[ Critical Care Original Research ]

VTE Prophylaxis
in Critically Ill Adults
A Systematic Review and Network Meta-analysis
Shannon M. Fernando, MD; Alexandre Tran, MD; Wei Cheng, PhD; Behnam Sadeghirad, PharmD, MPH, PhD;
Yaseen M. Arabi, MD; Deborah J. Cook, MD; Morten Hylander Møller, MD, PhD; Sangeeta Mehta, MD;
Robert A. Fowler, MDCM; Karen E. A. Burns, MD; Philip S. Wells, MD; Marc Carrier, MD; Mark A. Crowther, MD;
Damon C. Scales, MD, PhD; Shane W. English, MD; Kwadwo Kyeremanteng, MD, MHA;
Salmaan Kanji, PharmD; Michelle E. Kho, PT, PhD; and Bram Rochwerg, MD

BACKGROUND: Critically ill adults are at increased risk of VTE, including DVT, and pul-
monary embolism. Various agents exist for venous thromboprophylaxis in this population.
RESEARCH QUESTION: What is the comparative efficacy and safety of prophylaxis agents for
prevention of VTE in critically ill adults?
STUDY DESIGN AND METHODS: Systematic review and network meta-analysis of randomized
clinical trials (RCTs) evaluating efficacy of thromboprophylaxis agents among critically ill pa-
tients. We searched six databases (including PubMed, EMBASE, and Medline) from inception
through January 2021 for RCTs of patients in the ICU receiving pharmacologic, mechanical, or
combination therapy (pharmacologic agents and mechanical devices) for thromboprophylaxis.
Two reviewers performed screening, full-text review, and extraction. We used the Grading of
Recommendations Assessment, Development, and Evaluation to rate certainty of effect estimates.
RESULTS: We included 13 RCTs (9,619 patients). Compared with control treatment (a
composite of no prophylaxis, placebo, or compression stockings only), low-molecular-weight
heparin (LMWH) reduced the incidence of DVT (OR, 0.59 [95% credible interval [CrI], 0.33-
0.90]; high certainty) and unfractionated heparin (UFH) may reduce the incidence of DVT
(OR, 0.82 [95% CrI, 0.47-1.37]; low certainty). LMWH probably reduces DVT compared
with UFH (OR, 0.72 [95% CrI, 0.46-0.98]; moderate certainty). Compressive devices may
reduce risk of DVT compared with control treatments; however, this is based on low-
certainty evidence (OR, 0.85 [95% CrI, 0.50-1.50]). Combination therapy showed unclear
effect on DVT compared with either therapy alone (very low certainty).
INTERPRETATION: Among critically ill adults, compared with control treatment, LMWH reduces
incidence of DVT, whereas UFH and mechanical compressive devices may reduce the risk of
DVT. LMWH is probably more effective than UFH in reducing incidence of DVT and should be
considered the primary pharmacologic agent for thromboprophylaxis. The efficacy and safety of
combination pharmacologic therapy and mechanical compressive devices were unclear.
TRIAL REGISTRY: Open Science Framework; URL: https://osf.io/694aj
CHEST 2022; 161(2):418-428

KEY WORDS: critical care medicine; DVT; pulmonary embolism; VTE

FOR EDITORIAL COMMENT, SEE PAGE 305

418 Original Research [ 161#2 CHEST FEBRUARY 2022 ]

 resulting from PE among ICU patients has been
Take-home Points estimated to be as high as 12%.6 As such, clinicians
Study Question: What is the comparative efficacy caring for critically ill patients often have a low
and safety of prophylaxis agents for prevention of threshold to investigate cases of suspected VTE.
VTE in critically ill adults?
To reduce the incidence of VTE, critically ill patients
Results: Compared with control treatment, low-
typically receive prophylaxis in the form of
molecular-weight heparin (LMWH) reduces the
incidence of DVT (OR, 0.59 [95% credible interval pharmacologic or mechanical therapy, or both.7
[CrI], 0.33-0.90]; high certainty) and unfractionated Prophylaxis is recommended by guidelines for medical
heparin (UFH) may reduce the incidence of DVT and surgical critically ill patients,8-11 and omission of
(OR, 0.82 [95% CrI, 0.47-1.37]; low certainty). thromboprophylaxis in the first 24 h after ICU
LMWH probably reduces DVT compared with UFH admission has been associated with higher mortality.12
(OR, 0.72 [95% CrI, 0.46-0.98]; moderate certainty). Still, a degree of uncertainty remains regarding which
Interpretation: Among critically ill adults, LMWH agents or combination of agents are best in the
reduces incidence of DVT as compared with control prevention of VTE among critically ill patients, and
treatment and is probably more effective than UFH. substantial practice variation exists.13 Although the use
of antithrombotic medications such as unfractionated
Critically ill patients are at increased risk of VTE, heparin (UFH) and low-molecular-weight heparin
including DVT and pulmonary embolism (PE).1,2 This (LMWH) is preferred, these drugs may be associated
increased risk is attributable to factors associated with with important adverse effects, namely major bleeding.2
hospitalization (immobilization), as well as those unique In contrast, mechanical devices such as intermittent
to the critically ill (high severity of illness and pneumatic compression (iPC) and sequential
inflammation, sedation exacerbating immobility, compressive devices (SCDs) have been suggested for
indwelling venous catheters, etc.). The incidence of VTE patients with contraindications to antithrombotic
among patients in the ICU varies substantially, ranging agents,8 such as those at high risk of bleeding. To
from 10% based on clinical findings1,3 to more than summarize the available evidence, we conducted a
90% when examined through autopsy studies.4 These systematic review and network meta-analysis to evaluate
rates are highest in particular populations such as those the comparative efficacy and safety of these therapies for
experiencing major surgery or trauma.5 Mortality prevention of DVT and PE in critically ill adults.

