建立預測抗結核藥物治療期間肝損傷之評分系統暨評估基因型危險因子 之重要性
結核病 (Tuberculosis ， TB) 在世界各地都是導致死亡率的主要原因之ㄧ，在台灣也是傳染病致死 原因的第一位。目前結核病的治療需要合併使用多種藥物且至少需治療六個月以上的時間，而抗結 核藥物治療期間最常發生的不良反應即是肝毒性。肝毒性除了會導致病人服藥配合度不高，也間接 導致抗結核藥物的抗藥性問題。至目前為止用來預測抗結核藥治療期間肝毒性的危險因子仍尚未有 定論，且缺乏能實際應用在臨床上之相關建議。
本研究分為兩大部份，第一部份分析抗結核藥物治療期間肝損傷的流行病學上常見之危險因子，例
如：年齡、性別、營養狀況、合併疾病、生活習慣、第一線抗結核藥物使用品項及劑量、併用藥物 個數及肝功能檢測基準值；而第二部份則加入 NAT2 (N-acetyltransferase 2) 、 OATP1B1(Human org anic anion-transporting polypeptides 1B1) 及 UGT1A1 (UDP-Glucuronosyl Transferase1A1) 等基因型，
經 logistic regression 校正分析，根據其影響程度分別設計成兩個量表；利用 Hosmer and Lemeshow test 檢視其配適度，透過 AUC of ROC curve 檢視量表的鑑別力。
全部研究對象 (n=594) 的流行病學之危險因子分析及校正後，具有肝臟疾病、 rifampin 劑量≧ 12 m g/kg ， AST 、 ALT 基準值大於兩倍正常值上限和 ALP 基準值大於三倍正常值上限為顯著的危險因 子，此量表 AUC of ROC curve 為 0.694 ，在最佳切點 3 分時， sensitivity 和 specificity 分別為 64.2
% 和 67.4% ；而包含基因型研究對象的部份 (n=97) 校正後女性、 AST 基準值高於正常值、帶有基 因型 NAT2*7 、 OATP1B1*1a/*1a 和 OATP1B1*1a/*15 為顯著的危險因子，其量表之 AUC of ROC curve 為 0.862 ，最適當切點為 6 分，其 sensitivity 和 specificity 分別為 81.3% 和 80.2% 。
透過這兩個量表之比較，評估基因型危險因子在預測抗結核藥物治療期間肝傷的重要性，顯示出僅
使用一般流行病學之危險因子預測肝毒性時，縱使增加人數依舊無法提高預測的準確性，反之基因 型危險因子能大幅增加量表的鑑別力，顯示基因型在預測抗結核藥物治療期間發生肝損傷中扮演很 重要的角色。
Development of Risk Scales for Predicting Liver Injury during Anti tuberculosis Therapy and Evaluation the Importance of Genetic Ri
sk Factors
Tuberculosis (TB), one of the major causes of mortality throughout the world, was the greatest infectious c ause of death in Taiwan. The standard therapy of TB requird multiple medications, and patients should be t reated for at least 6 months. Hepatotoxicity occured duing anti-TB therapy decreased patients’ adherence a nd increased the drug resistance. The risk factors in predicting hepatotoxicity were still not clear, and lack of recommendations in clinical practice.
The study was composed of two sections. The traditional risk factors such as age, gender, nutritional status, concomitant diseases, social history, the usage and dosage of first line anti-TB drugs, and baseline liver fun ction were analyzed in section one. Section two, the genotypes of NAT2, OATP1B1 and UGT1A1 were ad ded. The risk factors were adjusted and weighted by logistic regression, and developing two scales. One of the two scales was only traditional risk factors included, and the other was added by genetic risk factors. T he goodness of fit was examed by Hosmer and Lemeshow test. The discriminatory power of scale was exa med by AUC of ROC curve.
Liver disease, dosage of rifampin 12 mg/kg, baseline AST and ALT 2x ULN (upper limit of normal) a≧ ≧ nd ALP 3x ULN were the risk factors in section 1 (n=594). The AUC of ROC curve was 0.694, the sensi≧ tivity was 64.2% and specificity was 67.4%. In section 2(n=97), female, AST abnormal, with genotype of NAT2*7, OATP1B1*1a/*1a and OATP1B1*1a/*15 were the significant risk factors. The AUC of ROC cur ve was 0.862, the sensitivity was 81.3% and specificity was 80.2%.
By comparing the two scales, the genetic risk scale showed better discriminative power in predicting liver i njury during anti-TB therapy. Removing the genetic risk factors through the genetic risk scale, the AUC of ROC curve was 0.728 in 97 study population while it was 0.649 in 594. It showed the limitation of traditio nal risk factors, even increased the study population, it still can’t predict hepatotoxicity precisely. The gene tic risk factors play important roles in predicting liver injury during anti-TB therapy.
