The review arrives at a moment of growing frustration in breast cancer management. Although surgery, chemotherapy, radiotherapy, endocrine therapy, and targeted agents have dramatically improved outcomes for many patients, a substantial proportion still experience recurrence and progression. One reason is that breast cancer is not a single disease. Its molecular subtypes—luminal A, luminal B, HER2-positive, and triple-negative breast cancer (TNBC)—each carry distinct biological behaviors, respond differently to the same drugs, and recur at different rates. Clinicians have long sought biomarkers that can be measured after diagnosis to guide treatment decisions in real time, and metabolites are emerging as unusually informative candidates.

The logic behind metabolomics is rooted in a fundamental feature of cancer biology. Tumor cells rewire their metabolic machinery to sustain energy production, maintain redox balance, and fuel relentless biosynthesis even under the hostile conditions of hypoxia and nutrient scarcity that characterize the tumor microenvironment. Because metabolites sit at the very end of the chain linking genes to proteins to cellular function, they offer a dynamic and sensitive readout of what a tumor is actually doing—often a more faithful snapshot of phenotype than genomic or proteomic data alone. Metabolites also participate directly in signaling, immune evasion, and epigenetic modification, meaning they are not merely passive byproducts but active participants in malignant progression.

To build their evidence map, the researchers conducted a systematic PubMed search covering studies published between 2006 and 2025, screening 445 initial hits down to 53 clinical studies involving human biological samples. Of these, 36 addressed metabolomics for monitoring therapeutic response, 9 focused on prognostic markers, and 8 examined signatures of disease progression. The studies drew on a variety of biological materials—serum most commonly, followed by plasma, tumor tissue, urine, and feces—and employed a range of analytical platforms. Liquid chromatography-mass spectrometry (LC-MS) dominated the field, with nuclear magnetic resonance (NMR) spectroscopy and gas chromatography-mass spectrometry (GC-MS) as important alternatives. Most studies (41) used untargeted approaches that survey the metabolome broadly, while 7 used targeted methods and 5 combined both strategies.

One of the review’s most striking findings is how rapidly cancer treatments themselves reshape the metabolic landscape. Within the first 24 hours of paclitaxel administration, patients show significant changes in plasma concentrations of 2-hydroxybutyrate, 3-hydroxybutyrate, pyruvate, and several amino acids involved in the TCA cycle and glycolysis. Longer courses of chemotherapy perturb sphingolipid metabolism and the biosynthesis of phenylalanine, tyrosine, and tryptophan, while adjuvant regimens alter tyrosine metabolism, lysine degradation, and branched-chain amino acid synthesis. Targeted therapies leave their own fingerprints: anti-HER2 treatment elevates plasma methionine in metastatic patients, and trastuzumab increases pantothenic acid, taurine, and L-histidine in early breast cancer. Even surgery and radiotherapy produce detectable shifts. Post-surgical plasma shows rises in sucrose—possibly reflecting prolonged physiological stress—and dodecanoic acid, an apoptosis-inducing fatty acid suggesting metabolic recovery after tumor removal. Remarkably, radiotherapy shifted several serum metabolites, including leucine, isoleucine, and lactate, toward levels observed in healthy individuals, hinting at partial metabolic normalization.

Beyond documenting these shifts, the review highlights metabolomics’ real clinical promise: predicting who will respond to neoadjuvant chemotherapy (NAC), the treatment given before surgery to shrink tumors. Achieving a pathological complete response (pCR) after NAC strongly predicts better survival, so knowing in advance who will benefit is invaluable. Here, the studies reveal subtype-specific patterns. In HER2-positive breast cancer, two independent studies found that elevated pre-treatment serum spermidine predicted good response to NAC combined with anti-HER2 agents. This polyamine likely works through antitumor immunity—intratumoral spermidine accumulation correlates with activated CD8+ T cells, and high tumor-infiltrating lymphocytes are known to predict better NAC response in this subtype.

