Population Pharmacokinetics of Rifabutin in Human Immunodeficiency Virus-Infected Patients

Abstract

Rifabutin pharmacokinetics were studied by the population approach (NONMEM) with 40 human immunodeficiency virus-infected patients receiving rifabutin at different doses for prophylaxis or therapy of mycobacterial infections. A two-compartment open model with first-order absorption was used as the structural pharmacokinetic model. Parameter estimates were the absorption rate constant (0.201/h), clearance/bioavailability (CL/F; 60.9 liters/h), volume of the central compartment/bioavailability (231 liters), intercompartmental clearance (60.3 liters/h), and volume of the peripheral compartment/bioavailability (V_p/F; 1,050 liters). The distribution and elimination half-lives were 1.24 and 25.4 h, respectively. The covariates tested for influence on CL/F and V_p/F were sex, age, weight, height, body surface area, tobacco smoking, drug addiction, alanine aminotransferase levels, creatinine clearance, total protein, bilirubin, numbers of CD4⁺ cells, presence of diarrhea, cachexia index, rifabutin use (prophylaxis versus therapy), rifabutin dose, study site, and the concomitant administration of clarithromycin, fluconazole, phenobarbital, ciprofloxacin, azithromycin, or benzodiazepines. The only statistically significant effects on rifabutin pharmacokinetic parameters were a 27% decrease in V_p/F due to the concomitant administration of azithromycin and a 39% increase in V_p/F due to tobacco smoking. Such effects may be considered clinically unimportant. Our results confirm the lack of a correlation of rifabutin pharmacokinetic parameters with parameters of disease progression and gastrointestinal function. Also, the lack of a correlation with covariates which were previously found to be significant, such as concomitant fluconazole and clarithromycin use, may suggest that the effect of such covariates may be less important in the real clinical setting, in which several concomitant factors may influence pharmacokinetic parameters, with an overall effect of no apparent correlation.

Rifabutin has been shown to be an effective agent both for prophylaxis (18) and treatment (23) of disseminated Mycobacterium avium complex disease in patients infected with human immunodeficiency virus (HIV). Rifabutin may also be used for treatment of multidrug-resistant Mycobacterium tuberculosis, even though a formal study has never been carried out. The clinically significant interactions of HIV protease inhibitors with rifampin (13) may foster the use of rifabutin in some clinical settings. For example, it has been hypothesized that rifabutin could be used for the treatment of tuberculosis in patients concomitantly treated with indinavir (2).

The absorption and disposition of drugs in the HIV-positive patient population are usually characterized by a greater variability than those in other patient populations due to the broad heterogeneity of the patient’s clinical status and concomitant drug intake. The effect of HIV infection on drug pharmacokinetics appears to be unpredictable (9, 11, 12, 20, 29). For the majority of drugs investigated regarding such aspect, a decrease of drug exposure has been reported (11, 12, 20), possibly due to a decrease of absorption as a consequence of gastrointestinal dysfunction, even though other mechanisms involving changes in protein binding, volume of distribution (V), or clearance (CL) cannot be excluded a priori (9, 29). It has been reported that the parameters of rifabutin exposure, such as the area under the plasma concentration-versus-time curve (AUC) and the peak concentration in plasma (C_max), may be independent of HIV disease stage when patient’s CD4⁺ cell counts are used as a marker of disease progression (4, 16). However, parameters such as the presence of diarrhea and cachexia index (1 − actual patient weight/ideal body weight) may be more predictive of differences in rifabutin absorption and disposition than the CD4⁺ cell count. We prospectively studied the influence of such parameters as well as other covariates on rifabutin pharmacokinetics with minimal restrictions to patient enrollment in an attempt to obtain data from a broadly heterogeneous patient population.

(This study was presented at the 37th Interscience Conference on Antimicrobial Agents and Chemotherapy, Toronto, Ontario, Canada, 28 September to 1 October 1997.)

