This study is a retrospective, two-center study. The case data of 2,665 patients with solid tumors after receiving anti-PD-1/PD-L1 immunotherapy at the Affiliated Hospital of Qingdao University and the Eighth People's Hospital of Qingdao from October 2018 to October 2023 were screened, 60 patients with anti-PD-1/PD-L1 treatment-related cardiovascular toxicity were screened, and patients who did not experience any immune-related adverse events during the same period were screened in a 1:2 ratio 120 cases were included in the control group. Controls were matched 1:2 to cases based on age (±5 years), sex, tumour type, and treatment line, in addition to being irAE-free during follow-up. A total of 2,665 patients receiving ICI therapy were screened. During follow-up, 314 patients were lost to follow-up and excluded. Among the remaining patients, 60 developed cardiovascular toxicity and were included as cases. From the 2,291 patients without cardiovascular toxicity, 120 controls were selected at a 1:2 ratio and matched to cases. This matching design was chosen to enhance statistical power and reduce potential bias, as the number of cases was relatively limited, and increasing the number of controls improves the efficiency of estimation without substantially increasing cost or workload. Age 31–92 years, median age 65 (57, 70) years. Inclusion criteria: 1. Age ≥18 years; 2. Solid malignant tumor diagnosed by pathology or cytology; 3. Received at least 1 treatment with PD-1/PD-L1 inhibitors; 4. Relatively complete medical records with 1 year of follow up. Exclusion criteria: 1. Uncertain primary tumor, unclear pathological results and clinical stage; 2. Combination of multiple primary malignant tumors; 3. Acute infection during anti-PD-1/PD-L1 treatment; 4. Immunodeficiency or combination of autoimmune diseases; 5. Elevated cardiac biomarkers or cardiovascular symptoms and signs after immunotherapy, Symptoms and signs attributed to other diagnosed diseases[a multidisciplinary review (cardiology and oncology) attributed those findings clearly to pre-existing or alternative diagnoses rather than to ICI therapy]. With 7 potential independent variables, a minimum sample size of 70 was required. A total of 180 patients were included in this study. In the overall study population (n = 180), the underlying cancer types included lung cancer (n = 103), gastrointestinal malignancies (n = 51), gynecologic cancers (n = 9), and melanoma (n = 17).This retrospective study was approved by the hospital's medical ethics committee (approval number: QYFYWZLL29401) and conducted in accordance with the Declaration of Helsinki. Since this was a retrospective study, the Investigational Review Board waived the requirement for written informed consent.

A total of 180 patients were included in this study, of which 136 (75.56%) were male and 44 (24.44%) were female. The median age of the included patients was 65 (57,70) years old; there were 85 cases (47.22%) with a history of smoking, 120 patients who did not experience cardiovascular toxicity, and 60 patients who experienced cardiovascular toxicity. Among the 60 patients with cardiovascular toxicity, myocarditis was the most common event (n = 35), followed by heart failure (n = 13), arrhythmias (n = 8), and acute coronary syndrome (n = 4). Table 1 list the demographic data and baseline characteristics. The results of the analysis of variance of the influential factors included in the study showed that the differences in age, sex, SIRI, ECOG, tumor stage, combined hypertension, combined diabetes, and tumor metastasis between the non-cardiovascular toxicity group and cardiovascular toxicity group were statistically significant (all p < 0.05). See Table 1.

