Identifying patients suitable for targeted adjuvant therapy: advances in the field of developing biomarkers for tumor recurrence following irradiation

ABSTRACT

Introduction

Radiation therapy (RT) is commonly used to treat cancer in conjunction with chemotherapy, immunotherapy, and targeted therapies. Despite the effectiveness of RT, tumor recurrence due to treatment resistance still lead to treatment failure. RT-specific biomarkers are currently lacking and remain challenging to investigate with existing data since, for many common malignancies, standard of care (SOC) paradigms involve the administration of RT in conjunction with other agents.

Areas Covered

Established clinically relevant biomarkers are used in surveillance, as prognostic indicators, and sometimes for treatment planning; however, the inability to intercept early recurrence or predict upfront resistance to treatment remains a significant challenge that limits the selection of patients for adjuvant therapy. We discuss attempts at intercepting early failure. We examine biomarkers that have made it into the clinic where they are used for treatment monitoring and management alteration, and novel biomarkers that lead the field with targeted adjuvant therapy seeking to harness these.

Expert Opinion

Given the growth of data correlating interventions with omic analysis toward identifying biomarkers of radiation resistance, more robust markers of recurrence that link to biology will increasingly be leveraged toward targeted adjuvant therapy to make a successful transition to the clinic in the coming years.

KEYWORDS:

• Radiation therapy (RT)
• adjuvant therapy
• biomarkers
• tumor recurrence
• radiation resistance
• precision medicine

1. Introduction

Radiation therapy (RT) employs high-energy radiation to damage DNA, intending to elicit cancer cell kill over the impact on normal tissues. Approximately 50% of cancer patients will receive RT during their treatment [1]. RT is mainly used for solid tumors such as glioblastoma, lung, breast, rectal, prostate, colorectal, cervical, esophageal, and head and neck cancers and can be delivered through either external beam therapy or internal brachytherapy and can be used as treatment with curative intent, as adjuvant therapy, neoadjuvant therapy, or as palliative care [2]. RT is often administered following surgical intervention or in conjunction with other agents based on patient circumstances.

While irradiation effectively elicits cell kill, tumor recurrence, often due to treatment resistance, occurs in most malignancies with mechanisms that can be tumor-type dependent or generalized as hallmarks of cancer. The redundancy and dynamic nature of mechanisms that contribute to radiation resistance, coupled with the lack of data on how therapy alters RT response based on molecular and proteomic features, emphasizes the need for research correlating multi-omic data to RT [3].

The optimization of tumor detection before overt progression is itself a critical step since identifying patients suitable for targeted adjuvant therapy depends on correlating outcome measures, specifically progression-free survival, to any intervention that is likely to produce a beneficial effect.

Currently, upfront standard of care (SOC) management of tumors is tumor-specific and reasonably well-defined. In contrast, tumor progression criteria carry tumor-specific definitions, with the designation of progression usually occurring at the time of follow-up after the standard of care management has been completed. Most often, progression is captured after some time has passed following completion of SOC, thus incorporating the test of time and/or the presence of active tumor tissue as might be exemplified by obtaining tissue from suspicious areas identified on physical exam or imaging, further solidifying that progression has occurred. Upon recurrence, treatment is typically far more individual and secondarily heterogeneous than management at diagnosis, leading to smaller data sets that need to be harnessed effectively, as will be discussed in our companion paper on small data sets and strategies for maximizing their use.

This leads to several problem areas in gathering and analyzing data that would identify patients suitable for targeted adjuvant therapy: 1) dealing with imperfect and heterogenous post-treatment data that is employed as a surrogate of treatment efficacy; 2) the lack of linkage of existing ‘during treatment’ and post-treatment data to pretreatment data which is that type of data typically present in large scale data repositories and finally 3) lack of biomarkers that predict progression linked to SOC and novel interventions.

