Therapeutic Efficacy of 29 Antimicrobial Regimens in Experimental Intraabdominal Sepsis

REVEWS OF NFECTOUS DSEASES. VOL. 3, NO.3. MAY-JUNE 1981 1981 by The University of Chicago. 0162-0886/81/0303-0009$02.00 Therapeutic Efficacy of 29 Antimicrobial Regimens in Experimental ntraabdominal Sepsis John G. Bartlett, Thomas J. Louie, Sherwood L. Gorbach, and Andrew B. Onderdonk From the nfectious Disease Research Laboratory, Veterans Administration Hospital, and the Department of Medicine, Tufts University School of Medicine, Boston, Massachusetts An animal model of colonic perforation was used to examine the efficacy of 29 antimicrobial regimens in the treatment of intraabdominal sepsis. Efficacy was judged on mortality during the first 12 days after challenge and on the incidence of intraabdominal abscess noted at necropsy upon completion of the experiment. n general, antimicrobial agents that are active against coliform bacteria prevented early mortality, whereas drugs that are active against Bacteroidesfragilis were most effective in reducing the incidence of late abscess formation. Exceptions were metronidazole, which produced a significant reduction in early mortality, and chloramphenicol, which caused only a modest reduction in the incidence of abscess. Optimal results were obtained with several regimens that showed good in vitro activity against both coliforms and B. fragilis. ntraabdominal infections often involve a complex microflora composed of both aerobic and anaerobic bacteria [1]. The virtual impossibility of differentiating pathogens, commensals, and symbionts in these mixed infections poses a major dilemma. Such distinctions have important implications for our understanding of the pathophysiology of mixed infections, and these distinctions are critical for a rational approach to selection of antimicrobial agents. n an effort to simulate sepsis following colonic perforation, we have developed an animal model of intraabdominal sepsis by use of an intraperitoneal challenge with pooled cecal contents. Our initial studies with this model indicated a twostage disease. At first, we observed acute peritonitis, coliform bacteremia, and a 35%-450/0 mortality. All animals that survived the early acute peritonitis developed the second stage of the disease, which was characterized by formation of intraabdominal abscesses at five to seven days after Received for publication September 16, 1980, and in revised form November 20, 1980. This study was supported by grants from The Upjohn Company; Pfizer Laboratories; Eli Lilly and Company; Smith, Kline & French; Merck Sharp & Dohme; Schering; Bristol Laboratories; Hoechst-Roussel Pharmaceuticals; and Hoffmann-LaRoche. Please address requests for reprints to Dr. John Bartlett, Division of nfectious Diseases, The John Hopkins Hospital, Baltimore, Maryland 212Q5. challenge [2]. Bacteriologic studies have shown that Escherichia coli is the numerically dominant organism in exudate obtained during the early, peritonitis stage, whereas Bacteroides fragilis is the numerically dominant microbe in the contents of abscesses [3]. n subsequent experiments inocula composed of various bacterial species in pure culture were used in place of the mixed fecal flora. The results showed that E. coli caused early lethality, and B. fragilis appeared to be responsible for formation of intraabdominal abscesses [4, 5]. Our conclusions regarding the role of coliform bacteria and anaerobes in this animal model were further supported with antimicrobial trials [6]. We noted that gentamicin protected against lethality, but had no effect on the incidence of intraabdominal abscess; by contrast, clindamycin failed to alter mortality, but produced a significant reduction in the incidence of abscess. The initial selection of antimicrobial agents for testing in this model was based on the restricted spectra of activity of gentamicin and clindamycin. These agents permitted comparison of results of treatment directed against coliforms and treatment directed against anaerobes. However, numerous other antimicrobial agents and combinations of agents are commonly used for patients with intraabdominal sepsis. The present study is a comparative trial of 29 chemotherapeutic regimens in this model. Results with 12 of these regimens 535

