To the Editor: Owing to the heterogeneity of congenital heart disease-associated pulmonary hypertension (CHD-PH) disease and the development of the pulmonary vascular system in pediatric patients, the management of CHD-PH in children is significantly different from that in adults.[1] Early identification and intervention of CHD-PH are of great significance for disease management in children with CHD, and non-invasive, accurate clinical tools are urgently needed to predict PH risk in children with CHD. However, the applicability of the current risk-scoring system to pediatric CHD-PH remains unclear. Furthermore, there are presently no user-friendly, non-invasive online tools available for predicting CHD-PH, and there is a scarcity of related studies in Asian populations. This study aimed to develop and validate the prediction tools for estimating the risk of pulmonary hypertension and clinical deterioration in children with congenital heart disease. A total of 1983 children with CHD diagnosed in the Children's Hospital of Chongqing Medical University, the First Affiliated Hospital of Guangzhou Medical University, and the Guangdong Women and Children Hospital in China from January 2016 to June 2022 were included. The inclusion criteria were clinically diagnosed CHD patients aged 3 months to 7 years and underwent right heart catheterization (RHC) and interventional therapy or heart repair operation according to a scientific statement from the American Heart Association. The exclusion criteria were as follows: (1) left obstructive lesions, including pulmonary vein obstruction, mitral stenosis, cor triatriatum, obstruction of the left ventricular outflow tract and aortic valve, and aortic coarctation, (2) Eisenmenger syndrome, and (3) lack of data on RHC and/or echocardiography. A CHD-PH model was designed to predict the risk of PH in pediatric CHD non-invasively. A total of 1517 patients diagnosed between January 2016 and December 2019 were assigned to the training cohort for the development of the prediction model. The 466 cases diagnosed between January 2020 and June 2022 were assigned to the validation cohort. PH was defined as an age of >3 months with a mean pulmonary artery pressure (mPAP) of >20 mmHg, as determined by RHC.[2] Clinical deterioration was defined as one of the following events occurring within one month after cardiac repair: (1) death, (2) use of norepinephrine and adrenaline, (3) respiratory failure, and (4) heart failure according to the Age-based Ross classification for heart failure in children.[3] Categorical variables are presented as frequencies with percentages, and continuous variables are presented as median (interquartile range [IQR]). Univariate logistic regression was used to screen variables significantly related to CHD-PH and clinical deterioration in the training cohort. Predictive variables with P <0.05 were applied to the multivariable logistic regression model. A multivariable logistic regression of backward stepwise selection and Akaike's information criterion were used to select significant predictors for nomogram construction. All variance inflation factors were <4, indicating no collinearity among the screened variables. The consistency index, the area under the receiver operating characteristic (ROC) curve (AUC), calibration curve, and decision curve analysis (DCA) were used to evaluate the predictive ability of the nomogram. The Hosmer-Lemeshow goodness-of-fit test is used to evaluate the calibration of the model in multivariable logistic regression analysis, and a calibration curve is plotted. For missing data, choose multiple imputation method to interpolate missing values. The CHD-PH non-invasive predictive nomogram was developed as an online tool (https://chd-ph.shinyapps.io/CHD-PH/). The nomogram to predict clinical deterioration was constructed as a practical and accurate online tool (https://chd-ph-pre.shinyapps.io/CHD-AE/). A value of P <0.05 was considered statistically significant. A total of 1983 children with CHD were recruited, with a median age of 2 (2–4) years and a female proportion of 59.3% (1175/1983). Among them, 701 (35.4%) children with CHD-PH were diagnosed by RHC, with a median age of 2 (1–3) years. Additionally, 93.9% [658/701] of CHD-PH patients had Ross class I or II, despite the presence of PH in pediatric patients, most of them have good cardiac functional status and mild symptoms of heart failure. The number of female children with CHD-PH was 1.68 times higher than that of male children, and the mPAP of female children was higher than that of male children (28 [24–32] mmHg vs. 26 [24–30] mmHg, P = 0.009). Down syndrome was the most common chromosomal