Validation of Limited Sampling Strategy for the Estimation of Mycophenolic Acid Exposure in Chinese Adult Liver Transplant Recipients

Transcription

1 LIVER TRANSPLANTATION 13: , 2007 ORIGINAL ARTICLE Validation of Limited Sampling Strategy for the Estimation of Mycophenolic Acid Exposure in Chinese Adult Liver Transplant Recipients Chen Hao, 1 Chen Erzheng, 1 Mao Anwei, 1 Yu Zhicheng, 2 Shen Baiyong, 1 Deng Xiaxing, 1 Zhang Weixia, 2 Peng Chenghong, 1 and Li Hongwei 1 1 Center of Organ Transplantation and 2 Institute of Clinical Pharmacology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, P.R. China Mycophenolate mofetil (MMF) is indicated as immunosuppressive therapy in liver transplantation. The abbreviated models for the estimation of mycophenolic acid (MPA) area under the concentration-time curve (AUC) have been established by limited sampling strategies (LSSs) in adult liver transplant recipients. In the current study, the performance of the abbreviated models to predict MPA exposure was validated in an independent group of patients. A total of 30 MPA pharmacokinetic profiles from 30 liver transplant recipients receiving MMF in combination with tacrolimus were used to compare 8 models performance with a full 10 time-point MPA-AUC. Linear regression analysis and Bland-Altman analysis were used to compare the estimated MPA-AUC 0-12h from each model against the measured MPA-AUC 0-12h. A wide range of agreement was shown when estimated MPA-AUC 0-12h was compared with measured MPA-AUC 0-12h, and the range of coefficient of determination (r 2 ) was from to The model based on MPA pharmacokinetic parameters C 1h,C 2h,C 6h, and C 8h had the best ability to predict measured MPA-AUC 0-12h, with the best coefficient of determination (r ), the excellent prediction bias (2.18%), the best prediction precision (5.11%), and the best prediction variation (2SD 7.88 mg h/l). However, the model based on MPA pharmacokinetic sampling time points C 1h,C 2h, and C 4h was more suitable when concerned with clinical convenience, which had shorter sampling interval, an excellent coefficient of determination (r ), an excellent prediction bias (3.48%), an acceptable prediction precision (14.37%), and a good prediction variation (2SD mg h/l). Measured MPA-AUC 0-12h could be best predicted by using MPA pharmacokinetic parameters C 1h,C 2h,C 6h, and C 8h. The model based on MPA pharmacokinetic parameters C 1h,C 2h, and C 4h was more feasible in clinical application. Liver Transpl 13: , AASLD. Received March 16, 2007; accepted July 15, Mycophenolate mofetil (MMF) as a part of immunosuppressive regimen has been widely used for the prevention of acute rejection in organ transplantation. 1 MPA, the active compound of MMF, inhibits the proliferation of B and T lymphocytes through inhibiting inosine monophosphate dehydrogenase, a key enzyme involved in the synthesis of nucleotides. 2 MMF is free of toxicities that affect long-term patient and graft survival and negative effects on blood pressure, lipids, and carbohydrates. MMF-based regimens with low doses of calcineurin inhibitors and corticosteroid reduction or removal are recommended after organ transplantation. 3 When MMF was introduced, the manufacturer and the health authorities considered therapeutic drug monitoring unnecessary because of its favorable safety profile under the standard dosage (1 g twice a day). In kidney, heart, and liver transplant patients, the relationship between MPA pharmacokinetics and clinical outcomes (acute rejection and/or adverse effects) has been reported. 4-6 The pharmacokinetics of MPA exhibit Abbreviations: MMF, mycophenolate mofetil; MPA, mycophenolic acid; AUC, area under the concentration-time curve; LSS, limited sampling strategy; TAC, tacrolimus; HPLC, high-performance liquid chromatography; r 2, coefficient of determination; 95% CI, 95% confidence interval. The first 2 authors contributed equally to this study. Address reprint requests to Chen Hao, Center of Organ Transplantation, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, 197 Ruijin Er Road, Shanghai , P.R. China. Telephone: 86(0) ; haochendr@yahoo.com.cn DOI /lt Published online in Wiley InterScience ( American Association for the Study of Liver Diseases.

