Association between serum inhibitory and bactericidal concentrations and therapeutic outcome in bacterial endocarditis

REVIEW

Association between Serum Inhibitory and Bactericidal Concentrations and Therapeutic Outcome in Bacterial Endocarditis

DAVID L. COLEMAN, M.D. RALPH I. HORWITZ, M.D. VINCENT T. ANDRIOLE, M.D.

Several recent reviews on the therapy of bacterial endocardltis have recommended that a serum inhibitory and/or bactericidal concentration (SlC/SBC) of 1:8 or more be achieved to ensure successful therapeutic outcome. We conducted a methodokgic and statlstical analysis of the available literature on endocardltls to determine the association between SICYSBC tlters of 1:8 or more and therapeutic outcome. We reviewed 17 studies pubfished between 1948 and 1980 in which both SIC/SBC and therapeutic outcome were available. Factors that affect outcome, such as age, duration of symptoms, organism, and valve status, varied widely among the 226 patients. The methods used to measure StC/SBC differed with respect to the time of obtainkg the blood specimen relative to the antibiotic dose, size of the bacterial inoculum, type of broth, and definition of the bactericidal end-point. None of the 17 studies showed a significant association between SIC/SBC titers of 1:8 or more and survival or bacteriologic cure. Flfteen of the 17 also failed to demonstrate a significant association between SIC/SBC titers of 1:8 or more and medical cure. However, each of the studies that failed to demonstrate an association between SIC/SBC titers of 1:8 or more and improved therapeutic outcome had an insufficient sample size to confkentty exclude a falsenegative result. Analysis of the published data reveats insufficient evidence to demonstrate that SIC and SBC titers are of prognostic value in the therapy of patients wtth bacterial endocardltls.

New Haven,Connecticut

From the Infectious Disease and General Mediiine Sections, Department of Internal Medicine, Yale University School of Medicine. New Haven, Connecticut. This wc& was presented in abstract form at the 21st Interscience Conference on Antimicrobial Agents and Chemotherapy, Chicago. Illinois, November 1981. Dr. Horwitz is a Henry J. Kaiser Family Foundation Faculty Scholar in General Internal Medicine. Requests for reprints should be addressed to Dr. David L. Coleman, LCI 208, Infectious Disease Section, Yale University School of Medicine, 333 Cedar Street, New Haven. Cmnecticut 06510. Manuscript accepted on January 26. 1982.

260

Augusll982

The Amarken

The successful treatment of bacterial endocarditis requires several weeks of highdose antimicrobial therapy. The adverse consequences of inadequate treatment and toxicity of antimicrobial agents have led to the development of in vitro tests to monitor therapy. Although the disk agar diffusion method developed by Kirby, Bauer, and coworkers [l] is a useful guide to the selection of antimicrobial therapy for many infections, its usefulness is limited in the treatment of endocarditis because it does not adequately quantitate the concentration of antibiotic necessary to achieve an inhibitory or bactericidal effect. Tube dilution tests to determine the capacity of the patient’s serum to inhibit growth of or kill bacteria have been thought to be an important guide to the therapy of bacterial endocarditis. In 1947, Schlichter and McLean [2] published a report on 10 patients with streptococcal endocarditis whose therapy was monitored by testing the antimicrobial activity of their serum. The serum specimens, obtained just prior to

