delayed treatment, ciprofloxacin showed the highest cure rate. ... related infections with drugs which kill not only growing but also nongrowing and adherent ...
Vol. 35, No. 4
ANTIMICROBIAL AGENTS AND CHEMOTHERAPY, Apr. 1991, p. 741-746
0066-4804/91/040741-06$02.00/0 Copyright © 1991, American Society for Microbiology
Killing of Nongrowing and Adherent Escherichia coli Determines Drug Efficacy in Device-Related Infections ANDREAS F. WIDMER,1t ADRIAN WIESTNER,1 RENO FREI,2 AND WERNER ZIMMERLI'* Departments of Internal Medicine and Research' and Central Laboratory of Bacteriology,2 University Hospital,
Basel, Switzerland Received 18 September 1990/Accepted 18 January 1991
Antimicrobial therapy of device-related infections often fails, despite the in vitro susceptibility of the infecting strain. Therefore, alternative laboratory-based in vitro tests are required to predict the outcome. Fleroxacin, ciprofloxacin, aztreonam, and co-trimoxazole were tested against Escherichia coli ATCC 25922 in vitro and in the tissue-cage animal model. The importance of early treatment was evaluated by starting the drugs either 30 min before or 4, 12, and 24 h after bacterial challenge. Results were compared with the in vitro drug efficacy against nongrowing and adherent Escherichia coil ATCC 25922. The alternative in vitro tests correlated highly with the outcome in the tissue-cage animal model. In the prophylaxis group (drug given 30 min before bacterial challenge), co-trimoxazole was less efficacious than the other three drugs (P < 0.001). In delayed treatment, ciprofloxacin showed the highest cure rate. It was also more potent than the other drugs against nongrowing and adherent E. coil ATCC 25922. The efficacies of aztreonam, fleroxacin, and ciprofloxacin dropped significantly (P < 0.01) when the time interval between bacterial challenge and the start of treatment was delayed to >4 h. These data emphasize (i) the need for proper timing of prophylaxis in patients undergoing implant surgery, and (ii) the possibility of successful treatment of established devicerelated infections with drugs which kill not only growing but also nongrowing and adherent bacteria.
Infections associated with prosthetic devices remain a major clinical problem (19). Antimicrobial therapy usually fails to cure such infections, unless the device has been removed (4). The high susceptibility of implants to pyogenic infections is due in part to impaired local host defense mechanisms (24, 25). Bacterial persistence is mainly related to the resistance of surface-adherent microorganisms to phagocytosis and antimicrobial treatment (7, 10, 11, 21, 23). The choice of a therapeutic antimicrobial agent is usually guided by the in vitro susceptibility of the anticipated (prophylaxis) or the isolated (treatment) microorganism to the drug (9). Importantly, however, bacteria in devicerelated infections are likely to be in the stationary phase of growth (silent infection), against which beta-lactam antibiotics show a diminished activity (20). Furthermore, adherence of bacteria alters their susceptibility to antimicrobial agents (7, 23). Additionally, a delay of prophylaxis in implant surgery has been shown to increase the risk of infection (13). Having previously shown that staphylococcal device-related infections can be cured with rifampin (23), we evaluated the efficacy of antimicrobial drugs against Escherichia coli. Four antimicrobial agents were tested by different in vitro tests. These results were compared with the outcome of infections in the tissue-cage animal model. (This study was presented in part at the 30th Interscience Conference on Antimicrobial Agents and Chemotherapy [7a, 23a].)
