Introduction

Top Abstract Introduction Materials & Methods Results Discussion References

Infections caused by Streptococcus pneumoniae resistant to -lactam antimicrobial agents are an increasingly frequent problem in clinical practice (Boswell et al., 1994; Austrian, 1994; Finch, 1995). The breakpoint for penicillin resistance in pneumococcal pneumonia, as well as the ability of MICs and MBCs to predict in vivo activity, remains to be established. Therefore, the optimal antibiotic therapy for P^R S. pneumoniae pneumonia is not clear (Boswell et al., 1994; Austrian, 1994; Finch, 1995). Numerous factors intrinsic to the interaction of antimicrobial agents with microorganisms, such as pharmacokinetics, pharmacodynamics and susceptibility to antimicrobial resistance, determine the therapeutic efficacy of the antimicrobial agents. Few animal studies have examined the relationship between in vitro susceptibility testing and in vivo efficacy for S. pneumoniae (Frimodt-Moller et al., 1986, 1987; Frimodt-Moller and Thomsen, 1987). Those studies examined the relationship between MIC and in vivo efficacy by testing a single bacterial strain against a large number of antimicrobial agents with various MICs. The results of those studies do not account for the differences between the various strains of S. pneumoniae with regard to penicillin resistance. Based on these considerations, we asked whether the MIC results themselves would be predictive of clinical outcome or therapeutic efficacy. In this study, we examined the in vivo activity of amoxicillin against 12 strains of S. pneumoniae with penicillin MICs of <0.01 to 16 µg/ml in a neutropenic murine pneumonia model. A wide range of doses was used to describe a dose-response curve and to determine quantitative parameters of in vivo activity. The relationships between MICs, MBCs and KRs determined from in vitro time-kill curves and parameters of in vivo activity for amoxicillin are described.

	Materials and Methods

Top Abstract Introduction Materials & Methods Results Discussion References

Challenge Organisms

The following 12 S. pneumoniae clinical strains isolated from sinus, ear or bronchopulmonary samples, pleural effusions or blood cultures and provided by the National Reference Center for Pneumococci (P. Geslin, Créteil, France) were used for these experiments: two strains were P^S (P52181 and P30923), four were P^I (P31192, P30189, P40225 and P54B) and six were P^R (P54988, P12698, P15986, P40422, P41375 and P53681).

Antibiotics

Amoxicillin sodium salt (Beecham Laboratories, Paris, France) was used in this study. Amoxicillin was reconstituted according to the instructions on the package insert and diluted in sterile water to the desired concentrations.

In Vitro Studies

Antibiotic susceptibility tests. MICs and MBCs were determined for each strain in Mueller-Hinton infusion broth, supplemented with 5% lysed horse blood (Bio-Mérieux, Lyon, France), by means of the tube dilution method (National Committee for Clinical Laboratory Standards, 1993). Each tube contained 2-fold dilutions of antibiotic and a final bacterial concentration of 10⁶ CFU/ml. After aerobic incubation for 18 hr at 37°C, the MIC was defined as the lowest concentration of antibiotic at which no turbidity was visible to the naked eye. The MBC was determined by plating 0.01-ml aliquots from tubes with no visible growth onto Columbia agar supplemented with 5% sheep blood (Bio-Mérieux). The plates were incubated overnight at 37°C, and the MBC was defined as the lowest concentration of antibiotic killing 99.9% of the original inoculum.

Killing curves. Time-kill curves were plotted for an inoculum of 10⁷ S. pneumoniae, in brain-heart infusion broth, and amoxicillin concentrations equivalent to 4 times their MICs. Samples were removed after 1, 3 and 6 hr of incubation and diluted serially. Viable counts of bacteria were determined by plating appropriately diluted cultures on Columbia agar supplemented with 5% sterile sheep blood. Each strain was tested in triplicate, and the mean value of bacterial reduction was calculated (log CFU/ml in vitro). The KR for each strain was defined as the log CFU/ml between the initial inoculum and the viable bacteria after 1, 3 or 6 hr of incubation. The 1-hr, 3-hr and 6-hr KRs were determined and analyzed.

