Machine Learning Models for ASCVD Risk Prediction in an Asian Population - How to Validate the Model is Important.

Introduction: Atherosclerotic cardiovascular disease (ASCVD) is prevalent worldwide including Taiwan, however widely accepted tools to assess the risk of ASCVD are lacking in Taiwan. Machine learning models are potentially useful for risk evaluation. In this study we used two cohorts to test the feasibility of machine learning with transfer learning for developing an ASCVD risk prediction model in Taiwan. Methods: Two multi-center observational registry cohorts, T-SPARCLE and T-PPARCLE were used in this study. The variables selected were based on European, U.S. and Asian guidelines. Both registries recorded the ASCVD outcomes of the patients. Ten-fold validation and temporal validation methods were used to evaluate the performance of the binary classification analysis [prediction of major adverse cardiovascular (CV) events in one year]. Time-to-event analyses were also performed. Results: In the binary classification analysis, eXtreme Gradient Boosting (XGBoost) and random forest had the best performance, with areas under the receiver operating characteristic curve (AUC-ROC) of 0.72 (0.68-0.76) and 0.73 (0.69-0.77), respectively, although it was not significantly better than other models. Temporal validation was also performed, and the data showed significant differences in the distribution of various features and event rate. The AUC-ROC of XGBoost dropped to 0.66 (0.59-0.73), while that of random forest dropped to 0.69 (0.62-0.76) in the temporal validation method, and the performance also became numerically worse than that of the logistic regression model. In the time-to-event analysis, most models had a concordance index of around 0.70. Conclusions: Machine learning models with appropriate transfer learning may be a useful tool for the development of CV risk prediction models and may help improve patient care in the future.

A risk stratification model modified from the U.S. guideline could be applied in an Asian population with or without ASCVD.

AIM: This study aimed to evaluate the performance of a modified US (MUS) model for risk prediction of cardiovascular (CV) events in Asian patients and compare it to European and Japanese models. MATERIAL AND METHODS: The MUS model, based on the US ACC/AHA 2018 lipid treatment guideline, was employed to stratify patients under primary or secondary prevention. Two multi-center prospective observational registry cohorts, T-SPARCLE and T-PPARCLE, were used to validate the scoring system, and the primary outcome was the time to first occurrence/recurrence of major adverse cardiac events (MACEs). The MUS model's performance was compared to other models from Europe and Japan. RESULTS: A total of 10,733 patients with the mean age of 64.2 (SD: 11.9) and 36.5% female were followed up for a median of 5.4 years. The MUS model was validated, with an AUC score of 0.73 (95% CI 0.68-0.78). The European and Japanese models had AUC scores ranging from 0.6 to 0.7. The MUS model categorized patients into four distinct CV risk groups, with hazard ratios (HRs) as follows: very high-vs. high-risk group (HR=1.91, 95% CI 1.53-2.39), high-vs. moderate-risk group (HR=2.08, 95% CI 1.60-2.69), and moderate-vs. low-risk group (HR=3.14, 95% CI 1.63-6.03). After adjusting for the MUS model, a history of ASCVD was not a significant predictor of adverse cardiovascular outcomes within each risk group. CONCLUSION: The MUS model is an effective tool for risk stratification in Asian patients with and without ASCVD, accurately predicting MACEs and performing comparably or better than other established risk models. Our findings suggest that patient management should focus on background risk factors instead of solely on primary or secondary prevention.

Association of Major Adverse Cardiac Events and Beta-Blockers in Patients with and without Atherosclerotic Cardiovascular Disease: Long-Term Follow-Up Results of the T-SPARCLE and T-PPARCLE Registry in Taiwan.

Beta-blockers are widely used, but the benefit is now challenged in patients at risk of atherosclerotic cardiovascular disease (ASCVD) in the present coronary reperfusion era. We aimed to identify the risk factors of a major adverse cardiac event (MACE) and the long-term effect of beta-blockers in two large cohorts in Taiwan. Two prospective observational cohorts, including patients with known atherosclerosis cardiovascular disease (T-SPARCLE) and patients with at least one risk factor of ASCVD but without clinically evident ASCVD (T-PPARCLE), were conducted in Taiwan. The primary endpoint is the time of first occurrence of a MACE (cardiovascular death, nonfatal stroke, nonfatal myocardial infarction, and cardiac arrest with resuscitation). Between December 2009 and November 2014, with a median 2.4 years follow-up, 11,747 eligible patients (6921 and 4826 in T-SPARCLE and T-PPARCLE, respectively) were enrolled. Among them, 273 patients (2.3%) met the primary endpoint. With multivariate Cox PH model analysis, usage of beta-blocker was lower in patients with MACE (42.9% vs. 52.4%, p < 0.01). In patients with ASCVD, beta-blocker usage was associated with lower MACEs (hazard ratio 0.72; p < 0.001), but not in patients without ASCVD. The event-free survival of beta-blocker users remained higher during the follow-up period (p < 0.005) of ASCVD patients. In conclusion, in ASCVD patients, reduced MACE was associated with beta-blocker usage, and the effect was maintained during a six-year follow-up. Prescribing beta-blockers as secondary prevention is reasonable in the Taiwanese population.