2 STRATEGY FOR ESTIMATING MYCOPHENOLIC ACID EXPOSURE 1685 large interindividual variability. It is possible that clinical outcomes could be improved by monitoring the performance of MPA exposure. Hence, concentrationcontrolled dosing of MMF based on MPA pharmacokinetic profiles may be more advantageous than a fixeddose regimen in order to maximize efficacy and minimize the incidence of both short- and long-term toxic side effects. It has been confirmed that MPA area under the concentration-time curve (AUC) is closely related to acute rejection episodes 7-9 and adverse events. 10,11 Full-time analysis of MPA-AUC at 12-hour intervals requires at least 8-10 plasma samples, thus making it impractical for clinical application. Hence, the estimation of MPA-AUC 0-12h by using a LSS has been suggested as a sufficiently precise but still practical method for the prediction of MPA exposure in adult 12,13 and pediatric renal transplant recipients and other organ transplant recipients. 17,18 For LSS study, Ting et al. 19 have some important suggestions, as follows. (1) It is essential to validate the predictive performance of the LSS in other patient populations. The prediction bias and prediction precision of the LSS should be determined. (2) A clinically feasible LSS should use 3 blood samples, preferably within a short period of time, to reduce the inconvenience of therapeutic drug monitoring. (3) The application of a specific LSS is ideally limited to the population and drug formulation that is used to develop it. In adult liver transplant patients, we have first established the abbreviated MPA pharmacokinetic profiles for prediction of MPA-AUC 0-12h by LSS. 20 We found that the relationship between estimated MPA-AUC 0-12h and measured MPA-AUC 0-12h based on 3 or 4 MPA pharmacokinetic parameters was related in some abbreviated models, but we have not prospectively validated these models. In this study, we tested them in another independent group of Chinese adult liver transplant recipients to choose the best model that satisfied 2 criteria: excellent precision and appropriateness in a clinical setting. PATIENTS AND METHODS Thirty adult liver transplant recipients were enrolled onto the current study. Patients were subject to a triple drug immunosuppressive regimen (TAC with corticosteroid and MMF). Three hundred blood plasma samples were measured by high-performance liquid chromatography (HPLC) for MPA pharmacokinetic analysis. The study was approved by the independent ethics committee of Ruijin Hospital. The procedure was described in detail to all patients before admission, and informed consent was obtained. Patients A total of 30 adult patients (28 men) undergoing orthotopic liver transplantation were enrolled onto this study in Organ Transplantation Center in Ruijin Hospital. The average age of patients was 51 years (range, years). The primary hepatic diseases for clinical diagnosis were hepatitis B viral cirrhosis with hepatocellular carcinoma (16 patients), hepatitis B viral cirrhosis (10 patients), severe hepatitis B (1 patient), intrahepatic cholestasis (1 patient), hepatic bile duct carcinoma (1 patient), and hepatic amyloidosis (1 patient). Immunosuppressive Regimen Triple immunosuppressive regimen was used: MMF (CellCept, Roche), tacrolimus (TAC; Prograf, Fujisawa), and corticosteroid. A total of 1.0 g of MMF was provided within 6 hours before liver transplantation and 1.0 g twice daily after surgery. The dose of MMF was reduced according to the occurrence of related side effects. TAC was provided 2 hours later after the administration of MMF. Trough concentration of TAC was monitored daily 2 hours after the morning dose of MMF. The dose of TAC was adjusted according to the target concentration of 5-10 ng/ml. A total of 500 mg of methylprednisolone was injected in all patients during the anhepatic period. For hepatocellular or hepatic bile duct carcinomas, 500 mg of methylprednisolone was provided intravenously on the first day after liver transplantation, and no more any corticosteroids were provided after this. For benign hepatic diseases, the dose of methylprednisolone was tapered after surgery and after 1 week; prednisone was provided at a dosage of 20 mg daily. Daclizumab (Zenapax, Roche) or basiliximab (Simulect, Novartis) were provided for immune induction therapy in some cases. Pharmacokinetic Protocol Several blood samples were taken from every recipient without side effects of MMF within 14 days (range, 6-14 days) after liver transplantation. Every pharmacokinetic profile consisted of 10 blood samples collected during the 12-hour dosing interval (before administration, and at 0.5, 1, 1.5, 2, 4, 6, 8, 10, and 12 hours after receipt of the dose). Patients were required to fast from 10:00 PM the night before sampling and then allowed to take food 30 minutes after sampling the next morning. The volume of blood taken for each pharmacokinetic parameter was 2 ml. All blood samples were collected in tubes containing ethylenediaminetetraacetic acid as an anticoagulant. All samples were stored at 20 C until analysis. A total of 30 pharmacokinetic profiles were available. Determination of Total Plasma MPA Concentration by HPLC Plasma concentration of total MPA was measured instantaneously by a validated HPLC method described previously. 21 The method of HPLC used in Institute of Clinical Pharmacology of Ruijin Hospital was ensured by participating in the MPA Proficiency Testing Scheme, provided by Cardiac and Vascular Sciences Analytic Unit, St. George s Hospital Medical School (London, UK). The software package WinNonlin Program, version 4.1, was used to calculate total MPA-AUC 0-12h,C max, and t max. Full 12-hour MPA-AUC was determined according to the linear trapezoidal rule.