Experimental Pneumococcal Pneumonia in Mice

We failed to induce pneumonia with any P^I or P^R strain in immunocompetent Swiss mice. All of these strains belonged to serotypes 6, 9, 14, 19 and 23, which are naturally avirulent for mice independently of their isolation sites in humans (Bédos et al., 1991; Briles et al., 1992). Thus, pneumonia was induced in leukopenic Swiss mice. Female Swiss mice (body weight, 20-24 g) were obtained from Iffa-Credo Laboratories (Lyon, France). We induced sustained leukopenia in Swiss mice by daily i.p. injections (150 mg/kg of body weight) of cyclophosphamide (Endoxan; Sarget Laboratories, Mérignac, France) starting 3 days before infection. Counts of circulating leukocytes per cubic millimeter of blood were reduced from about 7000/mm³ to 800/mm³ on the day of infection. Animals were infected by intratracheal instillation via the mouth, as described in detail elsewhere (Azoulay-Dupuis et al., 1991). Briefly, animals were anesthetized i.p. with 0.2 ml of 0.65% sodium pentobarbital and were suspended by the upper incisors. The trachea was cannulated via the mouth with a blunt needle, and 50 µl of bacterial suspension (10⁷ CFU of P^S, P^I or P^R S. pneumoniae per mouse) was instilled. Leukopenic mice developed acute pneumonia and died within 2 to 3 days. All animals quickly became bacteremic. Bacterial counts exceeded 10⁸ CFU/lungs of infected mice at the time of their death (10⁶ CFU/ml of blood).

Survival Studies

Therapy was initiated 1 hr after bacterial challenge. A total of six, 12-hourly, s.c. amoxicillin injections were administered over 72 hr. Controls animals received saline. Fifteen animals per treatment group were used, and in each experiment the animals were infected simultaneously. Experiments were repeated at least twice. The observation period was 14 days. Death rates were recorded daily, and cumulative survival rates were compared.

With P^S strain P52181, we previously showed that amoxicillin treatment at 5 mg/kg (treatment schedule consisting of s.c. injections of 5 mg/kg, at 12-hr intervals, over 3 days) was associated with an 87% survival rate [vs. untreated control group (0%)] (Moine et al., 1994). Moreover, the survival rate was not significantly improved with larger doses of amoxicillin. The experiments were repeated to confirm these results. Then, we determined for each strain the MTD of amoxicillin required to achieve the same efficacy, i.e., 75 to 85% survival.

Bactericidal Activity In Vivo

Amoxicillin was assessed for its ability to eradicate bacteria from the lungs. Drug administration was started 1 hr after infection and consisted of a single dose of amoxicillin. Various doses (5-3000 mg/kg of body weight) of amoxicillin were administered in 0.5 ml of sterile water, according to the infective strain. Four to six serial 2-fold increases in dose were used for each drug-organism combination, to describe the dose-response relationship. Controls received the same volume of isotonic saline. The total CFU recovered from whole-lung homogenates was determined 1, 3, 6 and 9 hr after one injection. Mice were killed by CO₂ asphyxiation and exsanguinated by cardiac puncture. The lungs were removed and homogenized in 1 ml of saline. Serial 10-fold dilutions of the homogenates were plated onto Columbia agar (0.1 ml/9-cm-diameter plate). The lower limit of detection was 2 log CFU/lung, which corresponded to the weakest dilution of tissue homogenates (10¹) that avoided significant drug carryover with control inocula. In controls, mean bacterial counts increased from 7.3 to 7.6 log CFU/lung during the time of observation (1-9 hr after infection). For any given dose of amoxicillin, for each strain of S. pneumoniae, efficacy or observed maximal effect (log CFU_max per lung) was defined by the maximal reduction obtained (in log CFU per lung) for each treated mouse at any time point, compared with the mean log CFU per lung of control mice just before therapy (0 hr). Representative dose-effect relationships for amoxicillin against the 12 S. pneumoniae strains were then extrapolated.