This review aims to identify the challenges in intercepting early progression to identify patients suitable for targeted adjuvant therapy, discuss existing and emerging markers that have made it into the clinic, and examine advances in targeted therapies aimed at RT-resistant disease. The goal is to explore barriers and advancements to biomarkers for tumor recurrence and radiation to generate a plan to leverage data in this space to move targeted therapy to the forefront of standard-of-care management.

2. Intercepting progression - challenges in detecting early recurrence and developing targeted adjuvant therapy

Tumor recurrence is typically detected by several concurrent and tumor-specific mechanisms: clinical, imaging, and, for some malignancies – combining one or several laboratory parameters as an additive feature (Figure 1). Currently, the detection and capture of progression most often occur after SOC management. As a result, it is nearly impossible to intercept progression in real-time, and this would be needed to identify, validate, and optimize the type and timing of additional therapy that can delay or prevent treatment failure. Tumor recurrence as a parameter captured after RT is completed poses several challenges that persist in employing this time point to advance biomarkers: 1) normal and abnormal tissue changes are difficult to distinguish from progression on imaging; 2) recurrence is a tumor and patient-specific, leading to data heterogeneity; 3) clinical deterioration is not always an indication of disease progression but is most often the captured entity in electronic health records (EHR) in conjunction with imaging. Examples of a clinically observable response include head and neck cancers, cervix and bladder cancer, and other external or superficially observable malignancies where the response can be visualized or quantified during treatment or even immediately upon completion. The challenge in the early interception of recurrence or the ability to capture this during treatment is grounded in the need for more data on the tumor state while treatment is administered. Most progression criteria are not engaged to operate while treatment is ongoing unless overt clinical decline or extreme mass enlargement is noted. RT fields can be adapted during the treatment when tumor growth is observed clinically or radiographically on imaging acquired on the LINAC, such as, for example, cone beam CTs acquired during RT for image-guided treatment. This type of imaging is not currently employed as part of tumor progression criteria in guidelines. However, when overt progression is noted, radiation oncologists may carry out additional imaging, adapt treatment volumes and RT plans, or alter treatment altogether depending on observable progression or response.

Figure 1. Current approaches to assessing tumor recurrence in the clinic involve clinical aspects (history, physical exam, patient-reported outcomes/toxicity assessments, laboratory assessments, and imaging).

We should note that tumor recurrence parameters can rarely be linked exclusively to RT, given that a combination of RT and chemotherapy or other systemic therapy may be administered neoadjuvant or, more often, concurrently with increasing heterogeneity over the course of the disease (Figure 2). The role and intent of radiation shift over the natural history of the malignancy from curative to radical palliative to palliative and from the treatment of tissues that RT naïve to re-irradiation with the treatment of tissues that have either fully or partially been previously radiated (Figure 2). Scenarios where treatment is limited to only RT, compared to tumors where the standard of care is concurrent chemoradiation or radiation and another agent, would provide an avenue for RT-specific markers of failure. However, data that originates from patients where RT is the exclusive SOC management is not as common (e.g. Stage 1A lymphoma, early head and neck cancer, low or favorable intermediate prostate cancer) and is typically defined by early-stage disease or tumor biology that already renders the tumor highly responsive to RT administered alone with curative intent. As a result, the bulk of data that emerges from SOC will originate from malignancies where RT is augmented or affected by other treatments, e.g. androgen deprivation therapy and chemotherapy. Relevant biomarkers are TGF-β1 as linked to CNS, pancreatic, thoracic, gastric, and breast cancer, IL-1α and IL-6 associated with radiation-induced DNA damage in head and neck cancer, and prostate cancer, COX2 and CRP associated with inflammation and immune response in breast, prostate and lung cancer and HIF1α responsible for hypoxia signaling in most cancers though little data exists for patients only receiving RT [6–10]. Targeted treatments for PD-L1, COX 2, IL-6, and HiF1, implicated after RT in various cancers, have shown promising results.

Figure 2. Oncologic management modalities and their concurrent utilization over the course of malignancy illustrate the evolution of management components (surgery, radiation, systemic management) and existing biomarkers utilized in the clinic in relation to sample acquisition [4,5].