the initial strain, stored in skim milk at -70°C, was cultured every 3 months. The following antimicrobial agents, provided by the indicated manufacturers, were evaluated: ciprofloxacin (Bayer AG, Wuppertal, Federal Republic of Germany), aztreonam (Squibb AG, Zurich, Switzerland), co-trimoxazole and fleroxacin (Hoffmann-La Roche & Co., Ltd., Basel, Switzerland). Trimethoprim and sulfamethoxazole for in vitro testing were also supplied by Hoffmann-La Roche & Co., Ltd. Animal model. We used a previously described foreignbody animal model (25). In brief, four sterile polytetrafluoroethylene (Teflon) tubes (32 by 10 mm), perforated by 130 regularly spaced 1-mm-diameter holes (CIBA-GEIGY Ltd., Basel, Switzerland) were aseptically implanted into the flanks of albino guinea pigs (weight, 800 to 1,100 g). Experiments were started after complete healing of the wound, i.e., 3 to 6 weeks after surgery. Prior to each experiment, the interstitial fluid that accumulated in the tissue cages was checked for sterility. At day zero, tissue cages of the treatment group were infected by local inoculation of 7 x 104 CFU of E. coli ATCC 25922. Established infection was confirmed by quantitative cultures of the tissue-cage fluid before the first drug injection. Short-term antibiotic therapy was started at 4, 12, and 24 h after bacterial challenge. Each drug was given intraperitoneally (i.p.) every 12 h for 4 days, for a total of eight doses. Tissue-cage fluid was aspirated 2 days and 7 to 10 days after the completion of antibiotic therapy. Three animals in the aztreonam group and one in the fleroxacin group died as a result of the treatment (weight loss). Therefore, in these animals a sample for the final culture was taken at the time of death, i.e., 3 to 4 days after the end of therapy. A positive culture identified as E. coli was defined as treatment failure. In most experiments, one untreated animal with four cages served as a control. Bacterial counts of tissue-cage fluid of controls were quantitatively determined on days 1 to 4, 7, 11, and 14. None of these guinea pigs died. Animal experiments were performed with
MATERIALS AND METHODS Test strain and drugs. All experiments were performed with E. coli ATCC 25922. Working cultures were maintained on blood agar and were transferred weekly. A new aliquot of Corresponding author. t Present address: Department of Medicine, Division of Epidemiology, University Hospital, Iowa City, IA 52242. *
741
742
WIDMER ET AL.
ANTIMICROB. AGENTS CHEMOTHER. TABLE 1. Pharmacokinetics of the drugs examined in this study Drug level (,ug/ml)
Drug ~~~~Dose DoPeak in guinea pig serum (mgrug) (human) serumc
Co-trimoxazole (TMP/SMX)d Aztreonam Fleroxacin Ciprofloxacin
10/50 200 10 10
2.7/60 (1.7/60) 183 (242) 4.9 (4.2) 3.1 (2)
In tissue-cage fluid Peak Trough
MIC (pg/mW'
Ratio of trough level/MIC
1.5/27 8.8 3.1 0.95
0.06/1.2 0.125 0.07 0.02
40/>800 Aztreonam 0.125 200 400 Fleroxacin 0.07 0.62 20 Ciprofloxacin 0.02 0.07 10 a Trimethoprim-sulfamethoxazole (TMP/SMX) was tested in vitro with a mixture of 1:20.
confirming that most colonies gave a copy of themselves on the agar. The two other slides were transferred with the rack in Ca2E (50 mg/liter)-supplemented Trypticase soy broth containing antibiotics at a concentration corresponding to the twofold standard MBC. After another 24 h, they were processed as described above. Colonies of controls (adherent inoculum) and samples were counted after pressing them onto agar and expressed as the number of CFU per slide. Experiments were excluded if the adherent CFU of the controls were 103 CFU per slide, in order to have similar adherent inocula in each experiment. The test with four samples each was repeated at least three times. The kill rate (in percent) was calculated as follows: 100 - (CFU per slide on drug incubated slides/CFU per slide on control slides) x 100. For better comparison with the results of the time-kill curves, these values were converted into the loga-it rithm of killing as follows: 2 - [log (100 - %killing)]. t Statistics. The Fisher exact test with the Bonferroni ad-t justment, when appropriate, was used for examining proportions in the animal studies. If no P value is indicated, they difference was not significant. Variance component analysis was done for adherence results by using statistics software (SAS; SAS Institute Inc., Cary, N.C.). RESULTS Pharmacokinetics. In the pharmacokinetic studies (Tabl4 1), the peak levels corresponded to those suggested fox humans by the manufacturer. Tissue-cage fluid trough levels exceeded the MBCs of aztreonam, fleroxacin, and ciprofloxacin. The tissue-cage peak level of co-trimoxazole showed a ratio very similar to that of trimethoprim-sulfamethoxazole, as tested in vitro (1:18 versus 1:20, respectively). Based on this ratio, the trough level would also exceed the MBC of trimethoprim-sulfamethoxazole. The trough level of sulfamethoxazole was above the MIC (4.8 ,ug/ml), but was below the MBC (38 ,ug/ml) of sulfamethoxazole as a single agent. The half-lives of ciprofloxacin and fleroxacin were 1.5 and 1.8 h, respectively. Animal model. Mean bacterial counts in tissue-cage fluid TABLE 6. Killing of glass-adherent E. coli ATCC 25922 Drug
