- Email: [email protected]
Clinical importance of “breakthrough” bacteremia
CLINICAL STUDIES
Clinical Importance of “Breakthrough” Bacteremia
MELVIN L. BARTH
P. WEINSTEIN, RELLER,
M.D.*
M.D.
Denver, Colorado
Fifty-one episodes of bacteremia and a single episode of fungemia occurred during treatment with seemingly adequate doses of appropriate antibiotics. Clinical findings in these “breakthrough” bacteremias and fungemia were compared with those in 448 nonbreakthrough episodes. Breakthrough was more likely to be caused by facultative or aerobic gram-negative rods (e.g., Enterobacteriaceae and Pseudomonas species) than by anaerobes. Of the underlying conditions examined, immunosuppressive doses of glucocorticosteroids, diabetes mellitus, and moderate renal failure were significantly more frequent in patients with breakthrough. A significant association was also observed between an lntra-abdominal primary focus of infection (abscesses, biliary tract or bowel infections) and the occurrence of breakthrough. Mortality in breakthrough bacteremia was 81 percent compared with 40 Percent in non-breakthrough episodes. The phenomenon of breakthrough bacteremia shows the potential limitations of antibiotic therapy alone. The term “breakthrough” bacteremia has been used to describe those bloodstream infections that occur during treatment with presumably adequate doses of appropriate antibiotics, i.e., ones to which the isolated organism is susceptible in vitro [ 11. The outcome of therapy in this syndrome has been discouraging, with mortality higher than that associated with other bacteremias [2]. Anderson et al [i] have suggested that poor outcome in gram-negative breakthrough bacteremia is related to adverse host factors, some of which they have identified, and possibly to inadequate serum antibiotic levels despite usual doses of these agents. We identified 52 episodes of breakthrough bacteremia and fungemia during a review of 500 cases of septicemia in adults [2,3] and report herein a detailed clinical evaluation of this syndrome. PATIENTS AND METHODS
From the Division of Infectious Diseases, Department of Medicine, University of Colorado School of Medicine, Denver, Colorado. Current address and address for reprint requests: University of Medicine and Dentistry of New Jersey-Rutgers Medical School, Academic Health Science Center, CN 19, New Brunswick, New Jersey 08903. Manuscript accepted August 11, 1983. l
The hospital records of all patients reported to have positive results of blood, culture at the University of Colorado Hospital and the Denver Veterans Administration Medical Center were reviewed according to the method of Feinstein et al [4,5] during a 21-month period. Full details have been reported previously [2,3] and are only summarized here. Blood culture methods at the two hospitals were identical. Each culture set was obtained by a separate venipuncture and consisted of 15 ml of blood divided equally into three culture bottles containing 45 ml of supplemented peptone broth. All blood cultures were incubated for 14 days, and a terminal subculture was carried out before results were reported as negative.
February 1984
The American Journal of Medicine
Volume 78
175
BREAKTHROUGH
TABLE I
BACTEREMIA-WEINSTEIN
Microorganisms isolated in Breakthrough and Non-Breakthrough Bacteremia
Microorganisms
l
and RELLER
Single
Breakthrough Episodes* Multiple
4 Staphylococci 8 Streptococci 14 Enterobacteriaceae 5 Pseudomonas species 2 Gram-positive anaerobes 1 Gram-negative anaerobes 3 Other bacteria 0 Fungi Number of episodes with percent in parentheses.
1
1 13 2 1 2 3 1
All 5 (8) 7 (12) 27 (46)+ 7 (12)+ 3 (5) 3 (5) 6 (10)
1 (a+
Non-Breakthrough Episodes’ 83 (14) 118 (21) 181 (32) 34 (6) 34 (6) 39 (7) 32 (5) 50 (9)
t p <0.05.
All information extracted from patient charts was coded for computer entry. Data were analyzed with programs of the Statistical Package for the Social Sciences [6]. Each blood culture with a positive result was evaluated by criteria published previously [3] to determine whether it more likely represented true infection or contamination. The primary source of infection was determined when possible. Body temperature and the presence or absence of hypotension, leukopenia, and absolute granulocytopenia, as defined earlier [3], were recorded. Underlying factors that might predispose to sepsis or influence outcome were noted; these included hematologic and nonhematologic neoplasms, immunosuppressive doses of corticosteroids, trauma, surgery, diabetes meliitus, renal failure, and hepatic cirrhosis, as defined previously [2]. The outcome of each episode of bacteremia and fungemia as well as the ultimate hospital course of each patient was determined. When death occurred, it was designated “directly related,” “indirectly related,” or “unrelated” to sepsis on the basis of criteria established earlier [2]. in accord with Anderson et al [ 11, breakthrough bacteremia was defined as persistent or recurrent bacteremia that occurred despite “appropriate” antimicrobial therapy. In addition, we required that the organism(s) isolated be sensitive by standardized in vitro disk susceptibility testing and that the antimicrobials being given at the time blood culture specimens were drawn were of adequate dose and route of administration as established by criteria in the MedicalLetter [7]. Serum antibiotic levels were not determined prospectively to confirm therapeutic adequacy. An episode of bacteremia or fungemia was defined by the first positive blood culture result or by a new positive blood culture result occurring more than two days after the last preceding positive result [3]. Any positive culture result obtained within 48 hours of a previous positive result was considered to represent the same episode. Statistical analysis was carried out with the modified chisquare test described by O’Brien and Shampo [8].
