematic review and meta-analysis.
Lorenzo Manganaro, Tommaso Russo, Lisa Novello et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis assessed prognostic models and recurrence predictors after curative-intent locoregional treatment of colorectal liver metastases. It included 293 studies, with 85,150 patients in model-development cohorts and 5549 in validation cohorts. External validation was uncommon, performance reporting varied, and methodological quality was generally limited. Across the pooled evidence, postoperative circulating tumor DNA positivity was the strongest identified predictor of recurrence (HR 4.78); several clinicopathological and treatment-response factors were also associated with recurrence-free survival. The authors conclude that existing models incompletely represent the biological heterogeneity of recurrence.
Key findings
- Postoperative circulating tumor DNA positivity was the strongest identified recurrence predictor (HR 4.78).
- Other associated factors included postoperative carcinoembryonic antigen > 5 ng/mL (HR 2.79), poor neoadjuvant response (HR 2.76), and positive resection margins (HR 1.73).
- External validation was reported in only 33/293 studies, and heterogeneous performance reporting limited cross-study comparison.
- Adjuvant and peri-operative systemic therapies were associated with lower recurrence risk.
Why it matters
Recurrence risk after curative-intent treatment is associated with tumour, biomarker, and treatment-response features, but current models have limited validation and inconsistent reporting. This review helps distinguish predictors with repeated evidence from models whose performance and clinical usefulness remain difficult to compare. It also highlights the potential value of dynamic biomarkers in future risk-stratification approaches.
What to keep in mind
The review reports limited overall methodological quality, uncommon external validation, and heterogeneous performance reporting, restricting cross-study comparability and clinical translation.