Co-trimoxazole Aztreonam Fleroxacin
Ciprofloxacin
CFU/slide (mean ± SE) Controls After treatmenta
Killing
%
Log killing
153 241 338 531
0 94.3 88.4 >99.9
0 1.25 0.93 >3
± ± ± ±
19 17 10 56
576 14 39 0
± ± ± ±
129 7 20 0
a Adherent bacteria were incubated at drug concentrations corresponding to twice the MBC determined in the logarithmic growth phase (see text).
744
WIDMER ET AL.
of untreated animals are shown in Fig. 1. Within 24 h after inoculation, the bacterial density rose from 7 x 104 to 1.35 x 107 CFU/ml and remained at this level during the whole experiment. In untreated animals, most tissue cages were spontaneously expelled 3 weeks after infection. Table 2 summarizes the results of the prophylaxis of tissue-cage infections with a single dose of an antimicrobial agent. Co-trimoxazole was significantly less efficacious than the other three drugs were, whether they were tested early or late after prophylaxis. Several tissue-cage fluids became sterile >1 day after the single-dose prophylaxis. Table 3 shows the results of a 4-day antimicrobial treatment of tissue-cage infections, starting at different time intervals after bacterial inoculation. All agents, except cotrimoxazole, were similarly effective if treatment was started 4 h after bacterial inoculation. A delay to 12 h significantly lowered the cure rates by aztreonam, fleroxacin, and ciprofloxacin. At this interval, ciprofloxacin was significantly more efficacious than the other drugs tested. At the 24-h interval, ciprofloxacin still sterilized 44% of the tissue-cage fluids. MBCs in the stationary and logarithmic phases of bacterial growth. The MBCs of fleroxacin, aztreonam, and co-trimoxazole for E. coli were 6 to 1,300 times higher when they were tested in the stationary growth phase than when they were tested in the logarithmic growth phase (Table 4). In contrast, ciprofloxacin killed growing and nongrowing E. coli equally well. Inoculum effect observed with E. coli ATCC 25922. Bacterial counts in tissue-cage fluid of untreated animals rose from 7 x 10' (inoculum) to approximately 1 x 107 CFU/ml within 24 h (Fig. 1). Therefore, MICs were also tested with inocula corresponding to these in vivo conditions (Table 5). Whereas the MICs of both quinolones remained in the susceptible range at 106 CFU/ml, the test strain was highly resistant to co-trimoxazole and aztreonam at this inoculum. At 107 CFU/ml, the MICs of the quinolones also increased to the resistant range. Time-kill studies. All antimicrobial agents except ciprofloxacin were less active against stationary-phase bacteria (Fig. 2 and 3). Drug efficacy on glass-adherent E. coli ATCC 25922. Adherent microorganisms are not killed as well as planktonic bacteria are (11, 23). We therefore tested the efficacy of the drugs against glass-adherent E. coli ATCC 25922 (Table 6). Co-trimoxazole completely failed to kill adherent bacteria in any experiment. In contrast, ciprofloxacin eradicated all adherent bacteria in three of three tests. This clear cutoff was not observed with fleroxacin or aztreonam, which reduced adherent bacteria in a range of 0.7 to 3 log units in different experiments. However, these two drugs were also more effective than co-trimoxazole (P < 0.01). DISCUSSION The ratio of the serum drug level of an antimicrobial agent to the MIC or the local antibiotic level to the MBC has been shown to correlate with the outcome in experimental infections (1, 5, 8, 17, 22). In a recently published study (14), the in vitro susceptibility predicted the clinical outcome. We analyzed the efficacy of four antimicrobial agents on adherent, growing, and nongrowing E. coli ATCC 25922 in vitro and compared these results with prophylaxis and treatment studies using an animal model for device-related infections (24, 25). Drug levels in tissue-cage fluid exceeded the MIC at any time during the treatment. Thus, a similar outcome
ANTIMICROB. AGENTS CHEMOTHER.