Methods registered our protocol with the Open Science Framework (https://osf.
io/694aj). Institutional review board approval was not required,
We followed the Preferred Reporting Items for Systematic Review and because all study data had been published previously, and we did
Meta-Analysis statement extension for network meta-analysis.14,15 We not include individual patient data.

ABBREVIATIONS: CrI = credible interval; GRADE = Grading of Rec- Health System, the Institute of Health Policy, Management and Eval-
ommendations Assessment, Development, and Evaluation; HIT = uation (R. A. Fowler, D. C. Scales), Dalla Lana School of Public Health,
heparin-induced thrombocytopenia; iPC = intermittent pneumatic University of Toronto, the Department of Critical Care Medicine (R. A.
compression; LMWH = low-molecular-weight heparin; PE = pulmo- Fowler, D. C. Scales), Sunnybrook Health Sciences Centre, the Keenan
nary embolism; RCT = randomized clinical trial; SCD = sequential Research Centre for Biomedical Science (K. E. A. Burns and D. C.
compression device; UFH = unfractionated heparin Scales), Li Ka Shing Knowledge Institute, St. Michael’s Hospital, Tor-
AFFILIATIONS: From the Division of Critical Care (S. M. Fernando, A. onto, ON, Canada; the Department of Biostatistics (W. Cheng), Yale
Tran, S. W. English, K. Kyeremanteng, and S. Kanji), Division of School of Public Health, Yale University, New Haven, CT; the Inten-
Hematology (P. S. Wells and M. Carrier), Department of Medicine, the sive Care Department (Y. M. Arabi), King Abdulaziz Medical City, the
Department of Emergency Medicine (S. M. Fernando), the Department College of Medicine (Y. M. Arabi), King Saud Bin Abdulaziz University
of Surgery (A. Tran), the School of Epidemiology and Public Health (P. for Health Sciences, Riyadh, Kingdom of Saudi Arabia; and the
S. Wells, M. Carrier, and S. W. English), University of Ottawa, the Department of Intensive Care (M. H. Møller), Copenhagen University
Clinical Epidemiology Program (P. S. Wells, M. Carrier, S. W. English, Hospital Righospitalet, Copenhagen, Denmark.
K. Kyeremanteng, and S. Kanji), Ottawa Hospital Research Institute, Drs Tran and Chen contributed equally to this manuscript.
Ottawa, the Department of Health Research Methods, Evidence, and FUNDING/SUPPORT: The authors have reported to CHEST that no
Impact (B. Sadeghirad, D. J. Cook, K. E. A. Burns, M. A. Crowther, and funding was received for this study.
B. Rochwerg), the Department of Anesthesia (B. Sadeghirad), the CORRESPONDENCE TO: Shannon M. Fernando, MD; email: sfernando@
Department of Medicine (D. J. Cook, M. A. Crowther, and B. Roch- qmed.ca
werg), the School of Rehabilitation Science (M. E. Kho), McMaster
University, Hamilton, the Interdepartmental Division of Critical Care Copyright Ó 2021 American College of Chest Physicians. Published by
Medicine (S. Mehta, R. A. Fowler, K. E. A. Burns, D. C. Scales), Uni- Elsevier Inc. All rights reserved.
DOI: https://doi.org/10.1016/j.chest.2021.08.050
versity of Toronto, the Department of Medicine (S. Mehta), Sinai

