Search

Background: Advanced gastro-oesophageal cancer (GEA) treatment has been improved by the introduction of immune checkpoint inhibitors (CPIs), yet identifying predictive biomarkers remains a priority, particularly in patients with a combined positive score (CPS) < 5, where the benefit is less clear. Our study assesses certain immune microenvironment features related to sensitivity or resistance to CPIs with the aim of implementing a personalised approach across CPS < 5 GEA. Design: Through integrative transcriptomic and clinicopathological analyses, we studied in both a retrospective and a prospective cohort, the immune tumour microenvironment features. We analysed the cell types composing the immune infiltrate highlighting their functional activity. Results: This integrative study allowed the identification of four different groups across our patients. Among them, we identified a cluster whose tumours expressed the most gene signatures related to immunomodulatory pathways and immunotherapy response. These tumours presented an enriched immune infiltrate showing high immune function activity that could potentially achieve the best benefit from CPIs. Finally, our findings were proven in an external CPI-exposed population, where the use of our transcriptomic results combined with CPS helped better identify those patients who could benefit from immunotherapy than using CPS alone (p = 0.043). Conclusions: This transcriptomic classification could improve precision immunotherapy for GEA. [ABSTRACT FROM AUTHOR]

Purpose: The aim of this study was to determine the prevalence of distress and unmet supportive care needs in post-treatment colorectal cancer (CRC) survivors. Also, to explore the association between both variables and to identify potential associated sociodemographic and cancer-related risk factors. Methods: A cross-sectional study of 200 CRC survivors who at least 1 month before had completed the primary treatment for CRC was conducted. The Brief Symptom Inventory-18 (BSI-18) and the Spanish version of Cancer Survivors' Unmet Needs (S-CaSUN) were used. Results: One in five CRC survivors showed clinical distress and 86% expressed at least one unmet need. Distress was positively associated with the prevalence of needs in all domains. All comprehensive care and information needs were expressed by at least 20% of survivors and some by more than 50%. Other needs also mentioned by 20% of survivors were financial support, ongoing case manager, and concerns about cancer recurrence. The risk factors associated were lower socioeconomic status, younger age, and a primary treatment that includes more than surgery. Conclusions: The findings highlight the relevance of extending psychosocial care beyond the CRC primary medical treatment. A person-centered approach that addresses informational, emotional, social, and physical needs can increase satisfaction with care and also prevent psychological morbidity in CRC survivors. [ABSTRACT FROM AUTHOR]

Background: Budigalimab is a humanized, recombinant immunoglobulin G1 monoclonal antibody targeting programmed cell death protein 1 (PD-1). We present the safety, efficacy, pharmacokinetic (PK), and pharmacodynamic data from patients enrolled in the head and neck squamous cell carcinoma (HNSCC) and non-small cell lung cancer (NSCLC) expansion cohorts of the phase 1 first-in-human study of budigalimab monotherapy (NCT03000257; registered 15 December 2016). Patients and methods: Patients with recurrent/metastatic HNSCC or locally advanced/metastatic NSCLC naive to PD-1/PD-1-ligand inhibitors were enrolled; patients were not selected on the basis of oncogene driver mutations or PD-L1 status. Budigalimab was administered at 250 mg intravenously Q2W or 500 mg intravenously Q4W until disease progression/unacceptable toxicity. The primary endpoints were safety and PK; the secondary endpoint was efficacy. Exploratory endpoints included biomarker assessments. Results: In total, 81 patients were enrolled (HNSCC: N = 41 [PD-L1 positive: n = 19]; NSCLC: N = 40 [PD-L1 positive: n = 16]); median treatment duration was 72 days (range, 1–617) and 71 days (range, 1–490) for the HNSCC and NSCLC cohorts, respectively. The most frequent grade ≥ 3 treatment-emergent adverse event was anemia (HNSCC: n = 9, 22%; NSCLC: n = 5, 13%). Both dosing regimens had comparable drug exposure and increased interferon gamma-induced chemokines, monokine induced by gamma interferon, and interferon-gamma-inducible protein 10. Objective response rates were 13% (90% CI, 5.1–24.5) in the HNSCC cohort and 19% (90% CI, 9.2–32.6) in the NSCLC cohort. Median progression-free survival was 3.6 months (95% CI, 1.7–4.7) and 1.9 months (95% CI, 1.7–3.7) in the HNSCC and NSCLC cohorts. Conclusions: The safety, efficacy and biomarker profiles of budigalimab are similar to other PD-1 inhibitors. Development of budigalimab in combination with novel anticancer agents is ongoing. [ABSTRACT FROM AUTHOR]