Across most solid tumors, imaging is the most reliable method for definitively detecting recurrence both while treatment is ongoing and after it is completed; however, due to the practical constraints (cost and accessibility of frequent imaging) and the problem of distinguishing tumor progression from tissue and anatomical changes that are secondary to treatment, a multimodal approach is most often utilized with imaging being carried out sometime after treatment is completed except for special circumstances. Radiological interpretation is challenged by surgical manipulation and RT and chemotherapy via inflammation, changes in tumor vasculature, and or infiltration of immune cells into the tumor. This can lead to imaging changes requiring more clinical features or alterations in the laboratory parameters to code as progression. Protocol for surveillance varies based on patient and cancer characteristics but often involves laboratory tests and imaging data. This makes it difficult for only the scan to be used as the source of data that designates progression. Though imaging data is often seen as superior for detecting early recurrence, machine learning (ML) models combining proteomic and imaging data, CT, MRI, and PET, to detect early recurrence are better. Combining quantitative and qualitative MRI features with pathologic data (i.e. VEGFR) effectively predicts early recurrence [11]. Other markers studied in imaging to detect early recurrence are CD147 in head and neck cancers, VEGF in liver and ovarian cancer, EGFR, HER2, and Ki67 in breast cancer, and TEM8 in colorectal cancer, all of which require more data for the development of reliable surveillance protocols [12–14]. Novel imaging techniques developed using biomarkers aggregated with imaging modalities transitioning to the clinic notably include PSMA-PET to detect early recurrence in prostate cancer [15]. Ongoing research focuses on using TEM8, VEGF, and EGF in combination with PET to detect early recurrence [11,14,16].

