Methods

Top Methods Results Discussion Author & Article Info References

From 1 June 1987 to 31 May 1989, patients who had catheters and ports inserted at Memorial Sloan-Kettering Cancer Center were followed prospectively to identify possible device-related morbidity. Baseline demographic information was obtained from the patients' medical records and operative reports. During subsequent admissions, the chart was reviewed by the vascular access nurse clinician. Daily microbiology reports were cross-checked against the database. Any study patient not seen by a principal investigator for 3 months received a questionnaire to identify any device-related morbidity in the intervening period. Patients were contacted if questionnaires were not returned or additional information was required. Follow-up continued until the device was removed or the patient died.

Definitions of Infection

Preset criteria for the diagnosis of device-related infectious morbidity were established [9, 19].

Device-related bacteremia or fungemia was defined using the following criteria: 1) a 10-fold or greater increase in colony-forming units of organism per milliliter of blood obtained through the device compared with simultaneous peripheral blood cultures; 2) in the absence of peripheral blood cultures, more than 1000 colony-forming units of organism obtained through the device; and 3) a positive result of catheter tip culture when the device was removed specifically for suspected device-related infection in the absence of cultures as stated above.

Device-related bacteremia or fungemia was considered cured when culture results were negative at the termination of antibiotic therapy and no evidence of clinical infection occurred at least 2 weeks later.

Catheter tunnel infection was defined as induration, tenderness, and erythema beginning at least 1 cm from the catheter exit site and tracking up the catheter tract.

Port pocket infection was defined as induration, erythema, and tenderness around the port with culture-positive material aspirated from the port pocket.

Cutaneous site infection was defined as erythema, induration, or tenderness and exudate at the catheter cutaneous exit site or at the port surface needle access site.

Microbiologic Methods

Blood for culture, including quantitation, was collected into an Isolator 10 tube (Wampole Laboratories, Cranbury, New Jersey) and a 100-mm by 16-mm Vacutainer tube containing 5.95 mg of sodium polyethanol sulfonate (Becton-Dickinson Medical Systems, Rutherford, New Jersey). The Isolator tube was processed according to the manufacturer's instructions. After centrifugation, concentrated sediment from the Isolator tube was inoculated in equal portions onto two Columbia sheep blood agar plates and two chocolate agar plates, which were incubated aerobically at 37 °C in 5% CO₂ in air for 5 days. Blood from the Vacutainer tube was inoculated into one bottle containing 50 mL of Columbia broth with increased cysteine and carbon dioxide (Becton-Dickinson). The bottle remained unvented during incubation. Cultures from catheter tips were taken by vortexing for 30 seconds in 30 mL of trypticase soy broth followed by inoculation of 1 drop on Columbia sheep blood plates as above with incubation for 3 days. Specimens from suspected site infections were plated on chocolate agar, MacConkey agar, and Columbia sheep blood agar with colistin and nalidixic acid.

Care of Devices

All patients with catheters required home device care. After insertion, the catheter exit site was covered with an occlusive gauze dressing until the site had healed and the Dacron cuff had engrafted (a minimum of 1 month). At this point, patients were given the option to continue to use an occlusive gauze dressing or to wear no dressing other than an adhesive bandage or a piece of tape to secure the device and prevent motion and inadvertent pulling. Each catheter lumen, when not in continual use, was flushed daily with 5 mL of heparinized saline (10 U/mL). Devices were inspected each day, and exit sites were cleaned with soap and water and then povidone-iodine solution. Catheter injection caps were changed twice each week. Patients with catheters were instructed not to swim. Patients with ports did no procedures on the device at home. Ports were flushed after each use and every 4 to 6 weeks when not in use with 5 mL of normal saline followed by 3 mL heparinized saline (100 U/mL).

Statistical Analysis

Descriptive statistical analysis consisted of paired and unpaired t-tests, analysis of variance, and chi-square analysis when indicated. For patients who had more than one device, only the first of each type was included in the analysis.

