View
2
Download
0
Category
Preview:
Citation preview
Prosthetic Valve Endocarditis Resulting from Nosocomial Bacteremia A Prospective, Multicenter Study
Guodong Fang, MD; Thomas F. Keys, MD; Layne O. Gentry, MD; Alan A. Harris, MD; Nilda Rivera, RN; Karen Getz, RN; Peter C. Fuchs, MD; Marie Gustafson, RN; Edward S. Wong, MD; Angella Goetz, RN, MNEd; Marilyn M. Wagener, MPH; and Victor L. Yu, MD
• Objective: To determine the incidence of endocarditis in bacteremic patients with prosthetic heart valves and the risk factors for and the effect of duration of antibiotic therapy on development of endocarditis in such patients. • Design: Multicenter, prospective observational study. • Setting: Six university teaching hospitals with high-volume cardiothoracic surgery. • Participants: One hundred seventy-one consecutive patients with prosthetic heart valves who developed bacteremia during hospitalization. • Measurements and Main Results: Patients were evaluated when they were identified as having bacteremia and 1,2,6, and 12 months after its occurrence. Of 171 patients, 74 (43%) developed endocarditis: Fifty-six (33%) had prosthetic valve endocarditis at the time bacteremia was discovered ("endocarditis at outset"), whereas 18(11%) developed endocarditis a mean of 45 days after bacteremia was discovered ("new endocarditis"). Mitral valve location and staphylococcal bacteremia (Staphylococcus aureus or S. epidermidis) were significantly associated with the development of "new" endocarditis. All 18 cases of new endocarditis were nosocomial, and in 6 of these cases (33%) bacteremia was acquired via intravascular devices. Twenty-one patients without evidence of endocarditis at the time of bacteremia received short-term antibiotic therapy (<14 days); 1 patient (5%) developed endocarditis. Eleven of 70 patients (16%) who received long-term antibiotic therapy (>14 days) developed endocarditis (P > 0.2). • Conclusions: Bacteremic patients with prosthetic heart valves were at notable risk for developing endocarditis, even when they received antibiotic therapy before endocarditis developed and regardless of the duration of such therapy. Intravascular devices were a common portal of entry.
M ore than 100 000 artificial heart valves are implanted annually in the United States, and the number is increasing (1). Because the number of invasive therapeutic and diagnostic procedures done during hospitalization is also increasing, the risk for bacteremia is correspondingly greater.
A common dilemma for the clinician is determining the optimal therapeutic approach to a patient with a prosthetic valve who develops bacteremia. There has been controversy regarding the optimal duration of antibiotic therapy for bacteremic patients with prosthetic valves. An informal survey of infectious disease and cardiology specialists revealed that the recommended duration of therapy ranged widely, from 10 days to 8 weeks.
To clarify this issue, we conducted a prospective, observational, multicenter study during a 3-year period in six university teaching hospitals with high-volume cardiac surgery. Our objectives were to determine 1) the incidence of endocarditis in bacteremic patients with an implanted heart valve, 2) the risk factors for development of subsequent prosthetic valve endocarditis, and 3) the optimal duration of antibiotic therapy in this select group of patients who experience bacteremia. We also report mortality rates.
Methods
We did a prospective, observational study at six university teaching hospitals: The Cleveland Clinic, Cleveland, Ohio; St. Luke's Episcopal Hospital, Houston, Texas; Rush-Presbyteri-an-St. Luke's Medical Center, Chicago, Illinois; St. Vincent Hospital and Medical Center, Portland, Oregon; Hunter-McGuire Veterans Affairs Medical Center, Richmond, Virginia; and Presbyterian University Hospital and Veterans Affairs Medical Center, Pittsburgh, Pennsylvania. The study lasted approximately 2 years at each hospital and was conducted in the period from 1986 to 1989; start dates differed among hospitals. Bacteremic patients were identified through a daily review of blood culture results in the microbiology laboratory; these patients were then screened by the infection control practitioner for presence of a prosthetic heart valve. The study group comprised bacteremic patients with a prosthetic valve who were prospectively monitored during the hospitalization and after discharge for 1 year. All treatment decisions were made by the attending physicians without intervention by the investigators.
Bacteremia was considered to be present if one or more blood cultures yielded an organism. Endocarditis was classified as "definitive," "presumptive," and "suspicious." Definitive endocarditis was defined by 1) culture positivity or proven pathology (demonstration of vegetations or isolation of the organism from a prosthetic valve obtained at autopsy or during
Ann Intern Med. 1993;119:560-567.
From the University of Pittsburgh and Veterans Affairs Medical Center, Pittsburgh, Pennsylvania; The Cleveland Clinic, Cleveland, Ohio; Baylor University, Houston, Texas; Rush Presbyterian-St. Luke's Medical Center, Chicago, Illinois; St. Vincent Hospital and Medical Center, Portland, Oregon; Hunt-er-McGuire Veterans Affairs Medical Center, Richmond, Virginia. For current author addresses, see end of text.
