Neoadjuvant immunotherapy, in which immune checkpoint inhibitors are administered before surgical resection, has transformed the management of locally advanced non-small cell lung cancer (NSCLC). When the treatment works exceptionally well, the tumor may vanish entirely under the microscope — a phenomenon called pathological complete response, or pCR — which is strongly associated with improved long-term survival and recurrence-free outcomes. The problem, however, is that pCR can only be confirmed definitively after the tumor has been surgically removed and examined by a pathologist. Patients who are unlikely to respond may undergo months of immunotherapy and its associated costs and toxicities for little benefit, while those most likely to benefit cannot be reliably identified in advance with current imaging or clinical assessments.

The research team, led by investigators at Tianjin Medical University Cancer Institute and Hospital in collaboration with Shandong Cancer Hospital and Institute, addressed this challenge using radiomics — a computational approach that extracts hundreds of quantitative features from medical images, capturing subtle patterns of texture, intensity, shape, and spatial relationships that are invisible to the human eye. Rather than focusing solely on the primary tumor, as most prior radiomic studies have done, the researchers deliberately analyzed the whole lung, reasoning that the tumor’s immunological microenvironment and the surrounding lung tissue might harbor additional predictive signals that a tumor-only analysis would miss.

The study enrolled 284 patients with stage II to IIIB NSCLC from the first center, treated between August 2018 and June 2025, who were randomly divided into a training set and an internal test set in a 7:3 ratio stratified by pCR status. An additional 82 patients from the second center, treated between April 2023 and March 2026, served as an external validation cohort — a critical step for confirming that the model’s performance generalizes beyond a single institution. All patients underwent baseline computed tomography before starting neoadjuvant immunotherapy, and radiomic features were extracted from two regions of interest: the primary tumor itself and the entire lung parenchyma.

To build the predictive models, the team employed LightGBM, a gradient-boosting decision tree algorithm known for its efficiency and accuracy with high-dimensional data. Two independent models were trained: one using only primary tumor (PT) features and one using only whole-lung (WL) features. The predicted probabilities from both models were then fused into a combined PTWL model. The results were striking. On the internal test set, the PT model achieved an area under the receiver operating characteristic curve (AUC) of 0.701, while the WL model reached 0.676 — comparable but individually modest. When the two were combined, however, the PTWL model’s AUC jumped to 0.788 internally and 0.787 on the external validation set, demonstrating that whole-lung information adds genuine predictive value beyond what the tumor alone can reveal.

Among the 366 patients in the overall cohort, whose median age was 63 years and who were predominantly male (317 of 366), 151 — or 41.3 percent — achieved a pathological complete response after neoadjuvant immunotherapy. The researchers further enhanced their model by incorporating carcinoembryonic antigen (CEA), a well-established serum tumor marker, into the PTWL signature. This combined model showed improved positive predictive value, reaching 0.769 in the internal test set and 0.636 in the external cohort, meaning that when the model predicted pCR, it was increasingly likely to be correct.

Perhaps the most clinically innovative aspect of the study is its hierarchical decision-making framework. Rather than relying on a single model output, the researchers designed a two-tiered system: the PTWL radiomic model performs initial risk stratification, sorting patients into preliminary risk categories, and then the combined model — incorporating both radiomics and CEA — serves as a secondary confirmation step for patients in specific risk tiers. This layered approach mirrors how clinicians actually make decisions, using multiple lines of evidence rather than a single test result, and it allows for more nuanced, tiered clinical decision-making than a one-size-fits-all prediction.

The framework’s clinical relevance was validated through survival analysis. Patients classified as low-risk by the hierarchical framework exhibited significantly prolonged recurrence-free survival, as demonstrated by Kaplan-Meier analysis with a log-rank P value below 0.05. This finding suggests that the model is not merely predicting a pathological endpoint but is also capturing information relevant to real-world oncological outcomes — a critical distinction for any tool intended to guide treatment decisions. Recurrence-free survival, the period after treatment during which no cancer returns, is one of the most meaningful endpoints for patients and oncologists alike.

From a technical standpoint, the study’s methodology reflects several best practices in radiomics research. The team assessed feature reproducibility using interclass correlation coefficients (ICC), a standard measure of how consistently features can be extracted across repeated measurements. Model discrimination was evaluated using receiver operating characteristic curves, and decision curve analysis (DCA) was employed to assess clinical utility — determining at which probability thresholds the model would provide net benefit if used in practice. The researchers also reported net reclassification improvements (NRI) and integrated discrimination improvements (IDI), sophisticated statistical measures that quantify how much better one model is at correctly classifying patients compared to another, providing rigorous evidence that the whole-lung integration genuinely improves prediction rather than merely adding complexity.

The implications of this work extend beyond NSCLC. The finding that whole-lung radiomic features — capturing the broader pulmonary environment, including emphysematous changes, inflammatory patterns, and parenchymal heterogeneity — contribute predictive information suggests that the tumor’s surroundings matter in immunotherapy response. This aligns with growing biological evidence that the immune landscape of the entire organ, not just the tumor itself, influences how checkpoint inhibitors perform. Patients with chronic obstructive pulmonary disease (COPD), for instance, may have a different immune contexture in their lungs that affects treatment response, and whole-lung analysis may implicitly capture such factors. As the authors conclude, their hierarchical framework facilitates individualized neoadjuvant treatment by reinforcing pCR identification, potentially sparing patients unlikely to benefit from unnecessary immunotherapy while confidently guiding those most likely to achieve complete pathological response toward the most effective treatment pathway. With further prospective validation, this approach could become a routine part of the pre-treatment workup for lung cancer patients worldwide.

