Bloodstream Infections After Invasive Nonsurgical Cardiologic Procedures

Objective  To define the incidence, risk factors, and characteristics of bloodstream infections (BSIs) after invasive nonsurgical cardiologic procedures (ICPs).

Methods  Retrospective case-control study; multivariate analysis.

Results  Between January 1991 and December 1998, 22 006 ICPs were performed in our hospital and 25 BSIs were documented within 72 hours after ICP. Overall incidence of bacteremia was 0.11% (25 cases) (0.24% after percutaneous transluminal coronary angioplasty [14 cases of 5625 patients], 0.6% after diagnostic cardiac catheterization [9 cases of 14 034 patients], and 0.8% after electrophysiologic studies [2 cases of 2347 patients]). These 25 patients with bacteremia were compared with 50 controls randomly selected among patients who underwent an ICP but did not have BSIs. Patient-related risk factors for BSI were age older than 60 years (20 cases [80%] vs 28 controls [56%]), valvular disease (4 [16%] vs 1 [2%]), congestive heart failure (7 [28%] vs 1 [2%]), indwelling bladder catheter before the ICP (5 [20%] vs 1 [2%]), more than 1 puncture for the ICP (5 [20%] vs 3 [6%]), a prolonged procedure (83.7 vs 65.1 minutes); and/or more than 1 ICP performed (2 [8%] vs 0). Multivariate analysis identified the presence of congestive heart failure (odds ratio, 21; 95% confidence interval, 6.8-66.0) and age older than 60 years (odds ratio, 1.9; 95% confidence interval, 1.9-6.3) as independent risk factors for BSI after ICP. Bloodstream infection was detected a median of 1.7 days after the procedure. Gram-negative bacteremia accounted for 17 cases (68%) of the BSIs. Among the patients with BSI, the duration of hospital stay was significantly increased (21 vs 6 days). The overall mortality rate was 0.009% for patients who underwent an ICP (8.0% for the 25 patients with bacteremia documented within 72 hours after ICP).

Conclusions  Bloodstream infection should be included among the potential complications of ICP. Elderly patients with recent congestive heart failure episodes constitute a subgroup with a higher risk of postprocedure bacteremia. Therapy with antimicrobial agents against gram-positive and gram-negative bacteremia should be initiated after performing blood cultures in patients with signs suggestive of infection.

IN MODERN hospitals many invasive nonsurgical cardiologic procedures (ICPs) are performed every day. Although these techniques are safe, a wide array of systemic and local potential complications associated with these procedures have been reported. Systemic adverse effects include cerebrovascular accidents (0.07%-0.30%), myocardial infarction (0.06%-4.80%), and even death (0.10%-1.10%).^1-8

Related infectious complications are not commonly reported, and local problems predominate.^6,8-24 However, patients undergoing ICP may develop bloodstream infections (BSIs) that may cause death or significantly extend the hospital stay, which increases the cost.

An ICP-related BSI is an infrequent finding, and most of the available literature consists of isolated case reports that do not allow for an accurate definition of risk factors.^{14,16,17,19-22} Herein we report what is to our knowledge the largest case-control series of early BSI after ICP from a single center. Our objective was to define the incidence, risk factors, and clinical characteristics of early BSI in patients who have undergone ICP.

Our institution is a 2000-bed teaching hospital with a very active invasive cardiology department and heart transplantation program. The records of ICPs performed in the Cardiology Department were cross referenced with the records of BSI detected in the Clinical Microbiology Laboratory. Patients with BSI detected within 72 hours after ICP were selected as "cases." This period was chosen to maximize the likelihood of only including BSIs related to the ICPs. For each case subject with ICP-related BSI, 2 "control" subjects were elicited from a random number table from patients who had undergone ICP in the same period but had not developed a subsequent BSI.

Blood cultures were ordered when the patients experienced fever, chills, or others signs suggestive of infection. Only significant bacteremic episodes were included, and all blood cultures were obtained through direct venipuncture rather than through an intravascular device. When a case of Staphylococcus epidermidis bacteremia was found, the episode was only included when at least 2 blood culture sets yielded the same microorganism and the patient showed clinical signs suggestive of infection. Blood samples were processed with an automated monitoring system (BACTEC NR; Becton, Dickinson and Company, Franklin Lakes, NJ). Isolates were identified with standard microbiological tests and automated methods (MicroScan; Dade International Inc, West Sacramento, Calif). None of the patients received prophylactic treatment, and ICPs were performed using standard sterile techniques and infection prevention measures.²⁵ Femoral access was undertaken in all instances.