560 1 October 1993 • Annals of Internal Medicine • Volume 119 • Number 7 (Part 1)
Downloaded From: http://annals.org/ by a University of Pittsburgh User on 05/25/2013
Table 1. Criteria for Diagnosis of Endocarditis
New endocarditis (n = 18)* No evidence of endocarditis (see definition in Methods) at
the time of positive blood culture An interval of 1 week or more between a positive blood
culture and appearance of endocarditis Nosocomial bacteremia or a defined portal of entry for the
blood-borne organism Endocarditis (n = 56) at the outsetf
Evidence for endocarditis was present at the time of the first positive blood culture, or "suspicious" evidence of endocarditis developed shortly after blood cultures were obtained
* Development of endocarditis after bacteremia. t The positive blood culture was the initial manifestation of endocar
ditis.
open heart surgery) or 2) vegetations documented on echocardiogram. Presumptive endocarditis was defined by peripheral embolic signs, including Osier nodes, Janeway lesions, Roth spot, petechiae, splenomegaly, acute cerebrovascular accident; or a new or changing murmur. The criterion for suspicious endocarditis was sustained bacteremia (defined by three positive blood cultures of three or the occurrence of at least four positive blood cultures within 72 hours) with no identified portal of entry. Thus, four positive cultures obtained within 72 hours fulfilled the criterion for "suspicious" endocarditis, whereas two positive of two or three positive of four blood cultures obtained within 72 hours did not.
Demonstration of vegetations by transthoracic echocardiography in a bacteremic patient with a prosthetic valve was arbitrarily considered as definitive evidence for endocarditis; transesophageal echocardiography was not routinely done at the time of study.
Patients were classified as having bacteremia only, as having bacteremia with endocarditis at the outset, or as having bacteremia with subsequent development of new endocarditis (Table 1). Patients who were identified as having endocarditis at the time the initial blood culture was obtained were classified as having endocarditis at the outset. Patients in whom bacteremia led to the subsequent development of prosthetic valve endocarditis were classified as having new endocarditis.
The portal of entry for bacteremic infection was defined by
isolation of the same organism from a source other than blood or cardiac vegetation and by review of the patient record by the investigators. Subtyping of the isolates was not done.
We assessed severity of illness for each patient at the time of bacteremia using an index based on mental status, vital signs, the need for respiratory support, and the occurrence of cardiac arrest; previous investigators have found this index to be highly predictive of outcome (2-4).
All bacteremic patients with prosthetic valves were prospectively studied for up to 1 year after the onset of bacteremia. Outcome and cardiac status were assessed 1, 2, 6, and 12 months after the detection of bacteremia. Discharged patients had telephone follow-up on chart review at 6 and 12 months.
Clinical and laboratory data for analysis were entered into a computer database (Prophet Systems, Division of Research Resources, National Institutes of Health, Bethesda, Maryland). Categorical data were analyzed using a chi-square or Fisher exact test. Continuous variables (age, days hospitalized, and laboratory values) were compared using the Mest or Newman-Keuls test. A multiple regression model was used to examine the effects of several factors on the development of endocarditis. The regression model was also used to evaluate the risk factors for death. Factors used in the regression model were those found to be significant by univariate analysis and those hypothesized to seriously affect outcome.
Results
One hundred seventy-one patients with bacteremia and a prosthetic heart valve constituted the study group. The mean number of positive blood cultures per patient was 3.0 and the median number was 3.6 (range, 1 to 14 blood cultures). All patients had fever, which was usually the motivating factor for obtaining the blood culture. At 6 months, follow-up data were complete for all 171 study patients. By 12 months, 4 patients with bacteremia had been lost to follow-up, but follow-up data were obtained on all 74 patients with endocarditis. Patient age ranged from 23 to 89 years (mean age, 65 years). Bacteremic patients were significantly older (mean age, 69 years) than those with endocarditis (mean age, 60 years) (P < 0.001). Other clin-
Table 2. Demographic and Clinical Characteristics of 171 Patients with Bacteremia Who Had Prosthetic Valves
Variable Bacteremia Only New Endocarditis Endocarditis at the Outset P Value* (n = 97) (n = 18) (n = 56)
Sex, n (%) Male 59 (61) 9(50) 42 (75) 0.09 Female 38 (39) 9(50) 14 (25)
Immunosuppressive therapy, n (%) 8(8) 4(22) 4(7) 0.14 Malignancy, n (%) 12 (12) 1(5) 3(6) >0.2 Severity of illness, n (%)
+ 1 31 (32) 3(17) 37 (66) <0.001 +2 21 (21) 9(50) 12 (21) 0.03 +3 17 (17) 4(20) 4(7) 0.14 > + 4 28 (29) 2(11) 3(5) 0.001
Antibiotic duration, n/n (%)"$ <2 weeks 20/79 (25) 1/16 (6) 0/43 (0) 0.001 >2 weeks 59/79 (75) 15/16 (94) 43/43 (100)
Endocarditis onset, n {%)% Early NA§ 12 (67) 11 (20) <0.001 Late NA 6(33) 45 (80)
Cause, n (%) Polymicrobial 20 (21) 3(17) 4(7) 0.06 Monomicrobial 77 (79) 15 (83) 52 (93)
* By chi-square test (two-tailed). t Patients who die within 2 weeks of detection of bacteremia were omitted from the analysis. $ Endocarditis was classified as having an early onset when it occurred within 60 days of valve insertion and as having a late onset when it occurred
more than 60 days after valve insertion. § NA = not applicable.
