Development and external validation of a multivariable nomogram for predicting severe immune checkpoint inhibitor-associated myocarditis in advanced lung cancer

Introduction

In recent years, immunotherapy has emerged as a promising treatment approach for various types of cancer, including advanced lung cancer. The use of immune checkpoint inhibitors (ICIs), such as programmed cell death protein 1 (PD-1) inhibitors, has shown significant efficacy in improving the outcomes of patients with advanced non-small cell lung cancer (NSCLC) (1,2). However, along with their therapeutic benefits, ICIs have been associated with immune-related adverse events (irAEs) that can affect different organs, including the heart (3).

One of the potentially life-threatening irAEs associated with immunotherapy is immune myocarditis, an inflammatory condition affecting the heart muscle (4,5). Immune myocarditis can presented with a wide range of symptoms, including chest pain, dyspnea, palpitations, and heart failure, and may lead to severe cardiac dysfunction and even death if not promptly recognized and managed (6,7). Given the increasing use of ICIs in the treatment of lung cancer patients, the risk of developing immune myocarditis has become a topic of significant clinical relevance (8).

Several studies have investigated the incidence, risk factors, clinical presentation, and management of immune myocarditis in cancer patients receiving immunotherapy (9-11). However, although registry-based and multicenter studies have proposed clinical risk stratification approaches incorporating troponin elevation, ECG/conduction abnormalities, and ventricular dysfunction, most tools are applied at the time of suspected or established ICIM and have limited validation in Asian lung cancer real-world cohorts (10). Therefore, a pragmatic baseline risk-prediction tool that can be applied around ICI initiation—distinct from diagnostic evaluation for active myocarditis—may help triage monitoring intensity and enable earlier recognition of high-risk patients in routine practice (12).

Baseline abnormalities in cardiac biomarkers and routine tests may also have clinical plausibility for risk stratification (13). Pre-treatment troponin or BNP elevation and ECG/echocardiographic abnormalities can reflect subclinical myocardial injury or reduced cardiac reserve rather than autoimmune susceptibility per se, potentially lowering the threshold for clinically overt myocarditis once immune activation occurs during ICI therapy. In line with cardio-oncology guidance recommending baseline ECG and cardiac biomarkers before ICI initiation, we aimed to develop and externally validate a parsimonious model based on routinely available assessments (troponin, BNP, ECG, and echocardiography) to predict subsequent ICIM risk in advanced lung cancer. We present this article in accordance with the TRIPOD reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-aw-1252/rc).

Methods

Patient data

This retrospective analysis included 834 patients with advanced lung cancer who received ICIs at The First Affiliated Hospital of Guangzhou Medical University, Zhejiang Cancer Hospital, and The Affiliated Hospital of Kunming University of Science and Technology between January 2018 and June 2024 (Figure 1). Data from the Guangzhou center were randomly split at a 7:3 ratio into a training cohort (n=306) and an internal test cohort (n=131); the remaining 397 patients constituted the external validation cohort (Figure 1). Given the retrospective design and the rarity of severe ICIM, the study size was determined pragmatically by including all consecutive eligible patients during the prespecified period; therefore, no a priori sample size calculation was performed (14).

Figure 1 Flowchart of patient selection and Risk-Scoring model development. Flow diagram of patient selection and cohort splitting. From all ICI-treated patients, exclusions yielded 834 eligible cases; these were partitioned into a development cohort (patients without severe ICIM, CTCAE grade 0–2, n=403; severe ICIM, CTCAE v5.0 grade ≥3, n=34) and an external validation cohort (patients without severe ICIM, CTCAE grade 0–2, n=366; severe ICIM, CTCAE v5.0 grade ≥3, n=31). The development set was used to build the logistic prediction model; the external set was used for out-of-sample validation. CTCAE, Common Terminology Criteria for Adverse Events; ICI, immune checkpoint inhibitor; ICIM, immune checkpoint inhibitor-associated myocarditis.