Subject of Research: Whole-lung radiomics for predicting pathological complete response to neoadjuvant immunotherapy in non-small cell lung cancer

Article Title: Enhancing prediction of pathological complete response to neoadjuvant immunotherapy in non-small cell lung cancer by whole-lung radiomics

Article References: Cai, X., Yang, J., E, J., Wang, D., Qu, F., Wang, Q., Ning, Z., Liu, F., Zhang, Y., Gao, B., Sun, B., & Liu, Y. (2026). Enhancing prediction of pathological complete response to neoadjuvant immunotherapy in non-small cell lung cancer by whole-lung radiomics. BMC Cancer. https://doi.org/10.1186/s12885-026-16939-7

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16939-7

Keywords: non-small cell lung cancer, neoadjuvant immunotherapy, radiomics, pathological complete response, artificial intelligence, LightGBM, CT imaging, CEA, recurrence-free survival, predictive medicine, tumor microenvironment, personalized oncology

Intrahepatic cholangiocarcinoma, often abbreviated ICC, is a primary malignancy of the liver that originates from the epithelial cells lining the intrahepatic bile ducts. It is the second most common primary hepatic malignancy, trailing only hepatocellular carcinoma, and its incidence has been rising in many parts of the world over recent decades. The clinical stakes of distinguishing ICC from hepatocellular carcinoma are considerable. Hepatocellular carcinoma, particularly when diagnosed early, can often be managed with established protocols ranging from ablation and resection to liver transplantation and, more recently, immunotherapy-based systemic regimens. Intrahepatic cholangiocarcinoma, by contrast, typically demands aggressive surgical resection with careful attention to lymph node status and biliary anatomy, and its response to treatments designed for hepatocellular carcinoma is generally poorer. When a liver mass cannot be definitively characterized before surgery, the surgical team may be forced to plan for both possibilities, and misclassification can lead to suboptimal operative strategy and delayed access to the most appropriate therapy.

Conventional tumor markers offer only partial help. Carbohydrate antigen 19-9, commonly known as CA19-9, and carcinoembryonic antigen, or CEA, are the most widely used serological markers in biliary tract malignancies, but neither is specific enough to reliably separate intrahepatic cholangiocarcinoma from hepatocellular carcinoma, particularly in patients who also have chronic hepatitis or cirrhosis, conditions in which these markers can be elevated for reasons unrelated to cancer. Imaging, too, has its limits; small tumors, atypical enhancement patterns, and underlying liver disease can all blur the radiological distinction. This diagnostic gray zone is precisely where the new study positions itself, asking whether the lipid composition of blood plasma might carry a signature specific enough to resolve what imaging and protein markers cannot.

The research team, led by Anrui Mao and Jie Yang, with corresponding authors Kun Tong and Hongkun Zhou, drew on clinical expertise from the Department of Hepatobiliary and Pancreatic Surgery at the First Hospital of Jiaxing and collaboration with Zhejiang Chinese Medical University. The investigators assembled three groups of participants: fifty healthy normal controls, fifty patients with hepatocellular carcinoma, and thirty patients with intrahepatic cholangiocarcinoma. Plasma samples were collected from each participant and subjected to untargeted lipidomic analysis, a technique that does not presuppose which lipids might be relevant but instead surveys the entire lipid complement of a sample in an unbiased fashion. This approach is powerful precisely because it allows the data to reveal which molecular species separate the groups, rather than testing a preselected hypothesis about a single candidate molecule.

The technology underlying this survey is liquid chromatography coupled to mass spectrometry, a workhorse of modern metabolomics. In essence, the lipid extracts from each plasma sample are separated on a chromatographic column according to their chemical properties and then ionized and weighed with extraordinary precision by the mass spectrometer. Each lipid species produces a characteristic retention time and mass-to-charge ratio, and with the aid of spectral databases, these features can be assigned to specific molecular identities, including the class of lipid, the total number of carbon atoms in its fatty acyl chains, the number of double bonds, and additional structural features such as oxygen substitutions. The resulting dataset, containing thousands of features per sample, is then interrogated with statistical tools to find species whose abundance differs significantly between diagnostic groups. It is a molecular census of the blood, conducted at a level of detail that was simply unattainable a generation ago.

What the census revealed was striking in its internal logic. When the researchers compared patients with intrahepatic cholangiocarcinoma to healthy controls, the lipid species that changed most prominently belonged to three classes: phosphatidylethanolamine, sphingomyelin, and ceramide. These are not random molecules. Phosphatidylethanolamine and sphingomyelin are major structural phospholipids of cellular membranes, and ceramide sits at the hub of sphingolipid metabolism, a pathway intimately involved in cell death, proliferation, and stress responses. Cancer cells are known to remodel their membrane lipid composition extensively to support rapid growth, altered signaling, and survival under hypoxic and nutrient-poor conditions. The fact that the cholangiocarcinoma signature concentrated in the sphingolipid pathway suggests that bile duct-derived tumors may rewire lipid metabolism in a fundamentally different way than hepatocyte-derived tumors do.

That difference became even clearer when the hepatocellular carcinoma group was examined. Patients with this tumor type showed their most prominent lipid alterations in a different trio of classes: phosphatidylcholine, triglycerides, and acylcarnitines. Phosphatidylcholine is the counterpart membrane phospholipid to phosphatidylethanolamine, while triglycerides and acylcarnitines reflect fatty acid storage and mitochondrial fatty acid transport, respectively. Hepatocellular carcinoma arises from hepatocytes, the liver’s metabolic workhorses, and its lipid perturbations appear to echo the metabolic identity of the cell of origin. In other words, the two tumors leave distinct lipidological calling cards in the plasma, and the divergence is not merely a matter of degree but of pathway.

At the center of the study stands one molecule. Sphingomyelin (30:1; O2) was significantly upregulated in the plasma of patients with intrahepatic cholangiocarcinoma relative to healthy controls, but significantly downregulated in patients with hepatocellular carcinoma, with both differences reaching statistical significance at P less than 0.05. This bidirectional behavior is what gives the molecule its diagnostic power: it does not simply flag the presence of disease, it points toward one disease and away from the other. The nomenclature itself encodes the molecule’s structure, with 30:1 indicating thirty carbon atoms and one double bond across its acyl chains, and O2 denoting two additional oxygen atoms, features that define a specific and measurable molecular entity rather than a vague class of compounds.

The quantitative performance of this single lipid was remarkable. When the researchers constructed receiver operating characteristic analyses to test how well SM (30:1; O2) discriminated intrahepatic cholangiocarcinoma from the other groups, it achieved an area under the curve of 0.8793 with P less than 0.001, a level of diagnostic accuracy that exceeded the conventional markers CA19-9 and CEA in the same comparison. An area under the curve approaching 0.88 indicates substantially better-than-chance discrimination, and the fact that a single lipid species outperformed established protein biomarkers underscores the untapped diagnostic information carried in the plasma lipidome. Perhaps even more intriguingly, correlation analysis revealed that circulating levels of SM (30:1; O2) were strongly and significantly positively correlated with the TNM stage of intrahepatic cholangiocarcinoma, with a Spearman correlation coefficient of 0.721 and P less than 0.001. A correlation of that magnitude suggests that the lipid is not merely a passive bystander but tracks closely with tumor burden and progression, raising the possibility that it could eventually contribute not only to diagnosis but also to staging and monitoring.