Data collected included age; sex; underlying diseases (diabetes mellitus, malignancy, central nervous system disorders, chronic renal failure, liver disease, human immunodeficiency virus, valvular heart disease, congestive heart failure in the previous 14 days, presence of intravascular prosthetic material, chronic obstructive pulmonary disease, and previous myocardial infarction); risk factors (alcoholism, current smoking, drug abuse, steroid treatment, central venous catheters, and indwelling urinary tract catheter); type of ICP (percutaneous transluminal coronary angioplasty [PTCA], diagnostic cardiac catheterization, or electrophysiologic studies); priority of the ICP (urgent or not); procedure variables (duration of the procedure, number of punctures, and number of days the sheath was left in place); complications and outcome during or after ICP (fever, chills, vascular lesion, neurologic lesion, hematoma at the site of puncture, primary BSI, catheter-related BSI, urosepsis, pneumonia, meningitis, endocarditis, shock, intensive care unit admission, and death); analytical data (white blood cell count, left deviation, platelet count); microorganism identified from blood cultures (gram-positive, gram-negative, or polymicrobial); duration of hospital stay; and empirical treatment (adequate or inadequate). Antimicrobial therapy was considered adequate if 1 or more antimicrobial agents with in vitro activity against the corresponding isolate were administered for a minimum of 5 days. Death was attributed to the infectious process if the patient died within 10 days of the bacteremic episode with a clinical course suggesting persistent infection; it was always attributed if the patient died during the phase of acute infection and death could not be clearly attributed to any other cause.

Assessment of significant risk factors for BSI was performed by univariate and multivariate analyses. Quantitative variable analysis was analyzed with the Mann-Whitney test. The study of qualitative variables was performed with the χ² test with the Yates correction or the Fisher exact test (2-tailed) whenever possible. Adjusted risk ratios with 95% confidence intervals were assessed by stepwise logistic regression analysis (SPSS software; SPSS Inc, Chicago, Ill). Nonrelated variables identified as significant in the univariate analysis were entered into the logistic regression analysis. Statistical significance was defined as P<.05.

During the study period (January 1991 to December 1998), 68 patients had an episode of BSI among the 22 006 who underwent an ICP. Of these episodes, 25 occurred within 72 hours after the procedure, resulting in an overall incidence of 0.11%. Incidence of early BSI was significantly higher after PTCA: 14 (0.24%) of 5625 patients who underwent PTCA followed by 9 (0.06%) of 14 034 patients who underwent cardiac catheterization and 2 (0.08%) of 2347 patients who underwent electrophysiologic studies (P = .001). However, the type of cardiologic procedure was not an independent risk factor in the multivariable study.

Risk factors predisposing to BSI were grouped into those present before the procedure (the first group) and those related to the procedure itself (the second group) (Table 1). Among the first group, age older than 60 years (20 cases [80%] vs 28 controls [56%]), prior valvular disease (4 [16%] vs 1 [2%]), congestive heart failure (7 [28%] vs 1 [2%]), and indwelling bladder catheterization (5 [20%] vs 1 [2%]) were more common among patients with BSI. Among the second group, more than 1 ICP performed (2 [8%] vs 0), number of punctures needed (mean of 1.2 vs 1.05), duration of the procedure (mean of 83.7 vs 65.1 minutes), and development of inguinal hematoma (4 [16%] vs 1 [2%]) were associated with a higher risk of developing BSI in the univariate analysis. Multivariate analysis identified the presence of congestive heart failure (odds ratio [OR], 21; 95% confidence interval, 6.8-66.0) and age older than 60 years (OR, 1.9; 95% confidence interval, 1.9-6.3) as independent risk factors for BSI after ICP.

Bloodstream infection was detected a median of 1.7 days after ICP (range, 0-3 days); 9 (36%) of them were detected in the first 24 hours after ICP and 21 (84%) in the first 48 hours. The main clinical signs of BSI were fever (25 [100%]) and chills (16 [64%]), which were both only present in patients with BSI. One patient developed endocarditis, which was possibly related to an episode of Staphylococcus aureus BSI after an electrophysiologic study. However, it was not included in this analysis because the BSI occurred 7 days after the procedure. None of the patients with both bacteremia and valvular heart lesions who were included in this study developed endovascular infection. No local infection was documented in this series.