3. Existing and emerging biomarkers in the clinic

While the validation and clinical transition of biomarkers predictive of tumor recurrence would require a reliable correlation to tumor burden and the tumor state, data currently obtained during patient treatment or follow-up falls broadly into three categories: clinical with associated patient outcomes/toxicity, laboratory, and imaging (Figures 1 and 2) with existing biomarkers employed for tumor monitoring examples including PSA for prostate cancer, CA-125 for ovarian cancer, AFP for liver cancer, CEA for colorectal cancer, CA 19–9, CA 15–3, and CA 27–29 for breast cancer all of which are measured in serum levels and are sensitive but not specific markers of recurrence [17–23] (Table 1). In addition to the biomarkers measured in serum used to detect recurrence, additional features originating in tumor tissue analysis, such as HER2, IDH-1, and MGMT methylation status, are employed to guide treatment decisions based on their prognostic and, in some cases, predictive performance, and several are indicators of treatment response [23,26,43] (Table 1). Other evolving biomarkers of tumor recurrence, measured in serum, that require more research are CA 27–29 and VEGF [34,35]. Given the mechanism of action of RT with the differential repair of DNA damage between normal and abnormal cells, cell cycle redistribution, repopulation, and reoxygenation of hypoxic areas and ensuing radiosensitivity, biomarkers involved in DNA repair, hypoxia, alteration of the immune response, and tumor microenvironment can help link radiation response and radiation resistance [42]. Evolving biomarkers of radiation resistance are KRAS, TP53, IL-6, NRF2, HiF-1a, COX2, IGF-1 R, CRP, PD – L1, and EGFR, and some can, in certain circumstances, guide treatment selection, although this is, investigational [7,9,10,33–36,38]. Cancer stem cells (CSCs) are also responsible for mechanisms involving radiation resistance due to their ability to survive and repopulate after exposure to cytotoxic agents; thus, markers of stemness CD133, CD44, ALDH1, and several heat shock proteins (HSPs) involved in regulating cell proliferation, resistance, and evasion of immune destruction have been associated with radiation resistance as well [42,44,45]. In theory, proteomic panels comprised of molecular features that define resistance can allow for the selection of patients in whom the administration of adjuvant therapy can increase response to RT. Specifically for endometrial cancer, the integration of molecular characteristics (MSI/MMR status, TP53, and Polymerase epsilon (POLE)) allows for the development of prognostic groups, which makes it easy to identify patients with the worst prognosis that would likely benefit most from adjuvant therapy [46,47]. We should also note that few studies at this time attempt to implement a focus on personalizing RT based on tumor-specific proteomic and molecular features, especially features of recurrence. Previous research has found germline biomarkers associated with radiation treatment response, specifically radiation-related toxicity, in patients with head and neck cancers [48]. Studies have shown that detecting circulating tumor DNA (ctDNA) and microRNAs (miRNAs) can be reliable indicators of tumor recurrence, though specificity and sensitivity vary based on the tumor. Mismatch repair (MMR) and microsatellite instability (MSI) genes are useful biomarkers in DNA repair mechanisms. MMR/MSI status is used as a diagnostic tool for Lynch syndrome (higher risk of colorectal cancer), as a prognostic indicator for colorectal cancers, and for screening and surveillance for colorectal and endometrial cancer [49]. Mechanisms of radiation resistance are highly variable and dependent on disease type. Previous research published by Wu et al. in June 2023 discusses our current knowledge of molecular mechanisms related to tumor resistance, such as DNA damage repair, cell cycle arrest, apoptosis escape, modification of cancer cells by tumor microenvironment, presence of exosomal, non-coding RNA, and metabolic reprogramming [50]. The most studied pathways related to radiation resistance are p53, KRAS, and PI3K/Akt/mTOR. P53 stimulates apoptosis in response to RT and is linked to radiation sensitivity. In contrast, mutant forms of p53 have been linked to RT resistance, though this can be inconsistent in a few disease types [51]. Therapies targeting p53, such as Gendicine, are effective for breast, pancreatic, cervical, and ovarian cancers [51]. Furthermore, several transcription factors are linked to radiation resistance, like the immune modulators activated by RT, STAT3, and NFkB, which upregulate anti-apoptotic genes and promote cell proliferation. Also linked to radiation resistance are NF-kB, NRF2, and HIF-1, which maintain stemness, increase Epithelial-to-mesenchymal transition (EMT), and protect from free oxygen radicals necessary for RT-induced cell kill [52]. Il-6 induces glycolysis, which protects from radiation-induced mitochondrial damage by activating PI3K/Akt/mTORpathway [53]. RT activates KRAS signaling, which elevates CD44 expression and (EMT) and stemness [40]. The identification and effective utilization of these biomarkers in clinics and the robust interest in existing and evolving biomarkers (Figure 3) illustrate that proteins can act as biomarkers in cancer, and given sufficient data and appropriate labeling, biomarker panels can be identified that more deeply connect to biological tumor response and clinical data. More established biomarkers of tumor recurrence are measured in serum compared to tissue samples due to the accessibility of sample acquisition (Figure 2). While the standard-of-care laboratory tests alone are seldom sufficient to conclusively detect recurrence or the extent of the recurrence due to their lack of specificity, they are a cost-effective example of data acquisition on a large scale and given linkage to emerging and existing biomarkers early indication of recurrence may be identifiable mainly as aggregated with imaging data that can generate features of progression. Though diagnostic criteria using these biomarkers can be problematic due to the high rate of heterogeneity within populations, tumor-specific molecular features can be used in conjunction with clinical data to classify tumor recurrence and make treatment decisions, and work in this space is actively evolving.

Figure 3. Existing and evolving biomarkers and their literature footprint. The biomarker literature footprint was determined based on a web of science literature search. Terms used in the search are ‘tumor recurrence and ‘radiation’ and the biomarker name [54].

Table 1. Existing and evolving biomarkers of tumor recurrence, radiation resistance, and stemness.