The authors conclude that SM (30:1; O2) is a promising biomarker for intrahepatic cholangiocarcinoma, capable of effectively distinguishing it from hepatocellular carcinoma and providing valuable guidance for preoperative diagnosis and therapeutic decision-making. The study, which was approved by the Ethics Committee of the First Hospital of Jiaxing and conducted in accordance with the Declaration of Helsinki, was supported by the Key Medical Discipline of Jiaxing City in the hepatopancreatobiliary direction, the National Oncology Clinical Key Speciality, and the Zhejiang Provincial Medical and Health Science and Technology Projects. As with any biomarker discovery study, the findings will need validation in larger, independent, and multi-center cohorts before a sphingomyelin assay could enter routine clinical practice, and questions about how the lipid behaves in other liver diseases, such as cirrhosis, cholangitis, and steatohepatitis, will need careful answers. But the conceptual advance is clear and consequential: the metabolic identity of a liver tumor, written in the language of lipids, can be read from a simple blood draw. If subsequent studies confirm the performance of SM (30:1; O2), surgeons could one day walk into the operating room knowing, before the first incision, which cancer they are about to face, and patients could receive the right operation, the right therapy, and the right prognosis far earlier in their disease course.

Subject of Research: Lipidomic identification of sphingomyelin (30:1; O2) as a plasma diagnostic biomarker for intrahepatic cholangiocarcinoma

Article Title: Sphingomyelin (30:1; O2) as a novel lipid biomarker associated with the diagnosis of intrahepatic cholangiocarcinoma

Article References: Mao, A., Yang, J., Bian, C., Jia, Y., Tong, K., & Zhou, H. (2026). Sphingomyelin (30:1; O2) as a novel lipid biomarker associated with the diagnosis of intrahepatic cholangiocarcinoma. BMC Cancer. https://doi.org/10.1186/s12885-026-17142-4

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17142-4

Keywords: intrahepatic cholangiocarcinoma, hepatocellular carcinoma, sphingomyelin, lipidomics, biomarker, liquid chromatography-mass spectrometry, CA19-9, CEA, liver cancer, sphingolipids, TNM stage, plasma diagnostics

The study enrolled 139 patients with adenocarcinoma of the esophagogastric junction who underwent curative resection. Of these, 65 patients experienced a disease-free survival event, meaning their cancer returned or they died during the follow-up period, while 74 patients remained free of disease. The researchers set out to determine whether imaging-derived measures could add predictive power beyond the conventional clinical factors that oncologists already use, such as tumor stage and serum tumor markers. To do this, they systematically analyzed clinical variables, body composition parameters, and spectral CT parameters, then used Cox regression analysis to identify which factors independently predicted worse disease-free survival.

Spectral CT, sometimes called dual-energy or energy-resolving CT, is a technology that distinguishes tissues not only by their conventional X-ray attenuation but also by how they behave at different X-ray energy levels. One of its most useful quantitative outputs is the extracellular volume fraction, a measure that reflects the proportion of extracellular space within a tissue. In tumors, an elevated extracellular volume fraction is generally interpreted as a sign of increased stromal content, tissue disorganization, and possibly more aggressive biology. Because spectral CT can quantify this property noninvasively from contrast-enhanced scans that patients already receive as part of standard staging, the researchers reasoned that it might serve as a preoperative window into tumor aggressiveness without requiring any additional invasive procedure.

The second pillar of the model is the skeletal muscle index, a body composition measure derived from the cross-sectional muscle area at a defined anatomical level, normalized for the patient’s height. Low skeletal muscle index, often described as sarcopenia, is a well-recognized marker of frailty and poor physiological reserve in cancer patients. In cancers of the upper gastrointestinal tract, where malnutrition and weight loss are common, muscle depletion has repeatedly been linked to worse surgical outcomes and survival. By quantifying muscle mass objectively from the same CT scan, the investigators could capture an aspect of the patient’s condition that is often missed by simple body weight or body mass index measurements.

The results of the Cox regression analysis identified four independent predictors of worse disease-free survival. Elevated carcinoembryonic antigen, or CEA, above 5 nanograms per milliliter carried the largest effect, with a hazard ratio of 3.31, indicating that patients with elevated CEA faced more than three times the hazard of a disease-free survival event compared with those below the threshold. Clinical T3 or T4 stage, reflecting deeper tumor invasion, more than doubled the hazard with a hazard ratio of 2.63. Low skeletal muscle index was associated with a hazard ratio of 2.18, and a high extracellular volume grade carried a hazard ratio of 1.98. All four associations reached statistical significance, with p values ranging from less than 0.001 to 0.026, confirming that both the tumor-derived spectral parameter and the patient-derived body composition measure contributed independent prognostic information.

To translate these findings into a practical tool, the researchers constructed three nested models of increasing complexity. Model 1 incorporated only the conventional clinical factors, CEA and clinical T stage. Model 2 added the body composition measure, the skeletal muscle index, and Model 3 further incorporated the spectral CT-derived extracellular volume grade. The models were assessed using time-dependent receiver operating characteristic analysis, which evaluates how well a model discriminates between patients who do and do not experience events at specific time points, along with calibration curves and decision curve analysis to judge clinical usefulness.

The performance gains from adding the imaging measures were consistent across all three time horizons evaluated. At one year, the area under the curve rose from 0.715 for Model 1 to 0.747 for Model 2 and 0.762 for Model 3. At two years, the corresponding values were 0.738, 0.759, and 0.773. At three years, the models achieved 0.768, 0.810, and 0.818 respectively. Although the incremental improvements in discrimination may appear modest in absolute terms, they were achieved using information that is already available before surgery, at no additional cost or risk to the patient. Calibration curves showed acceptable agreement between predicted and observed disease-free survival probabilities, and decision curve analysis supported the net clinical benefit of the fuller models across a range of threshold probabilities.