1089 -
-80
io7 u
10 5
_
10 2 0
10 9
2
4
6
8
10
12
14
16
18
20
22
24
2
4
6
8
10
12
14
16
18
20
22
24
B
108 1067-
E 10o5 10 4
.
102 0
Tile (Hours) FIG. 2. Time-kill studies of E. coli ATCC 25922 in the logarithmic and stationary phases of growth. The concentrations of the drugs correspond to the MBC in the logarithmic growth phase. (A) Aztreonam at 0.125 ,ug/ml); (B) co-trimoxazole (trimethoprim-sulfamethoxazole) at 0.06/1.2 jig/ml. Controls (L) and antibiotics (U) in the logarithmic growth phase and controls (0) and antibiotics (0) in the stationary growth phase are indicated.
could be expected. Surprisingly, the in vivo efficacy of the four antimicrobial agents differed widely. In contrast, the described alternative in vitro tests highly correlated with the outcome in the animal model. The properties of the antimicrobial agent needed for the optimal outcome were different whether the drug was administered early or late after bacterial challenge. In the prophylaxis group, the bacterial density in tissue cages was 7 x 104 CFU/ml; i.e., it was similar to that of the regular MIC test conditions. All drugs except co-trimoxazole were equally efficacious. The complete failure to prevent infection by co-trimoxazole was unlikely to be related to the thymidine content of interstitial fluid, since MICs were not influenced by adding 10% pooled, heat-inactivated guinea pig serum to MHB (data not shown). It also cannot be explained by the inability of co-trimoxazole to kill nongrowing bacteria, since aztreonam was prophylactically efficient, despite the same drawback. Furthermore, different betalactam antibiotics were shown to prevent infection in patients who underwent orthopedic surgery (13). However, it is conceivable that even in patients treated prophylactically, some microorganisms adhered to the implant and therefore could escape killing by co-trimoxazole (Table 6). If short-term treatment was started 4 h after bacterial challenge, all drugs except co-trimoxazole were equally effective. Four hours after inoculation, bacteria grew exponentially (Fig. 1), but their density was not yet higher than 3 x 105 CFU/ml. For these reasons, neither the inability to kill nongrowing bacteria nor the inoculum effect could be an
ANTIBIOTIC TREATMENT OF DEVICE-RELATED INFECTIONS
VOL. 35, 1991
nongrowing and adherent bacteria and high inocula also determined the outcome. In conclusion, the outcome of antimicrobial therapy in this animal model depended on the potency of the drug against nongrowing and adherent bacteria and the timing of drug administration. Established infections require long-term treatment with such drugs for complete healing.
109
108 / 3 107 i07~~~~~~~~0
E
10
u
10
6
104
103
102 0
E
109
-
107
-
2
4
6
8
10
12
14
16
18
20
22
24
106 -__
104
-
C
103102 0
2
4
6
8
745
10
12
14
16
18
20
22
24
Time (Hours)
FIG. 3. Time-kill studies of E. coli ATCC 25922 in the logarithmic and stationary phases of growth. The concentrations of the drugs correspond to the MBC in the logarithmic growth phase. (A) Fleroxacin at 0.07 ,ug/ml; (B) ciprofloxacin at 0.02 ,ug/ml. Controls (l) and antibiotics (U) in the logarithmic growth phase and controls (0) and antibiotics (0) in the stationary growth phase are indicated.