MATERIALS AND METHODS

Five medical centers (centers 1, 3, 4, 5, and 6) participated in the study. The numbers of patients enrolled at each center were as follows: 6 at center 1, 15 at center 4, 9 at center 3, 7 at center 5, and 3 at center 6. The patients’ demographics are presented in Table 1.

TABLE 1.

Patient demographics^a

Characteristic	Value
Age (yr)	34 (25–47)
Wt (kg)	52 (34–76)
Ht (cm)	167 (149–180)
Body surface area (m2)	1.56 (1.22–1.89)
No. of patients
Gender
Male	28
Female	12
Risk factor
Drug addiction	36
Other	4
Rifabutin use
Prophylaxis	13
Therapy	27
Rifabutin dose
150 mg twice daily	8
150 mg three times daily	2
300 mg daily	13
450 mg daily	10
600 mg daily	7
Diarrhea	8
Tobacco smoker	27
Receiving clarithromycin	9
Receiving fluconazole	23
Receiving phenobarbital	10
Receiving ciprofloxacin	10
Receiving azithromycin	7
Receiving benzodiazepines	9
Cachexia index	17.06 (−15.3–44.6)
Serum creatinine concn (mg/dl)	0.89 (0.30–1.89)
ALT concn (IU/liter)	50.6 (8–296)
Total protein concn (g/dl)	7.1 (4.8–11.5)
Bilirubin concn (mg/dl)	1.4 (0.2–18.3)
CD4+ lymphocyte count (cells/μl)	44.8 (0–230)

^aContinuous covariates are indicated as mean (range). 

A complete medical history, a physical examination, and a panel of laboratory tests consisting of a chemistry screen and a complete blood cell count with differential and platelet count were available within a week of the pharmacokinetic study. Medications administered concomitantly or within 2 weeks before the first study day were recorded.

Patient demographics are summarized in Table 1. Rifabutin (Pharmacia and Upjohn S.p.A., Milan, Italy) was administered at dosages of 150 mg twice daily to 8 patients, 150 mg three times a day to 2 patients, 300 mg daily to 13 patients, 450 mg daily to 10 patients, and 600 mg daily to 17 patients. Thirteen patients received rifabutin for prophylaxis of M. avium complex infection. Twenty-three of the remaining patients received rifabutin for the treatment of suspected or culture-proven atypical mycobacteriosis, while four were treated for extrapulmonary tuberculosis. Patients were instructed to take rifabutin tablets at least 1 h before or 2 h after a meal. Patient compliance was assessed by questioning them. Only patients who assured the investigators that they had complied with the dosage regimen were studied. In order to describe the concentration-versus-time curves following steady-state dosing of rifabutin (at least 2 weeks of administration of the same dosage regimen), an attempt was made to collect plasma samples at time zero (before drug administration) and, subsequently, at random within each of the following time intervals after drug administration: 0 to 4, 4 to 12, 12 to 24, 24 to 48, and 48 to 96 h. Each patient was studied once.

Blood samples were collected in tubes containing heparin, placed on ice, and centrifuged within 30 min after collection. Plasma was harvested and stored at −70°C until it was assayed for rifabutin concentrations.

The levels of rifabutin in plasma were determined by a high-pressure liquid chromatography assay with clothiapine (Merck, Rahway, N.J.) as an internal standard. Rifabutin powder was kindly provided by Pharmacia and Upjohn S.p.A.