1 October 1993 • Annals of Internal Medicine • Volume 119 • Number 7 (Part 1) 561
Downloaded From: http://annals.org/ by a University of Pittsburgh User on 05/25/2013
ical and demographic characteristics are shown in Table 2.
Endocarditis
Seventy-four of 171 patients (43%) were classified as having prosthetic valve endocarditis. Of these 74 patients, 42 (57%) were classified as having definitive endocarditis, 22 (30%) as having presumptive endocarditis, and 10 (13%) as having suspicious endocarditis.
Of the 171 study patients, 56 (33%) were classified as having endocarditis at the time bacteremia occurred (endocarditis at the outset) and 18 (10%) as having endocarditis that developed after bacteremia (new endocarditis). The remaining 97 patients (57%) with bacteremia had no evidence of endocarditis (see Table 2).
Twenty-three patients had vegetations detectable by transthoracic echocardiography: Seven patients were classified as having "definitive" prosthetic valve endocarditis based solely on echocardiographic demonstration of vegetations; another 6 patients had abnormal echocardiographic findings but also showed other objective evidence of endocarditis (including new or changing murmur and peripheral stigmata); and 10 patients with abnormal echocardiographic findings had the diagnosis confirmed at open heart surgery or at autopsy. Transesophageal echocardiography was not routinely used at the time of our study.
Of the 74 patients with prosthetic valve endocarditis, 2 (3%) had peripheral embolism (cerebral emboli or cerebrovascular accident), 15 (20%) had stigmata (Osier nodes, Janeway lesions, Roth spot, petechiae), 4 (5%) had splenomegaly, and 23 (31%) had new or changing murmur. No significant difference was found regarding the presence of these signs between patients with new endocarditis and those with endocarditis at the outset (data not shown).
Location and Type of Valvular Prosthesis
Eighty-seven of the 171 study patients (51%) had prosthetic mitral valves, 102 (60%) had prosthetic aortic valves, and 8 (5%) had prosthetic tricuspid valves (the number of patients totals more than 171 because 25 patients (15%) had more than one prosthetic valve)
(Table 3). Presence of a prosthetic mitral valve was seen significantly more often in patients with new endocarditis (P = 0.005), whereas prosthetic aortic valves were seen significantly more often in patients with endocarditis at outset (P < 0.001) (Table 3). Of the 171 study patients, 82 (48%) had mechanical valves and 87 (51%) had biological valves (including 5 patients with both mechanical and biologic valves). In seven cases (4%), the exact valve type was not known because the operation had been done elsewhere and past medical records were not available (see Table 3). No association was observed between type of valve and the occurrence of endocarditis.
Timing of Endocarditis
Prosthetic valve endocarditis has been arbitrarily classified as "early" when it occurs within 60 days of valve insertion and " la te" when it occurs more than 60 days after valve insertion (5). Of the 74 patients with prosthetic valve endocarditis, 23 (31%) were classified as having early endocarditis and 51 (69%) as having " la te" endocarditis. Twelve of the 18 patients (67%) with new endocarditis were classified as having "early onset" disease compared with 11 of 56 patients with endocarditis at the outset (difference, 48%; 95% CI, 19 to 75; P < 0.001).
Cause of Bacteremia
As shown in Table 4, 206 organisms were isolated from the blood of 171 patients. Staphylococci classified as coagulase-negative were considered S. epidermidis for the purposes of our study. Staphylococcus epidermidis was the most common cause of bacteremia among all study patients (43 of 171 [25%]) and among patients with endocarditis (30 of 74 [41%]) (Table 4). The most common causes of bacteremia among patients with new endocarditis were Staphylococcus epidermidis (6 of 18 [33%]) and S. aureus (4 of 18 [22%]), whereas the most common cause among patients with endocarditis at the outset was S, epidermidis (24 of 56 [42%]). Twenty-seven of 171 patients (16%) had more than one isolate.
Table 3. Valvular Location and Type in 171 Bacteremia Patients with Prosthetic Valves
Variable Bacteremia Only New Endocarditis Endocarditis at Outset Total (n = 97) (n = 18) (n = 56) (n = 171)*
M / ' < n (/u/
Valve location Mitral 50 (52) 15 (83) 22 (39) 87 (51) Aortic 57 (59) 3(17) 42 (75) 102 (60) Tricuspid 3(3) 3(17) 2(4) 8(5)
Valve type Mechanical 51 (53) 9(50) 22 (39) 82 (48) Biological 46 (47) 9(50) 32 (57) 87 (51) Unknown 4(4) 0(0) 3(5) 7(4)
* One hundred seventy-one patients had 197 valves. Twenty-four patients had two prosthetic valves, and one patient had three prosthetic valves. Among the 25 patients with multiple valves, 12 had bacteremia only, 3 had new endocarditis, and 10 had endocarditis at outset.
t Twenty-five patients had multiple valves: Eleven patients had two mechanical valves, 9 had two biologic valves, 4 had one mechanical valve and one biologic valve, 1 had one mechanical valve and two biologic valves. Eighty-two patients had 93 mechanical valves and 87 patients had 97 biologic valves.