Summary: Background Overcoming resistance to anti-epidermal growth factor receptor (EGFR) monoclonal antibodies (mAbs) in patients with KRAS wildtype (WT) metastatic colorectal cancer (mCRC) could help meet the needs of patients with limited treatment options. Methods In this phase 1b study, patients with N/KRAS WT, MET-positive mCRC who had progressed following anti-EGFR mAb treatment received escalating oral doses of capmatinib (150, 300, and 400 mg) twice daily plus weekly intravenous cetuximab (at the approved dose). The primary objective was to establish a recommended dose for expansion (RDE) of capmatinib in combination with cetuximab. Safety, preliminary activity, pharmacokinetics, and pharmacodynamics were also explored. Results Thirteen patients were enrolled. No patients experienced a dose-limiting toxicity at investigated doses; the RDE was established as capmatinib 400 mg twice daily plus cetuximab. All patients experienced adverse events (AEs) suspected to be related to the study treatment. Five patients (38.5%) reported study-drug–related AEs of grade 3/4 in severity. No patients achieved a complete or partial response according to RECIST v1.1; however, tumor shrinkage of 29–44% was observed in 4 patients. Conclusions Capmatinib plus cetuximab was well tolerated. Preliminary signs of activity were observed. Further investigation is warranted to obtain efficacy data and refine predictive biomarkers of response. Clinical trial registration NCT02205398. [ABSTRACT FROM AUTHOR]

Background: This Phase 1 dose-escalation/expansion study assessed safety/tolerability of sapanisertib, an oral, highly selective inhibitor of mTORC1/mTORC2, in advanced solid tumours.Methods: Eligible patients received increasing sapanisertib doses once daily (QD; 31 patients), once weekly (QW; 30 patients), QD for 3 days on/4 days off QW (QD × 3dQW; 33 patients) or QD for 5 days on/2 days off QW (QD × 5dQW; 22 patients). In expansion cohorts, 82 patients with renal cell carcinoma (RCC), endometrial or bladder cancer received sapanisertib 5 mg QD (39 patients), 40 mg QW (26 patients) or 30 mg QW (17 patients).Results: Maximum tolerated doses of sapanisertib were 6 mg QD, 40 mg QW, 9 mg QD × 3dQW and 7 mg QD × 5dQW. Frequent dose-limiting toxicities (DLTs) included hyperglycaemia, maculo-papular rash (QD), asthenia and stomatitis (QD × 3dQW/QD × 5dQW); expansion phase doses of 5 mg QD and 30 mg QW were selected based on tolerability beyond the DLT evaluation period. One patient with RCC achieved complete response; nine experienced partial responses (RCC: seven patients; carcinoid tumour/endometrial cancer: one patient each). Sapanisertib pharmacokinetics were time-linear and supported multiple dosing. Pharmacodynamic findings demonstrated treatment-related reductions in TORC1/2 biomarkers.Conclusions: Sapanisertib demonstrated a manageable safety profile, with preliminary antitumour activity observed in RCC and endometrial cancer.Clinical Trial Registration: ClinicalTrials.gov, NCT01058707. [ABSTRACT FROM AUTHOR]

