Machine learning for prediction of immunotherapeutic outcome in non-small-cell lung cancer based on circulating cytokine signatures

Characterization of pre- and on-treatment soluble immune mediators and the tumor microenvironment in NSCLC patients receiving PD-1/L1 inhibitor monotherapy

BACKGROUND

Despite the favorable therapeutic efficacy observed with ICI monotherapy, the majority of non-small cell lung cancer (NSCLC) patients do not respond. Therefore, identifying patients who could optimally benefit from ICI treatment remains a challenge.

METHODS

Among 183 patients with advanced or recurrent NSCLC who received ICI monotherapy, we analyzed 110 patients whose pre- and post-treatment plasma samples were available. Seventy-three soluble immune mediators were measured at ICI initiation and 6 weeks later. To identify useful biomarkers, we analyzed the association of pre-treatment levels and on-treatment changes of soluble immune mediators with survival of patients. The associations of pre-treatment or on-treatment biomarkers with irAE development, PD-L1 expression, CD8+ TIL density, and neutrophil to lymphocyte ratio (NLR) were also analyzed.

RESULTS

Univariate analysis showed that pre-treatment biomarkers included 6 immune mediators, whereas on-treatment biomarkers included 8 immune mediators. Multivariate analysis showed that pre-treatment biomarkers included 4 immune mediators (CCL19, CCL21, CXCL5, CXCL10), whereas on-treatment biomarkers included 5 immune mediators (CCL7, CCL19, CCL23, CCL25, IL-32). IrAE development was associated with on-treatment change in CCL23. PD-L1 expression was associated with the pre-treatment levels of TNFSF13B and the on-treatment change in CCL25. CD8+ TIL density was associated with the pre-treatment CXCL10 level, whereas NLR was correlated with pre-treatment levels of CCL13 and CCL17.

CONCLUSION

We identified several soluble immune mediators as pre-treatment and on-treatment biomarkers of survival in patients with NSCLC treated with ICI monotherapy. Some of these biomarkers were associated with other possible predictors, including irAE development, PD-L1 expression, CD8+ TIL density and NLR. Further large-scale studies are needed to establish biomarkers for patients with NSCLC who received ICI monotherapy.

Construction of a risk prediction model for lung infection after chemotherapy in lung cancer patients based on the machine learning algorithm

Purpose

The objective of this study was to create and validate a machine learning (ML)-based model for predicting the likelihood of lung infections following chemotherapy in patients with lung cancer.

Methods

A retrospective study was conducted on a cohort of 502 lung cancer patients undergoing chemotherapy. Data on age, Body Mass Index (BMI), underlying disease, chemotherapy cycle, number of hospitalizations, and various blood test results were collected from medical records. We used the Synthetic Minority Oversampling Technique (SMOTE) to handle unbalanced data. Feature screening was performed using the Boruta algorithm and The Least Absolute Shrinkage and Selection Operator (LASSO). Subsequently, six ML algorithms, namely Logistic Regression (LR), Random Forest (RF), Gaussian Naive Bayes (GNB), Multi-layer Perceptron (MLP), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN) were employed to train and develop an ML model using a 10-fold cross-validation methodology. The model's performance was evaluated through various metrics, including the area under the receiver operating characteristic curve (ROC), accuracy, sensitivity, specificity, F1 score, calibration curve, decision curves, clinical impact curve, and confusion matrix. In addition, model interpretation was performed by the Shapley Additive Explanations (SHAP) analysis to clarify the importance of each feature of the model and its decision basis. Finally, we constructed nomograms to make the predictive model results more readable.