To determine whether ports are better than catheters with respect to infection and adjusting for other possible risk factors (covariates), we did a survival-type regression analysis comparison. The dependent variable was defined as the number of days until the first documented device-related infection of any kind or time of removal or last follow-up (censored observations). To avoid difficulty in interpreting results due to correlated observations for the same patient, only the first device inserted was used for the analysis, regardless of whether the patient had more than one of the same or different devices. Only patients with solid tumors, leukemia, or lymphoma or myeloma were included in the analysis of time to first infection because tumor type was likely to be a predictor of infections and because only a small proportion of patients with ports had other tumor types. Variables tested for relation to the outcome variable were device type, tumor type, age, device breakage, and clot (defined as clinical necessity to administer urokinase for an occluded device lumen or a positive venogram). Only catheter clot and breakage that occurred before the first infection or censoring time were considered and treated as time-dependent variables. The strength of the relation between infection-free time and each explanatory variable was assessed parametrically by the Wilcoxon-Gehan test statistic rather than by the log-rank test or Cox regression because the log-logistic distribution gave a better parametric fit to the data than did the Weibull distribution, and data did not follow a proportional hazards model. The Kaplan-Meier method was used to estimate time to first infection within subgroups to plot curves. The best parametric fit was obtained using the log-logistic distribution in SAS procedure LIFEREG (Statistical Analysis Systems Institute Inc., Cary, North Carolina), which allowed us to quantify the relation between risk factors and time to first infection and to give a simple predictive model. In this model, the proportion of patients in a population who have not had an infection at T days after insertion is 1/(1 + [T/A]^b), where A is the baseline median duration multiplied by proportionality coefficients for relevant risk factors that significantly prolong or shorten time to infection, and b is an exponent that is inversely related to the dispersion of times to infection in the population. In all cases, a P value less than 0.05 was considered significant. All summary data are presented as mean ±SD unless otherwise specified.

Results

Top Methods Results Discussion Author & Article Info References

During a 2-year period beginning 1 June 1987, 1630 venous access devices were inserted in 1430 patients: 707 ports were inserted in 680 patients, and 923 catheters were inserted in 788 patients (38 patients had a catheter and a port on separate occasions). All patients were followed for a potential minimum of 500 days. Table 1 shows that the average individual number of days that devices were in situ until removal or last follow-up, without censoring for infection, is significantly longer for ports than for catheters (P 0.001). Total port days were 277 853 at last follow-up compared with 165 710 catheter days; 171 (25%) of the ports were still in situ compared with 54 (7%) of the catheters. On average, ports remained in situ 408 ± 309 days (median, 339 days), whereas catheters remained in situ 210 ± 239 days (median, 130 days).

At least one device-related infection occurred in 341 (43%) of the catheters compared with 57 (8%) of the ports (P 0.001) (Table 2). Both site infections and device-related bacteremia or fungemia occurred more frequently with catheters than with ports. No difference was seen in the incidence of catheter tunnel or port pocket infections. More catheters (154 [20%]) than ports (25 [3%]) were ultimately removed because of infectious morbidity (P 0.001); these figures represent slightly fewer than one half the devices that ever caused infection. There were 2.767 infections per 1000 catheter days compared with only 0.211 infections per 1000 port days (P 0.001).

Device-related bacteremia or fungemia was the predominant type of infection with catheters, whereas ports had a more equal distribution of pocket, site, and device-related bacteremia or fungemia (see Table 2). Without accounting for removed devices, device-related bacteremia or fungemia tended to occur later than site or tunnel tract infection. Overall, port infections occurred later than catheter infections. Port pocket infections tended to occur earlier than site infections or device-related bacteremia or fungemia; however, no statistically significant differences appeared within this device category referable to the onset of infection.

Patients with ports were older than patients with catheters, (49 ± 18 years [median, 52 years] compared with 29 ± 20 years [median, 26 years]). Differences in age were noted between patients with catheters and those with ports if they had solid tumors or leukemias (Table 3). For patients with catheters, those with lymphoma or myeloma were older than patients in all other groups except those with the acquired immunodeficiency syndrome (AIDS); those with AIDS were older than those with solid tumors and bone marrow transplant; and those with leukemia were older than those with bone marrow transplant (P < 0.05). For patients with ports, those with solid tumors were older than those with leukemia; patients with leukemia were older than those with lymphoma or myeloma (P < 0.05). Ports were inserted predominantly for treatment of adult solid tumors (breast, colon, lung), whereas catheters were used in patients with bone marrow transplants, hematologic cancers, and childhood cancers. Ports were inserted in patients with leukemia only after their acute induction and were primarily for maintenance chemotherapy. In general, patients with catheters received more aggressive chemotherapy associated with a relatively greater degree and duration of neutropenia and thrombocytopenia than did patients with ports, and their devices had to be accessed more frequently for phlebotomy or blood product transfusion. Although a wide range of leukocyte counts was found at device insertion, no statistically significant difference in counts was found for all or for specific patient categories (Table 3). The precise degree and duration of neutropenia and the frequency of device use are not available for direct comparison.