3 1686 HAO ET AL. TABLE 1. Selected Models for the Prediction of MPA-AUC 0-12h by Limited Sampling Strategies 20 Model Sampling time (h) Model equation r 2 1 2, *C 2h 4.136*C 6h , 0.5, *C 0h 0.731*C 0.5h 1.877*C 2h , 1, *C 0h 0.713*C 1h 1.752*C 2h , 2, *C 1h 1.604*C 2h 4.116*C 4h , 2, *C 1h 1.750*C 2h 4.586*C 6h , 4, *C 2h 2.078*C 4h 3.085*C 6h , 2, 6, *C 1h 1.871*C 2h 3.176*C 6h 3.664*C 8h , 2, 4, *C 1h 1.573*C 2h 2.057*C 4h 3.543*C 6h Abbreviations: MPA, mycophenolic acid; AUC, area under the concentration-time curve; r 2, coefficient of determination. Statistical Analysis In our previous study, 20 we established a series of abbreviated models for estimation of MPA-AUC 0-12h by LSS (Table 1). Validation of 8 models by a new independent group of adult liver transplant recipients was undertaken. Linear regression analysis was used to compare the estimated MPA-AUC 0-12h from each model against the measured MPA-AUC 0-12h. After that, according to the method of Sheiner and Beal, 22 prediction bias was quantified as the percentage of prediction error 100 [(estimated MPA-AUC 0-12h MPA-AUC 0-12h)/MPA-AUC 0-12h ]. Prediction precision was quantified as the percentage of absolute prediction error 100 [(estimated MPA-AUC 0-12h MPA-AUC 0-12h )/ MPA-AUC 0-12h ]. Mean prediction bias and prediction precision (with 95% confidence intervals [95% CIs]) were calculated as the arithmetic mean of the prediction bias and the prediction precision for each patient for above models. The agreement between estimated MPA-AUC 0-12h and MPA-AUC 0-12h was assessed by a Bland-Altman plot, and the difference expressed with mean 2SD. To compare the Bland-Altman analysis derived from the 2 models, we used a mountain plot analysis. A mountain plot analysis (or folded empirical cumulative distribution plot) is created by computing a percentile for each ranked difference between a new method and a reference method. For a folded plot to be obtained, the following transformation is performed for all percentiles 50: percentile 100 percentile. These percentiles are then plotted against the differences between the 2 methods. 23 RESULTS In previous studies, some models have been reported that allow the prediction of MPA-AUC at 12-hour intervals by LSS. In this study, another independent group of 30 MPA pharmacokinetic profiles was used to test the predictive performance of 8 models. MPA Pharmacokinetic Parameters and Relationship Between MPA Concentration and MPA-AUC 0-12h In 30 pharmacokinetic profiles, mean t max was hours (range, hours). Mean C max was 7.45 TABLE 2. Relationship Between MPA-C 0h C 12h and MPA-AUC 0-12h Single MPA concentration Equation for estimation of MPA-AUC 0-12h r 2 P value C 0h 2.949E-02*C 0h C 0.5h 3.230E-02*C 0.5h C 1h 9.162E-02*C 1h C 1.5h 0.247*C 1.5h C 2h 0.190*C 2h C 4h 0.101*C 4h C 6h 7.205E-02*C 6h C 8h 4.829E-02*C 8h C 10h 4.296E-02*C 10h C 12h 4.068E-02*C 12h Abbreviation: MPA, mycophenolic acid; AUC, area under the concentration-time curve; r 2, coefficient of determination mg/l (range, mg/l). Mean MPA-AUC 0-12h was mg h/l (range, mg h/l). These parameters were similar to our previous data. The relationship between single time-point MPA concentration and measured MPA-AUC 0-12h was shown in Table 2. In agreement with previously published data, the correlation of MPA plasma concentrations from C 0h to C 12h with the corresponding MPA-AUC 0-12h was poor to moderate. Comparison of Each Model s Ability to Accurately Predict MPA-AUC 0-12h A comparison of measured MPA-AUC 0-12h vs. calculated MPA-AUC 0-12h for each model is shown in Figure 1. In Figure 1. Comparison of each selected model s ability to accurately predict MPA-AUC 0-12h. A total of 30 pharmacokinetic profiles were analyzed. (A) MPA-C 2h, C 6h, r , P < (B) MPA-C 0h,C 0.5h,C 2h, r , P < (C) MPA-C 0h,C 1h,C 2h, r , P < (D) MPA-C 1h,C 2h, C 4h, r , P < (E) MPA-C 1h, C 2h, C 6h, r , P < (F) MPA-C 2h,C 4h,C 6h, r , P < (G) MPA-C 1h,C 2h,C 6h,C 8h, r , P < (H) MPA-C 1h,C 2h,C 4h,C 6h, r , P <

