代谢组学作为预测代偿期肝硬化患者失代偿或肝病相关死亡风险的工具
本研究嵌套于PREDESCI随机对照试验,纳入167例代偿期肝硬化且临床显著门脉高压(CSPH)患者,评估靶向代谢组学能否提升临床模型预测失代偿或肝病相关死亡的能力。结果显示,在HVPG/临床模型基础上加入神经酰胺(d18:1/22:0)和蛋氨酸后,C-index由0.748显著提高至0.808(p=0.032),且不含HVPG的临床/代谢物模型预测能力与含HVPG模型无显著差异。
要点
- 在代偿期肝硬化伴CSPH患者中,HVPG/临床模型预测主要终点的C-index为0.748。
- 加入神经酰胺(d18:1/22:0)和蛋氨酸后,模型C-index提高至0.808,差异有统计学意义。
- 不含HVPG的临床/代谢物模型C-index为0.785,与含HVPG模型无显著差异。
核心结论
- 在代偿期肝硬化伴CSPH患者中,代谢组学可提升临床模型的预测能力,并达到与含HVPG模型相似的预测效能。
研究设计与样本
- 本研究为PREDESCI队列(非选择性β受体阻滞剂 vs 安慰剂随机对照试验,纳入201例代偿期肝硬化伴CSPH患者)的嵌套研究,共纳入167例有血标本的患者。
- 采用超高效液相色谱-质谱联用进行靶向代谢组学血清分析,代谢物经单变量时间-事件Cox回归分析,以Log-Rank p值筛选排名靠前的代谢物构建逐步Cox模型,模型比较采用DeLong检验。
主要发现
- 82例CSPH患者随机分配至非选择性β受体阻滞剂组,85例分配至安慰剂组;33例发生主要终点(失代偿/肝病相关死亡)。
- 包含HVPG、Child-Pugh及治疗方案的模型(HVPG/临床模型)C-index为0.748(95%CI 0.664-0.827)。
- 加入神经酰胺(d18:1/22:0)和蛋氨酸后(HVPG/临床/代谢物模型),模型效能显著提升,C-index为0.808(95%CI 0.735-0.882),p=0.032。
- 上述2种代谢物联合Child-Pugh及治疗方案(临床/代谢物模型)的C-index为0.785(95%CI 0.710-0.860),与含或不含代谢物的HVPG模型均无显著差异。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND AND AIMS: Patients with compensated cirrhosis with clinically significant portal hypertension (CSPH: HVPG > 10 mm Hg) have a high risk of decompensation. HVPG is, however, an invasive procedure not available in all centers. The present study aims to assess whether metabolomics can improve the capacity of clinical models in predicting clinical outcomes in these compensated patients. APPROACH AND RESULTS: This is a nested study from the PREDESCI cohort (an RCT of nonselective beta-blockers vs. placebo in 201 patients with compensated cirrhosis and CSPH), including 167 patients for whom a blood sample was collected. A targeted metabolomic serum analysis, using ultra-high-performance liquid chromatography-mass spectrometry, was performed. Metabolites underwent univariate time-to-event cox regression analysis. Top-ranked metabolites were selected using Log-Rank p -value to generate a stepwise cox model. Comparison between models was done using DeLong test. Eighty-two patients with CSPH were randomized to nonselective beta-blockers and 85 to placebo. Thirty-three patients developed the main endpoint (decompensation/liver-related death). The model, including HVPG, Child-Pugh, and treatment received ( HVPG/Clinical model ), had a C-index of 0.748 (CI95% 0.664-0.827). The addition of 2 metabolites, ceramide (d18:1/22:0) and methionine (HVPG/Clinical/Metabolite model), significantly improved the model's performance [C-index of 0.808 (CI95% 0.735-0.882); p =0.032]. The combination of these 2 metabolites together with Child-Pugh and the type of treatment received (Clinical/Metabolite model) had a C-index of 0.785 (CI95% 0.710-0.860), not significantly different from the HVPG-based models including or not metabolites. CONCLUSIONS: In patients with compensated cirrhosis and CSPH, metabolomics improves the capacity of clinical models and achieves similar predictive capacity than models including HVPG.
引用
- Nicoară-Farcău O, Lozano JJ, Alonso C, et al. Metabolomics as a tool to predict the risk of decompensation or liver-related death in patients with compensated cirrhosis. Hepatology. 2023. PMID: 36811400. doi:10.1097/HEP.0000000000000316
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。