explanation for the treatment failures. Therefore, the inability of co-trimoxazole to kill surface-adherent E. coli best explains its lack of efficacy in vivo. The crucial property of an antimicrobial agent in the prophylaxis and early treatment group was the ability to kill adherent bacteria. If therapy was delayed to 12 h after bacterial challenge, the cure rate of aztreonam dropped from 84 to 8%. The considerable inoculum effect of aztreonam (Table 5) could be a trivial explanation for the failure of delayed treatment, suggesting a clinical significance of the inoculum effect (3). At 12 h after inoculation, bacterial counts in tissue-cage fluid were near the plateau. The low activity of aztreonam against nongrowing E. coli therefore offers another explanation for its poor efficacy in the delayed treatment of tissue-cage infections. Ciprofloxacin had the best efficacy against high inocula and nongrowing and adherent E. coli ATCC 25922 and was also most successful in the delayed treatment study. The ability of ciprofloxacin to sterilize tissue-cage abscesses remained similar whether the treatment was started at 12 or 24 h after infection. This indicates that an antimicrobial agent with the described properties does not necessarily loose its activity if the start of the treatment is further delayed. The main reason for the treatment failure of ciprofloxacin in 56% of the inoculations can be explained by the short treatment course. Unfortunately, longer treatment courses cannot be tested, since guinea pigs do not tolerate prolonged antimicrobial therapy (2, 15). In these late-treatment groups, the potency of an antimicrobial agent against
ACKNOWLEDGMENTS W.Z. was the recipient of a Career Development Award from the Swiss National Science Foundation. A.F.W. is the recipient of fellowship award 27922 from the Swiss National Science Foundation. This work was partially supported by a grant from Hoffmann-La Roche & Co. Ltd. The skillful help of Zarko Rajacic is gratefully acknowledged. REFERENCES 1. Blaser, J., B. B. Stone, M. C. Groner, and S. H. Zinner. 1987. Comparative study with enoxacin and netilmicin in a pharmacodynamic model to determine importance of ratio of antibiotic peak concentrations to MIC for bactericidal activity and emergence of resistance. Antimicrob. Agents Chemother. 31:10541060. 2. Bouchenaki, N., P. E. Vaudaux, E. Huggler, F. A. Waldvogel, and D. P. Lew. 1990. Successful single-dose prophylaxis of Staphylococcus aureus foreign body infections in guinea pigs by fleroxacin. Antimicrob. Agents Chemother. 34:21-24. 3. Brook, I. 1989. Inoculum effect. Rev. Infect. Dis. 11:361-368. 4. Dougherty, S. H. 1988. Pathobiology of infection in prosthetic devices. Rev. Infect. Dis. 10:1102-1117. 5. Drusano, G. L. 1988. Role of pharmacokinetics in the outcome of infections. Antimicrob. Agents Chemother. 32:289-297. 6. Edberg, S. C. 1986. The measurement of antibiotics in human body fluids: techniques and significance, p. 381-476. In V. Lorian (ed.), Antibiotics in laboratory medicine. The Williams & Wilkins Co., Baltimore. 7. Farber, B. F., M. H. Kaplan, and A. G. Clogston. 1990. Staphylococcus epidermidis extracted slime inhibits the antimicrobial action of glycopeptide antibiotics. J. Infect. Dis. 161:3740. 7a.Frei, R., A. F. Widmer, and W. Zimmerli. 1990. Program Abstr. 30th Intersci. Conf. Antimicrob. Agents Chemother., abstr. 31. 8. Frimodt-Moller, N., M. W. Bentzon, and V. F. Thomsen. 1986. Experimental infection with Streptococcus pneumoniae in mice: correlation of in vitro activity and pharmacokinetic parameters with in vivo effect for 14 cephalosporins. J. Infect. Dis. 154:511-517. 