Journal Ot Medklna

Volume 73

BACTERIAL

the administration of the antibiotic dose, were serially diluted in broth. A known inoculum of the infecting organism was then introduced into the broth-serum mixtures and incubated overnight. The highest serum dilution that achieved inhibition of growth on overnight culture was determined and called the serum inhibitory concentration (SIC). The Schlichter method was subsequently modified by Fisher [3] who determined the serum bactericidal concentration (SBC) by subculturing the tubes that failed to exhibit macroscopic evidence of growth after overnight incubation. Measurement of the serum inhibitory and/or bactericidal concentration (SlC/SBC) in patients during therapy for bacterial endocarditis has become common practice. However, since the original description of the test, controversy has arisen over the association between SIC/SBC and therapeutic outcome. Schlichter and McLean [2] recommended that the SIC be maintained at a titer of 1:2 or greater in the treatment of streptococcal endocarditis. Geraci [4] suggested that SBC titers greater than or equal to 1:4 would lead to an optimal outcome in enterococcal endocarditis. A subsequent review [6] suggested a peak SBC of 1:8 or more. A recent textbook of infectious disease [S] , citing evidence in the rabbit model of endocarditis [7], has recommended a peak SBC of 1:8 or more and that the antibiotic dosage be reduced when levels of 1:16 or more are achieved. A subcommittee of the American Heart Association [8] also recently recommended that the peak SBC be maintained at 1:8 or higher in the treatment of streptococcal endocarditis. However, other reports on treatment of bacterial endocarditis in humans have failed to demonstrate an association between SIC/SBC and therapeutic outcome [9-l 11. The conflicting results and recommendations on the usefulness of this widely employed test are confusing. Therefore, we reviewed the literature on endocarditis to determine whether SIC/SBC titers of 1:8 or more were associated with improved therapeutic outcome. We also attempted to identify and evaluate those factors that may have contributed to the confusion about the value of this test. We found significant variability among laboratories in performance of the serum dilution test, important differences in the clinical characteristics of the patients studied, and insufficient evidence to support the association between SIC/SBC titers of 1:8 or more and outcome in the treatment of bacterial endocar-

ditis. METHODS We identified 90 studies published between 1948 and 1980 that reported on two or more patients with bacterial endocarditis. Twenty of these studies reported both the SIC/SBC and therapeutic outcome (defined later) and were selected for further analysis. The authors of the 20 studies were

Augud

ENDOCARDITIS-COLEMAN

ET AL.

contacted if additional information was needed for this review. Three of the 20 studies were excluded because we were unable to establish both SIC/SBC and therapeutic outcome for individual patients in these studies. Therefore, the data base was composed of the 17 remaining studies. The authors of four of these 17 studies [ 9,10,12,13] provided additional information on the clinical characteristics, SIC/SBC, and therapeutic outcome for individual patients in their studies. In addition, the authors of 14 of the studies [9-221 further clarified the methods used to determine the SIUSBC. The following information was extracted from each of the 17 studies: Study Characteristics. The total number of patients, number of patients for whom both SIC/SBC measurements and outcome were reported, final year of the patient sampling period, and recommended SICXSBC were determined. Laboratory Methods for Measuring SlC/SBC. The size of the bacterial inoculum, definition of the bactericidal end-point, type of broth, and whether the blood specimen was obtained within one hour after (peak) or before (trough) the administration of the antibiotic dose were recorded. Whenever timing of the blood specimen relative to the antibiotic dose was not specified, we classified the specimen as a trough to avoid underestimating the antibacterial activity achieved during a treatment interval. For example, if we recorded a SBC of 1:4 as a peak level in a patient with a successful therapeutic outcome, when in reality it was trough value, we would have incorrectly associated survival with a peak SBC that was both less than 1:8 and lower than that actually achieved. Association of SICISBC with Therapeutic tktcome for Individual Patients within Each Study. The patient’s age, gender, cardiac valve status, duration of symptoms prior to hospitalization, organism(s), SIC, and SBC were noted. Therapeutic outcome was defined for each patient according to three general categories: Survival: If the patient was alive, at the conclusion of the study, the period of follow-up was noted; if the patient died, the timing of death relative to the course of therapy was recorded. Medical cure: A medical cure was entered if the patient was alive after a course of medical (nonsurgical) therapy. If the patient died, or required a change in antimicrobial therapy or surgery because of lack of clinical improvement (from causes related to endocarditis), this patient was classified as a medical failure. Bacteriologic cure: A bacteriologic cure was counted if the patient was alive without evidence of relapse for 12 or more months after the conclusion of medical therapy, or if both Gram’s stain and culture of the valve at surgery gave negative results. If the patient died, a bacteriologic cure was counted if both Gram’s stain and culture of the valve at necropsy gave negative results. The patient was considered a bacteriologic failure if blood cultures gave positive results for the original organism at the time of death, or if either Gram’s stain or culture of the infected valve at surgery or autopsy gave positive results. Individual patients were included only if sufficient information was provided to determine both survival and at least one SIUSBC. The therapeutic outcomes for patients from each study with SIC/SBC titers of less than 1:8 were com-

1992

The American Journalot l#ediclne

Volume 73

261

BACTERIAL ENDOCARDITIS-CDLEMAN ET AL.