536 Bartlett et al. have been reported previously and are included here for comparison [6-10]. Materials and Methods Animal Model. Details of this animal model and the natural course of the disease have been summarized previously [2, 3]. Briefly, male Wistar rats (Charles River Breeding Laboratories, Wilmington, Mass.) weighing 160-180 g were used for all experiments. The inoculum was prepared in an anaerobic chamber by use of techniques designed to insure a uniform challenge to all recipients. Cecal contents from meat-fed rats were pooled and then mixed with an equal volume of prereduced peptone yeast glucose broth and sterile barium sulfate (10070 wt/vol), The pooled inoculum was filtered through surgical gauze, placed into 5-ml vials, quick-frozen in liquid nitrogen, and then stored at - 40 C until used. Previous studies indicated that the concentrations of bacterial species of this inoculum did not change after prolonged periods of storage in a frozen state [3]. At the time of challenge, the frozen inoculum was thawed in an anaerobic chamber. Aliquots of 0.5 ml were placed into sterile gelatin capsules for intraperitoneal insertion, by midline abdominal incision, into rats. Antimicrobial agents. Seventeen antimicrobial agents were tested either alone or in various combinations. These agents were clindamycin (The Upjohn Company, Kalamazoo, Mich.), chloramphenicol (Parke, Davis & Company, Detroit, Mich.), doxycycline (Pfizer Laboratories, New York, N.Y.), penicillin G (Squibb & Sons, Princeton, N.J.), carbenicillin (Pfizer Laboratories), rosamicin (Rosaramicin'P; Schering, Kenilworth, N.J.), metronidazole (Searle & Co., Chicago, ll.), cephalothin (Eli Lilly and Company, ndianapolis, nd.) moxalactam (Eli Lilly and Company), cefazolin (Smith Kline & French, Philadelphia, Pa.), cefoxitin (Merck Sharp & Dohme, West Point, Pa.), gentamicin (Schering), amikacin (Bristol Laboratories, Syracuse, N. Y.), tobramycin (Eli Lilly and Company), spectinomycin (The Upjohn Company), erythromycin (Abbott Laboratories, Chicago, ll.), cefotaxime (Hoechst-Roussel Pharmaceuticals, Somerville, N.J.), and sulfamethoxazole-trimethoprim (Hoffmann-LaRoche, Nutley, N. J.). Antimicrobial dosage. Antimicrobial dosage was selected according to levels of drug in the serum of healthy Wistar rats after administration of arbitrarily selected doses. n most instances, this approximated the usual recommended dose for patients in mg/kg multiplied by six to adjust for the body surface area of a 160- to 180-g animal [11]. The antimicrobial agents were given im to a minimum of five healthy control animals, and serum was obtained by cardiac puncture at intervals of 0.5, 1, 4, and 8 hr. Levels of drug in serum were measured by agar diffusion assay with Bacillus subtilis ATCC (American Type Culture Collection, Rockville, Md.) no. 6633 as the test organism [12], or by the hemolysis inhibition assay with Clostridium perfringens Boston Veterans Administration strain no. 249 as the test organism [13]. Levels of trimethoprim were determined by spectrophotometry [14]. Susceptibility to antimicrobial agents. n vitro susceptibility of strains of three bacterial species to the antimicrobial agents was tested. E. coli, Streptococcus faecalis, and B. fragilis were selected because, in terms of both frequency of recovery and numerical concentration, they were the predominant isolates recovered from infected sites of untreated animals [3]. By use of microtiter techniques, susceptibility tests were done with serial twofold dilutions of the antimicrobial agents [15]. E. coli and S. faecalis were grown in Mueller Hinton broth with an inoculum of 10 4-105 cfu/ml and incubated in air. B. fragilis was tested using an inoculum of 10 5-106 cfu/ml after growth in prereduced brain-heart infusion supplemented broth (Scott Laboratories, Fiskeville, R..). MCs were defined as the lowest concentrations giving no visible growth after a 24-hr incubation period for aerobes and after incubation for 48 hr in an anaerobic chamber for anaerobes. Treatment. The experiments were conducted sequentially using groups of 100-130 animals tested simultaneously. Each experimental group included 15-30 rats given an antimicrobial regimen and 10-30 untreated control animals. All antimicrobial agents were given by im injection. Treatment was initiated 4 hr after implantation and then was given at 8-hr intervals for 10 days. An exception was carbenicillin, which was given at 4-hr intervals. Evaluation of treatment. Results of antimi-

