Use of Pharmacokinetic-Pharmacodynamic Target Attainment Analyses To Support Phase 2 and 3 Dosing Strategies for Doripenem

Abstract

A doripenem population pharmacokinetic model and Monte Carlo simulations were utilized for dose regimen decision support for future clinical development. Simulation results predict that 500 mg of doripenem administered over 1 h every 8 h would be effective against bacterial strains with MICs less than 2 μg/ml and that less susceptible strains could be treated with prolonged infusions.

There has been a growing appreciation of the value of incorporating knowledge from nonclinical models of infection into the early stages of clinical drug development. As described by Drusano et al., the integration of pharmacokinetic-pharmacodynamic (PK-PD) targets derived from nonclinical exposure-response data with phase 1 pharmacokinetic data can be used to optimize antimicrobial dosing regimens for phase 2 and 3 studies (2). Recently, the Food and Drug Administration has implemented end-of-phase-2a meetings with sponsors to review the pharmacokinetics and PK-PD of investigational agents and discuss data supporting dose selection in order to increase the likelihood of regulatory success (4).

Doripenem, a parenteral carbapenem currently in clinical development, demonstrates a broad spectrum of in vitro and in vivo microbiological activity against most clinically relevant gram-negative and -positive pathogens, including commonly occurring extended-spectrum beta-lactamase and AmpC producing strains (3; H. Huynh, P. R. Rhomberg, and R. N. Jones, Abstr. 43rd Intersci. Conf. Antimicrob. Agents Chemother., abstr. F-528, 2003). Like previous studies with beta-lactams in neutropenic mouse-thigh infection models (1), a study by Craig and Andes found the proportion of the dosing interval for which drug concentrations exceed the MIC of the targeted microorganism (T>MIC, in percent) for doripenem to be the PK-PD measure that correlated best with change in bacterial count (log₁₀ CFU/thigh) for Streptococcus pneumoniae, Staphylococcus aureus, and Klebsiella pneumoniae (D. R. Andes and W. A. Craig, Abstr. 43rd Intersci. Conf. Antimicrob. Agents Chemother., abstr. A-308, 2003). The mean (SD) T>MIC associated with 1- and 2-log reductions were 21.1% (8.9%) and 27.3% (11.9%) for S. pneumoniae, 32.3% (6.7%) and 35.4% (5.0%) for S. aureus, and 36.1% (7.4%) and 43.3% (7.1%) for gram-negative bacilli, respectively.

Using the above-described PK-PD targets and phase 1 pharmacokinetic data, we carried out the analyses described herein in order to (i) develop a population pharmacokinetic model to describe the disposition of doripenem; (ii) assess the performance of various doripenem dosing regimens in attaining PK-PD targets over a range of MICs using a Monte Carlo simulation; and (iii) identify potential dosing regimens for phase 2 and 3 studies.

Phase 1 data were obtained from a double-blind dose escalation study of intravenous doripenem in 24 healthy subjects between 18 and 65 years of age who received one of four regimens for 7 days, 500 or 1,000 mg given every 12 or 8 h (D. Thye, T. Kilfoil, A. Leighton, and M. Wikler, Abstr. Intersci. Conf. Antimicrob. Agents Chemother., abstr. A-21, 2003). Blood samples were collected at prespecified time points on days 1 (n = 12), 4 to 6 (n = 6), 7 (n = 13), 8 (n = 1), and 11 (n = 1).

Population pharmacokinetic analyses were performed using NONMEM, version 5.1.1. As demonstrated by the individual predicted (i.e., fitted based on Bayesian estimation using the post hoc option in NONMEM) versus observed concentrations in Fig. 1, data were well fitted by a two-compartment model with linear elimination and an additive-plus-proportional residual error model. The degree of heteroscedasticity seen along the line of identity was well counteracted by the weighted model as evidenced by looking at the weighted residuals versus predicted concentrations (data not shown). Pharmacokinetic parameter estimates and measures of interindividual variability for the final model are shown in Table 1. The estimated terminal half-life of doripenem was approximately 0.95 h. Interindividual variability and the covariance between parameters were estimated for clearance (CL), volume of the central compartment (V_c), and volume of the peripheral compartment (V_p). The interindividual variabilities (percent coefficient of variation) of CL, V_c, and V_p were estimated to be 13, 14, and 10, respectively. The coefficients of determination (r²) of the interindividual variabilities of CL and V_c, CL and V_p, and V_c and V_p were estimated to be 0.54, 0.51, and 0.37, respectively.

