Beyond Circulating Tumor DNA for Efficacy: Can We Use Cell-Free DNA to Detect and Monitor Toxicity Signals? Authors: Yiqing Zhao, John Gallon, Max H Kushner, Brad Jakubison, Jorge S Reis-Filho, Darren Hodgson, Julia Elvin, Yuval Dor, James Hadfield Published: 2026 Introduction Therapy-induced toxicity presents significant challenges to patients, physicians, and drug developers. Chemotherapy, radiotherapy, targeted therapies, and immunotherapies can cause adverse events necessitating treatment discontinuation and potentially causing life-threatening serious adverse events (SAEs; generally referred to as toxicities in this study1). Clinicians may avoid otherwise effective regimens when the perceived toxicity risk outweighs the risk of progression. For drug developers, therapy-induced toxicity can result in clinical failure, regulatory rejection, or market withdrawal, incurring financial and logistical costs. Current toxicity detection methods include medical imaging, tissue biopsy, and blood tests, which, though proven to be effective in various settings, all have limitations. Tissue biopsy is risky and invasive, whereas blood tests (eg, troponin or liver enzymes) and imaging can lack sensitivity and specificity.2,3 Moreover, testing requires prior knowledge of toxicity-affected organs, information that is unavailable during early-stage clinical trials or for novel toxicities or therapy combinations triggering multiple toxicities.4 A minimally invasive assay capable of earlier detection and monitoring of therapy-induced toxicity across multiple tissues remains an unmet need. Liquid biopsy is a promising tool for this purpose. Advances in circulating tumor DNA (ctDNA) have led to diagnostics, response monitoring, and minimal residual disease detection uses and revealed ctDNA's capability in multicancer early detection (MCED).5,6 Circulating cell-free DNA (cfDNA) from noncancerous cells, currently used for noninvasive prenatal testing and detection of transplant rejection,7,8 recently proved to be effective for monitoring tissue-specific damage via cell type–specific epigenetic markers,9,10 making cfDNA a potent candidate for earlier detection of therapy-induced toxicity. Existing toxicity monitoring methods do not detect life-limiting SAEs early enough and require toxicity-specific tests alongside efficacy monitoring. As ctDNA for efficacy becomes widely adopted, exploring circulating biomarkers for toxicity is logical. This commentary provides an early overview of the use of cfDNA for therapy-induced toxicity detection and highlights the opportunities and challenges of this new paradigm. Epigenetic Liquid Biopsy for Toxicity—Approaches and Opportunities Elevated tissue-specific cfDNA indicates cell death or increased turnover and may signal therapy-induced toxicity.11,12 Recent studies revealed that cfDNA epigenetic profiles contain tissue-of-origin (TOO) information that enables quantitative analysis of tissue- or cell type–specific cfDNA in the plasma.9,10,13 CpG methylation has proven to be superior to fragmentomic patterns and histone modifications for identifying TOO, because of higher cell and tissue type specificity, empowering accurate TOO assignment of cfDNA fragments9 (Fig 1). An overview of the application of cfDNA for monitoring tissue health and disease. Cancer cells release ctDNA, allowing for cancer monitoring through SNVs, CNVs, and methylation patterns. Noncancerous cells release circulating cfDNA on damage, which carries cell type–specific epigenetic markers that can be analyzed for tissue-of-origin information. A significant increase in the level of cell type–specific cfDNA can be indicative of tissue toxicity. This approach allows for earlier detection of therapy-induced toxicity and cfDNA-guided therapeutic regimen. This affects preclinical work, clinical trials, and postapproval stages to improve treatment efficacy and safety and expand patient populations. Image created using Biorender.com. cfDNA, cell-free DNA; ctDNA, circulating tumor DNA; CNV, copy number variants; SNVs, single-nucleotide variants. Recent cell type methylome atlases revealed thousands of differentially methylated regions, 97% of which are unmethylated and typically overlap cell type–specific enhancers, as biomarkers that could be used to quantify cell-specific cfDNA.9,10 These developmentally linked loci appear to be universal across individuals regardless of age, sex, or ethnicity and are evolutionarily conserved.10 Detection approaches for cfDNA methylation range from polymerase chain reaction amplification of cell type–specific markers for single toxicities (eg, FAM101A for cardiotoxicity)13,14 to genome-wide analyses (eg, sequencing of whole-genome methylation or the unmethylated fraction alone), enabling deconvolution to apportion cfDNA by cell type for multitissue toxicity monitoring.15-17 cfDNA methylation analysis has emerged as one of the leading approaches