In this multicenter longitudinal cohort study, a simplified risk prediction model for MACE in LVNC was developed using four independent risk factors: age at diagnosis, NT-pro BNP levels, LA enlargement, and LVEF ≤ 40%. The model demonstrated excellent discrimination and calibration ability by both internal and external validations, and was further converted into a risk score, the ABLE-SCORE, allowing clinicians to identify high-risk patients and provide targeted management. Derived from easily accessible laboratory and echocardiographic variables, the ABLE-SCORE showed efficient predictive performance and satisfactory net benefits compared with the current risk score for MACE in LVNC [22], indicating its reliability and convenience in clinical application.
LVNC is a heterogeneous entity with relatively poor prognosis. A recent large scale meta-analysis has suggested that the cardiovascular and all-cause mortality rates of LVNC are approximately 25- and 5-fold higher than the general population [29]. However, current data regarding the prognosis and risk stratification of LVNC remains scarce. A risk prediction model for MACE has been previously reported, and it was composed of six variables: age, gender, cardiovascular risk factors, family aggregation by genetic testing, ECG abnormality, and LVEF on CMR [22]. The model has contributed important insights into the association between clinical/genetic information and MACE among LVNC individuals, but the implementation of genetic testing and CMR scans as regular procedures for risk assessment has precluded its widespread clinical use, especially in some developing countries and regions with limited healthcare resources. As a result, there is a crucial need to establish more simplified and widely applicable risk models for LVNC patients using easily accessible variables. Moreover, since the previous study also aimed to construct a LVNC safety algorithm to screen patients free from events during follow-up, some minor clinical events such as non-sustained ventricular tachycardia and HF hospitalization have been included in its endpoints of MACE [22]. In this current study, the main purpose of our risk prediction model was to identify true high-risk LVNC patients with possibilities of experiencing life-threatening adverse events. Therefore, our primary outcome predominantly encompassed severe clinical conditions.
Enrolled by consistent imaging diagnostic criteria, the baseline characteristics of patients in our study were generally comparable to the previously reported LVNC cohorts, including age, cardiac function, size of the left ventricle, as well as LVEF [19–21, 30–35]. Over a median follow-up of nearly 5 years, MACE occurred in approximately 30% of the entire population, with decompensated HF and malignant ventricular arrhythmias being the leading cause of mortality. On multivariable Cox regression analysis, age at diagnosis, NT-pro BNP levels, LA enlargement and LVEF ≤ 40% by TTE were demonstrated to be independent risk factors for MACE in LVNC.
Previous investigations have shown a potential connection between higher NT-pro BNP levels and increased risk of death or heart transplantation in individuals with LVNC [36, 37]. To the best of our knowledge, NT-pro BNP, a biomarker reflecting the extent of cardiac dysfunction, was included for the first time in the risk prediction model for MACE in LVNC as a candidate variable. In our cohort, only 9 (5.8%) patients experienced MACE in the group with NT-pro BNP levels < 300 pg/mL. In contrast, the incidence of MACE elevated to 3.1-times and 8.6-times higher in groups with NT-pro BNP levels of 300–1000 pg/mL and > 1000 pg/mL, indicating a strong prognostic relevance between elevated NT-pro BNP and MACE among LVNC patients. LA enlargement was also retained as an independent risk factor in our multivariable prediction model. LVNC patients with dilated LA presented a higher prevalence of HF, atrial fibrillation, and cardiac thrombosis, consequently resulting in a greater number of thromboembolic events and increased mortality rates. According to performance comparison among different risk models, both IDI and NRI revealed that incorporating NT-pro BNP, LA enlargement, or a combination of the two variables into the existing MACE prediction model [22] could significantly enhance its original predictive performance. Consistent with prior studies [20, 22, 29, 30, 32, 35, 36, 38, 39], our data further highlighted the importance of impaired left ventricular systolic function in the prognosis of LVNC, and LVEF ≤ 40% on TTE at baseline was demonstrated to be a significant predictor for MACE in patients with LVNC.
Based on our simplified risk prediction model, a risk scoring system for MACE in LVNC was established. The ABLE-SCORE displayed good discrimination and calibration capability, which enabled the accurate risk stratification for up to 5 years of follow-up, with significant differences in the cumulative incidence of MACE among low-, intermediate-, and high-risk groups of LVNC patients. DCA indicated that the ABLE-SCORE possessed greater net benefits than the previously reported risk score [22], suggesting its strength in determining high-risk patients and facilitating decision-making in a clinical setting. To enable the widely use of this new risk score, we designed an online calculator to assist both clinicians and patients in the risk assessment of LVNC (www.able-score.com), and this internet-based calculator could be easily integrated into hospital information systems or intelligent diagnosis and treatment platforms in the future. The ABLE-SCORE might be a promising tool for the MACE prediction in LVNC, allowing for more individualized, precise patient surveillance and management, and ultimately, a better outcome.
Several potential limitations of our study warrant discussion. First, owing to the retrospective nature of the study, potential bias caused by incomplete follow-up could not be excluded. Encouragingly, our study yielded relatively low rates of missing data and follow-up failures. All participants have received in-depth clinical and imaging assessment according to consistent standards. Second, similar to prior LVNC studies [22, 39], not all participants underwent CMR scans. Given an absence of CMR data in nearly 15% of the participants, LGE was not included in the multivariable analysis. Third, our risk prediction model has not taken into account of genetic testing. On one hand, it is hard to regularly perform genetic testing for patients given the current economic and healthcare realities in China. On the other hand, the value of genetic testing for the diagnosis and prognosis of LVNC still remains controversial. Several genetic variants within the sarcomere, cytoskeleton, Z-disc, and nuclear envelope have been reported to be tentatively associated with the LVNC phenotype, but most identified mutations have strong overlaps with other cardiomyopathies or inherited conditions, bringing into question their relevance in the pathogenesis of LVNC [12, 33, 35, 40–42]. In our study, we included LVNC family history as one of the candidate variables in the risk model, yet it was eliminated after conducting the multivariable Cox regression analysis.