Extraction of analytes from plasma samples was done by adding 1 ml of phosphate buffer (pH 7.0) to 1 ml of plasma and then by adding 6 ml of the mixture benzene-methylene chloride (9:1; vol/vol) and vortexing for 3 min. The mixture was then centrifuged, and the organic layer was taken to dryness in a gentle stream of nitrogen. The residue was then reconstituted with the mobile phase and injected into the column. Separation of analytes was obtained with a Merck Select B column (catalog no. 19608; Merck, Darmstadt, Germany) by using as the mobile phase a mixture of acetonitrile-water (55:45; vol/vol) containing sodium dodecyl sulfate (0.01 M) and glacial acetic acid (0.5%) at a flow rate of 1 ml/min. Detection was done with a diode array detector (1040A; Hewlett-Packard, Palo Alto, Calif.) set at a wavelength of 280 nm. The standard curve was linear in the range of 20 to 400 ng/ml. Interday and intraday relative standard deviations determined at concentrations of 30, 150, and 350 ng/ml (n = 7) were less than 10%. The lower limit of quantitation was 20 ng/ml. Plasma samples in which rifabutin concentrations exceeded the calibration range were diluted with blank plasma and reanalyzed.

The software NONMEM (version IV, double precision, level 2.0) (19) was run on a Sun Sparc10 computer and was used to fit the 213 concentrations in plasma obtained for 40 patients. NONMEM allows the estimation of (i) the parameters of the structural model, i.e., the pharmacokinetic parameters; (ii) the parameters characterizing the fixed effects on the pharmacokinetic parameters, i.e., the coefficients describing the relationship between a pharmacokinetic parameter and covariates such as age, sex, and stage of disease (such coefficients allow the estimation of the “typical” pharmacokinetic parameter for an individual with certain values for, e.g., age and weight); (iii) the size of the interindividual variability in the pharmacokinetic parameters, i.e., the magnitude of variability which is not explained by the regression formula relating the pharmacokinetic parameter to the significant covariates affecting it; and (iv) the size of the intraindividual (residual) variability, i.e., the difference between the concentration which has been observed and the concentration which has been predicted by the model. All of the parameters listed above are estimated simultaneously. The final model which describes the data best is achieved through a model-building process which is based on evaluation of different structural models, the addition or deletion of covariates, and the evaluation of different models (e.g., additional or proportional) for random variables (inter- and intraindividual variability). Model selection is done with the objective function, an estimate of goodness of fit, which is minus twice the log likelihood of the data. The difference in objective function between a full and a reduced model is approximately distributed in a chi-square manner with q degrees of freedom, where q is the number of parameters whose values are fixed in the reduced model.

RESULTS

The number of plasma drug concentrations available for analysis were 36 at time zero and 39, 38, 34, 38, and 28 at each of the following sampling time intervals: 0 to 4, 4 to 12, 12 to 24, 24 to 48, and 48 to 96 h, respectively. All six planned samples were drawn from 30 patients. For the remainder of the patients the number of samples drawn per patient was five samples for each of six patients, one sample for each of four patient, one sample for each of two patients, and two samples for one patient. The decline in the concentration of rifabutin following oral administration was best described by a two-compartment open model with first-order rate constants for absorption and elimination by using the program-supplied routine ADVAN4 TRANS4. Such a structural model was chosen on the basis of visual inspection of the data and on the basis of the information gathered from the literature pertaining to rifabutin pharmacokinetics. Also, an informal test of the different structural models (e.g., one versus two compartments) was carried out with the NONMEM objective function. The model was parameterized in terms of the absorption rate constant (K_a); the volume of the central compartment/absolute bioavailability (V_c/F); the volume of the peripheral compartment/absolute bioavailability (V_p/F), which is the volume of distribution at steady state (V_SS)/absolute bioavailability minus V_c/F; intercompartmental clearance (Q/F); and CL/absolute bioavailability (CL/F). Interpatient variability in the pharmacokinetic parameters and residual variability were modeled with a constant coefficient of variation: A = A′(1 + B). In the case of interpatient variability A is θ_j, the typical value of a pharmacokinetic parameter in the jth individual; A′ is θ′_j, the predicted pharmacokinetic parameter; and B is ηθ_j, the random difference between the true value of the pharmacokinetic parameter for an individual and the value predicted by the model. In the case of residual variability, A is C_ij, the observed ith concentration in the jth individual; A′ is C′_ij, the predicted ith concentration in the jth individual; and B is ɛ_ij, the random difference between the predicted and the observed concentrations. Such models were found to be superior in minimizing the objective function compared with the other models tested, e.g., the proportional-plus-constant (tested only for residual variability), additive, or exponential model. Preliminary evaluation suggested that the data supported parameterization of interindividual variability only for CL/F and V_p/F. First-order approximation was assumed. The first-order conditional estimation method was used to confirm the most relevant models of the model-building process including the final model. This was done because with the first-order method the interindividual variability is assumed to be represented by a distribution with a mean equal to 0 and a variance of ω², while the first-order conditional estimation method allows first-order expansions about the values of the Bayes estimates of the interindividual errors and may provide less biased estimates than the first-order method in selected situations such as in the case of nonlinear systems.

