KEY POINTS
- This systematic review searched PubMed, Embase, and the Cochrane Library through February 2026. Of 840 identified records, 38 studies were included after screening, according to the PRISMA diagram.
- Approximately 20% of patients with locally advanced rectal cancer achieve pathological complete response after neoadjuvant chemoradiotherapy. Reliable pretreatment prediction could support organ-preservation strategies, but the review found no single sufficiently accurate predictor.
- Consistently favourable clinicopathological features included lower clinical T and N stage, node-negative disease, tumour length below 5 cm, and smaller gross tumour volumes, with proposed thresholds of approximately 37–40 cm³. Tumours located 6–8 cm from the anal verge may have higher response rates than very low or high lesions.
- Blood-based candidates included baseline and dynamic carcinoembryonic antigen, CA19-9, inflammatory indices, circulating tumour DNA clearance, and serum metabolites. However, findings were inconsistent across studies, and circulating tumour DNA alone showed limited specificity.
- Functional imaging and radiomics may add information beyond conventional MRI. Diffusion-weighted MRI, amide proton transfer-weighted imaging, PET-based imaging, delta-radiomics, and multimodal MRI models were repeatedly associated with improved prediction, but acquisition and analysis methods were not standardized.
- Molecular candidates included mismatch-repair status, immune-cell composition, KRAS and TP53 alterations, metabolomic signatures, and gut microbiota. Increased CD8-positive immune infiltration and reduced regulatory T-cell populations were associated with favourable response, while many genomic associations remained conflicting.
- Integrated clinical, blood, radiomic, deep-learning, and multi-omics models generally outperformed single-domain models. Most supporting studies were retrospective, small, or single-centre, and the review did not identify a model sufficiently validated to determine watch-and-wait eligibility in routine practice.
CLINICAL TAKEAWAY
Pathological complete response should not be predicted from one clinical variable, biomarker, or radiomic score. Multimodal models are the most credible direction, but prospective multicentre validation and standardized imaging and biomarker workflows are required before they can guide organ preservation.