TABLE I

Characteristics of Studles Reviewed

First Author

FinalYear d Patbnt 9amplePeriod

TotalNumber d Patlent

Number of Palieets wltk9fc/96c and TheraPauPc outcaew (Percent c4total)

Recomnended !Mum DFutlotl

Group A: Authorssuggestinga positiveassociationbetween SIC/SBC and outcome Schlichter [ 21 Geraci [ 141 Geraci (41 Jawetz [ 151 Quinn [ 161 BlourM[ 131 Bryan ]251 Wilson [ 12)

1948 1953 1957 1959 1961 1964 1971 1974

10 33 172 14 27 89 3 45 393

10 (100) 3 (9) 35 (20) 3 (26) 12 (44) lO(l1) 3 (100) 22 (49) 98 (25)

Watanakunakorn[ 191 Reyes [ 171 Witchitz [ 201 Mills [9] Reyes [ 181 Chunn [21] Melacoff [lo] Ellner [ 221 Cooper illI

1971 1972 1973 1974 1975 1975 1977 1977 1979

3 23 9 19 14 7 68 9 17 169

3 (100) 22 (96) 5 (67) 13 (100) 14 (100) 7 (100) 45 (66) 4 (44) 15 (88) 128 (76)

Total

1:2 (SIC) 1:2 (SBC) 1:4 (SBC) 1:10 (SlC/SBC) 1:16 (SIC) 1:5-1:lO (SIC) 1:8 (SBC) 1:8’ (SBC)

Group B: Authorsnot suggesting an asscciatioobetween SlC/SBC and outcome

Total l

- (SBC) -

(SBC) (SBC) (SIUSBC) (SICXBC) (SICKSBC) (SlC/SBC) (SBC) (SBC)

Positiveassociationclaimed only for patientswith “late onset” prostheticvalve endccarditis(more thantwo monthsafter valve replace-

tTmw.

pared with those with SIC/SBC titers of I:8 or more. For purposes of this comparison, the SIC/SBC titers were subdivided into peak SIC, trough SIC, peak SBC and trough SBC. All data were collected by a single observer (D.L.C.). The Fisher exact test for categorical data was used to measure statistical significance between groups. Data were considered to have achieved statistical significance at a p value <0.05. Each studythat failed to show a statistically significant association between SICXSBC titers of I:8 or more and improved therapeutic outcome was further analyzed to determine the probability (ps) that a statistically significant difference might have been missed because of an inadequate sample size [23]. We specified that each negative study should possess adequate statistical power (1 - pi) to have at least an 80 percent chance of detecting a 50 percent reduction in adverse outcome for the patients with SIC/SBC titers of or more 1:8. If the statistical power of the study was less than 0.80, a false-negative result could not be exc)&ed 1241. RESULTS

The 17 studies are divided into two groups based on whether the author(s) did (Croup A) or did not (Croup B) suggest that the SIC/SBC was associated with therapeutic outcome (Table I). Sufficient information for

262

August 1982

associating the SIC/SW with therapeutic outcome for individual patients was avaifable significantly more often in the Croup B studies (76 percent) than in the Group A studies (25 percent) (x2 = 126.9, p
The serum specimen used in the tube dilution test was obtained only at the peak antibiotic level in 62

Yhe Amerlcen Journal ef Medfcfne Volwne 73

BACTERIAL

_._____-___-----

-_---~

Age (yr, mean Gender Male Female

f

Group A N=98(%) 46f

S.D.)

18

Total species

Streptococci (Croup D) Staphylococcal species Cram-negative Polymicrobial Total Valve status Natural valve Normal History of prior murmur Rheumatic heart disease Congenital heart disease

111

70 (72)’

45 (35)

7 9 10 2

3 0 68 12

(7) (9)’ (10) (2)

98

valve

6 (8) 15 (21)

58 (52)’ 19 (17)

27 (37)’

15 (14)

3 (4)

10 (9) 9 (8)

73

Total

(2) (0) (53)’ (10)

128

22 (30)

Total Symptom interval Natural valve 12mo >2 mo

18

85 (77) 26 (23)

79

111

10 (23) 33 (77)’

44 (80)’ 11 (20)

43

55

Because information describing clfnlcal characteristics of study patients was occasionally unavailable, some of the comparisons do not contain 226 patients. 7 p <0.05 l

