Professional Documents
Culture Documents
net/publication/327870508
CITATIONS READS
0 312
14 authors, including:
Some of the authors of this publication are also working on these related projects:
The Molecular Mechanism of the Anti-tumoral Activity of an Iron Chelator in Mantle Cell Lymphoma View project
All content following this page was uploaded by Shlomit Tamir on 27 November 2018.
Abstract
Anticoagulation is often used in superior vena cava syndrome (SVCS) associated with cancer (i.e malignant SVCS), even
without thrombosis, but its effect on outcomes has not been reported. We aimed to determine factors and outcomes associ-
ated with thrombosis and anticoagulation in malignant SVCS. Patients with malignant SVCS diagnosed on computerized
tomography (CT) were retrospectively included, indexed at diagnosis and followed for 6 months using medical records. The
cohort included 183 patients with malignant SVCS of which 153 (84%) were symptomatic. Thirty of the 127 patients (24%)
with a reviewable baseline CT had thrombosis of the SVC or tributaries at diagnosis. Patients with baseline thrombosis
more often had symptomatic SVCS (p < 0.01). 70% (21/30) of patients with thrombosis and 52% (49/97) of those without
thrombosis at baseline received anticoagulation, most often at therapeutic doses. Thrombosis occurred in 5/39 patients with
anticoagulation (13%) compared to 2/18 (11%) of those without, during follow-up (p = 0.85). Anticoagulation was associated
with a reduction in risk of SVC stent placement during follow-up that did not reach statistical significance (HR 0.47, 95%
CI 0.2–1.13, p = 0.09). Major bleeding occurred in 7 (4%) patients, six of whom received anticoagulation (four therapeutic
and two intermediate dose). Neither thrombosis nor anticoagulation affected survival. Anticoagulation is commonly used
as primary prevention but its benefit remains to be proven. The role of reduced-dose anticoagulation in non-thrombotic
malignant SVCS should be prospectively assessed.
Highlights
5
* Avi Leader Institute of Oncology, Davidoff Cancer Center, Rabin
avileader@yahoo.com Medical Center, Petah Tikva, Israel
6
1 Institute of Hematology, Meir Medical Center, Kfar Saba,
Institute of Hematology, Davidoff Cancer Center, Rabin
Israel
Medical Center, 39 Jabotinsky, Petah Tikva 4941492, Israel
7
2 Cardiology Department, Meir Medical Center, Kfar Saba,
Sackler School of Medicine, Tel Aviv University, Tel Aviv,
Israel
Israel
8
3 Cardiovascular Research Institute Maastricht (CARIM),
Radiology Department, Rabin Medical Center, Petah Tikva,
Maastricht University, Maastricht, The Netherlands
Israel
9
4 Thrombosis Expert Center, Maastricht University Medical
Thoracic Oncology Unit, Rabin Medical Center, Petah Tikva,
Center (MUMC+), Maastricht, The Netherlands
Israel
13
Vol.:(0123456789)
R. Ratzon et al.
• Anticoagulation is commonly used as primary prevention (ICD-9), initial screening was done using the following
of SVC thrombosis, especially in patients with high grade codes: 4592 (‘compression of vein’ at any site, which
occlusions includes, but isn’t restricted to, SVCS) and 45387 (‘acute
• Thrombosis rates were similar with and without antico- venous embolism and thrombosis of other thoracic veins’).
agulation, and major bleeding (4%) occurred mainly with Patients were included in this study if they were aged
therapeutic dose anticoagulation 18 years or older and diagnosed with malignant SVCS
• This analysis was also weakly suggestive of a protective between January 1st 2007 and December 31st 2016. A
effect for anticoagulation on salvage SVC stenting; how- diagnosis of malignant SVCS had to fulfill all the follow-
ever, the significance threshold was not reached ing criteria: (1) Diagnosis of SVCS on chest computed
• The role of reduced dose anticoagulation for primary pre- tomography (CT) performed with intravenous (IV) con-
vention of SVCS thrombosis in high risk patients with trast media, as well as presence of a thoracic mass; (2) His-
malignant SVCS should be prospectively assessed tological evidence of active cancer from the mass or from
another site considered to be the same disease. Patients
whose diagnostic CT was not available for review, but had
Introduction a written radiology report confirming the above findings
were also eligible. Subjects with no clinical follow-up data
Superior vena cava (SVC) syndrome is caused by a reduc- after diagnosis were excluded. Patients were indexed at the
tion in the blood flow through the SVC or its tributary veins. time of SVCS diagnosis, defined as the date of the imaging
In 60–80% of cases, the SVC syndrome (SVCS) is a result study leading to SVCS diagnosis. Records were followed
of external compression by a mediastinal tumor [1–4]. The from the index date for 6 months or until death.