Results

The integration of Boruta and LASSO methodologies identified Gender, Smoke, Drink, Chemotherapy cycles, pleural effusion (PE), Neutrophil-lymphocyte count ratio (NLR), Neutrophil-monocyte count ratio (NMR), Lymphocytes (LYM) and Neutrophil (NEUT) as significant predictors. The LR model demonstrated superior performance compared to alternative ML algorithms, achieving an accuracy of 81.80%, a sensitivity of 81.1%, a specificity of 82.5%, an F1 score of 81.6%, and an AUC of 0.888(95%CI(0.863-0.911)). Furthermore, the SHAP method identified Chemotherapy cycles and Smoke as the primary decision factors influencing the ML model's predictions. Finally, this study successfully constructed interactive nomograms and dynamic nomograms.

Conclusion

The ML algorithm, combining demographic and clinical factors, accurately predicted post-chemotherapy lung infections in cancer patients. The LR model performed well, potentially improving early detection and treatment in clinical practice.

Machine learning models for predicting of PD-1 treatment efficacy in Pan-cancer patients based on routine hematologic and biochemical parameters

Immune checkpoint blockade therapy targeting the programmed death-1(PD-1) pathway has shown remarkable efficacy and durable response in patients with various cancer types. Early prediction of therapeutic efficacy is important for optimizing treatment plans and avoiding potential side effects. In this work, we developed an efficient machine learning prediction method using routine hematologic and biochemical parameters to predict the efficacy of PD-1 combination treatment in Pan-Cancer patients. A total of 431 patients with nasopharyngeal carcinoma, esophageal cancer and lung cancer who underwent PD-1 checkpoint inhibitor combination therapy were included in this study. Patients were divided into two groups: progressive disease (PD) and disease control (DC) groups. Hematologic and biochemical parameters were collected before and at the third week of PD-1 therapy. Six machine learning models were developed and trained to predict the efficacy of PD-1 combination therapy at 8-12 weeks. Analysis of 57 blood biomarkers before and after three weeks of PD-1 combination therapy through statistical analysis, heatmaps, and principal component analysis did not accurately predict treatment outcome. However, with machine learning models, both the AdaBoost classifier and GBDT demonstrated high levels of prediction efficiency, with clinically acceptable AUC values exceeding 0.7. The AdaBoost classifier exhibited the highest performance among the 6 machine learning models, with a sensitivity of 0.85 and a specificity of 0.79. Our study demonstrated the potential of machine learning to predict the efficacy of PD-1 combination therapy based on changes in hematologic and biochemical parameters.

Differences in systemic immune parameters in individuals with lung cancer according to race

Introduction

Racial and ethnic disparities in the presentation and outcomes of lung cancer are widely known. To evaluate potential factors contributing to these observations, we measured systemic immune parameters in Black and White patients with lung cancer.

Methods

Patients scheduled to receive cancer immunotherapy were enrolled in a multi-institutional prospective biospecimen collection registry. Clinical and demographic information were obtained from electronic medical records. Pre-treatment peripheral blood samples were collected and analyzed for cytokines using a multiplex panel and for immune cell populations using mass cytometry. Differences between Black and White patients were determined and corrected for multiple comparisons.

Results

A total of 187 patients with non-small cell lung cancer (Black, 19; White, 168) were included in the analysis. There were no significant differences in baseline characteristics between Black and White patients. Compared to White patients, Black patients had significantly lower levels of CCL23 and CCL27, and significantly higher levels of CCL8, CXCL1, CCL26, CCL25, CCL1, IL-1 b, CXCL16, and IFN-γ (all P <0.05, FDR<0.1). Black patients also exhibited greater populations of non-classical CD16+ monocytes, NKT-like cells, CD4+ cells, CD38+ monocytes, and CD57+ gamma delta T cells (all P <0.05).

Conclusions

Black and White patients with lung cancer exhibit several differences in immune parameters, with Black patients exhibiting greater levels of numerous pro-inflammatory cytokines and cell populations. The etiology and clinical significance of these differences warrant further evaluation.

