Clinical-genetic profiles and risk prediction model in childhood restrictive cardiomyopathy: a national cohort study of China

Abstract

Background

Understanding restrictive cardiomyopathy (RCM) in children is limited, and currently, no prognostic model is available for assessing risk stratification in pediatric RCM. The authors elaborated on the clinical and genetic characteristics of pediatric RCM and developed a prediction model to assess the 1-year risk of major adverse cardiovascular events (MACE), aiding in determining the optimal timing for heart transplantation (HTx).

Methods

This multicenter retrospective cohort study collected data on children with RCM from 14 centers in China, with patient enrollment from January 1, 2013, to December 31, 2022, and follow-up concluding on August 1, 2023. It analyzed clinical and genetic characteristics and followed patients longitudinally for the development of MACE (including cardiac death, HTx, or equivalent events). A logistic regression model was developed to predict MACE one year post-diagnosis, with internal validation using bootstrapping. The model’s performance was evaluated in terms of its discrimination, calibration, and clinical utility.

Results

The study included 185 children with a median diagnosis age of 5.4 years (IQR, 3.1–9.4), and 110 (59%) were male. Significant heart failure was the primary clinical feature. TNNI3 mutations were present in 61% of cases, the most common in pediatric RCM. During the follow-up period, 114 patients (62%) experienced MACE, with the median MACE-free survival time for the entire cohort being 2.1 years post-diagnosis (IQR, 0.6–5.4). A prediction model was developed to estimate the one-year risk of MACE using four easily accessible clinical parameters: heart failure classification, brain natriuretic peptide levels, cardiac troponin levels, and a modified score for ST-segment deviation. Internal validation with bootstrapping confirmed accuracy, showing an optimism-adjusted C statistic of 0.78 (95% CI, 0.72–0.85) and a Brier score of 0.17 (95% CI, 0.14–0.21), with a calibration slope of 0.90 (95% CI, 0.63–1.27). Decision curve analysis indicated high net benefit across HTx treatment thresholds from 8.5% to 78.3%.

Conclusions

This model utilizes accessible clinical parameters to assess individual risk for MACE in pediatric RCM, potentially improving precision in healthcare strategies and supporting more informed clinical decision-making.

Background

Restrictive cardiomyopathy (RCM) is a rare disorder characterized by diastolic ventricular dysfunction and biatrial enlargement [1]. Its etiology is diverse, including idiopathic, genetic, storage, infiltrative, fibrotic, and oncological causes [2]. In children, RCM is primarily idiopathic or genetic, often associated with pathogenic variants in genes encoding sarcomeric, cytoskeletal, and cytosolic proteins [3, 4]. The understanding of pediatric RCM is limited, largely due to its rarity among cardiomyopathies in children. Despite its rarity, pediatric RCM presents significant healthcare challenges and is associated with a poor prognosis, often resulting in heart failure, arrhythmias, and sudden cardiac death, with a 5-year transplant-free survival rate of about 30% [5, 6]. Heart transplantation (HTx) is the preferred long-term survival option [7]. However, the shortage of pediatric heart donors and delays in listing significantly limit the feasibility of HTx and create substantial challenges in determining the appropriate timing for listing and prioritizing candidates [8].

This study aims to define the clinical and genetic profiles of pediatric RCM and to develop a new prediction model for individualized risk assessment of major adverse cardiovascular events (MACE), aiding in determining the optimal timing for HTx.

