The incidence of central venous catheter (CVC)-related deep venous thrombosis (DVT) varies considerably in the ICU, depending on the population included and the detection methods 12. The rate of subclavian central venous catheter (SCVC)-related DVT found by routine Doppler ultrasound in the literature ranges from 4 to 67% with a mean incidence of 30%, of whom only 2% were symptomatic 3. Several thrombotic risk factors have been identified and are related either to the patient's condition or to the catheter 4.

With regard to insertion sites, a femoral insertion is known to increase the risk of infection or thrombosis. The incidence of CVC-related DVT in the upper extremities is estimated to be 2 to 6% when symptomatic and 11 to 19% when asymptomatic 5. In their conclusion, the experts of the Cochrane review group stated that it is probably reasonable to prefer subclavian access to femoral access 6. The CVC equipment may also be a risk factor. The incidence of thrombosis is higher with a CVC made of polyethylene than one made of polyurethane 7. Most CVC are now composed of polyurethane. In our hospital, two brands are available and both are made of polyurethane. One CVC is made of aromatic polyurethane (Ar) and is the most commonly used. Our Medical Devices Vigilance has collected reports about thrombosis associated with its use. Another brand is made of aliphatic polyurethane (Al). This Al CVC is marketed as being more thermosensitive than Ar, so its thrombogenic potential is theoretically lower.

Trauma patients have specific risk factors and the use of anticoagulant is often problematic in the initial phase because of the risk of bleeding. Studies in trauma patients have found widely varying rates of overall DVT from 0.36 to 58% depending on the patients studied and the means used to diagnose DVT 8910. However, most studies in the specific setting of trauma patients are not limited to the ICU.

The aim of this study was to identify SCVC-related DVT risk factors in ICU trauma patients and to determine whether the type of polyurethane used could be an independent risk factor. The study was performed regardless of the patient's clinical signs with systematic venous ultrasound screening.

Four hundred and eighty-six patients were screened during 21 months. Two hundred and ninety-five patients were not included for various reasons: cancer history, associated anticoagulation, CVC set up before transfer to our hospital, CVC in a femoral vein, estimated required duration of CVC <5 days, or logistic reasons (impossible to follow more than 10 patients at the same time on a weekly basis). Five patients were secondarily excluded (early death or early CVC removal). Finally, 186 patients were included in the analysis (84 in the Ar group and 102 in the Al group). Most were male (80%). Known DVT risk factors related to the patients' medical history were uncommon (Table 1). The median ISS was 30, meaning a majority of polytraumatized patients. The rate of severe brain-injured patients was high. One-half of the included patients had an initial Glasgow Coma Score <8, 37% of patients had a maximum head AIS equal to 5, and 21% developed severe intracranial hypertension (ICHT), one-half of them needing barbiturates to control their ICHT. The median length of stay was 20 days (IQR: 13 to 31) and the median length of catheterization was 13 days (IQR: 8 to 19). The CVC were mostly three lumina (16G/18G/18G), with a diameter of 7 Fr and a length of either 15 or 20 cm, depending on the placement side, respectively right or left.

The incidence of subclavian SCVC-related DVT concerned one-third of these patients. The hypothesis that Ar might lead to more thrombosis than Al is not confirmed. Only 186 patients were included in the study instead of the 224 originally planned. The clinical impress during the study was equivalence between the two products. An independent statistical team was consulted and found that the difference in SCVC-related DVT incidence between the Ar group and the Al group was as low as 2%. The number needed to have a good probability to demonstrate that this difference was significant should have been too high. Because such a small difference was not expected to have any impact on clinical practice, we decided to stop the study.

Other SCVC characteristics, such as multiple lumina and left-side SCVC placement, also did not appear associated with a higher risk of SCVC-related DVT in trauma patients, although they have been reported in other studies 111213. Parenteral nutrition was associated with a higher rate of SCVC-related DVT. Propofol seemed to have an opposite effect. Parenteral nutrition may have some impact on coagulation parameters 14. We did not find any report of a protective effect of propofol. In our study these relations might be due to a statistical bias because patients who received parenteral nutrition had a higher frequency of ICHT while those sedated with propofol were the least severely injured. These two factors did not remain significant in the multivariate analysis. A CVC in the internal jugular vein ipsilateral to the SCVC was an independent risk factor for thrombosis. These jugular catheters were mostly set up to monitor cerebral blood saturation, which required retrograde cannulation into the jugular bulb. The analysis confirmed the high risk of not centrally positioning the tip of the subclavian CVC 815. The CVC placement conditions in our emergency room may also explain the high rate of SCVC-related DVT 16.

Our results also confirm the recent report that massive transfusion (defined as >10 packed red blood cells. seems to favor SCVC-related DVT, although no threshold has been identified to date 17. In this study there was no evidence of a relationship between infection during catheterization time and SCVC-related DVT. The infection rate was high but consisted mostly of pneumonia. Bacteremia was rare and occurred only once in association with SCVC-related DVT and that very low incidence may have a bearing on our results. Raad and colleagues found a relationship between mural thrombosis found at postmortem examination and catheter-related septicemia, but catheterization was of long duration before the examination (mean 63 days) and all patients had cancer 18. Timsit and colleagues found that the relative risk of catheter-related septicemia was threefold higher when a thrombosis was present in critically ill patients 19. The duration of catheterization was comparable with our study while their rate of colonization was >20%. However, 35% of their patients were admitted with an ongoing infectious disease 19. Our DVT occurred very early and seemed to be associated more with inflammation due to trauma than to intercurrent infection.

The patients included in our study had both a SCVC and a major trauma. The following risk factors were identified in the various studies on trauma patients: older age (>40 or >60 years old), transfusion, surgery, spinal cord injury, pelvic fracture, head injury, and mobility score. Our patients were particularly vulnerable to DVT. First, the patients were in an ICU for severe polytrauma and during the first week they were all immobilized by sedation to allow iterative surgical treatment and mechanical ventilation. Secondly, 70% of these patients had severe brain injury. A brain injury was demonstrated to increase threefold to fourfold the risk of developing a clinically diagnosed DVT, whatever the pharmacological prophylaxis, in two large cohorts of patients admitted to trauma centers 820. In a study on 677 moderately and severely brain-injured patients monitored by duplex scans, a femoral DVT was present in 24% and an upper extremity DVT in 16%. The prevalence of SCVC was not detailed and the authors did not study incidence by severity level of the brain injury 21. In our patients, brain injury without ICHT was not associated with a higher rate of SCVC-related DVT, whereas ICHT increased the risk fivefold and this association had not yet been demonstrated. This increase in SCVC-related DVT incidence in case of ICHT may be related to coagulation disorders that occur after head injuries 2223. Coagulation activation from cerebral blood has been demonstrated very early after a head trauma 24, probably induced by tissue factor release from the injured brain. Sustained coagulopathy is more frequent in patients with tomographic signs of intracranial hypertension (compressed cisterns or midline shift) 25. In other studies, fibrin degradation products levels were demonstrated to be related to the severity of brain damage 22.

SCVC-related DVT is a particularly underestimated complication. Clinical signs of SCVC-related thrombosis are neither sensitive nor specific 252627. In polytraumatized patients the diagnosis of thrombosis is especially difficult in view of the edema related to their orthopedic injuries. In a study of 208 patients with femoral or subclavian CVC screened by ultrasound in an ICU, 33% of the patients developed CVC-related DVT but none were symptomatic 19. This was the case in our study. The clinical screening therefore appears ineffective. On the other hand, ultrasound screening is a noninvasive efficient bedside tool, as already demonstrated in orthopedic surgery 2829. The clinical importance of asymptomatic DVT remains to be demonstrated. PE occurred in 0 to 17% of cases after an upper-arm DVT in two studies 3031. In both studies, however, 10 to 20% of patients had a neoplasm and the responsibility of the catheter is thus difficult to determine. In the RIETE registry of patients with symptomatic upper-extremity DVT or PE, recurrent PE occurred in 3.8% of patients with cancer and 1.6% in those without 32. In all these cases, however, the DVT was symptomatic.

The second complication may be a post-thrombotic syndrome. Prandoni and colleagues found a 17% cumulative incidence of post-thrombotic syndrome in patients with vein catheter at 1 year but their study concerned only symptomatic upper-arm DVT 33. Few studies have reported data on the incidence and the range of estimates is wide, reflecting the heterogeneity of patients studied 34. Whether SCVC-related DVT patients have the same risk of developing long-term post-thrombotic syndrome as other DVT patients is not clear. All of these complications have been studied in symptomatic patients. The complications that may be linked to asymptomatic SCVC-related DVT remain to be confirmed. In Malinoski and colleagues' study, DVT was diagnosed by systematic ultrasound examination and the rate of PE after catheter-associated upper extremity DVT was 1.3% despite generalized thromboprophylaxis 35. In our study, a diagnosis of PE was confirmed twice. One PE occurred 20 days later than the occurrence of the SCVC-related DVT and was associated with a lower-limb DVT; both DVTs had disappeared when the patient left the ICU. The involvement of the SCVC-related DVT was uncertain. In the second case, the patient was free of DVT at the end of the ICU stay and PE occurred later. The DVT was found near the site of SCVC placement. Three other patients experienced brief episodes of hypoxia when the SCVC was removed. Although these events may seem anecdotal, we think that clinicians should keep in mind the potential harmfulness of SCVC-related DVTs.

High-risk patients should benefit from improved preventive measures. Clinicians are hesitant to use chemoprophylaxis because of the risk of intracranial hemorrhage. LMWH has proved effective as pharmacologic thromboprophylaxis in ICU trauma patients 2917. However, a recent study demonstrated that pharmacologic thromboprophylaxis use is associated with a 13-fold increased odds of further hemorrhage progression in patients whose follow-up computed tomography within 1 day of admission showed an intracranial hemorrhage increase 36. Patients at high risk of SCVC-related DVT but presenting a high risk of bleeding could thus not benefit from the anticipated pharmacologic thromboprophylaxis. The benefit of therapeutic anticoagulation was not demonstrated in our patients who underwent examinations after SCVC-related DVT had been diagnosed. At the time of the study, we were very careful about prescribing anticoagulants. When prescribed, the decision was made on the basis of the extent of the thrombosis, and the severity of intracranial, hepatic or pulmonary hemorrhagic contusions. The doses were possibly underestimated in patients with a severe inflammatory condition. However, in the study by Malinoski and colleagues on surgical and trauma patients, anticoagulation was not associated with a higher rate of resolution 35. Our decision to remove the SCVC was also taken on a case-by-case basis according to the extent of the thrombosis and according to the technical possibility of using another infusion mode. One of our main considerations was not to send a significant clot into the blood flow.

These findings have led us to implement preventive measures to decrease SCVC thrombosis. An educational film about good practices in CVC insertion has been made, which is now available for both junior and senior practitioners. Second, any SCVC whose tip is misplaced in the internal jugular vein or in the innominate vein is removed. The use of ultrasound for SCVC placement decreases the number of punctures and the failure rate, and is now encouraged 3738. Third, Duplex ultrasound examinations are now performed by most physicians in our service. They are systematic in all patients at the end of the first week since our study demonstrated that 60% of DVTs occurred within this delay, and then before removing the SCVC. Last, the time of chemoprophylaxis is discussed earlier, on a case-by-case basis, taking into account the risk of intracranial bleeding propagation. At the beginning of the study, thromboprophylaxis was prescribed late. Nowadays, we assess the risk-benefit ratio of beginning prophylaxis much earlier.

This study demonstrates that SCVC-related DVT is a frequent and early complication in severe polytraumatized patients, and its incidence does not depend on the type of polyurethane. SCVC-related DVT concerns one-third of patients and occurs early within the first week. The main risk factors were intracranial hypertension, a misplaced SCVC, or a massive transfusion. SCVC-related DVT is often underdiagnosed because it is mostly asymptomatic. Further large prospective studies are required to confirm our findings, to identify other risk factors, to study long-term morbidity linked to these SCVC-related DVTs and to evaluate the cost-effectiveness of systematic screening by venous ultrasound in these patients in whom pharmacologic thromboprophylaxis may be hazardous.

• The incidence of SCVC-related DVT does not depend on the kind of polyurethane.

• The main risk factors are intracranial hypertension and misplaced SCVC.

• One-half of CVC-related DVTs were found at the end of the first week after trauma.

• Routine coagulation tests do not help to diagnose these DVTs.

• The long-term consequences of these asymptomatic SCVC-related DVTs would be investigated to determine their clinical impact.

AIS: Abbreviated Injury Score; Al: aliphatic polyurethane; Ar: aromatic polyurethane;CVC: central venous catheter; DVT: deep venous thrombosis; ICHT: intracranial hypertension;IQR: interquartile range; ISS: Injury Severity Score; LMWH: low molecular weight heparin; PE: pulmonary embolism; SCVC: subclavian central venous catheter.

The authors declare that they have no competing interests.

LP, FM, VC, JB-B and GF contributed to the study conception. AG, LP, FM, VC and FS contributed to the data acquisition, and writing. LP and VC performed ultrasonographic examinations. CP, AL and J-FC contributed to clinical examinations of patients to determine the probability of thrombosis. FM contributed to the data analysis. All authors have made substantial contributions to the conception, the design and interpretation of results, have been involved in drafting the manuscript or revising it critically for important intellectual content, and have given final approval of the version to be published.