CASCADE: Context-Aware Data-Driven AI for Streamlined Multidisciplinary Tumor Board Recommendations in Oncology

This study addresses the potential of machine learning in predicting treatment recommendations for patients with hepatocellular carcinoma (HCC). Using an IRB-approved retrospective study of patients discussed at a multidisciplinary tumor board, clinical and imaging variables were extracted and used in a gradient-boosting machine learning algorithm, XGBoost. The algorithm's performance was assessed using confusion matrix metrics and the area under the Receiver Operating Characteristics (ROC) curve. The study included 140 patients (mean age 67.7 ± 8.9 years), and the algorithm was found to be predictive of all eight treatment recommendations made by the board. The model's predictions were more accurate than those based on published therapeutic guidelines by ESMO and NCCN. The study concludes that a machine learning model incorporating clinical and imaging variables can predict treatment recommendations made by an expert multidisciplinary tumor board, potentially aiding clinical decision-making in settings lacking subspecialty expertise.

Artificial Intelligence and Machine Learning in Predicting the Response to Immunotherapy in Non-small Cell Lung Carcinoma: A Systematic Review

Non-small cell lung carcinoma (NSCLC) is a prevalent and aggressive form of lung cancer, with a poor prognosis for metastatic disease. Immunotherapy, particularly immune checkpoint inhibitors (ICIs), has revolutionized the management of NSCLC, but response rates are highly variable. Identifying reliable predictive biomarkers is crucial to optimize patient selection and treatment outcomes. This systematic review aimed to evaluate the current state of artificial intelligence (AI) and machine learning (ML) applications in predicting the response to immunotherapy in NSCLC. A comprehensive literature search identified 19 studies that met the inclusion criteria. The studies employed diverse AI/ML techniques, including deep learning, artificial neural networks, support vector machines, and gradient boosting methods, applied to various data modalities such as medical imaging, genomic data, clinical variables, and immunohistochemical markers. Several studies demonstrated the ability of AI/ML models to accurately predict immunotherapy response, progression-free survival, and overall survival in NSCLC patients. However, challenges remain in data availability, quality, and interpretability of these models. Efforts have been made to develop interpretable AI/ML techniques, but further research is needed to improve transparency and explainability. Additionally, translating AI/ML models from research settings to clinical practice poses challenges related to regulatory approval, data privacy, and integration into existing healthcare systems. Nonetheless, the successful implementation of AI/ML models could enable personalized treatment strategies, improve treatment outcomes, and reduce unnecessary toxicities and healthcare costs associated with ineffective treatments.

Rapidly Evolving Pre- and Post-surgical Systemic Treatment of Melanoma

With the development of effective BRAF-targeted and immune-checkpoint immunotherapies for metastatic melanoma, clinical trials are moving these treatments into earlier adjuvant and perioperative settings. BRAF-targeted therapy is a standard of care in resected stage III-IV melanoma, while anti-programmed death-1 (PD1) immunotherapy is now a standard of care option in resected stage IIB through IV disease. With both modalities, recurrence-free survival and distant-metastasis-free survival are improved by a relative 35-50%, yet no improvement in overall survival has been demonstrated. Neoadjuvant anti-PD1 therapy improves event-free survival by approximately an absolute 23%, although improvements in overall survival have yet to be demonstrated. Understanding which patients are most likely to recur and which are most likely to benefit from treatment is now the highest priority question in the field. Biomarker analyses, such as gene expression profiling of the primary lesion and circulating DNA, are preliminarily exciting as potential biomarkers, though each has drawbacks. As in the setting of metastatic disease, markers that inform positive outcomes include interferon-γ gene expression, PD-L1, and high tumor mutational burden, while negative predictors of outcome include circulating factors such as lactate dehydrogenase, interleukin-8, and C-reactive protein. Integrating and validating these markers into clinically relevant models is thus a high priority. Melanoma therapeutics continues to advance with combination adjuvant approaches now investigating anti-PD1 with lymphocyte activation gene 3 (LAG3), T-cell immunoreceptor with Ig and ITIM domains (TIGIT), and individualized neoantigen therapies. How this progress will be integrated into the management of a unique patient to reduce recurrence, limit toxicity, and avoid over-treatment will dominate clinical research and patient care over the next decade.

