以放射治療引起的肝臟病變建立生物性的正常組織併發症機率模型

Abstract

放射治療是癌症病人治療種類中相當重要的方式。放射線可能產生正常器官併發症的科學化評估方式，是由正常組織在放射治療電腦設計規劃系統中計算出的劑量體積資料，訂定出特定的閥值或臨界點。現行常用的劑量學指標存在著未能將體積化效應溶入放射線併發症風險估計的缺失，而正常組織併發症機率（NTCP）模型則是被發展出的一個較完善方式，能利用連續化的劑量體積資料及特定參數的搭配，來加權計算出高劑量與低劑量放射線可能造成的器官傷害。在我們過去已完成三度空間順形放射治療肝臟腫瘤的病人中，我們發現正常組織併發症機率模型較傳統的劑量學指標，更能準確估計放射線引起肝併發症的風險，然而以這個美國團隊研發的模型預估的風險，明顯低估了台灣肝臟放射治療發生肝併發症的機率，這樣的現象呈現了台灣病人肝臟對放射線的忍受度，是明顯低於西方國家病人的肝忍受度，這也說明了發展出因應台灣病人獨特生物性背景的肝併發症，所衍生出的專屬模型的必要參數。 研究的第一階段由89位肝癌病人及62位胃癌病人接受肝臟放射治療後，是否發生放射線引起肝併發症之結合生物性的正常組織併發症機率模式開始，而Ｂ型肝炎病毒帶原者則是台灣病人在肝忍受度上不同於西方國家的特點。我們首先以三參數的萊曼正常組織併發症機率模型，重算89位肝癌病人原始放射治療順形設計中的劑量體積數據，並以事實上發生肝併發症與否的根據，運用最大機率統計法來求取最佳化的模型參數。在多變項分析中，Ｂ型肝炎帶原是統計中最具顯著性意義與肝併發症有關的變項，而最佳化下的萊曼正常組織併發症機率模型參數（體積效應參數，曲線斜率參數，半機率併發症劑量）分別為0.35，0.39，49.4Gy，相較於非Ｂ型肝炎帶原組的參數為0.86，0.31，46.1Gy，其主要的組間差別是在於體積效應參數在Ｂ型肝帶原者明顯較小，顯示Ｂ型肝炎帶原的生物性因素對正常組織併發機率模型的最佳化有重大的影響。研究的第二階段是以萊曼正常組織併發症機率模型最佳化62位接受放射治療的胃癌病人，Ｂ型肝炎帶原這個變項仍然是相關於發生肝併發症的具統計意義相關因素，三個模型參數在Ｂ型肝炎帶原組為0.11，6.88，20.5Gy，相較於非帶原組的1.99，0.09，21.5Gy，在體積效應參數上的明顯差別再度顯示Ｂ型肝炎帶原的生物因素影響了肝臟的放射線忍受度。研究的第三階段使用四參數的平行性結構正常組織併發症機率模型，來針對其適用於平行性結構特性的器官，以合併肝癌與胃癌的151位病人資料庫測試這個模型的最佳化。Ｂ型肝炎帶原仍然是合併資料庫多變項分析中最具統計意義的相關於肝併發症的變項，而平行性結構正常組織併發症模型四參數，平均功能體積比率，功能比率曲線標準差，半肝臟單元受損劑量，以及肝臟單元的劑量效應曲線斜率，在整個資料庫最佳化結果為0.54，0.14，50Gy，0.13；在Ｂ型肝炎組為0.53，0.07，50Gy，4.6

Radiotherapy has been one of the most important treatment modalities in cancer patients. The scientific method to estimate the risk of radiation-induced organ complication is using the dose-volume data from the computerized treatment planning system to perform calculations with certain thresholds and criteria. The current commonly used dosimetric parameters have the defects of non-volumetric criteria and the lack of volume effect integrated into the radiation-related organ damage. Normal tissue complication probability (NTCP) model has been proposed as a more comprehensive way to calculate the risk of complication by the use of the serial dose-volume data with a few parameters to weigh the risk between low-dose and high-dose damage. In our past patients with radiation-induced liver disease (RILD) after three-dimensional conformal radiotherapy (3DCRT), we found that the NTCP was more useful than the conventionally used parameters. However, the risk of RILD in Taiwan seemed underestimated with the NTCP model parameters developed in the Unites States. This means the tolerance of liver to radiation for patients in Taiwan different from the patients’ tolerance in the western countries, and the indication of generating the unique model parameters based on the biological features of RILD in Taiwanese patients. Our first step was to establish the biology-integrated NTCP in the two different databases, 89 patients with hepatocellular carcinoma (HCC) and 62 patients with gastric carcinoma (GC) undergoing 3DCRT. Hepatitis B viral (HBV) carriers have been the unique feature of Taiwanese patients in their liver tolerance as compared to the western countries. We first used the three-parameter Lyman NTCP model to recalculate the NTCP of RILD in 89 HCC patients by their original dose-volume data retrieved from the conformal design of 3DCRT. Logistic regression was used for significant factors of RILD. Maximal likelihood analysis was conducted to obtain the best estimates of NTCP model parameters based on the true occurrence of RILD in 17 of 89 HCC patients. In multivariate analysis, HBV carrier remained statistically significant as the susceptible factor to RILD. The best estimates of NTCP parameters (n, m, TD50(1)) were 0.35, 0.39, and 49.4 Gy. The parameters specifically estimated from HBV carriers were 0.26, 0.40, and 50.0 Gy, as compared to 0.86, 0.31, and 46.1 Gy for non-carrier patients. The main difference in volume effect parameter (n) between the two subgroups indicated the impact of this biological factor (HBV carrier) on modeling NTCP. The second step was to apply the Lyman NTCP model in 62 GC patients. HBV carrier status was the only independent factor associated with RILD. The parameters (n, m, TD50(1)) specifically estimated from HBV carriers were 0.11, 6.88, and 20.5 Gy, as compared to 1.99, 0.09, and 21.5 Gy for non-carrier patients. The difference in volume effect parameter similarly described the biological integration of HBV carrier into the NTCP model. The third step was to use the four-parameter parallel-architecture NTCP model, specifically designed for the organ with parallel feature like liver, in a combined group of 151 patients with either HCC or GC. HBV carrier was the only independent factor with statistically significant susceptibility to RILD in multivariate test. The NTCP model parameters, mean functional reserve (v50), width of functional reserve distribution ( ), dose at which half of liver subunits are damaged (d1/2), slope parameter for subunit dose response (k), were 0.54, 0.14, 50Gy, 0.13 (whole group); 0.53, 0.07, 50Gy, 4.6 10-7 (HBV carriers); 0.59, 0.12, 25Gy, 59.8 (non-HBV carriers), respectively. The main difference in slope parameter demonstrated the biological influence of HBV carrier on RILD. The threshold effect of fraction of liver damaged (f) became evident after integrating biological factor (HBV carrier) into the modeling process. We concluded the effectiveness of the two NTCP models in RILD, and the unique importance of HBV carrier in estimating the two NTCP model parameters. It is emphasized that physical and mathematical NTCP methods should be cautiously used with appropriate integration of biological factors. The biology integrated NTCP models are extremely important for HBV carrier patients undergoing 3DCRT or the other new technology of radiotherapy to the liver. Such importance of biological factor in radiation-induced liver damage also implies the corresponding biological pathogenesis and warrants the ongoing basic cellular or molecular research on RILD.