RkSULTS There were 51 episodes of breakthrough bacteremia and one of fungemia in 40 patients. In 41 instances, patients were receiving parenteral antimicrobials; 11
176
February 1964
episodes occurred in patients receiving oral agents. There were 25 episodes of gram-negative breakthrough bacteremia. Five patients were being treated with aminoglycosides alone, and five other patients were receiving combinatioti therapy in which aminoglycosides were the only agents administered that were effective in vitro against the infecting microorganism. The organisms isolated from patients with breakthrough bacteremia and fungemia were compared with those isolated from persons with non-breakthrough septicemias, i.e., bacteremia or fungemia that occurred in the absence of antibiotic therapy or in the presence of aniibiotics to which the isolated organism was subsequently shown to be resistant in vitro (Table I). Patients with breakthrough bacteremia had significantly more facultative and aerobic gram-negative rod isolates and fewer fungal isolates. Staphylococci and streptococci, which were uncommon breakthrough isolates, were even less common proportionately in patients with multiple breakthrough episodes. The frequency of polymicrobial episodes in breakthrough bacteremia (15 percent) was not different from that in non-breakthrough bacteremias (19 percent). Predisposing and other clinical and laboratory factors were compared in patients with breakthrough bacteremia and non-breakthrough bacteremia (Table II). Patients with breakthrough bacteremia were twice as likely to have been receiving immunosuppressive doses of corticosteroids (p
The American Journal of Medicine Volume 76
BREAKTHROUGH
TABLE Ii
BACTEREMIA-WEINSTEIN
and RELLER
Predisposing Factors and Clinical and Laboratory Findings in Patients with Breakthrough and NonBreakthrough Bacteremia
Single Predisposing factors Hematologic malignancy Other neoplasms Corticosteroids Trauma Surgery Diabetes mellitus Creatinine 2-5 mg/dl >5 mgldl Cirrhosis Clinical and laboratory Leukocytes <4,000/@ Granulocytes Hypotension Temperature
BreakthroughEpisodes* hlulliple
All
Non-Breakthrough Episodes’
3 3 5
2 3 19
5 14 5 11 5
0 3 9 2 0
1
4
5 (10)
31 (7)
5
1
6 (12)
53 (12)
3 5 5
1 0 0
4 (8) 5 (lo)+ 5 (lo)+
41 (9) 90 (20) 88 (20)
5 (10) 6 (12) 24 (46)t
50 (11) 93 (21) 108 (24)
5 17 14 13 5
32 129 65 59 44
(10) (33) (27)+ (25)+ (10)
(7) (29) (15) (13) (10)
findings
< 1,000/~1 <38’C
Number of episodes with percent in parentheses. + p <0.05. i: p <0.005. l
underlying factors predisposing to sepsis; only six patients (15 percent) had none of the defined predisposing factors, compared with 116 patients (26 percent) with non-breakthrough bacteremia. The primary sources of infection in breakthrough and non-breakthrough bacteremias are shown in Table III. intra-abdominal foci (abscesses, biiiary tract and bowel infections) accounted for 33 percent of all episodes of breakthrough bacteremia compared with 13 percent of non-breakthrough episodes. A possible related factor was the finding of multiple breakthrough bacteremias in liver transplant recipients. A total of 10 breakthrough episodes occurred in three such patients. Six of seven patients with more than one episode of breakthrough bacteremia were hospitalized at the
TABLE Ill
University of Colorado Hospital, and ail were male; only 11 of 29 patients (38 percent) with single breakthrough episodes at the same hospital were male. Details of these cases are shown in Table IV. The mean age (50 years) of the patients with multiple breakthrough episodes was no different from that of patients with single episodes. Six of the patients with multiple breakthrough bacteremias were infected with gram-negative microorganisms. Serious underlying diseases or factors known to be associated with diminished host defenses were present in ail patients with multiple episodes. There was one episode of breakthrough fungemia in this group of patients; the involved patient had undergone liver transplantation and also had multiple episodes of breakthrough bacteremia.
Sources of Breakthrough and Non-Breakthrough Bacteremia Non-Breakthrough
Breakthrough Episodes’ Source
Single
Multiple
All
Episodes*
Genitourinary tract Respiratory tract
5 2
1 4
6 (12) 6 (12)
85 (19) 80 (18)
Abscess Bowel Biliary tract Surgical wound Skin Bone Central nervous system Intravenous catheter Two or more possible sources Unknown
3 2 0
6 3 3
0 14
3 2
9 5 3 0 0 0 0 0 3 16
(17)+ (lo)+ (6)+
(6) (31)
32 20 10 18 16 2 5 14 27 137
(7) (4) (2) (4) (4) (0) (1) (3) (6) (31)
Number of episodes with percent in parentheses. + p <0.05. l
February 1984
The American Journal of Medicine
Volume 76
177
BREAKTHROUGHBACTEREMIA-WEINSTEIN and RELLER
TABLE IV
Characteristics and Outcome in Seven Patients with Multiple Episodes of Breakthrough Bacteremia Antimicrobial Therapy
Patient Number
Age/Sex
1
36/M
Corticosteroids, diabetes mellitus
1. Gentamicin 2. Ampicillin
2
52/M
Carcinoma of pancreas, diabetes mellitus, corticosteroids, surgery, granulocytopenia (third breakthrough episode only)
1. Chloramphenicol Gentamicin 2. Ampicillin Gentamicin
3
66/M
4
53/M
UnderlyingProblems
Hematologic malignancy, corticosteroids Liver transplant, diabetes mellitus, corticosteroids
46/M
Corticosteroids, renal failure
6
29/M
Liver transplant, corticosteroids
23/M
Hepatic cirrhosis, liver transplant, corticosteroids
1, Methicillin Gentamicin 2. Carbenicillin Cotrimoxazole
(6 days) 1. Staphylococcus epidermidis 2. Corynebacterium species, Micrococcus species 6
4. Cefazolin Chloramphenicol Gentamicin Cotrimoxazole 1. Carbenicillin 2. Carbenicillin Gentamicin 1. Cephalothin
Number of days after first breakthrough
1. Gentamicin 2. Carbenicillin Gentamicin 3. Gentamicin
February 1984
1. Bacteroides fragilis 2. Escherichia coli (6 days) 1. Escherichia coli
Source
Outcome
Respiratory tract Intra-abdominal abscess
Died
Intra-abdominal abscess lntra-abdominal abscess
Died
Respiratory tract
Respiratory tract Respiratory tract, intraabdominal abscess Unknown
Died
Died
Unknown
Genitourinary tract Respiratory tract
Bowel infarction Bowel infarction
Died
Biliary tract, surgical wound Biliary tract, surgical wound Bowel Liver abscess
Survived
Died
Liver abscess Liver abscess
episode.