4 STRATEGY FOR ESTIMATING MYCOPHENOLIC ACID EXPOSURE 1687 Figure 1.

5 1688 HAO ET AL. TABLE 3. Statistics for Residuals in 8 Models Model Minimum Maximum Mean SD agreement with previously published data, the best model (model 7) for assessment of MPA-AUC 0-12h was achieved by an abbreviated profile based on the 4 pharmacokinetic parameters C 1h,C 2h,C 6h, and C 8h, which comprises both early (C 1h,C 2h ) and late (C 6h,C 8h ) pharmacokinetic parameters (Fig. 1G). The predictive performance was also excellent in models 4 and 5, which were based on 3 pharmacokinetic parameters C 1h,C 2h,C 4h or C 1h,C 2h,C 6h (Fig. 1D,E). However, the former was superior to the latter in the clinical setting as a result of the shorter sampling interval. The ability to predict MPA-AUC 0-12h was moderate in model 2 and 3 based on the early MPA concentrations (C 0h,C 0.5h,C 2h and C 0h,C 1h,C 2h ) (Fig. 1B,C). The residuals in all 8 models were calculated (Table 3). All models have good mean residuals and symmetry. Model 7 had the smallest SD of residual. Comparison of the Estimated MPA-AUC 0-12h with the Measured MPA-AUC 0-12h by Prediction Bias and Prediction Precision A summary of 8 models respective performances in the estimation of measured MPA-AUC 0-12h is shown in Table 4 and Table 5. For the prediction bias, 8 models had CIs that included zero, thereby providing unbiased estimates of the MPA-AUC 0-12h. Model 8 had the best correlation and the smallest 95% CI. Models 4, 5, and 7 also had excellent performance. Other models showed poorer correlation and wider CIs. The median prediction bias for all models was 9.64% 7.52% (range, 1.43%-19.67%). For the prediction of precision, model 7 had the best correlation and the smallest 95% CI. Model 8 also had excellent performance. The range of 95% CI was acceptable in models 1, 4, 5, and 6. Poorer correlation and wider CIs were found in models 2 and 3. The median prediction precision for all models was 13.47% 5.63% (range, 5.11%-21.69%). Comparison of the Calculated MPA-AUC 0-12h With the Measured MPA-AUC 0-12h by the Bland-Altman Method A summary of 8 models detailed analysis by the Bland- Altman method is shown in Figure 2. In all models except for model 5, the estimated MPA-AUC 0-12h tended to underestimate the measured MPA-AUC 0-12h in the lower range of measured MPA-AUC 0-12h values, whereas in the upper range of measured MPA-AUC 0-12h values, the estimated MPA-AUC 0-12h tended to overestimate the measured MPA-AUC 0-12h. Four pharmacokinetic parameter algorithms (models 7 and 8) had the ability to predict the measured MPA-AUC 0-12h accurately, with the corresponding prediction variation of 7.88 mg h/l and 8.55 mg h/l ( 2SD; Fig. 2G,H). For considering 3 pharmacokinetic parameters, models 4 and 6 also had the corresponding prediction variation mg h/l and mg h/l ( 2SD; Fig. 2D,F). In mountain plot analysis (Fig. 3), model 7 was superior than the other 7 models. According to our previous study, calculated MPA- AUC 0-12h within 15% of measured MPA-AUC 0-12h is excellent. In this study, the results of models 5, 7, and 8 met our requirements (Table 6). In the report by Weber et al., 24 variation within 33% in MPA exposure is clinically acceptable, which is also defined in the FDCC trial. According to this standard, all models passed muster. DISCUSSION The studies of therapeutic drug monitoring of MPA exposure are reported mainly in renal transplantation. It has been shown that the best methodology is based on abbreviated pharmacokinetic parameters that use limiting sampling strategy (LSS). The basic data originate from a randomized, concentration-controlled trial in adult renal transplant patients 7,25 and pediatric renal transplant patients. 26,27 Recently some abbreviated equations adopting 2 or 4 time-point MPA concentrations within the sampling interval of 2 or 3 hours have been confirmed and validated The FDCC trial, in which MPA exposure is estimated by an abbreviated pharmacokinetic profile, and the OptiCept trial, in which MPA exposure is estimated by a 12-hour predose concentration, have finished and results will soon be published. In liver transplantation, there were some studies of MPA pharmacokinetics, 20,31-33 but none of them established of abbreviated MPA profiles for the estimation of MPA-AUC by LSS. Aw with coauthors 34,35 reported that MPA-C 0h was close to MPA-AUC 0-7h in children receiving liver transplants. Tredger et