Inclusion criteria

• Patients with advanced stage lung cancer receiving immunotherapy;
• Patients with available data on immunotherapy treatment;
• Patients who have undergone cardiac evaluation with electrocardiogram (ECG) and echocardiogram;
• Patients with pathology confirmation of lung cancer;
• Patients with complete data on gender, age, tumor-node-metastasis (TNM) staging, and smoking status;
• Patients with follow-up data on the occurrence of immune myocarditis.

Exclusion criteria

• Patients with early stage lung cancer;
• Patients who have not received immunotherapy treatment;
• Patients with incomplete data on cardiac evaluation;
• Patients without confirmed pathology of lung cancer;
• Patients with missing data on predictive factors;
• Patients lost to follow-up for monitoring immune myocarditis occurrence.

Pre-existing cardiovascular comorbidities were not used as exclusion criteria and were recorded as baseline characteristics. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of The First Affiliated Hospital of Guangzhou Medical University (approval No. KX-2023-009-01 approval date: 2023-09-01). Because this was a retrospective analysis of de-identified data, the requirement for written informed consent was waived by the ethics committee. All participating hospitals were informed and agreed to the study.

Immune checkpoint inhibitor-associated myocarditis (ICIM) diagnosis and grading

ICIM was adjudicated by two experienced physicians. After excluding alternative etiologies, diagnosis integrated medical history and clinical symptoms, the temporal relationship between ICI exposure and symptom onset, dynamic changes in troponin and other myocardial injury biomarkers, imaging abnormalities, cardiac magnetic resonance (CMR), and endomyocardial biopsy when indicated (15). For prediction modeling, candidate predictors were restricted to baseline/routine cardiac assessments at ICI initiation prior to any ICIM event; abnormalities recorded during suspected myocarditis were used only for endpoint adjudication and CTCAE grading. Toxicity grading followed CTCAE v5.0 (16), with ICIM classified as mild (grades 1–2) or severe (grade ≥3) (17). For model development and validation, the primary endpoint was severe ICIM (CTCAE v5.0 grade ≥3); patients without severe ICIM (no ICIM or mild ICIM, CTCAE grade 0–2) constituted the reference group (18).

Data collection

Demographic and clinical variables were collected, including age, sex, smoking status, comorbidities, lung cancer subtype, clinical stage, and treatment regimen. Candidate predictors were obtained from baseline/routine assessments at ICI initiation, prior to ICIM occurrence, including high-sensitivity troponin, BNP, ECG findings, and echocardiography results.

For laboratory predictors, results were recorded as normal versus abnormal based on the assay-specific upper reference limit (99th percentile/ULN) used at each participating center and operationalized using the local laboratory flag (normal vs abnormal) to accommodate inter-center assay heterogeneity. Index-center reference ranges for selected laboratory tests (CK: 50–310 U/L; BNP: <100 pg/mL) are provided in the Supplementary Materials.

ECG was categorized as abnormal if the baseline report documented any rhythm or conduction abnormality or repolarization change (e.g., ST-T changes, atrioventricular/bundle-branch block, atrial/ventricular arrhythmias); otherwise, it was categorized as normal (17). Echocardiography was categorized as abnormal if baseline transthoracic echocardiography documented ventricular systolic dysfunction [e.g., reduced (LVEF)], regional wall-motion abnormality, clinically relevant pericardial effusion, or other structural/functional abnormalities; otherwise, it was categorized as normal (17).

Outcome data were collected according to the adjudication and grading process described above.

Statistical analysis

Data from the Guangzhou center were randomly divided into a training cohort and internal validation cohorts at a 7:3 ratio for model development and internal evaluation. Non-normal distributed data were summarized as median (interquartile ranges). Categorical variables were analyzed by Chi-squared or Fisher’s exact test, and continuous variables by Student’s t-test or rank-sum test.