According to the relevant literature search and the recommendations of relevant experts this study intends to include the variables of one-factor logistic regression analysis p < 0.05 into the multifactorial logistic regression analysis, and among these variables, age, BMI, and SIRI are continuous numerical variables, and use the RCS (Restricted Cubic Spline) to evaluate whether the relationship between them and whether cardiovascular toxicity outcome occurs or not is nonlinear, and to determine its relationship with the Increased risk-related cutoff points, see Figures 1–3, age, BMI, SIRI p for nonlinear >0.05 (0.259, 0.836, 0.504), and therefore do not have a nonlinear relationship with cardiovascular toxicity outcomes. In order to construct the nomogram with high clinical utility, the cut-off values for age, BMI, and SIRI were determined based on clinical relevance, prior literature, and ROC curve analysis (Youden's index), relevant variables were converted into categorical formats with the following coding: age (<65 years old assigned 1, ≥65 years old assigned 2), SIRI (×10⁹) (<1.25 assigned 1, ≥1.25 assigned 2, and BMI (<22.48 kg/m² assigned 1, ≥22.48 kg/m² assigned 2) The other variables were assigned values: sex (female = 2, male = 1); ECOG (>1 assigned 1, ≤1 assigned 0), tumor stage (II-III assigned 1, Ⅳ assigned 2) combined with hypertension (no = 0, yes = 1); tumor metastasis (no = 0, yes = 1); These coded categories correspond directly to the variable axes (and their labels) shown in the nomogram (Figure 4). The re-assigned variables were subjected to single + multiple-factorial logistic regression analysis (the variables of single-factorial logistic regression analysis with p < 0.05 were included in the multiple-factorial logistic regression). factor logistic regression), and the results of the one-way logistic regression analysis showed that SIRI (OR = 2.26, 95% CI 1.19–4.27, p = 0.012), ECOG (OR = 9.67, 95% CI 3.04–30.69, p < 0.001), combined hypertension (OR = 3.50, 95% CI 1.78–6.88, p < 0.001), combined diabetes (OR = 2.52, 95% CI 1.13–5.66, p = 0.025), tumor metastasis (OR = 0.17, 95% CI 0.08–0.39, p < 0.001), tumor stage (OR = 0.40, 95% CI 0.21–0.76, p = 0.006), and sex (OR = 0.43, 95% CI 0.19–0.96, p = 0.40) were predictors of the occurrence of cardiovascular toxicity. The results of multifactorial logistic regression analysis showed that two indicators, ECOG (OR = 9.81, 95% CI 2.73–35.25, p < 0.001) and tumor metastasis (OR = 0.26, 95% CI 0.10–0.71, p = 0.008), were independent influences on cardiovascular toxicity, with ECOG as an independent cardiovascular toxicity risk factor and tumor metastasis was an independent protective factor for cardiovascular toxicity. See Table 2. Therefore, the seven variables of univariate logistic regression were used as input features to construct the cardiovascular toxicity risk prediction model by incorporating them into the nomogram.

The seven variables of one-way logistic regression were included in the nomogram to construct the cardiovascular toxicity risk prediction model, see Figure 4, with different values corresponding to different variables, and then the scores corresponding to each variable were summed up, and the resulting total score could be read on the total score value below, which corresponded to the probability associated with the occurrence of cardiovascular toxicity below. Thus, this model allows for individualized prediction for each patient. The cardiovascular toxicity prediction model constructed by the nomogram was validated, and the ROC curve was plotted, see Figure 5, with an AUC (Area Under Curve) of 0.77 and a 95% CI of 0.70–0.85, and its sensitivity, specificity, positive predictive value, negative predictive value and other indexes are shown in Table 3, and the predictive model was subjected to the Hosmer-Lemeshow test of the fitting effect, with a χ² = 12.934 and a p = 0.053,the fit of the model is better (p > 0.05), which indicates that the accuracy and differentiation of the nomogram model is better, and the calibration curve of the internal evaluation and internal validation is better fit, revealing that there is a good agreement between the predictive probability of the model and the actual situation, showing the accuracy of the predictive model, see Figure 6.The DCA curve shows that this model has a better effect on the clinical benefit, see Figure 7.

Higher ECOG scores were associated with worse 12-month overall survival (OS) and progression-free survival (PFS) in elderly patients with advanced cutaneous squamous cell carcinoma treated with ICIs in a previous real-world study (21). It is well known that ECOG physical status is an established prognostic marker for patients with advanced malignancies and correlates with OS (22).In the present study, we demonstrated that ECOG > 1 was an independent risk factor for cardiovascular toxicity ECOG (OR = 9.81, 95% CI 2.73–35.25, p < 0.001), which means that cardiovascular toxicity occurred 9.81 times more frequently in patients with ECOG > 1 than in patients with ECOG ≤ 1. This finding is clearly important when balancing the discussion of treatment benefits vs. risks or impact on quality of life. Therefore, patients with ECOG > 1 should fully consider the therapeutic benefit vs. the risk of cardiovascular toxicity when undergoing PD-1/PD-L1 therapy. This highlights the relevant role that ECOG still plays in treatment decisions.

