Catheter-Related Bacteremia and Outcome of Attempted Catheter Salvage in Patients Undergoing Hemodialysis

See related articles on pp 257-266 and 267-274. Single-lumen central venous catheters were introduced in the 1970s. In the 1980s, dual-lumen, tunneled, cuffed catheters were developed and became an accepted form of temporary and permanent vascular access in patients who lack other options [1]. With increasing numbers of debilitated patients undergoing dialysis, the use of dual-lumen cuffed catheters has become more common. These catheters are used for vascular access in 15% of patients undergoing dialysis in the United States [2]. Complications of vascular access devices are a major cause of illness and death in patients undergoing hemodialysis [3]. Treatment of the two major complications, thrombosis and catheter-related bacteremia, can necessitate removal of the catheter. Techniques developed in recent years have dramatically decreased the rate of catheter failure from thrombosis [4]. Infection, therefore, remains the leading cause of catheter loss. Previous studies have shown that the incidence of bacteremia is higher among patients in whom dialysis is done with noncuffed temporary catheters than among those in whom dialysis is done with cuffed, tunneled catheters [5-15]. However, the true incidence of bacteremia associated with dual-lumen, tunneled, cuffed catheters is unknown: No large prospective studies have been done, and smaller studies have not reported rates of bacteremia by using the recommended denominator of catheter-days. The appropriate management of catheter-related bacteremia in patients undergoing dialysis has also not been clearly defined. Although clinicians generally agree about the need for intravenous antibiotics in patients with catheter-related bacteremia, they disagree about the need for catheter removal. Because many patients undergoing long-term hemodialysis have exhausted all other options for vascular access by the time a tunneled, cuffed catheter is placed, it often seems preferable to treat catheter-related bacteremia without removing the device. The results of several recent small or retrospective studies suggest that many episodes of bacteremia can be successfully treated with the catheter in place (catheter salvage), although success rates have varied. In a prospective study of 13 episodes of catheter-related bacteremia [16], use of antibiotics both systemically and locked into the catheter had a 100% success rate. In contrast, two larger studies [13, 17] found that only 25% to 31% of episodes of catheter-related bacteremia could be successfully treated with systemic antibiotics alone. Many nephrologists allow 24 to 72 hours for a salvage attempt before removing an colonized catheter, provided that there is no evidence of tunnel infection or hemodynamic compromise. The safety and efficacy of this practice is not known because no large prospective studies have examined the outcome of bacteremia in patients undergoing hemodialysis in whom catheter salvage was attempted. To determine the incidence and outcome of catheter-related bacteremia, we conducted a prospective, observational study of outpatients undergoing hemodialysis at four dialysis centers. We also assessed the outcome of various methods of managing catheter-related bacteremia, especially attempted catheter salvage. Methods We enrolled patients who underwent hemodialysis with tunneled, cuffed catheters at four dialysis centers affiliated with Duke University Medical Center between April 1995 and January 1996. Catheters were inserted in the operating room (Perm-Caths, Quinton Instruments, Seattle, Washington) or by vascular radiologists (Tessio catheters, Med-Comp, Inc., Hadeville, Pennsylvania). For each patient, we collected demographic data and clinical information (number of years undergoing hemodialysis, history of bacteremia, and underlying medical conditions) by reviewing inpatient and outpatient charts. Patients suspected of having catheter-related bacteremia were reported to one of the authors within 24 hours of admission to the inpatient service; this author had daily contact with inpatient consulting nephrologists. Patients with bacteremia who were not admitted to the hospital were identified through interviews with dialysis personnel and through regular review of culture results from each dialysis center. The clinical response to treatment and all complications (infective endocarditis, osteomyelitis, septic arthritis, and death) were recorded for each patient. All patients with bacteremia were followed for at least 3 months after the initial episode. Patients were considered immunocompromised if they had HIV infection; cancer; or dependence on steroid treatment for systemic lupus erythematosus, chronic transplant rejection, or pulmonary disease. We defined suspected catheter-related bacteremia as fever (body temperature >38 C) or nonspecific systemic symptoms in a patient with a dual-lumen, tunneled, cuffed catheter for whom no other source of infection