Checkpoint inhibitors have revolutionised cancer treatment, yet only 20-30 % of patients achieve durable responses, highlighting the critical need for predictive models. This review focuses on PD-1/PD-L1 pathway inhibitors as monotherapy, examining current prediction frameworks spanning biomarker-based approaches, multi-omics integration, mathematical modelling, and artificial intelligence applications. Recent advances include SCORPIO and LORIS machine learning systems demonstrating superior statistical performance compared to traditional biomarkers, with area under curve values of 0.763. However, critical analysis reveals significant limitations in external validation across diverse healthcare settings, with many promising models failing to maintain performance outside their development institutions. Traditional pathological assessment by expert pathologists, including standardised PD-L1 scoring and tumour-infiltrating lymphocyte quantification, continues to form the foundation of clinical decision-making and provides essential validation for emerging AI approaches. Despite extensive research, established biomarkers show limited predictive accuracy, with PD-L1 demonstrating predictive value in only 28.9 % of FDA approvals. Multi-feature models incorporating genomic and clinical data show improved accuracy but face substantial validation challenges. Integration of spatial biomarkers and digital pathology has enhanced capabilities, achieving area under curve values of 0.84 in select studies. The most critical challenge is the "validation gap", many models show excellent single-institution performance but fail external validation, limiting clinical translation. Current obstacles include inadequate standardisation, interpretability concerns, and healthcare system integration difficulties. Future directions must prioritise rigorous multi-institutional validation studies, development of clinically implementable frameworks, and addressing practical deployment challenges to realise precision immunotherapy's potential.

Cancer reversion therapy represents a paradigm shift in oncology, focusing on reprogramming malignant cells to a non-malignant state rather than destroying them. This narrative review synthesizes current evidence, emerging technologies, and future directions in this promising field. Cancer reversion is founded on key biological observations: somatic cell reprogramming, spontaneous cancer regression, and microenvironmental influences on malignant behavior. Current approaches include epigenetic reprogramming using HDAC inhibitors and DNA methyltransferase inhibitors; microenvironmental modulation through extracellular matrix manipulation and vascular normalization; differentiation therapy exemplified by all-trans retinoic acid in acute promyelocytic leukemia; and targeting oncogene addiction as demonstrated in BCR-ABL-driven leukemias. Emerging technologies accelerating progress include single-cell analyses that reveal cancer heterogeneity and cellular state transitions; CRISPR-based approaches enabling precise genetic and epigenetic manipulation; patient-derived organoids that model tumor complexity; and artificial intelligence applications that identify novel reversion-inducing agents. Critical evaluation reveals that many reported “reversion” phenomena represent stimulus-dependent plasticity or transient growth arrest rather than stable phenotypic normalization. True cancer reversion requires durable, heritable phenotypic changes that persist after treatment withdrawal, with evidence of epigenetic consolidation and functional restoration. Despite promising advances, significant challenges remain: cancer cell plasticity facilitating therapeutic escape, difficulties in establishing stable reversion states, delivery challenges for solid tumors, and the need for combination approaches to address tumor heterogeneity. Future directions include integrated multi-omics analyses to comprehensively map cellular state transitions, studies of natural regression phenomena to identify reversion mechanisms, advanced nanodelivery systems for targeted therapy, and synthetic biology approaches creating intelligent therapeutic systems. By redirecting rather than destroying cancer cells, reversion therapy offers the potential for reduced toxicity and resistance, potentially transforming cancer from a deadly disease to a manageable condition.