Bactsricidal End-Point (prcsnf rsd&km i+l kK+cuknn)

TknhgofEksod Fksl Author

T P P P TIP T

Bryan [25) Wilson [ 121 Watanakunakcm Reyes [ 171 Witchitz [ 201 Mills 191 Reyes [ 181 Chunn [21] Malacoff [lo] Ellner [22] Cooper [11]

lnocufem S&e (s?Js~~mf)

Sdmpb (P=psk;T=tro@)

Schlichter [ 21 Geraci [ 141 Geraci [4] Jawetz [ 15) Quinn [ 161 Blount [ 131

“Mileu

39*

--

Methods Usad to Measure SIWSBC in 17 Studies

TABLE Ill

l

ET AL.

Group B N = 128(%)

55 (70) 24 (30)

Organism Streptococcal

Prosthetic

CDLEMAN

Cofnparlm of 226 Patlents In 17 Studies’

TABLE II

percent of the patients (nine studies). The trough level was used exclusively in 33 percent of the patients (two studies). The remaining studies used both peak and trough serum levels. The size of the bacterial inoculum used to test the serial dilutions of serum varied 1,OOO-fold from lo3 to lo6 organisms per ml (Table Ill). The types of broth used to culture the organism and to dilute the serum specimens also varied. Either Trypticase soy cr Mueller-f-finton broth was used in more than half the patients (eight studies). At least five different types of broth were used in the remaining patients. In only one study was the broth supplemented with pooled human serum in the serial dilution of the patients’ serum specimens. The definition of the bactericidal end-point ranged from a 99.0 to a 100 percent reduction in the original inoculum. Therefore, the number of viable organisms remaining after overnight incubation in what were labeled “bactericidal” concentrations of serum varied between zero and 1,000 among the 15 studies in which this information was available. A 99.9 percent reduction in inoculum size was used in more than half the patients (nine studies). Comparison of the methods used in the studies in Croups A and B did not demonstrate any important differences or trends. The methods used by the various authors were so divergent that substantive generalizations about each group could not be made. Association of SWSBC wlth Therapeutic Outcome for lndvklual Patknts wlthin Each Study. Survival and medical outcome were determined for all 226 patients.

ENDDCARDITIS-

[ 191

liqulde pour I’etufe

10’ 103 103 10’ 105-107 103

Type of Brotk

-

_Heart infusion Heart infusion Trypticase soy -

P P

103 105

Penassay Trypticase

TIP P P P P T/P TIP TIP TIP

10s 103 106 105 105 105 103 106 105

Mueller-Hinton Nutrient

des associations

99.9 99.9 99.9 99.2 100.0 99.9

soy

. Trypticase soy Mueller-Hinton Trypticase soy/human Mueller-Hinton Tryptose phosphate Trypticase soy

serum

99.9 100.0 100.0 99.9 99.0 106.0 999 99.9 99.9

d’antibiotiques.”

Augusll982

The Atnwlcan Journal ol Medlclna

Volume 73

283

BACTERIAL

ENDOCARDITIS-COLEMAN

TIMING OF BLOOD SAMPLE

ET AL

INOCULATION

SIZE

I org/dl

h

k

TYPE OF BROTH

Figure 7. The methods used to measure SICI SBC in each patient: timing of the blood sample relative to the antibiotic dose, size of the bacterial inoculum, type of broth, and definition of the bactericidal end-point. The number at the top of each bar refers to the percentage of patients with the respective technique. The ihoculum size in the study by Quinn and Colville [I61 was rounded off to lo6 organisms/ml.