chestjournal.org 419
Data Sources and Search Strategy concealment, blinding, missing outcome data, and other features of
We searched six databases (Medline, PubMed, EMBASE, Scopus, Web bias. We judged each criterion for each trial as definitely or probably
of Science, and the Cochrane Database of Systematic Reviews) from at low risk of bias or definitely or probably at high risk of bias.
inception through January 14, 2021. An experienced health sciences Reviewers resolved disagreements through discussion.
librarian assisted in the development of the search strategy. The
search strategy and results are shown in e-Figure 1. We conducted Data Synthesis and Analysis
further surveillance searches using the Related Articles feature.16
For each clinical outcome and each pair of interventions, we performed a
Study Selection random-effects meta-analysis to explore heterogeneity (based on the I2
statistic and visual inspection of forest plots). For pairwise meta-
Two reviewers (S. M. F. and A. T.) independently screened titles and
analyses, we calculated and reported ORs for all dichotomous outcomes
abstracts identified through the searches using Covidence (Melbourne,
and mean differences for continuous outcomes with corresponding
Australia). The same two reviewers independently assessed full texts of
95% CIs. We aimed to perform Egger’s tests to assess the small-study
the selected articles from phase 1. Disagreements were resolved by
effects when 10 or more studies were available for a comparison.
discussion. We included English-language randomized clinical trials
(RCTs; parallel, cluster, or cross-over) meeting the following criteria: (1) A network diagram was used to assess whether the interventions in the
enrolled mostly ($ 80%) adult patients ($ 16 years of age); (2) network were connected adequately to permit network meta-analyses.
conducted primarily ($ 80% of patients) in an ICU (either medical, We performed both fixed-effect and random-effects Bayesian network
surgical, or mixed); (3) randomized patients to receive pharmacologic meta-analysis to compare interventions among included studies. We
thromboprophylaxis (UFH or LMWH), mechanical assessed the convergence based on the Brooks-Gelman-Rubin
thromboprophylaxis (SCD or iPC), a combination of pharmacologic and diagnostics and the potential scale reduction factor. We used
mechanical thromboprophylaxis, placebo, or no thromboprophylaxis; noninformative priors for all treatment comparisons in fixed-effects
and (4) evaluated at least one of the outcomes of interest. and random-effects models and assumed a common between-trial
The critical outcome of interest was the incidence of all (symptomatic SD that existed only in random-effects models. Estimates of the
plus asymptomatic, screening detected) lower-extremity DVT (either effect sizes were expressed as the ORs for dichotomous outcomes
proximal or distal) at the longest available follow-up, but within and mean differences for continuous outcomes with corresponding
90 days of ICU admission. DVT had to be confirmed using 95% credible intervals (CrI; for network meta-analysis) or CIs (for
ultrasonography or venography. We expected that most studies would pairwise meta-analysis). Both consistency models and unrelated
use surveillance imaging for detection of DVT, given that symptoms means models were fitted to the data to detect possible violations of
and clinical examination are unreliable in critically ill patients (because the consistency assumption. The deviance information criteria were
of sedation, recumbent positioning, etc.) and that ICU clinicians are used to select between models, with smaller values being preferred
interested in the detection of VTE events that might impact morbidity and a difference of five points suggesting an important difference.
and mortality, regardless of symptomatology or clinical signs.17 Other For each outcome, we also estimated ranking probabilities using the
outcomes included incidence of PE at the longest available follow-up surface under the cumulative ranking curve and mean treatment
and incidence of any new VTE (defined as any upper or lower rankings. Surface under the cumulative ranking curve values range
extremity DVT or any segmental or more proximal PE) at the longest between 0 and 1, with values nearer to 1 indicative of preferred
available follow-up. PE had to be confirmed through CT imaging, interventions. Smaller values of mean rank suggest preferred
ventilation-perfusion scan, or autopsy. We also included short-term interventions. Network meta-analyses in the main text were
mortality (closest to either in-hospital, 28-day, 30-day, or 90-day), performed using OpenBUGS software version 3.2.3 and the
major bleeding (as defined by study authors), incidence of heparin- R2OpenBUGS package version 3.2-3.2 in R software (R Foundation
induced thrombocytopenia (HIT; diagnosed by enzyme-linked
for Statistical Computing).19
immunosorbent assay), and ICU length of stay.
In the primary analysis, we evaluated the comparative efficacy and
Data Extraction safety of the following nodes: LMWH, UFH, mechanical compressive
Two investigators (S. M. F. and A. T.) abstracted the following devices (SCD and iPC), combination of pharmacologic and
variables from the included articles: authors, year of publication, mechanical compressive devices, and control node (a composite of
study design, study dates, type of ICU (surgical, medical, mixed), either no prophylaxis, placebo, or graduated compression stockings
eligibility criteria, number of patients, and interventions and because these have not been demonstrated to reduce incidence of
comparators (name, dose, frequency, administration route, and VTE in the critically ill).20 In the secondary analysis, we applied a
duration) using a predesigned data extraction sheet (e-Table 1). Two hierarchical modeling approach21 to evaluate the comparative
investigators (S. M. F. and A. T.) independently collected outcome efficacy of the following treatment categories (above the treatment
information. All data were verified independently by a third level): any pharmacologic therapy, mechanical compression devices,
investigator (W. C.). Disagreements were resolved through discussion. combination of pharmacologic and mechanical compressive devices,
and control treatment. Where network meta-analysis could not be
Risk of Bias Assessment conducted, we present the results of the pairwise meta-analyses. We
Two reviewers (S. M. F. and A. T.) independently assessed the risk of used the Grading of Recommendations, Assessment, Development,
bias of the included studies using a modified Cochrane Collaboration and Evaluation (GRADE) approach to assess the certainty of
tool,18 which included sequence generation, allocation sequence evidence for each comparison.22