Six patients with breakthrough bacteremia had been hospitalized for less than 48 hours at the time bacteremia occurred, and none died. In contrast, mortality was 70 percent in nosocomial breakthrough bacteremia. When breakthrough bacteremia occurred in patients receiving oral antimicrobial agents, mortality was lower (two of 11) than in patients who were receiving parenteral antimicrobials (30 of 41). Breakthrough bacteremia in patients receiving oral agents was more likely to be due to gram-positive cocci (four of 11; parenteral eight of 41) and was never due to Pseudomonas (parenteral seven of 41). Patients with breakthrough bacteremia
178
(6 days) 3. Candida parapsilosis (21 days) 4. Corynebacterium species (27 days)
2. Escherichia coli (21 days) 1. Escherichia coli 2. Escherichia coli (36 days) 3. Pseudomonas aeruginosa (46 days) 4. Escherichia coli, Bacteroides species (54 days)
4. Carbenicillin Gentamicin
l
2. Escherichia coli, enterococcus
1. Cefazolin 2. Gentamicin
2. Cefazolin 7
1. Escherichia coli 2. Proteus mirabilis, Peptostreptococcus species (3 days)’ 1. Escherichia coli
(7 days) 3. Escherichia coli, Pseudomonas aeruginosa, Citrobacter freundi (20 days) Klebsiella pneumoniae 2. Klebsiella pneumoniae
3. Ampicillin Gentamicin
3. Amphotericin
5
Microorganisms
The American Journal of Medicine
taking oral agents were also less likely to have been receiving corticosteroids (one of 11; parenteral 23 of 41) or to have a diagnosis of cirrhosis (zero; parenteral five of 41) or diabetes mellitus (one of 11; parenteral eight of 41). Indeed, patients with breakthrough bacteremia receiving oral agents were more likely to be free of all predisposing factors (four of 11) than were patients receiving parenteral therapy (two of 41). Overall mortality in breakthrough bacteremia was 81 percent, compared with 40 percent in non-breakthrough episodes. However, the proportion of septic (directly related) deaths was the same (20 percent). Thus, the
Volume 78
BREAKTHROUGHBACTEREMIA---WEINSTEIN and RELLER
difference in overall outcome between breakthrough and non-breakthrough episodes was due to a much greater likelihood of death not directly related to sepsis itself in breakthrough (41 percent) as compared with non-breakthrough (21 percent) episodes. COMMENTS
Breakthrough bacteremia accounted for 22 percent of gram-negative rod bacteremias at the University of California at Los Angeles Medical Center [l] and 10.4 percent of all septicemic episodes (9.5 percent of gram-negative rod bacteremias) at our hospitals. Nonetheless, little has been written about this phenomenon. Our findings suggest that breakthrough bacteremia is more likely to occur in patients whose host defenses are compromised, in particular those receiving immunosuppressive doses of corticosteroid therapy [2]. Liver transplant recipients also are at great risk. In general, these patients are among those with the most severe immunocompromise, and three such persons accounted for 10 episodes of breakthrough bacteremia in this series. Only 15 percent of patients with breakthrough bacteremia had none of the predisposing factors evaluated. Hematologic malignancies were no more common, and other neoplasms were less common (although not significantly) in patients with breakthrough bacteremia than in other bacteremic patients. Clearly, factors other than malignancy alone predispose to breakthrough. The observations in this study differ from the finding of Anderson et al [l] that leukopenia predisposes to breakthrough bacteremia. In our patients, neither leukopenia (leukocyte count below 4,OOO/pl) nor absolute neutropenia (granulocyte count below l,OOO/pl)predisposed to breakthrough compared with other bacteremias. The reasons for this difference are not clear. The underlying illnesses in the UCLA study [l] are not specified, and the patient populations may differ. Our definition of leukopenia required a lower absolute granulocyte count [3, g-151; therefore, we might have expected a stronger correlation between breakthrough and leukopenia than was observed in the UCLA series [ 11. Failure to find this correlation suggests that other factors are important in determining both the occurrence and outcome of breakthrough bacteremia. The increased frequency with which gram-negative rods caused breakthrough bacteremia (Table I) indirectly supports the findings by Anderson and colleagues [1] that subinhibitory levels of antimicrobials may play a role in the occurrence of breakthrough. Although serum antibiotic levels were not measured prospectively in our patients at the time of breakthrough, many patients with gram-negative bacteremia were receiving aminoglycosides, the serum levels of which have been found to
vary considerably after standard therapeutic doses 116-181. We believe that the occurrence of breakthrough bacteremia, therefore, constitutes a strong indication for the determination of serum antibiotic levels. Since suboptimal antibiotic levels may occur with greater frequency in patients receiving oral agents, the occurrence of breakthrough in this setting also constitutes an indication for measuring serum levels. Yet another factor associated with breakthrough bacteremia is a primary focus of infection that is difficult to treat. The increased frequency of intra-abdominal abscesses in this and the UCLA series [l] emphasizes this fact. Indeed, intra-abdominal foci were nearly three times more likely to be present in breakthrough compared with non-breakthrough bacteremia, and intraabdominal foci were the source of sepsis in three patients with multiple breakthrough episodes. Breakthrough bacteremia with gram-negative enteric rods, therefore, may provide an important clue to the presence of an intra-abdominal focus of infection, especially when multiple breakthrough episodes occur. The increased mortality associated with breakthrough bacteremia has been shown previously, but multivariate analysis suggested that breakthrough itself was relatively less important than other factors [2]. However, several factors associated with an increased risk of death in bacteremia are also found with greater frequency in breakthrough than in non-breakthrough episodes. These include gram-negative bacteria, hospital-acquired infection, and an intra-abdominal primary focus of infection (e.g., abscess) or an unknown source
[21.