Regimensfor ntraabdominal Sepsis 537 crobial treatment were evaluated by two criteria: mortality and incidence of intraabdominal abscesses. Mortality was determined by the number of animals that expired within 12 days after implantation. All surviving animals were sacrificed at 12 days for necropsy to detect intra-abdominal abscesses. The criterion for abscess was a loculated, grossly purulent collection that demonstrated numerous polymorphonuclear cells and bacteria by direct gram stain. Animals were considered cured if they had survived the duration of the experiments and if necropsy revealed no intraabdominal abscesses. Treatment groups were compared with a normal approximation technique using the binomial distribution with 95070 confidence intervals [16]. Results Levels ojdrug in serum. Peak levels of drug in serum obtained with the antimicrobial test agents are summarized in table 1. The values given represent the mean levels obtained with doses used in Table 1. Antimicrobial doses and levels of drug in serum of rats. Mean level of drug in serum" at Antimicrobial Dose agent (mg) 0.5 hr 1 hr 4 hr 8 hr Amikacin 7 NT 40 NT 2 Gentamicin 2 NT 6.9 NT <0.5 Tobramycin 2 NT 7.1 NT <0.5 Cephalothin 20 35 6 0.5 <0.5 Cephalothin 55 113 51 3 <0.5 Cefazoiin 20 107 60 3 <0.5 Cefazolin 55 385 237 7 <0.5 Cefamandole 20 63 23 2 <0.5 Cefamandole 55 128 81 4 <0.5 Cefoxitin 20 71 18 4 <0.5 Cefotaxime 20 68 24 2 <0.5 Moxalactam 20 86 38 4 <0.5 Penicillin G 60 128 50 10 <0.5 Carbenicillin 90 510 290 2 <0.5 Doxycycline 4 NT 2.5 1 <0.5 Chloramphenicol 16 22 19 0.5 <0.5 Ciindamycin 16 5.5 3.1 NT <0.5 Rosamicin 0.2 0.2 0.1 <0.1 Erythromycin 16 1.6 3.2 1.6 <0.5 Metronidazole 15 26 33 12 4 Trimethoprim 2 2.1 3.2 2 <0.5 NOTE. NT = not tested. * Mean level of drug io serum in Ag/ml with ~5 healthy animals for each determination. subsequent experiments to compare efficacy. Two dose regimens were tested with cephalothin, cefazolin, and cefamandole because of the wide variations in dose recommendations for these agents in clinical practice. Sera obtained at 4 hr showed detectable levels ofall drugs except cephalothin (low-dose regimen), chloramphenicol, and rosamicin. Sera obtained at 8 hr showed no detectable bioactivity with any drug except for low levels with doxycycline, metronidazole, and amikacin. Subsequent testing of these agents after five doses at 8-hr intervals showed evidence of accumulation of doxycycline only, with a mean peak level of 4.2 ~g/ml and a mean trough level of 2.2 ug/ml. n vitro susceptibility. n vitro susceptibility tests showed E. coli to be susceptible to readily achieved levels of aminoglycosides, cephalosporins, cefoxitin, moxalactam, carbenicillin, doxycycline, and chloramphenicol (table 2). As anticipated, the strain of S. jaeca/is proved resistant to most agents tested. B. jragilis was sensitive to clindamycin, cefoxitin, cefotaxime, moxalactam, chloramphenicol, rosamicin, erythromycin, and metronidazole. Susceptibility was not tested for sulfamethoxazole-trimethoprim. Table 2. n vitro susceptibility of three organisms to antimicrobial drugs. Susceptibility of indicated organisms Antimicrobial Escherichia Streptococcus Bacteroides agent coli faecalis fragilis Amikacin 2 >128 >128 Gentamicin 0.1 >32 >32 Tobramycin 0.1 >32 >32 Spectinomycin 64 64 32 Cephalothin 4 32 128 Cefazolin 2 64 128 Cefamando1e 1 64 128 Cefoxitin 4 >64 8 Cefotaxime 2 >64 8 Moxalactam 1 >64 2 Penicillin G 32 8 16 Carbenicillin 8 64 8 Doxycycline 1 1 4 Chloramphenicol 4 8 8 Clindamycin >64 >64 0.5 Rosamicin 16 1 0.1 Erythromycin >64 32 0.1 Metronidazole >256 NT 4 NOTE. NT = not tested. Minimum inhibitory concentration in Ag/mt.

