Uncovering the Understanding of the Concept of Patient Similarity in Cancer Research and Treatment: Scoping Review

Patient similarity is a fundamental concept in precision oncology, offering a pathway to personalized medicine by identifying patterns and shared characteristics among patients. This concept enables stratification into clinically meaningful subgroups, prediction of treatment responses, and the tailoring of therapeutic interventions to individual needs. Despite its transformative potential, the definition, measurement, and clinical application of patient similarity remain inconsistently established, creating challenges in its integration into cancer research and clinical practice.

This study aimed to synthesize evidence on the multidimensional concept of patient similarity in cancer research by analyzing its application across different points of possible data types, methodological frameworks, biological contexts, and commonly studied cancer types.

This scoping review followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) framework and the Joanna Briggs Institute guidelines. A systematic search was conducted across PubMed, MEDLINE, LIVIVO, and Web of Science (covering the period from 1998 to February 2024) and was supplemented by snowball sampling and manual searches. Duplicate records were removed, and study selection was carried out in 3 phases: title and abstract screening, disagreement resolution, and full-text screening. Each step was independently performed by 2 reviewers in Rayyan, with conflicts resolved by a third reviewer. Data extraction was performed using a predefined template to capture methodological approaches, data types, cancer types, and research objectives related to similarity in patients with cancer.

This scoping review synthesized evidence from 137 studies, emphasizing the multidimensional concept of patient similarity in cancer research, which integrates diverse data types, methodological frameworks, research objectives, and cancer types. Transcriptomic data (92/137, 67.1%) and clinical data (65/137, 47.4%) were the most frequently used, often combined to enhance the comprehensiveness of similarity analyses. Machine learning (76/137, 55.5%) and network-based approaches (72/137, 52.5%) were prominent methods, reflecting their capacity to handle complex, high-dimensional data and uncover intricate relationships. Cancer subtype identification (70/137, 51.1%) and biomarker discovery (41/137, 29.9%) were the primary research objectives, underscoring the centrality of patient similarity in precision oncology. Breast, lung, and brain cancers were the most frequently studied, benefiting from established research frameworks and abundant datasets. Conversely, rare cancers were underrepresented, highlighting a critical gap in the generalizability of current methodologies.

This comprehensive scoping review examines the concept of patient similarity in cancer research and highlights the critical role of a multilayered perspective in capturing its complexity and identification to enhance understanding and application in precision oncology.