most common etiology of malignant SVCS is lung cancer
(75%) followed by non-Hodgkin lymphoma (20%) and meta-
static cancer (usually breast cancer) [1, 4, 5]. Surprisingly, Variables and measurement
little is known regarding the coexistence of both external
compression and thrombosis in malignant SVCS. Table 1 presents an overview of baseline variables.
In most patients the initial management includes diag-
nosing and then treating the underlying malignancy, while
relieving any symptoms. Endovascular stent implantation Endovascular stenting
is recommended in life threatening cases or refractory dis-
ease [6–8]. Anticoagulation is recommended for patients “Baseline SVC stent” was defined as endovascular SVC
undergoing SVC stenting and uniformly used in those with stent placement within the first week after SVCS diagno-
an intravascular device-associated thrombus [9]. In addi- sis. These stents were most probably placed as a result of
tion, some physicians prescribe anticoagulation for primary the initial clinical (or radiological) presentation. Stents
and secondary prevention of SVC thrombosis in malignant placed beyond the first week were classified as “salvage
SVCS (1,14). However, data on the efficacy and safety of stent” and are usually placed due to progression of SVCS
anticoagulation in this setting is lacking [10]. symptoms [11]. Therefore they were considered to be sur-
In this study of patients with malignant SVCS we aimed rogates for failure of initial management.
to: (1) assess the prevalence of thrombosis at diagnosis; (2)
determine current practice regarding anticoagulation; (3)
evaluate the effect of anticoagulation on clinical outcomes. Radiological review procedures and definitions
13
Thrombosis, anticoagulation and outcomes in malignant superior vena cava syndrome
13
R. Ratzon et al.
Thrombosis-related factors
Intravascular device 34 (19) 9 (30) 16 (17) 0.119
Hormonal therapy 6 (3) 2 (7) 3 (3) 0.338
Prior venous thromboembolism 11 (6) 4 (13) 5 (5) 0.214
Khorana score (Khorana et al. 2008)f
Data available 150 26 80
Intermediate risk (1–2) 116 (77) 18 (69) 65 (81) 0.045
High (≥ 3) 34 (23) 8 (31) 15 (19)
Thrombosis statusa,b
Data available 127
Thrombosis of SVC or tributariesa,b 30 (24)
Thrombosis of SVCa,b 7 (5)
This table shows the baseline characteristics overall and stratified by thrombosis status at SVC syndrome
diagnosis. Univariate analysis was performed, comparing baseline characteristics among patients with
thrombosis vs. without thrombosis at SVC syndrome diagnosis
CT computerized tomography; IQR inter-quartile range; SVC superior vena cava
*This column shows p values from a univariate analysis comparing the “yes” and “no” thrombosis sub-
groups. Each analysis considered the variable in the same row as the p value vs. all other variables in that
group
a
Thrombosis was defined as a thrombus in a reviewable contrast enhanced CT performed at SVC syndrome
diagnosis. Tributaries were defined as the brachiocephalic, jugular or subclavian veins
b
This subgroup analysis only includes the 127 patients with a reviewable CT at SVC syndrome diagnosis
c
Anticoagulation prescribed within 21 days of SVC syndrome diagnosis. The date of anticoagulation was
not known for two patients
d
The dose of anticoagulation was not known for 12 patients
e
Baseline SVC stent was defined as stent placement within the 1st week (< 7 days) after the diagnosis of
SVC syndrome (vs. none or ≥ 7days)
f
The data needed to calculate the Khorana score were missing in 33 patients. The statistical analysis was
performed for increasing Khorana scores. No patient had a score of 0
The intervention, “anticoagulation”, was defined as any Baseline variables were compared according to the status
class, dose or duration of anticoagulation prescribed within of anticoagulation and thrombosis at diagnosis (yes vs.
