B型肝炎病毒慢性感染之肝硬化引發肝細胞癌之決定因素與風險預測模型

Hepatocellular Carcinoma in Hepatitis B-Related Liver Cirrhosis: Determinants and Predictive model

曾鈺霖(Yu-Lin Tseng)

指導教授 ： 于明暉

國立臺灣大學/公共衛生學院/流行病學與預防醫學研究所/碩士(2016年)

https://doi.org/10.6342/NTU201601490

摘要

背景:
許多B型肝炎帶原者是具有臨床前期肝硬化，而這類臨床前期肝硬化者是肝細胞癌的風險族群。現今超音波影像被廣泛應用在偵測肝硬化與肝細胞癌上，在臨床上每次以超音波檢測肝硬化的結果，可能會隨著時間而有所變動，對於此現象是否會影響後續發生肝細胞癌的風險目前仍是未知的。本篇研究利用幾個長期追蹤的B型肝炎帶原者研究世代，分別有以下幾點目的: 1) 評估在不同特性的研究世代中，以超音波檢測出的肝硬化者，其短期、長期的肝細胞癌累積發生率。 2) 評估先前從B型肝炎帶原者中，所發展出的肝細胞癌危險分數，是否可以有效應用在以超音波檢測的肝硬化者上。 3) 針對以超音波檢測的肝硬化者，發展出肝細胞癌風險預測模式，並且評估此模式的預測能力。

材料與方法:
所有的資料是來自於三個長期追蹤的B型肝炎追蹤世代，年齡為20-80歲、無肝細胞癌病史，並且至少有一次(或以上)的超音波檢測，最後總共有5577位B型肝炎帶原者進入研究當中，研究個案資料的收集，包含基本人口學、生活史、超音波檢測、肝功能指標。在追蹤期間總共有24627次超音波檢查，其中有467位個案是以超音波檢測出一次(或以上)肝硬化。使用Kaplan-Meier來估計肝細胞癌的累積發生率，以及利用Cox proportional hazards regression來找出肝細胞癌的危險因子。在三個研究世代當中，選擇其中一個作為training set用以建構預測模型，另外兩個則作為validation set，用以驗證並且評估預測模式。利用Area under receiver operating characteristic (AUROC)、integrated discrimination improvement (IDI)、net reclassification improvement (NRI)來評估預測模型的表現，此外透過Nomogram的方法來建構危險分數，用以估計肝硬化者後續發生肝細胞癌的風險。

結果:
B型肝炎帶原者在平均追蹤12.75年後，總共有246位研究個案發生肝細胞癌。而利用超音波偵測之肝硬化者，10年之肝細胞癌之累積發生率在三個研究世代中為15至23%，其差異主要來自於各追蹤世代中，性別與年齡的分布不同。在肝硬化者中，以先前研究發展出的肝細胞癌危險分數、追蹤時是否有再次檢測出肝硬化、追蹤時丙胺酸轉氨酶(Alanine Aminotransferase; ALT)異常與否皆能有效預測肝細胞癌的發生，並且有達到統計顯著。單獨以先前研究發展出的肝細胞癌危險分數，預測肝硬化者5年、10年的肝細胞癌之AUC為0.71 (95% confidence interval [CI]: 0.60-0.82)與0.66 (95% CI: 0.57-0.74)，而在加入追蹤時是否有再次檢測出肝硬化與ALT異常與否後，預測肝硬化者5年、10年的肝細胞癌之AUC上升至0.81 (0.74-0.89)與0.73 (0.66-0.81)，此外，模式預測5年、10年肝細胞癌之IDI為0.125 (p=0.0019)與0.078 (p=0.0002)，而NRI則為0.500 (p=0.0010)與0.351 (p=0.0020)。在validation set當中，以nomogram轉換之危險分數的預測效果良好，預測5年、10年肝細胞癌之AUC分別為0.75 (95% CI: 0.67-0.83)與0.77 (95% CI: 0.71-0.84)。

結論:
本研究針對臨床前期肝硬化者所發展出來的危險分數，在不同特性族群上都有的表現，並且應該能應用至其他B型肝炎的族群中。使用此危險分數可以區分出具有不同肝細胞癌風險的B型肝炎肝硬化者，並能藉此來協助臨床上的篩檢與治療策略。