Microbiology

Device-related bacteremia or fungemia occurred as the initial infection with 264 catheters and 26 ports. A 10-fold difference in colony-forming units between the device and peripheral blood was found for 153 catheters and 13 ports; more than 1000 colony-forming units from the device in the absence of peripheral blood were found in 107 catheters and 12 ports; and in the remaining 4 catheters and 1 port, a positive result of a catheter tip culture was diagnostic of device-related bacteremia or fungemia. More than one organism was recovered from 63 catheters and 3 ports. The bacteria and fungi causing the first device-related bacteremia or fungemia are listed in Table 4. The type of organisms causing device-related bacteremia or fungemia differed between catheters and ports, with gram-negative bacilli (55%) isolated predominantly in catheters and gram-positive cocci (65.5%) isolated in ports. Tunnel infection was the initial infection for 17 catheters, but organisms were not isolated in 12 of them. Staphylococcus aureus was the causative organism in two tunnel infections and coagulase-negative staphylococci were the causative organisms in three. A gram-positive bacterium caused each of the 13 port pocket infections: coagulase-negative staphylococci in 1; Streptococcus pneumoniae in 1; diphtheroids in 1; and Staphylococcus aureus in 10. Isolated catheter-site infections without bacteremia or fungemia occurred with 60 catheters. Most catheter site infections were caused by gram-positive cocci, accounting for 61% of episodes. Gram-positive bacilli accounted for 8% of pure site infections, yeast caused 11% of infections, and the remaining 20% were caused by gram-negative bacilli. Twenty-four patients with catheter-related bacteremia or fungemia had an associated site infection. No relation existed between cutaneous cultures and blood isolates because the spectrum of those isolates mimicked those of pure device-related bacteremia or fungemia and site infections. Site infections occurred less frequently in ports than in catheters and were all caused by gram-positive cocci, predominantly coagulase-negative staphylococci.

Clinical Outcomes

Twenty-five of 26 ports associated with device-related bacteremia or fungemia as the first infection were sterilized without removal. Subsequently, port pocket infections occurred that required removal of two devices; three devices were further associated with bacteremia but did not require removal. Catheter-related bacteremia or fungemia occurred as the first infection with 264 catheters. One hundred ninety-one catheters were treated successfully and 73 others were removed. Subsequently, a second infection occurred with 83 catheters, and 25 of them were removed. Nineteen catheters were associated with a third or fourth bacteremia, with all devices ultimately removed (10 with the third bacteremia and 9 with the fourth). The organism responsible for the first infection was not predictive of the second organism when repeated infection occurred. One patient died of a catheter-related Klebsiella bacteremia. Sixteen of 17 catheter tunnel infections and 12 of 13 port pocket infections required device removal. Initial site infections occurring in patients with catheters usually responded well to antibiotics, with only 6 of 60 catheters removed. Although 4 of 18 ports were removed for a first site infection, removal of ports for a site infection did not occur statistically more often than removal of catheters. As noted, significantly more catheters (20%) than ports (3%) ultimately were removed for documented infectious morbidity.

Statistical Modeling of Risk Factors for Device-related Infection

Only the first devices inserted in patients who had solid tumors, leukemia, and lymphoma or myeloma were included in the statistical modeling analysis.

Clot occurred in 29 catheters not associated with infection (26 catheter luminal clots, 3 vessel thromboses), in 9 before the first infection (8 catheter luminal clots, 1 vessel thrombosis), and in 16 devices after an infection. Port clots occurred in 17 devices not associated with infection and in 1 port (luminal occlusion) before infection. Forty-three catheters broke and were repaired (19 not associated with infection, 8 repaired before first infection, and 16 repaired after an infection). Neither clot nor repair was related to the probability of device infection.

