|Home | About | Journals | Submit | Contact Us | Français|
The frequency and risk factors for central venous catheter-related thrombosis (CRT) during prolonged intravenous (i.v.) antibiotic therapy have rarely been reported. The primary objective of this study was to evaluate the frequency, incidence, and risk factors for CRT among patients being treated with prolonged i.v. antibiotic therapy. The secondary objective was to describe the clinical manifestations, diagnostic evaluation, and clinical management. This cohort study was conducted between August 2004 and May 2010 in a French referral center for osteoarticular infections. All patients treated for bone and joint infections with i.v. antimicrobial therapy through a central venous catheter (CVC) for ≥2 weeks were included. Risk factors were identified using nonparametric tests and logistic regression. A case-control study investigated the role of vancomycin and catheter malposition. A total of 892 patients matched the inclusion criteria. CRT developed in 16 infections occurring in 16 patients (incidence, 0.39/1,000 catheter days). The median time to a CRT was 29 days (range, 12 to 48 days). Local clinical signs, fever, and secondary complications of CRT were present in 15, 8, and 4 patients, respectively. The median C-reactive protein level was 95 mg/liter. The treatment combined catheter removal and a median of 3 months (1.5 to 6 months) of anticoagulation therapy. The outcome was good in all patients, with no recurrence of CRT. Three risk factors were identified by multivariate analysis: male sex (odds ratio [OR], 5.4; 95% confidence interval [CI], 1.1 to 26.6), catheter malposition (OR, 5.3; 95% CI, 1.6 to 17.9), and use of vancomycin (OR, 22.9; 95% CI, 2.8 to 188). Catheter-related thrombosis is a rare but severe complication in patients treated with prolonged antimicrobial therapy. Vancomycin use was the most important risk factor identified.
Prolonged high-dose intravenous (i.v.) antimicrobial therapy is necessary during difficult-to-treat infections such as endocarditis, osteoarticular infections, and sepsis secondary to foreign materials. Central venous catheter (CVC) placement facilitates therapy. This procedure is not risk free, however, and certain mechanical (pneumothorax during insertion), infectious (catheter-related infection), or thromboembolic (catheter-related thrombosis) complications may result in significant morbidity and mortality (1,–4).
There are limited data on central venous catheter-related thrombosis (CRT) occurring during prolonged i.v. antimicrobial therapy (2, 5). The existing data are primarily from patients in oncological cohorts for whom the use of repeated cycles of i.v. chemotherapy requires central venous catheter placement (6,–9). In this population, many risk factors for thrombosis have already been identified: hypercoagulability related to prolonged bed rest during repeated hospitalizations, malnutrition, and cancer progression, particularly with metastatic disease. Chemotherapy-related toxicities (6, 7, 9, 10) and male gender have also been noted (9). The prevalence of symptomatic CRT in these populations is variable, ranging from 0 to 30% (6, 10, 11). If there is systematic screening for thrombosis using venography or Doppler ultrasound, the prevalence is notably higher, ranging from 27 to 66% (6, 11, 12). In a prospective observational cohort of 48 consecutive patients with central venous catheter-associated Staphylococcus aureus bacteremia, Crowley et al. (13) found that 71% of them had an asymptomatic thrombosis detected by ultrasonography.
The recommendations for prevention of CRT-related thrombosis by the French National Federation of Cancer Centers underline the importance of catheter positioning: the distal tip must be placed at the junction between the superior vena cava and the right atrium, and anticoagulant drugs are not recommended (14, 15). Multiple studies have shown that the systematic use of low-molecular-weight heparin or long-term use of a vitamin K antagonist did not reduce the risk of thrombosis (6, 14,–16). The treatment of CRT-related thrombosis should be based on low-molecular-weight heparin. Maintenance of the catheter is justified if the catheter is mandatory, functional, in the right position, and not infected and has a favorable clinical evolution under close monitoring. Anticoagulant treatment should be maintained as long as the catheter is present. If the catheter is removed, there are no reliable data on the optimal duration of anticoagulant therapy (14, 15).