It has been shown that grade 1 and 2 hypertension without medication is associated with a higher risk of heart failure and other cardiovascular events in patients with breast, colorectal or gastric cancer (23, 24). Similarly, hypertension is a risk factor for chemotherapy-induced cardiomyopathy in oncology patients (25, 26). For patients with cardiovascular disease, hypertension and diabetes are a very important risk factor that promotes accelerated coronary atherosclerosis, plaque progression and rupture, myocardial remodeling, and myocardial ischemia and PD-1/PD-L1 has an important role in the above process. Atherosclerotic diseases are chronic inflammatory diseases, and PD-1/PD-L1-activated T-lymphocytes attack endothelial cells in addition to cardiomyocytes, promoting plaque progression and rupture as well as generating inflammation in the cardiac conduction system (15, 27). Therefore, PD-1/PD-L1 induces acute coronary syndrome and acute heart failure by damaging endothelial cells, accelerating plaque progression and rupture, and amplifying myocardial inflammation and myocardial ischemia (27–29). Our composite cardiovascular endpoint included myocarditis, arrhythmias, heart failure, acute coronary syndrome (ACS), venous thromboembolism, and others. While hypertension and diabetes are established risk factors for atherosclerotic events such as ACS, their association with other immune-related cardiotoxicities (e.g., myocarditis) may involve distinct pathways, such as systemic inflammation or endothelial dysfunction. Thus, the observed association likely reflects a heterogeneous risk profile across the composite outcome. The nomogram aims to capture shared clinical predictors that are readily available in practice, facilitating early risk stratification regardless of the specific cardiotoxicity subtype. Tumor Metastasis.

Unlike naïve T cells that circulate between secondary lymphoid tissues (i.e., lymph nodes) and blood, memory T cells circulate between lymphoid tissues, blood, and peripheral tissues (e.g., lungs, intestines, or skin) (30, 31). This allows memory T cells to respond directly and rapidly to the presence of pathogens in peripheral tissues. When tumors metastasize, memory T cells rapidly undergo activation and transfer to the corresponding tissues and organs to produce the appropriate anti-tumor effects. Recent studies have shown that CD4+ TCM cells (central memory CD4+ T cells) play a crucial role in cardioprotection during the treatment of ICIs (32), which is consistent with the results of the present study. In our analysis, tumor metastasis emerged as an independent protective factor against cardiovascular toxicity (OR = 0.26). While we hypothesize that memory T-cell redistribution and cardioprotective mechanisms may play a role, several non-causal explanations must be considered. These include selection bias (controls were irAE-free), competing risks such as shorter survival or earlier treatment discontinuation in metastatic patients, and potential confounding by tumor subtype, treatment regimen, or follow-up intensity. Therefore, this finding should be interpreted as hypothesis-generating rather than definitive. Future prospective studies with detailed time-to-event data, treatment duration adjustments, and larger sample sizes are needed to clarify whether metastasis truly confers biological protection or reflects methodological artifacts.

We observed a high risk of immune-related myocarditis, pericarditis, arrhythmias, coronary artery disease, and myocardial infarction in males in a retrospective study of immune checkpoint inhibitor-associated immune-associated adverse cardiac reactions, and we also found a sex-specific male pattern in adverse cardiac reports of arrhythmias, OR = 0.81 (95% CI 0.75–0.87, p < 0.001) (33, 34). The cardiovascular protective effects of estrogen have been confirmed by numerous studies both nationally and internationally (35, 36), which may predict that the protective effects of estrogen on the cardiovascular system may likewise be resistant to cardiovascular toxicity produced by immune checkpoint inhibitors.

Several limitations of this study warrant consideration. First, the control group was selected from patients without immune-related adverse events (irAEs) during follow-up and matched to cases based on age, sex, tumor type, and treatment line. While this matched design improves comparability, it differs from a full cohort risk model and may influence the absolute risk estimates derived from the nomogram. Consequently, the model should be interpreted as a tool for relative risk stratification rather than for estimating absolute risk, with its clinical utility primarily focused on identifying high-risk individuals for intensified cardiovascular monitoring during anti-PD-1/PD-L1 therapy. Additionally, the retrospective nature of the study, the limited sample size, and the internal validation within a single cohort raise concerns regarding potential selection bias and overfitting. Therefore, external validation in prospective, independent, and larger populations is essential before the nomogram can be recommended for clinical application.