Methods

Study population and design

The study cohort consisted of patients under the age of 18 who were diagnosed with RCM between January 1, 2013, and December 31, 2022, across 14 participating pediatric cardiology centers in China (Additional file 1: Figure S1). These hospitals are leading centers for pediatric cardiovascular disease treatment, with most being national or regional children’s medical centers, which ensures a high level of authenticity and credibility in the research data. Patients who were newly diagnosed with RCM according to the diagnostic criteria established by the American Heart Association and the Pediatric Cardiomyopathy Registry at participating centers were eligible for inclusion [1, 9]. Pediatric RCM is characterized by echocardiographic findings of ventricular diastolic dysfunction (both qualitative and quantitative analyses, as children cannot uniformly use certain echocardiographic parameters for diastolic function) and significant atrial enlargement (Z-score ≥ 2), along with a normal or mildly reduced ejection fraction (LVEF ≥ 40%) [1, 10, 11]. The diagnosis is established after excluding specific secondary causes of diastolic dysfunction and atrial enlargement. This exclusion process involves biochemical tests to rule out inflammatory infections, confirm elevated brain natriuretic peptide (BNP) or N-terminal pro-brain natriuretic peptide (NT-proBNP) levels, and ensure normal results in blood and urine tandem mass spectrometry, as well as normal autoimmune antibodies related to rheumatic diseases. Additionally, some patients undergo cardiac magnetic resonance imaging, cardiac catheterization, and genetic testing to further exclude secondary factors. Conditions excluded in this diagnostic process include constrictive pericarditis, valvular heart disease, cardiotoxic exposure, cardiac tumors, rheumatic or metabolic disorders, and significant left ventricular hypertrophy with restrictive filling dysfunction. This study was approved by the Institutional Review Board of Shanghai Children’s Medical Center (SCMCIRB-K2023079-1) and registered with the Chinese Clinical Trial Registry (ChiCTR2300071701). Informed consent was not required because the data used in this retrospective, observational study were deidentified.

Data collection

Clinical characteristics of the enrolled children at their first hospitalization were retrospectively collected. These included demographics, clinical symptoms, the New York Heart Association functional classification or modified Ross heart failure classification for children (NYHA/Ross FC), biomarkers such as BNP or NT-proBNP, cardiac troponin T or I (cTn), and creatine kinase-MB. Genetic results from whole-exome sequencing (conducted in recent years) or targeted cardiomyopathy gene panels (used in earlier years) based on Illumina platforms, along with data from echocardiography, electrocardiography (ECG), and late gadolinium enhancement (LGE) cardiac magnetic resonance imaging, were also collected. Height, weight, and echocardiographic variables were standardized to Z-scores using the Boston Children’s Hospital Z-Score Calculator (https://zscore.chboston.org). Similarly, ECG variables were standardized to Z-scores to identify abnormal changes [12]. To ensure comparability among patients from different centers, biomarker results (e.g., BNP and NT-proBNP) were standardized by dividing by the upper limit of their normal reference range, thus expressing them as ratios and eliminating differences caused by varying assay methods. According to the standards of the American College of Medical Genetics and Genomics, patients with pathogenic or likely pathogenic variants in genes associated with RCM are classified as having a positive genotype. Those with variants that are likely benign or benign are classified as having a negative genotype. Variants of uncertain significance and those with an unclear relationship to RCM are defined as inconclusive [4, 13]. Based on previous reports, clinical experience, and expert consultation, the better analysis of ST-segment elevations and depressions in a 12-lead-electrocardiogram RS-scaled (BASEL-RS) was used to quantify the degree of QRS amplitude-corrected ST-segment changes in each patient [14,15,16]. For pediatric patients undergoing right heart catheterization, a regression equation combined catheter-measured pulmonary artery systolic pressure with echocardiographic estimates of right ventricular systolic pressure to calculate estimated pulmonary artery systolic pressure (ePASP). If tricuspid regurgitation velocity was undetectable, ePASP was assumed to be 20 mmHg [17]. Patients with more than 20% missing data at initial diagnosis were excluded from the study.

Outcomes and follow-up

The primary endpoint was a composite outcome of MACE, including cardiac death, HTx, or equivalent events such as the appropriate use of an implantable cardioverter defibrillator (ICD), extracorporeal membrane oxygenation (ECMO), or sustained ventricular tachycardia associated with hemodynamic compromise [5, 18, 19]. Experienced cardiologists at each center verified clinical outcomes and administered treatments classified as MACE. Patients were followed until the first occurrence of any defined MACE or until the last follow-up on August 1, 2023. Patients without follow-up data were excluded from the study. The study process is summarized in Fig. 1.