As our center manages patients with osteoarticular infections, prolonged i.v. antimicrobial therapy is common. The occurrence of CRT is a rare but serious event, requiring hospitalization, CVC removal, and prolonged anticoagulation therapy.
The primary objective of this study was to evaluate the frequency, incidence, and risk factors for symptomatic CRT among patients being treated for an osteoarticular infection using prolonged i.v. antimicrobial therapy. The secondary objectives were to describe the clinical manifestations of CRT, our diagnostic evaluation, and the clinical management and outcomes of these patients.
This was a retrospective observational cohort study, carried out from August 2004 until May 2010 at a regional referral center for osteoarticular infections in Paris. All patients with bone and/or joint infections were included in the study if they were hospitalized in our unit, had a central venous catheter placed upon admission, and completed a prolonged course (≥2 weeks) of i.v. antimicrobial therapy. The patients were included only once in the study. All of the patients were watched during treatment and seen at the end of i.v. treatment in the outpatient clinic by an infectious disease specialist. A surgical intervention was not required for inclusion.
The following data, prospectively collected, were extracted and analyzed from the center's database: age, gender, length of stay, type of osteoarticular infection, surgical treatment, antimicrobial therapy (route, duration, and adverse events), and CVC-related complications.
In accordance with the usual management, the patient was hospitalized 48 h prior to surgery (if indicated) and a nontunneled CVC was placed in the left (or, rarely, the right) subclavian vein. If subclavian venous access was impossible, the CVC was placed in the right or left internal jugular vein. A subcutaneous port was placed when i.v. antimicrobial therapy for >8 weeks was planned. Antimicrobial therapy was determined according to the center's protocol, which is based on the French Infectious Disease Society guidelines (17). The initial antimicrobial therapy was started in the operating room and always consisted of a combination of 2 agents with at least one i.v. antibiotic. Only the following drugs were administered orally if indicated: fusidic acid, minocycline, levofloxacin, pefloxacin, and linezolid. Vancomycin was given as a high-dose continuous infusion, with target serum concentrations of 30 to 40 mg/liter, as previously described (18). Cefazolin (19), ceftazidime, piperacillin-tazobactam, and clindamycin (20) were also administered via continuous infusion.
The patients were hospitalized for 10 to 20 days, depending on the clinical response and tolerance to therapy. Parenteral antimicrobial therapy was then continued either at home or in a skilled nursing facility. All patients received venous thromboembolic disease prophylaxis with preventive doses of low-molecular-weight heparin, typically enoxaparin sodium (40 mg/day, given subcutaneously). Patients already receiving therapeutic doses of anticoagulation preoperatively continued this treatment postoperatively.
A CRT was suspected in the presence of at least one of the following clinical symptoms and signs (Fig. 1): pain at the CVC site; ipsilateral jugular venous distention; ipsilateral subclavicular fossa, upper extremity, neck, and/or facial swelling; development of collateral venous circulation; fever; an increase in the C-reactive protein (CRP) level after an initial decline; or signs and symptoms of pulmonary embolism. The diagnosis of a CRT was confirmed either by venous duplex ultrasound or by computed tomography angiography. A CRT was confirmed by venous duplex ultrasound if a mural thrombus not limited to the catheter tip was seen or if the vein was noncompressible along with a signal loss. The computed tomography angiography diagnosed a CRT if venous opacification was absent after contrast medium injection into the vein containing the CVC or, additionally, if no opacification was seen after contrast medium injection of the superior vena cava (SVC) or the homolateral internal jugular vein or signs of pulmonary embolism supported the diagnosis.
Charts of patients for whom the diagnosis of a CRT was confirmed were retrospectively examined for the following additional data: personal or family history of thromboembolic disease; catheter type (CVC or port); vessel accessed (subclavian or internal jugular); side accessed (right or left); CVC malpositioning (i.e., on plain chest films, the distal catheter tip can be seen pressing on the wall of the SVC and positioned above the SVC and right atrial junction) (Fig. 2A and andB);B); results of the subsequent coagulation workup; CRT management, including anticoagulation (type and duration), time until catheter removal; modifications of antibiotic therapy; and clinical outcomes of CRT.