Several covariates were tested for their influence on CL/F and V_p/F by addition to the basic model. The influence of the covariates (categorical and continuous) on the pharmacokinetic parameters was modeled to be a percentage as exemplified in the following manner for CL/F: CL/F = θ_n + θ_n × θ_m × cov_m, in which θ_n is the typical value for the pharmacokinetic parameter in the patients without the covariate m (cov_m), and θ_m is the percent modification of θ_n when the covariate m is in the model. Each covariate was considered significant if its addition to the model resulted in a reduction of the objective function of a factor of greater than 3.8 (P < 0.05 compared with a chi-square distribution with 1 degree of freedom). Covariates were considered borderline and were not included in the final model if they significantly reduced the objective function but the null value was included within the 95% confidence interval. Plots of weighted residuals versus predicted concentrations were also used as an additional help in model building. All significant covariates were included in a full model and were deleted one by one in order to reevaluate significance when other covariates were included in the model. A covariate was not retained in the model if its deletion did not lead to an increase in the objective function of greater than 3.8. The following covariates were tested for their effect on CL/F and V_p/F: gender, age, weight, height, body surface area, tobacco smoking (>10 cigarettes/day), drug addiction, alanine aminotransferase (ALT) levels, creatinine clearance calculated by using the serum creatinine level (3), total protein, total bilirubin, CD4⁺ cell count, the presence of diarrhea, cachexia index, rifabutin use (prophylaxis versus therapy), dose, study site, and concomitant administration of clarithromycin, fluconazole, phenobarbital, ciprofloxacin, azithromycin, or benzodiazepines.

A summary of the model-building process is reported in Table 2. In the phase of addition of the individual covariates to the model, the effects of the concomitant administration of azithromycin and the use of rifabutin (prophylaxis versus therapy) on CL/F as well as the effect of concomitant administration of ciprofloxacin on V_p/F were found to be of borderline significance and were not retained in the model. Covariates with a significant effect on pharmacokinetic parameters were concomitant administration of azithromycin and tobacco smoking on V_p/F. These fixed effects were retained in the final model. The values of the parameters obtained with the final model are reported in Table 3.

TABLE 2.

Summary of model-building process: addition of covariates to the basic model (step 1) and deletion of covariates from the full model (step 2)

Step and covariate	Change in objective function, answera
	CL/F	Vp/F
Step 1
Age	<3.8, NS	<3.8, NS
Wt	<3.8, NS	<3.8, NS
Ht	<3.8, NS	<3.8, NS
Body surface area	<3.8, NS	<3.8, NS
Sex	<3.8, NS	<3.8, NS
Risk factor	<3.8, NS	<3.8, NS
Prophylaxis vs. therapy	4.87, B	<3.8, NS
Diarrhea	<3.8, NS	<3.8, NS
Cahexia index	<3.8, NS	<3.8, NS
Tobacco smoking	<3.8, NS	7.7, S
Study site	<3.8, NS	<3.8, NS
Clarithromycin coadministration	<3.8, NS	<3.8, NS
Fluconazole coadministration	<3.8, NS	<3.8, NS
Phenobarbital coadministration	<3.8, NS	<3.8, NS
Ciprofloxacin coadministration	<3.8, NS	5.20, B
Azithromycin coadministration	7.21, B	4.14, S
Benzodiazepine coadministration	<3.8, NS	<3.8, NS
CD4+ lymphocyte count	<3.8, NS	<3.8, NS
CLCR	<3.8, NS	<3.8, NS
ALT level	<3.8, NS	<3.8, NS
Bilirubin level	<3.8, NS	<3.8, NS
Total protein concn	<3.8, NS	<3.8, NS
Step 2
Tobacco smoking		7.35, S
Azithromycin coadministration		3.92, S