Results underwent full-text review. We included 13 RCTs23-35

Search Results and Study Characteristics examining 9,619 patients. The findings extracted from
We identified 4,084 citations (Fig 1). After exclusion of the Xigris and Prophylactic Heparin Evaluation in
duplicates, we screened 3,307 citations, and 48 Severe Sepsis (XPRESS) trial were documented in three

420 Original Research [ 161#2 CHEST FEBRUARY 2022 ]

 Figure 1 – Flow chart summarizing evidence search and
Records identified study selection. RCT ¼ randomized clinical trial.
Identification through database
search, Inception
to January 2021
(N = 4,084)

Duplicates
removed
(n = 777)
Screening

Articles excluded by
Articles screened
title and abstract screening
(n = 3,307)
(n = 3,259)

Reasons for exclusion (n = 33)

Eligibility

Articles selected
Wrong population (n = 24)
for full-text review
Wrong outcome (n = 4)
(n = 48)
Wrong design (n = 5)

Studies included
for systematic
Secondary analyses of RCTs (n = 2)
review analysis
(n = 15)
Included

Studies included
in quantitative
synthesis /
meta-analysis
(n = 13)

separate publications.33,36,37 Study characteristics are interventions, 10 studies evaluated the efficacy of
shown in Table 1, with detailed review in e-Table 2. LMWH,24-29,31-34 six evaluated UFH,24-26,28,33,34 five
Screening methods for diagnosis of DVT and PE in evaluated compression devices (iPC or SCD),29-32,35 one
included studies are depicted in e-Table 3, and evaluated the combination of pharmacologic and
justification for excluded studies after full-text review are mechanical prophylaxis,23 and four evaluated control
included in e-Table 4. therapies.27,30,33,35 For the critical outcome of lower-
extremity DVT, all of the included studies conducted
Six of the included trials were conducted exclusively in
surveillance screening at regular intervals (e-Table 3).
North America,24,28-31,34 three were conducted
All studies used lower-extremity compression
exclusively in Europe,27,32,35 one was conducted in
ultrasonography with or without Doppler analysis for
Asia,26 and three were multicontinental.23,25,33 Five trials
included medical or mixed medical and surgical diagnosis of DVT (either proximal or distal), except for
patients,23,25,27,33,35 six exclusively included patients with two trials that used venography.27,28 Diagnostic testing
major trauma,24,28,29,31,32,34 and two exclusively included for PE was performed largely on the basis of clinical
postoperative ICU patients.26,30 In terms of signs and symptoms and not using surveillance imaging.

chestjournal.org 421
TABLE 1 ] Characteristics of the 13 Randomized Clinical Trials
Overall (13 studies; N ¼ 9,619)
Description No. of Studies (%) No. of Patients (%)
Continent of study ... ...
North America 6 (46.2) 1,858 (19.3)
Europe 3 (23.1) 652 (6.8)
Asia 1 (7.7) 156 (1.6)
Multinational 3 (23.1) 6,953 (72.3)
Year range of publication ... ...
1995-1999 4 (30.8) 980 (10.2)
2000-2004 3 (23.1) 731 (7.6)
2005-2009 1 (7.7) 1,186 (12.3)
2010-2014 3 (23.1) 4,283 (44.5)
2015-2020 2 (15.4) 2,439 (25.4)
ICU population ... ...
Major trauma 6 (46.2) 1,648 (17.1)
Mixed medical/surgical 5 (38.5) 7,485 (77.8)
Surgical (postoperative) 2 (15.4) 486 (5.1)
Interventions studied ... ...
LMWH 10 (76.9) 3,293 (34.2)
UFH 6 (46.2) 2,834 (29.5)
LMWH or UFH 1 (7.7) 985 (10.2)
Mechanical compressive devices 5 (38.5) 688 (7.2)
Adjunctive pharmacologic and compressive 1 (7.7) 957 (9.9)
devices
Control 4 (30.8) 1,394 (14.5)

LMWH ¼ low-molecular-weight heparin; UFH ¼ unfractionated heparin.

Risk-of-bias assessment of the included studies is shown comparative efficacy of the various interventions (with
in e-Table 5. associated GRADE certainty) is shown in Table 1.
Surface under the cumulative ranking curve rankings are
Incidence of DVT displayed in e-Table 6, and rankogram for incidence of
All 13 RCTs evaluated incidence of DVT.23-35 The DVT is show in e-Figure 2. Use of LMWH resulted in a
relevant network plot is shown in Figure 2A, and the reduction in DVT compared with UFH (OR, 0.72

A B C
LMWH UFH LMWH UFH UFH
LMWH

UFH or LMWH UFH or LMWH UFH or LMWH

(only in Arabi) (only in Arabi) (only in Arabi)
Control Control Control

SCD or iPC SCD or iPC SCD or iPC

UFH or LMWH + IPC UFH or LMWH + IPC UFH or LMWH + IPC
Combination Combination Combination

Figure 2 – A-C, Network plots for prevention of DVT (A), prevention of pulmonary embolism (PE) (B), and prevention of any VTE (DVT or PE) (C).
The size of the node corresponds to the number of patients randomized to that intervention. The thickness of the line and the associated numbers
correspond to the number of studies comparing the two linked interventions. iPC ¼ intermittent pneumatic compression; LMWH ¼ low-molecular-
weight heparin; SCD ¼ sequential compression device; UFH ¼ unfractionated heparin.