was apparent after a complete history was taken and a complete physical examination was done. We defined confirmed catheter-related bacteremia as growth of the same organism in blood cultures taken from two peripheral locations or from the catheter and a peripheral site. We provided treatment guidelines to all clinicians caring for patients. These guidelines recommended prompt removal of the catheter in patients who developed clinical signs of sepsis, tunnel infection, or persistent fever (>72 hours' duration) or bacteremia during antibiotic therapy. Clinical management of each patient, including the choice of antibiotic and the decision to keep or remove the catheter, was determined by the attending nephrologist. Vancomycin (loading dose, 20 mg/kg of body weight) and gentamicin (loading dose, 2 mg/kg) were the antibiotics most commonly used for empirical treatment. We defined short-course antibiotic therapy as antibiotic therapy lasting 2 weeks or less, and we defined long-course antibiotic therapy as antibiotic therapy lasting more than 2 weeks. Other antibiotics were sometimes used when results of culture and susceptibility testing became available. Blood cultures were repeated 1 to 2 weeks after the initial bacteremic episode, 1 week after completion of antibiotic therapy, and when clinically indicated. Patients with catheter-related bacteremia whose catheters were still in place 3 days after initial clinical recognition of bacteremia were considered to have had attempted salvage. Salvage was considered successful if the catheter that was in place at the time of the episode of catheter-related bacteremia was still in place at the end of 3 months or if the catheter had been removed for a reason other than persistent or recurrent bacteremia caused by the same organism (for example, if the catheter was no longer medically necessary because of maturation of alternate vascular access). Patients whose catheters were removed because of persistent fever or recurrent bacteremia caused by the same organism that was originally identified were considered to have had salvage failure. Patients who had their catheters removed in the setting of an infection caused by a new organism were excluded from the analysis. Data analysis (relative risks and 95% CIs) was performed by using Epi Info software (USD, Inc., Stone Mountain, Georgia). Patient characteristics potentially associated with rates of infection (age; duration of hemodialysis; history of bacteremia; presence of a Gore-Tex graft, diabetes mellitus, or HIV infection; injection drug use; and immunocompromised status) and rates of infection were evaluated by using chi-square analysis. Data from the four hemodialysis centers were analyzed both separately and together, with similar results; thus, we present results from analysis of all patients in all centers. All reported P values are two tailed. Results During the 9-month study period, 102 patients had dual-lumen, cuffed, tunneled catheters in place for 16 081 catheter-days. Fifty-five patients (54%) were female. Median age was 56 years (range, 24 to 87 years), and median duration of hemodialysis before the study was 2 years (range, 1 to 18 years). Twenty-three patients (23%) had conditions that met our case definition of immunocompromised status. Of these patients, 5 were infected with HIV, 3 had cancer, and 15 had long-term dependence on steroids. Forty-eight patients (47%) had diabetes mellitus, 77 (75%) had hypertension, and 6 (6%) used injection drugs on an ongoing basis. Forty-one patients (40%) developed 62 episodes of bacteremia. The incidence of catheter-related bacteremia was 3.9 episodes per 1000 catheter-days (95% CI, 3.0 to 4.9 episodes per 1000 catheter-days). Among the four hemodialysis units, the incidence of catheter-related bacteremia ranged from 2.2 episodes per 1000 catheter-days (CI, 0.4 to 3.2 episodes per 1000 catheter-days) to 5.5 episodes per 1000 catheter-days (CI, 2.3 to 6.6 episodes per 1000 catheter-days). Episodes of bacteremia did not cluster temporally at any hemodialysis center during the study period. Microbiology Thirty-nine (63%) of the 62 confirmed episodes of bacteremia were caused by a gram-positive coccus, 15 (24%) were caused by a gram-negative rod, and 1 (2%) was caused by a gram-positive rod. Seven episodes (11%) were caused by more than one organism. The most commonly isolated organism was Staphylococcus aureus (Table 1). Table 1. Microbiological Isolates from 62 Episodes of Bacteremia Risk Factors for Catheter-Related Bacteremia Table 2 shows the association between patient characteristics and the first episode of catheter-related bacteremia during the study period. After testing multiple factors (age, duration of hemodialysis, immunocompromised status, history of bacteremia, injection drug use, and presence of diabetes mellitus), we found that the risk for bacteremia was higher in patients who had a history of bacteremia (relative risk, 1.9 [CI, 1.2 to 2.9]) and in patients who were immunocompromised (relative risk, 1.