for MCED.5,18 While the recent NHS-Galleri results underscore limitations in current implementations and trial design, the broader literature continues to indicate that cfDNA methylation remains one of the strongest signals for early detection of pathology. This approach also identifies noncancerous tissue damage including pulmonary, cardiac, or hepatic injury; graft-versus-host disease; viral infection, and brain, vascular, renal, and intestinal damage.10,13,14 These studies highlight the potential of leveraging highly sensitive and specific methylation signatures for the earlier detection of toxicity where traditional diagnostic measures are unsatisfactory. The same approach can also be applied for toxicity biomarker discovery, multitissue toxicity screening (eg, for immunotherapies),17 and mechanistic analyses where causes are elusive (eg, combination therapies). Impact on Patients and Drug Development cfDNA epigenetic analyses can potentially enable earlier detection and improved monitoring of toxicity, even for novel treatment combinations or early-stage trials with unknown toxicity. It could allow more accurate safety profiling, improved cost efficiency, and opportunities for personalized toxicity mitigation. Its versatility in detecting any tissue toxicity caused by any treatment could shift paradigms in therapy-induced toxicity diagnosis and management, substantially benefiting patients and drug development. One key advantage of cfDNA-based methods is their potential for earlier adverse event detection in clinical trials. Unlike standard methods, which typically detect toxicity after clinical manifestation, cfDNA may enable real-time monitoring of subclinical tissue injury.10 This is critical, for example, in the case of antibody-drug conjugate (ADC)–induced interstitial lung disease (ILD), observed in up to 12% of treated patients. ILD is often diagnosed by computed tomography scans after pneumonitis symptoms, when patients might already have grade 2 or 3 ILD, necessitating treatment cessation. Accelerated lead times offered by cfDNA may allow diagnosis of asymptomatic ILD/pneumonitis before progression to higher grades.19 Proactive liquid biopsy monitoring could enable tailored treatment modulation, empowering clinicians to use near-real-time toxicity readouts for high-risk monitoring and initiating early interventions. This would allow patients to remain on life-saving medications for longer, enhancing therapeutic outcomes while mitigating the risk of SAEs. Beyond direct patient benefits, enhanced toxicity detection could deliver more cost-efficient clinical trials with higher success rates. With novel drug modalities and complex combinations where toxicity risks cannot be fully anticipated, cfDNA-based multitissue monitoring, empowered by cell methylation atlases, could prevent the spotlight effect in early-phase trials assessing only expected toxicities or those seen in preclinical models. A comprehensive, hypothesis-free early toxicity assay would ensure detection of any safety signal in the earliest phase trials to inform mitigations in later phase studies. This minimizes costly late-stage failures and candidate attrition while reducing reliance on expensive, time-consuming, and invasive methods. Early drug candidate selection could also be derisked with cfDNA-based preclinical monitoring, accelerating the development and translation of novel therapies that might otherwise be abandoned because of known but previously unmanageable, or new but unforeseen, toxicity. Epigenetic liquid biopsies may offer unique mechanistic insights into adverse events beyond detection. For example, increased hepatocyte cfDNA associated with liver injury might be accompanied by elevated CD8+ T-cell cfDNA, suggesting an immune-related mechanism, or by increased fibroblast cfDNA, indicating fibrotic pathogenic processes damaging the liver. Such rich mechanistic information is crucial for developing robust toxicity risk classifiers, refining precision toxicity monitoring and enabling targeted mitigation or prevention strategies. The positive implications for patients are profound and multifaceted. Detecting toxicities before they become irreversible would dramatically reduce SAE risk, improving patient safety and quality of life. Clinicians may be more willing to prescribe beneficial therapies previously associated with high toxicity risk if early toxicity detection enables mitigation before higher-grade adverse events develop. In addition, for symptomatic patients, cfDNA analysis could potentially differentiate drug-related damage from other unrelated causes of symptoms, allowing some patients to remain on effective treatment for longer and facilitating the safer reintroduction of life-saving treatment in patients for whom toxicity risk has been managed. This paradigm shift could alter the risk-benefit equation for drugs previously reserved for advanced or aggressive