RESUMO
A blood test identifying patients at increased risk of pulmonary hypertension (PH) could streamline the investigative pathway. The prospective, multicenter CIPHER study aimed to develop a microRNA-based signature for detecting PH in breathless patients and enrolled adults with a high suspicion of PH who had undergone right heart catheterization (RHC). The CIPHER-MRI study was added to assess the performance of this CIPHER signature in a population with low probability of having PH who underwent cardiac magnetic resonance imaging (cMRI) instead of RHC. The microRNA signature was developed using a penalized linear regression (LASSO) model. Data were modeled both with and without N-terminal pro-brain natriuretic peptide (NT-proBNP). Signature performance was assessed against predefined thresholds (lower 98.7% CI bound of ≥0.73 for sensitivity and ≥0.53 for specificity, based on a meta-analysis of echocardiographic data), using RHC as the true diagnosis. Overall, 926 CIPHER participants were screened and 888 were included in the analysis. Of 688 RHC-confirmed PH cases, approximately 40% were already receiving PH treatment. Fifty microRNA (from 311 investigated) were algorithmically selected to be included in the signature. Sensitivity [97.5% CI] of the signature was 0.85 [0.80-0.89] for microRNA-alone and 0.90 [0.86-0.93] for microRNA+NT-proBNP, and the corresponding specificities were 0.33 [0.24-0.44] and 0.28 [0.20-0.39]. Of 80 CIPHER-MRI participants with evaluable data, 7 were considered PH-positive by cMRI whereas 52 were considered PH-positive by the microRNA signature. Due to low specificity, the CIPHER miRNA-based signature for PH (either with or without NT-proBNP in model) did not meet the prespecified diagnostic threshold for the primary analysis.
RESUMO
BACKGROUND: This study aimed to develop a machine learning (ML) model to identify patients who are likely to have pulmonary hypertension (PH), using a large patient-level US-based electronic health record (EHR) database. METHODS: A gradient boosting model, XGBoost, was developed using data from Optum's US-based de-identified EHR dataset (2007-2019). PH and disease control adult patients were identified using diagnostic, treatment and procedure codes and were randomly split into the training (90%) or test set (10%). Model features included patient demographics, physician visits, diagnoses, procedures, prescriptions, and laboratory test results. SHapley Additive exPlanations values were used to determine feature importance. RESULTS: We identified 11,279,478 control and 115,822 PH patients (mean age, respectively: 62 and 68 years, both 53% female). The final model used 165 features, with the most important predictive features including diagnosis of heart failure, shortness of breath and atrial fibrillation. The model predicted PH with an area under the receiver operating characteristic curve (AUROC) of 0.92. AUROC remained above 0.80 for the prediction of PH up to and beyond 18 months before diagnosis. Among the PH patients, we also identified 955 pulmonary arterial hypertension (PAH) and 1432 chronic thromboembolic pulmonary hypertension (CTEPH) patients, and the range of AUROCs obtained for these cohorts was 0.79-0.90 and 0.87-0.96, respectively. CONCLUSIONS: This model to detect PH based on patients' EHR records is viable and performs well in subgroups of PAH and CTEPH patients. This approach has the potential to improve patient outcomes by reducing diagnostic delay in PH.