Predictive modelling of tumor response and progression in cervical and head and neck cancers: a retrospective and prospective validation study

Background: Precision oncology aims to tailor treatment based on individual tumor biology and patient-specific factors. Predictive modeling using imaging-derived parameters such as fat score, blood vessel density, necrosis, and immune cell infiltration may enhance early assessment of therapeutic response. This study aims to assess the predictive validity of a novel scoring system for tumor shrinkage, progression-free survival (PFS), and time to maximum response using retrospective data from 500 patients and validate these findings prospectively in 200 patients with cervical and head and neck cancers. Methods: A retrospective cohort of 500 subjects (50% cervical cancer, 50% head and neck cancer) with varied staging was analyzed to correlate imaging and pathological scores with actual treatment outcomes. A predictive model was applied and validated prospectively on an independent cohort of 200 patients. Parameters included fat content, blood vessel density, necrosis, and immune cell density scores, culminating in an overall score. Model-predicted vs actual outcomes were compared using 90% concordance threshold. Results: Retrospective data revealed strong correlation between high overall scores and favorable treatment response, including earlier time to maximum shrinkage (mean: 12.1 weeks), higher tumor regression (>50%), and longer PFS (mean: 18.3 months). The prospective cohort confirmed these findings with a 91% model concordance for time to response and 89% for PFS prediction. Multivariate regression highlighted blood vessel density and immune infiltration as the strongest predictors. Conclusions: The proposed composite scoring system shows promise in predicting therapeutic outcomes and could guide early adaptive therapeutic strategies. Further multi-center validation is warranted.