Blood-based biomarkers in patients with non-small cell lung cancer treated with immune checkpoint blockade

The paradigm of non-small cell lung cancer (NSCLC) treatment has been profoundly influenced by the development of immune checkpoint inhibitors (ICI), but the range of clinical responses observed among patients poses significant challenges. To date, analyses of tumor biopsies are the only parameter used to guide prognosis to ICI therapy. Tumor biopsies, however, are often difficult to obtain and tissue-based biomarkers are limited by intratumoral heterogeneity and temporal variability. In response, there has been a growing emphasis on the development of "liquid biopsy"‒ derived biomarkers, which offer a minimally invasive means to dynamically monitor the immune status of NSCLC patients either before and/or during the course of treatment. Here we review studies in which multiple blood-based biomarkers encompassing circulating soluble analytes, immune cell subsets, circulating tumor DNA, blood-based tumor mutational burden, and circulating tumor cells have shown promising associations with the clinical response of NSCLC patients to ICI therapy. These investigations have unveiled compelling correlations between the peripheral immune status of patients both before and during ICI therapy and patient outcomes, which include response rates, progression-free survival, and overall survival. There is need for rigorous validation and standardization of these blood-based assays for broader clinical application. Integration of multiple blood-based biomarkers into comprehensive panels or algorithms also has the potential to enhance predictive accuracy. Further research aimed at longitudinal monitoring of circulating biomarkers is also crucial to comprehend immune dynamics and resistance mechanisms and should be used alongside tissue-based methods that interrogate the tumor microenvironment to guide treatment decisions and may inform on the development of novel therapeutic strategies. The data reviewed here reinforce the opportunity to refine patient stratification, optimize treatments, and improve outcomes not only in NSCLC but also in the wider spectrum of solid tumors undergoing immunotherapy.

Small extracellular vesicles in plasma carry luminal cytokines that remain undetectable by antibody-based assays in cancer patients and healthy donors

Background

Small (30-150nm) extracellular vesicles (sEV), also known as exosomes, play a key role in cell-to-cell signaling. They are produced by all cells, circulate freely and are present in all body fluids. Evidence indicates that cytokines are present on the surface and/or in the lumen of sEV. The contribution of intravesicular cytokines to cytokine levels in plasma are unknown.

Methods

sEV were isolated by ultrafiltration/size exclusion chromatography from pre-cleared plasma obtained from patients with head and neck squamous cell carcinoma (HNSCC) and healthy donors (HDs). Multiplex immunoassays were used to measure cytokine levels in paired untreated and detergent-treated (0.5% Triton X-100) plasma and plasma-derived detergent-treated sEV. Non-parametric tests were used to assess differences in cytokine levels.

Results

The presence of cytokines in sEV isolated from patients' and HDs' plasma was confirmed by immunoblots and on-bead flow cytometry. sEV-associated cytokines were functional in various in vitro assays. Levels of cytokines in sEV varied among the HNSCC patients and were generally significantly higher than the levels observed in sEV from HDs. Compared to untreated plasma, levels for the majority (40/51) of the evaluated proteins were significantly higher in detergent-treated plasma (P<0.0001-0.03). In addition, levels of 24/51 proteins in sEV, including IL6, TNFRII, IL-17a, IFNa and IFNg, were significantly positively correlated with the difference between levels detected in detergent-treated plasma and untreated plasma.

Discussion

The data indicate that sEV-associated cytokines account for the differences in cytokine levels measured in detergent-treated versus untreated plasma. Ab-based assays using untreated plasma detect only soluble cytokines and miss cytokines carried in the lumen of sEV. Permeabilization of sEV with a mild detergent allows for Ab-based detection of sEV-associated and soluble cytokines in plasma. The failure to detect cytokines carried in the sEV lumen leads to inaccurate estimates of cytokine levels in body fluids.