Purpose: This open-label, phase Ib, dose-escalation, and dose-expansion study (NCT01862081) evaluated taselisib with a taxane in locally advanced or metastatic breast cancer (BC) and/or non-small cell lung cancer (NSCLC). Methods: Patients received taselisib (2–6 mg tablet or 3–6 mg capsule) plus docetaxel or paclitaxel. Primary endpoints were safety, dose-limiting toxicities, maximum tolerated dose, and identification of a recommended phase II dose. Secondary endpoints included pharmacokinetics and antitumor activity assessment. Results: Eighty patients (BC: 72; NSCLC: 7; BC/NSCLC: 1) were enrolled (docetaxel-receiving arms: 21; paclitaxel-receiving arms: 59). Grade ≥ 3 adverse events (AEs), serious AEs, and AEs leading to death were reported in 90.5%, 42.9%, and 14.3% of patients, respectively (docetaxel-receiving arms), and 78.9%, 40.4%, and 3.5% of patients, respectively (paclitaxel-receiving arms). Eight patients experienced dose-limiting toxicities. The maximum tolerated dose was exceeded with 3 mg taselisib (capsule) for 21 consecutive days plus 75 mg/m2 docetaxel and not exceeded with 6 mg taselisib (tablet) for 5 days on/2 days off plus 80 mg/m2 paclitaxel. Objective response rates and clinical benefit rates were 35.0% and 45.0%, respectively (docetaxel-receiving arms), and 20.4% and 27.8%, respectively (paclitaxel-receiving arms). Exposure for paclitaxel or docetaxel plus taselisib was consistent with the single agents. Conclusions: Taselisib in combination with a taxane has a challenging safety profile. Despite evidence of antitumor activity, the benefit–risk profile was deemed not advantageous. Further development is not planned. [ABSTRACT FROM AUTHOR]

Purpose: To determine the effect achieved from a single dose of anti-VEGF treatment switch, in patients with nAMD previously treated with bevacizumab, switched to either aflibercept or ranibizumab, and to compare the response between aflibercept and ranibizumab. Methods: In retrospective, observational, and comparative study, patients were divided into two groups: Group 1, patients switched to aflibercept; Group 2, patients switched to ranibizumab. Paired samples t test was performed to measure differences in central macular thickness (CMT). To compare whether there were differences between groups mixed-design ANOVA was used. Results: In Group 1, CMT changed from 360.51 to 260.16 µm, presenting a significant mean difference from PreSwitch to PostSwitch of 100.34 µm (p = 0.002, paired samples t test). In Group 2, CMT changed from 366.33 to 260.72 µm, showing a significant difference from PreSwitch to PostSwitch of 105.61 µm (p ≤ 0.000, paired samples t test). The mixed-design ANOVA compared both groups and resulted in a nonsignificant value of 0.90. Conclusion: The effect achieved from a single dose in patients switched to aflibercept or ranibizumab reduced significantly CMT measurements. Comparing aflibercept and ranibizumab, the effect appears to be similar in both drugs, in terms of reduction of CMT. [ABSTRACT FROM AUTHOR]

Background: There is increasing interest in the molecular profiling of tumour tissues in order to investigate alternative breast cancer (BC) therapies. However, the impact of genomic screening for druggable mutations with targeted gene panel sequencing (TGPS) in routine practice remains controversial. Methods: This is a retrospective analysis of data from a genomic screening programme at our institution, in which we performed simplified TGPS for mutations in PIK3CA, AKT1, KRAS, NRAS, and BRAF in order to select patients for targeted therapy clinical trials. The genomes of archived samples of primary (PT) and/or metastatic (MT) tumours from advanced BC patients were analysed with MassARRAY technology (Sequenom MassARRAY, OncoCarta v1.0). The level of PTEN expression was assessed by immunohistochemistry. The primary endpoint was to identify the proportion of BC patients with PI3 K and MAPK alterations who were included in clinical trials using targeted therapies against these pathways. Results: Two hundred and fifteen metastatic BC patients (65 PT and 168 MT) were included. Fifty-two patients (24.19 %) were enrolled in tailored clinical trials, of whom 29 (55.77, 13.49 % of all patients screened) harboured mutations targeted by the study drug. Moreover, 12 wild-type patients out of the 215 (5.58 %) were included in the clinical trials for which mutation analysis was an inclusion criteria. All the patients received drugs targeting the PI3K-AKT pathway and only two were given combinations directed against the PI3K and MAPK pathways. PI3KCA mutations were present in 33.7 % (61/181) of the patients, 45.83 % in PTs and 29.32 % in MTs. AKT1 mutations were detected in 5.48 % (8/146) of patients and PTEN loss in 34.67 % (52/150). KRAS, NRAS, and BRAF mutations were present in 12.06, 5.67, and 3.18 % of patients, respectively. Conclusions: Genomic screening with a simplified TGPS is feasible, and was used to identify 13.49 % of patients who were included in clinical trials using targeted therapy against the mutations they harboured; PI3KCA mutations were the most frequent aberration in our series. [ABSTRACT FROM AUTHOR]