In the training cohort, least absolute shrinkage and selection operator (LASSO)-logistic regression was used to select candidate predictors derived from baseline assessments at ICI initiation and to develop a multivariable logistic model, which was visualized as a nomogram. Model performance was assessed by discrimination (ROC/AUC), calibration (calibration curves with Brier score), and clinical utility [decision curve analysis (DCA) across clinically plausible threshold probabilities] (19).

For clinical classification, an operating probability cutoff was prespecified in the training cohort and applied unchanged to the internal test and external validation cohorts; however, threshold-based classification metrics [sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV)] were not reported in this manuscript and are planned for prospective validation.

Two-sided P<0.05 was considered statistically significant. Analyses were performed using R (version 4.2.2) together with MSTATA (www.mstata.com).

Model development and validation

Predictors selected by LASSO-logistic regression in the training cohort were used to construct the prediction nomogram and multivariable logistic model which was then evaluated in the internal test and external validation cohorts by ROC/AUC, calibration, and DCA. Point assignment for the nomogram/risk score was derived from model coefficients, and the total score mapped to the predicted probability of severe ICIM. Model building and visualization were implemented in R 4.2.2 and MSTATA.

Results

Patient characteristics

In this study, we analyzed the baseline demographic and clinical characteristics of a total of 834 individuals across different cohorts. The mean age of the overall cohort was 60 years, with a standard deviation of 14. When stratified by cohort, the mean age was similar across the training cohort (59 years) and the external test cohort (61 years), while the internal test cohort had a slightly higher mean age of 61 years. In terms of gender distribution, 79.5% of the overall cohort was male, with a slightly higher proportion in the internal test cohort (83.2%) compared to the training cohort (75.5%) and external test cohort (81.4%). The prevalence of smoking was found in 61.0% of the overall cohort, with a similar distribution across the training, internal test, and external test cohorts. When considering comorbidities, 52.3% of the overall cohort had comorbid conditions, with comparable rates among the different cohorts. Regarding lung cancer types, adenocarcinoma was the most common subtype (39.8%), followed by squamous cell carcinoma (35.4%), with similar distribution across the cohorts (Table 1).

Clinical staging revealed a significant difference between cohorts, with 81.7% of the overall cohort classified as stage IV. Treatment patterns varied across the different cohorts, with 78.5% of individuals did not receive first-line treatment. The prevalence of heart failure was 11.2% in the overall cohort, with consistent rates in the training, internal test, and external test cohorts. Biomarker analysis indicated abnormal levels of troponin, creatine kinase, BNP, and IL-8 in a subset of individuals, with significant differences noted in their distribution among the cohorts. Cardiac parameters assessed by ECG and echocardiography showed variations in abnormalities among the cohorts, with no distinct patterns observed. Levels of inflammatory markers IL-1β and IL-6 did not exhibit significant differences among the cohorts.

Univariate analysis of influencing factors for severe ICIM

Univariate analysis of factors influencing severe ICIM revealed significant associations (P<0.05) with the following variables: troponin, B-type natriuretic peptide (BNP), and echocardiography, totaling 3 factors (Table 2).

Lasso regression was employed to further analyze the aforementioned factors. Considering the importance of electrocardiography in the diagnosis of myocarditis, the model identified the following as the best matching factors: troponin, BNP, echocardiography, and electrocardiography, totaling 4 factors (Figure 2A,2B).

Figure 2 LASSO-logistic predictor selection. (A) Coefficient trajectories for candidate variables across log(λ); numbers on top indicate the count of non-zero coefficients. (B) Ten-fold cross-validated binomial deviance versus log(λ); vertical dotted lines mark λ_min and λ_1se. The parsimonious λ was chosen to construct the prediction model. LASSO, least absolute shrinkage and selection operator.