Prediction of Effectiveness and Toxicities of Immune Checkpoint Inhibitors Using Real-World Patient Data

PURPOSE

Although immune checkpoint inhibitors (ICIs) have improved outcomes in certain patients with cancer, they can also cause life-threatening immunotoxicities. Predicting immunotoxicity risks alongside response could provide a personalized risk-benefit profile, inform therapeutic decision making, and improve clinical trial cohort selection. We aimed to build a machine learning (ML) framework using routine electronic health record (EHR) data to predict hepatitis, colitis, pneumonitis, and 1-year overall survival.

METHODS

Real-world EHR data of more than 2,200 patients treated with ICI through December 31, 2018, were used to develop predictive models. Using a prediction time point of ICI initiation, a 1-year prediction time window was applied to create binary labels for the four outcomes for each patient. Feature engineering involved aggregating laboratory measurements over appropriate time windows (60-365 days). Patients were randomly partitioned into training (80%) and test (20%) sets. Random forest classifiers were developed using a rigorous model development framework.

RESULTS

The patient cohort had a median age of 63 years and was 61.8% male. Patients predominantly had melanoma (37.8%), lung cancer (27.3%), or genitourinary cancer (16.4%). They were treated with PD-1 (60.4%), PD-L1 (9.0%), and CTLA-4 (19.7%) ICIs. Our models demonstrate reasonably strong performance, with AUCs of 0.739, 0.729, 0.755, and 0.752 for the pneumonitis, hepatitis, colitis, and 1-year overall survival models, respectively. Each model relies on an outcome-specific feature set, though some features are shared among models.

CONCLUSION

To our knowledge, this is the first ML solution that assesses individual ICI risk-benefit profiles based predominantly on routine structured EHR data. As such, use of our ML solution will not require additional data collection or documentation in the clinic.

Circulating IL-6 and not its circulating signaling components sIL-6R and sgp130 demonstrate clinical significance in NSCLC patients treated with immune checkpoint inhibitors

Introduction: Clinical roles of plasma IL-6 levels have been reported in patients with various cancers, including non-small cell lung cancer (NSCLC), treated with immune checkpoint inhibitors (ICIs). However, the roles of other IL-6 signaling components, soluble IL-6 receptor (sIL-6R) and soluble gp130 (sgp130), in the plasma have not been elucidated. Methods: Blood was collected from 106 patients with NSCLC before initiation of ICI treatment (anti-PD-1 or anti-PD-L1 antibody). Plasma levels of IL-6, sIL-6R, sgp130, and their complexes were assessed by Cox regression hazard model to evaluate their clinical significance. The clinical role of IL-6 or IL-6R genetic polymorphisms was also analyzed. Results: Cox regression analysis showed that higher plasma IL-6 levels significantly predicted unfavorable overall survival (OS; hazard ratio [HR] 1.34, 95% confidence interval [CI] 1.05-1.68, p = 0.012) in NSCLC patients treated with ICIs. However, plasma sIL-6R and sgp130 levels showed no prognostic significance (p = 0.882 and p = 0.934, respectively). In addition, the estimated concentrations of binary IL-6:sIL-6R and ternary IL-6:sIL-6R:sgp130 complexes and their ratios (binary/ternary complex) were not significantly associated with OS (p = 0.647, p = 0.727, and p = 0.273, respectively). Furthermore, the genetic polymorphisms of IL-6 (-634G>C) and IL-6R (48892A>C) showed no clinical role by Kaplan-Meier survival analysis (p = 0.908 and p = 0.639, respectively). Discussion: These findings demonstrated the clinical significance of plasma levels of IL-6, but not of other IL-6 signaling components, sIL-6R and sgp130, suggesting that classical IL-6 signaling, but not trans-signaling, may be related to anti-tumor immune responses in cancer patients treated with ICIs.