The phenomenon of breakthrough bacteremia shows the potential limitations of antimicrobial therapy alone for sepsis, especially with single agents such as the aminoglycosides that have narrow therapeutic margins. Indeed, in such settings, therapeutic monitoring with antibiotic assays or serum bactericidal tests can provide evidence of adequate therapeutic margins [ 191, which may increase the likelihood of a successful outcome. Other forms of therapy may be required if the awesome associated mortality is to be reduced. For example, the need for surgical or catheter drainage of intra-abdominal abscesses is recognized [20]. Antiserum therapy has been shown recently to reduce mortality in gram-negative sepsis, but experience is limited [ 2 I]. Lastly, the role of blood culture methods that use resin removal of antibiotics to enable earlier detection of breakthrough bacteremia and improved therapy remains to be fully defined [ 221. ACKNOWLEDGMENT
We thank Drs. K. A. Lichtenstein and J. R. Murphy for their help in the initial design of this study, and Janet Gross for secretarial assistance.
February 1984
The American Journal of Medicine
Volume 78
179
BREAKTHROUGH
BACTEREMIA-WEINSTEIN
and RELLER
REFERENCES 1.
2.
3.
4.
5. 6. 7.
8. 9.
10.
11.
180
Anderson ET, Young LS, Hewitt WL: Simultaneous antibiotic levels in “breakthrough” gram-negative bacteremia. Am J Med 1976; 61: 493-497. Weinstein MP, Murphy JR, Reller LB, Lichtenstein KA: The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. II. Clinical observations with special reference to factors influencing prognosis. Rev Infect Dis 1983; 5: 54-70. Weinstein MP, Reller LB, Murphy JR, Lichtenstein KA: The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. I. Laboratory and epidemiologic observations. Rev Infect Dis 1983; 5: 35-53. Feinstein AR, Pritchett JA, Schimpff CR: The epidemiology of cancer therapy. Ill. The management of imperfect data. Arch Intern Med 1969; 123: 448-46 1. Feinstein AR, Pritchett JA, Schimpff CR: The epidemiology of cancer therapy. IV. The extraction of data from medical records. Arch Intern Med 1969; 123: 571-590. Nie NH, Hull CH, Jenkins JG, Steinbrenner K, Bent DH: Statistical package for the social sciences. 2nd ed. New York: McGraw-Hill, 1975. Medical Letter on Drugs and Therapeutics: Handbook of antimicrobial therapy, rev. ed. New Rochelle, New York: Medical Letter, Inc., 1976. O’Brien PC, Shampo MA: Comparing two proportions: the relative deviate test and chi-square equivalent. Mayo Clin Proc 1981; 56: 513-515. lssell BF, Keating MJ, Valdivieso M, Bodey GP: Continuous infusion tobramycin combined with carbenicillin for infections in cancer patients. Am J Med Sci 1979; 277: 311-318. Alavi BJ, Root RK, Djerassi I, et al: A randomized clinical trial of granulocyte transfusions for infection in acute leukemia. N Engl J Med 1977; 296: 706-711. Bodey GP, Buckley M, Sathe YS: Quantitative relationships between circulating leukocytes and infection in patients with acute leukemia. Ann Intern Med 1966; 64: 328340.
February 1984
The American Journal of Medicine
12. 13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Volume 78
Goodall PT, Vosti KL: Fever in acute myelogenous leukemia. Arch Intern Med 1975; 135: 1197-1203. Gurwith MJ, Brunton JL, Lank BA, Ronald AR, Harding GKM: Granulocytopenia in hospitalized patients. 1. Prognostic factors and etiology of fever. Am J Med 1978; 674: 121-126. Herzig RH, Herzig GP, Graw RG Jr, Bull MI, Ray KK: Successful granulocyte transfusion therapy for gram-negative septicemia: a prospective randomized controlled study. N Engl J Med 1977; 296: 701-705. Sickle EA, Greene WH, Wiernik PH: Clinical presentation of infection in granulocytopenic patients. Arch Intern Med 1975; 135: 715-719. Jackson G. Riff LJ: Pseudomonas bacteremia: pharmacologic and other bases for failure of treatment with gentamicin. J Infect Dis 1971; 124(suppl): S185-S191. Winters RE, Litwack KD, Hewitt WL: Relation between dose and levels of gentamicin in blood. J Infect Dis 1971; 124(suppl): S90-S95. Kaye D, Levison MB, Labovitz ED: The unpredictability of serum concentrations of gentamicin. Pharmacokinetics of gentamicin in patients with normal and abnormal renal function. J Infect Dis 1974; 130: 150-154. Stratton CW, Weinstein MP, Reller LB: Correlation of serum bactericidal activity with antimicrobial agent level and minimum bactericidal concentration. J Infect Dis 1982; 145: 160-168. Gerzof SG, Robbins AH, Johnson WC, Birkett DH, Nabseth DC: Percutaneous catheter drainage of abdominal abscesses: a five-year experience. N Engl J Med 1981; 305: 653657. Ziegler EJ, McCutchan AJ, Fierer J, et al: Treatment of gram-negative bacteremia and shock wtth human antiserum to a mutant Escherichia coli. N Engl J Med 1982; 307: 12251230. Appelbaum PC, Beckwith DG, Dipersio JR, Dyke JW, Salventi JF, Stone LL: Enhanced detection of bacteremia with a new BACTEC resin blood culture medium. J Clin Microbial 1983; 17: 48-51.