Perhaps the most striking finding emerged from an exploratory subgroup analysis that crossed the two imaging-derived variables. Patients who had both a high extracellular volume grade and a low skeletal muscle index showed the worst disease-free survival of any subgroup in the study. This combination is biologically plausible: it pairs a tumor that appears stroma-rich and potentially more aggressive on spectral imaging with a host whose muscle depletion suggests diminished reserve and possibly a more inflammatory, catabolic state. The convergence of adverse tumor biology and adverse host condition appears to compound risk in a way that neither measure captures alone, and the authors suggest this combined signature may support risk-adapted perioperative assessment and postoperative surveillance.

The practical implications of the study are considerable for the management of this tumor type. Adenocarcinoma of the esophagogastric junction sits at a diagnostic and therapeutic crossroads, treated variably by thoracic, upper gastrointestinal, and surgical oncology teams, and its prognosis after resection varies widely. A preoperative model built entirely from data available at staging, the CT scan and routine blood work, could help clinicians decide which patients warrant intensified neoadjuvant treatment, closer postoperative imaging surveillance, or earlier nutritional and physical interventions to address sarcopenia before it compounds recurrence risk. Because the model is preoperative, it does not depend on pathological examination of the resected specimen, which is only available after the operation when many treatment decisions have already been made.

The authors and readers alike should note the limitations inherent in the study design. It was a single-center retrospective analysis of 139 patients from Lanzhou University Second Hospital, and the model has so far been internally validated rather than tested on an independent external cohort. The exploratory subgroup findings, while compelling, require confirmation in larger and more diverse populations before they can change practice. The study was approved by the Institutional Review Board of the Second Hospital of Lanzhou University, with all clinical and imaging data fully anonymized before analysis, and the authors declare no competing interests. The research was supported by several Chinese funding bodies, including the National Natural Science Foundation of China and provincial science and technology programs. Even with these caveats, the study represents a concrete step toward a future in which the standard staging CT scan, already performed on nearly every patient, yields far more than tumor size and location, providing quantitative, reproducible measures of both the tumor’s microenvironment and the patient’s physical condition that together refine the forecast for life after cancer surgery.

Subject of Research: Preoperative spectral CT imaging biomarkers for predicting disease-free survival in esophagogastric junction adenocarcinoma

Article Title: Preoperative integration of spectral CT-derived extracellular volume and skeletal muscle index for disease-free survival prediction in adenocarcinoma of the esophagogastric junction

Article References: Liu, X., You, H., Ren, T., Chen, Q., Xu, M., Ma, L., Peng, Y., Zhang, C., Bao, J., Zhang, W., & Zhou, J. (2026). Preoperative integration of spectral CT-derived extracellular volume and skeletal muscle index for disease-free survival prediction in adenocarcinoma of the esophagogastric junction. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02873-4

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02873-4

Keywords: esophagogastric junction adenocarcinoma, spectral CT, extracellular volume fraction, skeletal muscle index, sarcopenia, disease-free survival, prognosis, medical imaging, body composition, CEA, cancer staging, Cox regression

The patient’s story began deceptively. He had been managed for bronchial asthma for two years, but over the six months before presentation he developed worsening breathlessness on exertion and wheezing. He had no cough and no coughing up of blood, and there were no laboratory signs of disseminated intravascular coagulation, the catastrophic clotting disorder that often accompanies advanced malignancy. When his symptoms suddenly accelerated, imaging became the first clue that something far more sinister than an asthma flare was underway. A computed tomography scan of the chest revealed diffuse centrilobular nodules distributed throughout both lungs, a so-called tree-in-bud pattern, and mild bilateral ground-glass opacities, the hazy areas that suggest partial filling or thickening of the air spaces.

That tree-in-bud appearance is classically the signature of infectious bronchiolar disease, particularly tuberculosis and nontuberculous mycobacterial infection, and it initially steered clinicians toward an infectious explanation. The patient was treated for a presumed asthma exacerbation with infection, but his respiratory status continued to deteriorate. Laboratory testing added a crucial piece of the puzzle: there was no marked inflammatory response, yet his serum carcinoembryonic antigen, or CEA, a tumour marker, was dramatically elevated at 34.7 nanograms per millilitre against a normal ceiling of 5.0. Microbiological and serological investigations, including tests for tuberculosis, failed to support any infectious cause. Together, the elevated tumour marker and the atypical radiology pointed toward a noninfectious, malignant vascular process.

Positron emission tomography using fluorodeoxyglucose showed mild tracer uptake in a pulmonary ground-glass lesion, with no abnormal activity elsewhere in the body to suggest a primary tumour outside the chest. Within two weeks, the man’s breathing had deteriorated so severely that he required urgent hospitalisation for hypoxaemia, a dangerous drop in blood oxygen. Contrast-enhanced CT found no large pulmonary artery thrombi, which might have suggested ordinary clot embolism, but lung perfusion scintigraphy revealed widespread bilateral perfusion defects, meaning large regions of lung tissue were receiving little or no blood flow. Echocardiography demonstrated mild pulmonary hypertension with strain on the right ventricle, the chamber that must pump blood through the lungs and which fails early when pulmonary vessels become obstructed.

On hospital day four, bronchoscopy with bronchoalveolar lavage and transbronchial lung biopsy was performed. Cultures of the lavage fluid were negative for bacteria and mycobacteria, further excluding infection. The biopsy revealed adenocarcinoma, but critically, the histopathology showed preserved vascular architecture without definitive tumour emboli, so PTTM could not be confirmed under the microscope. Right heart catheterisation on day five showed pressures that, while abnormal, had not yet crossed the formal diagnostic threshold for pulmonary hypertension, likely because the procedure was performed early in the course. Faced with a patient in rapid decline, the clinical team made the judgment call to diagnose PTTM associated with carcinoma of unknown primary on clinical grounds alone and to begin treatment before completing a full diagnostic workup, a decision that almost certainly saved his life.

After discussion with the patient and his family, the team selected nivolumab monotherapy, an immune checkpoint inhibitor that blocks the PD-1 receptor and unleashes T cells against tumour cells, initiated on hospital day eleven. Molecular testing had shown no EGFR mutation and no ALK or ROS1 rearrangements, ruling out targeted therapies, but PD-L1 immunostaining revealed strong expression in more than 75 percent of tumour cells, a biomarker associated with responsiveness to checkpoint blockade. Immunohistochemistry showed nuclear positivity for thyroid transcription factor-1, suggesting the tumour likely originated in the lung, though the small biopsy specimen could not definitively establish the primary site, and upper gastrointestinal endoscopy, performed because gastric cancer is the malignancy most often linked to PTTM, found nothing.