422 Original Research [ 161#2 CHEST FEBRUARY 2022 ]

 Boldface values are statistically significant findings. CrI ¼ credible interval; GRADE ¼ Grading of Recommendations Assessment, Development, and Evaluation; LMWH ¼ low-molecular weight heparin; iPC ¼
[95% CrI, 0.46-0.98]; moderate certainty) or control

GRADE
therapy (OR, 0.59 [95% CrI, 0.33-0.90]; high certainty).

Low
Low
Low

Low
Low

Low
Compared with control treatment, UFH (OR, 0.82
[95% CrI, 0.47-1.37]; low certainty) and compressive
Incidence of Any VTE

0.63 (0.18-1.59) devices (iPC or SCD; OR, 0.85 [95% CrI, 0.50-1.50])
0.79 (0.22-2.28)
0.82 (0.27-2.84)

0.80 (0.33-1.72)
0.76 (0.20-2.01)

0.96 (0.22-3.05)
may reduce the risk of DVT; however, these

No estimate

No estimate

No estimate
No estimate
OR (95% CrI)

comparisons were based on low-certainty evidence,

limited by imprecision.

Incidence of PE or Any VTE

Of the 13 included RCTs, five investigated the
incidence of PE among critically ill adults.23,25,29,32,34
The relevant network plot is shown in Figure 2B, and
Very low
Very low
Very low

Very low
Very low
Very low
Very low
Very low
GRADE

the GRADE results table is shown in Table 2.

Low

Low

Rankogram is displayed in e-Figure 3. Compared with

control treatment, LMWH (OR, 0.47 [95% CrI, 0.03-
3.91]) may reduce the incidence of PE, but this is based
Incidence of PE

on low quality of evidence that does not rule out the

0.29 (0.01-10.99)

2.30 (0.26-28.06)
1.59 (0.04-34.15)
3.52 (0.26-94.64)
2.41 (0.14-36.24)
1.44 (0.04-118.9)

intermittent pneumatic compression; PE ¼ pulmonary embolism; SCD ¼ sequential compression device; UFH ¼ unfractionated heparin.
0.47 (0.03-3.91)
0.70 (0.05-7.95)
0.20 (0.01-2.87)

0.65 (0.08-3.65)

possibility of important harm. Compared with control

OR (95% CrI)

treatment, the effect of UFH (OR, 0.70 [95% CrI, 0.05-

7.95]; very low certainty) and iPC or SCD (OR, 0.20
] Network Estimates for Incidence of DVT, PE, or Any VTE (Composite of DVT or PE)

[95% CrI, 0.01-2.87]; very low certainty) on PE is

uncertain. Similarly, the effect of LMWH
compared with UFH on PE is uncertain (OR, 0.65
[95% CrI, 0.08-3.65]) on the basis of very low quality of
evidence.
Moderate
GRADE

Incidence of any VTE (composite of DVT, PE, or both)

High
Low
Low

Low

Low

was evaluated in five RCTs.23,25,33-35 The relevant

network plot is shown in Figure 2C, and the GRADE
Incidence of DVT

results table is shown in Table 2. Rankogram is

0.59 (0.33-0.90)

0.72 (0.46-0.98)

displayed in e-Figure 4. Compared with control

0.82 (0.47-1.37)
0.85 (0.50-1.50)

0.69 (0.33-1.25)

0.96 (0.46-1.87)
No estimate

No estimate

No estimate
No estimate
OR (95% CrI)

treatment, both LMWH (OR, 0.63 [95% CrI, 0.18-

1.59]; low certainty) and UFH (OR, 0.79 [95% CrI,
0.22-2.28]; low certainty) may reduce the incidence of
any VTE.

Major Bleeding and HIT

Simultaneous mechanical and pharmacologic agents.

Seven RCTs investigated the incidence of major

bleeding.24,26-29,33,34 Unfortunately, data were not
sufficiently connected to perform network meta-
analyses; therefore, we conducted pairwise meta-
Combinationa vs SCD or iPC