562 1 October 1993 • Annals of Internal Medicine • Volume 119 • Number 7 (Part 1)
Downloaded From: http://annals.org/ by a University of Pittsburgh User on 05/25/2013
Table 4. Cause of Bacteremia in 171 Patients with Prosthetic Valves
Organism Bacteremia New Endocarditis Endocarditis at Outset Total
Staphylococcus epidermidis 13 6 24 43 S. aureus 19 4 11 34 Enterococcus faecalis 18 3 4 25 Viridans streptococci 2 0 5 7 Other streptococcal species 6 2 8 16
Streptococcus bovis 1 2 S. sanguis 1 S. sanguis 5. acidominimus 1 S. mitis 1 S. anginosus-constellatus 1 Group D streptococci (nonenterococcal) 1 4 S. pneumoniae 1 S. salivarius 1 1 Gamma-streptococcus (not group D) 1
Micrococcus species 0 0 1 1 Pseudomonas aeruginosa 11 2 0 13 Escherichia coli 9 0 1 10
8 1 0 9 Klebsiella pneumoniae 7 0 0 7 Serratia marcescens 4 0 0 4 Proteus mirabilis 1 2 0 3 Enterobacter aerogenes 3 0 0 3 Citrobacter freundii 2 0 0 2 K. oxytoca 1 1 0 2 Other gram-negative rods 6 0 1 7
Acinetobacter calcoaceticus subsp. anitratus 1 Acinetobacter calcoaceticus subsp. Iwoffi 1 Bacteroides fragilis 1 Morganella morganii 1 Pseudomonas cepacia 1 Pseudomonas diminuta 1 Salmonella species 1
Candida albicans 6 2 3 11 C. neoformans 1 0 0 1 Aspergillus niger 1 0 0 1 Torulopsis glabrata 1 0 0 1 Miscellaneous 4 0 2 6
Corynebacterium jeikeium (JK diphtheroids) 2 Actinobacillus actinomycetemcomitans 1 Listeria monocytogenes 1 Clostridium perfringens 1 Propionibacterium acnes 1
Total 123 23 60 206*
* Twenty-seven patients had polymicrobial bacteremia with 2 to 4 organisms isolated.
One patient had four organisms, 6 patients had three organisms, and 21 patients had two organisms. In cases of polymicrobial bacteremia, the most common organisms were S. epidermidis (10 cases), Enterococcus faecalis (8 cases), Candida species (7 cases), S. aureus (6 cases), Pseudomonas aeruginosa (5 cases), Enterobacter species (5 cases), and Klebsiella pneumoniae (4 cases). The rate of polymicrobial infection by patient group is shown in Table 2.
Portal of Entry
The most common portals of entry for bacteremia that led to the development of new endocarditis were intravascular catheters and wounds and skin infections (Table 5). The portal of entry in most of the cases of endocarditis at the outset was unknown.
Table 5. Portal of Entry of Bacteremia in 115 Patients with Prosthetic Heart Valves*
Portal of Entry Bacteremia New Total Only Endocarditis
Intravascular cathetert 26 6 32 Wound or skin 13 5 18
infection Urinary tract infection 12 0 12 Pneumonia 9 0 9 Disseminated 5 2 7 Abdominal or biliary 4 0 4
infection Miscellaneous^ 2 1 3 Unknown 26 4 30
Total 97 18 115
* Fifty-six patients with prosthetic valve endocarditis at the outset were excluded from this analysis.
t One patient had pacemaker-associated bacteremia. X Oropharynx infection (one patient); meningitis (one patient); disc
space infection (one patient).
1 October 1993 • Annals of Internal Medicine • Volume 119 • Number 7 (Part 1) 563
Downloaded From: http://annals.org/ by a University of Pittsburgh User on 05/25/2013
Table 6. Eighteen Patients with New Endocarditis Caused by Nosocomial Bacteremia
Patient Age Sex Organism Frequency of Positive Blood
Culture
Portal of Entry Days between Valve Insertion and
Positive Blood Culture
y n/n
1 50 M Staphylococcus epidermidis 1/10 Wound 4 2 75 F Candida albicans 4/4 Intravascular line 10 3 67 M Methicillin-resistant S. aureus
Pseudomonas aeruginosa 4/4 Disseminated 14
4 58 M S. epidermidis 3/8 Intravascular line 2 5 54 F P. aeruginosa; two streptococci 2/4 Intravascular line 10 6 65 M S. aureus 6/12 Wound 15 7 72 M S. aureus
E. cloacae 4/4 Wound 5
8 63 F S. epidermidis 7/12 Intravascular line 12 9 51 M S. faecalis 4/4 Vertebral osteomyelitis 2246
10 72 F S. faecalis 3/3 Wound 14 11 61 F Proteus mirabilis 2/2 Unknown 53 12 57 M S. aureus 2/4 Intravascular line 163 13 76 F P. mirabilis 3/3 Wound 75 14 65 F S. epidermidis 2/2 Unknown 54 15 75 M S. epidermidis 5/6 Unknown 41
16 72 F C. albicans 3/3 Unknown 2541 17 76 F S. epidermidis
S. faecalis Klebsiella oxytoca
1/1 Disseminated 10
18 47 M S. aureus 6/6 Intravascular line 6453
* Long-term = more than 2 weeks; short-term = 2 weeks or less.
