Optimisation of antimicrobial therapy using pharmacokinetic and pharmacodynamic parameters

Authors


Corresponding author and reprint requests: M. R. Jacobs, Department of Pathology, University Hospitals of Cleveland, 11100 Euclid Ave, Cleveland, OH 44106, USA
Tel: +1 216 844 3484
Fax: +1 216 844 5601
E-mail: mrj6@po.cwru.edu

Abstract

To understand the relationship between drug dose and efficacy, pharmacokinetic (PK) and pharmacodynamic (PD) characteristics need to be integrated. Patterns of antimicrobial activity fall into one of two major patterns: time-dependent killing and concentration-dependent killing. Time-dependent killing is characteristic of many antibiotic classes, such as β-lactams and macrolides, and seeks to optimise the duration of exposure of a pathogen to an antimicrobial. The major PK/PD parameter correlating with efficacy of time-dependent antimicrobials is the serum concentration present for 40–50% of the dosing interval, and this concentration is the susceptibility limit or breakpoint for the dosing regimen used. The second pattern, concentration-dependent killing, seeks to maximise antimicrobial concentration and is seen with aminoglycosides, quinolones and azalides. The major PK/PD parameter correlating with efficacy of these agents is the 24-h area under the curve to MIC ratio, which should be ≥25 for less severe infections or in immunocompetent hosts, and ≥100 in more severe infections or in immunocompromised hosts. PK/PD breakpoints for concentration-dependent agents can therefore be calculated from the formula AUC ÷ 25. This enables development of PK/PD breakpoints based on the above parameters for time- and concentration-dependent agents for defined dosing regimens. For an antimicrobial to be useful empirically, the MIC90s of the agent against the common pathogens responsible for the disease being treated should be below the PK/PD breakpoint. This is particularly important for oral dosing regimens for treating emerging resistant respiratory tract pathogens, where efficacy against the predominant pathogens, Streptococcus pneumoniae and Haemophilus influenzae, is required.

Introduction

Do drug pharmacokinetics have any relation to patient care? They do, and the relationships between pharmacokinetics and pharmacodynamics, learned during the previous 20–30 years, apply to the design of rational and optimal therapeutic regimens [1]. Pharmacokinetic and pharmacodynamic characteristics both influence dosing regimens of antimicrobials. For two decades, the focus has been on pharmacokinetic characteristics — what the body does to the drug and the overall disposition of the drug in the body. This is reflected most often by the serum concentration profile over time. Of particular interest as well is the penetration of drug into sites of infection.

Medications are administered, however, for their pharmacodynamic characteristics — what the drug does in the body [1–10]. Susceptibility of the pathogen to the drug, determined by measuring the minimum inhibitory concentration (MIC), is a reflection of the potency of a drug. However, to be able to understand the application or the relevance of drug dose to efficacy, we have to integrate pharmacokinetic characteristics with pharmacodynamic characteristics.

The increasing occurrence of antibiotic-resistant pathogens complicates the integration of pharmacokinetics and pharmacodynamics, and has an impact on treatment approaches to respiratory tract infections [11–25]. Clearly, as pathogens become more resistant to antimicrobial agents, the efficacy of standard dosing regimens may be reduced. This increases the need for newer regimens and newer antimicrobials. This review will focus on looking at how integration of pharmacokinetics and pharmacodynamics offers newer ways to evaluate susceptibility data and dosing regimens.

Patterns of antimicrobial activity

The lack of clear pharmacokinetic/pharmacodynamic endpoints has been a challenge in antimicrobial therapy. Although serum concentrations have been measured for years, the clinical significance of these has often been unclear. More recently, many pharmacokinetic/pharmacodynamic studies of antimicrobials showed that the magnitude of the pharmacokinetic/pharmacodynamic parameter required for efficacy is similar in various animal species and in humans. Thus, results from animal studies could predict antimicrobial activity in humans. This would be useful for dosing regimen design in situations in which it is difficult to collect sufficient clinical data, such as in instances of newly emerging resistance, and in clinical syndromes where spontaneous resolution rates are high, such as otitis media and sinusitis.

Despite the large number of classes of antimicrobial agents, patterns of antimicrobial activity fall into one of two major patterns: time-dependent activity and concentration-dependent killing.