Statistical Analysis

A dose-effect model (E_max model) was used to characterize in vivo antimicrobial efficacy (Gibaldi and Perrier, 1982). The pooled-data pharmacodynamic technique was used, i.e., all maximal reductions obtained in log CFU per lung for each treated mouse, compared with the mean log CFU per lung of control mice just before therapy (log CFU_max per lung), were computed and further used without any averaging or outlier deletion. The log CFU_max per lung vs. dose data were fitted to the simple E_max model described by the following equation (Gibaldi and Perrier, 1982):

(1)

where E is the observed effect (maximal reduction in log CFU per lung, compared with control mice just before therapy), D is the administered dose, E_max is a measure of relative efficacy and P₅₀ is a measure of potency. Data were fitted to the model using MK MODEL (Holford, 1994). Because the asymptotic error calculated by any nonlinear least-squares minimization technique or logarithm-likelihood maximization technique greatly underestimates true S.E., no attempt has been made to compare the different doses. Results are reported as estimated parameters and their coefficient of variation, which is an underestimation of the magnitude of the error relative to the nominal value. To allow more meaningful comparison of efficacy, we estimated the ampicillin P(1log) by deriving the following equation from eq. 1:

(2)

Correlation analysis between in vitro susceptibility data and in vivo parameters was done using the Spearman rank test.

	Results

Top Abstract Introduction Materials & Methods Results Discussion References

MIC and MBC determinations. The properties of the 12 strains are shown in table 1. The amoxicillin MBCs were equal to or twice the amoxicillin MICs in all 12 strains.

Killing curves. The rate at which each S. pneumoniae strain was killed by amoxicillin at concentrations equal to 4 × MIC is shown in figure 1.

View larger version (26K): [in this window] [in a new window]   Fig. 1.   In vitro killing of 12 strains of S. pneumoniae with amoxicillin MICs of 0.01 to 8 µg/ml. Time-kill curves were plotted for an inoculum of 107 S. pneumoniae and amoxicillin concentrations equivalent to 4 times their MICs.

Bactericidal activity in lung, and correlation between in vitro susceptibility data and parameters of in vivo efficacy. The E_max, P₅₀ and P(1log) for each strain are shown in table 2. Significant differences in E_max were noted among the 12 strains. Moreover, as expected, important differences in P₅₀ and the estimated P(1log) were also noted.

View larger version (11K): [in this window] [in a new window]   Fig. 2.   Relationship between P50 and MIC values for 12 S. pneumoniae strains. P50 was correlated with MIC (Spearman rank correlation: r = 0.98, P < .0001). Data are plotted on a logarithmic-logarithmic scale to allow better representation.

Therapeutic efficacy in experimental pneumonia. The survival data obtained have been summarized in table 3. With P^I strains, 2- to 5-fold increased doses of amoxicillin (10-50 mg/kg) gave protection similar to that against the P^S strain, whereas 10- to 60-fold increased doses were required with P^R strains. For each studied strain, the survival rate obtained was not significantly improved with larger doses of amoxicillin. For strains such as P41375 and P50681 (amoxicillin MICs of 4 and 8 µg/ml, respectively), considerably higher doses were required to achieve the same effect and could not be obtained.

View larger version (10K): [in this window] [in a new window]   Fig. 3.   Correlation between the MTDs of amoxicillin in survival studies (treatment schedule consisting of six s.c. injections at 12-hr intervals over 72 hr) and MICs for 10 S. pneumoniae strains with amoxicillin MICs ranging from 0.01 to 2 µg/ml. For strains such as P41375 and P50681 (amoxicillin MICs of 4 and 8 µg/ml, respectively), considerably higher doses were required and could not be obtained. MTD was correlated with MIC (Spearman rank correlation: r = 0.98, P < .0001). Data are plotted on a logarithmic-logarithmic scale to allow better representation.

View larger version (11K): [in this window] [in a new window]   Fig. 4.   Correlation between P50 values and the MTDs achieving the same efficacy (75-85% survival rate) in survival studies (treatment schedule consisting of six s.c. injections at 12-hr intervals over 72 hr) for 10 S. pneumoniae strains with amoxicillin MICs ranging from 0.01 to 2 µg/ml. MTD was correlated with P50 (Spearman rank correlation: r = 0.98, P < .0001). Data are plotted on a logarithmic-logarithmic scale to allow better representation.