Multivariate analysis of factors influencing severe ICIM

Multivariate logistic regression analysis identified the following as independent risk factors for severe ICIM: troponin [odds ratio (OR) =3.88, 95% confidence interval (CI): 1.41–10.66, P=0.008], BNP (OR =14.97, 95% CI: 3.29–68.21, P<0.001), ECG (OR =2.90, 95% CI: 0.75–11.23, P=0.12), and echocardiography (OR =7.86, 95% CI: 1.87–33.02, P=0.005) (Figure 3A). Despite the borderline statistics, ECG was retained in the final nomogram because of its clinical salience, high accessibility, and consistent directional signal in prior cardio-oncology literature, thereby adding pragmatic triage value within the multivariable framework. The accuracy and discriminative power of these four factors were validated using receiver operating characteristic (ROC) curves (Figure 3B).

Figure 3 Predictive contribution and single-marker discrimination. (A) Multivariable logistic coefficients retained after LASSO for troponin, BNP, ECG, and echocardiography (abnormal vs. normal). (B) ROC curves of each single marker for predicting severe ICIM with AUCs shown on the plot. AUC, area under the curve; BNP, B-type natriuretic peptide; CI, confidence interval; ECG, electrocardiogram; ICIM, immune checkpoint inhibitor-associated myocarditis; LASSO, least absolute shrinkage and selection operator; ROC, receiver operating characteristic.

Construction of a clinical nomogram model for severe ICIM prediction

Based on the results of multivariate logistic regression, a nomogram for predicting severe ICIM outcomes was constructed. In personalized medicine, troponin, BNP, echocardiography, and ECG can be used to determine the corresponding scores on the nomogram. By summing the scores of each variable, the nomogram enables the estimation of severe ICIM probability in patients receiving ICI treatment (Figure 4).

Figure 4 Nomogram of the multivariable logistic prediction model for severe ICIM. Points are assigned for troponin, BNP, ECG, and echocardiography (abnormal vs. normal). The total points map to the linear predictor and the predicted probability of severe ICIM at baseline. BNP, B-type natriuretic peptide; ECG, electrocardiogram; ICIM, immune checkpoint inhibitor-associated myocarditis.

Validation of the accuracy and discriminative power of the severe ICIM nomogram model

Random sampling was used, with a ratio of 7:3 between the training set (n=306) and the internal validation set (n=131). Additionally, 397 cases of clinical data from Zhejiang Cancer Hospital and The Affiliated Hospital of Kunming University of Science and Technology were used as the external validation set. ROC curves and calibration curves were employed to validate the accuracy and discriminative power of the nomogram. Results showed that the area under the curve (AUC) was 0.926 (95% CI: 0.880–0.971) for the training set, 0.885 (95% CI: 0.742–1.000) for the internal validation set, and 0.949 (95% CI: 0.884–1.000) for the external validation set (Figure 5A). Meanwhile, the calibration curves of the training set, internal validation set, and external validation set exhibited good agreement (Figure 5B-5D). The corresponding Brier scores were 5.0 (95% CI: 3.4–6.7) in the training set, 4.8 (95% CI: 2.5–7.0) in the internal validation set, and 3.3 (95% CI: 2.0–4.6) in the external validation set (Figure 5B-5D). This indicates that the predictive model has strong discriminative power.

Figure 5 Model performance across cohorts. (A) ROC curves for the development (training), internal test, and external validation cohorts with corresponding AUCs. (B-D) Calibration plots for the same cohorts; grey dashed line indicates perfect calibration. Brier scores are displayed on each panel. AUC, area under the curve; CI, confidence interval; ROC, receiver operating characteristic.

Evaluation of the clinical utility of the severe ICIM nomogram model

Clinical DCA was used to analyze the clinical utility of this model. Results showed maximum benefit in the training set (threshold probability: 0.00–0.62), internal validation set, and external validation set (threshold probability: 0.16–0.96) within these threshold probability ranges. DCA showed that the nomogram yielded a positive net benefit over the ‘treat-all’ and ‘treat-none’ strategies across clinically relevant threshold probabilities, supporting potential clinical utility and informing threshold selection in practice (Figure 6A-6C).