Clinical Importance of “Breakthrough” Bacteremia
MELVIN L. BARTH
P. WEINSTEIN, RELLER,
M.D.*
M.D.
Denver, Colorado
Fifty-one episodes of bacteremia and a single episode of fungemia occurred during treatment with seemingly adequate doses of appropriate antibiotics. Clinical findings in these “breakthrough” bacteremias and fungemia were compared with those in 448 nonbreakthrough episodes. Breakthrough was more likely to be caused by facultative or aerobic gram-negative rods (e.g., Enterobacteriaceae and Pseudomonas species) than by anaerobes. Of the underlying conditions examined, immunosuppressive doses of glucocorticosteroids, diabetes mellitus, and moderate renal failure were significantly more frequent in patients with breakthrough. A significant association was also observed between an lntra-abdominal primary focus of infection (abscesses, biliary tract or bowel infections) and the occurrence of breakthrough. Mortality in breakthrough bacteremia was 81 percent compared with 40 Percent in non-breakthrough episodes. The phenomenon of breakthrough bacteremia shows the potential limitations of antibiotic therapy alone. The term “breakthrough” bacteremia has been used to describe those bloodstream infections that occur during treatment with presumably adequate doses of appropriate antibiotics, i.e., ones to which the isolated organism is susceptible in vitro [ 11. The outcome of therapy in this syndrome has been discouraging, with mortality higher than that associated with other bacteremias [2]. Anderson et al [i] have suggested that poor outcome in gram-negative breakthrough bacteremia is related to adverse host factors, some of which they have identified, and possibly to inadequate serum antibiotic levels despite usual doses of these agents. We identified 52 episodes of breakthrough bacteremia and fungemia during a review of 500 cases of septicemia in adults [2,3] and report herein a detailed clinical evaluation of this syndrome. PATIENTS AND METHODS
From the Division of Infectious Diseases, Department of Medicine, University of Colorado School of Medicine, Denver, Colorado. Current address and address for reprint requests: University of Medicine and Dentistry of New Jersey-Rutgers Medical School, Academic Health Science Center, CN 19, New Brunswick, New Jersey 08903. Manuscript accepted August 11, 1983. l
The hospital records of all patients reported to have positive results of blood, culture at the University of Colorado Hospital and the Denver Veterans Administration Medical Center were reviewed according to the method of Feinstein et al [4,5] during a 21-month period. Full details have been reported previously [2,3] and are only summarized here. Blood culture methods at the two hospitals were identical. Each culture set was obtained by a separate venipuncture and consisted of 15 ml of blood divided equally into three culture bottles containing 45 ml of supplemented peptone broth. All blood cultures were incubated for 14 days, and a terminal subculture was carried out before results were reported as negative.
February 1984
The American Journal of Medicine
Volume 78
175
BREAKTHROUGH
TABLE I
BACTEREMIA-WEINSTEIN
Microorganisms isolated in Breakthrough and Non-Breakthrough Bacteremia
Microorganisms
l
and RELLER
Single
Breakthrough Episodes* Multiple
4 Staphylococci 8 Streptococci 14 Enterobacteriaceae 5 Pseudomonas species 2 Gram-positive anaerobes 1 Gram-negative anaerobes 3 Other bacteria 0 Fungi Number of episodes with percent in parentheses.
1
1 13 2 1 2 3 1
All 5 (8) 7 (12) 27 (46)+ 7 (12)+ 3 (5) 3 (5) 6 (10)
1 (a+
Non-Breakthrough Episodes’ 83 (14) 118 (21) 181 (32) 34 (6) 34 (6) 39 (7) 32 (5) 50 (9)
t p <0.05.
All information extracted from patient charts was coded for computer entry. Data were analyzed with programs of the Statistical Package for the Social Sciences [6]. Each blood culture with a positive result was evaluated by criteria published previously [3] to determine whether it more likely represented true infection or contamination. The primary source of infection was determined when possible. Body temperature and the presence or absence of hypotension, leukopenia, and absolute granulocytopenia, as defined earlier [3], were recorded. Underlying factors that might predispose to sepsis or influence outcome were noted; these included hematologic and nonhematologic neoplasms, immunosuppressive doses of corticosteroids, trauma, surgery, diabetes meliitus, renal failure, and hepatic cirrhosis, as defined previously [2]. The outcome of each episode of bacteremia and fungemia as well as the ultimate hospital course of each patient was determined. When death occurred, it was designated “directly related,” “indirectly related,” or “unrelated” to sepsis on the basis of criteria established earlier [2]. in accord with Anderson et al [ 11, breakthrough bacteremia was defined as persistent or recurrent bacteremia that occurred despite “appropriate” antimicrobial therapy. In addition, we required that the organism(s) isolated be sensitive by standardized in vitro disk susceptibility testing and that the antimicrobials being given at the time blood culture specimens were drawn were of adequate dose and route of administration as established by criteria in the MedicalLetter [7]. Serum antibiotic levels were not determined prospectively to confirm therapeutic adequacy. An episode of bacteremia or fungemia was defined by the first positive blood culture result or by a new positive blood culture result occurring more than two days after the last preceding positive result [3]. Any positive culture result obtained within 48 hours of a previous positive result was considered to represent the same episode. Statistical analysis was carried out with the modified chisquare test described by O’Brien and Shampo [8].