analyses (Fig 3). Compared with UFH, LMWH had an

uncertain effect on major bleeding (OR, 1.71 [95% CI:
Combinationa vs control

a
LMWH vs combination
SCD or iPC vs control

UFH vs combinationa
LMWH vs SCD or iPC

0.49-5.95]; very low certainty). Only a single trial

UFH vs SCD or iPC

compared LMWH with iPC and SCD regarding the

LMWH vs control
UFH vs control

LMWH vs UFH

incidence of major bleeding and did not show a

difference.29 Three RCTs compared the incidence of
Comparison

HIT between UFH and LMWH.25,28,34 LMWH may

TABLE 2

result in a lower incidence of HIT (OR, 0.38 [95% CI,

0.15-0.98]; low certainty), although certainty is limited
a

chestjournal.org 423
Major Bleeding
Intervention Comparator OR OR
Study or Subgroup Events Total Events Total Weight M-H, Random, 95% CI M-H, Random, 95% CI
2.1.1 LMWH vs UFH
Cohn 1999 5 34 5 32 41.5% 0.93 [0.24-3.58]
De 2010 1 81 2 75 20.0% 0.46 [0.04-5.14]
Geerts 1996 5 171 1 173 23.6% 5.18 [0.60-44.81]
Olson 2015 4 216 0 220 14.8% 9.34 [0.50-174.51]
Subtotal (95% CI) 502 500 100.0% 1.71 [0.49-5.95]
Total events 15 8
Heterogeneity: Tau2 = 0.51; χ2 = 4.33, df = 3 (P = .23); I 2 = 31%
Test for overall effect: Z = 0.84 (P = .40)

2.1.2 LMWH vs Control

Fraisse 2000 6 108 3 113 100.0% 2.16 [0.53-8.85]
Subtotal (95% CI) 108 113 100.0% 2.16 [0.53-8.85]
Total events 6 3
Heterogeneity: Not applicable
Test for overall effect: Z = 1.07 (P = .29)

2.1.3 LMWH vs iPC

Ginzburg 2003 4 218 4 224 100.0% 1.03 [0.25-4.16]
Subtotal (95% CI) 218 224 100.0% 1.03 [0.25-4.16]
Total events 4 4
Heterogeneity: Not applicable
Test for overall effect: Z = 0.04 (P = .97)

2.1.4 UFH/LMWH vs Control

Levi 2007 38 976 50 959 100.0% 0.74 [0.48-1.13]
Subtotal (95% CI) 976 959 100.0% 0.74 [0.48-1.13]
Total events 38 50
Heterogeneity: Not applicable
Test for overall effect: Z = 1.39 (P = .16)

0.01 0.1 1 10 100

Favors intervention Favors comparator

Heparin-Induced Thrombocytopenia
Intervention Comparator OR OR
Study or Subgroup Events Total Events Total Weight M-H, Random, 95% CI M-H, Random, 95% CI
1.1.1 LMWH vs UFH
Cook 2011 5 1,873 12 1,873 81.7% 0.42 [0.15-1.18]
Geerts 1996 0 129 2 136 9.6% 0.21 [0.01-4.37]
Olson 2015 0 216 1 220 8.7% 0.34 [0.01-8.34]
Subtotal (95% CI) 2,218 2,229 100.0% 0.38 [0.15-0.98]
Total events 5 15
Heterogeneity: Tau2 = 0.00; χ2 = 0.18, df = 2 (P = .91); I 2 = 0%
Test for overall effect: Z = 2.00 (P = .05)

1.1.2 LMWH vs Control

Fraisse 2000 1 108 1 113 100.0% 1.05 [0.06-16.95]
Subtotal (95% CI) 108 113 100.0% 1.05 [0.06-16.95]
Total events 1 1
Heterogeneity: Not applicable
Test for overall effect: Z = 0.03 (P = .97)

0.01 0.1 1 10 100

Favors intervention Favors comparator

Figure 3 – Pairwise meta-analyses among therapies for thromboprophylaxis comparing incidence of major bleeding and heparin-induced thrombo-
cytopenia among LMWH, UFH, mechanical compression, or control treatment. iPC ¼ intermittent pneumatic compression; LMWH ¼ low-molecular-
weight heparin; UFH ¼ unfractionated heparin.

