From the Departments of Radiology (A.D.S., K.A.Z., E.F., R.S., D.S., F.S., M.R.) and Data Science (M.G., S.T.L.), University of Mississippi Medical Center, 2500 N State St, Jackson, MS 39216; and ImageIQ, Cleveland, Ohio (J.B., A.V.).
Radiology. 2017 Jun;283(3):711-722. doi: 10.1148/radiol.2016160799. Epub 2016 Nov 3.
Purpose To determine whether use of the liver surface nodularity (LSN) score, a quantitative biomarker derived from routine computed tomographic (CT) images, allows prediction of cirrhosis decompensation and death. Materials and Methods For this institutional review board-approved HIPAA-compliant retrospective study, adult patients with cirrhosis and Model for End-Stage Liver Disease (MELD) score within 3 months of initial liver CT imaging between January 3, 2006, and May 30, 2012, were identified from electronic medical records (n = 830). The LSN score was measured by using CT images and quantitative software. Competing risk regression was used to determine the association of the LSN score with hepatic decompensation and overall survival. A risk model combining LSN scores (<3 or ≥3) and MELD scores (<10 or ≥10) was created for predicting liver-related events. Results In patients with compensated cirrhosis, 40% (129 of 326) experienced decompensation during a median follow-up period of 4.22 years. After adjustment for competing risks including MELD score, LSN score (hazard ratio, 1.38; 95% confidence interval: 1.06, 1.79) was found to be independently predictive of hepatic decompensation. Median times to decompensation of patients at high (1.76 years, n = 48), intermediate (3.79 years, n = 126), and low (6.14 years, n = 152) risk of hepatic decompensation were significantly different (P < .001). Among the full cohort with compensated or decompensated cirrhosis, 61% (504 of 830) died during the median follow-up period of 2.26 years. After adjustment for competing risks, LSN score (hazard ratio, 1.22; 95% confidence interval: 1.11, 1.33) and MELD score (hazard ratio, 1.08; 95% confidence interval: 1.06, 1.11) were found to be independent predictors of death. Median times to death of patients at high (0.94 years, n = 315), intermediate (2.79 years, n = 312), and low (4.69 years, n = 203) risk were significantly different (P < .001). Conclusion The LSN score derived from routine CT images allows prediction of cirrhosis decompensation and death. RSNA, 2016 Online supplemental material is available for this article.
目的 旨在确定源自常规计算机断层扫描(CT)图像的定量生物标志物肝脏表面结节度(LSN)评分是否可用于预测肝硬化失代偿和死亡。
材料与方法 本研究为机构审查委员会批准的符合 HIPAA 规定的回顾性研究,纳入 2006 年 1 月 3 日至 2012 年 5 月 30 日期间初始 CT 肝脏成像后 3 个月内患有肝硬化且终末期肝病模型(MELD)评分在范围内的成年患者(n = 830)。通过 CT 图像和定量软件测量 LSN 评分。采用竞争风险回归来确定 LSN 评分与肝失代偿和总生存的相关性。创建了一种结合 LSN 评分(<3 或≥3)和 MELD 评分(<10 或≥10)的风险模型,用于预测与肝脏相关的事件。
结果 在代偿性肝硬化患者中,40%(129/326)在中位随访时间为 4.22 年期间经历了失代偿。在校正了包括 MELD 评分在内的竞争风险后,LSN 评分(危险比,1.38;95%置信区间:1.06,1.79)被发现是肝失代偿的独立预测因素。高(1.76 年,n = 48)、中(3.79 年,n = 126)和低(6.14 年,n = 152)肝失代偿风险患者的中位失代偿时间明显不同(P <.001)。在伴有代偿或失代偿性肝硬化的全队列中,61%(504/830)在中位随访时间 2.26 年内死亡。在校正竞争风险后,LSN 评分(危险比,1.22;95%置信区间:1.11,1.33)和 MELD 评分(危险比,1.08;95%置信区间:1.06,1.11)被发现是死亡的独立预测因素。高(0.94 年,n = 315)、中(2.79 年,n = 312)和低(4.69 年,n = 203)风险患者的中位死亡时间明显不同(P <.001)。
结论 源自常规 CT 图像的 LSN 评分可用于预测肝硬化失代偿和死亡。RSNA,2016 在线补充材料可用于本文。