The early course was harrowing. Despite starting nivolumab, the man’s respiratory condition continued to worsen, and by day fourteen he needed high-flow nasal cannula oxygen. By day twenty-six he had developed severe disturbance of consciousness, likely reflecting the combined effects of profound hypoxaemia and the metabolic strain of a failing pulmonary circulation. Then, remarkably, both his consciousness and his oxygenation improved by day twenty-nine, and he was discharged on day fifty-four without any neurological or respiratory after-effects. Serial CEA measurements told the story of the tumour’s retreat: the marker fell progressively from 34.7 nanograms per millilitre before treatment to within the normal range two months after discharge, and it has remained normal throughout follow-up. Serum BNP, a marker of cardiac strain, transiently rose and then normalised, mirroring the resolution of right ventricular stress seen on serial echocardiography.

The treatment was not without cost. During follow-up the patient developed a series of immune-related adverse events, the flip side of a successfully activated immune system, including thyroid dysfunction, renal dysfunction and hepatic dysfunction. These were managed with close monitoring, and nivolumab was continued until after seven cycles, when treatment had to be stopped because of grade 3 pneumonitis, a significant inflammation of the lung tissue itself. Despite that complication, the underlying disease has never returned. The patient remains progression-free more than three years after symptom onset, joining only two previously reported patients who survived longer than three years with PTTM, one with stage IV breast cancer treated with chemotherapy and another with ROS1-rearranged lung cancer treated with crizotinib under extracorporeal membrane oxygenation.

The authors of the report argue that the case carries two broad lessons. First, the vascular tree-in-bud pattern, in which neoplastic emboli rather than infected secretions produce the characteristic branching nodules on CT, deserves wider recognition as a potential marker of PTTM, particularly when accompanied by rising tumour markers, absent inflammatory response and rapidly progressive hypoxaemia. Because transbronchial biopsy specimens are small and subject to sampling error, pathological confirmation of PTTM is often impossible in life, and clinicians may need to act on integrated clinical, radiological and haemodynamic evidence rather than waiting for a definitive histological diagnosis. Second, and perhaps more importantly, the case suggests that when PTTM is strongly suspected, prompt systemic treatment of the underlying malignancy may be justified even before the primary tumour is identified. No standard therapy exists for PTTM, and reported approaches, from anticoagulation and corticosteroids to pulmonary vasodilators and imatinib, rest on little more than case reports. But in all three long-term survivors to date, including this one, durable control of the underlying cancer was accompanied by resolution of the pulmonary vascular catastrophe. The exceptionally high PD-L1 expression in this patient may have contributed to the response to nivolumab, and the authors suggest that early recognition of PTTM combined with effective anticancer therapy may be the key to converting a disease measured in weeks into one measured in years.

Subject of Research: Long-term survival in pulmonary tumour thrombotic microangiopathy treated with nivolumab for cancer of unknown primary

Article Title: Long‐Term Survival Following Nivolumab Treatment in Pulmonary Tumour Thrombotic Microangiopathy Associated With Cancer of Unknown Primary: A Case Report

Article References: Fujishima, K., Osugi, J., Noma, S., Higuchi, M., & Suzuki, H. (2026). Long‐Term Survival Following Nivolumab Treatment in Pulmonary Tumour Thrombotic Microangiopathy Associated With Cancer of Unknown Primary: A Case Report. Respirology Case Reports, 14(10), Article e70763. https://doi.org/10.1002/rcr2.70763

Image Credits: AI Generated

DOI: 10.1002/rcr2.70763

Keywords: pulmonary tumour thrombotic microangiopathy, nivolumab, cancer of unknown primary, immunotherapy, pulmonary hypertension, PD-L1, CEA, tree-in-bud pattern, checkpoint inhibitor, case report, lung adenocarcinoma, immune-related adverse events

Gastric cancer remains one of the most lethal malignancies worldwide, and its burden is particularly heavy in China, where it ranks among the leading causes of cancer death. For decades, one bacterium has dominated the conversation about stomach cancer: Helicobacter pylori, the spiral-shaped microbe notorious for driving chronic inflammation and ulcers. But scientists increasingly recognize that the stomach is home to entire communities of microbes, and that tumors themselves — once thought to be sterile — harbor their own distinctive microbiota. What those intratumoral communities look like, and whether they matter clinically, has remained murky.

The Nantong team, led by Haonan Zhou, Wei Feng, Yuxi Li and senior authors Wei Zong and Yan Huang, set out to resolve that ambiguity by combining two complementary DNA-sequencing approaches. The first, 16S ribosomal RNA sequencing, is the workhorse of microbiome research: it reads a conserved genetic marker present in all bacteria, allowing researchers to identify which microbial groups are present in a sample. The second, a technique called 2bRAD-M, takes a different route. Instead of targeting a single marker gene, it uses a restriction enzyme to slice the entire genomes in a sample into tiny, standardized fragments, which are then sequenced and mapped back to reference genomes. That gives the method the power to profile not just bacteria but also fungi and archaea, and to reach species-level resolution even when microbial DNA is scarce or degraded — a common problem in tumor tissue.

By running both methods on the same samples, the researchers could cross-validate their findings and capture a fuller picture than either technique alone could provide. They analyzed sixty paired samples: tumor tissue and adjacent normal tissue taken from the same patients with gastric cancer. Sequencing the pairs side by side allowed the team to control for each patient’s individual background and isolate the microbial changes specifically tied to the tumor itself.

The results revealed a clear signature of dysbiosis — a destabilized, imbalanced microbial community — within the tumors. At the phylum level, the firmicutes, a broad group of mostly Gram-positive bacteria, were markedly enriched in cancer tissue, while proteobacteria and several other major phyla declined. Zooming in to the genus and species levels, the team found that Lactobacillus, a genus usually considered benign or even beneficial in the gut, was enriched in tumor tissue, as was Streptococcus pneumoniae, a pathogen better known for causing pneumonia. Strikingly, Helicobacter pylori — the microbe most famously associated with stomach cancer — was actually reduced in the tumor samples relative to adjacent normal tissue.