When many studies with nonsignificant results indicate a consistent trend in favor of a particular therapy, the studies are sometimes analyzed after the data are combined. However, the results of a pooled analysis may be misleading unless certain methodologic requirements are met [ 261. In addition to a consistent trend, requirements for a pooled analysis include: comparable and homogeneous clinical groups; similar criteria for diagnosing the disease and detecting the outcome; comparable methods for treating and following the patients; and similar methods for determining and interpreting SIC/SBC titers. Since the 17 studies in this review did not fulfill these methodologic requirements, a pooled analysis was not performed. In two of the studies, there was a significant association between SIC/SBC titers of 1:8 or more and medical cure. An improved likelihood of medical cure for patients with a trough SIC of 1:8 or more was found in the study by Quinn and Colville [IS] and for patients with a peak SBC of 1:8 or more in one of the studies by Reyes and co-workers [ 171. Patients with prosthetic valve endocarditis differ in their response to antimicrobial therapy from patients with natural valve endocarditis. Therefore, the five studies [lo-12,18,21] containing a total of 31 patients with prosthetic valve infection were further analyzed to determine the association between SIC/SBC titers of 1:8 or more and therapeutic outcome. The patients in each study with natural valve infection were evaluated separately from those with prosthetic valve infection. In addition, the two groups of patients in each study were analyzed together. SIC/SBC titers of 1:8 or more were not associated with improved therapeutic outcome for patients with natural or prosthetic valve endocarditis using either method of analysis. COMMENTS

The bacteriologic outcome was determined for 134 of the 226 patients (59 percent). Three of the 17 studies did not contain sufficient information to ascertain the bacteriologic status of any of the patients [ 2,14,25]. A significant association between SIC/SBC titers of 1:8 or more and either survival or bacteriologic cure was not found in any of the 17 studies. In addition, 15 of the 17 studies failed to demonstrate a significant association between SIUSBC titers of 1:8 or more and medical cure. However, none of these 15 studies had sufficient statistical power to detect a 50 percent reduction in adverse therapeutic outcome for the patients with SIC/SBC titers of 1:8 or more. Thus, although an overwhelming majority of studies did not show a significant association between SICXBC titers of 1:8 or more and improved therapeutic outcome, false-negative results could not be confidently excluded because of inadequate sample size.

264

Augusl1992

The

American Journal of Medclna

Several current sources have recommended that SIC/SBC titers of 1:8 or more be achieved to ensure the successful treatment of bacterial endocarditis [5,6,8]. Despite the assumption that SIC/SBC titers are useful in monitoring therapy of bacterial endocarditis, a detailed analysis of the endocarditis literature indicates that these tests have not been shown to predict survival, medical cure, or bacteriologic cure. The failure to employ standardized methods in the measurement of SIC/SBC impairs comparison of the available studies. Our review disclosed five important areas of methodologic variation that may markedly influence the results of SIC/SBC determinations: time of obtaining the serum specimen relative to the antibiotic dose, size of the bacterial inoculum, type of broth, supplementation of the broth with human serum, and definition of the bactericidal end-point. Serum specimens were obtained at the anticipated

Volume 73

BACTERIAL

peak and trough of antimicrobial activity. However, the optimal time for obtaining the specimen is uncertain. Therapeutic studies in animal models of endocarditis have reached conflicting conclusions. Although one study in dogs [27] suggested that trough levels were most useful in predicting outcome, others in rabbits [7,28] supported the use of peak levels. The efficacy of antimicrobial therapy during the interval between antibiotic doses depends on the drug’s serum and tissue half-life, volume of distribution, rate of bacterial killing, and mechanism of action, the organism being treated, the location of the infected valve, and the presence or absence of a prosthetic valve [29]. These factors should be considered in future studies on the role of peak or trough SIC/SBC. The results of SIC/SBC determinations have been markedly influenced by the 1,OOO-fold variation in the size of the bacterial inoculum used in the 17 studies. For example, a l,OOO-fold increase in the inoculum of Staphylococcus aureus or Streptococcus viridans may reduce the SBC by a factor of eight [30]. In addition, the SBC for Escherichia coli may be reduced four-fold and the SIC two-fold by a similar increase in inoculum size [ 301. Furthermore, large inocula (e.g., more than 106) may enhance the rate of spontaneous mutation to resistant forms 13 1,32]. The use of a low inoculum may result in the transfer of an insufficient number of organisms to solid media to ensure accurate measurement of the bactericidal level [33]. The use of five different types of broth among the 17 studies may also have produced marked variation in the results of the tube dilution tests. For example, the minimal inhibitory concentration of tobramycin for Pseudomonas aeruginosa may vary lOO-fold when resuits using Mueller-Hinton and Nutrient broth are compared [ 171. The minimal bactericidal concentration of penicillin for group B streptococci may vary between 0.15 and 1.04 pg/ml in tryptose phosphate, MuellerHinton, or Todd-Hewitt broths [34]. In addition, increased concentrations of calcium and magnesium may decrease the activity of aminoglycosides against P. aeruginosa [ 35-371 and enterobacteriaceae [38]. In only one study in our review was human serum added to the broth. Supplementation of the broth used in tube dilution tests with human serum has been recommended to ensure a more physiologic environment in vitro [ 33,391. The addition of human serum to the media may alter antibiotic performance in inverse proportion to the degree of protein binding. For example, the use of human serum as the diluent produced a four-fold decrease in the activity of highly protein-bound drugs such as rifampin [32] and oxacillin [39] against Staph. aureus. The definition of the bactericidal end-point also varied between studies. In order to be truly “bactericidal,” the