disease, affecting 3.7% (26/701) of children with CHD-PH. Atrial septal defect (34.7%, 688/1983), ventricular septal defect (32.7%, 650/1983), and patent ductus arteriosus (PDA) (92.0%, 1825/1983) were the most common disease subtypes. In this study, 143 children with CHD had events of clinical deterioration, including deaths due to pulmonary hypertensive crisis (2.1%), respiratory failure (21%), and heart failure (51.7%). Supplementary Tables 1 and 2, https://links.lww.com/CM9/B951 summarize the clinical characteristics of the training and validation cohorts. The results of the univariate and multivariable logistic analysis of pediatric CHD-PH are shown in Supplementary Table 3, https://links.lww.com/CM9/B951. In multivariable analysis, left ventricular hypertrophy (odds ratio [OR], 3.315; 95% confidence interval [CI], 2.126–5.171; P <0.001), right ventricular hypertrophy (OR, 2.887; 95% CI, 1.556–5.357; P = 0.001), growth retardation (OR, 2.101; 95% CI, 1.162–3.799; P = 0.014), PDA vs. PDA with other cardiac (OR, 2.552; 95% CI, 1.923–3.388; P <0.001), Down syndrome (OR, 6.742; 95% CI, 1.822–24.948; P = 0.004), age (OR, 0.818; 95% CI, 0.750–0.893; P <0.001), echocardiographic right ventricular systolic pressure (RVSP) (OR, 1.065; 95% CI, 1.036–1.094; P <0.001), platelets (OR, 0.998; 95% CI, 0.997–1.000; P = 0.025), and albumin levels (OR, 0.938; 95% CI, 0.903–0.975; P = 0.001) were identified as independent predictors of pediatric CHD-PH. A nomogram for non-invasive CHD-PH prediction, with the independent predictors identified above, was generated [Supplementary Figure 1A, https://links.lww.com/CM9/B951]. The calibration curve of the CHD-PH nomogram showed a high consistency between predicted and observed probabilities [Supplementary Figure 1B and Supplementary Figure 2A, https://links.lww.com/CM9/B951], demonstrating good calibration. The AUC values were 0.747 (95% CI, 0.721–0.772) and 0.756 (95% CI, 0.706–0.806), respectively [Supplementary Figure 1C and Supplementary Figure 2B, https://links.lww.com/CM9/B951]. The sensitivity and specificity of the nomogram were 64.3% and 73.8%, respectively, in the training cohort; and in the validation cohort, they were 57.2% and 82.9%, respectively. The DCA curve showed that the nomogram added more net benefits for almost all threshold probabilities [Supplementary Figure 1D and Supplementary Figure 2C, https://links.lww.com/CM9/B951], illustrating that the nomogram could predict the probability of CHD-PH occurrence with improved clinical benefits and practicality. The nomogram predicted the probabilities of CHD and CHD-PH groups as follows: 0.245 (IQR, 0.183–0.397) vs. 0.455 (IQR, 0.272–0.694) in the training cohort [Supplementary Figure 1E, https://links.lww.com/CM9/B951], and 0.192 (IQR, 0.149–0.263) vs. 0.331 (IQR, 0.206–0.682) in the validation cohort [Supplementary Figure 2D, https://links.lww.com/CM9/B951] (P <0.001 for both). The PH probability in the training cohort was 59.9% (348/581) for high-risk patients and 23.0% (215/936) for low-risk patients, and 57.4% (78/136) for high-risk patients and 18.2% (60/330) for low-risk patients in the validation set [Supplementary Figure 1F and Supplementary Figure 2E, https://links.lww.com/CM9/B951]. Supplementary Figure 2F, https://links.lww.com/CM9/B951 gives an example of application of dynamic CHD-PH nomogram. The univariable and multivariable logistic regression analysis results of clinical deterioration are exhibited in Supplementary Table 4, https://links.lww.com/CM9/B951. In multivariable logistic regression analysis, hemoglobin (OR, 0.964; 95% CI, 0.947–0.983; P <0.001), albumin (OR, 0.791; 95% CI, 0.748–0.837; P <0.001), mPAP (OR, 1.045; 95% CI, 1.025–1.065; P <0.001), pulmonary arterial oxygen saturation (OR, 0.838; 95% CI, 0.750–0.935; P = 0.002), red cell distribution width (RDW) (OR, 1.244; 95% CI, 1.086–1.426; P = 0.002), and creatinine (OR, 1.038; 95% CI, 1.006–1.071; P = 0.021) were significantly associated with clinical deterioration in children with CHD. Six independent predictors were used to construct a nomogram to predict the clinical deterioration of pediatric CHD [Supplementary Figure 1G, https://links.lww.com/CM9/B951]. The calibration curves showed good consistency between predicted and observed events of clinical deterioration [Supplementary Figure 1H and Supplementary Figure 3A, https://links.lww.com/CM9/B951], and the Hosmer–Lemeshow test indicated no significant deviations (χ2 = 59.687, P = 0.100; and χ2 = 35.078, P = 0.725, respectively). The AUCs of the training and validation cohorts were 0.845 (95% CI, 0.796–0.893) [Supplementary Figure 1I, https://links.lww.com/CM9/B951] and 0.753 (95% CI, 0.668–0.839) [Supplementary