That last finding is a reminder that the microbiology of established tumors may differ fundamentally from the microbiology of cancer initiation. H. pylori is a powerful driver of the chronic inflammation that sets the stage for gastric cancer, but once a tumor has formed, the microenvironment it creates — altered oxygen levels, changed acidity, disrupted immune surveillance — may favor entirely different microbial colonizers. The enrichment of Lactobacillus in particular raises intriguing questions: some studies have suggested that certain lactobacilli can promote tumor cell survival and proliferation, while others report protective effects. The new data add weight to the idea that in gastric tumors, this genus is part of a disease-associated community rather than a harmless passenger.

Critically, the microbial shifts were not just statistical noise. The team found that the abundance of specific microbial taxa correlated with clinical features of the disease, including TNM stage — the standard system for describing how far a tumor has grown and spread — and levels of carcinoembryonic antigen, or CEA, a protein biomarker routinely measured in patients with gastrointestinal cancers. In other words, the composition of the tumor microbiome tracked with measures of disease severity, raising the possibility that microbial profiles could one day complement existing biomarkers in assessing patients.

To verify that the sequenced DNA was not simply contamination introduced during sample handling, the researchers turned to independent methods. Fluorescence in situ hybridization, or FISH, uses fluorescently labeled probes that bind to specific bacterial RNA sequences, allowing intact microbial cells to be visualized directly inside tissue sections under the microscope. Immunohistochemistry for lipopolysaccharide, a molecule embedded in the outer membrane of many Gram-negative bacteria, provided a second line of visual confirmation. Together, these techniques demonstrated that bacteria genuinely reside within the tumor tissue, bolstering the sequencing results and addressing one of the most persistent criticisms of intratumoral microbiome studies.

Beyond cataloguing which microbes are present, the team used computational prediction tools to infer what the tumor-dwelling communities might be doing. Functional profiling suggested that microbial pathways related to metabolism and proliferation were significantly upregulated in gastric cancer tissue compared with normal tissue. While such predictions are indirect — they extrapolate function from the genes carried by the detected microbes rather than measuring activity directly — they hint that the intratumoral microbiota could be metabolically active participants in the tumor microenvironment, potentially supplying metabolites, modulating inflammation, or interacting with immune cells in ways that influence tumor behavior.

The study’s authors are careful to frame their conclusions appropriately. The work confirms that intratumoral microbiota dysbiosis exists in patients with gastric cancer and links specific taxa and functional pathways to clinical characteristics, but it stops short of proving causation. Whether the altered microbial communities help drive tumor progression, merely reflect it, or do some of both remains an open question that will require larger prospective cohorts and direct experimental studies. Still, the findings provide a foundation for future therapeutic strategies — from microbiome-based biomarkers that could refine risk assessment to interventions that might one day reshape the tumor’s microbial residents as part of cancer treatment. As the science of intratumoral microbiomes matures, gastric cancer is emerging as a prime example of how deeply intertwined microbes and malignancy can be.

Subject of Research: Intratumoral microbiota dysbiosis and its clinical associations in gastric cancer

Article Title: Integrated 2bRAD-M and 16S rRNA sequencing reveals intratumoral microbiota dysbiosis and clinical associations in gastric cancer

Article References: Zhou, H., Feng, W., Li, Y., Xu, Y., Zong, W., & Huang, Y. (2026). Integrated 2bRAD-M and 16S rRNA sequencing reveals intratumoral microbiota dysbiosis and clinical associations in gastric cancer. Journal of Translational Medicine. https://doi.org/10.1186/s12967-026-09025-w

Image Credits: AI Generated

DOI: 10.1186/s12967-026-09025-w

Keywords: gastric cancer, intratumoral microbiota, 2bRAD-M, 16S rRNA sequencing, dysbiosis, Helicobacter pylori, Lactobacillus, tumor microenvironment, biomarkers, CEA, TNM stage, Journal of Translational Medicine

The assay at the heart of the study targets methylation changes in five genes: SEPTIN9, BCAT1, IKZF1, BCAN and VAV3. DNA methylation is an epigenetic modification in which methyl groups are added to cytosine bases, typically at CpG dinucleotides, altering gene expression without changing the underlying genetic code. Cancer cells acquire highly characteristic methylation patterns, including widespread silencing of tumor-suppressor genes through promoter hypermethylation. Because these patterns are stable, reproducible and often tissue- or tumor-specific, methylated DNA fragments released into the bloodstream by dying tumor cells serve as a molecular fingerprint of malignancy. Circulating cell-free DNA, or cfDNA, carries these marks, and polymerase chain reaction-based assays can be designed to detect them even when tumor-derived fragments represent only a tiny fraction of the total cfDNA pool, which is otherwise dominated by DNA from blood cells.

SEPTIN9 is the most familiar name on the list, having already underpinned a commercially available plasma test for colorectal cancer screening in several countries. BCAT1, IKZF1, BCAN and VAV3 add complementary signals, and the logic of a multi-gene panel is straightforward: no single methylation marker is shed by every tumor, so combining several independent targets increases the chance that at least one will be detectable in a given patient. The five-gene assay evaluated here converts methylation measurements across these targets into a single composite result, classified as positive or negative, which was then benchmarked against pathology, the gold standard that definitively establishes whether a patient has cancer after tissue examination.

The study population came from a single colorectal surgery department, a design choice with important implications for how the results should be read. All 412 patients had plasma cfDNA methylation testing and routine serum tumor-marker evaluation performed either before surgery or before starting treatment, and each case was adjudicated by pathology. The final analysis compared 219 patients with confirmed colorectal cancer against 193 patients whose diagnoses were not colorectal cancer, a mix that reflects the genuine clinical question a hospital faces: distinguishing malignant from non-malignant disease among symptomatic patients referred to a surgical service. This is a harder and more realistic test than comparing cancer patients against healthy volunteers, because the control group includes people with polyps, inflammatory conditions and other colorectal pathologies that can confound biomarker performance.