Outcome
The mortality rates among patients with endocarditis were 41% (30 of 74), 46% (34 of 74), and 47% (35 of 74) at 2, 6, and 12 months, respectively. The mortality rate at 6 months among patients with new endocarditis (67%, 12 of 18) was higher than that of patients with endocarditis at the outset (39%, 22 of 56) (P < 0.05, chi-square). The mortality rate at 12 months for the patients with bacteremia only was 58% (54 of 93 [four patients lost to follow-up]). Eighty-nine percent of the deaths (31 of 35) among patients with endocarditis were attributed to infection, whereas 50% of the deaths (27 of 54) among patients with bacteremia only were attributed to infection. Mortality rates at 6 and 12 months were significantly higher among patients with early-onset endocarditis (65%, 15 of 23) than among patients with late-onset endocarditis (37%, 19 of 51) (difference, 28%; CI, 1 to 55; P < 0.04).
Outcomes at 6 months in 115 patients without endocarditis at outset (97 cases of bacteremia only and 18 cases of "new" endocarditis) were assessed using stepwise multiple regression. Factors included in the model were severity of illness (> +4 versus < +4), polymicrobial infection, development of endocarditis, type of valve (mechanical versus biological), and location of valve (mitral versus not mitral). The following variables proved to be significant: a severity of illness score greater than or equal to -1-4 (odds ratio, 1.6; CI, 1.4 to 2.0, P = 0.001); polymicrobial infection (odds ratio, 1.3; CI, 1.1 to 1.6; P = 0.01); and development of "new" endocarditis (odds ratio, 1.3; CI, 1.03 to 1.62; P = 0.03).
All patients received antimicrobial therapy. Of the 74
patients with endocarditis, 22 (30%) also underwent surgery to replace the infected artificial valve. Surgical replacement of the infected valve was significantly associated with improved outcome (mortality rate, 23%) when compared with medical therapy only (mortality rate, 56%) (P = 0.009). Details of this analysis will be published in the future.
Development of New Endocarditis
We assessed the risk for development of endocarditis in patients with a prosthetic valve who had bacteremia. The 56 patients who were classified as having endocarditis at the outset were excluded from the analysis because the first positive blood culture was indicative of endocarditis, not bacteremia, leaving 115 patients for analysis. Of these 115 patients, 18 met the predefined criteria for development of new endocarditis and 97 remained free of valvular infection (Table 6). Of the 18 cases of new endocarditis, 13 (72%) were classified as definitive, 4 (22%) as presumptive, and 1 (6%) as suspicious (Table 6). The mean and median intervals between documentation of bacteremia and subsequent development of endocarditis were 45 days and 28 days, respectively (range, 7 to 170 days).
Streptococcal species (excluding enterococci) were a common cause of endocarditis at outset but were rarely a cause of new endocarditis (one case of polymicrobial endocarditis). Staphylococcus infection (either S. aureus or S. epidermidis) occurred more frequently in patients developing new endocarditis (56% [10 of 18]) than in patients with bacteremia only (32% [31 of 97], P = 0.07).
564 1 October 1993 • Annals of Internal Medicine • Volume 119 • Number 7 (Part 1)
Downloaded From: http://annals.org/ by a University of Pittsburgh User on 05/25/2013
Table 6. (Cont'd)
Days between Duration of Valve Diagnosis of Endocarditis Outcome during Positive Blood Antibiotic Replacement Endocarditis Classification! Hospitalization Culture and Therapy* Endocarditis
30 Short-term No Findings at autopsy Definitive Died 7 Long-term No Echocardiography abnormality, emboli Definitive Died
28 Long-term No Findings at autopsy plus murmur Definitive Died
18 Long-term No Findings at autopsy Definitive Died 35 Long-term No Findings at autopsy plus murmur Definitive Died 69 Long-term Yes Valvular vegetation Definitive Survived 14 Long-term No Autopsy Definitive Died
165 Long-term No Echocardiographic abnormality, murmur Definitive Survived 170 Long-term No Splenomegaly Presumptive Survived 72 Long-term No Echocardiographic abnormality Definitive Died 94 Long-term No Valvular vegetation Definitive Died 7 Long-term No Echocardiographic abnormality, stigmata Definitive Survived 7 Long-term No Petechiae Presumptive Survived
28 Long-term No Findings at autopsy Definitive Died 33 Long-term No 6 positive of 7 blood cultures and multiple
positive blood cultures after therapy Suspicious Died
22 Long-term No Findings at autopsy Definitive Died 7 Long-term No Stigmata Presumptive Died
8 Long-term No Stigmata Presumptive Survived
The presence of a prosthetic valve at the mitral location (83% [15 of 18]) was significantly associated with development of new endocarditis (P = 0.005) (see Table 3). No association was found between valve type or presence of multiple prosthetic valves and the development of new endocarditis.