538 Bartlett et al. Results of antimicrobial treatment. The comparative efficacy of the 29 antimicrobial regimens tested in the rat model is summarized in table 3. With all regimens except clindamycin, mortality of treated animals was significantly less than that of untreated controls. Mortality with regimens that included an aminoglycoside was 5.6070 (24 of 425 animals); with cephalosporins it was 7.60/0 (29 of 379 animals). Mortality was comparable with penicillin, spectinomycin, chloramphenicol, metroni- dazole, and sulfamethoxazole-trimethoprim. With all these regimens, the reduction in mortality was significant when compared with untreated controls. The incidence of intraabdominal abscess noted at necropsy in animals that survived for 12 days after implantation showed considerable variation. Optimal results were obtained with clindamycin, cefoxitin, clindamycin plus gentamicin, cefotaxime, moxalactam, rosamicin plus gentamicin, cefamandole plus erythromycin, and carbenicillin Table 3. Results of treatment with antimicrobial agents for rats with intraabdominal sepsis. No. of animals No. of animals No. of animals that died/total no. of with abscesses/no. of cured/total no. of Treatment group animals in group (070) surviving animals (070) animals in group (070) Untreated controls 108/295 (37) 187/187 (100) 0/295 Gentamicin 2157 (4) 54/55 (98) 1/57 (2) Amikacin 2130 (7) 27/28 (96) 1/30 (3) Penicillin 4/30 (13) 24/26 (92) 2/30 (7) Sulfamethoxazoletrimethoprim 0/30 23/30 (77) 7/30 (23) Penicillin plus amikacin 0/30 22/30 (73) 8/30 (27) Chloramphenicol 2/60 (3) 34/58 (59) 24/60 (40) Cefazolin (L)* 2/30 (7) 14/28 (50) 14/30 (47) Cefamandole (L) plus tobramycin 4/30 (13) 10/26 (38) 16/30 (54) Doxycycline 7/30 (23) 7/23 (30) 16/30 (54) Clindamycin 37/89 (42) 3/53 (6) 49/89 (55) Cefamandole (L) 4/30 (13) 9/26 (35) 17/30 (57) Cephalothin (L) 1/30 (3) 11/29 (38) 18/30 (60) Erythromycin 16/60 (27) 6/44 (14) 38/60 (63) Carbenicillin 11/60 (18) 7/49 (14) 42/60 (70) Cefazolin (H) t 0/30 8/30 (27) 22/30 (73) Cefamandole (H) 1/30 (3) 7/29 (24) 22/30 (73) Cephalothin (H) 8/50 (16) 4/42 (10) 38/50 (76) Metronidazole 5/50 (10) 6/45 (13) 39/50 (78) Doxycycline plus gentamicin 2130 (7) 4/28 (14) 24/30 (80) Cefamandole (L) plus erythromycin 7/60 (12) 2/53 (4) 51/60 (85) Rosamicin plus gentamicin 2130 (7) 2/28 (7) 26/30 (87) Spectinomycin 1/52 (2) 6/51 (12) 45/52 (87) Clindamycin plus gentamicin 10/158 (7) 6/148 (5) 1421158 (90) Carbenicillin plus gentamicin 2160 (3) 4/58 (7) 54/60 (90) Moxalactam (L) 2/59 (3) 4/57 (7) 53/59 (90) Cefoxitin (L) 0/30 2/30 (7) 28/30 (93) Cefotaxime (L) 1/29 (3) 1/28 (4) 27/29 (93) Spectinomycin plus clindamycin 0/25 1/25 (4) 24/25 (96) Cefamandole (H) plus erythromycin 1/30 (3) 0129 29/30 (97) * L = low dose regimen (20 mg). t H = high dose regimen (55 mg).