Figure 6 Clinical utility of the model by decision curve analysis. (A-C) Net benefit across threshold probabilities in the development, internal test, and external validation cohorts, comparing the prediction model (red) against “treat-all” and “treat-none” strategies.

Discussion

Key findings

In a cohort of advanced lung cancer patients treated with ICIs, we developed and validated a risk-prediction model for severe ICIM. The model integrates four readily obtainable clinical variables—high-sensitivity troponin, BNP, ECG, and transthoracic echocardiography—and provides individualized risk stratification via a nomogram. Across the training and external validation cohorts, the model demonstrated robust discrimination (AUC >0.80) and good calibration, indicating stable predictive performance. High-risk patients exhibited a significantly higher incidence of severe ICIM than low-risk patients; baseline troponin elevation was most strongly associated, with BNP, ECG abnormalities and echocardiographic findings providing complementary predictive value in risk stratification. Notably, ECG abnormality showed an OR of 2.90 (95% CI: 0.75–11.23; P=0.12). Although not statistically significant, its direction was consistent and clinically meaningful, and we therefore retained ECG in the nomogram to reflect the value of electrical changes in risk identification (16,20). These observations align with prior evidence that nearly all patients with severe ICIM present with elevated troponin (~94%) and the majority with new ECG abnormalities (~88%) (4).

Comparison with existing literature

Our findings are highly concordant with contemporary guidelines and large cohorts. The 2022 ESC cardio-oncology guidelines recommend routine ECG and troponin monitoring before and early during ICI therapy, with echocardiography for higher-risk patients (17,21). We confirm that baseline troponin abnormalities predict subsequent severe ICIM, consistent with clinical observations. Although ECG did not reach statistical significance in our dataset, its effect direction mirrors published data: multicenter reports indicate that ~90% of severe ICIM cases develop new ECG changes, including ST-T abnormalities, conduction blocks, and arrhythmias (4), underscoring ECG’s role as a pragmatic “sentinel” sign. Given its high accessibility and sensitivity to electrical disturbances, we judged ECG to have clinical utility within a multivariable framework despite borderline statistics. We also observed higher severe ICIM risk among older and male patients, in line with systematic evidence (4,22). By contrast, BNP contributed less to prediction—likely reflecting limited specificity and susceptibility to comorbidities—whereas troponin, a more specific marker of myocyte injury, remained central (18,19,23). Unlike diagnostic studies emphasizing CMR or endomyocardial biopsy, our approach prioritizes low-barrier markers for early risk triage, which enhances real-world scalability.

Pathophysiological rationale

Severe ICIM arises when blockade of PD-1/PD-L1 or CTLA-4 pathways disrupts immune tolerance and triggers T-cell-mediated myocardial inflammation (22,24-27) (4,20-22). Myocyte injury leads to early troponin release, while conduction system involvement causes arrhythmias and conduction delays—mechanistically consistent with our ECG signal despite limited power for statistical significance (23,24). With disease progression, ventricular dysfunction emerges, BNP levels rises, and echocardiography may reveal regional wall-motion abnormalities (28). Importantly, approximately half of severe ICIM patients maintain preserved ejection fraction early in the course (4), indicating that imaging derangements often lag behind biomarker and electrical changes (29). The model’s architecture—“troponin → ECG → echocardiography” aligns with a pathophysiological continuum—thus confers clinical interpretability.

Clinical implications

Although severe ICIM is uncommon (<1%) (4,11), its carries substantial mortality (30–50%) (24) and poses major threats to treatment continuity (28). Our model supports early identification of high-risk patients during ICI initiation, enabling intensified monitoring, multidisciplinary co-management, and timely initiation of immunosuppression when appropriate (25,30). In practice, ECG abnormalities—even with modest standalone effect—add triage value as part of a composite tool, especially in resource-constrained settings (24,26). Because the model relies on routine tests rather than repeated CMR or prophylactic biopsy, it offers favorable accessibility and cost-effectiveness (27,30), positioning it as a clinical “gatekeeper” for real-world deployment.