In none of the 18 patients with new endocarditis did the urinary tract serve as the portal of entry. In contrast, 12 of the 97 cases of bacteremia without endocarditis (12%) originated in the urinary tract (see Table 5); this difference was not statistically significant.
In assessing the 115 patients without endocarditis at the outset (97 patients with bacteremia only plus 18 cases of new endocarditis), factors thought to contribute to the development of new endocarditis were assessed in a stepwise multiple regression model. These factors included valve location, valve type, staphylococcal cause, portal of entry, and time to bacteremia after valve insertion (within 60 days of valve replacement versus more than 60 days after valve insertion). The following factors proved to be significant: mitral valve location (odds ratio, 1.2; CI, 1.1 to 1.4; P = 0.01) and staphylococcal cause (odds ratio, 1.15; CI, 1.0 to 1.3; P = 0.05).
Duration of Antibiotic Therapy
We evaluated duration of antibiotic therapy and its effect in preventing endocarditis. Fifty-six patients with endocarditis at the outset and 18 patients who died within 2 weeks after the onset of bacteremia were excluded. Endocarditis was diagnosed in six patients within 2 weeks of initiating antibiotic therapy; because these six patients were automatically given long-term therapy, they were also excluded from analysis. Data on the remaining 91 patients are summarized in Table 7.
Of the 21 patients receiving short-term antibiotic therapy (<2 weeks), 1 (5%) developed endocarditis. Of the 70 patients receiving long-term therapy (>2 weeks), 11 (16%) developed endocarditis (P > 0.2). Thus, no association was found between duration of antibiotic therapy and subsequent development of endocarditis. The five patients who developed new S. aureus endocarditis received at least 4 weeks of antibiotic therapy (see Table 6).
Discussion
We enrolled 171 consecutive patients with a prosthetic heart valve who developed bacteremia. These patients were followed prospectively for 1 year and monitored closely for development of endocarditis. Our study differed from other studies of prosthetic valve endocarditis in that it was large and prospective and maintained follow-up for 1 year in most patients.
Surprisingly, 74 of 171 patients (43%) with bacteremia
1 October 1993 • Annals of Internal Medicine • Volume 119 • Number 7 (Part 1)
Table 7. Duration of Antibiotic Therapy in 91 Evaluable Patients with Bacteremia Who Had a Prosthetic Valve
Patient Group Short-term Long-term Total Therapyt Therapyt
Bacteremia only, n 20 59 79 New endocarditis, n 1 11 12
Total 21 70 91
* The following patients were excluded from the analysis: 56 patients with endocarditis at outset ; 18 patients who died within 2 weeks after the onset of bacteremia; and 6 patients with I lew endocarditis who were diagnosed before 2 weeks of therapy was cc >mpleted. Duration of anti-biotic therapy was found i not to be a determining factor in the devel-opment of endocarditis.
t Short-term = 14 days or less. t Long-term = more than 14 days.
565
Downloaded From: http://annals.org/ by a University of Pittsburgh User on 05/25/2013
developed prosthetic valve endocarditis, either at the time of bacteremia (56 patients) or later (18 patients). This rate of endocarditis may seem high when compared with the overall incidence of 1% to 4.4% among patients who undergo prosthetic valve replacement (6), but Sande and colleagues (7) found prosthetic valve endocarditis in 11 of 24 (46%) hospitalized patients with a preexisting prosthetic valve and sustained bacteremia. Alvarez-Elcoro and colleagues (8) observed an endocarditis rate of 89% in a study of nine patients with prosthetic heart valves and community-acquired bacteremia.
Our most important finding was that 18 of the 115 patients (16%) with prosthetic valves and bacteremia (excluding the 56 patients with endocarditis at the outset) developed subsequent prosthetic valve endocarditis. To ensure that these 18 patients had developed "new" endocarditis and had not actually manifested bacteremia at an early stage of endocarditis, we formulated strict criteria to distinguish "new" and "outset" endocarditis {see Tables 1 and 6). All cases of new endocarditis proved to be nosocomial, with a mean duration of 45 days (range, 7 to 170 days) between detection of bacteremia and the diagnosis of prosthetic valve endocarditis. The most common portals of infection were intravascular catheters (6 of 18 cases [33%]) and skin infections and wounds (5 of 18 cases [28%]) {see Table 6). A mitral prosthesis and staphylococcal bacteremia {S. aureus or S. epidermidis) were significant risk factors for subsequent development of endocarditis. Twelve of 18 cases (67%) were classified as early-onset endocarditis (that is, they occurred within 60 days of valve insertion). Two patients had had bacteremia at a previous hospital admission. Both were re-admitted 1 and 5 months later, respectively, with multiple blood cultures showing the same organism that had caused the initial episode of bacteremia. In two other studies of "nosocomial" prosthetic valve endocarditis (7, 9), all cases were caused by gram-positive cocci. In one of the studies, Sande and colleagues (7) concluded that gram-negative bacilli from a suspected source could be treated with a short course of antibiotics (10 to 14 days) combined with local interventions directed at the suspected source. Both studies were limited in that they were retrospective, follow-up was restricted to the period of hospitalization, and criteria for excluding endocarditis were weak (echocardiography, for example, was not available).