	Biomarker	Full Name	Details	Journal
Established biomarkers employed for clinical decision making	HER2	Human Epidermal Growth Factor Receptor 2	breast cancer	Zubair et al. 2021 [24]
	IDH-1	Isocitrate Dehydrogenase 1	gliomas	Stewart et al. 2022 [25]
	MGMT methylation	O6-Methylguanine-DNA Methyltransferase	gliomas	Wang et al. 2015 [26]
Established biomarkers used to detect recurrence	PSA	Prostate-Specific Antigen	prostate cancer	Wirth et al. 2005 [27]
	CA-125	Cancer Antigen-125	ovarian cancer	Giuliani et al. 2018 [28]
	AFP	Alpha-Fetoprotein	liver cancer	Xu et al. 2022 [29]
	CEA	Carcinoembryonic Antigen	Colorectal, lung, and Endometrial cancer	Lee et al. 2015 [20], Kozakiewicz et al. 2016 [30], Shintani et al. 2017 [31]
	CA 19–9	Cancer Antigen 19–9	Pancreatic cancer	Micke et al. 2003 [32]
	CA 15–3	Cancer Antigen 15–3	breast cancer	Zubair et al. 2021 [24]
	CA 27–29	Cancer Antigen 27–29	breast cancer	Zubair et al. 2021 [24]
Evolving biomarkers of radiation resistance	KRAS	Kirsten rat sarcoma viral oncogene homolog	Involved in uncontrolled cell proliferation	Toulany et al. 2023 [33]
	VEGF	Vascular Endothelial Growth Factor	Involved in angiogenesis	Mortexaee et al. 2019 [34]
	TP53	Tumor Protein 53	Tumor suppressor gene	Ho et al. 2022 [35]
	IL-6	Interleukin-6	Involved in immune and response	Tsai et al. 2019 [7]
	NRF2	Nuclear factor erythroid 2-related factor 2	Transcription factors involved in oxidative stress response	Jeong et al. 2017 [36]
	HiF-1a	Hypoxia-Inducible Factor 1α	Transcription factors involved in hypoxia response	Brown et al. 2020 [37]
	COX2	Cyclooxygenase-2	Enzymes involved in inflammation	Laube et al. 2016 [8]
	IGF-1R	Insulin-like Growth Factor 1 Receptor	Cell surface receptor; cell growth and survival	Breen et al. 2017 [38]
	CRP	C-reactive Protein	Acute phase reactant protein response to inflammation	Katano et al. 2017 [39]
	PD-L1	Programmed Death Ligand	Inhibiting immune response	Tsai et al. 2019 [7]
	TGF- β	Transforming growth factor-beta	Cellular proliferation	Wang et al. 2021 [6]
	EGFR	Epidermal growth factor receptor	Uncontrolled proliferation	Mortezaee et al. 2019 [34]
Markers of Cancer Stem Cells	CD44	Cell surface glycoprotein	Cell surface glycoprotein; tumor metastasis	Zhao et al. 2021 [40]
	CD133	Prominin-1	Cell surface marker; cell renewal and differentiation	Kamamoto et al. 2019 [41]
	ALDH1	Aldehyde Dehydrogenase 1	Enzymes involved in cellular detoxification	Arnold et al. 2020 [42]