Regimens/or lntraabdominal Sepsis 539 plus gentarmcm. The incidence of abscess in surviving animals treated with these regimens varied from 0 to 7010, and there was no statistically significant difference between regimens (figure 1). Animals treated with cephalothin, cefazolin, cefamandole, and cefamandole plus tobramycin showed abscess rates of 24070-50070 with the higher rates in the low-dose regimens; an exception was the group given a high dose of cephalothin, in which abscesses were noted in only 10010 of surviving animals. The incidence of abscess with chloramphenicol- 59% - was significantly lower than that of untreated controls but higher than that found in 18 of the 28 other treatment regimens. Overall efficacy of the regimens was evaluated on the basis of cure rates (figure 2). Superior single-agent regimens in this assessment were cefoxitin, cefotaxime, moxalactam, and spectinomycin. Comparable results were noted with erythromycin plus cefamandole and with each of the following when combined with gentamicin: carbenicillin, clindamycin, rosamicin, and doxycycline. Cure rates for these treatment groups ranged from 80010 to 97010, and there was no statistically significant difference between them. Treatment with cephalosporins was somewhat inferior primarily because of an increased incidence of intraabdominal abscess. The least effective regimens were aminoglycosides alone, penicillin, sulfamethoxazole-trimethoprim, chloramphenicol, and penicillin combined with amikacin, primarily because of the increased incidence of intraabdominal abscess. Discussion Animal models have been used extensively to study in vivo response to antimicrobial agents. n order for such models to have clinical application, they must simulate disease processes encountered in patients. Our previous studies with this model of intraabdominal sepsis indicate that several important criteria have been fulfilled. Bacteriologic studies of the inoculum show it to be similar to the fecal flora of humans. Also, the sequence of pathologic events simulates the septic complications often seen after colonic perforation in patients. Additionally, the bacteriology of the infection and the susceptibility profiles of these microbes are similar "to those reported in clinical studies of intraabdominal sepsis. <;efamandole (Hl+ Erythromycin Cefotaxime ~ Cefamardoleru "Erythromycin Ci~damlcin+SpeCtinOmYCin ~lin.dam~cin+ Gentam icin Clindam~cin Cefoxitin Moxalactam ---4 Carbenicillin+ Gentamicin ---4 Rosa~amicln+Gentamic in Cephalothin (H) Srctin:>mYCir Metronidazole Erlythro~ycin Doxycycline+Gentamicin Carbenicillin Cef;mandol1e (H Cefazolin(H DOxycy~line Cefe mandole u) CePh:lothin (Ll CTfamandole (L)+Tobramycin Cefazo~in (Ll Chloramphenicol peniciain+ Amikacin J Sui arethoxazol:-t rimethopr ir Pe~icillin Amikacin...- Gentamicin... o 10 20 30 ~o 50 60 70 ~o 90 100 % Of Surviving AnimalsWith Abscesses Figure 1. ncidence of intraabdominal abscess in surviving rats when sacrificed 12 days after intraperitoneal challenge with pooled cecal contents. Point indicates percent with abscesses; bar indicates 950,10 confidence interval constructed around portions using a normal approximation technique. Most animal models of therapeutic efficacy utilize a challenge of a single microbe in pure culture. n general, correlation between in vitro susceptibility and in vivo response has been good. ntraabdominal sepsis, however, poses far more complex questions because there are multiple bacterial species in both the inoculum and at infected sites. The problem is to determine which microbes are true pathogens and therefore require specific treatment. Because of variables in the hosts, differences in disease processes, lack of untreated controls, and the often critical role of surgical intervention, clinical trials have not provided definitive answers to these questions. These considerations make the animal model of intraabdominal sepsis particularly appropriate. Our previous studies suggested that primarily coliforms were responsible for early lethality in this model. The present study showed that, except