We found that 8 of 18 cases (44%) of new endocarditis were caused by fungi or gram-negative bacilli; 6 of these 8 cases were classified as "definitive": In four cases, patients had vegetations at autopsy, and in two cases, patients had vegetations on echocardiograms plus peripheral stigmata {see Table 6). The previous suggestion that endocarditis would be unlikely if the cause of bacteremia were a gram-negative bacillus or if a portal of entry could be established (7) was not supported by our findings.
We confirmed that S. epidermidis was the most common causative organism among both patients with bacteremia and those with endocarditis {see Table 4) (10-15). Staphylococcal bacteremia {S. epidermidis and S. aureus) was most commonly associated with intravascular lines and skin infections and wounds.
We originally hypothesized that the duration of antibiotic therapy would be a factor in preventing the development of endocarditis in bacteremic patients with prosthetic valves. However, this proved not to be the case; of the 18 patients who developed new endocarditis because of nosocomial bacteremia, 6 proved to be non-evaluable because of early death or because early diagnosis favored the institution of long-term therapy; 11 of the remaining 12 patients had received antibiotic therapy for more than 2 weeks (Table 7).
Our case definition for endocarditis may have been overly strict, such that a few patients classified as having bacteremia may actually have had endocarditis. Von Reyn and colleagues (16) introduced criteria for endocarditis that they also considered strict. Had those criteria been applied in our study, however, all 171 patients in our study would have been classified as having endocarditis. Von Reyn and colleagues' criteria differed from ours in that patients were classified as having "probable" endocarditis if two of two blood cultures were positive and a prosthetic valve was present; we required that at least three of three blood cultures be positive, with such cases considered only as "suspicious." In our study, patients who had three of three blood cultures that were positive or a greater number of positive blood cultures (one patient had six consecutive positive blood cultures) with a defined portal of entry were classified as bacteremic if no other evidence for endocarditis could be found. Von Reyn and colleagues (16) did not use the concept of portal of entry in their criteria.
Some of the patients classified as having "new endocarditis" might actually have had "endocarditis at the outset," particularly four patients with S. epidermidis bacteremia who had extensive but unrevealing workups for endocarditis at the time of bacteremia {see Table 6). Nevertheless, clinical criteria coupled with cardiac and laboratory studies suggested that these patients developed endocarditis after bacteremia from a defined infection. Definitive evaluation would have required microscopic examination of the prosthetic valve, which was not feasible.
Although we used an objective severity of illness index that was applied prospectively, the ideal index would have included a rigorous assessment of cardiac status at the time of positive blood culture. Nevertheless, all previous series of prosthetic valve endocarditis have had this shortcoming.
Because we did not evaluate hospitalized patients with prosthetic valve who did not have bacteremia, we were unable to assess the overall risk for development of new endocarditis.
Bacteremia occurring in a patient with a preexisting prosthetic valve is an ominous sign. An evaluation should be done for existing endocarditis, but physicians should also monitor patients for development of subsequent endocarditis. Duration of antimicrobial therapy was not a key factor in prevention of endocarditis in bacteremic patients with prosthetic valves; this finding was disappointing because duration of therapy is amenable to manipulation. Nevertheless, given the high incidence of endocarditis in our patients with prosthetic
566 1 October 1993 • Annals of Internal Medicine • Volume 119 • Number 7 (Part 1)
Downloaded From: http://annals.org/ by a University of Pittsburgh User on 05/25/2013
valves, long-term antibiotic therapy must still be considered the prudent course.
Prevention may be the factor most amenable to intervention by physicians. Because intravascular catheters were the source of bacteremia in 33% of the patients with "new" endocarditis (see Table 6), such invasive procedures must be approached with extreme caution in patients with prosthetic valves. The single most common portal of infection for the bacteremic patients was also the intravascular catheter (see Table 5). Intravascular catheters, if used, should be inserted using meticulously sterile technique, should remain in place for as brief a duration as possible, and should be discontinued at the earliest sign of phlebitis or inflammation. The issue of prophylactic antibiotic therapy for patients with prosthetic valves who undergo intravascular catheter placement has not been addressed in detail (15, 17-19).
Acknowledgments: The authors thank Rebecca A. Barto, PhD, Michael Scheld, MD, Donald Grandis, MD, Randall Starling, MD, L. Douglas Cowgill, MD, David Durack, MD, and Stephen B. Calderwood, MD, for their critical review of the manuscript.
Requests for Reprints: Victor L. Yu, MD, Division of Infectious Diseases, University of Pittsburgh Medical Center, Montefiore Building W931, Pittsburgh, PA 15213.
