以電腦模擬建構對照組評估族群肝癌篩檢效益

Abstract

研究背景 運用腹部超音波進行族群肝癌篩檢對於降低肝癌死亡風險之短期效益在過往的研究中已藉由非隨機分派研究之實驗設計得到證據。然而如何運用合宜對照組對於此腹部超音波族群篩檢策略以及近年來發展之抗病毒治療在長期追蹤之整體與邊際效益與成本效益評估扮演重要的角色。此一評估方法在過往的研究中曾以預測模式之建構結合電腦模擬之研究設計以到建立在未進行腹部超音波與抗病毒藥物之情境下之模擬對照組達到評估之目的。 目的 本研究論文之目的在於 (1) 運用肝病毒以及非病毒之因子發展肝癌預測模型； (2) 在(1)之風險預測模型下發展肝癌疾病自然進展模型以及肝癌存活之預後模型作為建構未進行腹部超音波篩檢之狀態下的模擬對照組之基礎； (3) 運用上述之模擬對照組評估腹部超音波篩檢策略對於降低肝癌死亡之長期效益； (4) 運用模擬對照組評估腹部超音波篩檢結合抗病毒治療對於降低肝癌死亡之長期追蹤效益； (5) 以模擬對照組評估腹部超音波篩檢結合抗病毒治療對於不同風險族群在 降低肝癌死亡之長期追蹤效益。 材料與方法 本研究所運用之資料乃源於以風險分數為導向之腹部超音波篩檢介入。此一介入計畫於2008年10月開始於彰化縣執行。該介入策略之目標族群為45-74歲，參與彰化社區整合式萬人健檢(Changhua community-based integrated screening, CHCIS)計畫之民眾。各鄉診中之目標族群民眾在運用肝病毒感染、麩丙酮酸轉胺脢(ALT)、麩草醋酸轉胺脢(AST)、第二型糖尿病、血小板計數構成之風險分數區辨之高風險民眾，將邀請接受由腸胃科專科醫師執行之腹部超音波篩檢，以達到偵測民眾中具有肝硬化以及疑似肝癌病患之目的。 本論文發展電腦模擬架構結合肝癌疾病自然史及實證資料發展出的預測模式及存活資料以建構對照組，進一步進行不同肝癌防治策略之成本效益分析。 結果 本論文主要發現歸納以下五點： (1)本研究所建構之肝癌危險預測模型，其預測力相當佳，作業接受曲線下的面積達0.89 (95% CI: 0.85-0.93)。 (2)以危險分層為邀請基礎之超音波肝癌篩檢，若再加上30%順從率之抗病毒藥物治療，經八年追蹤觀察與模擬控制組相較，可避免60%肝癌死亡率(RR=0.39, 95% CI: 0.32-0.46)。 (3)藉由與模擬控制組相較之模擬結果，僅考量抗病毒藥物使用之順從率在30%時，可貢獻約22%之肝癌死亡避免，若順從率提高至50%及70%時，避免肝癌死亡之貢獻可分別增加至30%至35%。 (4)藉由與模擬控制組相較之模擬結果，若將中風險族群超音波篩檢涵蓋率自25%增加至75%時，避免肝癌死亡之貢獻僅能多增加3%。 (5)本研究經濟決策模式結果顯示，與模擬控制組相較，運用超音波進行肝癌篩檢，每挽救一個人年命，其增加成本效果比為美金22,849元，除超音波篩外，若再增加30%順從率之抗病毒藥物治療，每挽救一個人年命，其增加成本效果比為美金101,849元，若抗病毒藥物治療順從率分別提高至50%及70%時，每挽救一個人年命，其增加成本效果比則分別為美金141,805及181,919元。 結論 本論文以彰化縣腹部超音波肝癌篩檢實證資料驗證使用電腦模擬建構對照組評估族群肝癌篩檢效益的效度及可行性，進一步將此設計結合肝癌疾病自然史及實證資料發展出的預測模式及存活資料，進行腹部超音波結合抗病毒藥物療法之效益評估與成本效益分析。

Background Mass screening for hepatocellular carcinoma (HCC) with abdominal ultrasound (AUS) has been demonstrated in short-term follow-up without using a randomized controlled design. How to have a suitable control group plays an important role in the evaluation of long-term effectiveness of the overall and marginal effectiveness and cost-effectiveness of AUS as well as the recently emerging anti-viral therapy. Such an evaluation has been addressed before by the development of predictive model and also the creation of a computer simulation design characterized by the pseudo-control group in the absence of intervention on AUS and anti-viral therapy. Aims The objective of this thesis is to (1) develop a predictive model based on viral and non-viral factors for the risk of HCC; (2) develop a natural history model for the disease progression of HCC embedded with the risk prediction model developed in (1) and the prognosis model in relation to the survival of HCC in order to form a pseudo-control group in the absence of AUS screening; (3) estimate long-term effectiveness and cost-effectiveness of AUS in reducing HCC mortality based on the pseudo-control group; (4) estimate long-term effectiveness and cost-effectiveness of AUS and anti-viral therapy in reducing HCC mortality based on the pseudo-control group; (5) estimate long-term effectiveness and cost-effectiveness of AUS and anti-viral therapy in reducing HCC mortality by risk groups based on the pseudo-control group; Data and methods Data used here are based on a risk score-guided invitation for abdominal ultrasound which has been launched since October, 2008 in Changhua. Subjects aged 45-74 years attended the Changhua community-based integrated screening (CHCIS) program were targeted. Those with high score based on hepatitis virus infection, ALT, AST, type 2 diabetes, platelet count were invited to receive ultrasonography screening performed by board-certified gastroenterologists in town-based health center to identify liver cirrhosis, and suspected HCC cases. We developed a pseudo-control group by building up the disease natural history of progression of HCC embedded with the risk prediction model for HCC and the survival of HCC by detection modes. The pseudo-control model was further used to the development of health economic decision model for cost-effectiveness of various preventive strategies. Results The main results of this thesis include (1) The predictive validity of the risk prediction model for HCC is very good on the basis of ROC curve performance with AUC higher up to 0.89 (95% CI: 0.85-0.93); (2) The observed 8-year HCC mortality reduction with AUS by risk groups together with around 30% coverage rate of anti-viral therapy was around 60% (RR=0.39, 95% CI: 0.32-0.46); (3) The simulated results by using the pseudo-control group indicate additional contribution of 30% compliance rate of anti-viral therapy (empirical estimate) to HCC mortality reduction was around 22%. The corresponding figures are be raised to 30% and 35% when the compliance rate of anti-viral therapy is enhanced to 50% and 70%, respectively. (4) The simulated results by using the pseudo-control group show 3% HCC mortality reduction attributable to additional contribution of screening intermediate risk group with AUS when the coverage rate of this group is enhanced from 25% to 75%.; (5) The results of health economic decision model show the ICER (incremental cost-effectiveness ratio) values were $22,849 for the administration of AUS, $101,849 for the administration of AUS plus 30% compliance to anti-viral therapy. The corresponding figure for 50% and 70% compliance rate to anti-viral therapy were $141,805 and $181,919, respectively. Conclusions The population-based screening programme for HCC in Changhua confirmed the validity and feasibility for the developed risk score applied in our screening programme. We further used a computer study design with a pseudo-control group that was developed on the basis of disease natural history of HCC embedded with the predictive model and the survival part of HCC to estimate long-term effectiveness of the overall and marginal effectiveness and cost-effectiveness of AUS and anti-viral therapy.