al. 36 set up the range of MPA-C 0h as the target of therapeutic drug monitoring for MMF. In our previous study, we had established a series of abbreviated models in adult liver transplant recipients for prediction of full-time MPA-AUC by LSS. From a total of 72 pharmacokinetic profiles, it was shown that measured MPA-AUC 0-12h could be predicted accurately by using MPA pharmacokinetic parameters C 1h,C 2h,C 6h, and C 8h. Some other models were less predictive but would be clinically useful. It is essential to validate the developed abbreviated models by LSS in another cohort of patient profiles (the validation or testing group) to ensure that its predictions are reliable. The predicted MPA-AUC should be compared with the measured value. We thus prospectively validated an algorithm for the estimation of MPA exposure

6 STRATEGY FOR ESTIMATING MYCOPHENOLIC ACID EXPOSURE 1689 TABLE 4. Prediction Bias of Limited Sampling Strategies to Estimate MPA-AUC 0-12h Model r 2 95% CI (%) Median Minimum Maximum 5th 25th 50th 75th 95th to to to to to to to to Abbreviations: MPA, mycophenolic acid; AUC, area under the concentration-time curve; r 2, coefficient of determination; 95% CI, 95% confidence interval. TABLE 5. Prediction Precision of the Limited Sampling Strategies to Estimate MPA-AUC 0-12h Model r 2 95% CI (%) Median Minimum Maximum 5th 25th 50th 75th 95th Abbreviations: MPA, mycophenolic acid; AUC, area under the concentration-time curve; r 2, coefficient of determination; 95% CI, 95% confidence interval. based on a LSS in an independent group of Chinese liver transplant recipients. In the current study, we observed that an algorithm (model 7) based on 4 pharmacokinetic sampling timepoint MPA concentrations during a 7-hour interval is able accurately predict the measured MPA-AUC 0-12h with the best correlation of determination (r ) when compared with other models. Sheiner and Beal 22 suggest that 2 main criteria are essential for the assessment of prediction: bias and precision. The lower the prediction bias and prediction precision, the more accurate and precise the predictions are. It is also suggested that an acceptable range of relative prediction bias and prediction precision in clinical studies is 15%- 20%. 37 In this model, median prediction bias and median prediction precision were excellent (2.18% and 5.11%, respectively). This algorithm can also be reliably applied to predict MPA exposure when considering the best agreement between estimated MPA-AUC 0-12h and full-time MPA-AUC 0-12h. Model 8, similar to model 7, also showed excellent performance of prediction. These results suggest the importance of enterohepatic cycling of MPA in liver transplant patients. In some liver transplant patients, the enterohepatic cycling reestablishes at 4-8 hours after administration of MMF. 31 Receipients of TAC have longer enterohepatic cycling than those who receive cyclosporine. 38,39 It is reasonable that the best model should include one time-point MPA concentration during the interval 4-12 hours after receiving the drug. However, these 2 models for the estimation of MPA- AUC 0-12h required 4 pharmacokinetic sampling time points with longer interval after dosing, which is inconvenient in the routine clinical setting. In renal transplantation, the abbreviated equation that uses only 3 MPA sampling time points during the first 2 hours could precisely predict the MPA exposure. 29 We had hypothesized that the algorithm based on 3 MPA concentrations within 2 hours after MMF dosing could be established in adult liver transplantation. However, the results in this study do not support our hypothesis. Two models including 3 MPA sampling time points (C 0h, C 0.5h,C 2h or C 0h,C 1h,C 2h ) had poor predictive performance and poor correlation of determination (r and 0.481). The mean 95% CIs for prediction bias and prediction precision were higher than other 6 models. The agreement and variation between estimated MPA-AUC 0-12h and measured MPA-AUC 0-12h were also poor. Among all models, only model 4 containing MPA C 1h, C 2h, and C 4h had an advantage for clinical convenience. This algorithm had shorter sampling intervals and

7 1690 HAO ET AL. Figure 2. Comparison between measured MPA-AUC 0-12h and estimated MPA-AUC 0-12h in selected models according to the method of Bland-Altman analysis. The 2SD of models 1-8 were, respectively, 12.94, 21.28, 21.55, 13.23, 19.14, 10.72, 7.88, and 8.55.