ENOOCARDITIS-COLEMAN

ET AL

diluted serum should kill 100 percent of the initial bacterial inoculum [40]. However, many laboratories define the bactericidal end-point as a 99.9 percent reduction in the initial inoculum because a few bacterial colonies may survive even at relatively high antibiotic concentrations. Interpretation of the bactericidal end-point is further complicated by the phenomenon of intermittent growth (“skip tubes”) of large numbers of bacteria at high antibiotic concentrations [41]. The “skip tube” phenomenon may occur because of in vitro mutation to a resistant form [32] and is a common source of interobserver variation in the determination of the bactericidal end-point in our laboratory. The in vitro measurement of SIC/SBC has theoretic limitations that may affect extrapolation to the process of antibacterial activity in vivo. For example, platelets release bactericidal proteins during the clotting process that may alter the results in vitro [42]. Although the role of the leukocyte in eradicating endovascular infections is uncertain, the performance of the serum dilution test in a cell-free environment in vitro may fail to measure the effects of interactions between antimicrobial agents and leukocytes [ 431. We did not find a statistically significant association between SIC/SBC titers of 18 or more and survival or bacteriologic cure. All but two studies [ 16,171 failed to demonstrate an association with medical cure. However, the studies that did not find a positive association between SIC/SBC titers of 1:8 or more and therapeutic outcome may have missed clinically important differences due to inadequate sample size [24]. Interpretation of these 17 studies is further impaired by the failure to report SICXBC titers for more than half the patients. In addition, the patients with SIC/SBC titers of less than 1:8 and those with SIC/SBC titers of I:8 or more were not similar with respect to factors that may affect outcome such as age, underlying medical illness, type of heart value, organism, duration of symptoms, and prior cardiac disease [44,45]. The risks of overemphasizing the need for an arbitrary SICISBC in the absence of supporting evidence are considerable. Drug toxicity might be needlessly increased in a clinically stable patient with SIC/SBC titers of less than 1:8. Antimicrobial efficacy might be mistakenly reduced by lowering the antibiotic dosage in a patient with “unnecessarily high” SIC/SBC titers. Diagnostic and therapeutic vigilance might also be reduced in a patient with a SIC/SBC titer of 1:8 or more. CONCLUSIONS Insufficient data are available to support the recommendation that SlClSBC titers of 1:8 or more be achieved in the treatment of patients with bacterial endocarditis. The available studies employ widely di-

August 1992

The American Journal of Medicine

Volume 73

265

BACTERIAL

ENDDCARDITIS-COLEMAN

ET AL.

vergent methods and contain very heterogenous patient populations. None of the studies showed a statistically significant association between survival or bacteriologic cure and SIC/SBC titers of 1:8 or more. Fifteen of 17 studies did not show an association between SIC/SBC titers of 1:8 or more and medical cure. However, we are unable to exclude the possibility that SIC/SBC titers of 1:8 or more are associated with a favorable therapeutic outcome because of the small number of patients within each study. The prognostic value of SICYSBC titers in the therapy of bacterial endocarditis cannot be determined on the basis of the published data. Therefore, a prospective, cooperative study using standardized laboratory methods and a rigorous study design is needed to accurately assess the role of SIC/SBC determinations in

1.

2.

3.

4. 5. 6.

7.

6.

9.

10.

11. 12.

13. 14. 15. 16.