Fig. 1

Study flowchart. MACE, major adverse cardiovascular events

Statistical analysis

Continuous variables were described as mean ± standard deviation or median with interquartile range (IQR), as appropriate. Categorical variables were presented as frequencies and percentages. Group comparisons were conducted using the t-test or Mann–Whitney U test for continuous variables, and the χ² test or Fisher’s exact test for categorical variables. The Kaplan–Meier method assessed the rate of developing MACE. A two-sided P-value of < 0.05 was considered statistically significant. All analyses were performed using R version 4.3.3 (R Foundation).

Model development

The risk model was developed using data from patients whose MACE outcomes were known 1 year post-diagnosis. Survivors and individuals lost to follow-up with a follow-up period of less than 1 year were excluded, as their outcomes could not be determined at the 1-year mark (Fig. 1). The remaining cases were included in the model development. Continuous variables were adjusted as needed, based on graphical analyses of the log odds of MACE across clinically relevant ranges. Thirteen candidate variables were identified by plotting restricted cubic spline curves, which showed a monotonic trend with MACE (Additional file 1: Table S1 and Figure S2). Missing data were assumed to be missing at random and were addressed using multiple imputations with chained equations, generating 25 imputed datasets [20]. The least absolute shrinkage and selection operator (LASSO) penalty was applied to each dataset to identify predictors strongly correlated with MACE [21]. Variable selection was performed independently in each of the 25 datasets using LASSO regression. To ensure accurate estimation of regression coefficients, the model included four predictors, allowing for more than ten MACE events per variable [22]. The final prediction model was developed by applying logistic regression to each imputed dataset, with the estimates combined according to Rubin’s rules [23]. The probability of an individual patient experiencing MACE 1 year after diagnosis was calculated using the equation below:

$$\mathrm P=\frac{\mathrm e\mathrm x\mathrm p(\mathrm p\mathrm r\mathrm o\mathrm g\mathrm n\mathrm o\mathrm s\mathrm t\mathrm i\mathrm c\:\mathrm i\mathrm n\mathrm d\mathrm e\mathrm x)}{1+\mathrm e\mathrm x\mathrm p(\mathrm p\mathrm r\mathrm o\mathrm g\mathrm n\mathrm o\mathrm s\mathrm t\mathrm i\mathrm c\:\mathrm i\mathrm n\mathrm d\mathrm e\mathrm x)},$$

where the prognostic index represents the sum of each predictor multiplied by its associated coefficient.

Bootstrapping was used to evaluate the final model’s performance, generating 200 bootstrap samples from each imputed dataset and combining their estimates [24]. Model performance was assessed using the optimism-adjusted C statistic and Brier score [25]. Model calibration was evaluated graphically by comparing predicted and observed MACE rates across the cohort. The average calibration slope of the bootstrap samples estimated the degree of optimism. The model’s clinical utility for risk stratification was assessed by categorizing patients into risk groups: (1) lower-risk, with a predicted 1-year MACE risk of < 15%; (2) intermediate-risk, with a predicted risk between 15 and 40%; and (3) higher-risk, with a predicted risk of > 40%. A decision curve analysis was performed to evaluate the model’s clinical benefit, balancing conservative treatment against HTx preparation. The prediction model was developed in accordance with the TRIPOD statement [26].

Results

Baseline characteristics and outcomes

The cohort consisted of 185 patients with a median follow-up of 1.4 years (IQR, 0.5–3.4), totaling 406 patient-years. The median age at diagnosis was 5.4 years (IQR, 3.1–9.4), and 110 patients (59.5%) were male. All patients were of Asian ethnicity, with 10 individuals (5%) reporting a family history of cardiomyopathy. Children diagnosed with RCM showed significant heart failure symptoms, most commonly fatigue and hepatomegaly, along with markedly elevated BNP or NT-proBNP levels. Additionally, 43% of the patients were classified as NYHA/Ross FC III-IV. Table 1 presents a comparison of baseline characteristics among 175 individuals with a confirmed 1-year outcome post-diagnosis. It was observed that those who experienced MACE within 1 year were more likely to have hepatomegaly, a history of syncope, poorer cardiac functional classification, elevated BNP or NT-proBNP levels, more pronounced ST-segment changes on ECG, and echocardiographic findings of a larger left atrial long axis and lower fractional shortening.