Strengths and limitations

Strengths include a prespecified multicenter development-validation workflow (stratified internal split plus an independent external cohort) that supports transportability; TRIPOD-aligned, transparent reporting; and exclusive reliance on routine, low-cost cardio-oncology tests (ECG, echocardiography, hs-troponin/BNP) with formal calibration and decision-curve analysis, using thresholds fixed in training and applied unchanged to validation—facilitating bedside uptake.

Limitations merit caution. (I) Retrospective design with potential residual confounding and documentation bias. (II) Severe ICIM is a rare outcome, and the limited number of severe events implies a lower events-per-predictor ratio (EPV) increasing uncertainty and potential overfitting (29), therefore, discrimination estimates (including high AUCs) should be interpreted cautiously under marked class imbalance. (III) Heterogeneous diagnostics and case ascertainment across centers (e.g., variation in CMR/biopsy availability) may contribute to outcome misclassification. (IV) Predictors were restricted to baseline assessments at ICI initiation and did not model longitudinal trajectories (e.g., dynamic troponin/ECG changes). (V) ECG abnormality showed a nonsignificant association but was retained for clinical interpretability within a multivariable framework. (VI) For rare events, ROC may overstate clinical performance; calibration drift may occur at extreme risk deciles. (VII) External validity focused on advanced lung cancer cohorts; prospective, registry-embedded impact evaluation remains needed. (VIII) We did not perform a temporally separated validation: splitting by calendar periods would further reduce events per period and could yield unstable estimates; robust temporal validation would require harmonized timestamps across centers (ICI initiation date, baseline cardiac assessment date, follow-up window, and adjudication/diagnosis date), which were not consistently available without substantial re-curation. (IX) Threshold-based classification metrics (sensitivity, specificity, PPV/NPV) and calibration intercept/slope were not reported; future prospective validation will prespecify operating points and provide these metrics to support clinical implementation (31).

Future directions

Built from widely available clinical indicators, this nomogram offers an immediately implementable way to flag patients at heightened risk of severe ICIM and to guide monitoring intensity. To translate it responsibly into practice, actions should focus on:

• Incorporating longitudinal trajectories to enable dynamic risk prediction and to verify the time-dependent contribution of ECG and other variables;
• Broadening external validation across tumor types and diverse ICI regimens;

Integrating the tool with electronic health records (EHRs) for real-time risk scoring and automated alerts (17,32).

With continued optimization and prospective multicenter validation, the model could anchor standardized cardio-oncology safety workflows while preserving immunotherapy delivery.

Conclusions

We developed and externally validated a simple four-predictor model—high-sensitivity troponin, BNP, ECG, and echocardiography—to estimate the risk of severe ICI-related myocarditis (severe ICIM) in advanced lung cancer. The nomogram showed strong discrimination and good calibration across internal and external cohorts, supporting pre-ICI triage and risk-stratified monitoring using readily available tests. Prospective multicenter studies and EHR integration are warranted to confirm generalizability and facilitate clinical implementation.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-aw-1252/rc

Data Sharing Statement: Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-aw-1252/dss

Peer Review File: Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-aw-1252/prf

Funding: This work was supported by National Natural Science Foundation of China (No. 82570007), Development and Industrialization of a Robotic Bronchoscopy Platform (2025B1111010001), Plan on enhancing scientific research in GMU (GMUCR2024-01011), Major Project of Guangzhou National Laboratory (GZNL2023A02011), Study of the Carcinogenic Mechanisms and Molecular Characteristics in Primary Pulmonary Lymphoepithelioma-like Carcinom (2024B03J0862), and the Construction Fund of Key Medical Disciplines of Hangzhou (2025HZGF06).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-aw-1252/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of the First Affiliated Hospital of Guangzhou Medical University (Approval number: KX-2023-009-01 approval date: 2023-09-01). Because this was a retrospective analysis of de-identified data, the requirement for written informed consent was waived by the ethics committee. All participating hospitals were informed and agreed to the study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Brahmer J, Reckamp KL, Baas P, et al. Nivolumab versus Docetaxel in Advanced Squamous-Cell Non-Small-Cell Lung Cancer. N Engl J Med 2015;373:123-35. [Crossref] [PubMed]
2. Garon EB, Hellmann MD, Rizvi NA, et al. Five-Year Overall Survival for Patients With Advanced Non‒Small-Cell Lung Cancer Treated With Pembrolizumab: Results From the Phase I KEYNOTE-001 Study. J Clin Oncol 2019;37:2518-27. [Crossref] [PubMed]
3. Yin Q, Wu L, Han L, et al. Immune-related adverse events of immune checkpoint inhibitors: a review. Front Immunol 2023;14:1167975. [Crossref] [PubMed]
4. Mahmood SS, Fradley MG, Cohen JV, et al. Myocarditis in Patients Treated With Immune Checkpoint Inhibitors. J Am Coll Cardiol 2018;71:1755-64. [Crossref] [PubMed]
5. Zhan L, Qin Z, Yang L, et al. Immune-related myocarditis characterized by malignant arrhythmia caused by conduction system involvement: case report and literature review. Front Oncol 2025;15:1589990. [Crossref] [PubMed]
6. Moslehi JJ, Salem JE, Sosman JA, et al. Increased reporting of fatal immune checkpoint inhibitor-associated myocarditis. Lancet 2018;391:933. [Crossref] [PubMed]
7. Moslehi J, Lichtman AH, Sharpe AH, et al. Immune checkpoint inhibitor-associated myocarditis: manifestations and mechanisms. J Clin Invest 2021;131:e145186. [Crossref] [PubMed]
8. Munir AZ, Gutierrez A, Qin J, et al. Immune-checkpoint inhibitor-mediated myocarditis: CTLA4, PD1 and LAG3 in the heart. Nat Rev Cancer 2024;24:540-53. [Crossref] [PubMed]
9. Salem JE, Allenbach Y, Vozy A, et al. Abatacept for Severe Immune Checkpoint Inhibitor-Associated Myocarditis. N Engl J Med 2019;380:2377-9. [Crossref] [PubMed]
10. Laenens D, Yu Y, Santens B, et al. Incidence of Cardiovascular Events in Patients Treated With Immune Checkpoint Inhibitors. J Clin Oncol 2022;40:3430-8. [Crossref] [PubMed]
11. Nielsen DL, Juhl CB, Nielsen OH, et al. Immune Checkpoint Inhibitor-Induced Cardiotoxicity: A Systematic Review and Meta-Analysis. JAMA Oncol 2024;10:1390-9. [Crossref] [PubMed]
12. Mahmood SS, Fradley MG, Cohen JV, et al. Myocarditis in Patients Treated With Immune Checkpoint Inhibitors. J Am Coll Cardiol 2018;71:A699.
13. Rothman AMK, Vachiery JL, Howard LS, et al. Intravascular Ultrasound Pulmonary Artery Denervation to Treat Pulmonary Arterial Hypertension (TROPHY1): Multicenter, Early Feasibility Study. JACC Cardiovasc Interv 2020;13:989-99. [Crossref] [PubMed]
14. Riley RD, Ensor J, Snell KIE, et al. Calculating the sample size required for developing a clinical prediction model. BMJ 2020;368:m441. [Crossref] [PubMed]
