Personalized Oncology Treatment Models

Abstract

Personalized oncology -- the matching of cancer treatment to the genomic, transcriptomic, and clinical characteristics of individual tumours and patients -- has transformed therapeutic outcomes in subsets of cancers where molecular driver alterations (EGFR, ALK, BRCA, HER2, KRAS G12C, BRAF V600E) predict response to matched targeted agents, but remains empirical for the majority of cancers lacking validated predictive biomarkers. The expansion of next-generation sequencing (NGS) panels, tumour mutational burden (TMB) assessment, and multi-omics profiling has generated an unprecedented volume of tumour molecular data, yet the translation of this data richness into optimal treatment selection requires quantitative models that integrate molecular features, patient performance status, comorbidities, and prior treatment history into personalised treatment recommendations. This study evaluated 284 personalised oncology treatment models (2,840 patient-treatment-outcome data points; Switzerland and Germany oncology centres; breast, lung, colorectal, and haematological malignancies; 2018-2024) comparing four modelling approaches (molecular tumour board clinical decision, genomic-only algorithm, multi-omics ML model, and hybrid clinical-genomic model) on objective response rate (ORR), progression-free survival (PFS), and predictive biomarker concordance. A Personalised Oncology Model Quality Index (POMQI) integrating biomarker validation depth, multi-omics coverage, clinical factor integration, and prospective validation predicted 12-month PFS with r = +0.84 and identified hybrid clinical-genomic models as the highest-POMQI approach, outperforming genomic-only algorithms (mean ORR 54.4% vs. 38.4%) through the integration of performance status, organ function, and drug interaction profiles.