the first 21 days after SVCS diagnosis. This period (and not no), using t-test for continuous variables and Chi square for
less) was selected to account for delays in anticoagulation categorical variables or Fisher’s exact test in case of 2*2
which may occur when invasive diagnostic or therapeutic tables. Multivariable general linear models, including vari-
procedures are indicated or when the CT is performed on an ables with p < 0.1 in univariate analysis, were performed
outpatient basis. Data on anticoagulation started later dur- to identify baseline factors associated with thrombosis and
ing follow-up (≥ 21 days after diagnosis) were collected but anticoagulation (in separate models) at SVCS diagnosis.
were not considered in the analyses. Multivariable Cox regression models (including variables
with p < 0.1 in univariate Kaplan–Meier survival analysis
Outcomes of all baseline variables) were performed for salvage stent
placement and death. Hazard ratio (HR), 95% confidence
Efficacy outcomes included new or recurrent thrombosis of interval (CI) and p-value were presented. When time to
the SVC or tributaries, salvage stent placement and death salvage stent was analyzed, death was added to the model
from any cause. The safety outcome was ISTH-defined as a competing risk. Anticoagulation was treated as a fixed
major bleeding [12]. variable.
13
Thrombosis, anticoagulation and outcomes in malignant superior vena cava syndrome
All analyses were performed with IBM SPSS Statistics Current anticoagulation practice in malignant SVCS
for Windows, Version 24.0 (Armonk, NY: IBM Corp.),
except for competing risk models that were performed using Half of patients without thrombosis received
PROC PHREG with SAS 9.4 for windows (SAS Institute anticoagulation
Inc., Cary, NC, USA).
Anticoagulation was used in 92 (51%) of all patients for a
median duration of 68 days (range: 14–179). Seventy per
cent (21/30) of patients with thrombosis and 52% (49/97) of
Results those without thrombosis at baseline received anticoagula-
tion. The anticoagulant prescribed was low molecular weight
Initial screening using ICD-9 codes identified 278 poten- heparin (LMWH) in 89/92 (97%) and LMWH followed by a
tial subjects, of whom 95 were excluded on medical record vitamin K antagonist in 3/92 (3%).
review mainly because of extrathoracic venous compression
or absence of malignancy (details in Online Resource 1). Anticoagulation for higher grade venous occlusions
Thus, the study cohort included 183 patients with malignant
SVCS. The characteristics of patients with and without anticoag-
ulation, as well as univariate analysis of variables associ-
ated with anticoagulation are shown in Online Resource 3.
Patients receiving anticoagulation had higher grade venous
Sample characteristics occlusion more often than patients without anticoagulation
(p < 0.05), based on multivariate analysis (Online Resource
Baseline patient characteristics are detailed in full in Table 1. 2).
The majority of patients (80%, N = 146) presented with mild
to moderate symptoms, while 29 (16%) were asymptomatic Therapeutic anticoagulation doses were frequently used
and only 7 (4%) had severe symptoms. Endovascular SVC for primary prophylaxis
stents were placed in 28 (15%) at baseline (i.e. < 7 days
after the diagnosis) at a median of 2 days (IQR 0–4) from Full dose anticoagulation was used in 80% (16/20) of anti-
diagnosis. coagulated patients with baseline thrombosis and in 59%
(27/46) of anticoagulated patients without thrombosis,
while the remainder received reduced doses (Table 1). An
Thrombosis prevalence at SVCS diagnosis additional 26 patients (14%) started anticoagulation 3 or
more weeks after SVCS diagnosis, and were not classified
Thrombosis and the degree of venous occlusion at diagnosis in the anticoagulation group. The median time from SVCS
were evaluated in the 127 patients (69%) with a reviewable diagnosis to starting anticoagulation was 0 days (IQR 0–4)
baseline CT. Patients with and without a reviewable diag- in patients who started anticoagulation within 3 weeks of
nostic CT had comparable baseline clinical characteristics SVCS diagnosis compared to 46 days (IQR 29–78) in those
(data not shown). who received anticoagulation later. The indication for early
anticoagulation was the malignant SVCS itself (with or with-
out thrombosis), while the indication varied in patients with
Thrombosis of SVC tributaries is prevalent, especially anticoagulation later during follow-up (including pulmonary
in symptomatic patients embolism, atrial fibrillation).