cancers because of toxicity concerns, making them options for patients with earlier or less aggressive disease. Reducing high-grade toxicity risk could expand the therapeutic window, broadening eligible populations for a given therapy and allowing more patients to benefit from potentially life-saving treatments. Discussion cfDNA epigenome liquid biopsy presents a promising avenue for early multitissue toxicity detection and may evolve into an all-in-one tool for monitoring efficacy and toxicity. Empowered by highly stable cell type–specific epigenetic markers and advancing next-generation sequencing technology, this sensitive and specific diagnostic approach could significantly improve patient experience and drug development. However, our understanding of this approach is nascent and challenges remain that require concerted efforts by the medical and scientific community. While sample collection is easily standardized using Streck or ethylenediaminetetraacetic acid tubes, made familiar by wide-spread ctDNA analysis, remaining challenges include robust method standardization, rigorous clinical validation across diverse patient cohorts and drug classes, and establishment of clear regulatory pathways for clinical integration. As this technology is still early in its development, cfDNA-based drug safety monitoring will most likely start with simpler pathology and therapy use cases before expanding into more complex applications. cfDNA results are more straightforward to interpret in an off-target toxicity setting, that is, when the tumor TOO and the toxicity-affected tissue are different, for example, ADC-associated lung ILD in a patient with breast cancer. In comparison, detection of on-target toxicity, that is, from an organ simultaneously affected by malignancy (such as non–small cell lung cancer therapy–associated ILD), is more challenging. Because cancer cells shed cfDNA with the same TOO as cells experiencing toxicity, deconvoluting these signals will likely require emerging workflows that combine epigenetic and genetic analyses. Such analyses could potentially track both efficacy, by decreasing levels of mutant (eg, EGFRm) ctDNA, and toxicity by increasing levels of cell type–specific (eg, lung-specific) cfDNA, thus resolving cfDNA origins for unconfounded toxicity readouts. Apart from TOO of toxicity and tumor, additional complexities such as the medical history of patients, effects of concurrent treatments such as surgery and radiotherapy, comorbidities, and non–therapy-related inflammations or tissue injury must also be factored into the baselines against which post-treatment cfDNA changes are assessed as indicators of toxicity. Such incremental refinement of analysis strategies relies on an ever-growing understanding of cfDNA biology building on a decade of ctDNA studies. When accompanied by clinical assessments and established diagnostic markers, the full potential of cfDNA could be released to enable multitissue toxicity monitoring and provide biomarkers for tissues where these are currently lacking. It is important to assess how much lead time this approach can achieve in the diagnosis of toxicity. This will guide clinical practice (eg, blood collection frequency) and identify cases where lead time improvements do not justify the additional costs given the availability of current tests. For this purpose, the abundance, specificity, and sensitivity of detection of cfDNA for toxicity need extensive characterization, and feasibility of toxicity risk prediction using baseline cfDNA levels and epigenetic profiles should be confirmed. The fundamental assumption of this method is that cells release cfDNA into the blood when damaged. In reality, different tissues exhibit different extents of cfDNA shedding when incurring cell damage, with some always shedding to plasma (eg, liver, hematopoietic cells) and some never shedding (eg, skeletal muscle); some tissues shed only upon certain injuries (eg, intestinal epithelium, lung epithelium; unpublished information). False positives could also arise when elevated cfDNA is not related to tissue damage but is instead caused by other events such as local macrophage perturbation.20 To ensure robust toxicity detection by a post-treatment vs baseline model and to potentially expand to analyses that do not require a cfDNA baseline reference, the tissue-specific rules governing cfDNA turnover must be elucidated to properly interpret tissue-specific cfDNA levels and establish meaningful toxicity threshold. Complementary analysis of different circulating analytes, similar to integrative liquid biopsy analysis for cancer,6 may provide insights into distinct aspects of tissue damage. For example, while cfDNA is considered an indicator of cell death caused by acute tissue damage or ongoing turnover (because of its short half-life of 15-120 minutes), the presence of cytoplasmic markers in blood (eg, liver enzymes) may