	Discussion

Top Abstract Introduction Materials & Methods Results Discussion References

In vitro susceptibility tests are designed, for the most part, to determine the activity of antimicrobial agents in a static test situation that is convenient for the laboratory. However, will the test results be predictive of clinical outcome or therapeutic efficacy? Numerous factors intrinsic to the interaction of antimicrobial agents with microorganisms, i.e., pharmacokinetics, pharmacodynamics and susceptibility to microbial resistance mechanisms, determine the therapeutic efficacy of these agents.

Experimental animal models, which allow standardization of infection and treatment regimens (Barza, 1978; Bergeron, 1978), have been used to analyze the relationship between in vitro susceptibility tests and parameters of in vivo antimicrobial efficacy. One-third of such experiments found discrepancies between in vitro and in vivo data (Zak and Sande, 1982). Few animal studies have examined the relationship between in vitro susceptibility testing and in vivo efficacy for S. pneumoniae (Frimodt-Moller et al., 1986, 1987; Frimodt-Moller and Thomsen, 1987). Nevertheless, Frimodt-Moller et al. (1986), using a murine peritonitis model, demonstrated a significant correlation between MICs and the 50% effective dose for 14 cephalosporins against a single strain of serotype 3 S. pneumoniae. In contrast, in the same model, the 50% effective dose was lower for ampicillin than for piperacillin (12.5 mg/mouse vs. 65 mg/mouse), in spite of a higher MIC against the pneumococcus for ampicillin (ampicillin MIC = 0.1 µg/ml vs. piperacillin MIC = 0.05 µg/ml) (Frimodt-Moller and Thomsen, 1987). One of the limitations of such studies is the use of a single bacterial strain challenged with several antibiotics with different MICs and potentially different pharmacological properties. Furthermore, the results of these studies do not account for differences between S. pneumoniae with regard to penicillin resistance.

The methodology of our study differs markedly from those published previously. We studied 12 strains of S. pneumoniae with amoxicillin MICs of 0.01 to 8 µg/ml. For each strain, we used 4- to 6-fold increases in amoxicillin dose to describe the dose-response curve and to determine quantitative parameters of in vivo amoxicillin activity [E_max, P₅₀ and P(1log)]. An E_max model has been previously used (Leggett et al., 1989; Fantin et al., 1991) to investigate, in murine infection models, the influence of dosing regimens on the efficacy of antibiotics possessing different pharmacodynamic characteristics. We used an E_max model to investigate the impact of penicillin resistance on the in vivo relative efficacy of amoxicillin against different strains of S. pneumoniae with penicillin MICs of <0.01 to 16 µg/ml. We have demonstrated a highly significant correlation between P₅₀, P(1log) and MIC or MBC for amoxicillin against strains of S. pneumoniae with a wide range of amoxicillin MICs (0.01-8 µg/ml). Our findings may have implications for the clinical use of amoxicillin. This study demonstrated that the standard in vitro MIC was an excellent predictor of the relative in vivo potency of amoxicillin against pneumococcal species, including highly P^R strains. Moreover, based on this animal model, the estimated MIC breakpoints for amoxicillin against S. pneumoniae would be 2 µg/ml for P^I and 4 µg/ml for P^R. P₅₀ and P(1log) net changes were observed with organisms with MICs of 2 µg/ml. Pallares et al. (1987) suggested that patients with bacteremic pneumococcal pneumonia due to P^R strains for which MICs were 4 µg/ml would not respond to therapy with a penicillin agent. Moreover, based on the simulation of human pharmacokinetics in a neutropenic murine thigh model, Andes et al. (1995) recently suggested that the estimated breakpoints for amoxicillin against S. pneumoniae would be 4 µg/ml for P^R. Our results are in agreement with these suggestions. Furthermore, the amoxicillin MIC breakpoints recently published by the National Committee for Clinical Laboratory Standards (National Committee for Clinical Laboratory Standards, 1995) differ by only one 2-fold lower dilution from those we proposed. On the other hand, the National Committee for Clinical Laboratory Standards breakpoints could be viewed as safer, in light of the much greater amoxicillin dose required for at least one of the strains with an MIC of 2 µg/ml (table 3).