266

the treatment of patients with bacterial endocarditis. Review of the available literature in search of the “ideal” SIC/SBC appears futile until results of such a study are available. ACKNOWLEDGMENT We wish to express our gratitude to Drs. Chatrchai Watanakunakorn, John Mills, Milagros Reyes, Walter Wilson, Joseph Geraci, E. L. Quinn, Janine Witchitz, Elliot Frank, Janet Fischer, Myron Cohen, John Blount, David Gilbert, Jerrold Ellner, and Ernest Jawetz, for their assistance in providing additional information for this review. We would also like to thank Dr. Eugene Shapiro for his thoughtful review of the manuscript and Sarah Horwitz for her patient assistance in data analysis.

Bauer AW, Kirby WM, Sher~is X, Tuck M: Antibiotic susceptibility testing by a standardized single disk method. Am J Clin Pathol 1966: 45: 493-496. Schlichter JG, McLean H: A method of determining the effective therapeutic level in the treatment of bacterial endocarditis with penicillin. A preliminary report. Am Heart J 1947: 34: 209-211. Fisher AM: A method for the determination of antibacterial potency of serum during therapy of acute infections. A preliminary report. Bull Johns Hopkins tfosp 1952; 90: 313-320. Geraci JE: The antibiotic therapy of bacterial endoc&itk. Med Clin North Am 1958; 42: 1101-l 140. Mandell GL, Sande MA: Some newer aspects of infective endocarditis. Geriatrics 1975; 30: 97-102. Scheid WM, Sande MA: Endocarditis and intravascular infections In: Mandell GL, Douglas RG, Bennett JE, eds. Principles and practice of infectious diseases. New York: John Wiley and Sons, 1979: 653-690. Carrizosa J, Kaye D: Antibiitii concentrations in serum. serum bacteriocidaf activity and results of therapy of streptococcal encbcarditis in rabbits. Antimicrob Agents Chemother 1977; 12: 479-483. Bisno AL, Dismukes WE, Durack DT, et al.: AHA Committee Report. Treatment of infective endocarditis due to virkians streptococci. Circulation 196 1; 63: 730A. Mills J, Drew D: Serratia marcesens endocarditis: a regional illness associated with intravenous drug abuse. Ann Intern Med 1976; 84: 29-35; Malacoff RF, Frank E, Andrble VT: Streptococca I endocardiiis (nonenterococcal, non group A). Single vs. combination therapy. JAMA 1979; 241: 1807-1810. Cooper R, Mills J: Serratia endocarditis. A follow up report. Arch Intern Med 1980; 140: 199-202. Wilson WR, Jaumin PM, Danielson GK, Giuliani ER, Washington JA, Geraci JE: Prosthetic valve endocarditis. Ann Intern Med 1975; 82: 751-756. Blount JG: Bacterial endocarditis. Am J Med 1965; 38: 909-922. Geraci JE. Martin WJ: Antibiotic therapy of bacterial endocarditis. Circulation 1954: 10: 173-194. Jawetz E, Brainerd HD: Staphylococcal endocarditis. Am J Med 1962; 32: 17-24. Quinn EL, Colville JM: Subacute bacterial endocarditis. N Engl J Med 1961; 264: 835-842.

August

1982

The American

Journal of wkfne

Volume

17.

18.

19. 20.

21.

22.

23. 24.

25.

26.

27.

28.

29. 30. 31.