RkSULTS There were 51 episodes of breakthrough bacteremia and one of fungemia in 40 patients. In 41 instances, patients were receiving parenteral antimicrobials; 11
176
February 1964
episodes occurred in patients receiving oral agents. There were 25 episodes of gram-negative breakthrough bacteremia. Five patients were being treated with aminoglycosides alone, and five other patients were receiving combinatioti therapy in which aminoglycosides were the only agents administered that were effective in vitro against the infecting microorganism. The organisms isolated from patients with breakthrough bacteremia and fungemia were compared with those isolated from persons with non-breakthrough septicemias, i.e., bacteremia or fungemia that occurred in the absence of antibiotic therapy or in the presence of aniibiotics to which the isolated organism was subsequently shown to be resistant in vitro (Table I). Patients with breakthrough bacteremia had significantly more facultative and aerobic gram-negative rod isolates and fewer fungal isolates. Staphylococci and streptococci, which were uncommon breakthrough isolates, were even less common proportionately in patients with multiple breakthrough episodes. The frequency of polymicrobial episodes in breakthrough bacteremia (15 percent) was not different from that in non-breakthrough bacteremias (19 percent). Predisposing and other clinical and laboratory factors were compared in patients with breakthrough bacteremia and non-breakthrough bacteremia (Table II). Patients with breakthrough bacteremia were twice as likely to have been receiving immunosuppressive doses of corticosteroids (p
The American Journal of Medicine Volume 76
BREAKTHROUGH
TABLE Ii
BACTEREMIA-WEINSTEIN
and RELLER
Predisposing Factors and Clinical and Laboratory Findings in Patients with Breakthrough and NonBreakthrough Bacteremia
Single Predisposing factors Hematologic malignancy Other neoplasms Corticosteroids Trauma Surgery Diabetes mellitus Creatinine 2-5 mg/dl >5 mgldl Cirrhosis Clinical and laboratory Leukocytes <4,000/@ Granulocytes Hypotension Temperature
BreakthroughEpisodes* hlulliple
All
Non-Breakthrough Episodes’
3 3 5
2 3 19
5 14 5 11 5
0 3 9 2 0
1
4
5 (10)
31 (7)
5
1
6 (12)
53 (12)
3 5 5
1 0 0
4 (8) 5 (lo)+ 5 (lo)+
41 (9) 90 (20) 88 (20)
5 (10) 6 (12) 24 (46)t
50 (11) 93 (21) 108 (24)
5 17 14 13 5
32 129 65 59 44
(10) (33) (27)+ (25)+ (10)
(7) (29) (15) (13) (10)
findings
< 1,000/~1 <38’C
Number of episodes with percent in parentheses. + p <0.05. i: p <0.005. l
underlying factors predisposing to sepsis; only six patients (15 percent) had none of the defined predisposing factors, compared with 116 patients (26 percent) with non-breakthrough bacteremia. The primary sources of infection in breakthrough and non-breakthrough bacteremias are shown in Table III. intra-abdominal foci (abscesses, biiiary tract and bowel infections) accounted for 33 percent of all episodes of breakthrough bacteremia compared with 13 percent of non-breakthrough episodes. A possible related factor was the finding of multiple breakthrough bacteremias in liver transplant recipients. A total of 10 breakthrough episodes occurred in three such patients. Six of seven patients with more than one episode of breakthrough bacteremia were hospitalized at the
TABLE Ill
University of Colorado Hospital, and ail were male; only 11 of 29 patients (38 percent) with single breakthrough episodes at the same hospital were male. Details of these cases are shown in Table IV. The mean age (50 years) of the patients with multiple breakthrough episodes was no different from that of patients with single episodes. Six of the patients with multiple breakthrough bacteremias were infected with gram-negative microorganisms. Serious underlying diseases or factors known to be associated with diminished host defenses were present in ail patients with multiple episodes. There was one episode of breakthrough fungemia in this group of patients; the involved patient had undergone liver transplantation and also had multiple episodes of breakthrough bacteremia.
Sources of Breakthrough and Non-Breakthrough Bacteremia Non-Breakthrough
Breakthrough Episodes’ Source
Single
Multiple
All
Episodes*
Genitourinary tract Respiratory tract
5 2
1 4
6 (12) 6 (12)
85 (19) 80 (18)
Abscess Bowel Biliary tract Surgical wound Skin Bone Central nervous system Intravenous catheter Two or more possible sources Unknown
3 2 0
6 3 3
0 14
3 2
9 5 3 0 0 0 0 0 3 16
(17)+ (lo)+ (6)+
(6) (31)
32 20 10 18 16 2 5 14 27 137
(7) (4) (2) (4) (4) (0) (1) (3) (6) (31)
Number of episodes with percent in parentheses. + p <0.05. l
February 1984
The American Journal of Medicine
Volume 76
177
BREAKTHROUGHBACTEREMIA-WEINSTEIN and RELLER
TABLE IV
Characteristics and Outcome in Seven Patients with Multiple Episodes of Breakthrough Bacteremia Antimicrobial Therapy
Patient Number
Age/Sex
1
36/M
Corticosteroids, diabetes mellitus
1. Gentamicin 2. Ampicillin
2
52/M
Carcinoma of pancreas, diabetes mellitus, corticosteroids, surgery, granulocytopenia (third breakthrough episode only)
1. Chloramphenicol Gentamicin 2. Ampicillin Gentamicin
3
66/M
4
53/M
UnderlyingProblems
Hematologic malignancy, corticosteroids Liver transplant, diabetes mellitus, corticosteroids
46/M
Corticosteroids, renal failure
6
29/M
Liver transplant, corticosteroids
23/M
Hepatic cirrhosis, liver transplant, corticosteroids
1, Methicillin Gentamicin 2. Carbenicillin Cotrimoxazole
(6 days) 1. Staphylococcus epidermidis 2. Corynebacterium species, Micrococcus species 6
4. Cefazolin Chloramphenicol Gentamicin Cotrimoxazole 1. Carbenicillin 2. Carbenicillin Gentamicin 1. Cephalothin
Number of days after first breakthrough
1. Gentamicin 2. Carbenicillin Gentamicin 3. Gentamicin
February 1984
1. Bacteroides fragilis 2. Escherichia coli (6 days) 1. Escherichia coli
Source
Outcome
Respiratory tract Intra-abdominal abscess
Died
Intra-abdominal abscess lntra-abdominal abscess
Died
Respiratory tract
Respiratory tract Respiratory tract, intraabdominal abscess Unknown
Died
Died
Unknown
Genitourinary tract Respiratory tract
Bowel infarction Bowel infarction
Died
Biliary tract, surgical wound Biliary tract, surgical wound Bowel Liver abscess
Survived
Died
Liver abscess Liver abscess
episode.