A laboratory panel to look for factors predisposing the patient to venous thrombosis was performed for 9 of the 16 patients with CRT. The panel consisted of the levels of proteins C and S, antithrombin III, homocysteine, and antiphospholipid antibodies and mutations in the factor V Leiden and prothrombin G20210A genes.
A case-control study was also performed in order to investigate the roles of certain variables that had not been prospectively determined: catheter site and position, vancomycin daily dose, smoking, and history of thrombosis.
For the cohort study, the demographic and clinical characteristics of the patients who had CRT and the management of their infections were compared with those of control patients using nonparametric tests (Fisher's exact test for categorical variables and the Mann-Whitney U test for continuous variables). These analyses were carried out with Epi Info software version 3.5.3 (CDC).
For the case-control study, 48 control patients in the database were randomly chosen, and their characteristics were compared with those of patients who experienced a CVC thrombosis using Fisher's exact test or the Mann-Whitney U test. Variables associated with a P value of ≤0.25 were included in a logistic regression model (Statistica; StatSoft) for a multivariate analysis of risk factors for a CRT.
During the study period, 892 patients with a joint or bone infection were included in the study. The baseline characteristics, the osteoarticular infections treated, and the pathogens isolated are detailed in Table 1. The antimicrobials given for ≥2 weeks are shown in Table 2.
Twenty-seven catheter related infections were observed, but none was associated with symptomatic CRT.
Symptomatic CRT developed during treatment of 16 bone and joint infections, occurring in 16 patients, equivalent to an incidence of 0.39/1,000 catheter days.
The demographics and clinical characteristics of the patients are shown in Table 3. The median duration from the start of antibiotic therapy to the onset of CRT clinical signs was 29 days (range, 12 to 48 days), with fever in 8 patients and local signs in 15 patients (Fig. 1). One patient had neither fever nor local signs but had lower thoracic pain found to be due to a pulmonary embolism. An increase in CRP, which was returning toward normal, was observed among 15 patients with a median CRP of 95 mg/liter (4 to 231 mg/liter). Eosinophilia was observed in 1 patient. Catheter-related thrombosis complications were observed in 4 patients: 1 patient with a pulmonary embolism, 1 patient with SVC syndrome, and 2 patients with both aforementioned conditions. Peripheral blood cultures in 8 patients remained sterile, as did the cultures of the catheter tips in 11 patients.
Four of the 9 patients who were screened for thrombophilia had abnormal test results. Two patients were heterozygous for the factor V Leiden mutation, 1 patient was heterozygous for the prothrombin G20210A mutation, and 1 patient had a moderately elevated homocysteine level (16.4 μmol/liter).
The CRT required a new hospitalization in 12 patients for a median duration of 5 days (range, 3 to 9 days) or an extension of the hospital stay in 2 patients for 12 and 14 days, respectively. Two patients with a pulmonary embolism and superior vena cava syndrome required intensive care on admission. Anticoagulation was prescribed using therapeutic doses of low-molecular-weight heparin followed by a vitamin K antagonist for a median duration of 3 months (1.5 to 6 months). The precise duration of anticoagulation was determined by the response to treatment. Venous duplex ultrasound scanning was performed at 1 month after treatment onset. If the thrombosis was still present, anticoagulation was withheld, and the ultrasound examination was repeated at 3 months. Only 1 patient with recurrent thromboembolic events received >6 months of anticoagulation. Prolonged ad vitam treatment was decided on, but the patient stopped treatment after 7 months. Systematic catheter removal was performed after a median of 2 days (1 to 5 days) of effective anticoagulation. Among the 3 patients who had only been receiving antimicrobial therapy for 2 weeks, a new CVC was inserted on the contralateral side of the thrombus to complete 6 weeks of i.v. antimicrobial therapy. For all other patients, therapy was continued either using a peripheral i.v. catheter or by the oral route.
A median follow-up of 20 months (range, 6 to 52 months) revealed excellent outcomes for all patients, with no deaths and no recurrent thromboembolic events.