^aS, significant; B, borderline; NS, not significant. 

TABLE 3.

Population pharmacokinetic parameters of rifabutin administered to HIV-positive patients^a

Value	CL/F (liters/h)	Vc/F (liters)	Q/F (liters/h)	Vp/F (liters)	Ka (h−1)	θ6 (%)	θ7 (%)	CVCL/F (%)
Mean	60.9	231	60.3	1,050	0.201	0.39	0.27	32
95% CI	50.4, 71.4	75, 790	21.3, 99.3	728, 1,372	0.102, 0.895	0.076, 0.704	−0.109, −0.431	22, 49

^aθ₆, percent increase in the value of V_p/F in patients smoking tobacco; θ₇, percent decrease in the value of Vp/F in patients treated concomitantly with azithromycin; CV_CL/F, interpatient variability of CL/F; CI, confidence interval; CV, coefficient of variation. 

DISCUSSION

The disposition of rifabutin was biexponential, as reported previously (24). The mean values (95% confidence intervals) of the pharmacokinetic parameters obtained with the basic model without covariates were 61.0 liters/h (49.8 and 72.2 liters/h) for CL/F, 221 liters (73 and 780 liters) for V_c/F, 61.6 liters/h (24 and 99.2 liters/h) for Q/F, 1,230 liters (898 and 1,562 liters) for V_p/F, and 0.199 h⁻¹ (0.100 and 0.820 h⁻¹) for K_a. After the inclusion of covariates in the model the variability for each parameter decreased from 33 to 32% for CL/F and from 31% to nonsignificant for V_p/F. This finding should not be interpreted as a complete absence of variability in V_p/F. It is more likely that the extent of decrease in V_p/F variability due to the inclusion of significant covariates in the model was sufficient to cause a complete drop in the value of V_p/F because of model overparameterization. In other words, the variability in CL/F, retained in the model, explained all the variability because the data did not support a more elaborate model including variability in V_p/F. Residual variability remained at a high value of 44% (confidence interval, 33 and 53%) after covariate inclusion in the model.

Since the patient population enrolled in our study was represented by patients at a late stage of HIV disease, it may be interesting to compare the pharmacokinetic parameters obtained in our study with the ones obtained for healthy volunteers or patients at an early stage of HIV disease.

The value of CL/F has been reported to be 0.81 liters/h/kg of body weight in a single-dose study with healthy volunteers (17). In our steady-state study, the value of CL/F per body weight obtained by normalization of the mean CL/F value to mean patient body weight is 1.17 liters/h/kg, a value comparable to the one reported in the study with healthy volunteers, considering that rifabutin induces its own metabolism (24).

A study with radiolabelled rifabutin administered intravenously has found a value for rifabutin bioavailability of approximately 20% after the administration of a single dose and of 12% after the administration of multiple doses (24). However, it has been argued that such values may be underestimated (15). In the most conservative of the hypotheses, that is, in case of a bioavailability of 12% at steady state, V_c, V_p, and V_SS in the basic model of our study would have values of 27, 148, and 174 liters, respectively, which correspond to a high level of distribution throughout the body.

Since in most of the previous studies the value of the volume of distribution during the elimination phase (V_z) is reported, an approximate mean value for such a parameter may be estimated by dividing the mean value for CL/F by the mean value computed by the program for the postdistributional elimination rate constant. The values of V_z/F estimated in such a manner are 49 liters/kg for the basic model and 43 liters/kg for the final model for patients who are not smoking tobacco and not taking azithromycin. Such values are comparable to the value of 40 liters/kg reported for healthy volunteers (15).