Microorganisms responsible for BSI are listed in Table 2. Of the episodes, 17 (68%) were caused by gram-negative microorganisms. None of them were polymicrobial and no grouping of cases was detected. Our 4 patients with previous valvular heart lesions had a gram-negative bacteremia (Pseudomonas cepacia, Klebsiella pneumoniae, Enterobacter cloacae, and Escherichia coli), and none of them had endocarditis or endarteritis. The origins of the bacteremia in the patients with an indwelling bladder catheter were S aureus in 2; Enterococcus faecalis in 1; Serratia marcescens in 1; and E coli in 1.

Antimicrobial therapy was considered adequate in 75% of cases. Five patients did not receive therapy (1 patient died before treatment was initiated, and the other 4 had already been released from the hospital or were asymptomatic when the results were available). All of them survived, and the origins of these bacteremias were Klebsiella oxytoca in 2; E faecalis in 1; and E coli in 1.

Three patients (12%) developed septic shock, and 2 of them died (the overall mortality was 12%, and the related mortality was 8% for the 25 patients with bacteremia that was detected within 72 hours after ICP). The median hospital stay was significantly longer for patients with bacteremia (21 vs 6 days) (Table 1).

The use of ICP has experienced a rapid expansion in recent years, and a large number of patients undergo these procedures every day. More than 300 000 PTCAs were performed in the United States in 1990²³ and more than 600 000 in 1999.

Although infectious complications are uncommon, the great number of procedures performed every year implies that even a low incidence of infection may have significant consequences. We are reporting what is to our knowledge the largest series of early BSIs after ICP. In our hospital, 0.11% of all patients who had undergone an ICP had a clinically significant procedure-related BSI in the following 3 days. Incidence of BSI was significantly higher among patients who underwent PTCA (0.24%), although the type of procedure was not an independent risk factor in our multivariate model. In the literature, incidence of bacteremia after PTCA ranges from 0.20 to 0.80,^9-11,19 although the study by Samore et al¹¹ includes bacteremias that were detected in the first 7 weeks following PTCA.

To identify patients with a higher risk of severe infectious complications who would require a closer follow-up or even prophylaxis, we performed a case-control study and literature review. Risk factors proposed in previous studies (all but 2 were derived from series with less than 6 patients with bacteremia) are summarized in Table 3.

Congestive heart failure was the most significant risk factor in our multivariate model (OR, 21; 95% confidence interval, 6.8-66.0). This factor was also identified by Samore et al¹¹ in the only other comparative series of patients with bacteremia after PTCA (OR, 43.3). The reason for this association is not clear. In our series, patients with bacteremia and previous congestive heart failure were younger than those without BSI (median age, 61 vs 71 years; P = .04), had a greater incidence of valvular diseases (3 cases [42%] vs 1 control [5%]; P = .02) and diabetes mellitus (4 [57%] vs 1 [5%]; P = .003), were more commonly receiving mechanical ventilation (2 [28%] vs 0; P = .01), required more central venous catheters (3 [43%] vs 1 [5%]; P = .02), and needed more than 1 puncture for the ICP (4 [57%] vs 1 [5%]; P = .03).

The second independent risk factor was age older than 60 years, which was also suggested in previous works.¹¹ In our series, patients older than 60 years were more commonly heavy smokers (10 [50%] vs 0; P = .04). In the literature, cohorts of older patients who underwent PTCA include more women, more patients with unstable angina, and more patients with left ventricular dysfunction and congestive heart failure. Age older than 75 years was reported to be an independent risk factor for death, acute myocardial infarction, need for transfusion, and need for arterial repair after PTCA.³

In contrast to the results of Samore et al,¹¹ we could not find a correlation between the risk of bacteremia and any of the factors related to the procedure itself, although some of them were significant in the univariate analysis (Table 1). Samore et al found that duration of the procedure (OR, 6.8), number of catheterizations in the same site (OR, 4.0), difficult vascular access (OR, 15.0), and length of time the sheath was left in place (OR, 6.8) increased the risk of bacteremia. The significantly lower incidence of bacteremia after PTCA in our study compared with that of Samore et al¹¹ (14 cases [0.24%] vs 27 cases [0.64%]; P = .003) may be justified by technical factors. For example, 20 (80%) of our patients vs 12 (44%) of the patients required just 1 puncture, the duration of the procedure in our center was less than 1 hour in 36% vs 11% of the cases, and the sheath always remained in place less than 24 hours (mean time, 7.28 hours), while in the study by Samore et al¹¹ the sheath remained in place less than 24 hours in only 7 cases (26%). In another series, duration of the procedure was also not found to be a significant risk factor.⁹

