Imaging biomarkers for early detection, monitoring, and prediction of treatment- related side effects after radiotherapy: A scoping review

Technical Summary: Imaging Biomarkers for Radiotherapy-Related Side Effects

Problem Statement
Radiotherapy is a cornerstone of cancer treatment, yet radiation-induced normal tissue damage remains a significant clinical challenge, leading to substantial morbidity and reduced quality of life. Current assessment of these side effects relies heavily on subjective clinician-reported grading systems (e.g., CTCAE) and patient-reported outcome measures (PROMs), which suffer from inter-observer variability and often detect injury only after symptoms manifest. While imaging biomarkers offer the potential for objective, quantitative, and early detection of tissue injury, their translation from research to routine clinical practice has been limited. The specific problem addressed is the disconnect between the growing volume of translational research on imaging biomarkers for radiotherapy-induced side effects and their actual clinical implementation and regulatory approval.

Methodology
This study is a scoping review conducted in accordance with PRISMA-ScR guidelines. The authors performed systematic searches across MEDLINE, Embase, and the Cochrane Library from inception through January 2025.

• Selection Criteria: The review identified studies focusing on imaging biomarkers for the detection, monitoring, or prediction of radiotherapy-related side effects.
• Data Extraction: Two independent reviewers screened 2,733 unique records, resulting in 102 included studies (published 2007–2024). Data were extracted using a standardized template covering study design, cancer type, imaging modality, validation approach (internal/external/technical), clinical endpoints, and adherence to reporting frameworks (e.g., FDA-NIH BEST categories, EQUATOR checklists).
• Analysis: Quantitative analysis was performed to characterize the landscape of research, including the distribution of biomarker categories (predictive, diagnostic, prognostic, monitoring), validation status, and clinical implementation.

Key Results

• Study Landscape: The 102 included studies were predominantly single-institution (85%) and retrospective. Neurologic and head and neck cancers accounted for 67% of the studies.
• Biomarker Categories: Predictive biomarkers were the most common (53%), followed by diagnostic (25%), prognostic (22%), and monitoring (18%).
• Imaging Modalities: Magnetic Resonance Imaging (MRI) was the dominant modality (67.6%), followed by CT (19.6%) and PET-CT (12.7%).
• Validation and Standardization: A critical finding is the lack of rigorous validation. 57% of studies did not technically validate their biomarkers, and external validation occurred in only 6% of studies. No studies reported clinical implementation or regulatory approval.
• Reporting Quality: None of the included studies utilized an Enhancing the QUAlity and Transparency Of health Research (EQUATOR) checklist, despite 85% of studies having clear inclusion/exclusion workflows and many reporting guidelines being available at the time of publication.
• Specific Clinical Endpoints:
  • Neurologic: Research focused heavily on distinguishing radiation necrosis from tumor recurrence (diagnostic) and predicting cognitive decline. MRI, particularly advanced functional sequences, showed promise but relied on small, retrospective cohorts.
  • Head and Neck: Studies investigated xerostomia and dysphagia. While some imaging features (e.g., parotid CT texture, FDG-PET uptake) improved prediction over dose-based models, these findings were derived from small internal cohorts without external validation.
• Machine Learning: 34% of studies employed machine learning or deep learning, yet the limited use of external validation suggests these models are not yet ready for clinical deployment.

Key Contributions

• Mapping the Translational Gap: The review provides a comprehensive map of the current evidence base, explicitly highlighting that despite growing research interest, imaging biomarkers for radiotherapy toxicity remain almost entirely investigational.
• Framework Application: The authors apply established translational frameworks (O'Connor et al. roadmap, FDA-NIH BEST classification) to the literature, revealing that the majority of biomarkers are stuck at the earliest stages of development (technical validation) and have not progressed to clinical validation or utility demonstration.
• Identification of Systemic Barriers: The paper identifies specific methodological shortcomings hindering translation, including the predominance of single-center studies, lack of external validation, insufficient technical standardization (e.g., QIBA, IBSI), and poor adherence to reporting guidelines.
• Reclassification Proposal: The authors suggest that all imaging biomarkers developed for radiotherapy side effects should be explicitly framed as "safety biomarkers" under the BEST taxonomy, which may offer clearer regulatory pathways for qualification.

Significance and Claims
The paper claims that while imaging biomarkers hold considerable potential for precision radiation oncology, the field has not yet achieved the rigor required for clinical adoption. The significance of this review lies in its identification of the specific barriers preventing translation:

1. Lack of Validation: The scarcity of external and technical validation prevents the generalizability of findings.
2. Methodological Heterogeneity: Variability in image acquisition, processing, and reporting undermines reproducibility.
3. Regulatory Uncertainty: The absence of studies reporting regulatory approval or clinical implementation indicates a failure to engage with qualification pathways.

The authors conclude that bridging this translational gap requires a shift toward rigorous external validation, methodological standardization, and structured pathways toward regulatory qualification. They emphasize that future research must move beyond single-institution retrospective cohorts toward prospective, multi-center studies embedded in clinical trial infrastructure to demonstrate clinical utility. The paper does not propose new experimental applications but rather calls for a systematic overhaul of how these biomarkers are developed and validated to achieve meaningful clinical impact.