8 STRATEGY FOR ESTIMATING MYCOPHENOLIC ACID EXPOSURE 1691 Figure 3. Comparison of estimated MPA-AUC 0-12h between model 7 (open squares) and the other 7 models (black cross) by mountain plot analysis. fewer sampling time points than models 7 and 8 did. However, it also had excellent ability for prediction of full-time MPA-AUC with the correlation of determination (r ). The prediction bias was excellent, and the prediction precision was clinically acceptable. A total of 27 of 30 estimated MPA-AUC 0-12h were within 33% of measured MPA-AUC 0-12h. This model could therefore be applied in the outpatient setting. LSS offers a practical approach to investigate the potential of MPA AUC monitoring in routine clinical application. However, LSS has certain limitations. 37,40 It depends on timed concentrations for prediction of

9 1692 HAO ET AL. TABLE 6. Estimated MPA-AUC 0-12h Compared With Measured MPA-AUC 0-12h Compared with measured MPA-AUC 0-12h (n) Within 15% or Compared with measured MPA-AUC 0-12h (n) Within 33% or Model 15% (n) 15% (n) Percentage 15% (n) 33% (n) Percentage Abbreviations: MPA, mycophenolic acid; AUC, area under the concentration-time curve; n, profile number. MPA-AUC 0-12h, so sampling time is critical. Deviation from the exact sampling times may affect the accuracy of prediction for MPA-AUC 0-12h, especially in models that used early sampling times, because of rapid change of MPA concentration during the absorption and distribution phases. On the other hand, a Bayesian approach has been evaluated in renal transplantation. 41,42 Compared with LSS, the Bayesian approach has some advantages. 19 Sampling time is more flexible; and it is not so necessary to collect samples at exact time points. New data may be continuously added to the population parameters, so more accurate predictions may be made. To our knowledge, this is the first study to prospectively validate abbreviated MPA pharmacokinetic profiles for estimation of MPA exposure that is based on a LSS in adult liver transplant recipients. Because only a few MPA pharmacokinetic profiles were tested in this study, the models we present should be further tested with larger data sets from other centers. In conclusion, the results in the current study exhibited that measured MPA-AUC 0-12h could be accurately predicted by MPA pharmacokinetic parameters C 1h, C 2h,C 6h, and C 8h. The abbreviated equation based on MPA pharmacokinetic parameters C 1h, C 2h, and C 4h had excellent predictive performance and was more useful for clinical application. ACKNOWLEDGMENTS The authors did not receive any funding for this study. The authors have no conflicts of interest that are directly relevant to the content of this study. We thank Lu Hui for reviewing the article in manuscript and Tao Ma for his help with the statistical analysis. REFERENCES 1. Titte RS, Bruce K, Herwig-Ulf MK. Mycophenolate mofetil in solid-organ transplantation. Expert Opin Pharmacother 2003;4: Lipsky JJ. Mycophenolate mofetil. Lancet 1996;348: Kuypers DR, Vanrenterghem YC. Tailoring immunosuppressive therapy. Nephrol Dial Transplant 2002;17: Cox VC, Ensom MHH. Mycophenolate mofetil for solid organ transplantation: does the evidence support the need for clinical pharmacokinetic monitoring? Ther Drug Monit 2003;25: van Gelder T, Le Meur Y, Shaw LM, Oellerich M, DeNofrio D, Holt C, et al. Therapeutic drug monitoring of mycophenolate mofetil in transplantation. Ther Drug Monit 2006; 28: Grasser B, Iberer F, Schaffellner S, Kniepeiss D, Stauber R, Koshsorur G, et al. Trough level guided mycophenolate mofetil rejection prophylaxis in liver transplantation. Transplant Proc 2001;33: van Gelder T, Hilbrands LB, Vanrenterghem Y, Weimar W, de Fijter JW, Squifflet JP, et al. A randomized doubleblind, multicenter