The terminal half-life was 29.1 h for the basic model and 25.4 h for the final model in patients not smoking tobacco and not taking azithromycin. Such values are comparable to the values obtained for HIV-positive patients at the early stage of disease following the administration of multiple doses (24).

Overall, there is evidence that rifabutin pharmacokinetics in patients at later stages of HIV disease such as the ones enrolled in our study are not substantially different from the ones reported in the literature for healthy volunteers or patients at earlier stages of HIV disease.

In a review article (15) comparing pharmacokinetic data obtained for HIV-positive patients at early stages of HIV disease (24) and healthy volunteers (17, 27), it was shown that the HIV-positive patient population had a higher interpatient variability. Also, more recent pharmacokinetic data obtained for HIV-positive patients at a later stage of disease (16) confirmed that the variability of the rifabutin AUC was up to twice as much the one observed for healthy volunteers (26). HIV-positive patients at later stages of disease represent a more heterogeneous population since they can present with different degrees of gastrointestinal dysfunction as well as general alterations of normal physiologic conditions and usually receive concomitantly a larger number of drugs. Therefore, we performed a study of the effect of several covariates on rifabutin pharmacokinetic parameters in an attempt (i) to identify subpopulations of patients for whom a different dosage regimen could be more appropriate with respect to that used for the general HIV-positive patient population and (ii) to compare the results of more rigorous, controlled studies of the effects of covariates on rifabutin pharmacokinetic parameters with the results obtained in our study, which had the characteristics of an open, prospective, explorative design with a patient population representing those seen in a real clinical setting.

Weight, height, and body surface area did not have an influence on pharmacokinetic parameters.

In a previous study, C_max, AUC, and V_z were found to be higher in women than in men (5, 15). However, such differences disappeared when the data were normalized for body fat (15). It may be possible that we did not find an effect of gender in part because our population was severely wasted in terms of body weight. Therefore, differences due to body fat may have been blunted.

Study site and drug abuse did not have a significant effect on pharmacokinetic parameters. Drug abuse was entered into the analysis only as a risk factor for HIV disease since a urine drug screen was not performed.

There was no trend in the rifabutin pharmacokinetic parameters versus the rifabutin dose. This is confirmed by a previous observation showing no dose dependence of rifabutin pharmacokinetics (15).

As mentioned above, the issue of the effect of HIV disease on rifabutin pharmacokinetics has recently been addressed in a review article (15). Mean parameters of rifabutin exposure were comparable between healthy volunteers and patients at early stages of HIV disease. In two other studies, the parameters of rifabutin exposure did not differ between groups of HIV-infected patients stratified by their CD4⁺ cell counts (4, 15). However, CD4⁺ cell count may not be a precise index of the dysfunction of physiologic processes influencing the absorption and disposition of drugs, such as gastrointestinal malabsorption. In our analysis, diarrhea, cachexia index, use of rifabutin as prophylaxis or therapy, and CD4⁺ cell count were included as parameters of disease progression and organ dysfunction. The relationship between pharmacokinetic parameters, CD4⁺ cell count, and cachexia index was analyzed both linearly and by stratifying the patients in discrete intervals according to parameter values. None of the parameters listed above was significant. A cachexia index of greater than 35 resulted in a significant decrease in V_p/F of 46% (confidence intervals, 34 and 60%), with a drop of the objective function of 4.12. However, only three patients had body weights lower than one-third of their ideal body weight. It may be worthwhile to extend this analysis to a larger number of patients with extremely severe wasting to confirm our finding.

Patients who received rifabutin for therapy were estimated to have a higher CL/F value (8.63%), with a drop in objective function of 4.87. However, the confidence for such an estimate was not sufficient for inclusion of the parameter in the final model. CL/F should be determined by a formal pharmacokinetic study for patients receiving rifabutin for prophylaxis or therapy.

