Introduction

Reliable and safe vascular access is one of the most essential features of modern medical care. Unfortunately, the intravascular devices needed to establish stable access are associated with significant potential for producing iatrogenic disease, most often catheter-related bacteremia or candidemia [1, 2, 3]. More than 250,000 intravascular device-related bloodstream infections (BSIs) occur in the United States each year [1]; the majority are related to short-term noncuffed, percutaneously placed central venous catheters (CVCs). CVC-related BSIs are associated with 12–25% attributable mortality [4, 5, 6, 7], and prolongation of hospital stay [5, 6, 7, 8] and marginal cost to the health system of $33,000–35,000 per episode [5, 6, 8].

Understanding the pathogenesis of CVC-related BSIs is essential to devising strategies for prevention of these infections; however, few studies have determined the mechanism of CVC-related colonization and infection, particularly using molecular techniques to verify potential routes of infection [9, 10, 11, 12, 13, 14, 15]. We report a prospective study which employed molecular epidemiology to determine the pathogenesis of CVC-related BSI with short-term, noncuffed CVCs placed percutaneously in the internal jugular, subclavian, or femoral veins.

Patients

A total of 1,098 patients participating in two randomized trials during 1998–2000 [16, 17] formed the study population. One trial studied the effect of a 1% chlorhexidine–75% alcohol solution for cutaneous antisepsis for intravascular catheters [17], and the other evaluated the efficacy of a novel chlorhexidine-impregnated sponge dressing for prevention of catheter-related BSI [16]. The two trials were very similar in overall design and studied common patient populations (Table 1). The majority of patients studied were elderly and had one or more underlying diseases; the mean severity of illness score (Acute Physiology and Chronic Health Evaluation II) was 22. Most patients had a urinary catheter and were mechanically ventilated during the period in which they had a study CVC. Complete data were obtained for 1,263 CVCs, a total of 6,075 CVC days. Data were collected prospectively on study patients with newly inserted short-term, noncuffed and nontunneled central venous catheters, including demographic features, underlying diseases, Acute Physiology and Chronic Health Evaluation II [18], reason for placement of the catheter, service, antibiotic use, length of hospital stay, number of days the catheter remained in place, presence of other invasive devices (urinary catheters and endotracheal tubes), and all clinical and laboratory data pertaining to infection. Both studies were approved by the institutional review board, and written informed consent was obtained from all subjects prior to enrollment.

Table 1 Features of the study population (APACHE Acute Physiology and Chronic Health Evaluation; n=1,098)
Full size table

Microbiological methods

At the time of catheter removal, skin of the insertion site was cultured quantitatively, as previously described [19]. For each catheter two 5-cm segments, the intracutaneous segment and the tip, each transported in a sterile container, were cultured semiquantitatively; each hub and fluid aspirated aseptically from the most distal injection port of the line were also cultured quantitatively, as previously described [19]. Micro-organisms were identified according to standard criteria [20]. When catheter-associated BSI occurred, isolates recovered from the insertion site, catheter segments, infusate or hubs, and blood cultures that appeared similar phenotypically were subtyped by pulsed-field gel electrophoresis after digestion of genomic DNA with restriction endonucleases [21], using a computerized system and Centers for Disease Control criteria [22] for determining the relatedness of isolates (Gel Doc 2000, Bio-Rad Laboratories, Hercules, Calif., USA).

Definitions

Catheter-tip colonization was defined as a positive semiquantitative culture of an intravascular catheter segment (>15 colony-forming units) and is considered synonymous with local infection of the catheter [19]. Catheter-related BSI was defined as isolation of the same strain from the catheter segment, a hub, or infusate and from one or more blood cultures, as confirmed by restriction-fragment subtyping, with no other clearcut source for the BSI [19]. BSIs were considered to be extraluminally acquired if concordance was demonstrated solely between isolates from catheter segments, skin, and blood cultures and intraluminally acquired if concordance was demonstrated only between isolates from a hub or infusate and blood cultures; if the findings suggested that both routes or neither might be operative, the pathogenesis of infection was designated as indeterminate.