plasma concentration controlled study of the safety and efficacy of oral mycophenolate mofetil for the prevention of acute rejection after kidney transplantation. Transplantation 1999;68: Yamani MH, Starling RC, Goormastic M, Van Lente F, Smedira N, McCarthy P, et al. The impact of routine mycophenolate mofetil drug monitoring on the treatment of cardiac allograft rejection. Transplantation 2000;69: DeNofrio D, Loh E, Kao A, Korecka M, Pickering FW, Craig KA, et al. Mycophenolic acid concentrations are associated with cardiac allograft rejection. J Heart Lung Transplant 2000;19: Mourad M, Malaise J, Chaib Eddour D, De Meyer M, Konig J, Schepers R, et al. Correlation of mycophenolic acid pharmacokinetic parameters with side effects in kidney transplant patients treated with mycophenolate mofetil. Clin Chem 2001;47: Mourad M, Malaise J, Chaib Eddour D, De Meyer M, Konig J, Schepers R, et al. Pharmacokinetic basis for the efficient and safe use of low-dose mycophenolate mofetil in combination with tacrolimus in kidney transplantation. Clin Chem 2001;47: van Hest RM, Mathot RA, Vulto AG, Le Meur Y, van Gelder T. Mycophenolic acid in diabetic renal transplant recipients: pharmacokinetics and application of a limited sampling strategy. Ther Drug Monit 2004;26: Johnson AG, Rigby RJ, Taylor PJ, Jones CE, Allen J, Franzen K, et al. The kinetics of mycophenolic acid and its glucuronide metabolite in adult kidney transplant recipients. Clin Pharmacol Ther 1999;66:

10 STRATEGY FOR ESTIMATING MYCOPHENOLIC ACID EXPOSURE Filler G. Abbreviated mycophenolic acid AUC from C0, C1, C2, and C4 is preferable in children after renal transplantation on mycophenolate mofetil and tacrolimus therapy. Transpl Int 2004;17: Schutz E, Armstrong VW, Shipkova M, Weber L, Niedmann PD, Lammersdorf T, et al. Limited sampling strategy for the determination of mycophenolic acid area under the curve in pediatric kidney recipients. Transplant Proc 1998;30: David-Neto E, Pereira Araujo LM, Sumita NM, Mendes ME, Ribeiro Castro MC, Alves CF, et al. Mycophenolic acid pharmacokinetics in stable pediatric renal transplantation. Pediatr Nephrol 2003;18: Ng J, Rogosheske J, Barker J, Weisdorf D, Jacobson PA. A limited sampling model for estimation of total and unbound mycophenolic acid (MPA) area under the curve (AUC) in hematopoietic cell transplantation (HCT). Ther Drug Monit 2006;28: Dosch AO, Ehlermann P, Koch A, Remppis A, Katus HA, Dengler TJ. A comparison of measured trough levels and abbreviated AUC estimation by limited sampling strategies for monitoring mycophenolic acid exposure in stable heart transplant patients receiving cyclosporin A containing and cyclosporin A free immunosuppressive regimens. Clin Ther 2006;28: Ting LS, Villeneuve E, Ensom MH. Beyond cyclosporine: a systematic review of limited sampling strategies for other immunosuppressants. Ther Drug Monit 2006;28: Chen H, Peng C, Yu Z, Shen B, Deng X, Qiu W, et al. Pharmacokinetics of mycophenolic Acid and determination of area under the curve by abbreviated sampling strategy in Chinese liver transplant recipients. Clin Pharmacokinet 2007;46; Yu ZC, Cai WM, Xu D, Wang XH. HPLC determination of mycophenolate mofetil and its active metabolite mycophenolic acid in human plasma. Chin J Pharm Anal 2005;25: Sheiner LB, Beal SL. Some suggestions for measuring predictive performance. J Pharmacokinet Biopharm 1981; 9: Krouwer JS, Monti KL. A simple, graphical method to evaluate laboratory assays. Eur J Clin Chem Clin Biochem 1995;33: Weber LT, Hoecker B, Armstrong VW, Oellerich M, Tonshoff B. Validation of an abbreviated pharmacokinetic profile for the estimation of mycophenolic acid exposure in pediatric renal transplant recipients. Ther Drug Monit 2006;28: Hale MD, Nicholls AJ, Bullingham RE, Hene R, Hoitsma