540 Bartlett et al. Erythromycin Cephal~thin (L) Cefamandole (L) Clindamyci n Doxyc~cline Crfamandole ~L)+TobramYCir Cef a:olin (L) Chloramphenicol Penicillin+ Art i kacin Sulfamethoxazole-trimethoprim Penicillin A~ j1kafin G~amicin Cefamandole (H)+Erythromycin ClindamylmtlpFctinomycin Cefotaxime Cefoxitin Moxalactam Carbeni~iin+- ~entaticin Clindamyclin+~entamicin Spectinomycin Rosaramicin +-Gentamic i n Cefamandole(L) +-Erythromycin Doxycycline +- Gentamicin Metronidazole qephalo~hin (H) Cefamandole (H) Cefazolin (H),, o 10 20 30 40 50 60 70 80 90 100 % Cured for clindamycin, all antimicrobial regimens tested produced a significant reduction in mortality compared with untreated control animals. Mortality with regimens that included an aminoglycoside or a cephalosporin was only 5070-8%; mortality of untreated controls was 38070. Penicillin G also reduced mortality, but this result might be explained by the use of large doses, which resulted in peak serum levels of drug well in excess of the MC for E. coli. More difficult to explain, however, is the reduced mortality achieved with metronidazole. Only 10070 of the animals that received this drug expired, despite its poor in vitro activity against E. coli. This observation is of interest in view of clinical trials showing that metronidazole treatment of mixed aerobic-anaerobic infections resulted in eradication of resistant coliforms as well as susceptible anaerobes from the infected site [17]. More extensive studies with this model showed that treatment with metronidazole reduces the Figure 2. Cure rate, indicating survival without evidence of intraabdominal abscesses, in rats sacrificed 12 days after intraperitoneal challenge with pooled cecal contents. Point refers to percent cured and bar indicates 95070 confidence limits. incidence of bacteremia due to E. coli, but this effect is noted only when susceptible anaerobic bacteria are included in the inoculum [8]. Similar results were noted with mixed cultures in vitro. A conclusion from this experiment was that anaerobic bacteria may convert metronidazole to metabolites that are active against the coliforms. The major differences between regimens were seen with the incidence ofintraabdominal abscess. n general, there was a good correlation between prevention of abscess and in vitro activity of the antimicrobial test agent vs. B. fragilis. The overall incidence of abscess with aminoglycoside treatment alone was 98070; with three cephalosporin regimens (cephalothin, cefamandole, and cefazolin) it was 30070. Among these three cephalosporins, cephalothin appeared somewhat superior, possibly because of its relatively greater resistance to the fj-actamase of B. fragilis. The lowest incidence of abscess was noted with clindamycin,

