Division of Cardiovascular, Brigham and Women's Hospital, Boston, MA, USA.
Hospital de Clínicas de Porto Alegre and UFRGS Medical School, Porto Alegre, Brazil.
Eur J Heart Fail. 2021 Aug;23(8):1346-1356. doi: 10.1002/ejhf.2120. Epub 2021 Mar 9.
Prognostic models of sudden cardiac death (SCD) typically incorporate data at only a single time-point. We investigated independent predictors of SCD addressing the impact of integrating time-varying covariates to improve prediction assessment.
We studied 8399 patients enrolled in the PARADIGM-HF trial and identified independent predictors of SCD (n = 561, 36% of total deaths) using time-updated multivariable-adjusted Cox models, classification and regression tree (CART), and logistic regression analysis. Compared with patients who were alive or died from non-sudden cardiovascular deaths, patients who suffered a SCD displayed a distinct temporal profile of New York Heart Association (NYHA) class, heart rate and levels of three biomarkers (albumin, uric acid and total bilirubin), with significant differences observed more than 1 year prior to the event (P < 0.001). In multivariable models adjusted for baseline covariates, seven time-updated variables independently contributed to SCD risk (incremental likelihood chi-square = 46.2). CART analysis identified that baseline variables (implantable cardioverter-defibrillator use and N-terminal prohormone of B-type natriuretic peptide levels) and time-updated covariates (NYHA class, total bilirubin, and total cholesterol) improved risk stratification. CART-defined subgroup of highest risk had nearly an eightfold increment in SCD hazard (hazard ratio 7.7, 95% confidence interval 3.6-16.5; P < 0.001). Finally, changes over time in heart rate, NYHA class, blood urea nitrogen and albumin levels were associated with differential risk of sudden vs. non-sudden cardiovascular deaths (P < 0.05).
Beyond single time-point assessments, distinct changes in multiple cardiac-specific and systemic variables improved SCD risk prediction and were helpful in differentiating mode of death in chronic heart failure.
心脏性猝死(SCD)的预后模型通常仅纳入单一时间点的数据。我们研究了 SCD 的独立预测因素,探讨了整合时变协变量对改善预测评估的影响。
我们研究了 PARADIGM-HF 试验中的 8399 例患者,并使用时间更新的多变量调整 Cox 模型、分类和回归树(CART)以及逻辑回归分析,确定了 SCD 的独立预测因素(n=561,占总死亡人数的 36%)。与存活或死于非突发性心血管死亡的患者相比,SCD 患者的纽约心脏协会(NYHA)心功能分级、心率和三种生物标志物(白蛋白、尿酸和总胆红素)的时间变化模式明显不同,且在事件发生前 1 年以上就有显著差异(P<0.001)。在调整基线协变量的多变量模型中,有七个时间更新的变量独立导致 SCD 风险增加(增量似然卡方=46.2)。CART 分析确定,基线变量(植入式心脏复律除颤器的使用和 B 型利钠肽前体水平)和时间更新的协变量(NYHA 心功能分级、总胆红素和总胆固醇)改善了风险分层。CART 定义的最高风险亚组的 SCD 危险度几乎增加了八倍(危险比 7.7,95%置信区间 3.6-16.5;P<0.001)。最后,心率、NYHA 心功能分级、血尿素氮和白蛋白水平的时间变化与突然与非突然心血管死亡的风险差异相关(P<0.05)。
除了单一时间点评估外,多个心脏特异性和全身变量的明显变化可改善 SCD 风险预测,并有助于区分慢性心力衰竭的死亡模式。