Our statistical model (with all data) shows a significantly (P < 0.001) higher infection rate in patients who received catheters than in patients who received ports (Table 5). Among patients with solid tumors, young children (<7 years) and young adults (18 to 27 years) had significantly shorter times to infection than did other patients (P = 0.014 and P = 0.04, respectively). However, age was not a significant risk factor for hematopoietic tumors. Among the two hematopoietic groups, patients with leukemia had shorter infection-free times compared with patients with lymphoma or myeloma (P = 0.02). On average, devices inserted in patients with solid tumors had longer infection-free times compared with those in patients with hematologic diseases (P = 0.005). Among patients with hematopoietic diseases, those who were diagnosed with leukemia showed the shortest infection-free times compared with patients with lymphoma or myeloma (P = 0.02).

Kaplan-Meier estimates Figure 1 demonstrate graphically the marked differences in infection-free intervals among devices used in patients with the three tumor types and devices.

View larger version (21K): [in this window] [in a new window]   Figure 1. Comparison of infection-free interval of catheters compared with ports in patients with different cancers along with risk for infections.

Discussion

Top Methods Results Discussion Author & Article Info References

Tunneled, cuffed Silastic catheters, first described by Broviac and Scribner [20] and subsequently modified by Hickman and colleagues [21], were developed for long-term use in the outpatient setting and in recipients of bone marrow transplants. These catheters are used customarily in patients requiring repeated, intensive multilumen access and phlebotomy, continuous infusions of vesicant therapy, or long-term nutritional support [22]. Completely implanted subcutaneous ports offer a viable alternative to right atrial Silastic catheters in some patients and are relatively free of risk for infection when evaluated for their duration in situ [1, 3, 7, 8, 10, 13-15, 17, 18, 23-28]. Central access with ports is achieved in a manner similar to that with external catheters. However, rather than exiting the skin, the catheter is attached to a reservoir within a port that is surgically implanted in a subcutaneous pocket. Right atrial catheters require a protocol of irrigation and exit site care by the patient; they place limitations on patients' lifestyles, and they may alter patients' body image. Ports require no patient maintenance and no changes in activities of daily living. Ports are accessed by insertion of a noncoring Huber point needle through a septum located under the patient's skin and can be used for all of the same functions as a catheter, including blood sampling. Difficulty may be encountered when accessing ports in markedly obese persons as well as when there is hematoma formation in patients with severe thrombocytopenia. Manufacturers suggest that septum integrity is maintained after as many as 2500 to 3000 separate punctures.

Infection remains a significant cause of morbidity for patients with venous access devices. Device-related infection can be divided into exit site infection, tunnel tract or port pocket infection, and device-related bacteremia or fungemia. In a review of pediatric patients receiving chemotherapy, children with long-term, right atrial catheters experienced a 6.4 times higher rate of bacteremia and spent an average of 15.4 more days per year in the hospital for treatment-related complications than did a matched population of patients without catheters [29]. The incidence of device-related infection ranges from 2.7% to 60%, depending on the type of device, the criteria used for the diagnosis of device-related infection, and the patient's underlying disease [8, 19, 30-35]. Catheter-related infection occurs frequently in patients with cancer at our medical center, occurring five times more often than port infections, with a 12-fold difference when assessed by incidence of morbidity per device-day. Differences for bacteremia or fungemia are most dramatic, with a 21-fold difference between catheters and ports on a risk-per-day basis. Because the type of device used in the evaluation was not randomly assigned, a bias was introduced; patients with catheters inserted because of their disease process may have their devices accessed more frequently, may have a longer duration of neutropenia, and may require home care. In a small randomized study of 46 adults with hematologic cancers, ports were shown to be as safe as dual-lumen Hickman catheters [7]. Thrombocytopenia with predisposition to hematoma formation after access, frequent phlebotomy, large transfusion requirements, and multiluminal needs for hydration and antibiotics limit the use of dual-lumen ports in patients with hematologic cancers. In a randomized study of infectious morbidity, ports have been shown to be associated with fewer infections than were catheters in patients with solid tumors [8].