Regimensjor ntraabdominal Sepsis 541 erythromycin, carbenicillin, metronidazole, rosamicin, high-dose cephalothin, spectinomycin, cefotaxime, cefoxitin, and moxalactam. All these drugs, with the exception of cephalothin, showed good in vitro activity against the strain of B. fragilis employed in the inoculum. Regimens using these drugs alone or in combination with an aminoglycoside showed abscess rates of 0 to only 14070. A discrepancy was noted between activity in vitro and results in vivo with chloramphenicol treatment vs. B. fragilis. The incidence of abscess in chloramphenicol recipients was 59% despite a mean peak level of drug in serum of 22 ug/rnl, which is well in excess of the MC for B. fragilis. A possible explanation for this discrepancy between in vitro and in vivo response is that anaerobes may inactivate chloramphenicol by enzymatic reduction of the nitro group [9]. Because the process of inactivation is slow, the organisms appear susceptible according to the usual in vitro susceptibility tests. The clinical relevance of these observations is uncertain. Two agents of special interest were spectinomycin and sulfamethoxazole-trimethoprim, since previous studies have shown divergent results, which appear to reflect differences in assay methodology, with in vitro testing vs. B. fragilis [18-21]. The results obtained in the present experiments showed good in vivo activity against this organism for spectinomycin and relatively poor activity for sulfamethoxazole-trimethoprim. The conclusion of this study, in which multiple antimicrobial agents were tested in experimental intraabdominal sepsis, is that optimal results were obtained with drug regimens that were active against both coliforms and anaerobes. Enterococci did not appear to play an important role in this model despite their frequent presence in the mixed flora found at infected sites. This conclusion is based on the observation that treatment directed against enterococci, such as penicillin combined with an aminoglycoside, showed minimal difference compared with results with either penicillin or an aminoglycoside alone. Furthermore, our previous studies using inocula of various bacteria in pure culture have shown that enterococci failed to cause bacteremia, mortality, or abscesses in the rat model [4]. This point is emphasized because the necessity of selecting antimicrobial agents active against enterococci is a frequent topic of debate in the management of intraabdominal sepsis. Certain limitations in this study should be noted. First, the inoculum was composed of pooled cecal contents to insure that each animal received an identical challenge. However, a disadvantage of this experimental design is that conclusions are restricted to a limited number of microbial strains. Variations from our results might be anticipated with the use of inocula composed of other bacterial species or strains with different in vitro susceptibility profiles. Another limitation is that the experiment was standardized so that untreated animals had a mortality of 35%-45%, and all surviving animals had intraabdominal abscesses. Using a similar model, Nichols et al. [22] have shown results whose variation from ours depend largely on inoculum size. t is obvious that our model represents just one form in a spectrum of diseases associated with colonic perforation. Finally, antimicrobial treatment was initiated 4 hr after challenge. This time was arbitrarily selected because it proved sufficiently early to produce a therapeutic response, yet sufficiently delayed to simulate the realities of clinical practice. Our subsequent studies have shown that treatment of animals with clindamycin after abscesses have formed provides no apparent benefit (authors' unpublished observations). This indicates that the outcome may be quite different when antimicrobial treatment is initiated at a later time in the evolution of the infection. References 1. Thadepalli, H., Gorbach, S. L., Broido, P. W., Norsen, J., Nyhus, L. M. Abdominal trauma, anaerobes and antibiotics. Surg. Gynecol. Obstet. 137:270-276, 1973. 