Baseline thrombosis of the SVC or tributaries was found The effect of anticoagulation on clinical outcomes
in 30/127 (24%), while only 7/127 (5%) had a thrombus in
the SVC. High grade venous occlusion of the SVC or its Thrombosis rates were similar
tributaries was present in 101/127 (80%). Table 1 shows the with and without anticoagulation
characteristics of patients with and without baseline throm-
bosis, as well as univariate analysis of variables associated A follow-up CT was performed in 57 of the 127 patients
with thrombosis. Patients with baseline thrombosis more (45%) with a reviewable baseline CT, at a median of 138
often had symptomatic SVCS (vs. asymptomatic; p < 0.01) days (IQR 71–182) from diagnosis. The main reason for
and higher Khorana scores (p < 0.05) than patients without not performing a follow-up CT was early death (N = 43).
thrombosis, based on multivariate analysis (Online Resource Thrombosis of the SVC or tributaries during follow-up
2). was identified in 7/57 (12%) at a median of 73 days (IQR
13
R. Ratzon et al.
66–139) after SVCS diagnosis. Thrombosis occurred in 5/39 included solid malignancy and anticoagulation in the final
patients with anticoagulation (13%) compared to 2/18 (11%) model. Patients with solid malignancy (vs. hematological)
of those without anticoagulation (p = 0.85). Table 2 depicts had an increased risk of salvage stenting (HR 7.16, 95% CI
selected patient variables according to thrombosis status 1.36–37.68, p < 0.05). Anticoagulation prescribed within 21
during follow-up. days of diagnosis (vs. none) was associated with a reduction
in risk that did not reach statistical significance (HR 0.47,
Salvage stenting was more likely in solid cancers 95% CI 0.2–1.13, p = 0.09).
Table 2 Characteristics of Stratification based on patient characteristics All patients with a New thrombo-
patients with or without a new follow-up CTa,b, N (%) sis of SVC &
SVC thrombosis during follow- tributariesa,b, N (%)
upa,b
No Yes
This table shows selected baseline characteristics overall in patients who underwent follow-up CT. The
characteristics are stratified by thrombosis status in the follow-up CT
CT computerized tomography; SVC superior vena cava
a
Thrombosis of the SVC or tributaries during follow-up was defined as a thrombus in a reviewable contrast
enhanced CT performed after the diagnosis of SVC syndrome, provided there was no thrombus in this ves-
sel at diagnosis. Tributaries were defined as the brachiocephalic, jugular or subclavian veins
b
This analysis only includes patients the 57 patients with a reviewable post-diagnosis CT
c
SVC stent placement within the first week after the diagnosis of SVC syndrome (< 7 days)
d
Endovascular SVC stent placement ≥ 7 days after the diagnosis of SVC syndrome
e
Anticoagulation prescribed within 21 days of SVC syndrome diagnosis
13
Thrombosis, anticoagulation and outcomes in malignant superior vena cava syndrome
Neither anticoagulation nor thrombosis affected survival Accordingly, the effect of anticoagulation on outcomes is
of interest, but had not been previously reported. New or
Over a median follow-up of 44 (IQR 22–93) days, 90 (49%) recurrent cases of thrombosis during follow-up occurred in
patients died. Cancer-related factors were the main prog- 12% of the 57 patients with a reviewable follow-up CT. Of
nostic determinants as shown in Online Resource 4 which note, thrombosis rates were similar with and without antico-
depicts the multivariate analysis of factors associated with agulation. This comparison may have been confounded by
death from any cause. Neither thrombosis at diagnosis nor a higher prevalence of high grade occlusions in the antico-
anticoagulation was associated with death in univariate (data agulated group, which may theoretically confer a higher risk
not shown) or multivariate analysis. The cause of death was of thrombosis. However, the number of events and sample
not reliably documented. size in this subgroup did not allow us to adjust for baseline
variables or the competing risk of death. We also had no
data on anticoagulation status at the time of the follow-up
Discussion CT. Therefore, this analysis does not allow for definitive
conclusions on the role of anticoagulation in prevention of
To the best of our knowledge, this is the largest cohort thrombosis of the SVC or tributaries.