4. Advances in planning targeted adjuvant therapies involving RT

We discussed biomarkers employed in the clinic to guide clinical decision-making and monitor treatment. These biomarkers represent a guiding template for functionality at the bedside that can be enhanced with the omic data explosion both by aggregation with exiting SOC laboratory data and imaging data and by inclusion in prospective trials. The biomarkers PSA, CA-125, AFP, CEA, CA 19–9, and CA 15–3 are measured in serum and employed for monitoring tumor burden. Meanwhile, parameters such as HER2, measured in tissue, are employed to select management to improve outcomes. Radiation resistance-related biomarkers based on VEGF, EGFR, IL-6, NRF2, HiF-1a, COX2, and IGF-1 R are not well established to fulfill such roles. It is conceivable that targeted adjuvant therapy after RT failure may be linked directly or indirectly to markers of the immune response, including CD133, CD44, ALDH1, and CRP. Several clinical studies linking natural agents such as Quercetin, Apigenin, and Epigallocatechin-3-gallate (EGCG) as potential targets of markers of the immune response are ongoing and may eventually prove useful in the clinic [55]. Trifluridine (FTD) and Tipracil (TPI) are effective in targeting cells expressing CD44 and CD133 [56,57]. Triptolide (TPL) has also been linked to effective treatment for cells expressing CD 133 in colon cancer [58]. Targeting CD133 with antibody-conjugated SN-38 loaded nanoparticles effectively reduced growth and recurrence in CD 133 positive colon cancer [59]. Targeting PD-L1, which is upregulated by RT, with monoclonal antibodies effectively reduces tumor recurrence by targeting the mechanism of radiation resistance [60]. These therapeutic agents are effective because of their ability to alter the tumor microenvironment and target the immunomodulatory effects of RT, often increasing radiation sensitivity. The development of targeted drug therapies involving these markers shows promising results, though still in the new stages. With more data that can enable a realistic correlation of RT dose to specific tissues and targets with omic data, biomarkers of radiation resistance may become more realistic. Identifying such biomarkers can lead to druggable targets and agents being delivered in conjunction with RT and chemotherapy.

The current focus in developing adjuvant therapies involving RT focuses on optimizing early detection of recurrence, increasing radiation sensitivity often by targeting one of the known molecular features contributing to radiation resistance, and using machine learning to develop prognostic models that can help guide and select personalized radiation therapy. For example, classifying patients of endometrial cancer based on their molecularly defined subgroup can help clinicians pair certain targeted therapies or more aggressive treatments with patients who are more likely to resist radiation and experience recurrence [47]. Current ML applications are focused on developing accurate prognostic models using radiomic, proteomic, and imaging features. Due to the challenges in detecting recurrence and the prevalence of relatively small datasets, validating ML models for treatment selection is difficult. Progress is ongoing, as evidenced by a recent ML model employed in patients with prostate cancer, wherein using imaging features for RT treatment planning, 89% of clinically acceptable plans were obtained [61]. Automated treatment planning for RT can leverage deep learning algorithms for beam orientation selection, dose map prediction, fluence map generation, and delivery parameter generation [62]. One such example of machine learning (ML) for adjuvant therapy selection in patients with head and neck cancer was associated with a survival benefit, particularly for patients with an intermediate risk of recurrence [63]. Established ML models are emerging for more common cancers (e.g. breast, prostate) due to the availability of more extensive data in these settings. More complex models that aggregate clinical, imaging, omic, and radiation dosimetry data toward personalized treatment planning require ongoing research.

The detection of new biomarkers of radiation resistance and tumor recurrence (i.e. PD-L1 and Cav-1) have been linked to multimodal therapies, including radiation-sensitizing agents for lung and pancreatic cancer [64,65]. Treatments targeting these biomarkers of radiation resistance have shown promising results, indicating that acquired resistance to RT can be inhibited [9]. Drugs that increase radiation sensitivity when used in conjunction with chemoradiation have also been shown to be effective (e.g. Valproic Acid in conjunction with chemoradiation in GBM); however, further validation and prospective studies are needed before implementation can become standard of care in the clinic [66–68]. Statistical models that increasingly explore molecular features can successfully predict recurrence patterns and generate prognostic models, indicating that with further enhancement, they may be able to guide the optimal selection of adjuvant therapy [69].