Retention of the sheath (40 vs 22 hours) and bleeding around the insertion site (43% vs 9% transfusion rate) were risk factors for infection in a prospective series of 500 patients who underwent angioplasty with 6 related infections (4 bacteremias).¹⁰ These factors were not found to increase the risk of bacteremia in our series, but both our transfusion rate (4%) and the length of time the sheath was left in place are significantly less than those reported in the series by McHenry et al¹⁰.

Early repuncture of the ipsilateral femoral artery or repeated procedures through a previously traumatized femoral access site have been proposed as risk factors in some short series, mainly for local complications.^16,19-21,24 However, these findings were not confirmed by others.^6,12 Two of our 25 patients with bacteremia underwent a second cardiologic procedure, but this factor was not independently related to a higher risk of bacteremia.

Local infections are very uncommon⁸; there were none in our series, probably because we had focused on the early onset of bacteremia. Blood cultures may remain negative even after severe local infections,¹⁸ and femoral endarteritis may be associated with osteoarticular metastases and peripheral emboli.^22,23

Fever after ICP has been correlated with the use of contrast medium² or with the introduction of bacterial endotoxin from the surface of the catheter or from the skin.^12,24 However, the change to single-use pyrogen-free catheters has virtually eliminated these problems as well as the potential for BSI due to unsterile catheters. In our study, fever 25 (100%) of the cases vs none of the controls; P<.001) and chills were the most constant symptoms of BSI after ICP. The association of fever after an ICP with real infection has also been indicated by some authors (50% of cases vs 4% of controls).¹⁰

Gram-negative microorganisms were more commonly identified in our study (17 [68%]). This finding has also been observed by other authors,^15,26,27 while gram-positive bacteria (mainly Staphylococcus species) predominated in other reports.^9,11,19 In a series of 7 cases of bacteremia after coronary stenting, 3 episodes were caused by S aureus, 2 by E cloacae, and 1 each by Proteus mirabilis and S marcescens.¹⁵ Gram-negative bacteria are well-known colonizers of perineal and groin skin, mainly in patients with indwelling bladder catheters. However, this was not the only reason, since 3 of our 5 cases with indwelling bladder catheter had gram-positive bacteremia.

Other potential reasons for the importance of gram-negative bacteremia in our series is that our population was severely ill and older (mean age, 66.6 years), both well-recognized risk factors for colonization by gram-negative bacteria. In some of the published series in which gram-positive bacteremia predominated, blood cultures were obtained from the femoral catheter or from the vessel from which the catheter had just been removed. Thus, some of these isolates may represent contamination by skin pathogens.¹² This practice is not useful for the diagnosis of BSI unless lysis centrifugation techniques are used and another blood culture is obtained through the vein.

The overall mortality rate related to the bloodstream infection was 0.09 per 1000 ICPs (8% of the 25 patients with bacteremia that was detected within 72 hours after ICP). In other series, 1.7% of the deaths were caused by a related infection (0.16 per 1000 PTCAs).² Patients with bacteremia had a much more prolonged hospital stay (21 vs 6 days), which also increased the cost.

Bloodstream infection complicates 0.11% of ICPs. Elderly patients with recent congestive heart failure episodes constitute a subgroup with a higher risk of postprocedure bacteremia. Antimicrobial agents against gram-positive and gram-negative bacteria should be initiated after blood cultures are performed in patients with signs suggestive of infection. The selection of a subgroup of patients with a high risk of infection who could benefit from antimicrobial prophylaxis merits further study.

Accepted for publication February 12, 2001.

The authors are grateful for the help of the staff of the Cardiac Catheterization Department and Microbiology Laboratory of the Hospital General Universitario "Gregorio Marañón."

Corresponding author and reprints: Patricia Muñoz, MD, PhD, Servicio de Microbiología, Hospital General Universitario "Gregorio Marañón," Doctor Esquerdo 47, Madrid 28006, Spain (e-mail: pmunoz@micro.hggm.es).