Six patients with breakthrough bacteremia had been hospitalized for less than 48 hours at the time bacteremia occurred, and none died. In contrast, mortality was 70 percent in nosocomial breakthrough bacteremia. When breakthrough bacteremia occurred in patients receiving oral antimicrobial agents, mortality was lower (two of 11) than in patients who were receiving parenteral antimicrobials (30 of 41). Breakthrough bacteremia in patients receiving oral agents was more likely to be due to gram-positive cocci (four of 11; parenteral eight of 41) and was never due to Pseudomonas (parenteral seven of 41). Patients with breakthrough bacteremia
178
(6 days) 3. Candida parapsilosis (21 days) 4. Corynebacterium species (27 days)
2. Escherichia coli (21 days) 1. Escherichia coli 2. Escherichia coli (36 days) 3. Pseudomonas aeruginosa (46 days) 4. Escherichia coli, Bacteroides species (54 days)
4. Carbenicillin Gentamicin
l
2. Escherichia coli, enterococcus
1. Cefazolin 2. Gentamicin
2. Cefazolin 7
1. Escherichia coli 2. Proteus mirabilis, Peptostreptococcus species (3 days)’ 1. Escherichia coli
(7 days) 3. Escherichia coli, Pseudomonas aeruginosa, Citrobacter freundi (20 days) Klebsiella pneumoniae 2. Klebsiella pneumoniae
3. Ampicillin Gentamicin
3. Amphotericin
5
Microorganisms
The American Journal of Medicine
taking oral agents were also less likely to have been receiving corticosteroids (one of 11; parenteral 23 of 41) or to have a diagnosis of cirrhosis (zero; parenteral five of 41) or diabetes mellitus (one of 11; parenteral eight of 41). Indeed, patients with breakthrough bacteremia receiving oral agents were more likely to be free of all predisposing factors (four of 11) than were patients receiving parenteral therapy (two of 41). Overall mortality in breakthrough bacteremia was 81 percent, compared with 40 percent in non-breakthrough episodes. However, the proportion of septic (directly related) deaths was the same (20 percent). Thus, the
Volume 78
BREAKTHROUGHBACTEREMIA---WEINSTEIN and RELLER
difference in overall outcome between breakthrough and non-breakthrough episodes was due to a much greater likelihood of death not directly related to sepsis itself in breakthrough (41 percent) as compared with non-breakthrough (21 percent) episodes. COMMENTS
Breakthrough bacteremia accounted for 22 percent of gram-negative rod bacteremias at the University of California at Los Angeles Medical Center [l] and 10.4 percent of all septicemic episodes (9.5 percent of gram-negative rod bacteremias) at our hospitals. Nonetheless, little has been written about this phenomenon. Our findings suggest that breakthrough bacteremia is more likely to occur in patients whose host defenses are compromised, in particular those receiving immunosuppressive doses of corticosteroid therapy [2]. Liver transplant recipients also are at great risk. In general, these patients are among those with the most severe immunocompromise, and three such persons accounted for 10 episodes of breakthrough bacteremia in this series. Only 15 percent of patients with breakthrough bacteremia had none of the predisposing factors evaluated. Hematologic malignancies were no more common, and other neoplasms were less common (although not significantly) in patients with breakthrough bacteremia than in other bacteremic patients. Clearly, factors other than malignancy alone predispose to breakthrough. The observations in this study differ from the finding of Anderson et al [l] that leukopenia predisposes to breakthrough bacteremia. In our patients, neither leukopenia (leukocyte count below 4,OOO/pl) nor absolute neutropenia (granulocyte count below l,OOO/pl)predisposed to breakthrough compared with other bacteremias. The reasons for this difference are not clear. The underlying illnesses in the UCLA study [l] are not specified, and the patient populations may differ. Our definition of leukopenia required a lower absolute granulocyte count [3, g-151; therefore, we might have expected a stronger correlation between breakthrough and leukopenia than was observed in the UCLA series [ 11. Failure to find this correlation suggests that other factors are important in determining both the occurrence and outcome of breakthrough bacteremia. The increased frequency with which gram-negative rods caused breakthrough bacteremia (Table I) indirectly supports the findings by Anderson and colleagues [1] that subinhibitory levels of antimicrobials may play a role in the occurrence of breakthrough. Although serum antibiotic levels were not measured prospectively in our patients at the time of breakthrough, many patients with gram-negative bacteremia were receiving aminoglycosides, the serum levels of which have been found to
vary considerably after standard therapeutic doses 116-181. We believe that the occurrence of breakthrough bacteremia, therefore, constitutes a strong indication for the determination of serum antibiotic levels. Since suboptimal antibiotic levels may occur with greater frequency in patients receiving oral agents, the occurrence of breakthrough in this setting also constitutes an indication for measuring serum levels. Yet another factor associated with breakthrough bacteremia is a primary focus of infection that is difficult to treat. The increased frequency of intra-abdominal abscesses in this and the UCLA series [l] emphasizes this fact. Indeed, intra-abdominal foci were nearly three times more likely to be present in breakthrough compared with non-breakthrough bacteremia, and intraabdominal foci were the source of sepsis in three patients with multiple breakthrough episodes. Breakthrough bacteremia with gram-negative enteric rods, therefore, may provide an important clue to the presence of an intra-abdominal focus of infection, especially when multiple breakthrough episodes occur. The increased mortality associated with breakthrough bacteremia has been shown previously, but multivariate analysis suggested that breakthrough itself was relatively less important than other factors [2]. However, several factors associated with an increased risk of death in bacteremia are also found with greater frequency in breakthrough than in non-breakthrough episodes. These include gram-negative bacteria, hospital-acquired infection, and an intra-abdominal primary focus of infection (e.g., abscess) or an unknown source
[21.