The mechanisms of device-related infection may explain why ports are less likely to be associated with infection than are catheters [34]. Migration of skin flora through the cutaneous insertion site with catheter colonization [34, 36] is supported by our finding and that of others that gram-positive organisms, especially coagulase-negative staphylococci, are responsible for a significant percentage of the cases of device-related bacteremia in patients with catheters. Catheter irrigation with solutions containing 25 µg/mL vancomycin may decrease the frequency of bacteremia attributed to luminal colonization with vancomycin-sensitive bacteria [37], but it was not used in our study sample because of the potential for vancomycin-resistant organisms to emerge.

Addition of the Dacron cuff to catheters allows a significantly longer infection-free duration in situ. However, the cuff is not completely protective: Microcolonies of cutaneous bacteria enclosed in a sheath-like glycocalyx matrix adherent to the catheter surface have been seen on removal from patients with cancer [35]. In neutropenic patients, especially in those with altered mucosal barriers, there may be translocation of endogenous gut bacteria to the catheter [38]. Device-related bacteremias at our center support this mechanism because device infections are caused predominantly by gram-negative enteric organisms that occur in patients with leukemia who have catheters. Compared with catheters, ports are irrigated less frequently, require no home care, and are less prone to environmental or cutaneous contamination when not accessed. These factors may also contribute to the decreased incidence of infections associated with ports.

Initial antibiotic therapy for device-related bacteremia or fungemia should be based on knowledge of the patients' host defenses, presence of neutropenia, and type of device. Gram-negative organisms predominate at our medical center in patients with catheter-related bacteremia or fungemia. The experience at our center is that in most instances, catheters associated with bacteremia or fungemia can be sterilized with antibiotics without removal of the device [9, 39]. Specific predictors of treatment success are not available. Development of one catheter-related bacteremia does not always indicate that a second infection will occur, nor does knowledge of the organism of the first infection help the clinician to identify the second organism when repeated infection occurs. Antibiotic sterilization of ports after diagnosis of device-related bacteremia or fungemia also is usually successful without device removal. Ports are most often placed in patients with solid tumors, and coagulase-negative staphylococci are the predominant pathogens. Clinicians should always consider removing any device from patients in whom there is evidence of septic emboli, refractory hypotension, and persistent culture positivity without eradication of infection shown by decreasing serial colony-forming unit counts, and when the device is no longer required for therapy. Our study does not suggest that vessel or catheter thrombosis is a risk factor for infection. The mechanism of catheter tunnel infection remains uncertain because it can occur on a catheter segment on either side of the Dacron cuff and need not be associated with septicemia or with exit site infection. Skin organisms may be transported along the catheter by capillary action at the time of insertion [32]. Our study and that of Benezra and colleagues [9] show that the incidence of tunnel infection is low, accounting for fewer than 2% of all catheter-related infections, occur significantly earlier than device-related bacteremia or fungemia, can be associated with catastrophic local morbidity or death, and invariably require device removal. Addition of a second, more proximal silver-impregnated cuff to Hickman catheters does not alter the incidence of bacteremias, and because the incidence of tunnel infection is low, a large, multicenter study is needed to prove efficacy against tunnel infection [40]. Port pocket infections, usually caused by gram-positive cocci, suggest direct inoculation or migration of organisms along the accessing needle as the primary mechanism. Regardless of the underlying disease, all port pocket infections have been caused by Staphylococcus aureus and require removal.

Site infection can range from a small infection easily managed with local care and topical antibiotics to a more aggressive type of infection with progression to cellulitis, tunnel or pocket infection, or septicemia. Evidence of erythema and exudate may be subtle in the neutropenic, immunosuppressed patient. If possible, patients with port site infections should have the Huber point needle removed, and an intravenous access remote from the site should be established for delivery of antibiotic agents until all signs of site infection resolve.

For all categories of infection, risk is greater in patients with catheters compared with ports. At our center, implanted ports are used whenever possible. A randomized trial of ports compared with catheters in patients with similar infusion, transfusion, phlebotomy, and device-accessing requirements and on treatment regimens associated with similar degrees and duration of neutropenia still is needed to distinguish the influences of disease and therapy from specific devices with regard to infectious morbidity.