9. Greenwood, D. 1981. In vitro veritas? Antimicrobial susceptibility tests and their clinical relevance. J. Infect. Dis. 144:380385. 10. Gristina, A. G. 1987. Biomaterial-centered infection: microbial adhesion versus tissue integration. Science 237:1588-1595. 11. Gristina, A. G., R. A. Jennings, P. T. Naylor, Q. N. Myrvik, and L. X. Webb. 1989. Comparative in vitro antibiotic resistance of surface-colonizing coagulase-negative staphylococci. Antimicrob. Agents Chemother. 33:813-819. 12. Jones, C., D. L. Stevens, and 0. Ojo. 1987. Effect of minimal amounts of thymidine on activity of trimethoprim-sulfamethoxazole against Staphylococcus epidermidis. Antimicrob. Agents Chemother. 31:144-147. 13. Kaiser, A. B. 1986. Antimicrobial prophylaxis in surgery. N. Engl. J. Med. 315:1129-1138. 14. Lorian, V., and L. Burns. 1990. Predictive value of susceptibility tests for the outcome of antibacterial therapy. J. Antimicrob. Chemother. 25:175-181. 15. Manning, P. J., J. E. Wagner, and J. E. Harkness. 1986. Biology and diseases of guinea pigs, p. 149-181. In J. G. Fox, B. J. Cohen, and F. M. Loew (ed.), Laboratory animal medicine. Academic Press, Inc. (London), Ltd., London. 16. National Committee for Clinical Laboratory Standards. 1988. Methods for dilution antimicrobial susceptibility tests for bac-
746
17.
18.
19. 20.
21.
WIDMER ET AL. teria that grow aerobically, 2nd ed. Tentative standard. NCCLS document M7-T2. National Committee for Clinical Laboratory Standards, Villanova, Pa. Pangon, B., V. Joly, J. M. Vallois, L. Abel, A. Bure, N. Brion, A. Contrepois, and C. Carbon. 1987. Comparative efficacy of cefotiam, cefmenoxime, and ceftriaxone in experimental endocarditis and correlation with pharmacokinetics and in vitro efficacy. Antimicrob. Agents Chemother. 31:518-522. Schoenknecht, F. D., L. D. Sabath, and C. Thornsberry. 1985. Susceptibility tests: special tests, p. 1000-1008. In E. H. Lennette, A. Balows, W. J. Hausler, Jr., and H. J. Shadomy (ed.), Manual of clinical microbiology, 4th ed. American Society for Microbiology, Washington, D.C. Sugarman, B. 1986. Infections and prosthetic devices. Am. J. Med. 81(Suppl. 1A):78-84. Tuomanen, E., R. Cozens, W. Tosch, 0. Zak, and A. Tomasz. 1986. The rate of killing of Escherichia coli by P-lactam antibiotics is strictly proportional to the rate of bacterial growth. J. Gen. Microbiol. 132:1297-1304. Vaudaux, P. E., G. Zulian, E. Huggler, and F. A. Waldvogel.
ANTIMICROB. AGENTS CHEMOTHER. 1985. Attachment of Staphylococcus aureus to polymethacrylate increases its resistance to phagocytosis in foreign body infection. Infect. Immun. 50:472-477. 22. Vogelman, B., S. Gudmundsson, J. Leggett, J. Turnidge, S. Ebert, and W. A. Craig. 1988. Correlation of antimicrobial pharmacokinetic parameters with therapeutic efficacy in an animal model. J. Infect. Dis. 158:831-847. 23. Widmer, A. F., R. Frei, Z. Rajacic, and W. Zimmerli. 1990. Correlation between in vivo and in vitro efficacy of antimicrobial agents against foreign body infections. J. Infect. Dis. 162:96-102. 23a.Widmer, A. F., R. Frei, and W. Zimmerli. 1990. Program Abstr. 30th Intersci. Conf. Antimicrob. Agents Chemother., abstr. 30. 24. Zimmerli, W., P. D. Lew, and F. A. Waldvogel. 1984. Pathogenesis of foreign body infection. Evidence for a local granulocyte defect. J. Clin. Invest. 73:1191-1200. 25. Zimmerli, W., F. A. Waldvogel, P. Vaudaux, and U. E. Nydegger. 1982. Pathogenesis of foreign body infection: description of an animal model. J. Infect. Dis. 146:487497.