Despite recent diagnostic and therapeutic advances, the survival of patients with gastric cancer is still poor. The majority of patients are diagnosed with advanced disease and chemotherapy represents the only possible therapeutic approach. However, chemotherapy seems to have reached an efficacy plateau in this setting. Gastric cancer is a complex and heterogeneous disease because it emerges from multiple interactions of genetic, environmental and host factors. A better understanding of its molecular characteristics may lead to an improvement of outcomes. The recent molecular classification by The Cancer Genome Atlas project divides gastric cancer into four subtypes that could be taken into consideration in future clinical trials with targeted agents. So far trastuzumab, a monoclonal antibody addressing the HER2 receptor, is the only targeted agent approved in the first-line setting, but only in patients overexpressing HER2. Negative data have been obtained in first-line therapy when antiangiogenics, anti-EGFR or anti-MET monoclonal antibodies have been studied in randomised controlled trials. Ramucirumab, a monoclonal antibody binding to VEGFR2, is the only antiangiogenic agent currently recommended in patients progressing after first-line treatment. In this review, we discuss whether personalised therapy may have a role in gastric cancer. [ABSTRACT FROM AUTHOR]

This two-arm, randomised, multicentre, open-label, phase IIIb study investigated the safety and efficacy of a 3-h catumaxomab infusion with/without prednisolone premedication to reduce catumaxomab-related adverse events. Patients with malignant ascites due to epithelial cancer received four 3-h intraperitoneal catumaxomab infusions with/without intravenous prednisolone (25 mg) premedication before each infusion. The primary safety endpoint was a composite safety score calculated from the incidence and intensity of the most frequent catumaxomab-related adverse events (pyrexia, nausea, vomiting and abdominal pain). Puncture-free survival (PuFS) was a co-primary endpoint. Time to next puncture (TTPu) and overall survival (OS) were secondary endpoints. Prednisolone premedication did not result in a significant reduction in the main catumaxomab-related adverse events. The mean composite safety score was comparable in both arms (catumaxomab plus prednisolone, 4.1; catumaxomab, 3.8; p = 0.383). Median PuFS (30 vs. 37 days) and TTPu (78 vs. 102 days) were shorter in the catumaxomab plus prednisolone arm than in the catumaxomab arm, but the differences were not statistically significant ( p = 0.402 and 0.599, respectively). Median OS was longer in the catumaxomab plus prednisolone arm than in the catumaxomab arm (124 vs. 86 days), but the difference was not statistically significant ( p = 0.186). The superiority of catumaxomab plus prednisolone versus catumaxomab alone could not be proven for the primary endpoint. Prednisolone did not result in a significant reduction in the main catumaxomab-related adverse events. The study confirms the safety and efficacy of catumaxomab administered as four 3-h intraperitoneal infusions for the treatment of malignant ascites. [ABSTRACT FROM AUTHOR]

Oesophageal cancer is currently the eighth most common malignancy worldwide. The incidence varies enormously between geographic areas, with increased incidences (>100/100,000) in Asia, and central and south Africa. The incidence of oesophageal cancer in Spain has been approximately 8/100,000 among men and 1/100,000 among women. Oesophageal cancers are histologically classified as squamous cell carcinoma (SCC) or adenocarcinoma, both with different associated risk factors. While SCCs have become increasingly less common, the incidence of adenocarcinoma has risen steeply. In the SEOM Clinical Guidelines for Oesophageal Cancer we propose rules for adequate initial diagnosis and staging as well as treatment recommendations from early to advanced disease. [ABSTRACT FROM AUTHOR]

Metastatic colorectal cancer (CRC) represents an important health problem in which several biological predictive and prognostic factors have been identified, including clinical features and molecular markers that might influence the response to treatment. Actually, certain prognostic factors are considered key elements, along with disease extent, for deciding the therapeutic approach. However, a distinction between resectable/potentially resectable and unresectable patients must be made in order to establish an adequate therapeutic strategy. Different drugs and chemotherapy regimens are currently available, and their administration depends on patient characteristics, disease-related factors and the treatment objective. Moreover, special situations such as peritoneal carcinomatosis and local treatment of CRC in the setting of metastatic disease should be considered when deciding the most appropriate treatment strategy. This article reviews all the previously mentioned issues involved in the management of metastatic CRC and suggests some general recommendations for its treatment. [ABSTRACT FROM AUTHOR]

