Introduction.

Plasma is the traditional biological sample for PK studies and Therapeutic Drug Monitoring (TDM) of the pivotal anti-tuberculosis (TB) drug rifampicin. Saliva may be an attractive alternative matrix, also considering that it may reflect protein-unbound, active plasma concentrations. The objectives of this study were (i) to compare the PK of rifampicin in saliva and plasma and (2) to assess whether saliva could be an alternative matrix for PK studies and TDM with this drug.

Methods.

Adescriptive PK study was performed among 15 adult Tanzanian TB patients who were in the intensive phase of TB treatment. Time-matched samples of stimulated saliva (obtained with a Salivette® device containing citric acid) and plasma were collected at predose and at 1, 2, 3, 4, 6, 8, 10 and 24 hours after intake of rifampicin. Salivary, total (protein-unbound plus bound) and unbound plasma concentrations of rifampicin were measured with validated HPLC methods. Salivary and plasma PK parameters were assessed, ratios of salivary to (total and unbound) plasma concentrations were calculatedandthe performance of salivary concentrations to predict plasma concentrations was evaluated using the jackknife method.

Results.

The geometrie mean AUC0_24 of rifampicin in saliva (3.1 h*mg/L) was slightly but signifïcantly lower than the protein-unbound AUC0_24 in plasma (5.3 h*mg/L) and these were much lower than the plasma AUC0_24 based on total concentrations (32.7 h*mg/L). Tmax and elimination half-lives of rifampicin based on salivary, proteinunbound and total plasma concentrations were similar. Geometrie mean concentration ratios for salivary versus total and unbound plasma concentrations were 0.099 and 0.614 and these ratios were not dependent on time post dose or associated rifampicin plasma concentrations (repeated measures ANOVA). The prediction of total and unbound plasma concentrations based on salivary concentrations of rifampicin resulted in median percentage prediction errors (MPPE) of 13.4% and 6.0% and median absolute percentage prediction errors (MAPE) of 35.7% and 23.0%, respectively, which means that these predictions were sufficiently accurate (<15%) yet imprecise (>15%).

Condusions.

AUC0.24 and Cmax of rifampicin in saliva were much lower than those in plasma based on total plasma concentrations. The AUC0_24 of rifampicin in saliva was of the same order of magnitude as the protein-unbound plasma AUC0_24, but was signifïcantly lower. It is not possible to predict total or protein-unbound plasma concentrations from salivary concentrations, due to inadequate precision associated with this prediction.