The effect of tobacco smoking on rifabutin pharmacokinetics has never been evaluated. Tobacco smoking was found to increase the value of the rifabutin V_p/F by 39%. This finding may reflect an effect of tobacco smoking on bioavailability. A decrease in bioavailability may be caused by the enzymatic activity induced by tobacco smoking. The fact that tobacco smoking did not have an effect on CL/F may be explained by the existence of a substantial metabolism in the gut wall. In fact, it has never been verified whether the fraction of a rifabutin oral dose which does not reach the systemic circulation is unabsorbed or metabolized by gut microflora, the gut wall, or the liver. Cigarette smoking has been recognized to induce the activity of P-450 isoform CYP1A2 (25). Among several antipyrine metabolites, rifabutin only induces the formation of norantipyrine (21). CYP1A2 has been shown to play an important role in norantipyrine formation (8). It may also be possible that the variability of or the confidence in CL/F was higher than that in V_p/F, with the result of a nonsignificant answer by the program.

What was stated above seems to be true for the effect of the concomitant administration of azithromycin on rifabutin pharmacokinetic parameters. Both CL/F and V_p/F were affected by azithromycin coadministration. However, the confidence for the effect of azithromycin on CL/F was too low to be accepted, even though the drop in the objective function was greater than the one found for V_p/F. The effect of azithromycin on both CL/F and V_p/F may be due to an effect on bioavailability. The value of V_p/F decreased by 27%. This may be explained by metabolic inhibition and an increase in bioavailability. Protein binding displacement is less probable since azithromycin is not highly protein bound. The interaction between rifabutin and azithromycin has never been reported.

Clarithromycin and fluconazole have been shown to cause 82% (30) and 77% (6) increases in the rifabutin AUC, respectively. Our study failed to recognize such interactions. One of the reasons may be that the dose of fluconazole was, on average, lower than the one administered in the previous study. In fact, in our study the fluconazole doses were 400, 300, 200, 100, 150, and 50 mg/day for 1, 1, 4, 8, 3, and 5 patients, respectively. Both studies of the effect of fluconazole and clarithromycin on rifabutin pharmacokinetics were conducted with a randomized, cross-over, controlled design with patients selected by using inclusion criteria more stringent than the ones used in our study, which was an open, uncontrolled study with patients with highly heterogeneous demographic characteristics. This approach may result in the failure to recognize a significant effect. On the other hand, it may suggest that the significance of a pharmacokinetic interaction should undergo postmarketing reevaluation by a pharmacokinetic screening approach in order to confirm that a well-defined interaction is still significant in a heterogeneous population in which several concomitant factors may affect drug absorption and disposition. It must be pointed out that our results are limited by the fact that for this data set only eight patients were not receiving clarithromycin or fluconazole. However, it has been shown that such a number of patients may be sufficient to identify a significant covariate (10).

The concomitant administration of phenobarbital, ciprofloxacin, and benzodiazepines did not have a significant effect on rifabutin pharmacokinetics.

Rifabutin is highly metabolized by the liver. However, in patients with alcoholic liver disease only a slight increase in the rifabutin AUC has been observed (7). None of the patients enrolled in the study had overt cirrhosis. The effect of ALT was performed by subdividing patients into two groups by using twice the upper limit of a normal titer as a cutoff value. ALT levels did not correlate with rifabutin pharmacokinetic parameters.

It has been estimated that diseases affecting the kidney (28) or the gallbladder (22) may result in an alteration of rifabutin elimination. Creatinine clearance (CL_CR) analysis was performed both linearly and by comparing groups with discrete intervals of CL_CR values. CL_CR did not correlate with pharmacokinetic parameters. However, only six patients had values of CL_CR between 40 and 60 ml/min, 18 patients had CL_CR values of between 60 and 90 ml/min, and the rest of the patients had CL_CR values above 90 ml/min. Bilirubin levels did not correlate with pharmacokinetic parameters. Eight patients had bilirubin levels above 1.2 mg/dl. The analysis was performed by dividing the patients into two groups by using 1.2 mg/ml as a cutoff value.