The present review gives a perspective on the Aurora kinase family members, their function in normal cells, their role in cancer progression as well as their potential as target for anticancer treatment. Mitosis has been an important target for anticancer therapy development, leading to some specific drugs mainly addressing Tubulines, as a key structure of the mitotic spindle. Vinca alkaloids, taxanes or epotilones are good examples of conventionally developed antimitotic agents. However, novel classes of antineoplastic drugs are being studied, targeting the regulatory system that controls functional aspects of mitosis, such as Aurora or Polo-like kinases or Kinespondin inhibitors. The specific role of the different Aurora kinase proteins as regulator enzymes of the mitotic process in normal cells is discussed. Some of the mechanisms that link Aurora overexpression with cancer are also considered. Thereafter, the clinical and preclinical development of the different Aurora kinase inhibitors is presented. This is nowadays a very active area of therapeutic research and at least, sixteen new compounds are being studied as potential antineoplastic drugs. Most of them are in a very early phase of clinical development. However, we summarized the most recently published findings related with these drugs: main characteristics, way of administration, dose limiting toxicities and recommended doses for further studies. Another important aspect in Aurora kinase inhibition is the study and validation of potential biomarkers to optimize the clinical development. Several studies included pharmacodynamic assessments in normal blood cells, skin or/and tumor biopsies. Several proposals included a higher mitotic index, a decreased number of mitosis with bipolar spindles or normal alignment of chromosomes and inhibition of histone H3 phosphorylation. Future strategies and challenges for trials with Aurora kinase inhibitors are also discussed. [ABSTRACT FROM AUTHOR]

Copyright of Clinical & Translational Oncology is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Most of colorectal cancer CRC-related death is due to metastasis and the finding of markers for prognosis of invasiveness, constitutes an appealing challenge. Here, after analysing cDNA array containing 43 tumour and 5 normal mucosa samples, we report that the expression of the ZNF518B gene as a whole and that of its two major splicing isoforms are significantly increased in tumours. The canonical isoform was also up-regulated in a patients' cohort containing 70 tumour and 69 adjacent tissue samples. The effects of silencing ZNF518B on the phenotype of CRC cell lines were then studied. The gene does not affect cell proliferation, but plays a significant role in cell migration and invasiveness and induces changes in the epithelial-to-mesenchymal transition markers, suggesting that ZNF518B favours tumour cell dissemination. To study the regulation of the gene, transcription-related changes in nucleosomal organisation and epigenetic marks around the transcriptional start site were analysed. The positioning of a nucleosome over the transcription start site and the differential presence of the epigenetic marks H3K9ac, H3K27ac, H3K4me3 and H3K9me3 correlate with gene expression. Inhibition of histone deacetylases increases the transcription of ZNF518B, which may be a candidate for invasiveness prognosis in CRC and a target for epigenetic drugs. [ABSTRACT FROM AUTHOR]