Univariate analysis of prospectively collected variables revealed that male gender and treatment with vancomycin were the main risk factors for a CRT, with odds ratios (OR) of 5.2 (95% confidence interval [CI], 1.2 to 23.0) and 39.5 (95% CI, 5.19 to 300.84), respectively (see Table S1 in the supplemental material).
Comparing the use of vancomycin among patients with CRT versus controls confirmed the previous findings and demonstrated that a high daily dose of vancomycin was a risk factor for thrombosis (median of 3.5 [interquartile range, 3 to 4] versus median of 3 g [interquartile range, 2 to 3.5]; P = 0.04) (Table 4). Patients with improperly positioned catheters had a higher risk of developing CRT (OR of 5.3 [95% CI, 1.6 to 17.9]) than a patient with a properly positioned catheter. Multivariate analysis revealed that these variables were independently associated with the occurrence of a CRT.
There is a paucity of data (2, 5) regarding CRT outside the hematological-oncological literature. We have carried out an observational cohort study in a large population of patients treated for an osteoarticular infection with prolonged i.v. antimicrobial therapy. The incidence of symptomatic CRT among 892 patients treated for osteoarticular infections with prolonged i.v. antimicrobial therapy by CVC was 0.39/1,000 catheter days. This value is close to the 0.48/1,000 catheter days observed by Pulcini et al. (2) in patients treated for bone and joint infections. The incidence of this complication in our study was low, but it caused significant morbidity, increasing the length of stay and causing a pulmonary embolism and/or superior vena cava syndrome in 4 of 16 patients. These complications occurred at rates similar to those described in the literature (6, 7, 21).
Three CRT risk factors were identified in our population. The first was male gender. The influence of gender was described in two studies, although with contradictory results. Morazin et al. observed in a prospective monocentric study of 5,447 central venous catheters placed in patients with cancer (of whom half had breast cancer) that female gender was an independent risk factor for CRT (11). The authors pointed out, however, that 81.5% of the population studied was female and that not analyzing the individual chemotherapeutic agents used in breast cancers could have confounded the results. In another recent retrospective cohort study (9) in 400 patients undergoing chemotherapy using a subcutaneous port, the conclusion from the multivariate analysis was that male sex (OR of 2.17) was an independent risk factor for CRT. These results are in line with those of our study, although the number of instances of CRT was low in our population, so this point must be validated in a larger population.
The second risk factor found, CVC malpositioning, has already been observed in many studies (6, 7, 11, 22). Certainly, one could hypothesize that a CVC tip pressing against the vessel wall and the toxicity of the antimicrobial agent being infused could cause endothelial trauma. Other prior studies also cited the specific vein accessed as a risk factor. For example, CVC placement in the femoral vein has been shown to have a higher risk for development of a CRT than placement in the subclavian vein, which in turn has a higher risk than that for placement in the jugular vein. Additionally, the left subclavian position has shown to carry a higher risk than the right position (6, 7). Prior to this study, a left-sided subclavian CVC was placed in our center due to local habits, but we now favor right-sided placement.
The third and major risk factor for CRT uncovered in our study was the use of vancomycin. Fifteen out of 16 patients developing CRT were treated with high-dose continuous intravenous vancomycin. Even though the venotoxicity of vancomycin has been known for a long time (23,–25), publications on vancomycin-associated CRT are rare. A case-control study, including 400 patients with a peripherally inserted central catheter (PICC) concluded that the infusion of antibiotics, particularly vancomycin, was a risk factor for venous thrombosis (5). A recent randomized clinical trial on vancomycin administration by a novel midline catheter versus a PICC noted no thromboembolic event (26). Leroy et al. reported a case of extensive thrombophlebitis with gas associated with continuous vancomycin infusion through a central venous catheter (27). We systematically use prolonged courses of high-dose vancomycin in continuous i.v. infusion for the treatment of osteoarticular infections secondary to methicillin-resistant staphylococci. We adhere to this treatment regimen because of vancomycin's time-dependent antibacterial activity and our good clinical outcomes in difficult-to-treat infections. As previously published, our prospective study of 60 patients treated with high-dose continuous intravenous vancomycin infusion and another antibiotic for prosthetic hip infection resulted in only two relapses, with 41 (68%) patients considered cured after 2 years of follow-up (18). Furthermore, continuous i.v. infusion of vancomycin has several practical advantages: it simplifies nursing care, is very convenient for outpatient antibiotic parenteral therapy (OPAT), and allows easy drug dose monitoring and adjustment. However, by keeping the endothelial wall in continuous and prolonged contact with the agent, this modality for administration of vancomycin may certainly allow for vancomycin-associated venotoxicity. In fact, among our patients, the risk of CRT increased with a high daily dose of vancomycin (median of 3.5 g [interquartile range, 3 to 4 g]) and the duration of vancomycin treatment, and no signs of CRT were observed in patients with <12 days of treatment. Drouet et al. showed in a recent in vitro model that continuous vancomycin infusion induced greater cell toxicity than intermittent infusion at doses higher than 1 g/day (28).
Catheter-related infection was not associated with CRT in our study.
It is important for doctors treating patients with prolonged i.v. antimicrobial therapy to know the signs of central venous catheter-related thrombosis, allowing for early diagnosis and treatment in order to avoid secondary complications. The clinical and laboratory signs observed most often in our study were the development of local signs related to the CRT (pain and swelling) and an increase in the CRP level, despite an initial good response to therapy. These signs appeared with a median delay of 1 month and not before 12 days of treatment. In a recent clinical trial of short-term vancomycin treatment (<6 days), no thromboembolic event was observed (26) Complementary exams confirming the diagnosis are venous duplex ultrasonography and computed tomography angiography, especially when a pulmonary embolism is suspected.
A hypercoagulability workup performed in 9 patients revealed an abnormality among 4 patients. Three of these patients developed a severe complication of CRT. These data raise the question of the presence of a latent risk factor encouraging the development of CRT among certain patients and the possibility of early detection. The patient with extensive thrombophlebitis associated with continuous vancomycin infusion reported by Leroy et al. was also heterozygous for the factor V Leiden mutation (27).
The management of our patients included in all cases effective anticoagulation and CVC removal. The systematic removal of the CVC remains controversial in the literature (14,–16). Among our patients the retention of the central line was rarely needed as the thrombosis occurred most frequently near the end of i.v. treatment. The central line was replaced on the opposite site of the thrombosis in only 3 patients in order to continue i.v. therapy. This decision must be made on a case-by-case basis, taking into account the infection, alternative therapies, the thrombosis severity, and the clinical progression during anticoagulation therapy. Reliable data on the optimal duration of anticoagulation treatment after catheter removal are lacking (14, 15). We treated our patients for a median duration of 3 months (range, 1.5 to 6 months), depending on clinical and venous duplex ultrasound evolution during treatment.
The limitations of our study are its retrospective, monocentric characteristic with population-specific patients, including only bone and joint infections with a high proportion of chronic infections, our preferred method of administering vancomycin in a high-dose continuous fashion, and a low number of thromboembolic events. We cannot exclude the possibility that some of the control patients have had asymptomatic thrombosis. The conclusion drawn on the risk factors should take into count these elements. It is difficult to generalize our conclusions to other populations requiring long-term antibiotic therapy such as infective endocarditis or intensive care unit patients who have more acute and septic clinical presentations and frequent bacteremia or additional risk factors for thrombosis (bed rest and immobility).
In conclusion, central venous catheter-related thrombosis during long-term i.v. antimicrobial therapy is a rare but serious event. In our experience, this complication occurs after >12 days of treatment. The principal risk factor identified in our study is the use of high-dose continuous vancomycin. CRT prevention also relies on making sure the central venous catheter remains in the proper position. The prevention of CRT complications relies on early diagnosis and management.
We thank Alysa Krain for editorial assistance.
S. Guillet collected the data and contributed to manuscript writing. V. Zeller collected the data, designed and analyzed the study, and contributed to manuscript writing. V. Dubée performed the statistical analyses and contributed to the critical writing and revising of the scientific content of the manuscript. F. Ducroquet, N. Desplaces, M. H. Horellou, S. Marmor, and J. M. Ziza contributed to the critical writing and revised the scientific content.
The authors declare no competing financial interests.
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/AAC.00700-15.