Total protein concentration did not correlate with rifabutin pharmacokinetic parameters. Rifabutin is highly bound to plasma proteins with concentration-independent binding over the range of concentrations achieved clinically (1). Therefore, minor changes in protein binding are unlikely to affect the disposition of rifabutin.

The issue of a pharmacokinetic-pharmacodynamic relationship cannot be addressed for rifabutin since there is a lack of data correlating parameters of drug exposure in plasma (e.g., C_max, minimum concentration of drug in plasma [C_min], and AUC) or parameters of drug exposure normalized to bacteria susceptibility (e.g., C_max/MIC, C_min/MIC, AUC/MIC, and percentage of time between doses that the concentration is above the MIC) to parameters of drug efficacy such as prophylaxis or therapy outcome or risk of breakthrough. The concentrations in lungs and in the gastrointestinal tract, which are considered to be the portals of entry for mycobacteria, are severalfold higher than the concentrations in plasma (15). Also, the concentrations in macrophages and polymorphonuclear leukocytes have been shown to exceed by several times the concentrations in plasma (31). Therefore, in order to assess the pharmacokinetic-pharmacodynamic relationships, the parameters of rifabutin efficacy should be correlated to the concentrations at relevant body sites. On the other hand, if the concentration-versus-time profiles for tissues are similar to that for plasma, it should be possible to correlate the concentrations in plasma to efficacy. This may be true for lung tissue, in which it was shown by a naive pooling approach that the decay of the rifabutin concentration parallels that in plasma (14, 15).

The concentration-versus-time profiles for the typical patients with or without the covariates which were found to influence V_p/F are shown in Fig. 1. The clinical significance of such covariates appears to be negligible since a difference in the concentration in plasma was appreciable only after 24 h following administration. However, since rifabutin is highly distributed into tissues with a ratio of levels in plasma/levels in tissue of greater than 10/1 (15), the effect of the observed 39% increase and 27% decrease in V_p/F due to tobacco smoking and coadministration of azithromycin, respectively, may be of greater magnitude when the rifabutin concentrations at specific body sites are taken into consideration.

FIG. 1.

Simulation of the steady-state plasma concentration-versus-time profiles of rifabutin (300 mg daily) for the typical patient not smoking and not receiving azithromycin (——), smoking (—), receiving azithromycin (---), and smoking and receiving azithromycin (—-—).

In conclusion, among several covariates which were tested for their influence on rifabutin pharmacokinetic parameters, the effects of the concomitant administration of azithromycin and tobacco smoking were statistically significant in altering the rifabutin V_p/F. However, such effects are unlikely to be of clinical relevance unless the effect on the concentrations in tissue is greater than that on the concentrations in plasma.

Our study did not identify the well-characterized effects of clarithromycin and fluconazole on rifabutin pharmacokinetics. This may be due to the highly heterogeneous demographic characteristics of our study population. It may suggest that postmarketing population pharmacokinetic studies should be carried out in order to reevaluate the significance of pharmacokinetic interactions in real clinical settings in which several concomitant factors may influence a pharmacokinetic parameter.

Our study confirms previous observations of the lack of an effect of HIV disease on rifabutin pharmacokinetics. In fact, the pharmacokinetic parameters obtained for our study population, representative of patients at a later stage of HIV disease, were comparable to the ones obtained for healthy volunteers and patients at an early stage of HIV disease. Furthermore, covariates reflecting stage of disease and organ dysfunction such as diarrhea, cachexia index, use of rifabutin for prophylaxis or therapy, and CD4⁺ cell count were not found to correlate with rifabutin pharmacokinetic parameters. Therefore, rifabutin should be administered to patients at an advanced stage of disease without modification of the standard dosage regimen.