Time-dependent killing

Time-dependent killing refers to the time it takes for a pathogen to be killed by exposure to an antimicrobial (Figure 1). The goal of time-dependent killing is to optimise the duration of exposure [1]. With time-dependent killing, postantibiotic effects (persistence of antimicrobial action after the antimicrobial is removed) are minimal. Time-dependent killing is characteristic of β-lactam antibiotics (penicillins, cephalosporins, monobactams, and carbapenems), macrolides, and clindamycin. The major pharmacokinetic/pharmacodynamic parameter that correlates with clinical and bacteriologic efficacy of these drugs is the time for which the serum concentration exceeds the MIC of the pathogen. β-Lactam antibiotics are the most commonly used antimicrobials in clinical practice, especially in treating infections of the upper and lower respiratory tract. In animal models of human infection, different classes of β-lactams require different times above the MIC for net and maximum bactericidal activity. As would be expected, the required time above the MIC varies, depending on the pathogen, infection site, and drug, but is generally 40–50% of the dosing interval. Similar times above the MIC are required to achieve 80% or greater rates of bacteriologic cure in otitis media and in sinusitis caused by Haemophilus influenzae and Streptococcus pneumoniae with β-lactam antibiotics. Using mortality after 4 days of therapy as an endpoint, the relationship between time above the MIC and efficacy of penicillins and cephalosporins in animal models shows similar findings. These correlations serve as the foundation for defining the pharmacodynamic correlates between time above the MIC and bacteriologic and clinical outcome.

Figure 1.

Time above MIC — correlation of serum pharmacokinetics with MIC (susceptibility) of an organism. Drug A is present at a concentration of 2 mg/L for 50% of the dosing interval, while drug B is present at a concentration of 2 mg/L for 30% of the dosing interval.

The next step is to find out if dosing regimens are likely to achieve sufficiently high serum antibiotic concentrations to exceed the MICs of pathogens for 40–50% of the dosing interval. Tables 1 and 2 show conservative adult and pediatric dosing regimens for common parenteral and oral antimicrobials. Using this information, and knowing the antibiotic half-life, we can predict if a dosing regimen will be successful. Tables 1 and 2 also show the pharmacokinetic breakpoints for these agents and dosing regimens, reflecting the serum concentrations present for 40–50% of the dosing interval. As an example, the serum concentration of amoxicillin when 500 mg of this agent is administered orally at 8-h intervals during a 24-h period shows that amoxicillin has a half-life of 30–45 min. With this dosing regimen, amoxicillin achieves a concentration of 2 mg/L for 3.3 h of each 8-h dosing interval (or 9.9 h of a 24-h day), which is 41% of the dosing interval. Therefore, this regimen achieves an amoxicillin concentration of 2 mg/L for over 40% of the dosing interval, and should therefore be active against organisms with MICs ≤2 mg/L. If 875 mg is administered at 12-h intervals, the amoxicillin concentration exceeds 2 mg/L for 4.5 h of each dosing interval (9 h of a 24-h day). Therefore, this regimen achieves an amoxicillin concentration that exceeds 2 mg/L for approximately 40% of the dosing interval. Thus both dosing regimens achieve serum concentrations above 2 mg/L for about 40% of the dosing interval. The pharmacodynamic breakpoint of amoxicillin can therefore be determined to be 2 mg/L for both of these dosing regimens.

Table 1.  Pharmacokinetic/pharmacodynamic breakpoints — serum concentration of oral β-lactams present for >40–50% of dosing interval, and MIC90 values of US isolates of S. pneumoniae
DrugDosing regimenS. pneumoniae
AdultPediatricMIC90 (mg/L)Pharmacokinetic/pharmacodynamic
breakpoint (mg/L)
  • a

    Dosing regimen based on amoxicillin component.

  • t.i.d., three times a day; b.i.d., twice a day; q.d., once a day.

Amoxicillin500 mg t.i.d.40 mg/kg/day t.i.d.22
Amoxicillin875 mg b.i.d.45 mg/kg/day b.i.d.22
Amoxicillin–clavulanatea500 mg t.i.d.40 mg/kg/day t.i.d.22
Amoxicillin–clavulanatea875 mg b.i.d.45 mg/kg/day b.i.d.22
Cefaclor500 mg t.i.d.40 mg/kg/day t.i.d.>640.5
Cefuroxime500 mg b.i.d.30 mg/kg/day b.i.d.81
Cefprozil500 mg b.i.d.30 mg/kg/day b.i.d.161
Loracarbef400 mg b.i.d.30 mg/kg/day b.i.d.>640.5
Cefixime400 mg q.d.8 mg/kg/day q.d.320.5
Table 2.  Pharmacokinetic/pharmacodynamic breakpoints of parenteral β-lactams based on serum concentrations present for >40–50% of dosing regimens shown, and MIC90 values of US isolates of S. pneumoniae
DrugDosing regimenS. pneumoniae
MIC90 (mg/L)
Pharmacokinetic/pharmacodynamic
breakpoint (mg/L)
  • a

    Based on free serum level.