The phenomenon of breakthrough bacteremia shows the potential limitations of antimicrobial therapy alone for sepsis, especially with single agents such as the aminoglycosides that have narrow therapeutic margins. Indeed, in such settings, therapeutic monitoring with antibiotic assays or serum bactericidal tests can provide evidence of adequate therapeutic margins [ 191, which may increase the likelihood of a successful outcome. Other forms of therapy may be required if the awesome associated mortality is to be reduced. For example, the need for surgical or catheter drainage of intra-abdominal abscesses is recognized [20]. Antiserum therapy has been shown recently to reduce mortality in gram-negative sepsis, but experience is limited [ 2 I]. Lastly, the role of blood culture methods that use resin removal of antibiotics to enable earlier detection of breakthrough bacteremia and improved therapy remains to be fully defined [ 221. ACKNOWLEDGMENT
We thank Drs. K. A. Lichtenstein and J. R. Murphy for their help in the initial design of this study, and Janet Gross for secretarial assistance.
February 1984
The American Journal of Medicine
Volume 78
179
BREAKTHROUGH
BACTEREMIA-WEINSTEIN
and RELLER
REFERENCES 1.
2.
3.
4.
5. 6. 7.
8. 9.
10.
11.
180
Anderson ET, Young LS, Hewitt WL: Simultaneous antibiotic levels in “breakthrough” gram-negative bacteremia. Am J Med 1976; 61: 493-497. Weinstein MP, Murphy JR, Reller LB, Lichtenstein KA: The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. II. Clinical observations with special reference to factors influencing prognosis. Rev Infect Dis 1983; 5: 54-70. Weinstein MP, Reller LB, Murphy JR, Lichtenstein KA: The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. I. Laboratory and epidemiologic observations. Rev Infect Dis 1983; 5: 35-53. Feinstein AR, Pritchett JA, Schimpff CR: The epidemiology of cancer therapy. Ill. The management of imperfect data. Arch Intern Med 1969; 123: 448-46 1. Feinstein AR, Pritchett JA, Schimpff CR: The epidemiology of cancer therapy. IV. The extraction of data from medical records. Arch Intern Med 1969; 123: 571-590. Nie NH, Hull CH, Jenkins JG, Steinbrenner K, Bent DH: Statistical package for the social sciences. 2nd ed. New York: McGraw-Hill, 1975. Medical Letter on Drugs and Therapeutics: Handbook of antimicrobial therapy, rev. ed. New Rochelle, New York: Medical Letter, Inc., 1976. O’Brien PC, Shampo MA: Comparing two proportions: the relative deviate test and chi-square equivalent. Mayo Clin Proc 1981; 56: 513-515. lssell BF, Keating MJ, Valdivieso M, Bodey GP: Continuous infusion tobramycin combined with carbenicillin for infections in cancer patients. Am J Med Sci 1979; 277: 311-318. Alavi BJ, Root RK, Djerassi I, et al: A randomized clinical trial of granulocyte transfusions for infection in acute leukemia. N Engl J Med 1977; 296: 706-711. Bodey GP, Buckley M, Sathe YS: Quantitative relationships between circulating leukocytes and infection in patients with acute leukemia. Ann Intern Med 1966; 64: 328340.
February 1984
The American Journal of Medicine
12. 13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Volume 78
Goodall PT, Vosti KL: Fever in acute myelogenous leukemia. Arch Intern Med 1975; 135: 1197-1203. Gurwith MJ, Brunton JL, Lank BA, Ronald AR, Harding GKM: Granulocytopenia in hospitalized patients. 1. Prognostic factors and etiology of fever. Am J Med 1978; 674: 121-126. Herzig RH, Herzig GP, Graw RG Jr, Bull MI, Ray KK: Successful granulocyte transfusion therapy for gram-negative septicemia: a prospective randomized controlled study. N Engl J Med 1977; 296: 701-705. Sickle EA, Greene WH, Wiernik PH: Clinical presentation of infection in granulocytopenic patients. Arch Intern Med 1975; 135: 715-719. Jackson G. Riff LJ: Pseudomonas bacteremia: pharmacologic and other bases for failure of treatment with gentamicin. J Infect Dis 1971; 124(suppl): S185-S191. Winters RE, Litwack KD, Hewitt WL: Relation between dose and levels of gentamicin in blood. J Infect Dis 1971; 124(suppl): S90-S95. Kaye D, Levison MB, Labovitz ED: The unpredictability of serum concentrations of gentamicin. Pharmacokinetics of gentamicin in patients with normal and abnormal renal function. J Infect Dis 1974; 130: 150-154. Stratton CW, Weinstein MP, Reller LB: Correlation of serum bactericidal activity with antimicrobial agent level and minimum bactericidal concentration. J Infect Dis 1982; 145: 160-168. Gerzof SG, Robbins AH, Johnson WC, Birkett DH, Nabseth DC: Percutaneous catheter drainage of abdominal abscesses: a five-year experience. N Engl J Med 1981; 305: 653657. Ziegler EJ, McCutchan AJ, Fierer J, et al: Treatment of gram-negative bacteremia and shock wtth human antiserum to a mutant Escherichia coli. N Engl J Med 1982; 307: 12251230. Appelbaum PC, Beckwith DG, Dipersio JR, Dyke JW, Salventi JF, Stone LL: Enhanced detection of bacteremia with a new BACTEC resin blood culture medium. J Clin Microbial 1983; 17: 48-51.