Current Author Addresses: Dr. Fang: Box 485, Division of Geographic Medicine, Health Science Center, University of Virginia, Charlottesville, VA 22908. Dr. Keys: Department of Infectious Diseases, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44106. Dr. Gentry and Ms. Getz: Infectious Disease Section, St. Luke's Episcopal Hospital, 6720 Bertner Avenue, Box 233, Houston, TX 77030. Dr. Harris and Ms. Rivera: Hospital Epidemiology, Rush-Presbyterian-St. Luke's Medical Center, 1753 West Congress Parkway, Chicago, IL 60612. Dr. Fuchs and Ms. Gustafson: St. Vincent's Hospital, 9205 Southwest Barnes Road, Portland, OR 97225. Dr. Wong: Epidemiology, Hunter-McGuire Hospital, Veterans Affairs Medical Center, 1201 Broad Rock Road, Richmond, VA 23249. Ms. Goetz, Ms. Wagener, and Dr. Yu: University of Pittsburgh Medical Center, Division of Infectious Diseases, Montefiore W931, Pittsburgh, PA 15213.
References
1. Lazenby D, Gold JP. Prevention and management of prosthetic valve endocarditis. Infect Med. 1991;8:15-6, 42-5.
2. Korvick JA, Bryan CS, Farber B, Beam TR Jr, Schenfeld L, Muder RR, et al. A prospective observational study of Klebsiella bactere
mia in 230 patients: outcome for antibiotic combinations versus monotherapy. Antimicrob Agents Chemother. 1992;36:620-5.
3. Chow JW, Fine MJ, Shlaes DM, Quinn JP, Hooper DC, Johnson M, et al. Enterobacter bacteremia: clinical features and emergence of antibiotic resistance during therapy. Ann Intern Med. 1991;115:585-90.
4. Korvick JA, Peacock JE Jr, Muder RR, Wheeler RR, Yu VL. Addition of rifampin to combination antibiotic therapy for Pseudo-monas aeruginosa bacteremia: prospective trial using the Zelen protocol. Antimicrob Agents Chemother. 1992;36:620-5.
5. Wilson WR, Jaumin PM, Danielson GK, Giuliani ER, Washington JA II, Geraci JE. Prosthetic valve endocarditis. Ann Intern Med. 1975;82:751-6.
6. Calderwood SB, Swinski LA, Karchmer AW, Waternaux CM, Buckley MJ. Prosthetic valve endocarditis. Analysis of factors affecting outcome of therapy. J Thorac Cardiovasc Surg. 1986;92:776-83.
7. Sande MA, Johnson WD, Hook EW, Kaye D. Sustained bacteremia in patients with prosthetic cardiac valves. New Engl J Med. 1972; 286:1067-70.
8. Alvarez-Elcoro S, Mateos-Mora M, Mantecon Y. Community-acquired febrile illness in patients with prosthetic heart valves. Southern Med J. 1985;78:1431-4.
9. Friedland G, Von Reyn CF, Levy B, Arbeit R, Dasse P, Crumpacker C. Nosocomial endocarditis. Infect Control. 1984;5:284-8.
10. Invert TS, Dismukes WE, Cobbs CG, Blackstone EH, Kirklin J. Prosthetic valve endocarditis. Circulation. 1984;69:223-32.
11. Heimberger TS, Duma RJ. Infections of prosthetic heart valves and cardiac pacemakers. Infect Dis Clin North Am. 1989;3:221-44.
12. Calderwood SB, Swinski LA, Waternaux CM, Karchmer AW, Buckley MJ. Risk factors for the development of prosthetic valve endocarditis. Circulation. 1985;72:31-7.
13. Rossiter SJ, Stinson EB, Oyer PE, Miller DC, Schapira JN, Martin RP, et al. Prosthetic valve endocarditis. Comparison of heterograft tissue valves and mechanical valves. J Thorac Cardiovasc Surg. 1978;76:795-803.
14. Rocchiccioli C, Chastre J, Lecomte Y, Gandjbakhch I, Gibert C. Prosthetic valve endocarditis. The case for prompt surgical movement. J Thorac Cardiovasc Surg. 1986;92:784-9.
15. Karchmer AW, Bisno AL. Infections of the prosthetic heart valves and vascular grafts. In: Bisno AL, Waldvogel FA, eds. Infections Associated with Indwelling Medical Devices. Washington, DC: American Society of Microbiology; 1989.
16. Von Reyn CF, Levy BS, Arbeit RD, Friedland G, Crumpacker CS. Infective endocarditis: an analysis based on strict case definitions. Ann Intern Med. 1981;94:505-18.
17. Brooks RG, Notario G, McCabe RE. Hospital survey of antimicrobial prophylaxis to prevent endocarditis in patients with prosthetic heart valves. Am J Med. 1988;84:617-21.
18. Horstkotte D, Friedrichs W, Pippert H, Bircks W, Loogen F. Nutzen der Endokarditisprophylaxe bei patientein mit prothet ischen Herz-klappen [Benefits of endocarditis prevention in patients with prosthetic heart valves]. Z Kardiol. 1986;75:8-11.
19. Dismukes WE, Karchmer AW, Buckley MJ, Austen WG, Swartz MN. Prosthetic valve endocarditis. Analysis of 38 cases. Circulation. 1973;48:365-77.
1 October 1993 • Annals of Internal Medicine • Volume 119 • Number 7 (Part 1) 567
Downloaded From: http://annals.org/ by a University of Pittsburgh User on 05/25/2013
Recommended