  • q.i.d., four times a day; t.i.d., three times a day; q.d., once a day; b.i.d., twice a day.

Penicillin G2 × 106 U q.i.d.44
Ampicillin1 g q.i.d.42
Cefuroxime0.75 g t.i.d.84
Cefotaxime1 g t.i.d.22
Ceftriaxonea1 g q.d.22
Cefepime1 g b.i.d.44
Ceftazidime1 g t.i.d.328
Meropenem0.5 g t.i.d.21

How can these observations contribute to clinical decision-making? Taking into account the dosing regimens and accepting the defined pharmacodynamic correlate of the concentration present for 40–50% of the dosing interval, pharmacokinetic breakpoints can be determined for defined dosing regimens of various β-lactams (Tables 1 and 2). These pharmacokinetic breakpoints can then be compared to MICs of individual pathogens or to collections of strains, noting if the MIC of a strain or the MIC90 of a group of strains is at or below this breakpoint (i.e. susceptible) or above the breakpoint (i.e. resistant). For some parenteral β-lactams such as penicillin G, the MIC90 for current strains of S. pneumoniae in the USA is below the pharmacokinetic breakpoint, predicting clinical and bacteriologic success. For other agents, such as cefuroxime and ceftazidime, the MIC90 of strains in the USA is above the breakpoint, predicting clinical failure for the most resistant strains. It is important to note that these examples are based on serum concentrations and body sites where concentrations are in equilibrium with serum concentrations, and do not apply to sites where drug penetration is limited, such as the central nervous system. Lower breakpoints are therefore applicable to meningitis.

How do these breakpoints apply to common pathogens? Table 3 shows the percentage of dosing intervals for which serum concentrations of oral β-lactams are above the MIC90s against common respiratory pathogens. The calculations are based on the dosing regimens shown. Amoxicillin–clavulanate is the only oral β-lactam to exceed the MIC90s of all three pathogens for ≥40% of the dosing interval. Although the cephalosporins maintain good activity against penicillin-susceptible S. pneumoniae, they are inactive or inadequately active against penicillin-intermediate and penicillin-resistant strains. Some of these antibiotics are also inadequate against H. influenzae or Moraxella catarrhalis due to inadequate pharmacokinetics of these agents. Amoxicillin, however, has the same percentage of the dosing interval above the MIC90 of beta-lactamase negative isolates of H. influenzae, as does amoxicillin-clavulanate.

Table 3.  Percentages of dosage interval serum levels of oral β-lactams are above MIC90s of pathogens
 S. pneumoniaeH. influenzaeM. catarrhalis
Penicillin
susceptible
Penicillin
intermediate
Penicillin
resistant
Amoxicillin100594600
Amoxicillin–clavulanate10059464170
Cefpodoxime832108237
Cefuroxime axetil753503333
Cefprozil753202141
Cefixime59008848
Cefaclor6000035
Loracarbef5000926

Time above the MIC is also the important parameter for determining efficacy of the macrolides (but not azalides such as azithromycin). Macrolides provide unbound drug concentrations that are greater than the MIC90s against macrolide-susceptible strains of S. pneumoniae for at least 50% of the dosing interval (Table 4). However, unbound serum drug concentrations do not exceed MICs of H. influenzae or macrolide-resistant S. pneumoniae. This might reflect many of the problems that have been noted in the use of macrolides against H. influenzae infections.

Table 4.  Pharmacokinetic/pharmacodynamic breakpoints for macrolides
DrugRegimenBreakpoints (mg/L)
Pharmacokinetic/pharmacodynamicNCCLS
S. pneumoniaeH. influenzae
Erythromycin500 mg q.i.d.0.250.25
Clarithromycin250 mg b.i.d.0.250.258
Azithromycin500 mg o.d.0.120.54

