The performance of Bayesian feedback in estimating individual blood level, AUCoo and body clearance was evaluated with antimicrobial trimethoprim in a rabbit model using data obtained from intravenous administration of 15.3 mg/kg dose in nine male rabbits. The estimated equation parameters for a two-compartment body model by the method using one feedback of blood level showed considerable or large deviation from the individual fitting values. However, good estimation for forecasting blood levels, AUCoo and body clearance were resulted. Since clearance is an essential pharmacokinetic parameter to describe drug elimination in vivo, the optimization of dosage regimens can be easily determined by using the relationship of dose, dosing interval, blood level and clearance at steady-state. Some theoretical aspects on the Bayesian approach were also discussed. Since the Bayesian feedback method is dependent on some assumptions such as normal or log-normal distribution of residual error and knowledge of the prior distribution of the population pharmacokinetic parameters, it is suggested that the performance and prediction interval of the method should be evaluated before applying the method to a new drug. 劇毒藥物劑量的個體適量化及控制安全有效的血中濃度已成為藥物療法上的常識及必備條件。傳統的藥物動力學方法在測定參數時過於繁雜，而簡化法常導致較大的偏誤。近年來Bayes理論在臨床藥物動力學上的應用廣受重視，因為此方法只須病人服藥後之一個血液樣品及母集團之參數及其分佈，就可推定血中濃度之經時變化。本報告就Bayes理論的擴張方法一Bayes回饋推算法之內涵及適用性加以研討，利用trimethoprim快速靜脈注射於九隻雄兔所得之數據，檢討此方法對各個體的藥物動力學方程式，血中藥物濃度，曲線下全面積及體出清率的推定性。結果顯示此方法所推定的個體藥物動力學方程式的參數偏誤較大，但對血中藥物濃度，曲線下全面積及體出清率的推定則相當近似。此結果並非意外，蓋Bayes回饋推算法並不在確定各別的藥物動力學方程式中的各個參數，這些參數只是用來合成血中濃度的可變數。這些參數所組合的函數能相當近似地反映血中藥物濃度，故在臨床上常被應用來預估血中藥物濃度的經時變化。本報告進一步顯示這些參數所導出的個體出清率和直接以個體的血中濃度去迴歸計算所得者相當近似，利用出清率則能更有效地設計適宜的藥物治療法。因為Bayes回饋推算法本身有種種的假設前提，這些前提並不一定成立，故在利用此方法於臨床藥物控制之前，有必要對各劇毒藥物的適用性加以檢討。