Statistical methods

The significance of differences between groups was calculated by χ2 or Fisher’s exact test. We considered two-tailed p values less than 0.05 as statistically significant.

Results

Of the 1263 CVCs 333 (26.3%) were colonized at removal and 35 (2.7%) caused CVC-related BSI, an incidence density of 5.9 per 1000 CVC days. BSIs were caused by coagulase- negative staphylococci (n=27), enterococci (n=4), enteric Gram-negative bacilli (n=3), or Candida (n=1; Table 2).

Table 2 Microbiology and presumed pathogenesis of 35 central venous catheter-related bloodstream infections, based on DNA subtyping (extraluminal concordance demonstrated solely between skin, catheter segments, and blood cultures, intraluminal concordance demonstrated exclusively between a hub or infusate and blood cultures, indeterminate findings suggest that neither or both routes of infection operative)
Full size table

Recognizing that the database is comprised of patients who were participating in two randomized trials of novel strategies for prevention of CVC and arterial catheter-related BSI, and that both strategies were found to reduce the incidence of catheter colonization and CVC- and arterial catheter-related BSI [16, 17], the rate of CVC-related BSI was also calculated for the pooled data from the control groups [700 CVCs, 3,626 catheter days, 25 CVC-related BSI (3.6%), 7 per 1000 catheter days] and the treatment groups of the two trials [563 CVCs, 2,449 catheter-days, 10 CVC-related BSIs (1.7%), 4.0 per 1000 catheter days].

In the overall population studied 45% of CVC-related BSIs were extraluminally acquired, 26% appear to have had an intraluminal origin, and the mechanism of infection was indeterminate in 29% (Table 2). In a separate analysis of BSIs from the pooled data of the treatment and control groups (Table 3) it can be seen that the extraluminal route was most important (60%) for the catheters in the control groups, but with successful suppression of cutaneous colonization by the two strategies studied the CVC-related BSIs that occurred in the treatment groups were much more likely to be intraluminal (60%) or indeterminate (30%) in origin (p=0.006). A representative pulsed-field gel electrophoresis gel showing the pathogenesis of a CVC-related BSI is shown in Fig. 1.

Table 3 Contrasts in the pathogenesis of central venous catheter-related bloodstream infections in the control and treatment groups of the study population: 1% chlorhexidine–75% alcohol solution for cutaneous antisepsis for intravascular catheters [17] or a chlorhexidine-impregnated sponge dressing applied to the insertion site [16]
Full size table
Fig. 1
figure 1

Pulsed-field gel electrophoresis showing the probable pathogenesis of a central venous catheter-related bacteremia with coagulase-negative Staphylococcus. The isolates from the skin of the insertion site, catheter tip, and blood and were concordant, indicating an extraluminal route of infection

Full size image

The pathogenetic mechanisms were also analyzed according to the method of insertion: first catheter in a de novo site and second catheter inserted in an old site over a guidewire; there were no differences between the two groups (Table 4).

Table 4 The pathogenesis of central-venous catheter-related bloodstream infection, comparing de novo insertion (first catheter in a site) and guidewire exchange (second catheter in the site)*
Full size table