Figure 3B, https://links.lww.com/CM9/B951], respectively. The sensitivity and specificity of the nomogram were 75.0% and 82.2%, respectively, in the training cohort; and in the validation cohort, they were 61.7% and 82.8%, respectively. The DCA curve demonstrated that the nomogram exhibited a positive net benefit [Supplementary Figure 1J and Supplementary Figure 3C, https://links.lww.com/CM9/B951]. The nomogram scores of children without clinical deterioration and those with clinical deterioration in the training cohort were significantly different (0.078 [IQR, 0.053–0.119] vs. 0.181 [IQR, 0.085–0.624], P <0.001; Supplementary Figure 1K, https://links.lww.com/CM9/B951), which was confirmed in the validation cohort (0.090 [IQR, 0.064–0.130] vs. 0.154 [IQR, 0.097–0.389], P <0.001; Supplementary Figure 3D, https://links.lww.com/CM9/B951). The probability of clinical deterioration in the training cohort was 66.7% (64/96) for high-risk patients and 0.57% (8/1413) for low-risk patients, and 80.9% (38/47) for high-risk patients and 0.72% (3/416) for low-risk patients in the validation set [Supplementary Figure 1L and Supplementary Figure 3E, https://links.lww.com/CM9/B951]. Supplementary Figure 3F, https://links.lww.com/CM9/B951 gives an example of application of dynamic nomogram to predict the clinical deterioration of pediatric CHD. Usage of the web server: on the left side, patient characteristics are displayed. Select subgroups from the drop-down menu and adjust the progress bar to set indicator values. Once the relevant features are selected, click the "Predict" button. The right side will display disease risk predictions along with a 95% CI. Click "Numeric Summary" and "Model Summary" to access data tables and a comprehensive overview of the model [Supplementary Figures 2G and Supplementary Figures 3G, https://links.lww.com/CM9/B951. This study use multicenter data to develop and externally validate a convenient and practical nomogram to predict endpoints of clinical deterioration of pediatric CHD and a subset of them who may develop CHD-PH. Multivariable analysis identified nine independent predictors (left ventricular hypertrophy, right ventricular hypertrophy, growth retardation, PDA, Down syndrome, age, echocardiographic RVSP, platelets, and albumin) of pediatric CHD-PH. Moreover, our findings demonstrated that hemoglobin, albumin, mPAP, pulmonary arterial oxygen saturation, RDW, and creatinine were risk factors for clinical deterioration in pediatric CHD. Additionally, this study included a large cohort of pediatric CHD patients. The prediction nomograms showed good discrimination and calibration performance, and we further developed online dynamic web servers to make them convenient and valuable tools in clinical practice. The nomograms have good application scenarios, including long-term management of pediatric CHD in primary healthcare institutions and early identification of high-risk patients for early referral, evaluation, and treatment. These online tools could be integrated with a hospital case management system and used for artificial intelligence-aided diagnosis. Machine learning and feedback optimization in subsequent application processes will further improve the efficiency and accuracy of clinical predictions. This study has some limitations. First, multivariable models did not include some potential predictors, such as N-terminal pro-B-type natriuretic peptide and tricuspid annular plane systolic excursion, because this information was not uniform in the retrieved datasets. Second, young children cannot reliably perform the 6-minute walking distance test, widely used to measure motor capacity in adult CHD studies but is unsuitable for young children. In conclusion, this research may be important for individual monitoring, risk assessment, and treatment decisions for pediatric CHD. Ethical approval Ethical approval was approved by the Children's Hospital of Chongqing Medical University, with the number (2022) Ethics Review (Research) No. (8). All patients' guardian were informed and signed the written informed consent forms. Funding This project was supported by grants from the National Natural Science Foundation of China (Nos. 82000034, 82120108001, and 82170069), Chongqing Postdoctoral International Exchange Training Program (No. 2021JLPY001), National Clinical Medical Research Center (No. NCRC-2022-GP-08), Guangdong Department of Science and Technology (No. 2022A0505030017), Basic Science and Application of Guangzhou Science and Technology Plan (No. 202201010069). Conflicts of interest None.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2024) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: