Development of peripheral biomarker-based prognostic nomograms for short-term and long-term survival in immune checkpoint inhibitor-associated myocarditis

Abstract

Background: Immune checkpoint inhibitor-associated myocarditis (ICI myocarditis) is a rare but highly fatal immune-related adverse reaction. This study aimed to develop nomogram prognostic models for both short-term and long-term survival outcomes in patients with ICI myocarditis based on key biomarkers in peripheral blood.

Methods: In this single-center retrospective study, we included 90 patients with ICI myocarditis at the Fourth Hospital of Hebei Medical University. Critical peripheral biomarkers associated with 40-day and 1-year overall survival (OS) were identified. Two prognostic models were developed and evaluation of the models were performed with receiver operating characteristic (ROC) curves, C-index, calibration curves, and decision curve analysis (DCA).

Results: A total of 24 patients (26.7%) succumbed within 40 days, while 40 patients (44.4%) died within one year. Cardiac troponin-I (cTnI), N-terminal pro-brain natriuretic peptide (NTBNP) and lactic dehydrogenase-to-albumin ratio (LAR) were identified as critical prognostic factors for 40-day OS in patients with ICI myocarditis and utilized to develop a nomogram model. The model demonstrates an area under the curve (AUC) of 0.867 [95% confidence interval (CI): 0.774-0.960] and a C-index of 0.824. Another predictive model for the 1-year OS was developed based on cTnI, NTBNP, LAR and systemic inflammatory response index (SIRI) with an AUC of 0.765 (95% CI: 0.664-0.866) and a C index of 0.742. The calibration curve demonstrates that both models exhibit strong consistency. The results of the DCA further indicate that both nomograms possess substantial clinical utility.

Conclusions: These two prediction models will enable clinicians to more effectively utilize readily available peripheral blood biomarkers for the convenient and efficient identification of high-risk patients with poor prognoses, thereby facilitating early intervention.

Keywords: Immune checkpoint inhibitor (ICI); biomarkers; myocarditis; nomogram; prognosis.

Figure 1

Clinical characteristics of patients with ICI-myocarditis: flowchart of patient enrollment and grouping (A). The number of patients with other concomitant irAEs (B). The change of CK-MB (C), cTnI (D), NTBNP (E) and LVEF (F) from baseline to the onset of ICI myocarditis. CK-MB, creatine kinase isoenzyme; cTnI, cardiac troponin-I; ICI, immune checkpoint inhibitor; irAEs, immune-related adverse events; LVEF, left ventricular ejection fraction; NTBNP, N-terminal pro-brain natriuretic peptide.

Figure 2

The development and evaluation of a nomogram prognostic model for 40-day OS in patients with ICI myocarditis: LASSO-Cox regression analysis of 14 risk factors (A), and 3 risk factors were selected from tenfold cross-validation at the optimum value of the parameter λ by 1 − s.e. criteria (B). Nomogram for 40-day OS in patients with ICI myocarditis (C). ROC curve of the nomogram model (D), calibration curve for nomogram by bootstrap with 1,000 repetitions (E), and decision curve analysis of the nomogram (F). Kaplan-Meier survival curves of 40-day survival for patients with high and low risk stratified by the model (G). AUC, area under the curve; cTnI, cardiac troponin-I; ICI, immune checkpoint inhibitor; LAR, lactic dehydrogenase-to-albumin ratio; LASSO, least absolute shrinkage and selection operator; NTBNP, N-terminal pro-brain natriuretic peptide; OS, overall survival; pr 40-day, prediction model for 40-day OS; ROC, receiver operating characteristic; s.e., standard error.

Figure 3

The development and evaluation of a nomogram prognostic model for 1-year OS in patients with ICI myocarditis: LASSO-Cox regression analysis of 15 risk factors (A), and 4 risk factors were selected from tenfold cross-validation at the optimum value of the parameter λ by minimum criteria (B). Nomogram for 1-year OS in patients with ICI myocarditis (C). ROC curve of the nomogram model (D), calibration curve for nomogram by bootstrap with 1,000 repetitions (E), and decision curve analysis of the nomogram (F). Kaplan-Meier survival curves of 1-year survival for patients with high and low risk stratified by the model (G). AUC, area under the curve; cTnI, cardiac troponin-I; ICI, immune checkpoint inhibitor; LAR, lactic dehydrogenase-to-albumin ratio; LASSO, least absolute shrinkage and selection operator; NTBNP, N-terminal pro-brain natriuretic peptide; OS, overall survival; pr 1-year, prediction model for 1-year OS; ROC, receiver operating characteristic; SIRI, system inflammation response index.

Figure 4

The online dynamic nomogram for the prediction of 40-day (A) and 1-year (B) OS in patients with ICI myocarditis. The user can refer to the patient’s peripheral blood test results to adjust the values of various indicators on the left panel. Afterward, click the “Predict” button and select the corresponding button at the top right to obtain the predicted survival rate and the patient’s survival curve. Among these indicators, LAR = LDH/ALB, and SIRI = NE × MO/LY. Please note the units of measurement for each indicator: NTBNP (ng/L), cTnI (mg/L), LDH (U/L), ALB (g/L), NE (1×10⁹/L), MO (1×10⁹/L), LY (1×10⁹/L). Additionally, there are “time40” and “time360” buttons at the bottom of the left panel that allow users to view the patient’s survival probability at other time points, within 40 days and 1 year. To maintain the reliability of the prediction results, it is recommended to adjust these to the default maximum values. ALB, albumin; AUC, area under the curve; cTnI, cardiac troponin-I; ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; LAR, lactic dehydrogenase-to-albumin ratio; LDH, lactate dehydrogenase; LY, absolute lymphocyte count; MO, absolute mononuclear count; NE, absolute neutrophil count; NTBNP, N-terminal pro-brain natriuretic peptide; OS, overall survival; ROC, receiver operating characteristic; SIRI, system inflammation response index.

Figure 5

The online app webpage for external validation of the model. We developed a web application to facilitate the external validation of the model. Researchers from various institutions can upload their own data as an external validation dataset. By following the instructions provided on the website, they can organize the data into a .csv file and upload it. Afterward, they can click on the “40-day Survival Model” and “1-Year Survival Model” buttons to perform the external validation for both the 40-day and 1-year survival models. The website will automatically generate calibration and DCA curves. The figure shows a screenshot of an example where we used simulated data to perform external validation of the model. ALB, albumin; cTnI, cardiac troponin-I; DCA, decision curve analysis; LDH, lactate dehydrogenase; LY, absolute lymphocyte count; MO, absolute mononuclear count; NE, absolute neutrophil count; NTBNP, N-terminal pro-brain natriuretic peptide; pr 40-day, prediction model for 40-day overall survival.