The headline numbers tell a consistent story. The methylation assay achieved a sensitivity of 73.5 percent, meaning it correctly flagged roughly three out of four cancers, with a 95 percent confidence interval running from 67.7 to 79.4 percent. Its specificity of 91.2 percent, with a confidence interval of 87.2 to 95.2 percent, indicates that false positives were relatively uncommon. The area under the receiver operating characteristic curve, a summary of overall discriminatory ability that ranges from 0.5 for a useless test to 1.0 for a perfect one, reached 0.873. CEA, by contrast, detected only 35.9 percent of cancers, though its specificity was higher at 96.7 percent, yielding an AUC of 0.730. The difference in AUC between the two tests was statistically significant, with a p-value below 0.001 on the DeLong test, a standard method for comparing correlated receiver operating characteristic curves.

The researchers also explored whether combining the two markers would help. An either-positive rule, in which a patient is considered positive if either the methylation assay or CEA is positive, pushed sensitivity up to 79.9 percent but dragged specificity down to 88.6 percent, producing an AUC of 0.843. That trade-off illustrates a fundamental principle of diagnostic test design: combining tests with independent signals raises the chance of catching true cases but simultaneously accumulates false positives from each component. For a test intended as an adjunctive risk-stratification tool rather than a standalone diagnosis, the modest gain in sensitivity may or may not justify the additional false positives, a calculation that depends on the clinical setting, the cost of follow-up colonoscopy and the prevalence of disease in the tested population.

Perhaps the most biologically revealing results concern how performance scaled with tumor burden. Sensitivity climbed from 56.8 percent in stage I disease to 100 percent in stage IV, from 38.1 percent in T1 tumors to 94.1 percent in T4 tumors, and from 53.6 percent in tumors measuring two centimeters or less to 95.0 percent in tumors larger than six centimeters. This gradient is exactly what the biology of circulating tumor DNA predicts. Early tumors are small, have limited cell turnover and relatively poor blood supply at their margins, so they shed comparatively little DNA into circulation. Advanced tumors, with larger mass, greater necrosis and more active remodeling of the tumor microenvironment, release far more methylated DNA fragments, making them easier to detect. The pattern is a double-edged sword: it confirms the assay is genuinely measuring tumor-derived material, but it also means the test is weakest precisely where a screening tool would deliver the greatest benefit, at the earliest and most curable stages.

This stage dependence is not a flaw unique to this assay; it is an inherent constraint of liquid biopsy technology that the field has grappled with since its inception. Detecting a handful of tumor DNA fragments among the vast background of healthy cfDNA is an analytical challenge comparable to finding a needle in a haystack, and the needle count rises with tumor size and stage. Strategies to improve early detection include deeper sequencing, larger panels of methylation targets, optimized DNA extraction and amplification chemistry, and multimodal algorithms that integrate methylation with other features such as fragmentomics and mutation profiling. The authors of the current study are careful to frame their five-gene assay as a complement to existing modalities, including colonoscopy and established screening programs, rather than a replacement for them.

The real-world design of the study is both its greatest strength and its principal limitation. By testing the assay in a consecutive clinical cohort from a single surgical department, the researchers captured performance under the messier conditions of actual practice, where patient characteristics, sample handling and comorbidities deviate from the idealized settings of model-development studies. Yet the single-center, single-department, retrospective nature of the work means the findings cannot be generalized wholesale. A cohort drawn from a colorectal surgery clinic overrepresents symptomatic patients and advanced disease, so sensitivity in an asymptomatic screening population would likely be lower than the figures reported here. Specificity, too, depends on the composition of the control group, and a healthier population might shift the numbers in either direction. The authors themselves emphasize that large-scale, multi-center prospective studies are needed to define the assay’s utility across diverse screening and diagnostic settings before it can be recommended for broad clinical deployment.

Even with those caveats, the study adds meaningful weight to the case for methylation-based liquid biopsy in colorectal cancer, a disease that remains one of the most common and deadly malignancies worldwide despite being highly curable when caught early. A blood test that reliably outperforms CEA, a marker that has been in clinical use for decades yet misses nearly two-thirds of cancers in this cohort, represents a tangible step toward less invasive and more accessible detection pathways. The five-gene panel’s 0.873 AUC in a real-world surgical cohort, achieved with a technically simple PCR-based workflow on routine plasma samples, suggests that epigenetic blood testing is maturing from promising concept to clinically credible instrument. The road ahead runs through prospective validation in screening populations, standardization of laboratory protocols and careful health-economic analysis, but the direction of travel is now unmistakable: the chemical marks that tumors stamp on their DNA are becoming a readable, quantifiable and increasingly trustworthy signal in modern cancer medicine.

Subject of Research: Diagnostic performance of a five-gene plasma cell-free DNA methylation assay for colorectal cancer detection compared with CEA

Article Title: Performance of a five-gene plasma cfDNA methylation assay for colorectal cancer detection: a single-center real-world analysis

Article References: Chen, B., Xu, X., Yuan, H., Wan, Z., Du, W., Yu, Q., Qian, Y., Wang, J., Yu, Y., Lyu, J., Yang, Z., & Zheng, B. (2026). Performance of a five-gene plasma cfDNA methylation assay for colorectal cancer detection: a single-center real-world analysis. BMC Cancer. https://doi.org/10.1186/s12885-026-17047-2

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17047-2

Keywords: colorectal cancer, cfDNA methylation, liquid biopsy, SEPTIN9, CEA, early detection, epigenetics, tumor biomarkers, real-world study, BMC Cancer, diagnostic sensitivity, circulating tumor DNA

Radioligand therapy with 177Lu-PSMA-617 has transformed the treatment landscape for metastatic castration-resistant prostate cancer, a lethal stage of the disease in which roughly half of patients survive less than five years after their metastatic diagnosis. The therapy exploits prostate-specific membrane antigen, a protein abundantly displayed on the surface of prostate cancer cells and further upregulated as tumors become castration resistant. A ligand called PSMA-617 delivers the beta-emitting radioisotope lutetium-177 directly to tumor cells, irradiating metastases throughout the body while sparing much of the surrounding healthy tissue. Following the landmark phase III VISION trial, the approach gained approval from the United States Food and Drug Administration in March 2022 for PSMA-positive patients who had already received an androgen receptor pathway inhibitor and taxane chemotherapy.

Yet the therapy has an uncomfortable secret: only about half of eligible patients derive meaningful clinical benefit. Except for PSMA PET imaging criteria used to confirm that tumors carry the target molecule, clinicians currently lack validated biomarkers to identify who will respond and who will progress despite treatment. This gap matters both scientifically and economically. A recent cost-effectiveness analysis based on the VISION trial concluded that adding 177Lu-PSMA to standard care may only be cost-effective under specific clinical and economic assumptions, underscoring the urgency of better patient selection.