Table 1 Baseline characteristics of patients in the overall cohort, stratified by MACE at 1 year

Of the 123 subjects (66%) who underwent genetic testing, 81 (66%) had a positive genotype: 70 had variants in sarcomeric genes, with TNNI3 being the most prevalent (n = 49, 61% of positive genotypes), and 11 had variants in cytoskeletal or cytosolic genes. The remaining 42 (34%) had a negative genotype. All positive variants were heterozygous, and no patient carried multiple positive variants (Fig. 2). Genetic testing rates and mutation distribution across different regions are shown in Additional file 1: Figure S3, and detailed variants of patients with positive and inconclusive genotypes are presented in Additional file 1: Table S2.

Fig. 2

Genetic testing outcomes in pediatric restrictive cardiomyopathy (123 out of 185). The inner circle displays the distribution of positive, negative and inconclusive genotypes among patients who underwent genetic testing. The outer ring highlights the distribution of variant genes within the positive genotypes. Oceanic-gradient segments in the diagram denote sarcomeric genes, while sunlit-sands segments indicate cytoskeletal/cytosolic genes. MACE, major adverse cardiovascular events

In the pharmacological treatment of pediatric RCM in China, nearly every child received diuretics. This was followed by the use of angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, or angiotensin receptor neprilysin inhibitors, which were prescribed to about half of the patients. Approximately 30% of the children were also treated with beta-blockers and antiplatelet medications, such as aspirin or clopidogrel. A small number of children received calcium channel blockers and oral medications for reducing pulmonary arterial hypertension. Comparison between patients with and without MACEs at 1 year revealed no differences in the use of any class of medication between the two groups (Table 1).

During the follow-up period, 114 patients (62%) experienced MACE, with the median MACE-free survival time for the entire cohort being 2.1 years post-diagnosis (IQR, 0.6–5.4) (Fig. 3A). Cardiac death occurred in 74 patients (65%), while 32 patients (28%) underwent HTx. Eight patients reached an equivalent endpoint: five received an ICD (4%) due to recurrent syncope and evidence of myocardial ischemia; two experienced sustained ventricular tachycardia with hemodynamic compromise (2%), discovered during hospital monitoring and survived after resuscitation; and one underwent ECMO therapy (1%). Overall, 28% of patients (95% CI, 23%–33%) experienced a MACE each year. The cohort’s 1-, 2-, and 5-year MACE-free survival rates were 70%, 51%, and 29%, respectively. Further analysis indicated that overall patient prognosis was not significantly associated with sex, completion of genetic testing, or genetic phenotype (Additional file 1: Figure S4). However, landmark analysis suggested that patients with TNNI3 mutations might experience a lower MACE-free survival rate than those with other mutations, starting from 1.5 years post-diagnosis. And LGE-positive patients might have a lower MACE-free survival rate than LGE-negative patients, starting from nine months post-diagnosis (Additional file 1: Figure S5).

Fig. 3

Survival of the cohort and performance of the prediction model. A MACE-free survival for the study cohort. B Calibration plot illustrating the agreement between predicted and observed 1-year MACE risk for one imputed dataset. The spike histogram on the x-axis represents the number of patients with a predicted risk at each x-axis value. C Patients were stratified into 3 predicted risk groups (< 15%, 15%–40%, > 40%) using the 1-year MACE model, and the MACE-free survival probability within 5 years post-baseline assessment was plotted for one imputed dataset. D Decision curve analysis demonstrates a high net benefit across most HTx treatment thresholds, ranging from 8.5% to 78.3%. MACE, major adverse cardiovascular events; HTx, heart transplantation