In TNBC, the picture is more complex but equally intriguing. Poor responders showed increases in chlorokynurenine, anthranilic acid, and 3-hydroxykynurenine in pre-treatment plasma, along with elevated acetylated polyamines—pointing to altered tryptophan and polyamine metabolism, both deeply intertwined with immune regulation. Another study found that responders had decreased plasma trimethylamine N-oxide (TMAO), a gut microbiota-produced metabolite previously shown to activate endoplasmic reticulum stress kinase PERK, triggering gasdermin E-mediated pyroptosis in tumor cells and enhancing CD8+ T cell-mediated antitumor immunity. Even fecal metabolites have entered the picture: an NMR study of luminal breast cancer found that good NAC responders excreted higher levels of amino acids such as methionine, valine, alanine, and isoleucine—possibly reflecting reduced tumor demand for these building blocks as the cancer shrank. This noninvasive sampling approach also underscores the interplay between gut microbiota and chemotherapy efficacy.

Metabolomics may also forecast the dark side of treatment. The review cataloged studies linking metabolic signatures to chemotherapy-induced peripheral neuropathy, hypersensitivity reactions, cardiometabolic complications, pain, fatigue, and long-term neurologic toxicity. Histidine emerged as a recurring culprit: levels of this essential amino acid predicted the severity of paclitaxel-induced neuropathy and differed between patients who experienced doxorubicin-related hypersensitivity and those who did not. Mechanistically, histidine is converted by histidine decarboxylase into histamine, the classic mediator of allergic responses and an inflammatory neuromodulator. Aromatase inhibitor-related musculoskeletal symptoms—common in postmenopausal patients on long-term endocrine therapy—were associated with upregulated organic acids and downregulated lipid and sphingolipid pathways. Even radiotherapy-induced skin reactions showed a metabolic signature involving 13 markers, including ethanolamine and thymine, with alanine, aspartate, and glutamate metabolism most significantly altered. Such pharmacometabolomics could one day enable early intervention and dose modification before toxicity becomes debilitating.

For disease monitoring, metabolomics offers the tantalizing prospect of catching recurrence before imaging can. Patients with recurrent breast cancer exhibited significantly lower serum levels of formate, histidine, proline, choline, glutamic acid, and other metabolites compared with non-recurrent patients, with branched-chain amino acid metabolism—specifically the degradation of valine, leucine, and isoleucine—showing significant disruption. A multicenter study of preoperative serum in ER-positive early breast cancer identified a metabolite signature that independently predicted recurrence regardless of clinicopathological factors, with recurrent patients showing elevated valine, leucine, isoleucine, choline, phenylalanine, histidine, glycine, tyrosine, and lactate. The involvement of branched-chain amino acids makes biological sense: they fuel the TCA cycle for ATP production, activate mTOR signaling to drive proliferation, and valine specifically promotes cell-cycle progression through translational regulation of cyclin D2. Metabolic signatures also shift across disease stages and metastatic sites. Early-stage disease shows predominant carbohydrate metabolism, stage II features disrupted glycerophospholipid remodeling, and metastatic patients display elevated acetoacetate, ketone bodies, phenylalanine, and glutamate—the latter fueling invasion through glutathione production and the system Xc-antiporter. A 15-metabolite panel predicted brain metastasis with 96.9% accuracy.

Prognostically, the most consistent signal across studies is lactate. Elevated lactate and glycine in tumor tissue, and elevated lactate and pyruvate in serum, correlate with reduced relapse-free survival and overall survival, particularly in ER-positive patients. Lactate is far more than waste: it acidifies the tumor microenvironment to promote invasion, stimulates angiogenesis through hypoxia-related pathways, suppresses cytotoxic T cells and natural killer cells, renders tumors resistant to radiotherapy, and even regulates gene expression through lactylation, a post-translational modification that drives tumor progression. Bile acids tell a contrasting story: glycochenodeoxycholate levels were positively associated with survival and inversely correlated with tumor proliferation scores. In TNBC, elevated plasma diacetylspermine, a spermine catabolite, marked increased metastasis risk and poorer survival.

The authors are candid about the field’s obstacles. Analytical platforms differ in sensitivity and metabolite coverage, sample handling varies widely, chemotherapy regimens are often pooled in ways that obscure drug-specific effects, and definitions of response differ between studies using pCR, residual cancer burden, RECIST criteria, or survival endpoints. Small sample sizes—ranging from 8 to 699 patients—compound the problem, and confounders such as diet, comorbidities, and smoking are often unaddressed. Only a minority of studies performed subtype-specific analyses or integrated metabolomics with other omics layers. The review calls for large, multi-institutional prospective trials with standardized protocols, longitudinal sampling designs, and multi-omics integration.