424 Original Research [ 161#2 CHEST FEBRUARY 2022 ]

because of the risk of bias, imprecision, and low number compared with UFH. The theoretical benefits of LMWH
of events (Fig 3). over UFH were described previously and include higher
bioavailability after subcutaneous dosing and longer
Other Outcomes plasma half-life.39 With regard to safety, the pairwise
Pairwise meta-analyses for incidence of DVT, PE, any meta-analysis did not show a clear difference between
VTE, hospital mortality, and ICU length of stay are UFH and LMWH in the incidence of major bleeding.
shown in e-Figures 5-9, and associated GRADE Furthermore, studies of patients outside of the ICU
confidence in estimates is shown in e-Tables 7-10. collectively suggest that incidence of HIT is lower in
patients receiving LMWH for VTE prophylaxis as
Secondary Analyses
compared with UFH,40 and we similarly found that
Secondary analysis based on a hierarchical modeling LMWH may result in a lower incidence of HIT among
approach comparing pharmacologic therapy alone the critically ill, although with low certainty of evidence.
(UFH or LMWH) vs mechanical compression alone Taken together, our results suggest that clinicians should
(iPC or SCD) vs combination (pharmacologic and consider using LMWH over UFH for VTE prophylaxis
mechanical compression) vs control treatment are in critically ill patients, particularly in light of evidence
shown in e-Figures 10-12. No differences between the supporting the cost-effectiveness of LMWH in this
interventions were identified. In particular, compared setting.41 Although UFH may be favored over LMWH in
with a combined strategy of pharmacologic and patients with renal failure because of presumed LMWH
compression therapy, pharmacologic therapy alone bioaccumulation,13 LMWH (particularly dalteparin) was
showed an uncertain effect on the incidence of DVT not associated with increased risk of bleeding in a
(OR, 1.08 [95% CrI, 0.53-2.19]; very low certainty). multicenter observational study of critically ill patients
Similarly, combination therapy had an uncertain effect with severe renal insufficiency,42 a finding that also has
on the incidence of DVT compared with compression been shown outside of the ICU.43 Therefore,
therapy alone (OR, 0.74 [95% CrI, 0.09-6.10]; very low randomized evaluation of thromboprophylaxis in
certainty). critically ill patients with renal insufficiency, among
other conditions that impair pharmacodynamics (eg,
Discussion severe shock), remains an important area for future
Pharmacologic agents are the most commonly used study. For most ICU populations, LMWH should be
therapy for prophylaxis against VTE in critically ill favored.
adults, and their use over no prophylaxis is In patients with a clear contraindication to
recommended by clinical practice guidelines.10,11 anticoagulation, mechanical devices (namely pneumatic
Indeed, the findings of our study provide evidence compression devices) have been favored for VTE
supporting the use of LMWH and UFH as compared prophylaxis,8 although their efficacy over no prophylaxis
with no pharmacologic prophylaxis in the prevention of has not been demonstrated clearly. We found that
DVT. Although the available trials did not find a similar mechanical compression devices may reduce the
certainty of effect in reducing PE, moderate certainty incidence of DVT, but this was based on low-certainty
evidence did suggest that LMWH and UFH probably evidence that does not rule out the possibility of harm.
reduce the incidence of any VTE compared with no Evidence is limited in critically ill adults to support the
pharmacologic prophylaxis. Although the use of these
notion that compression devices reduce the risk of
agents will result in additional costs, cost-effectiveness
bleeding when compared with pharmacologic
modeling still supports this practice when compared
prophylaxis, and this has been emphasized in clinical
with the increased incidence of VTE when no
practice guidelines.11 Taken together, the existing data
pharmacologic prophylaxis is used.38
provides very low confidence in the efficacy of
Perhaps the more controversial question is whether the mechanical devices, and the American Society of
existing literature supports the use of any particular Hematology guidelines provide a conditional
agent. The American Society of Hematology guidelines recommendation for their use over no prophylaxis.11
suggest that LMWH should be considered over UFH in Although the use of these devices poses seemingly little
critically ill patients.11 The findings of our study lend harm, they may be associated with additional resource
further support to this statement by identifying that use and cost, and their lack of demonstrated effect
LMWH probably reduces the incidence of DVT should prompt clinicians to treat critically ill patients

chestjournal.org 425
with pharmacologic prophylaxis as soon as it is safe to thromboprophylaxis in each of these populations
do so. separately. We relied on aggregate data extracted from
trials and did not have access to individual patient data,
Finally, we evaluated the efficacy of a combined strategy
which would have allowed for more extensive subgroup
of pharmacologic and mechanical prophylaxis against
analyses. Therefore, our findings may not extend across
the use of either of these methods alone, a question
recently investigated in The Pneumatic Compression all critically ill populations. Furthermore, although all
for Preventing Venous Thromboembolism (PREVENT) studies conducted surveillance imaging for detection of
trial.23 PREVENT did not demonstrate a reduction in DVT, heterogeneity existed regarding the screening
DVT with the use of adjunctive compression devices in period, and two trials used venography instead of
addition to pharmacologic prophylaxis as compared ultrasonography. Whether DVT discovered through
with a strategy of pharmacologic prophylaxis alone. Our routine surveillance holds the same prognostic
results are consistent with the findings of PREVENT, significance as clinically evident VTE is unknown.
with an unclear effect of combination therapy compared Although providers have stated the importance of
with either pharmacologic therapy alone or compression routine screening in identifying insidious DVT at risk of
therapy alone in the prevention of DVT, although on the embolization,17 it is worth noting that some of these
basis of low-certainty evidence limited by very serious clots may be small or distal, and therefore, it is unlikely
imprecision. Similarly, the American Society of that all screen-detected VTE carry the same significance
Hematology guidelines do not support the use of to patients. Our conclusions were limited by serious
combination therapy over either pharmacologic therapy imprecision in several effect estimates. However, where
or compression therapy alone.11 The PREVENT trial applicable, we downgraded GRADE certainty ratings
showed little effect of combination therapy on skin accordingly and used appropriate language to
breakdown, discomfort, or mobility,23 although adverse contextualize our conclusions. Finally, data for several
events represent an important avenue for future secondary outcomes (major bleeding, HIT, ICU length
research. of stay, and mortality) were insufficient to allow for
network meta-analysis, limiting comments on the
Our review has important limitations. We could not
relative efficacy and safety of these agents regarding
examine whether drug dose influenced efficacy; thus,
those outcomes, although we did perform pairwise
studies enrolling patients treated with both LMWH and
meta-analyses.
UFH at varying doses were included. Similarly, not
enough RCTs involving the various LMWH agents
(namely dalteparin and enoxaparin) were available for
Interpretation
us to compare their individual efficacy with other agents.
These questions related to specific agent and dose In critically ill adults, we found that LMWH reduces the
represent important avenues for future study. Our incidence of DVT and that UFH may reduce incidence
review did not capture any of the recent studies of DVT compared with control treatment. LMWH
examining therapeutic dose parenteral anticoagulation probably reduces incidence of DVT compared with UFH
for DVT prevention in seriously ill patients with with unclear effect on major bleeding. Therefore,
COVID-19, which may be associated with a higher LMWH should be considered as the primary agent for
thrombotic risk than other critically ill populations.44,45 pharmacologic thromboprophylaxis in most critically ill
Although COVID-19 may require a unique approach to populations. Mechanical compression devices may
anticoagulation, it is unlikely that such reduce incidence of DVT compared with control
recommendations ultimately will be extended to other treatment, but this was on the basis of low-certainty
ICU patients (for whom the existing evidence is evidence. As such, initiation of pharmacologic
summarized in this review). Finally, concerns persist prophylaxis should occur as soon as it is safe to do so.
regarding heterogeneity. Because we sought to evaluate Finally, combination (pharmacologic and mechanical)
exclusively critically ill patients, we included therapy has an unclear effect on incidence of DVT
heterogenous trials of predominantly medical, surgical compared with each method independently. Together,
(including multiple trauma populations), and mixed these results may inform clinical practice, guideline
populations and were unable to evaluate recommendations, and future research in this field.