Model development

Data from 175 patients, excluding ten survivors with a follow-up period of less than 1 year and including 55 events, were used for model development. A summary of missing data for the variables is presented in Additional file 1: Table S3. LASSO regression was performed separately on each imputed dataset, and the following four predictors were consistently selected across all 25 iterations as the final predictors: NYHA/Ross FC III-IV, ln(BNP or NT-proBNP ratio), adjusted cTn ratio [assigned a value of 0 if the cTn ratio is < 1; otherwise, converted to its natural logarithm, ln(cTn ratio)], and BASEL-RS. Estimates of odds ratios for the predictors and sensitivity analyses are detailed in Table 2. For an individual pediatric patient with RCM, the risk of experiencing MACE 1 year after diagnosis was calculated using the following equation:

Table 2 Pediatric restrictive cardiomyopathy risk prediction model and sensitivity analyses

$$\mathrm P\left(\mathrm M\mathrm A\mathrm C\mathrm E\:\mathrm a\mathrm t\:1\:\mathrm y\mathrm e\mathrm a\mathrm r\right)=\frac{\exp\left(\mathrm p\mathrm r\mathrm o\mathrm g\mathrm n\mathrm o\mathrm s\mathrm t\mathrm i\mathrm c\:\mathrm i\mathrm n\mathrm d\mathrm e\mathrm x\right)}{1+\exp\left(\mathrm p\mathrm r\mathrm o\mathrm g\mathrm n\mathrm o\mathrm s\mathrm t\mathrm i\mathrm c\:\mathrm i\mathrm n\mathrm d\mathrm e\mathrm x\right)},$$

where the prognostic index = 0.884 × (NYHA/Ross FC III-IV) + 0.885 × ln(BNP or NT-proBNP ratio) + 0.781 × (cTn ratio adjusted) + 1.168 × (BASEL-RS) − 4.644.

The optimism-adjusted C statistic was 0.78 (95% CI, 0.72–0.85), and the optimism-adjusted Brier score was 0.17 (95% CI, 0.14–0.21). Internal validation using bootstrapping showed a calibration slope of 0.90 (95% CI, 0.63–1.27), indicating minor over-optimism. Figure 3B illustrates the calibration, demonstrating good overall agreement between predicted and observed 1-year risk. The prediction model effectively stratified patient risk (Fig. 3C). It also showed a high net benefit across most treatment thresholds for HTx therapy (8.5% to 78.3%), suggesting its clinical superiority (Fig. 3D).

Discussion

Childhood RCM is rare, accounting for less than 5% of pediatric cardiomyopathy cases [27, 28], leading to a scarcity of large-scale, high-quality cohort studies in the past decade. Our study, involving 185 pediatric RCM patients, characterizes their clinical features at diagnosis and serves as a robust validation and complement to the Pediatric Cardiomyopathy Registry study from over a decade ago [5]. Our findings in the Chinese population align closely with those from the USA and Canada. Both studies observed a young median age at diagnosis—5.4 years in our study versus 6.2 years in the Registry study. Approximately 40% of patients showed congestive heart failure, with 43% classified as NYHA/Ross FC III-IV, comparable to the 37% in the Registry study. Baseline echocardiographic characteristics were similar, with atrial enlargement but normal left ventricular dimensions and systolic function. Both studies noted poor prognosis, with a median event-free survival of about 2 years and a 5-year event-free survival rate of around 30%. Our analysis also revealed conduction abnormalities and ST-segment depression in some patients, suggesting myocardial ischemia, which may contribute to cardiac events [16, 29]. LGE on cardiac magnetic resonance indicates severe myocardial fibrosis or scarring, linked to adverse outcomes in RCM [30]. Our research suggests that patients with LGE face a higher risk of MACE mid-to-long-term post-diagnosis. However, nearly 40% of patients could not complete this examination at their initial visit, potentially affecting result interpretation. Further research is needed to confirm LGE’s prognostic role in pediatric RCM.