Still, the trajectory is clear. Metabolomics offers something conventional biomarkers and imaging cannot: the ability to detect early biochemical perturbations that precede visible disease change, from a noninvasive blood sample, repeatedly over time. If the field can achieve the standardization the authors demand, metabolic fingerprints—especially when fused with genomic and transcriptomic data—could transform breast cancer from a disease managed by population averages into one monitored molecule by molecule, patient by patient.

For patients with early-stage disease, surgery can be curative, but it does not always eliminate the risk of relapse. Conventional scans provide an important view of anatomy, yet they may not detect a small population of cancer cells before it grows into a visible lesion. MRD testing approaches the problem from a different direction. Instead of searching for a mass, it looks for fragments of tumor DNA circulating in the blood. If those fragments persist after resection, they may indicate that microscopic disease remains, even when imaging appears clear.

The study’s focus on EGFR-mutated lung cancer is particularly significant. EGFR mutations can drive the uncontrolled growth of tumor cells and are found in a substantial proportion of lung adenocarcinomas, especially among people who have never smoked or have smoked lightly. These alterations also create an opportunity for precision medicine because they can be targeted by drugs known as EGFR tyrosine kinase inhibitors. The challenge is determining which patients need additional treatment after surgery and which may be spared months or years of therapy and its potential side effects.

Molecular residual disease detection is designed to make that decision more precise. After a tumor is removed, researchers can analyze its genetic profile and identify mutations or other molecular features unique to that cancer. Highly sensitive sequencing methods can then search for matching fragments in subsequent blood samples. The technical difficulty is considerable: tumor DNA may represent only a tiny fraction of all cell-free DNA in the bloodstream, while normal tissues continuously release their own genetic material. A reliable test must therefore distinguish a genuine cancer signal from background noise and laboratory artifacts.

The clinical value of MRD does not rest solely on whether a test can detect DNA. The crucial question is whether the result changes what doctors do and improves outcomes for patients. A positive result might identify people at particularly high risk of recurrence, supporting closer surveillance or consideration of adjuvant targeted treatment. A negative result could help define a group with a lower immediate risk, although it cannot guarantee that a relapse will never occur. The timing of blood collection, the depth of sequencing, the mutation selected for tracking and the duration of follow-up all influence the meaning of a result.

In EGFR-mutated disease, the stakes are amplified by the availability of effective targeted therapies. Drugs such as osimertinib have demonstrated benefits in the postoperative setting, but treatment decisions still require a balance between reducing recurrence risk and avoiding unnecessary exposure. MRD could eventually provide a dynamic measure of disease status, allowing care to become more responsive than a one-time decision based only on tumor stage and pathology. A rising molecular signal might prompt further investigation, while sustained clearance could help doctors assess whether treatment is suppressing residual disease.

The research also highlights why a blood-based test should be interpreted as part of a broader clinical framework rather than as an isolated verdict. A negative result may reflect the biological limits of detection, particularly when a tumor sheds little DNA into the bloodstream. A positive result may require confirmation, because technical contamination or clonal changes in non-cancerous cells can complicate genetic analysis. For this reason, the practical adoption of MRD testing depends on standardized laboratory methods, carefully defined thresholds and prospective evidence connecting test results with treatment decisions and long-term survival.

As precision oncology moves beyond matching drugs to mutations, it is increasingly turning toward the continuous monitoring of disease. The work by Zhou and colleagues places EGFR-mutated early-stage lung cancer within that wider transformation, where molecular information collected after surgery may help reveal what conventional scans cannot yet see. The promise is substantial: earlier recognition of recurrence, more individualized use of targeted therapy and a clearer understanding of who remains at risk. The field’s next challenge is ensuring that molecular signals translate into decisions that are not only technically accurate, but demonstrably better for patients.

Subject of Research: Molecular residual disease detection in early-stage resected EGFR-mutated non-small cell lung cancer

Article Title: Clinical utility of molecular residual disease detection in early-stage resected EGFR-mutated non-small cell lung cancer

Article References: Zhou, F., Su, C., Liang, W. et al. Clinical utility of molecular residual disease detection in early-stage resected EGFR-mutated non-small cell lung cancer. Nature Communications (2026). https://doi.org/10.1038/s41467-026-76392-9

Image Credits: AI Generated

DOI: 10.1038/s41467-026-76392-9

Keywords: Molecular residual disease, MRD, EGFR mutation, non-small cell lung cancer, lung cancer, liquid biopsy, circulating tumor DNA, precision oncology, cancer recurrence, targeted therapy