病毒性肝癌精準預防

研究背景肝癌過去在B型肝炎及C型肝炎高盛行的臺灣一向為主要的癌症死因，但在過去三十年一系列的初段、次段與末段預防的手段，其發生率及死亡率已經呈現下降的趨勢。這些預防手段主要是依循個案在疾病自然史的進展，包括急性肝炎、慢性肝炎、肝硬化、肝癌發生與復發及死亡，針對不同階段的防治手段，且已有許多流行病學實證顯示各項介入手段的效益。我的論文希望能系統性的分析不同階段疾病病程及介入措施的影響，進一步探討病毒性肝炎相關肝癌的精準預防。本論文的研究目的包括， 1. 利用貝氏圖形化因果模型(Bayesian causal graphic model)從台灣實際的肝癌發生率趨勢變化進行估計量化不同時間點針對不同年齡層實施的防治介入的直接效果(direct effect)、估計肝癌發生率年變化率(annual change)對於防治介入的效果的影響以及不同防治介入之間因為因果關係間接影響肝癌發生率變化的間接效果(indirect effect)； 2. 利用數位雙生(digital twin)的概念結合貝氏圖形化因果模型從實際肝癌發生率資料經由模型估計所得的參數創造一個電腦模擬的虛擬組(virtual group)，並以現行三種防治政策實施的狀況下為對照組進行比較，估計如何加強抗病毒藥的治療覆蓋以及在何時可以達成病毒性肝炎相關肝癌的消滅； 3. 利用決策樹(decision tree)結合馬可夫模型(Markov model)分別針對核苷酸類似物抗病毒藥(nucleos(t)ide analogues, NAs)治療慢性B型肝炎和全口服直接作用抗病毒藥(all-oral direct-acting antivirals, DAAs)治療慢性C型肝炎進行成本效益分析。材料與方法本研究將以數位雙生研究設計結合貝氏圖形化因果關係模型，透過國家癌登資料庫中台灣肝癌發生率的實際資料並考慮年齡層和不同縣市別進行模型估計得到在不同時間點開始執行的三個肝癌防治介入對於肝癌發生率趨勢變化的直接和間接作用之相關參數，這三個不同的肝癌防治介入包括1984年的新生兒B肝疫苗注射計畫(hepatitis B vaccination)、1995年的全民健康保險制度(universal health care)和2004年的慢性B和C肝試辦計畫(selective antiviral therapy)，接著利用虛擬試驗(in-silico experiment)中，針對除了上述三種不同肝癌防治介入外，再加上提升B肝和C肝抗病毒藥的覆蓋率(added-on large-scale antiviral therapy)的虛擬實驗組預測以台灣的情境何時可以達成病毒性肝炎相關肝癌消滅的目標，即發生率下降至4 /100,000以下，並和只延續上述三種不同肝癌防治介入的對照組進行比較，探討如何透過B肝和C肝抗病毒藥的治療來達成目標。最後利用馬可夫決策樹模型分別針對B肝和C肝抗病毒藥物治療進行成本效益分析。結果透過數位雙生的虛擬試驗預測，在提升慢性B肝和C肝抗病毒藥物治療的覆蓋率之後，若多預防10,042位病毒性肝炎相關肝癌發生個案的發生，則可以在西元2045年達成病毒性肝炎相關肝癌消滅的目標，相對應的慢性B肝和C肝需治療人數分別為159,992和34,624。在成本效益分析方面，長期使用B肝抗病毒藥物相較於完全未使用抗病毒藥物的策略，在成本增加39,942.53美元的情況下，可以多得到2.83質量調整壽命年，增量成本-效果比(incremental cost-effectiveness ratio, ICER)為14,110.96美元，在肝癌防治上合乎成本效益(cost-effective)。C肝抗病毒藥物治療策略有三種，分別為所有不論肝纖維化程度的C肝病毒血症給予抗病毒藥物治療(策略1)，只針對嚴重肝纖維化(at leat F3 and above)的C肝病毒血症給予抗病毒藥物治療(策略2)，以及完全沒有抗病毒藥物治療(策略3)，若以策略3為對照組，策略1在成本減少47,975.79美元的情況下，可以多得到5.15質量調整壽命年，增量成本-效果比為-9,315.69美元，而策略2則在成本減少38,354.96美元的情況下，可以多得到4.32質量調整壽命年，增量成本-效果比為-8,878.46美元。若以策略2為對照組，策略1在成本減少9,620.83美元的情況下，可以多得到0.83質量調整壽命年，增量成本-效果比為-11,591.36美元，故所有不論肝纖維化程度的C肝病毒血症皆給予抗病毒藥物治療的策略相較於其他兩種策略，在肝癌防治上達到成本節省(cost-saving)。結論利用數位雙生研究設計合併貝氏圖形化因果模型，釐清不同年齡層和不同地域執行的肝癌防治介入之間的因果關係，並搭配相對應的成本效益分析，台灣在肝癌防治上的經驗，可以作為未來全世界消滅病毒性肝炎相關肝癌制定相關公衛政策的指引。

