肝細胞癌放射治療後的縱貫性肝功能之影響因子以及對於死亡之影響

Abstract

目前對於肝細胞癌在接受肝臟放射治療後長期追蹤之縱貫性肝功能的重要性尚未充分認識。詳細了解此重要性包含了識別不良肝功能惡化的高風險族群，以及評估不良肝功能對於存活的影響。這些關聯性在長期追蹤過程中的時變性，以及分析縱貫性數據時流行病學上需要考量的因素，對於是否能夠正確評估不同階段肝功能的重要性相當重要且值得注意。 在論文的第一部分(第二章)，我們旨在尋找影響肝臟放射治療後長期波動的肝功能中有意義的肝功能惡化之關鍵風險因子和影響的時間變化效應。由於縱貫性肝功能是本部分的研究結果，我們在Cox模型中採用了穩健變異數(robust variance estimation)估計來更正確的檢驗所估計的風險比的虛無假設，以考慮同一受試者不同時間肝功能的相關性。我們還於模型中導入了時間依賴協變項(time-dependent covariates)，用於了解顯著違反比例風險假設的因子其效應的時變趨勢。我們發現性別是個重要的顯著因子，和男性相比，女性肝功能惡化的風險會隨時間每月增加。這個發現與另一個發現形成了對比，放射治療前較高的肝功能數值可能導致放射治療後肝功能惡化的效應在時間上幾乎保持恆定。 在論文的第二部分(第三章)，我們專注於縱貫性肝功能對於存活的影響。作為一種客觀且有鑑別力的方法，白蛋白-膽紅素分級(Albumin-Bilirubin grade, ALBI grade)在這裡代表了肝功能，計算了它對於死亡的時間依賴族群可歸因分率(time-dependent population attributable fraction, time-dependent PAF)。由於縱貫性肝功能是本部分的研究暴露變項，我們將時間依賴白蛋白-膽紅素分級納入了競爭風險的事件發生時間設置下(time-to-event settings with competing risks)的累積發生函數(cumulative incidence functions, CIFs)中以估計族群可歸因分率，我們也應用了逆機率加權(inverse probability weighting, IPW)來處理動態腫瘤狀態所引起的時間變化干擾效應。我們發現經過逆機率加權調整後的第三等級白蛋白-膽紅素分級在肝臟放射治療後第一年的死亡族群可歸因分率估計可高達90%，且兩年後仍然維持在60%。這樣的發現顯示縱貫性肝功能對肝細胞癌族群的存活具有時間依賴且顯著的影響，其重要性可能不亞於肝細胞癌本身。 總結來說，縱貫性肝功能可能不論在調節臨床影響因子和直接影響肝細胞癌患者的存活都具有顯著的角色。我們的研究突顯了在長期追蹤中監測肝臟放射治療後縱貫性肝功能的重要性，有助於促進建立肝衰竭的預防策略，最終能夠提升肝細胞族群的生存和福祉。

The importance of longitudinal liver function (LF) following hepatic radiotherapy (RT) in treating hepatocellular carcinoma (HCC) has not been fully recognized. Understanding this importance in detail involves identifying high-risk groups for adverse LF deterioration and evaluating the impact of adverse LF on survival. Time variation of the associations within long-term follow-up and epidemiological considerations in analyzing longitudinal data are noteworthy for accurately assessing this importance across different time periods. In the first part of this dissertation (Chapter 2), we aim to identify essential risk factors for LF deterioration within the fluctuating long-term LF following hepatic RT and their time-varying effects. Since longitudinal LF is the study outcome in this part, we use robust variance estimation within the Cox model to accurately test the null hypotheses regarding the estimated hazard ratios, accounting for LF correlation within subjects. We also introduce time-dependent covariates for the factors that significantly violate the proportional hazards assumption to understand their trends of effects over time. We find that gender is a significant factor, with the hazard of LF deterioration for females increasing monthly over time compared with males. This contrasts with another finding that a higher baseline LF score before RT may induce LF deterioration after RT that remains almost constant over time. In the second part of this dissertation (Chapter 3), we focus on the impact of longitudinal LF on survival. Albumin-Bilirubin (ALBI) grade, as an objective and discriminative approach, is used as the representative for LF, and its time-dependent population attributable fraction (PAF) for mortality is calculated. Since longitudinal LF is the study exposure in this part, we incorporate time-dependent ALBI grade into the cumulative incidence functions in time-to-event settings with competing risks to estimate the PAF. We also apply inverse probability weighting (IPW) to address time-varying confounding due to dynamic tumor status. We find that the estimated IPW-adjusted PAF of ALBI grade 3 for mortality is as high as 90% within the first year following hepatic RT and remains at 60% after two years. These findings indicate that longitudinal LF has a time-dependent and substantial impact on survival in HCC populations, possibly with importance no less than that of HCC itself. In summary, longitudinal LF may play a significant role in mediating clinical factors and directly affecting survival for HCC patients. Our work highlights the importance of monitoring longitudinal LF following hepatic RT during long-term follow-up, helps facilitate prevention strategies for hepatic decompensation, and ultimately enhances the survival and well-being of HCC populations.