To address this problem, investigators at IRCCS Istituto Romagnolo per lo Studio dei Tumori in Meldola, Italy, enrolled 145 patients between April 2017 and October 2022 in a prospective observational biological study nested within a single-arm phase II trial of 177Lu-PSMA-617. Of these, 142 patients received at least one dose and were fully evaluable. Within seven days of the first infusion, the researchers collected baseline blood samples and measured a panel of routinely available laboratory markers, including hemoglobin, alkaline phosphatase, prostate-specific antigen, chromogranin A, neuron-specific enolase, and CEA, along with inflammatory ratios derived from the complete blood count. All patients received treatment in four cycles administered every eight to twelve weeks, using either a lower activity of 3.7 to 4.4 gigabecquerels or a higher activity of 5.5 gigabecquerels per cycle.

At a median follow-up of 44 months, the cohort experienced a median progression-free survival of 7.0 months and a median overall survival of 16.8 months, outcomes broadly consistent with the pivotal VISION and TheraP trials. In univariable analyses, elevated platelet-to-lymphocyte ratio, alkaline phosphatase, PSA, and CEA each predicted shorter progression-free survival, while low hemoglobin and high neutrophil-to-lymphocyte ratio, along with the same markers, predicted shorter overall survival. But when the researchers integrated all clinical characteristics and serum biomarkers into multivariable Cox regression models, only a handful of factors retained independent significance: prior docetaxel exposure, low hemoglobin, high PSA, and, critically, elevated CEA. High CEA independently predicted both shorter progression-free survival, with a p-value of 0.026, and shorter overall survival, with a p-value of 0.003.

The statistical rigor of the analysis strengthens the claim. Multicollinearity checks using variance inflation factors confirmed that no variable simply duplicated another, with all values below two. When CEA was added to a model containing established prognostic factors, the overall model fit improved significantly according to a likelihood-ratio test, and Harrell’s concordance index rose from 0.732 to 0.742. Although PSA alone discriminated outcomes better than CEA alone, with C-indices of 0.652 versus 0.598, combining the two markers pushed the C-index to 0.694, demonstrating that CEA carries prognostic information that PSA does not capture. Exploratory correlation analyses reinforced this independence: baseline CEA showed essentially no relationship with PSA levels and only a weak association with chromogranin A, while it did not correlate with bone disease burden measured on PSMA PET scans.

The biological story behind elevated CEA may be as compelling as the statistics. CEA is an oncofetal glycoprotein long used as a tumor marker in colorectal and other cancers, and previous work has linked it to aggressive disease biology in prostate cancer. In particular, elevated serum CEA in castration-resistant prostate cancer has been associated with liver metastases, markedly shorter survival, and distinctive circulating tumor DNA alterations. It has also been tied to aggressive variant prostate cancer, a molecularly distinct and treatment-resistant phenotype. Although a single-arm design without a control group prevents the authors from claiming that CEA specifically predicts benefit from 177Lu-PSMA rather than simply flagging inherently aggressive disease, the finding raises the hypothesis that high CEA identifies a biologically distinct subgroup with intrinsic resistance to PSMA-targeted radioligand therapy. Notably, baseline FDG PET was not mandatory in this cohort, so the researchers could not test whether CEA-elevated patients harbored FDG-positive, PSMA-low lesions typical of aggressive variants.

Other markers behaved as expected. Lower hemoglobin, likely reflecting chronic inflammation and extensive bone marrow involvement rather than a direct effect on treatment response, remained independently associated with shorter overall survival. Alkaline phosphatase, a proxy for skeletal tumor burden, also independently predicted overall survival, consistent with three prior studies and a recent meta-analysis. Inflammatory ratios lost much of their significance after adjustment, although platelet-to-lymphocyte ratio retained an association with progression-free survival. An intriguing exploratory interaction suggested that the prognostic impact of CEA may differ according to the administered radiation activity, but the authors caution that this post hoc finding is hypothesis-generating only, since dose was not randomly assigned and lower activities were reserved for frailer patients.

The study’s limitations temper, but do not erase, its promise. The administered activities were lower than the currently approved Pluvicto regimen of 7.4 gigabecquerels every six weeks, potentially limiting generalizability, and no metastatic biopsies were available to histologically confirm neuroendocrine differentiation among patients with elevated chromogranin A or neuron-specific enolase, markers of limited specificity that can rise in many benign and malignant conditions. External validation in cohorts treated with the approved dosing schedule will be essential before CEA enters routine risk stratification. Still, the vision is clear: if validated, a marker already present on virtually every hospital laboratory requisition could, at essentially no added cost, help clinicians identify patients facing aggressive, inherently resistant disease, guide them toward alternative or combination strategies, and reserve this costly radioligand therapy for those most likely to benefit. In the precision oncology era, the humble CEA blood draw may prove an unexpectedly powerful ally.

Subject of Research: Baseline serum carcinoembryonic antigen as an independent prognostic biomarker for survival outcomes in metastatic castration-resistant prostate cancer patients treated with 177Lu-PSMA radioligand therapy

Article Title: Baseline serum carcinoembryonic antigen (CEA) is an independent prognostic biomarker for metastatic castration-resistant prostate cancer (mCRPC) patients receiving 177Lu-PSMA therapy

Article References: Giunta, E. F., Marini, I., Grassi, I., Nicolini, S., Foca, F., Celli, M., Caroli, P., Sarnelli, A., Brighi, N., Schepisi, G., Monti, M., De Giorgi, U., Lolli, C., & Sansovini, M. (2026). Baseline serum carcinoembryonic antigen (CEA) is an independent prognostic biomarker for metastatic castration-resistant prostate cancer (mCRPC) patients receiving 177Lu-PSMA therapy. European Journal of Nuclear Medicine and Molecular Imaging. https://doi.org/10.1007/s00259-026-08173-x

Image Credits: AI Generated

DOI: 10.1007/s00259-026-08173-x

Keywords: 177Lu-PSMA, prostate cancer, radioligand therapy, CEA, biomarker, metastatic castration-resistant prostate cancer, prognostic factors, PSMA, overall survival, progression-free survival, PSMA-617, serum tumor markers