Premium
Limited sampling strategy for predicting isoniazid exposure in patients with extrapulmonary tuberculosis
Author(s) -
Alshaikheid Mohammed,
Chaabane Amel,
Ben Fredj Nadia,
Ben Brahim Hajer,
Ben Fadhel Najah,
Chadli Zohra,
Slama Ahlem,
Boughattas Naceur A.,
Chakroun Mohamed,
Aouam Karim
Publication year - 2020
Publication title -
journal of clinical pharmacy and therapeutics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.622
H-Index - 73
eISSN - 1365-2710
pISSN - 0269-4727
DOI - 10.1111/jcpt.13098
Subject(s) - isoniazid , mean squared error , area under the curve , medicine , mathematics , confidence interval , statistics , sampling (signal processing) , linear regression , tuberculosis , pharmacokinetics , dosing , filter (signal processing) , pathology , computer science , computer vision
What is known and objective Limited sampling strategies (LSS), using few sampling times after dosing, have been used to reliably predict the isoniazid area under the 24‐hour concentration‐time curve (AUC). Experience with isoniazid is very limited, and no LSS has been developed in south‐Mediterranean populations. Hence, we aimed to develop an accurate and convenient LSS for predicting isoniazid AUC in Tunisian patients with extrapulmonary tuberculosis. Methods Pharmacokinetic profiles consisting of six blood samples each, collected during the 24‐hour dosing interval, were obtained from 25 (6 men and 19 women) Tunisian patients with extrapulmonary tuberculosis. The AUC was calculated according to the linear trapezoidal rule. The isoniazid concentrations at each sampling time were correlated by a linear regression analysis with the measured AUC. We analysed all the developed models for their ability to estimate the isoniazid AUC. Error indices including the percentage of Mean Absolute Prediction Error (%MAE) and the percentage of Root Mean Squared Prediction Error (%RMSE) were used to evaluate the predictive performance. The agreement between predicted and measured AUCs was investigated using Bland and Altman and mountain plot analyses. Results and discussion Among the 1‐time‐point estimations, the C 3 ‐predicted AUC showed the highest correlation with the measured one ( r 2 = .906, %MAE = 10.45% and %RMSE = 2.69%). For the 2‐time‐point estimations, the model including the C 2 and C 6 provided the highest correlation between predicted and measured isoniazid AUC ( r 2 = .960, %MAE = 8.02% and %RMSE = 1.75%). The C 0 /C 3 LSS model provided satisfactory correlation and agreement ( r 2 = .930, %MAE = 10.19% and %RMSE = 2.32%). The best multilinear regression model for predicting the full isoniazid AUC was found to be the combination of 3 time points: C 0 , C 1 and C 6 ( r 2 = .992, %MAE = 4.06% and %RMSE = 0.80%). The use of a 2‐time‐point LSS to predict AUC in our population could be sufficient. C 2 /C 6 combination has shown the best correlation but the use of the C 0 /C 3 combination could be more practical with an accurate prediction. Therapeutic drug monitoring of isoniazid based on the C 3 can be used also in daily clinical practice in view of its reliability and practicality. What is new and conclusion The LSS using C 0 and C 3 is reliable, accurate and practical to estimate the AUC of isoniazid. A 1‐time‐point LSS including C 3 had acceptable correlation coefficient and prediction error indicators could be used alternatively.