關鍵字

肝硬化 ； B型肝炎 ； 肝細胞癌 ； 預測模式

並列摘要

Background:
Many hepatitis B virus (HBV) carriers with cirrhosis are subclinical. Subclinical liver cirrhosis as measured by abdominal ultrasonography is strongly associated with an increased risk for hepatocellular carcinoma (HCC). Ultrasonography is now widely used in the diagnosis of liver cirrhosis for HCC screening. Whether a change over time in clinically measured ultrasonographic features influences HCC risk is unknown. With the use of a pooled longitudinal cohort database of HBV carriers, the purposes of this study are: 1) to establish the short-term and long-term cumulative risk for HCC in cirrhosis detected by ultrasonography across cohorts; 2) to investigate whether a HCC screening risk score previously associated with HCC in HBV carriers is prognostic for patients with cirrhosis detected by ultrasonogrpahy; and 3) to develop a novel prediction model for HCC in patients with cirrhosis detected by ultrasonography and evaluate its performance.

Materials Methods:
From 1988 to 2010, data on demographics, lifestyle, ultrasonography, and liver biochemistry were collected prospectively on 5577 HBV carriers aged 20-80 years from three cohorts, who were free of HCC and underwent at least one ultrasonography measurement. There were 24627 ultrasonography measurements from participants; of whom 467 were diagnosed with cirrhosis. The Kaplan-Meier method was used to estimate cumulative risk. Cox proportional hazards regression was used to identify predictors for HCC. A subcohort was used as the training set to guide the building of the risk model, and other two subcohorts were used to test the validity and transportability of the risk model. Area under receiver operating characteristic curve (AUROC), integrated discrimination improvement (IDI) and net reclassification improvement (NRI) were used to evaluate prediction model performance. A nomogram construction strategy that combined the HCC screening risk score with routine clinical tests were used to evaluate individual risk.

Results:
During a mean follow-up of 12.75 years, we identified 246 HCC cases. The 10-years cumulative HCC incidence was estimated to be 15%-23% in three cohorts, depending on distributions of age and sex. HCC screening risk score, the presence or absence of liver cirrhosis in repeated measurement, and follow-up alanine aminotransferase (ALT) activity were statistically significant independent predictors of HCC risk. AUROCs analyses revealed that using the HCC screening risk score alone predicted HCC with an accuracy of 0.71 (95% confidence interval [CI]: 0.60-0.82) and 0.66 (95% CI: 0.57-0.74) in 5- and 10-years, respectively. The addition of liver cirrhosis and ALT during follow-up into the nomogram improved the predictive performance, showing the corresponding AUCs of 0.81 (95% CI: 0.74-0.89, it is significant in comparison with the model including the HCC screening risk score) and 0.73 (95% CI: 0.66-0.81), respectively. In addition, the corresponding IDI were 0.125 (p=0.0019) and 0.078 (p=0.0002) and the corresponding NRI were 0.500 (p=0.0010) and 0.351 (p=0.0020) in 5- and 10-years, respectively. In the validation set, the nomogram worked well, and the AUROCs for 5- and 10-years prediction were 0.75 (95% CI: 0.67-0.83) and 0.77 (95% CI: 0.71-0.84), respectively.

Conclusions:
Our developed nomogram performed well across cohorts, and may be applied to other HBV carrier populations. The risk score may be able to identify HBV-related cirrhosis with differing risk for HCC to assist in clinical decision-making process.

並列關鍵字

Cirrhosis ； Hepatitis B ； Hepatocellular carcinoma ； Prediction model

參考文獻

1. 衛生福利部, 103年死因統計結果分析. 2014.

2. Bruix, J., M. Sherman, and D. American Association for the Study of Liver, Management of hepatocellular carcinoma: an update. Hepatology, 2011. 53(3): p. 1020-2.

3. El-Serag, H.B., Hepatocellular carcinoma. N Engl J Med, 2011. 365(12): p. 1118-27.

4. Venook, A.P., et al., The incidence and epidemiology of hepatocellular carcinoma: a global and regional perspective. Oncologist, 2010. 15 Suppl 4: p. 5-13.

5. Terrault, N.A., et al., AASLD guidelines for treatment of chronic hepatitis B. Hepatology, 2016. 63(1): p. 261-83.