FIG. 1.

Individual predicted versus observed concentrations based on the final population pharmacokinetic model developed using phase 1 data from 24 subjects. The line of best fit was described by the following equation: individual predicted concentration = (0.973 × observed concentration) + 0.497; r² = 0.972, P < 0.0001.

TABLE 1.

Final population pharmacokinetic parameter estimates for the model developed using phase 1 data from 24 subjects

Parameter	Estimate		Magnitude of interindividual variabilitya
	Mean	SEM (%)	CV (%)	SEM (%)
CL (liter/h)	14.5	2.6	13.2	32.5
Vc (liters)	9.43	6.4	14.4	23.6
Q (liters/h)	9.69	20.3	NE	NE
Vp (liters)	5.88	6.7	10.4	69.4
Residual variability (CV [%])b	20.5-10.9	10.3, 39.3

^aNE, not able to estimate.

^bEstimates of residual variability magnitude for predicted doripenem concentrations ranging from 1 to 20 μg/ml and higher. CV, coefficient of variation.

Using the above-described mean pharmacokinetic parameter estimates and covariance matrix, a 5,000-patient Monte Carlo simulation using SAS, version 8.2, was carried out to evaluate the probability of PK-PD target attainment for various doripenem dosing regimens (doses of 250, 500, 750, 1,000, 2,000, and 3,000 mg; intervals of 6, 8, 12, and 24 h and infusion durations of 1 to 6 and 24 h). PK-PD target attainment probabilities for a range of free-drug T>MICs (30 to 45%) (based on 8.5% protein binding [data on file, Peninsula Pharmaceuticals Inc.]) were evaluated for each dosing regimen across a range of doubling MICs from 0.25 to 16 μg/ml.

Figure 2 demonstrates the simulated plasma concentration-time profiles for an average subject in the population (based on central estimates for CL, V_c, V_p, and intercompartmental clearance [Q]) for different durations of infusion for 500-mg and 1,000-mg doses (Fig. 2a and b, respectively) and for 750-, 1,000-, 2,000-, and 3,000-mg doses administered every 24 h as a continuous infusion (Fig. 2c). Simulations demonstrated high probabilities of PK-PD target attainment for patients with infecting organisms manifesting MICs of 0.5 μg/ml or less across the majority of regimens evaluated, including those as low as 250 mg administered every 8 h. A subset of these results comparing PK-PD target attainment probabilities for selected regimens at MICs of 2, 4, and 8 μg/ml is shown in Table 2.

FIG. 2.

Simulated concentration-time profiles for 500-mg (a) and 1,000-mg (b) doses of doripenem with various durations of infusion and various doses of doripenem administered over 24 h as a continuous infusion (c). Profiles are for an average subject in the population.

TABLE 2.

Comparison of pharmacokinetic-pharmacodynamic target attainment probabilities by dosing regimen, duration of infusion, and MIC

Dosing regimen	Duration of infusion (hr)	MIC (μg/ml)	Probability of patients achieving target T>MIC
			30%	35%	40%	45%
500 mg q8h	1/2/3	1	1.00/1.00/1.00	1.00/1.00/1.00	1.00/1.00/1.00	0.99/1.00/1.00
500 mg q8h	1/2/3	2	1.00/1.00/1.00	0.99/1.00/1.00	0.77/1.00/1.00	0.25/0.90/1.00
500 mg q8h	3/4/5	4	1.00/1.00/1.00	1.00/1.00/0.99	0.84/0.99/0.99	0.26/0.90/0.95
1,000 mg q12h	4/5/6	4	1.00/1.00/1.00	1.00/1.00/1.00	0.92/1.00/1.00	0.23/0.96/1.00
1,000 mg q8h	1/2/3	4	1.00/1.00/1.00	0.99/1.00/1.00	0.77/1.00/1.00	0.25/0.90/1.00
1,000/2,000/3,000 mg q24h	24	4	0/0.98/1.00	0/0.98/1.00	0/0.98/1.00	0/0.98/1.00
1,000 mg q8h	3/4/5	8	1.00/1.00/1.00	1.00/1.00/0.99	0.84/0.99/0.99	0.26/0.90/0.95
1,000/2,000/3,000 mg q24h	24	8	0/0/0.46	0/0/0.46	0/0/0.46	0/0/0.46