Concentration-dependent killing

The goal of concentration-dependent killing is to maximise concentration and attain the highest possible antimicrobial concentration at the site of infection (Figure 2). With concentration-dependent killing, prolonged postantibiotic activity, persisting even when concentrations are below MICs, is also often present. Concentration-dependent killing is characteristic of aminoglycosides, quinolones, azalides (azithromycin), ketolides, and vancomycin. The major pharmacodynamic parameters that correlate with clinical and bacteriologic efficacy of these drugs are the 24-h area under serum drug concentration curve (AUC) to MIC ratio, or the peak drug concentration to MIC ratio, based on free or unbound serum concentration values. So, again, the MIC remains a primary correlate of pharmacodynamic potency when correlated with the appropriate parameter. The parameters that correlate with clinical and bacteriologic efficacy are 24-h AUC/MIC ratios of ≥25–30 in immunocompetent patients, ≥100–125 in immunocompromised patients, and peak/MIC ratios of ≥10–12 [1,5,17,25]. Pharmacodynamic breakpoints can therefore be determined by the formula AUC ÷ 25 for immunocompetent patients, or AUC ÷ 125 for immunocompromised patients. For azithromycin, with a 24-h AUC of 3 mg.h/L, the pharmacodynamic breakpoint for immunocompetent patients is therefore 0.12 mg/L (3 ÷ 25), which results in this agent being clinically effective against macrolide-susceptible S. pneumoniae (MIC90 of 0.12 mg/L), but not H. influenzae (MIC90 of 1–2 mg/L) or macrolide-resistant S. pneumoniae (MIC90 of >8 mg/L) [6,7]. For the quinolones, AUC/MIC ratios of 25 have also been used to determine breakpoints (Table 5, Figures 4 and 5). MICs of H. influenzae are considerably below these breakpoints for all quinolones. However, MIC90s of S. pneumoniae are above these breakpoints for older agents such as ciprofloxacin and ofloxacin, while MIC90s for newer agents are below these breakpoints [1,5,25,26] (Figure 3).

Figure 2.

AUC/MIC and peak/MIC ratios — correlation of serum pharmacokinetics with MIC (susceptibility) of an organism. The MIC at which the magnitudes of these ratios that are required for clinical success are achieved becomes the pharmacokinetic/pharmacodynamic breakpoint.

Table 5.  AUC values, pharmacokinetic/pharmacodynamic breakpoints and AUC/MIC90 ratios with S. pneumoniae for selected quinolones
Drug
Dose (mg)
24-h AUC
(mg.h/L)
Protein
binding (%)
Pharmacokinetic/pharmacodynamic
breakpoint (mg/L)a
S. pneumoniae
MIC90
(mg/L)
24-h AUC/MIC90
ratiob
  • a

    Values shown are based on total/free serum drug AUC values using AUC values of 25 as breakpoints, with values adjusted to correspond with MIC dilution series values.

  • b

    Values shown are based on total/free serum drug AUC values.

  • c

    After 400-mg loading dose.

Ciprofloxacin500 b.i.d.23301/0.5212/8
Ciprofloxacin750 b.i.d.40301/1220/14
Ofloxacin400 b.i.d.70312/2235/23
Levofloxacin500 q.d.48312/1150/35
Sparfloxacin200 q.d.c20450.5/0.250.2580/44
Gatifloxacin400 q.d.34201/10.569/55
Moxifloxacin400 q.d.48502/10.25192/96
Figure 4.

Correlation of pharmacokinetic/pharmacodynamic parameters in 134 hospitalised patients with respiratory tract, skin or complicated urinary tract infections treated with 500 mg intravenous levofloxacin daily for 5–14 days. Numbers above each bar are numbers of patients. Adapted from Preston et al.[25].

Figure 5.

Fluoroquinolone peak: MIC90 ratios for S. pneumoniae. Levofloxacin, moxifloxacin and gatifloxacin achieve target ratios for free drugs of 3, while moxifloxacin achieves a ratio of 9. Ciprofloxacin does not achieve the minimum target ratio of 3. Dosing regimens are in mg per dose at dosing frequencies shown. Adapted from Turnidge [29].

Figure 3.

Fluoroquinolone AUC: MIC90 ratios for S. pneumoniae at standard dosing regimens. Levofloxacin, moxifloxacin and gatifloxacin achieve target ratios for free drugs of 25, while moxifloxacin achieves a ratio of 100. Ciprofloxacin does not achieve the minimum target ratio of 25. Dosing regimens are in mg per dose at dosing frequencies shown. Adapted from Turnidge [29].

CIP, ciprofloxacin; LEV, levofloxacin; MOX, moxifloxacin; GAT, gatifloxacin; bid, twice a day; d, once a day.

Susceptibility of major respiratory tract infection (RTI) pathogens to oral agents

Susceptibility of the major RTI pathogens in the USA to oral agents at pharmacokinetic/pharmacodynamic breakpoints is shown in Table 6. Susceptibility variation in S. pneumoniae in different geographic regions is considerable, with many areas having isolates that are more susceptible to β-lactams, macrolides, doxycycline and trimethoprim–sulfamethoxazole [21,23]. However, fluoroquinolone resistance is emerging and has the potential to become significant [23]. Significant susceptibility variation in H. influenzae by geographic region is seen only for amoxicillin and trimethoprim–sulfamethoxazole, while there is virtually no geographic variation in susceptibility of M. catarrhalis.