Discussion

There are two major sources of intravascular-device-related BSI: (a) colonization of the device, catheter-related infection, and (b) contamination of the fluid administered through the device, infusate-related infection [23]. Contaminated infusate is the cause of most epidemic intravascular device-related BSIs [1]. In contrast, catheter-related infections are responsible for most endemic device-related BSIs. In order for micro-organisms to cause catheter-related infection they must first gain access to the extraluminal or intraluminal surface of the device where they can adhere and become incorporated into a biofilm that allows sustained infection and hematogenous dissemination [11]. Micro-organisms gain access by one of three mechanisms: skin organisms invade the percutaneous tract, probably facilitated by capillary action [24], at the time of insertion or in the days following; micro-organisms contaminate the catheter hub (and lumen) when the catheter is inserted over a percutaneous guidewire or later manipulated [25]; or organisms are carried hematogenously to the implanted device from remote sources of local infection, such as a pneumonia [26, 27] (Fig. 2). With short-term devices (in place <10 days) [28]—peripheral IV catheters, arterial catheters and noncuffed, nontunneled CVCs—the limited studies that have been carried out suggest that most catheter-related BSIs are of cutaneous origin, from the insertion site, and gain access extraluminally, occasionally intraluminally [12, 29, 30]. For long-term devices (in place >10 days) [28]—cuffed and tunneled central venous catheters, subcutaneous ports or peripherally inserted central venous catheters—luminal colonization appears to be the major mechanism of device-related BSI [10, 13, 31, 32].

Fig. 2
figure 2

Potential sources of infection of a percutaneous intravascular device: extraluminally, from the contiguous skin flora; intraluminally, by contamination of the catheter hub and lumen or contamination of infusate; and hematogenously, from distant, unrelated sites of infection. (From [28])

Full size image

In our large, prospective study of short-term, noncuffed CVCs used in patients in an ICU, the extraluminal route appears to have been the major (or a contributory) mechanism of infection of nearly two-thirds of CVC-related BSIs; however, in 26%, BSIs appeared to have derived from intraluminal contaminants, suggesting that both extra- and intraluminal routes are important with short-term noncuffed CVCs. Studies from Europe [15, 25, 33], where noncuffed, nontunneled CVCs are commonly left in place for longer than the 4–7 days representative of most centers in the United States (Table 1) [19, 34, 35], have found the intraluminal route to be more important, and we believe that the intraluminal route begins to predominate when a noncuffed and nontunneled catheter is left in place for much longer than 1–2 weeks.

Numerous studies performed in the United States have found heavy cutaneous colonization to be a powerful risk factor for CVC-related BSI [36, 37, 38, 39, 40]. Catheter insertion into the femoral vein [41] or internal jugular vein [42] rather than the subclavian vein also carries a higher risk of CVC-related BSI, probably because of heavier cutaneous colonization present at femoral or internal jugular insertion sites [43].

A number of measures have been found to be effective in prevention of CVC-related BSI. Foremost among these is education of personnel regarding catheter insertion technique and catheter care practices. Several large prospective studies have shown a 30–70% relative risk reduction with focused education programs [44, 45]. Tunneling a CVC also appears to reduce the risk of catheter-related BSI, both with catheters placed in the internal jugular or femoral veins [46, 47], and might be considered if circumstances favor cannulation of an internal jugular or femoral vein rather than a subclavian vein (e.g., severe coagulopathy or a hemodialysis catheter). Novel technologies which can reduce cutaneous colonization at the insertion site are most likely to be effective for preventing infection originating by the extraluminal route but can also provide benefit for intraluminally derived infections and have shown efficacy with short-term noncuffed CVCs, such as chlorhexidine for cutaneous antisepsis, chlorhexidine-impregnated sponge dressing, and anti-infective coated catheters [16, 17, 19, 48]. Our findings demonstrate, as presented in Table 3, that preventive strategies that effectively reduce cutaneous colonization markedly reduced the risk of CVC-related BSI with short-term noncuffed CVCs but shifted the pathogenesis of those CVC-related BSIs that still occur in treated patients to the intraluminal route.

In conclusion, in our study the majority of CVC-related BSIs with short-term noncuffed CVCs were extraluminally derived (45%), although 26% appeared to have had an intraluminal origin, and in 29% of catheters the mechanism of infection was indeterminate, suggesting that both routes may have been operative. The mechanism of infection with CVCs inserted in old sites over a guidewire did not differ materially from that seen with catheters inserted in de novo sites. This study points up the importance of a multifaceted approach for prevention of CVC-related BSI with short-term noncuffed CVCs: strategies designed to prevent extraluminally derived BSIs but also strategies to prevent infection that originates from intraluminal contaminants.