5. Conclusions

Given the multifactorial challenges described in the early identification of tumor failure and the persistence of imperfect progression parameters currently captured, the development of predictive markers of RT resistance that are clinically relevant and realistic will take time and far more data linkage than is currently implemented. Radiation therapy is rarely employed in isolation, and thus, biomarkers that define its effects are going to be the product of singling out pathways that are triggered when patients are radiated in distinction to pathways that are also affected by other interventions (chemotherapy, androgen deprivation therapy) and this may only be possible through in-depth omic analysis that is linked to clinical data and deeply linked to radiation dose to tissue. Though several challenges exist in detecting recurrence, a multimodal approach using a combination of data for laboratory tests and imaging is most efficient in detecting tumor recurrence early, and data in this space is growing. However, once actionable biomarkers are identified, for molecular features to be employed in the clinic, the sample wherein they are detected must be easy to access and measure as well as cost-effective, as exemplified by markers currently employed in the clinic. Future research in developing targeted adjuvant therapies should focus on optimizing the early detection of tumor recurrence and linking it biologically to adjuvant management, anticipating that personalized management that is biologically optimized will ideally be implemented immediately upon diagnosis and remain adaptable throughout the natural history of the disease.

6. Expert opinion

In the context of radiation therapy (RT), technological advances and the integration of imaging data have allowed for added precision with improvement in toxicity and cancer outcomes. Ongoing research in identifying therapeutic targets, exploring serum and tissue biomarkers, and developing prognostic models to select adjuvant targeted therapy have yet to reach the clinic or become embedded in treatment planning. This paper delves into why this is the case and examines how biomarker integration in the clinic may look like using existing advancements and how they transform patient care. Although large repositories of molecular and imaging feature data are emerging, RT-specific data parameters, particularly the radiation dose to structures and tumors in conjunction with imaging, need to be improved due to siloed data and limited data sharing. Data collection, standardization, and integration challenges persist due to a lack of real-time data, sample heterogeneity, and sampling bias, hindering progress. While some efforts are underway to harmonize data collection practices, standardized progression criteria remain firmly stuck in the past, with progression criteria generally applied downstream of the standard of care management and yet to be linked to robust data-sharing platforms that could unleash the potential of vast data repositories. Such data could be computationally optimized to capture when progression or, better yet, tumor resistance to treatment is identified. The lack of publicly available data raises the concern that it will take considerable time for clinical and computational investment to develop large, reliable repositories of progression data that can produce actionable clinical results. Identifying patients suitable for targeted adjuvant therapy requires understanding the tumor state defined by biological parameters, recognizing early progression, or, better yet, the upfront determination of aggressive tumor behavior.

In selecting adjuvant therapy after radiation failure, there are two areas of focus: detecting early progression and the rational selection of targeted agents. As noted, progression is currently captured at a time point beyond completion of the RT course, thus offering no prospect for its alteration. Such alteration could take several forms: 1) altering dose, 2) altering fractionation, 3) altering treatment volumes, and 4) the addition of a radiation sensitizer. Some or all of these strategies are currently employed in the clinic for some malignancies. However, on a case-by-case basis, this is more difficult to carry out in high workload, high volume clinical environments, and the capture of data involving such adaptation to tumor behavior could be better. Rational selection of targeted agents requires identifying biomarkers that can be used to survey for recurrence or indicate radiation resistance. No such markers are generally employed during standard-of-care treatment delivery in part because available features are confounded in their interpretation by the treatment itself and because the argument is made that, most often, treatment would not be altered if the marker were to indicate disease progression as no superior options might be available or recommended based on guidelines. PSA, CA-125, AFP, CEA, CA 19–9, CA 15–3, and HER2 biomarkers are widely used in the clinic. These biomarkers are partly used to detect recurrence or develop predictive models that aid in treatment decision-making, enabling clinicians to optimize therapy for individual patients. However, promising treatments that target markers of radiation resistance, such as PD-L1, COX2, IL-6, and HiF1, are the subject of ongoing research.