This study represents the first large-scale genetic analysis of pediatric RCM. In our cohort, nearly two-thirds of patients in mainland China underwent genetic testing, with the highest rates in East China and the lowest in the Northwest, reflecting regional disparities in economic and healthcare resources. Among those tested, about two-thirds had a positive genotype, primarily involving sarcomeric gene mutations, with fewer cytoskeletal or cytosolic mutations. TNNI3 was the most common mutation across regions in the Chinese cohort. A previous study from Australia found that pediatric RCM has a genetic detection rate of up to 80%, with one-third of the cases carrying TNNI3 mutations [31]. Similarly, a pediatric RCM cohort from Japan reported that TNNI3 mutations were the most common, accounting for approximately 30% of the cases [32]. Therefore, we speculate that TNNI3 mutations are likely the most prevalent in pediatric RCM. However, further research is needed to explore gene mutation distributions in other ethnic groups. Due to the rapid progression of RCM, our subgroup analysis showed no significant differences among genotypes. Landmark analysis indicated that children with TNNI3 mutations might have poorer mid- to long-term outcomes compared to those with other mutations. Interestingly, this finding is similar to the conclusion from the Australian cohort, which found that TNNI3 mutations were associated with more severe clinical outcomes in pediatric cardiomyopathy patients [31]. However, the small sample size may affect the robustness of this conclusion. Despite these challenges, the study highlights the critical role of genetic testing in diagnosing RCM and underscores the need for more extensive and diverse studies to validate these findings.

We developed a new prediction model, which is not yet registered, that combines clinical symptoms, biomarkers, and ECG data to effectively identify patients at increased risk of MACE 1 year after diagnosis. Our study underscores the crucial role of heart failure and myocardial ischemia in risk stratification for pediatric RCM. Key clinical indicators such as NYHA/Ross FC and BNP or NT-proBNP are commonly used to assess heart failure severity [33]. Elevated cTn levels indicate myocardial injury, highlighting ongoing damage and their importance in risk assessment. While previous studies have noted ST-segment depression in cardiomyopathy patients, they were limited to qualitative ECG analyses and did not assess the severity of these changes [16, 34, 35]. We employed the BASEL-RS scoring method to evaluate ST-segment depression in children with RCM quantitatively, demonstrating its strong risk stratification capabilities. These factors collectively emphasize their critical role in RCM risk assessment, illustrating the disease’s pathophysiological progression toward heart failure and myocardial ischemia.

HTx is an effective treatment for end-stage pediatric RCM [4, 36]. However, in China, the primary approach for treating children with end-stage RCM is medical management focused on heart failure, due to the scarcity of donors. Pediatric HTx began later than adult transplantation, and the growth in the number of surgeries has been relatively slow, resulting in a low probability of end-stage children receiving a heart transplant. The waiting time and queuing mechanism for pediatric HTx in China rely on the China Organ Transplant Response System, which follows the “children first” principle: children under 18 with end-stage heart failure who depend on mechanical assistance (such as ECMO) are prioritized [37]. Emergency cases can be matched within a few hours, while non-emergency children typically wait an average of 1–6 months. ECMO is currently the primary means of mechanical circulatory support for children with severe heart failure in China who do not respond well to medication, serving as a bridge to emergency HTx. In contrast, long-term mechanical circulatory devices, such as ventricular assist devices, have not yet been widely adopted for pediatric RCM patients in China [38]. Additionally, for children with RCM who exhibit evidence of myocardial ischemia and syncope, the implantation of an ICD may be considered to prevent sudden death caused by malignant ventricular arrhythmias [4]. However, this intervention does not slow the progression of heart failure associated with RCM. Given the current status of pediatric HTx—including surgical risks, postoperative care, rehabilitation, and the costs incurred by the patient’s family—clinicians often prioritize the use of conservative medications to prolong the function of the native heart before HTx. However, delaying HTx beyond the optimal window can increase reliance on inotropic drugs, mechanical ventilation, and circulatory support, raising the mortality risk during the wait [19]. Therefore, precisely evaluating the risk level for individual patients is essential. This model effectively categorizes patients into lower, intermediate, and higher-risk groups by assessing their risk 1 year after diagnosis. Although these probability divisions do not establish specific clinical decision thresholds, they do help identify high-risk patients, enabling targeted interventions such as risk management, close monitoring, and HTx strategies. This proactive approach encourages collaborative decision-making between clinicians and patients’ families, ultimately enhancing the prognosis for pediatric RCM.