Background: Definitive chemoradiation is the primary treatment for locally advanced head and neck carcinoma (LAHNSCC). Optimising outcome predictions requires validated biomarkers, since TNM8 and HPV could have limitations. Radiomics may enhance risk stratification.This single-centre observational study collected clinical data and baseline CT scans from 171 LAHNSCC patients treated with chemoradiation. The dataset was divided into training (80%) and test (20%) sets, with a 5-fold cross-validation on the training set. Researchers extracted 108 radiomics features from each primary tumour and applied survival analysis and classification models to predict progression-free survival (PFS) and 5-year progression, respectively. Performance was evaluated using inverse probability of censoring weights and c-index for the PFS model and AUC, sensitivity, specificity, and accuracy for the 5-year progression model. Feature importance was measured by the SHapley Additive exPlanations (SHAP) method and patient stratification was assessed through Kaplan–Meier curves.The final dataset included 171 LAHNSCC patients, with 53% experiencing disease progression at 5 years. The random survival forest model best predicted PFS, with an AUC of 0.64 and CI of 0.66 on the test set, highlighting 4 radiomics features and TNM8 as significant contributors. It successfully stratified patients into low and high-risk groups (log-rank p < 0.005). The extreme gradient boosting model most effectively predicted a 5-year progression, incorporating 12 radiomics features and four clinical variables, achieving an AUC of 0.74, sensitivity of 0.53, specificity of 0.81, and accuracy of 0.66 on the test set.The combined clinical-radiomics model improved the standard TNM8 and clinical variables in predicting 5-year progression though further validation is necessary.QuestionThere is an unmet need for non-invasive biomarkers to guide treatment in locally advanced head and neck cancer.FindingsClinical data (TNM8 staging, primary tumour site, age, and smoking) plus radiomics improved 5-year progression prediction compared with the clinical comprehensive model or TNM staging alone.Clinical relevanceSHAP simplifies complex machine learning radiomics models for clinicians by using easy-to-understand graphical representations, promoting explainability.Methods: Definitive chemoradiation is the primary treatment for locally advanced head and neck carcinoma (LAHNSCC). Optimising outcome predictions requires validated biomarkers, since TNM8 and HPV could have limitations. Radiomics may enhance risk stratification.This single-centre observational study collected clinical data and baseline CT scans from 171 LAHNSCC patients treated with chemoradiation. The dataset was divided into training (80%) and test (20%) sets, with a 5-fold cross-validation on the training set. Researchers extracted 108 radiomics features from each primary tumour and applied survival analysis and classification models to predict progression-free survival (PFS) and 5-year progression, respectively. Performance was evaluated using inverse probability of censoring weights and c-index for the PFS model and AUC, sensitivity, specificity, and accuracy for the 5-year progression model. Feature importance was measured by the SHapley Additive exPlanations (SHAP) method and patient stratification was assessed through Kaplan–Meier curves.The final dataset included 171 LAHNSCC patients, with 53% experiencing disease progression at 5 years. The random survival forest model best predicted PFS, with an AUC of 0.64 and CI of 0.66 on the test set, highlighting 4 radiomics features and TNM8 as significant contributors. It successfully stratified patients into low and high-risk groups (log-rank p < 0.005). The extreme gradient boosting model most effectively predicted a 5-year progression, incorporating 12 radiomics features and four clinical variables, achieving an AUC of 0.74, sensitivity of 0.53, specificity of 0.81, and accuracy of 0.66 on the test set.The combined clinical-radiomics model improved the standard TNM8 and clinical variables in predicting 5-year progression though further validation is necessary.QuestionThere is an unmet need for non-invasive biomarkers to guide treatment in locally advanced head and neck cancer.FindingsClinical data (TNM8 staging, primary tumour site, age, and smoking) plus radiomics improved 5-year progression prediction compared with the clinical comprehensive model or TNM staging alone.Clinical relevanceSHAP simplifies complex machine learning radiomics models for clinicians by using easy-to-understand graphical representations, promoting explainability.Results: Definitive chemoradiation is the primary treatment for locally advanced head and neck carcinoma (LAHNSCC). Optimising outcome predictions requires validated biomarkers, since TNM8 and HPV could have limitations. Radiomics may enhance risk stratification.This single-centre observational study collected clinical data and baseline CT scans from 171 LAHNSCC patients treated with chemoradiation. The dataset was divided into training (80%) and test (20%) sets, with a 5-fold cross-validation on the training set. Researchers extracted 108 radiomics features from each primary tumour and applied survival analysis and classification models to predict progression-free survival (PFS) and 5-year progression, respectively. Performance was evaluated using inverse probability of censoring weights and c-index for the PFS model and AUC, sensitivity, specificity, and accuracy for the 5-year progression model. Feature importance was measured by the SHapley Additive exPlanations (SHAP) method and patient stratification was assessed through Kaplan–Meier curves.The final dataset included 171 LAHNSCC patients, with 53% experiencing disease progression at 5 years. The random