15. Thuny F, Bonaca MP, Cautela J. What Is the Evidence of the Diagnostic Criteria and Screening of Immune Checkpoint Inhibitor-Induced Myocarditis? JACC CardioOncol 2022;4:624-8. [Crossref] [PubMed]
16. Freites-Martinez A, Santana N, Arias-Santiago S, et al. Using the Common Terminology Criteria for Adverse Events (CTCAE - Version 5.0) to Evaluate the Severity of Adverse Events of Anticancer Therapies. Actas Dermosifiliogr (Engl Ed) 2021;112:90-2. [Crossref] [PubMed]
17. Brahmer JR, Abu-Sbeih H, Ascierto PA, et al. Society for Immunotherapy of Cancer (SITC) clinical practice guideline on immune checkpoint inhibitor-related adverse events. J Immunother Cancer 2021;9:e002435. [Crossref] [PubMed]
18. Thygesen K, Alpert JS, Jaffe AS, et al. Fourth Universal Definition of Myocardial Infarction (2018). J Am Coll Cardiol 2018;72:2231-64. [Crossref] [PubMed]
19. Debray TPA, Collins GS, Riley RD, et al. Transparent reporting of multivariable prediction models developed or validated using clustered data (TRIPOD-Cluster): explanation and elaboration. BMJ 2023;380:e071058. [Crossref] [PubMed]
20. Zlotoff DA, Hassan MZO, Zafar A, et al. Electrocardiographic features of immune checkpoint inhibitor associated myocarditis. J Immunother Cancer 2021;9:e002007. [Crossref] [PubMed]
21. Lyon AR, López-Fernández T, Couch LS, et al. 2022 ESC Guidelines on cardio-oncology developed in collaboration with the European Hematology Association (EHA), the European Society for Therapeutic Radiology and Oncology (ESTRO) and the International Cardio-Oncology Society (IC-OS). Eur Heart J 2022;43:4229-361. [Crossref] [PubMed]
22. Mauriello A, Correra A, Maratea AC, et al. Immune Checkpoint Inhibitor-Associated Myocarditis: Risk, Diagnosis, and Clinical Impact. J Clin Med 2026;15:814. [Crossref] [PubMed]
23. Murtagh G, deFilippi C, Zhao Q, et al. Circulating biomarkers in the diagnosis and prognosis of immune checkpoint inhibitor-related myocarditis: time for a risk-based approach. Front Cardiovasc Med 2024;11:1350585. [Crossref] [PubMed]
24. Feng S, Zhao B, Sha S, et al. Overview of immune checkpoint inhibitor associated myocarditis mechanisms diagnostics and treatment. Front Immunol 2025;16:1677984. [Crossref] [PubMed]
25. Ma P, Liu J, Qin J, et al. Expansion of Pathogenic Cardiac Macrophages in Immune Checkpoint Inhibitor Myocarditis. Circulation 2024;149:48-66. [Crossref] [PubMed]
26. Won T, Kalinoski HM, Wood MK, et al. Cardiac myosin-specific autoimmune T cells contribute to immune-checkpoint-inhibitor-associated myocarditis. Cell Rep 2022;41:111611. [Crossref] [PubMed]
27. Axelrod ML, Meijers WC, Screever EM, et al. T cells specific for α-myosin drive immunotherapy-related myocarditis. Nature 2022;611:818-26. [Crossref] [PubMed]
28. Talukder S, Ghose A, Chakraborty T, et al. Evolving cardiac biomarkers for immune checkpoint inhibitor related myocarditis in cancer patients. Int J Cardiol Heart Vasc 2023;49:101278. [Crossref] [PubMed]
29. Chen YC, Dolladille C, Rao A, et al. Immune Checkpoint Inhibitor Myocarditis and Left Ventricular Systolic Dysfunction. JACC CardioOncol 2025;7:234-48. [Crossref] [PubMed]
30. Badaan S, Abbass A, Tov S, et al. How should immune checkpoint inhibitor myocarditis be treated? Cardiooncology 2025;11:108. [Crossref] [PubMed]
31. Van Calster B, McLernon DJ, van Smeden M, et al. Calibration: the Achilles heel of predictive analytics. BMC Med 2019;17:230. [Crossref] [PubMed]
32. Chandra A, Philips ST, Pandey A, et al. Electronic Health Records-Based Cardio-Oncology Registry for Care Gap Identification and Pragmatic Research: Procedure and Observational Study. JMIR Cardio 2021;5:e22296. [Crossref] [PubMed]