Sensitivity analyses, carried out to assess the magnitude of deviations in PK-PD target attainment for simulations based on the upper and lower confidence bounds versus those based on the central estimates for the pharmacokinetic parameters, failed to show a substantial impact of uncertainty in the estimation of pharmacokinetic parameter typical values or interindividual variabilities. The greatest deviations in PK-PD target attainment probabilities were noted for those closer to 0.50 based on central estimates or when Q was altered to the upper or lower bounds of the confidence interval.

As demonstrated by in vitro susceptibility data collected in the United States, the European Union, and Japan, MICs at which 90% of organisms are inhibited for doripenem are generally ≤1 μg/ml for most organisms including methicillin-susceptible staphylococci, penicillin-susceptible and -resistant S. pneumoniae, Haemophilus influenzae (ampicillin susceptible or β-lactamase positive), Moraxella catarrhalis, members of the Enterobacteriaceae, Pseudomonas aeruginosa, ceftazidime-susceptible Acinetobacter spp., Bacteroides spp., Prevotella spp., Clostridium spp., and other gram-positive anaerobes. For various strains of resistant P. aeruginosa, MICs from 2 to 8 μg/ml have been noted (5). A recent evaluation of 415 resistant isolates (extended-spectrum β-lactamase- and AmpC-producing strains) collected worldwide demonstrated β-lactamase stability, with MICs for doripenem generally being ≤1 μg/ml (Huynh et al., 43rd ICAAC). Thus, evaluation of PK-PD target attainment for various doripenem regimens in this MIC range is of interest.

Using these data and a PK-PD T>MIC target of 35%, high probabilities of PK-PD target attainment for all doses (250 to 1,000 mg) administered every 8 h for MICs of 1 μg/ml, regardless of length of infusion, were observed. For MICs of 2 μg/ml, doses of 500 mg or higher administered every 8 h as a 1-h infusion were associated with the same degree of target attainment. For organisms with higher MICs, high probabilities of PK-PD target attainment were achieved when the same total daily dose of doripenem was administered as a prolonged infusion. Thus, by varying the length of infusion, the same dosing regimen should allow for the effective treatment of pathogens with various MICs with little or no increase in drug exposure.

Although uncertainty in the estimation of typical values of pharmacokinetic parameter estimates or interindividual variability had little impact on PK-PD target attainment results, two notable and related limitations of these analyses were evident. First, the number of subjects in the pharmacokinetic model development data set was relatively small and demographically homogenous. Second, the population pharmacokinetic model was derived from healthy subjects with a restricted range of renal function (creatinine clearance: mean, 107 ml/min; range, 68 to 136 ml/min). Consequently, the interindividual variability described is likely an underestimate of the true variability observed in patient populations. In patient populations with enhanced renal function, the probability of attaining target exposures would be expected to be lower than those predicted. Similarly, in patient populations with compromised renal function, the probability of attaining target exposures would be expected to be greater than those predicted.

The integration of phase 1 pharmacokinetic data with nonclinical data from infection models, and the use of the Monte Carlo simulation, allowed us to construct doripenem dosing strategies for further clinical evaluation in phase 2 and 3 that both maximize the probability of therapeutic benefit while minimizing drug exposure. When sufficient phase 2 pharmacokinetic data are available, it will be important to use these data to refine the current population pharmacokinetic model and subsequently make more accurate inferences about PK-PD target attainment in patient populations of interest in order to further support dose selection for future clinical trials.