Demissei Biniyam G, Valente Mattia A E, Cleland John G, O'Connor Christopher M, Metra Marco, Ponikowski Piotr, Teerlink John R, Cotter Gad, Davison Beth, Givertz Michael M, Bloomfield Daniel M, Dittrich Howard, van der Meer Peter, van Veldhuisen Dirk J, Hillege Hans L, Voors Adriaan A
Department of Cardiology, University of Groningen, University Medical Centre Groningen, Hanzeplein 1, 9713 GZ, Groningen, the Netherlands.
Department of Epidemiology, University of Groningen, University Medical Centre Groningen, Groningen, the Netherlands.
Eur J Heart Fail. 2016 Mar;18(3):269-80. doi: 10.1002/ejhf.443. Epub 2015 Dec 3.
The clinical value of single biomarkers at single time-points to predict outcomes in patients with acute heart failure (AHF) is limited. We performed a multimarker, multi-time-point analysis of biomarkers for the prediction of post-discharge clinical outcomes in high-risk AHF patients.
A set of 48 circulating biomarkers were measured in the PROTECT trial which enrolled 2033 patients with AHF. Associations between baseline levels of biomarkers and outcomes (30-day all-cause mortality, 30-day death or rehospitalization for renal/cardiovascular causes and 180-day all-cause mortality) were evaluated. Prognostic accuracies of baseline, days 2 or 3, 7, and 14 biomarker measurements were estimated and compared utilizing a time-dependent area under the curve (AUC) analysis. Forty-four biomarkers were significantly associated with outcomes, but 42 had limited prognostic value (C-index < 0.70). However, multimarker models combining best-performing biomarkers from different clusters had a much stronger prognostic value. Combining blood urea nitrogen (BUN), chloride, interleukin (IL)-6, cTnI, sST-2 and VEGFR-1 into a clinical model yielded a 11% increase in C-index to 0.84 and 0.78 for 30-day and 180-day all-cause mortality, respectively, and cNRI of 0.86 95% CI [0.55-1.11] and 0.76 95% CI [0.57-0.87]. Prognostic gain was modest for the 30-day death/rehospitalization for cardiovascular or renal causes endpoint. Comparative time-dependent AUC analysis indicated that late measurements provided superior accuracy for the prediction of all-cause mortality over 180 days, with few exceptions including BUN and galectin-3. However, the predictive value of most biomarkers showed a diminishing pattern over time irrespective of moment of measurement.
Multimarker models significantly improve risk prediction. Subsequent measurements, beyond admission, are needed for majority of biomarkers to maximize prognostic value over time, particularly in the long term.
单一生物标志物在单一时间点预测急性心力衰竭（AHF）患者预后的临床价值有限。我们对生物标志物进行了多标志物、多时间点分析，以预测高危AHF患者出院后的临床结局。
在纳入2033例AHF患者的PROTECT试验中，检测了一组48种循环生物标志物。评估了生物标志物基线水平与结局（30天全因死亡率、30天因肾脏/心血管原因死亡或再次住院以及180天全因死亡率）之间的关联。利用曲线下时间依赖性面积（AUC）分析估计并比较了基线、第2或3天、第7天和第14天生物标志物测量的预后准确性。44种生物标志物与结局显著相关，但42种的预后价值有限（C指数<0.70）。然而，将来自不同簇的表现最佳的生物标志物组合成多标志物模型具有更强的预后价值。将血尿素氮（BUN）、氯、白细胞介素（IL）-6、肌钙蛋白I（cTnI）、可溶性生长刺激表达基因2蛋白（sST-2）和血管内皮生长因子受体1（VEGFR-1）纳入临床模型，30天和180天全因死亡率的C指数分别提高11%，达到0.84和0.78，净重新分类改善（cNRI）分别为0.86（95%CI[0.55-1.11]）和0.76（95%CI[0.57-0.87]）。对于30天因心血管或肾脏原因死亡/再次住院这一终点，预后改善幅度较小。比较性时间依赖性AUC分析表明，除了BUN和半乳糖凝集素-3等少数例外情况，后期测量对预测180天以上的全因死亡率具有更高的准确性。然而，大多数生物标志物的预测价值随时间呈递减模式，与测量时间无关。
多标志物模型显著改善风险预测。对于大多数生物标志物，需要在入院后进行后续测量，以随时间最大化预后价值，尤其是在长期。
