The prediction error in the Bayesian analysis program for digoxin was evaluated in Japanese patients, and factors influencing the accuracy were investigated. Serum concentrations of digoxin were monitored two times and were compared with the predicted values obtained by using the Bayesian analysis program. The prediction error at the first time was 43.1%. Although this estimation error was reasonably restored at the second time of monitoring, the prediction error remained at 26.6%. These data suggested that unknown factors not included in the program affected the serum concentration of digoxin. Retrospective research of the digoxin serum concentrations in the patients suggested the coadministration of the drugs, which were the P-glycoprotein modulators, as well as the unexpected alteration of the serum creatinine, were the important factors influencing the prediction of the drug serum concentrations. We next examined the inhibitory effect of quinidine, verapamil and spironolactone on the transcellular transport of digoxin by using human P-glycoprotein overexpressing LLC-GA5-COL150 cells. Quinidine, verapamil and spironolactone could inhibit the transcellular transport of digoxin by 50%. In addition, the reduction of the renal clearance by 50%, which could possibly be caused by this inhibition, led to the increase of 36% in the steady state through concentrations of digoxin in the physiological pharmacokinetic model. In conclusion, the prediction of long-term serum concentration-time profiles of digoxin, based on the Bayesian analysis, will be disturbed by the coadministration of the P-glycoprotein modulators and the unexpected alteration of the serum creatinine.