Table 6.  Susceptibility of major RTI pathogens in the USA to oral agents at pharmacokinetic/pharmacodynamic breakpoints [26]
AgentPercentage of strains susceptible
S. pneumoniaeH. influenzaeM. catarrhalis
  • a

    Based on NCCLS breakpoints. NA, not applicable.

Amoxicillin–clavulanate9097100
Amoxicillin906114
Cefaclor2725
Cefixime5799100
Cefpodoxime639964
Cefprozil64186
Cefuroxime647937
Macrolides670100
Clindamycina89NANA
Doxycycline762096
Levofloxacin99.810099
Trimethoprim–sulfamethoxazolea57759

Correlation of pharmacokinetic/pharmacodynamic parameters with bacteriologic and clinical outcome in humans

Although there are a limited number of studies in humans that have determined bacteriologic outcome, many studies in which this has been done show a striking correlation between bacteriologic outcome and pharmacokinetic/pharmacodynamic parameters. Examples of acute otitis media [6,7] and acute bacterial exacerbations of chronic bronchitis studies [13,27,28] where such correlations have been shown are listed in Tables 7 and 8.

Table 7.  Bacteriologic outcome studies in acute otitis media [6,7]
TreatmentDosing regimen
(mg/kg/day)
Bacteriologic failure rate (%)
S. pneumoniaeH. influenzae
Penicillin
susceptible
Penicillin
non-susceptible
  • NA, not applicable; ND, not done.

  • a

    For

  • β

    -lactamase-negative/positive strains.

PlaceboNA84ND52
Cefaclor40 t.i.d.106240
Cefuroxime axetil30 b.i.d.92115
Amoxicillin40–50 t.i.d. or b.i.d.102023/63a
Amoxicillin–clavulanate45 b.i.d.10023
Ceftriaxone 1 day50 q.d. IM0530
Ceftriaxone 3 day50 q.d. IM090
Azithromycin
susceptible
Azithromycin
resistant
 
Azithromycin10 q.d. × 3
or 10 QD day 1,
then 5 q.d. × 4
59257
Trimethoprim–sulfamethoxazole
susceptible
Trimethoprim–sulfamethoxazole
resistant
 
Trimethoprim–sulfamethoxazole 07918
Table 8.  Bacteriologic outcome studies in acute exacerbation of chronic bronchitis (AECB) [13,27,28]
TreatmentBacteriologic failure rate (%)
S. pneumoniaeH. influenzaeAll organisms
  1. P-values <0.01 for all organisms in all three studies.

{Azithromycin
Amoxicillin–clavulanate
30
0
50
0
33
1
{Ciprofloxacin
Cefuroxime axetil
10
0
0
14
4
18
{Ciprofloxacin
Clarithromycin
17
7
0
35
8
23

In a unique study of 134 hospitalised patients with bacteriologically proven respiratory tract, skin or complicated urinary tract infections treated with 500 mg of intravenous levofloxacin daily for 5–14 days, Preston et al. correlated clinical outcome with steady-state pharmacokinetics on day 3 of treatment with levofloxacin MICs of the pathogen from the corresponding patient [25]. This study showed that patients with AUC/MIC ratios >100 or peak/MIC ratios >12 had a 1% rate of clinical failure, those with AUC/MIC ratios of 25–100 or peak/MIC ratios of 3–12 had a 12% rate of clinical failure, and those with AUC/MIC ratios <25 or peak/MIC ratios <3 had a 43% rate of clinical failure (Figure 4).

Conclusions

Pharmacokinetic and pharmacodynamic characteristics are major determinants of efficacy of antimicrobial therapy and serve as a rational basis for determination of clinically relevant susceptibility breakpoints. Previously, in vitro susceptibility was not well integrated with drug pharmacokinetics. It is now clear that pharmacokinetic and pharmacodynamic characteristics have to be considered to enable the use of optimal dosing regimens for antimicrobials and in determining clinically relevant susceptibility breakpoints. The ability of an antimicrobial dosing regimen to meet the pharmacokinetic/pharmacodynamic parameter required for efficacy against emerging resistant bacteria needs to be considered in designing effective antimicrobial regimens and in selecting suitable empirical therapy. Basing susceptibility breakpoints on pharmacokinetic/pharmacodynamic parameters requires changing many of the breakpoints in current use, and this process was begun in 2000 in the USA with the revision of many breakpoints for S. pneumoniae and oral β-lactams. Additionally, recent guidelines for respiratory tract infections have been largely based on application of pharmacokinetic/pharmacodynamic parameters.

Ancillary