reveal information on either cell death or reversible cell injury in a more integrative and less time-restricted manner. Finally, regarding clinical implementation, we envision initial deployment in centralized, specialized laboratories to ensure rigorous standardization, quality control, and external proficiency testing, with subsequent transition to accredited institutional laboratories, as clinical evidence thresholds are met and validated, and distributable kits or the decentralized platforms required to deliver these tests have matured. In conclusion, the powerful toolbox behind this novel liquid biopsy strategy is only beginning to flourish. With advancement in next-generation sequencing technology, the paradigm shift to liquid biopsy–based drug safety monitoring appears to be imminent. The unique potential of cfDNA as a unified platform for monitoring therapy-induced toxicity across multiple tissues, from preclinical to clinical trials, through to on-market safety monitoring, stands to revolutionize drug development. The versatility of this assay allows for integration with other diagnostic procedures such as those for cancer and chronic diseases, thus transforming patient care. Yiqing Zhao Employment: AstraZeneca Stock and Other Ownership Interests: AstraZeneca Research Funding: AstraZeneca John Gallon Employment: AstraZeneca, Tagomics Ltd, AstraZeneca (I) Stock and Other Ownership Interests: AstraZeneca, Tagomics Ltd, AstraZeneca (I) Patents, Royalties, Other Intellectual Property: Patents for methods for identifying DNA from a specific tissue or cell type in a cell-free DNA (cfDNA) sample Max H. Kushner Employment: AstraZeneca Stock and Other Ownership Interests: Legend Biotech Travel, Accommodations, Expenses: AstraZeneca Brad Jakubison Employment: AstraZeneca Travel, Accommodations, Expenses: AstraZeneca Jorge S. Reis-Filho Employment: AstraZeneca Leadership: Grupo Oncoclinicas Stock and Other Ownership Interests: Grupo Oncoclinicas, Repare Therapeutics Consulting or Advisory Role: Repare Therapeutics, PAIGE.AI, Bain Capital, SAGA Diagnostics, MultiplexDx Darren Hodgson Employment: AstraZeneca Stock and Other Ownership Interests: AstraZeneca Julia Elvin Employment: Foundation Medicine, AstraZeneca Stock and Other Ownership Interests: Roche Yuval Dor Consulting or Advisory Role: Guardant Health Research Funding: AbbVie (Inst), Tagomics (Inst) Patents, Royalties, Other Intellectual Property: YD has filed patents related to cfDNA methylation technology (Inst) James Hadfield Employment: AstraZeneca Stock and Other Ownership Interests: AstraZeneca Consulting or Advisory Role: Tagomics Travel, Accommodations, Expenses: AstraZeneca No other potential conflicts of interest were reported. SUPPORT Supported by AstraZeneca (JH); Israel Science Foundation, Israel Innovation Authority, and Horizon Europe (YD). Y.Z. and J.G. contributed equally to this work. AUTHOR CONTRIBUTIONS Conception and design: All authors Financial support: Julia Elvin Collection and assembly of data: Yiqing Zhao, John Gallon Data analysis and interpretation: Yiqing Zhao, John Gallon, Brad Jakubison, Yuval Dor, James Hadfield Manuscript writing: All authors Final approval of manuscript: All authors Accountable for all aspects of the work: All authors AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center. Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments). Yiqing Zhao Employment: AstraZeneca Stock and Other Ownership Interests: AstraZeneca Research Funding: AstraZeneca John Gallon Employment: AstraZeneca, Tagomics Ltd, AstraZeneca (I) Stock and Other Ownership Interests: AstraZeneca, Tagomics Ltd, AstraZeneca (I) Patents, Royalties, Other Intellectual Property: Patents for methods for identifying DNA from a specific tissue or cell type in a cell-free DNA (cfDNA) sample Max H. Kushner Employment: AstraZeneca Stock and Other Ownership Interests: Legend Biotech Travel, Accommodations, Expenses: AstraZeneca Brad Jakubison Employment: AstraZeneca Travel, Accommodations, Expenses: AstraZeneca Jorge S. Reis-Filho Employment: AstraZeneca Leadership: Grupo Oncoclinicas Stock and Other Ownership Interests: Grupo Oncoclinicas, Repare Therapeutics Consulting or Advisory Role: Repare Therapeutics, PAIGE.AI, Bain Capital, SAGA Diagnostics, MultiplexDx Darren Hodgson Employment: AstraZeneca Stock and Other Ownership Interests: AstraZeneca Julia Elvin Employment: Foundation Medicine, AstraZeneca Stock and Other Ownership Interests: Roche Yuval Dor Consulting or Advisory Role: Guardant Health Research Funding: AbbVie (Inst), Tagomics (Inst) Patents, Royalties, Other Intellectual Property: YD has filed patents related to cfDNA methylation technology (Inst) James Hadfield Employment: AstraZeneca Stock and Other Ownership Interests: AstraZeneca Consulting or Advisory Role: Tagomics Travel, Accommodations, Expenses: AstraZeneca No other potential conflicts of interest were reported. REFERENCES