A, Squifflet JP, et al. The pharmacokinetic pharmacodynamic relationship for mycophenolate mofetil in renal transplantation. Clin Pharmacol Ther 1998;64: Weber LT, Shipkova M, Armstrong VW, Wagner N, Schutz E, Mehls O, et al. The pharmacokineticpharmacodynamic relationship for total and free mycophenolic acid in pediatric renal transplant recipients: a report of the German study group on mycophenolate mofetil therapy. J Am Soc Nephrol 2002;13: Weber LT, Shipkova M, Armstrong VW, Wagner N, Schutz E, Mehls O, et al. Comparison of the Emit immunoassay with HPLC for therapeutic drug monitoring of mycophenolic acid in pediatric renal-transplant recipients on mycophenolate mofetil therapy. Clin Chem 2002;48: Le Guellec C, Buchler M, Giraudeau B, Le Meur Y, Gakoue JE, Lebranchu Y, et al. Simultaneous estimation of cyclosporin and mycophenolic acid areas under the curve in stable renal transplant patients using a limited sampling strategy. Eur J Clin Pharmacol 2002;57: Pawinski T, Hale M, Korecka M, Fitzsimmons WE, Shaw LM. Limited sampling strategy for the estimation of mycophenolic acid area under the curve in adult renal transplant patients treated with concomitant tacrolimus. Clin Chem 2002;48: Payen S, Zhang D, Maisin A, Popon M, Bensman A, Bouissou F, et al. Population pharmacokinetics of mycophenolic acid in kidney transplant pediatric and adolescent patients. Ther Drug Monit 2005;27: Fatela-Cantillo D, Hinojosa-Perez R, Peralvo-Rodriguez MI, Serrano-Diaz Canedo J, Gomez-Bravo MA. Pharmacokinetic evaluation of mycophenolic acid profiles during the period immediately following an orthotopic liver transplant. Transplant Proc 2006;38: Pisupati J, Jain A, Burckart G, Hamad I, Zuckerman S, Fung J, Venkataramanan R. Intraindividual and interindividual variations in the pharmacokinetics of mycophenolic acid in liver transplant patients. J Clin Pharmacol 2005;45: Manzia TM, De Liguori Carino N, Orlando G, Toti L, De Luca L, D Andria D, et al. Use of mycophenolate mofetil in liver transplantation: a literature review. Transplant Proc 2005;37: Aw MM, Brown NW, Itsuka T, Gonde CE, Adams JE, Heaton ND, et al. Mycophenolic acid pharmacokinetics in pediatric liver transplant recipients. Liver Transpl 2003; 9: Brown NW, Aw MM, Mieli-Vergani G, Dhawan A, Tredger JM. Mycophenolic acid and mycophenolic acid glucuronide pharmacokinetics in pediatric liver transplant recipients: effect of cyclosporine and tacrolimus comedication. Ther Drug Monit 2002;24: Tredger JM, Brown NW, Adams J, Gonde CE, Dhawan A, Rela M, et al. Monitoring Mycophenolate in liver transplant recipients: toward a therapeutic range. Liver Transpl 2004;10: David OJ, Johnston A. Limited sampling strategies for estimating cyclosporin area under the concentration-time curve: review of current algorithms. Ther Drug Monit 2001;23: Hesselink DA, van Hest RM, Mathot RA, Bonthuis F, Weimar W, de Bruin RW, et al. Cyclosporine interacts with mycophenolic acid by inhibiting the multidrug resistanceassociated protein 2. Am J Transplant 2005;5: van Gelder T, Klupp J, Barten MJ, Christians U, Morris RE. Comparison of the effects of tacrolimus and cyclosporine on the pharmacokinetics of mycophenolic acid. Ther Drug Monit 2001;23: Rousseau A, Marquet P. Application of pharmacokinetic modeling to the routine Ther Drug Monit of anticancer drugs. Fundam Clin Pharmacol 2002;16: Premaud A, Le Meur Y, Debord J, Szelag JC, Rousseau A, Hoizey G, et al. Maximum a posteriori Bayesian estimation of mycophenolic acid pharmacokinetics in renal transplant recipients at different postgrafting periods. Ther Drug Monit 2005;27: van Hest R, Mathot R, Vulto A, Weimar W, van Gelder T. Predicting the usefulness of therapeutic drug monitoring of mycophenolic acid: a computer simulation. Ther Drug Monit 2005;27: