Abstract
Background
Venous thrombosis is a common occurrence in cancer patients, developing spontaneously or in combination with indwelling central venous devices (CVD).
Purpose
To analyze the multidetector CT (MDCT) prevalence, appearance, and significance of catheter-related thoracic venous thrombosis in oncologic patients and to determine the percentage of thrombi identified in the original reports.
Material and Methods
Five hundred consecutive patients were considered. Inclusion criteria were: presence of a CVD; availability of a contrast-enhanced MDCT; and cancer history. Exclusion criteria were: direct tumor compression/infiltration of the veins; poor image quality; device tip not in the scanned volume; and missing clinical data. Seventeen (3.5%) out of the final 481 patients had a diagnosis of venous thrombosis.
Results
Factors showing the highest correlation with thrombosis included peripherally-inserted CVD, right brachiocephalic vein tip location, patient performance status 3, metastatic stage disease, ongoing chemotherapy, and longstanding CVD. The highest prevalence was in patients with lymphoma, lung carcinoma, melanoma, and gynecologic malignancies. Eleven out of 17 cases had not been identified in the original report.
Conclusion
CVD-related thrombosis is not uncommon in cancer patients and can also be observed in outpatients with a good performance status and a non-metastatic disease. Thrombi can be very tiny. Radiologists should be aware of the possibility to identify (or overlook) small thrombi.
The association between cancer and thrombosis is well-established. It results from several factors such as production of procoagulating substances by tumor cells, direct vascular compression or infiltration by tumor masses, reduced mobility of cancer patients, surgery, and chemotherapy (1, 2). Deep venous thrombosis is observed in up to 43% of cancer patients under treatment, depending on the type of neoplasm and on the type of chemotherapy agent (3); pulmonary embolism is incidentally detected in up to 5% of patients with malignancies undergoing staging or follow-up computed tomography (CT) (4–6). On the other hand, occult malignancies are incidentally discovered in 13% of hospitalized patients undergoing pulmonary CT angiography and CT venography because of suspected thromboembolic venous disease (3) and 70% of the patients with incidentally detected pulmonary emboli on routine multidetector row CT (MDCT) do have cancer (7).
Cancer patients bearing a central venous device (CVD) are at increased risk of developing central venous thrombosis and pulmonary embolism. Risk factors for thrombosis are believed to include the following: type of malignancy (lung cancer and lymphoma, in particular), type of chemotherapy, type of CVD itself (catheter material, high catheter-to-vessel size ratio, etc.), insertion site (left-sided catheters clot more frequently than right-sided ones), and tip location (increased risk for catheter tips high in the superior vena cava) (2, 8).
The purpose of this retrospective study was two-fold. To analyze the MDCT appearance and prevalence of catheter-related thoracic venous thrombosis in cancer patients and to determine the percentage of unreported thrombi also discussing the possible reasons for missed detection.
Material and Methods
Patients
Five hundred consecutive patients were initially considered between July 2009 and December 2009 in a single cancer centre. Institutional Review Board approval was obtained, with particular reference to access to sensitive health data. Given the retrospective nature of the study, also patient consent was consequently waived.
Inclusion criteria were: presence of an indwelling CVD; availability of a contrast-enhanced multidetector CT study; and cancer history. Planned exclusion criteria were: direct tumor compression/infiltration of the veins (six subjects excluded for this reason); poor image quality (10 subjects); device tip not included in the scanned volume (no subject excluded for this reason); concurrent chest radiation therapy (one subject); and major clinical data unavailable (two subjects). The main reason for poor image quality-related exclusions was an inadequate contrast enhancement of the venous portion with the catheter. This was usually due to contralateral contrast medium injection combined with an early acquisition (i.e. not enough time given for enhanced blood to reach the veins containing the catheter). Two reviewers decided by consensus which cases had to be excluded.
CVD type, access, and tip location were taken in consideration. CVD types included peripherally-inserted CVDs, tunneled CVD, and implanted ports. Venous access was distinguished into right and left brachial vein, right and left subclavian vein, and right and left jugular vein. The location of the CVD distal extremity was classified as right and left subclavian vein, right and left jugular vein, right and left brachiocephalic vein, superior vena cava, inferior vena cava, right atrium (including cases close to the atriocaval junction), and right ventricle.
As to patient features, patient type (in- or outpatients), concurrent use of anticoagulants, and performance status were recorded. The East Cooperative Oncology Group (ECOG) performance status system was chosen, categorizing between performance status 0 (patient is fully active, able to carry on all pre-disease performance without restriction), 1–2 (1, patient is restricted in physically strenuous activity but is ambulatory and is able to carry out work of a light or sedentary nature; 2, patient is ambulatory and capable of all self-care but unable to carry out any work activities), 3 (patient is capable of only limited self-care and is confined to bed or chair more than 50% of waking hours), and 4 (patient is completely disabled, cannot carry on any self-care, and is totally confined to bed or chair) (9).
As to the underlying malignancy, the following elements were considered: tumor in metastatic stage or not, patients still on chemotherapy (including patients at less than one month from treatment), patients at more than one month from treatment, chemotherapy-naïve patients.
Time from CVD placement and CT study was also considered, distinguishing between short-time (<1 month), medium-time (1–6 months), and long-time (>6 months) CVDs. The number of patients with the CVD in place at the moment of the CT study was considered as well any case of catheter malfunctioning. Thrombus size was electronically measured, distinguishing <5 mm, 5–20 mm, >20 mm, and occlusive thrombi (full-lumen thrombosis, all lengths). Finally, the number of cases with pulmonary embolism signs on CT was retrospectively evaluated.
CT study
A variety of contrast-enhanced CT protocols were included in our series. These included chest studies (45 cases), neck and chest studies (11 cases), and whole body studies (425 cases). In five cases concomitant neck and chest CT was performed with chest acquisition first, followed by the neck. In the remaining six cases neck acquisition was obtained before the chest. Among the 425 whole-body studies, 69 included the neck (scanned before the chest in 55 cases and after the chest in 14) while 356 only included the abdomen, with possible multiphase acquisitions (chest scanned before the abdomen in 246 cases and after the arterial-phase abdomen acquisition in 110 cases). In most cases the acquisition was performed with the patient's upper limbs positioned above the head and only in a few cases scans were obtained with one or both upper limbs positioned the trunk.
All CT studies were performed with a Brilliance 16 scanner (Philips Medical Systems, Eindhoven, The Netherlands) by a group of radiologists and technologists experienced in cancer imaging. CT examinations were performed before and after IV injection of an iodized, non-ionic contrast medium, iomeprol (Iomeron, Bracco, Milan, Italy). The contrast medium was administered as a 100–150 mL bolus via a power injector, at a 350 mgI/mL concentration and at a 2–4 mL/s injection rate. A high collimation was employed (1–3 mm acquisition thickness). Depending on the kind of CT study, the images were obtained with a start delay ranging from 20–80 s from the beginning of the contrast injection, with or without a bolus tracking function. In the majority of the cancer patients enrolled the CT study was performed for a whole body survey, and consequently a standard 70-s delay was employed. The images were reconstructed with both body standard and lung kernels.
Image evaluation
The CT images were reviewed by two radiologists, blinded to the previous reports. A third opinion had originally been planned in case of major disagreement but this never happened in practice. The two radiologists, who reviewed the exams by consensus, were not involved in the original interpretation of CT studies and were not aware of patient data (except for those directly recognizable on the CT images themselves). In case of repeated CT exams over time, only the first MDCT study obtained after the CVD placement was considered for each patient: this was done to simplify the subsequent data analysis. The two reviewers evaluated the CT studies at a workstation monitor (Extended Brilliance Workspace, ver.4.3.0.1.5000, Philips Medical Systems, Eindhoven, The Netherlands). They were free to use any desired method to improve image interpretation, including variation in reconstruction thickness, image reformation in planes different from the axial one, change in the window setting, and measurement of attenuation values. The reviewers defined the presence and position of the thrombus and electronically measured its length. Then a distinction was made between thrombi detected at initial CT imaging and described in the original report and thrombi only observed at retrospective analysis.
Statistical analysis
A non-parametric Fisher exact test was used to compare groups and calculate the relative risk per each factor considered. The relative risk was expressed as the probability of rejecting the null hypothesis that the factor had no association with thrombosis. Values lower than 1.0 were considered to indicate no relationship while values higher than 1.0 were regarded as markers of a strong association (10). The Fisher exact test was performed using the hypergeometric distribution and the statistical analysis was carried out using Analyse-it v.2.2 (Analyse-it Software, Ltd., Leeds, UK).
Results
The final group included 481 patients. There were 216 men and 265 women aged 18–89 years (mean 63). The 481 primary tumors included breast carcinoma (77 cases), colorectal carcinoma (57 cases), lung carcinoma (55 cases), hematological malignancy (including Hodgkin disease, non-Hodgkin lymphoma, or leukemias, 44 cases), upper gastrointestinal tract carcinoma (40 cases), pancreatic tumors (adenocarcinoma or neuroendocrine tumor, 39 cases), ovarian carcinoma (31 cases), soft-tissue sarcoma (29 cases), uterine carcinoma (17 cases), transitional cell carcinoma (16 cases), head and neck carcinoma (12 cases), melanoma (11 cases), bone sarcoma (10 cases), gallbladder and biliary tract carcinomas (nine cases), renal carcinoma (five cases). The remaining 29 cases had a variety of rare tumors (27 cases) or tumor of unknown origin (two cases).
Seventeen (3.5%) out of 481 patients had a retrospective diagnosis of venous thrombosis (Fig. 1). Three out of these 17 subjects were symptomatic (upper limb swelling in two cases, and pain and CVD malfunctioning in the remaining one). One of these three patients had a color Doppler diagnosis of venous thrombosis.
A 61-year-old man with pancreatic carcinoma, liver metastasis and peritoneal carcinosis. (a) Axial and (b) coronal re-formatted MDCT images show an extended thrombus around a right-sided port. The clot is seen along the middle portion of the catheter but not around its distal extremity. Note the subcutaneous portion of the catheter (short arrow in A)
The CVD had been put in place 6–645 days before CT imaging (mean 166 days). Time from CVD placement was <1 month for 83 patients, 1–6 months for 326 patients and >6 months for 72 patients. In 309 cases the device was being used for chemotherapy and working correctly in the period of the CT study while in the remaining 172 patients the CVD was not in use (many ports are left in place for a variable time after the end of chemotherapy). One hundred and five patients were currently receiving a prophylactic treatment with anticoagulants at the moment of the CT study and 12 others (including the three symptomatic subjects) were receiving an anticoagulant therapy. Thrombus length was <5 mm in four out of 17 cases, between 5–20 mm in five cases, and >20 mm in four cases (Fig. 2). In the three remaining cases the thrombosis was occlusive.
A 69-year-old, non-hospitalized woman during follow-up for abdominal lymphoma (6 months after complete response to treatment). Tiny clot (arrow) seen attached to a tunneled catheter. The thrombus had been overlooked at initial evaluation
Factors best correlating with thrombosis included right brachial vein access (obviously only employed for peripherally-inserted central catheters, relative risk 2.8), ongoing chemotherapy (relative risk 2.6), performance status 3 (relative risk 2.5), hospitalized patient (relative risk 2.3), peripherally-inserted CVD (relative risk 2), right brachiocephalic vein tip location (relative risk 2), metastatic stage disease (relative risk 1.9), and longstanding CVD (relative risk 1.75) (Table 1). On the other hand, conditions rarely found in combination with thrombosis included jugular vein access, right atrial location of catheter tip (including, as described above, those cases with the extremity of the catheter at the atriocaval junction), outpatient status, performance status 0, non-metastatic tumor stage, no history of concurrent or previous chemotherapy, short time intervening between the CVD placement and the CT scan.
Correlation between the various parameters considered and the presence or absence of thrombosis
N/A= not applicable
Some histotypes carried a higher risk of CVD-related thrombosis than others. The number of cases for many primary sites was too low to allow an effective statistical analysis. Nevertheless, a higher relative risk was found in combination with melanoma (2.7), hematological malignancy (2.4) lung carcinoma (2.4), ovarian carcinoma (1.9), and uterine carcinoma (1.7) (Table 2).
Correlation between the tumor of origin and the presence and absence of thrombosis. The relative risk is calculated for histotypes with a sufficient number of cases
N/A= not applicable
The thrombus length was <5 mm in six out of 17 cases, between 5–20 mm in four cases, and >20 mm in three cases. The thrombosis was complete, involving the full lumen, in the remaining four cases. Only six of the 17 cases of thrombosis (2.3% of 481 patients) had originally reported and all these thrombi were over 5 mm. There was no complete occlusion among the 11 non-diagnosed cases. Pulmonary emboli were retrospectively recognized in 33 patients, three with a concomitant CVD-related thrombosis and 30 without any CVD thrombosis.
Discussion
A CVD is an indwelling catheter with its distal extremity placed into a large vein. Basically, there are three types of lines. A peripherally-inserted central venous catheter is a short-term device inserted in a peripheral vein and then advanced toward the heart. A tunneled catheter is a surgically-placed long-term device passing subcutaneously with only the end of the catheter brought through the skin. An implanted port is a small appliance installed entirely under the skin; a catheter connects its own port to a vein (1, 2, 11–13). Ideally, all devices are inserted with the tip down to the junction between the superior vena cava and the right atrium (12, 14).
CVDs can undergo various complications, both at the moment of their placement and in the subsequent period of time. The early and late complications include pneumothorax, hemorrhage, arrhythmia, malposition, detachment with embolization, malfunction, infection, drugs extravasation, catheter lumen and tip thrombosis, and catheter-related venous thrombosis (2). Besides causing catheter malfunctioning, CVD-related thrombosis increases the risk of infection and can cause postphlebitic syndrome in 15–30% of the cases and pulmonary embolism in 11% of the cases (half of which are asymptomatic) (2). In most patients the catheter surface is coated by a fibrin sheath within 24 hours from insertion and most thromboses occur within one month from placement (1, 2, 8). When the CVD-associated thrombosis becomes symptomatic or thrombi are detected incidentally during cancer staging or follow-up a treatment with low-molecular weight heparin is started, and continued until the catheter is left in place (14).
The radiologist plays a major role in pre-placement anatomical evaluation, in CVD positioning, and in complication recognition and management (12, 15, 16). By using venography, 12–74% (approximately 41% on average) of the patients with CVDs show evidence of thrombosis (2). Particularly, two venography studies have shown venous thromboembolism in 37% (17) and in 62% (18) of the cancer patients evaluated three months from placement of implanted port and not receiving any pharmacological prophylaxis (this percentage is 10% for subjects receiving warfarin and 6% for those receiving low-molecular weight heparin – although recent guidelines do not support anymore the use of anticoagulants for thrombosis prevention) (14). Color Doppler sonography showed upper extremity deep venous thrombosis in 12–44% of asymptomatic cancer patients with a CVD and in 7–48% of those with a clinical suspicion (19). Color Doppler sonography has proven sensitive (78–100%) and specific (82–100%) for CVD-related upper extremity thrombosis (20, 21) but it is unable to detect central thrombi.
Only few papers have investigated the role of CT. Forneris reported on the use of helical CT in 18 hemodialyzed patients with a complicated indwelling internal jugular vein catheter (21). Cannon et al. reported a case of CT-diagnosed superior vena cava thrombosis developing a short period after central venous catheterization and surgery for adenocarcinoma of Vater ampulla (22). Aggarwal and McCauley described a case of superior vena cava syndrome diagnosed by CT in a patient with tunneled CVD-induced thrombosis (23). Knutstad reviewed the role of imaging in the diagnosis and interventional management of CVD-related complications, underlying that the diagnostic approach includes radiography, fluoroscopy, CT and ultrasound (15). Díaz published a pictorial review on the normal and abnormal MDCT findings in cancer patients with CVDs (24). We are not aware of any article systematically focusing on CT findings in CVD-related thrombosis in cancer patients. CT accurately depicts the degree and extent of benign central venous obstruction, being comparable to venography (25).
In most of our cases the typical subject with CVD-related thrombosis was an inpatient with a bad performance status and undergoing chemotherapy for a metastatic stage cancer. Nevertheless, it is important to notice that thrombi were found also in some outpatients with good clinical status and non-advanced cancer disease.
Some tumors are known to carry an increased risk of thromboembolic events, such as melanoma, lung carcinoma, hematological malignancies, and gynecological malignancies (2, 5, 8). This correlation appears in the present study, too, but the small sample size of several histological subgroups and the relatively low overall frequency of thrombosis do not allow an adequate statistical analysis of this aspect for all histotypes.
In this study 65% of thrombi were noticed retrospectively. No precise reason for under-diagnosis has yet been definitely identified at this stage. Many hypotheses can be put forward as to the factors contributing to the missed recognition of thrombosis, including small size of thrombi, port-related or contrast-related beam-hardening artifacts, poor knowledge of this occurrence by radiologists, focus on oncologic aspects. In many cases the thrombus was masked by the intense and inhomogeneous enhancement of the vein when scanned at the moment of dense filling with the contrast medium (still compact and scarcely mixed with blood) (Fig. 3). The thrombus might be masked by the mix of unenhanced blood coming from a vein (e.g. the left brachiocephalic vein) and the enhanced blood coming from another confluent vein (e.g. the right brachiocephalic vein) (Fig. 4). The inflow artifact from unenhanced blood in the azygous vein should also be considered, although this occurrence is more likely to cause false-positive than false-negative results. The beam-hardening artifact frequently visible around the intravenous catheter, which is a dense structure, may masquerade the presence of a thrombus or may be mimicked from the thrombus itself. Since our study was aimed at thrombus detection, the two readers had a clear notion of what to look for and evaluated accurately the mediastinum veins, possibly changing the window setting (also employing very wide windows, to attenuate the contrast-induced dense enhancement). The importance of using appropriate window values in detecting CVD-associated complications has been underlined also in a recent review (24). We believe that, in cancer patients with a CVD, the radiologists should follow the veins appropriately, also changing the window setting. In an article on occult pulmonary embolism detected at MDCT in oncology patients (5) it was retrospectively calculated that 75% of emboli had been missed at initial clinical image interpretation. Typical missed emboli were solitary and located in smaller arteries. It was concluded that radiologists should not be distracted by the thoracic manifestations of cancer and should always carry out a complete assessment of pulmonary artery in this kind of population (5).
A 33-year-old man with rectal carcinoma metastatic to the liver. Axial MDCT images (a) with a conventional mediastinum window setting; and (b) with a wide window setting showing a thrombus (black arrow in (a) and small white arrow in (b)) adjacent to a central catheter (black arrow in (b)). The clot had not been recognized at initial evaluation, possibly because of masquerading within the densely enhanced blood at normal mediastinum window values. Note that the catheter itself is not visible in (a). Also note the superficial portion of the catheter – large white arrows in (a) and (b)
A 55-year-old man with melanoma metastatic to the lungs. Same axial MDCT image shown (a) with conventional mediastinum window values; and (b) with a wide window setting showing a thrombus adjacent to a central catheter. The clot had not been recognized at the initial evaluation, possibly because resembling the typical mixing of enhanced and unenhanced blood at normal mediastinum window setting
The clinical importance of unsuspected CVD-related thrombosis in asymptomatic cancer patients is not as well-known as the natural history of these thrombi if untreated. In this study, in all the cases diagnosed at initial clinical CT, the referring clinician was alerted and an anticoagulant therapy was started immediately. In our retrospective study we could not investigate the real significance and natural evolution of the findings identified. Some other limitations of this study should be acknowledged. First, retrospective studies are notoriously weaker than prospective ones. Second, the overall number of patients with thrombi was small. In patients undergoing repeated CT studies only the first scan was analyzed, while many data point to an increase risk of thrombosis with time: this had been planned initially to simplify the data collection. Furthermore, there are some missing data. No data on concurrent anticoagulant therapy at the moment of CT were collected and changes in patient management (beginning or change of anticoagulant therapy, catheter withdrawal, and so on) in patients with a CT diagnosis of thrombosis were not defined. The site of antecubital IV contrast medium injection was not recorded, particularly if it was on the same side of the CVD or opposite to the venous line. Finally, it should be considered that the two radiologists reviewed the CT studies by consensus, therefore the possibility of interobserver variability cannot be ruled out. Nevertheless, the fact that these two reviewers always reached a consensus without the need for a third opinion, indicates that the data were collected in a simple and objective way. This justifies, at least partially, the limitation of lacking a gold standard in our study.
In conclusion, CVD-related thrombosis is a relatively common occurrence in patients with cancer. Although this complication is well recognized, there does not appear to be any study looking specifically at the radiological interpretation of routine CT studies in patients with CVDs. We emphasize that CVD-related thrombi can be very small, possibly overlooked by the radiologist, and can be encountered also in outpatients with a good performance status and a non-metastatic disease. Radiologists should carefully scrutinize chest scans with different window settings to detect CVD-related thrombi.