Limitations

Our study population was drawn from academic centers across China. All patients were Asian, so our results should be extrapolated with caution to patients from other racial or ethnic backgrounds. We acknowledge that the genetic testing methodologies used in our study have inherent limitations. The sequencing platforms employed varied over time; earlier cases often used targeted cardiomyopathy gene panels, which may not have included all relevant genes. In contrast, more recent cases utilized whole-exome sequencing, providing a broader genetic analysis. Moreover, differences in the gene panels used across various centers and genetic testing companies may have led to inconsistencies in the genetic data obtained. These variations in testing approaches could result in the under- or over-detection of genetic variants.

This was a retrospective cohort study with variable patient follow-up times, and some patients may have been lost to follow-up before experiencing an initial MACE. Despite being the largest pediatric RCM cohort to date, the rarity of the disease limited the sample size. Given the extremely poor prognosis of pediatric RCM and these sample size limitations, our current data only support the construction of a 1-year post-diagnosis prediction model. Longer-term prediction models or those incorporating genotypic data cannot be developed with the current dataset. However, we believe that at this stage, a 1-year prediction model is sufficient to provide guidance for decision-making regarding HTx. In China, at least, 1 year is adequate to complete the necessary preparations. Recent guidelines recommend that external validation ideally be conducted by independent investigators not involved in the original model development and that it should be distinct from the development process [39]. Also, even if external validation is performed within this Chinese cohort, further validation is necessary when extending applications to other ethnic populations. To ensure modeling efficacy, we employed the bootstrapping method for internal validation, and external validation by different teams is needed to assess the model’s generalizability further.

As in similar studies, we selected MACE as the composite endpoint to ensure more than ten events per variable in the prediction model [18, 40]. Unlike cardiac death, HTx may be influenced by clinician preferences and donor availability, potentially introducing bias. Sensitivity analysis showed that NYHA/Ross FC, BNP or NT-proBNP, and cTn remained statistically significant even after excluding patients for whom HTx was the endpoint, indicating these variables have a strong impact on predicting cardiac death (Table 2). However, the significance of BASEL-RS diminished when HTx cases were excluded, suggesting its initial relevance might have been partly dependent on these events. We hypothesize that abnormal ST-segment changes may prompt some clinicians to decide on earlier HTx. Given the clinical importance and similar significance of each endpoint in decision-making and patient prognosis, we consider MACE a suitable observational endpoint. Future research should focus on analyzing these endpoints individually.

This study is the first to propose a risk prediction model for pediatric RCM. However, since the RCM patient group is inherently at a very high risk of MACE, the cut-off values for lower, intermediate, and higher risk divisions in this study are used solely to validate the model’s discriminatory ability, rather than to define thresholds for absolute RCM risk. Consequently, the current model cannot provide critical thresholds to support actual clinical decision-making, such as transplant listing. There remains a lack of large-scale research data on pediatric RCM, and defining absolute risk thresholds for this condition requires further exploration in future studies. Nonetheless, the model can calculate and identify patients with a high probability of MACE. Clinicians can consider the actual clinical situation and the MACE probability, along with the patient’s family’s willingness for transplantation, to jointly discuss and formulate the next treatment plan.

Conclusions

In this study, we analyzed the national cohort of children with RCM in China and identified heart failure as a clinical characteristic, with TNNI3 mutations being the most common genetic factor. We developed a new prediction model that utilizes accessible clinical data to provide personalized assessments of the risk of MACE in pediatric RCM patients. It has the potential to enhance precision in healthcare strategies and guide decisions regarding HTx for this rare but significant form of cardiomyopathy in children.