We also found a highly significant correlation between P₅₀, P(1log) and the MTD of amoxicillin required against strains of S. pneumoniae in survival studies (treatment schedule consisting of six s.c. injections at 12-hr intervals over 3 days). These results fully confirmed the reliability of the model. However, the data obtained by this method of analysis were difficult to compare with dose-survival studies. The schedule of drug administration may have an important influence on efficacy, particularly in this neutropenic model. Schmidt and colleagues (Schmidt et al., 1949; Schmidt and Walley, 1951), using rat peritonitis and pneumonia models, demonstrated that the effectiveness of sodium penicillin G was related to the frequency with which the drug was administered. Bakker-Woundenberg et al. (1984), using a rat pneumonia model, demonstrated that in rats with intact host defense mechanisms a given amount of penicillin was equally effective when administered continuously or at relatively long intervals (12 hr), whereas in rats with impaired host defenses maintenance of bactericidal levels of penicillin was particularly important for therapeutic efficacy. Moreover, we do not know how a longer course of therapy, i.e., 24 hr, and the dosing interval, i.e., 6, 8 and 12 hr, would alter the P₅₀ and P(1log) values obtained in our bactericidal study. The dosing interval has been shown to be a significant determinant of the dose required to produce the P₅₀ for -lactams in a Klebsiella pneumoniae mouse pneumonia model (Leggett et al., 1989). Additionally, we do not know which pharmacokinetic parameters correlated with efficacy in this infection model. The duration of time that serum levels exceeded the MIC has been shown to be the major pharmacokinetic parameter correlating with efficacy in different models (Frimodt-Moller et al., 1986; Vogelman et al., 1988; Andes et al., 1995). Frimodt-Moller et al. (1986) demonstrated a significant correlation between the 50% effective dose and the time the serum concentration remained above the MIC for 14 cephalosporins against a single strain of S. pneumoniae. Recently, in an abstract form, Andes et al. (1995) demonstrated that maximum efficacy was observed when serum amoxicillin levels exceeded the MIC for 40 to 50% of the dosing interval.

We observed differences in the P₅₀ values, relative to the MICs. These differences were particularly relevant to P^R strains. Against P15986 and P40422 strains (two highly P^R strains with amoxicillin MICs of 2 µg/ml), the P₅₀ of amoxicillin (achieving a 2.07- and 2.64-log unit reduction in CFU) was 1346 and 1361 mg/kg, respectively. In contrast, with the P12698 strain (also highly P^R, with the same amoxicillin MIC of 2 µg/ml), the P₅₀ required (achieving a 2.98-log unit reduction in CFU) was 301 mg/kg. On the other hand, with these three strains, the ratios of their P₅₀ values to the MIC ranged from 150 to 680 liters/kg. These discrepancies were also observed, to a lesser extent, in survival studies. Finally, our results showed that in vivo activities of amoxicillin against S. pneumoniae strains can be slightly different despite similar MICs. -Lactam-tolerant S. pneumoniae strains have already been reported (Liu and Tomasz, 1985). Antibiotic resistance and penicillin tolerance in clinical isolates of Streptococcus species have also been reported (Handwerger and Tomasz, 1985; Tuomanem et al., 1986). Because variations in potency of amoxicillin among the differents strains of S. pneumoniae could not be explained by differences in pharmacokinetic or pharmacodynamic activity, the antibiotic tolerance trait, in combination with increased levels of penicillin resistance of S. pneumoniae strains, would explain these discrepancies in vivo.

In conclusion, our study demonstrated that the standard in vitro MIC was an excellent predictor of the relative in vivo potency of amoxicillin against pneumococcal strains, including highly P^R strains. Furthermore, these data suggest that the estimated MIC breakpoint for amoxicillin resistance would be 4 µg/ml, although this remains to be confirmed in clinical studies.