survival forest model best predicted PFS, with an AUC of 0.64 and CI of 0.66 on the test set, highlighting 4 radiomics features and TNM8 as significant contributors. It successfully stratified patients into low and high-risk groups (log-rank p < 0.005). The extreme gradient boosting model most effectively predicted a 5-year progression, incorporating 12 radiomics features and four clinical variables, achieving an AUC of 0.74, sensitivity of 0.53, specificity of 0.81, and accuracy of 0.66 on the test set.The combined clinical-radiomics model improved the standard TNM8 and clinical variables in predicting 5-year progression though further validation is necessary.QuestionThere is an unmet need for non-invasive biomarkers to guide treatment in locally advanced head and neck cancer.FindingsClinical data (TNM8 staging, primary tumour site, age, and smoking) plus radiomics improved 5-year progression prediction compared with the clinical comprehensive model or TNM staging alone.Clinical relevanceSHAP simplifies complex machine learning radiomics models for clinicians by using easy-to-understand graphical representations, promoting explainability.Conclusion: Definitive chemoradiation is the primary treatment for locally advanced head and neck carcinoma (LAHNSCC). Optimising outcome predictions requires validated biomarkers, since TNM8 and HPV could have limitations. Radiomics may enhance risk stratification.This single-centre observational study collected clinical data and baseline CT scans from 171 LAHNSCC patients treated with chemoradiation. The dataset was divided into training (80%) and test (20%) sets, with a 5-fold cross-validation on the training set. Researchers extracted 108 radiomics features from each primary tumour and applied survival analysis and classification models to predict progression-free survival (PFS) and 5-year progression, respectively. Performance was evaluated using inverse probability of censoring weights and c-index for the PFS model and AUC, sensitivity, specificity, and accuracy for the 5-year progression model. Feature importance was measured by the SHapley Additive exPlanations (SHAP) method and patient stratification was assessed through Kaplan–Meier curves.The final dataset included 171 LAHNSCC patients, with 53% experiencing disease progression at 5 years. The random survival forest model best predicted PFS, with an AUC of 0.64 and CI of 0.66 on the test set, highlighting 4 radiomics features and TNM8 as significant contributors. It successfully stratified patients into low and high-risk groups (log-rank p < 0.005). The extreme gradient boosting model most effectively predicted a 5-year progression, incorporating 12 radiomics features and four clinical variables, achieving an AUC of 0.74, sensitivity of 0.53, specificity of 0.81, and accuracy of 0.66 on the test set.The combined clinical-radiomics model improved the standard TNM8 and clinical variables in predicting 5-year progression though further validation is necessary.QuestionThere is an unmet need for non-invasive biomarkers to guide treatment in locally advanced head and neck cancer.FindingsClinical data (TNM8 staging, primary tumour site, age, and smoking) plus radiomics improved 5-year progression prediction compared with the clinical comprehensive model or TNM staging alone.Clinical relevanceSHAP simplifies complex machine learning radiomics models for clinicians by using easy-to-understand graphical representations, promoting explainability.Key Points: Definitive chemoradiation is the primary treatment for locally advanced head and neck carcinoma (LAHNSCC). Optimising outcome predictions requires validated biomarkers, since TNM8 and HPV could have limitations. Radiomics may enhance risk stratification.This single-centre observational study collected clinical data and baseline CT scans from 171 LAHNSCC patients treated with chemoradiation. The dataset was divided into training (80%) and test (20%) sets, with a 5-fold cross-validation on the training set. Researchers extracted 108 radiomics features from each primary tumour and applied survival analysis and classification models to predict progression-free survival (PFS) and 5-year progression, respectively. Performance was evaluated using inverse probability of censoring weights and c-index for the PFS model and AUC, sensitivity, specificity, and accuracy for the 5-year progression model. Feature importance was measured by the SHapley Additive exPlanations (SHAP) method and patient stratification was assessed through Kaplan–Meier curves.The final dataset included 171 LAHNSCC patients, with 53% experiencing disease progression at 5 years. The random survival forest model best predicted PFS, with an AUC of 0.64 and CI of 0.66 on the test set, highlighting 4 radiomics features and TNM8 as significant contributors. It successfully stratified patients into low and high-risk groups (log-rank p < 0.005). The extreme gradient boosting model most effectively predicted a 5-year progression, incorporating 12 radiomics features and four clinical variables, achieving an AUC of 0.74, sensitivity of 0.53, specificity of 0.81, and accuracy of 0.66 on the test set.The combined clinical-radiomics model improved the standard TNM8 and clinical variables in predicting 5-year progression though further validation is necessary.QuestionThere is an unmet need for non-invasive biomarkers to guide treatment in locally advanced head and neck cancer.FindingsClinical data (TNM8 staging, primary tumour site, age, and smoking) plus radiomics improved 5-year progression prediction compared with the clinical comprehensive model or TNM staging alone.Clinical relevanceSHAP simplifies complex machine learning radiomics models for clinicians by using easy-to-understand graphical representations, promoting explainability. [ABSTRACT FROM AUTHOR]