Data must be acquired in real-time to capture the tumor state to be translated into tumor resistance markers and the progression measurement. Doing so requires seamless accessibility and cost-effective large-scale data output in clinical settings. Blood and urine are more accessible and cost-effective than tumor tissue or CSF. However, data from blood or urine is nonspecific, requiring more research to arrive at connections to the tumor state and generate clinically actionable biomarkers. Tissue samples often expose more information about tumor microenvironment and genetic alterations but can be unreliable due to tumor heterogeneity, sampling error, and single time point acquisition, with tissue scarcity additionally impeding progress. Ideally, more research will rely on acquiring biospecimens such as blood and urine, with more resources directed toward the reliable interpretation of omic data from multiple time points. The fact that few serum biomarkers made it into the clinic is not a testament to the inability to employ serum for this purpose but rather an excellent example of serum proteomic biomarker success, the byproduct of a time when the level of data available was a far cry from the unparallel omic data explosion we are privy to today. Harnessing omic data from accessible biospecimens will allow for a better understanding of the tumor state, contributing to adaptive RT treatment planning. Meanwhile, imaging is and will remain a pillar of data growth even as the analysis of imaging data acquired during RT courses needs to evolve to connect to accessible biomarkers such as blood and urine.

Comprehensive integration of imaging and biospecimen biomarkers is critical to detecting early progression and improving treatment outcomes. Tailoring adjuvant therapy to individual patients by incorporating patient-specific factors, such as genetic profiles, tumor characteristics, and biomarker expression, can provide clinicians the tool sets to rationally optimize RT features such as radiation doses, fractionation schedules, treatment techniques, and drug therapy, improving treatment efficacy while minimizing damage to healthy tissues. Notably, many studies focus on prognostic molecular features, which, while useful for decision-making, may not link mechanistically with the tumor state, limiting biological understanding. It is critical to focus on unraveling the molecular and genetic mechanisms that drive tumor growth and response to RT to refine existing treatment paradigms and identify novel targets for therapeutic intervention.

Traditionally, RT has relied on standardized treatment guidelines, offering a one-size-fits-all approach to patient care. Increased availability of molecular and genomic features should make it possible to transcend this paradigm to develop treatments tailored to patient characteristics. While guidelines will continue to provide a foundation for adjuvant therapy, tailoring treatment to the patient and tumor-specific features will improve outcomes, but the achievement of this new optimized level of care will require accessible, cost-effective biospecimen acquisition and analysis and the consistent enforcement of shared omic data originating in clinical trials and standard of care.

Acronyms

CNS	=	Central Nervous System
CSC	=	Cancer Stem Cells
CT	=	Computed Tomography
EHR	=	Electronic health records
GBM	=	Glioblastoma multiforme
LINAC	=	Linear accelerator
ML	=	Machine learning
MMR	=	Mismatch repair
MRI	=	Magnetic resonance imaging
MSI	=	Microsatellite instability
NCCN	=	National Comprehensive Cancer Network
PET	=	Positron emission tomography
PSMA-PET	=	Prostate-specific membrane antigen positron emission tomography
RT	=	Radiation therapy
SOC	=	Standard of care

Article highlights

• Optimization of tumor detection and progression due to treatment resistance is critical to identifying patients suitable for targeted adjuvant therapy.

• Imaging is the most reliable method for definitively detecting recurrence both while treatment is ongoing and after it is completed; however, there are practical constraints (cost and accessibility of frequent imaging) and the problem of distinguishing tumor progression from tissue and anatomical changes caused by treatment.

• Biomarkers analyzed in this review broadly fall under 4 categories: 1) established biomarkers employed for clinical decision-making; 2) established biomarkers used to detect recurrence; 3) evolving biomarkers of radiation resistance and 4) biomarkers of cancer stem cells.

• The current focus in developing adjuvant therapies involving RT focuses on optimizing early detection of recurrence, increasing radiation sensitivity often by targeting one of the known molecular features contributing to radiation resistance, and using machine learning to develop prognostic models that can help guide and select personalized radiation therapy.

• It is critical to focus on unraveling the molecular and genetic mechanisms that drive tumor growth and response to RT to refine and tailor existing treatment paradigms and identify novel targets for therapeutic intervention

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or material discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or mending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was funded by the National Cancer Institute, U.S. Department of Health and Human Services, National Institutes of Health Funding [ZID BC 010990].