關鍵字

肝癌；貝氏圖形化因果模型；數位雙生；馬可夫決策樹；成本效益分析

並列摘要

Background Hepatocellular carcinoma (HCC) was a leading cause of cancer-related death due to high prevalence of chronic hepatitis B and C in Taiwan. After a series of interventions across primary to tertiary prevention, both incidence and mortality of HCC has significantly declined in the past three decades. Nevertheless, each prevention program is directed and oriented by disease natural history from acute hepatitis, chronic hepatitis, cirrhosis, occurrence and relapse of HCC before death. Although numerous epidemiological data published before, the information is fragmentary rather than systematic. It is necessary to elucidating each part of disease natural history and the impact of interventions on HCC prevention in order to develop precision prevention of virus-related HCC. Study aims the aims of this study are as follows: 1. To develop a Bayesian causal graphic model for linking annual change of the secular trend of HCC incidence with three change-point interventions, taking the heterogeneity of demographic and geographic variations into account, to estimate the parameters governing the direct and indirect effects of three interventions; 2. To re-create a digital twin group using the parameters after learning from the causal model to predict when and how the elimination of virus-related HCC can be achieved by adding large-scale antiviral therapy to the existing interventions for HCC prevention; 3. To do cost-effectiveness analysis of nucleos(t)ide analogues (NAs) therapy and all oral direct-acting antivirals (DAAs) for the prevention of virus-related HCC by using Markov decision tree models. Materials and Methods The digital twin study design in conjunction with the statistical learning from Bayesian causal graphic model was envisaged to do in-silico experiment with the experimental group given by the added-on large-scale antiviral therapy until the time to eliminate virus-related HCC, and the control group given the current three change-point interventions including selective antiviral therapy. The virtual subject was further assigned by the administration of antiviral therapy to become the experimental subject with the initiation of DAAs for CHC and the long-term use of NAs therapy for CHB. How and when to achieve the elimination of virus-related HCC incidence less than 4 per 100,000 could be predicted by using such an in-silico experiment with the digital twin approach. Finally, cost-effectiveness analysis of NAs therapy and DAAs therapy for the prevention of virus-related HCC was conducted by using Markov decision tree models. Results Based on the incidence predicted by the virtual twin group, the elimination of virus-related HCC in Taiwan would be achieved by the end of 2045 when a further 10,042 virus-related HCC incident cases could be averted after treating 159,992 and 34,624 chronic hepatitis B and C patients with large-scale antiviral therapy, respectively. Compared with no NAs therapy, the base-case shows long-term NAs therapy cost more ($39,942.53) and earned 2.83 quality-adjusted life years (QALYs). The positive incremental cost-effectiveness ratio (ICER) of long-term NAs therapy was $14,110.96. Compared with no DAAs therapy (Strategy 3), DAAs for all viremic patients (Strategy 1) shows cost less ($47,975.79) but earned 5.15 QALYs. Otherwise, DAAs for viremic patients with at least F3 and above (Strategy 2) shows cost less ($38,354.96) but earned 4.32 QALYs. The ICERs of Strategy 1 and Strategy 2 were -$9,315.69 and -$8,878.46, respectively. Compared with Strategy 2, Strategy 1 cost less ($9,620.83) but earned 0.83 QALYs and the ICER was -$11,591.36. As a result, DAAs therapy for all viremic patients for the prevention of virus-related HCC was cost-saving. Conclusion We demonstrate how to use the digital twin design with Bayesian casual graphic model to achieve the elimination of virus-related HCC and conduct the following cost-effectiveness analysis. These findings may provide a state-of-the-art evidence-based estimate on the global elimination of virus-related HCC.

並列關鍵字

Hepatocellular carcinoma ； Bayesian causal graphic model ； Digital twin ； Markov decision tree ； Cost-effectiveness analysis

參考文獻

1. Lal, A., Y. Pinevich, O. Gajic, V. Herasevich, and B. Pickering, Artificial intelligence and computer simulation models in critical illness. World J Crit Care Med, 2020. 9(2): p. 13-19.

Google Scholar

2. Masison, J., J. Beezley, Y. Mei, H. Ribeiro, A.C. Knapp, L. Sordo Vieira, B. Adhikari, Y. Scindia, M. Grauer, B. Helba, W. Schroeder, B. Mehrad, and R. Laubenbacher, A modular computational framework for medical digital twins. Proc Natl Acad Sci U S A, 2021. 118(20).

Google Scholar

3. Fagherazzi, G., Deep Digital Phenotyping and Digital Twins for Precision Health: Time to Dig Deeper. J Med Internet Res, 2020. 22(3): p. e16770.

Google Scholar

4. Björnsson, B., C. Borrebaeck, N. Elander, T. Gasslander, D.R. Gawel, M. Gustafsson, R. Jörnsten, E.J. Lee, X. Li, S. Lilja, D. Martínez-Enguita, A. Matussek, P. Sandström, S. Schäfer, M. Stenmarker, X.F. Sun, O. Sysoev, H. Zhang, and M. Benson, Digital twins to personalize medicine. Genome Med, 2019. 12(1): p. 4.

Google Scholar

5. Lal, A., G. Li, E. Cubro, S. Chalmers, H. Li, V. Herasevich, Y. Dong, B.W. Pickering, O. Kilickaya, and O. Gajic, Development and Verification of a Digital Twin Patient Model to Predict Specific Treatment Response During the First 24 Hours of Sepsis. Crit Care Explor, 2020. 2(11): p. e0249.

Google Scholar

主題瀏覽