2. Weinstein, W. M., Onderdonk, A. B., Bartlett, J. G., Gorbach, S. L. Experimental intra-abdominal abscesses in rats: development of an experimental model. nfect. mmun. 10:1250-1255, 1974. 3. Onderdonk, A. B., Weinstein, W. M., Sullivan, N. M., Bartlett, J. G., Gorbach, S. L. Experimental intraabdominal abscesses in rats: quantitative bacteriologyof infected animals. nfect. mmun. 10:1256-1259, 1974. 4. Onderdonk, A. B., Bartlett, J. G., Louie, T., Sullivan Seigler, N., Gorbach, S. L. Microbial synergy in experimental intra-abdominal abscess. nfect. mmun. 13: 22-26, 1976. 5. Onderdonk, A. B., Kasper, D. L., Cisneros, R. L., Bartlett, J. G. The capsular polysaccharide of Bacteroides jragilis as a virulence factor: comparison of the pathogenic potential of encapsulated and unencapsulated strains. J. nfect. Dis. 136:82-89, 1977. 6. Weinstein, W. M., Onderdonk, A. B., Bartlett, J. G.,

542 Bartlett et al. Louie, T. J., Gorbach, S. L. Antimicrobial therapy of experimental intraabdominal sepsis. J. nfect. Dis. 132: 282-286, 1975. 7. Bartlett, J. G., Onderdonk, A. B., Louie, T. J., Kasper, D. L., Gorbach, S. L. Lessons from an animal model of intra-abdominal sepsis. Arch. Surg. 113:853-857, 1978. 8. Onderdonk, A. B., Louie, T. J., Tally, F. P., Bartlett, J. G. Activity of metronidazole against Escherichia coli in experimental intra-abdominal sepsis. J. Antimicrob. Chemother. 5:201-210, 1978. 9. Onderdonk, A. B., Kasper, D. L., Mansheim, B. J., Louie, T. J., Gorbach, S. L., Bartlett, J. G. Experimental animal models for anaerobic infections. Rev. nfect. Dis. 1:291-301, 1979. 10. Louie, T. J., Onderdonk, A. B., Gorbach, S. L., Bartlett, J. G. Therapy for experimental intraabdominal sepsis: comparison of four cephalosporins with c1indamycin plus gentamicin. J. nfect. Dis. 135(Suppl.):S8-S22, 1977. 11. Freireich, E. J., Gehan, E. A., Rail, D. P., Schmidt, L. H., Skipper, H. E. Quantitative comparison of toxicity of anticancer agents in mouse, rat, hamster, dog, monkey, and man. Cancer Chemotherapy Reports 50:219-244, 1966. 12. Sabath, L. D. A simple, rapid microassay for nephrotoxic antibiotics. Scope Monograph, The Upjohn Co., Kalamazoo, Mich., 1972, p. 7-33. 13. Louie, T. J., Tally, F. P., Bartlett, J. G., Gorbach, S. L. Rapid microbiological assay for chloramphenicol and tetracyclines. Antimicrob. Agents Chemother. 9: 874-878, 1976. 14. Schwartz, D. E., Koechlin, B. A., Weinfeld, R. E. Spectrofluorimetric method for the determination of trimethoprim in body fluids. Chemother 14(Suppl.): 22-29, 1969. 15. Rotile, C. A., Fass, R. J., Prior, R. B., Perkins, R. L. Microdilution technique for antimicrobial susceptibility testing of anaerobic bacteria. Antimicrob. Agents Chemother. 7:311-315, 1975. 16. Colton, T. Statistics in Medicine. Little, Brown, and Co., Boston, 1974, p. 153-158. 17. Eykyn, S. J., Phillips,. Metronidazole in surgical infections. J. Antimicrob. Chemother. 4(Suppl. C):75-81, 1978. 18. Rosenblatt, J. E., Gerdts, A. M. Activity of spectinomycin against anaerobes. Antimicrob. Agents Chemother. 12:37-39, 1977. 19. Phillips, 1., Warren, C. Susceptibility of Bacteroides fragilis to spectinomycin. J. Antimicrob. Chemother. : 91-95, 1975. 20. Rosenblatt, J. E., Stewart, P. R. Lack of activity of sulfamethoxazole and trimethoprim against anaerobic bacteria. Antimicrob. Agents Chemother. 6:93-97, 1974. 21. Phillips, 1., Warren, C. Activity of sulfamethoxazole and trimethoprim against Bacteroides fragilis. Antimicrob. Agents Chemother. 9:736-740, 1976. 22. Nichols, R. L., Smith, J. W., Fossedal, E. N., Condon, R. E. Efficacy of parenteral antibiotics in the treatment of experimentally induced intraabdominal sepsis. Rev. nfect. Dis. 1:302-309, 1979.