reporting patterns of thrombosis (N = 127) and anticoagu- Salvage SVC stenting, performed in 16% of patients,
lation (N = 183) in malignant SVCS. We found that throm- was far more common in solid malignancies than hemato-
bosis of the SVC or tributaries was prevalent at diagnosis logical. This possibly reflects the usually swift reduction in
(24%) especially in patients with symptoms of the syndrome. tumor mass with steroids and chemotherapy in aggressive
This suggests that the thrombosis was clinically relevant and non-Hodgkin lymphoma resulting in a rapid resolution of
not just an incidental finding. As shown in our study, radi- SVCS symptoms. This analysis was also weakly suggestive
ologists and clinicians should be aware that the thrombosis of a protective effect for anticoagulation on salvage SVC
occurs most often in the SVC tributaries (i.e. brachioce- stenting; however, the significance threshold was not reached
phalic, jugular and/or subclavian veins) and not necessarily (HR 0.47, 95% CI 0.2–1.13, p = 0.09) making this finding
in the SVC. The prevalence of thrombosis is not surprising hypothesis-generating at best.
and similar to smaller prior cohorts showing thrombosis in The overall incidence of major bleeding was high (4% at
26–57% [13, 14]. 6 months). Major bleeding occurred mainly in patients with
Several anticoagulation practice patterns were apparent therapeutic dose anticoagulation, suggesting that therapeu-
in this study. First, anticoagulation was used after SVCS tic dose anticoagulation may be harmful in patients without
diagnosis in half of all patients. Second, anticoagulation thrombosis in this setting, in the absence of evidence clearly
was commonly prescribed in the absence of thrombosis supporting efficacy.
(52%), while it was not universally used (70%) in patients This study also confirmed that the overall prognosis in
with thrombosis. Third, the degree of venous occlusion of malignant SVCS is dismal (51% alive at 6 months). Neither
the SVC or tributaries at diagnosis was the only clinical thrombosis nor anticoagulation affected mortality. Cancer-
or radiological variable associated with use of anticoagu- related factors were the main prognostic determinants, simi-
lation, whereby patients with higher grade occlusion (vs. lar to prior studies [11], and are unlikely to be overcome by
occlusions < 90%) were more likely to receive anticoagula- supportive management strategies (such as anticoagulation
tion. There is however no evidence to suggest whether the or stenting).
degree of venous occlusion affects the rate of subsequent One of the major strengths of this study is the review
thrombosis and justifies anticoagulation. Fourth, the anti- of radiological data in a significant proportion of patients
coagulation doses used were mostly full therapeutic doses, (69%), unlike prior cohorts which didn’t review any imag-
even in patients without thrombosis. Reduced doses were ing studies [13, 14]. Another strong point is that the patient
more frequent in patients without thrombosis than those with characteristics are in line with previous malignant SVCS
thrombosis. cohorts, strengthening the external validity of these findings.
Conflicting estimates of anticoagulation use (13% vs. There are several limitations which warrant discussion.
94%) have been reported in two retrospective cohorts of First, confounding by indication of anticoagulation is pos-
malignant SVCS (N ≤ 50) with very different study popula- sible, along with other limitations inherent to any retrospec-
tions [1, 14]. The 94% rate of anticoagulation was in a study tive study. The factors influencing use of anticoagulation
of 50 critically ill cancer patients with malignant SVCS in may differ between physicians, centers and clinical settings
the intensive care unit (26% with thrombosis), while the 13% [1, 14]. Second, the diagnosis of malignant SVCS relied
was in 47 patients with unselected malignant SVCS. upon the original radiology report in the 56 (31%) patients
These data suggest that anticoagulation is commonly used whose diagnostic CTs were performed at another center and
in malignant SVCS but that treatment is not standardized. were not available for review. Nevertheless, none of the 127
13
R. Ratzon et al.
13