73

Reyes MP. Pallutke WA, Wylin RF, Lerner AM: Pseudomonas endocarditis in the Detroit Medical Center 1969-1972. Medicine (Baftimore) 1973; 52: 173-194. Reyes MP, Brown WJ, Lerner AM: Treatment of patients with pseudomonas endocarditis with high dose aminoglycoside and carbenicillin therapy. Medicine (Baltimore) 1978; 57: 57-67. Watanakunakorn C: Streptococcus bovis endocarditis. Am J Med 1974; 56: 256-260. Witchitz S, Gilbert C, Witchitz J, Go@erot M. Vachon F. Vie-Dupont V: Pseudomas endocarditis. A report of nine cases. Eur J Cardiol 1976; 4: 91-97. Chunn CJ, Jones SR, McCutchan JA, Young EJ, Gilbert DN: Hemophilus parainfluenza infective endocarditis. Medicine (Baltimore) 1977; 56: 99-l 14. Ellner JJ, Rosenthal MS, Lerner PI. McHenry MC: Infective endocardftis caused by slow-growing, fastidious, gram negative bacteria. Medicine (Baltimore) 1979; 58: 145158. Feinstein AR: Clinical biostatistics. St. Louis: C. V. Mosby. 1977: 320-334. Freiman JA. Chalmers TC, Smith H, Kuebler RR: The importance of beta, the type II error and sample size in the design and interpretation of the randomized control triil. Summary of 71 “negative” trials. N Engl J Med 1978; 299: 690694. Bryan CS. Marney SR. Alford RH, Bryant RE: Gram negative bacillary endocarditis. Interpretation of the serum bactericidal test. Am J Med 1975; 56: 209-215. Goldman L, Feinstein AR: Anticoagulants and myocardial infarction. The problems of pooling, drowning, and floating. Ann Intern Med 1979; 90: 92-94. Hamburger M, Gerartcis JC, Scott NJ, Carleton J, Beasley JS: Studies in experimental staphylococcal endocarditis in dogs. V. Treatment with oxacillin. J Lab Clin Med 1967; 70: 786-789. Tight RR: Ampicillin therapy of experimental enterococcal endocarditis. Antimicrob Agents Chemother 1980; 18: 307-310. Kunin CM: Blood level measurements and antimicrobial agents. Clin Pharmacol Ther 1974; 16: 251-256. Dunlop SG: The serum dilution bacterlckial test for antibiotic effectiveness. Am J Med Technol 1965; 31: 69-76. Gould JC: The laboratory control of antibiotic therapy. Br Med Bull 1960; 16: 29-34.

BACTERIAL

32.

33.

34.

35.

36.

37.

38.

McCabe WR, Lorian V: Comparison of the antibacterial activity of rffampicin and other antibiotics. Am J Med Sci 1968; 256: 255-265. Reller LB, Stratton CW: Serum dilution test of bactericidal activity. II. Standardizaticn and correlation with antimicrobial assays and susceptibility tests. J Infect Dis 1977; 136: 196-204. Kwang SK, Yoshimura RN, lmagawa DT, Anthony BF: Importance of medium in demonstrating penicillin tolerance by group B streptococci. Antimicrob Agents Chemother 1979; 16: 214-216. Davis SD, lanetta A: Antagonistic effect of calcium in serum on the activity of tobramycin against pseudomonas. Antimicrob Agents Chemother 1972; 1: 466-469. Zimelis VM, Jackson GG: Activity of aminoglycosides against Pseudomonas aeruginosa: specificity and site of calcium and maonesium antagonism. J Infect Dis 1973; 127: 663-669. Gilbert DN, Kutscher E, Ireland P, Barr&t JA, Sanford JP: Effect of the concentrations of maonesium and calcium on the in vitro susceptability of PseGomonas aeruginosa to gentamicin. J Infect Dis 1971; 124 suppl: S37-S45. Fass RJ. Barnishan J: Effect of divalent cation concentrations on the antibiotic susceptibilities of nonfermenters other than

39.

40.

41.

42. 43.

44.

45.

August 1982

ENDOCARDITIS-COLEMAN

ET AL.

Pseudomonas aeruginosa. Antimicrob Agents Chemother 1979; 16: 434-438. Stratton CW, Reller LB: Serum dilution test for bactericidal activity. Selection of a physiologic diluent. J Infect Dis 1977; 136: 187-195. Palmer DW: Inadequate response to “adequate” treatment of bacterial infection: L forms and “bactericidal” antibiotic activity. J infect Dis 1979; 139: 725-727. Lorian V, Siletti RP, Biondo FX, DeFreitas CC: Paradoxical effect of aminoglycoskfe antibiotics on the growth of gram negative bacilli. J Antimicrob Chemother 1979; 5: 613615. Weksler BB, Nachman RL: Rabbit platelet bactericidal protein. J EXD Med 1971: 134: 1114-l 130. Root Rk. lsturiz R, Molavi A, Metcalf JA, Malach HL: Interactions between antibiotics and human neutrophils in the killing of staphylococci. J Clin Invest 1981; 67: 247259. McNeil1 AM, Strong JE, Lockwood WR: Bacterial endocarditis: an analysis of factors affecting long term survival. Am Heart J 1978: 95: 448-453. Weinstein L, Dalton AL: Host determinants of response to antimicrobial therapy. N Engl J Med 1968; 279: 467-473. 524-530.580-587.

The Amerkaa

Joumaf d Madkfne

vehDnle 73

267