426 Original Research [ 161#2 CHEST FEBRUARY 2022 ]

Acknowledgments 3. Hirsch DR, Ingenito EP, Goldhaber SZ. identified the need to update systematic
Prevalence of deep venous thrombosis reviews. J Clin Epidemiol. 2008;61(8):755-
Author contributions: S. M. F. and B. R. among patients in medical intensive care. 762.
conceived of the study idea. S. M. F., A. T., JAMA. 1995;274(4):335-337. 17. Cook D, Meade M, Guyatt G, et al.
W. C., and B. R. coordinated the systematic
4. Perkins GD, McAuley DF, Davies S, Clinically important deep vein thrombosis
review. S. M. F. and A. T. designed the search Gao F. Discrepancies between clinical and in the intensive care unit: a survey of
strategy. S. M. F. and A. T. screened abstracts postmortem diagnoses in critically ill intensivists. Crit Care. 2004;8(3):R145-
and full texts. S. M. F. and A. T. acquired the patients: an observational study. Crit Care. R152.
data and judged risk of bias in the studies. W. 2003;7(6):R129-R132.
C. verified the data and performed the 18. Guyatt GH, Busse JW. Modification of
5. Schultz DJ, Brasel KJ, Washington L, et al. Cochrane tool to assess risk of bias in
analyses. B. R. created the GRADE evidence
Incidence of asymptomatic pulmonary randomized trials. Evidence Partners, Inc.,
profiles. All authors interpreted the data embolism in moderately to severely website. https://www.evidencepartners.
analyses. All authors cowrote and revised the injured trauma patients. J Trauma. com/resources/methodological-resources/.
manuscript for intellectual content. All 2004;56(4):727-731; discussion 731-723. Accessed January 31, 2021.
authors provided their final approval for
manuscript submission. All authors agree to 6. Geerts W, Selby R. Prevention of venous 19. Sturtz S, Ligges U, Gelman A.
thromboembolism in the ICU. Chest. R2WinBUGS: a package for running
be accountable for all aspects of the work. S.
2003;124(6 suppl):357s-363s. WinBUGS from R. J Stat Soft. 2005;12(3):
M. F. is guarantor and takes responsibility for
7. McLeod AG, Geerts W. Venous 1-16.
the work.
thromboembolism prophylaxis in 20. Arabi YM, Khedr M, Dara SI, et al. Use of
Financial/nonfinancial disclosures: The critically ill patients. Crit Care Clin. intermittent pneumatic compression and
authors have reported to CHEST the 2011;27(4):765-780, v. not graduated compression stockings is
following: B. S. reports receiving funding 8. Geerts WH, Bergqvist D, Pineo GF, et al. associated with lower incident VTE in
from PIPRA AG (www.pipra.ch) outside of Prevention of venous thromboembolism: critically ill patients: a multiple propensity
the submitted work. D. J. C. is supported by a American College of Chest Physicians scores adjusted analysis. Chest.
Canada Research Chair in Critical Care Evidence-Based Clinical Practice 2013;144(1):152-159.
Knowledge Translation. K. E. A. B. is Guidelines (8th Edition). Chest. 21. Owen RK, Tincello DG, Keith RA.
supported by a Physician Services 2008;133(6 suppl):381s-453s. Network meta-analysis: development of a
Incorporated-50 Mid-Career Clinical 9. Kearon C, Akl EA, Comerota AJ, et al. three-level hierarchical modeling
Research Award. R. A. F. is the H. Barrie Antithrombotic therapy for VTE disease: approach incorporating dose-related
Fairley Professor of Critical Care at the Antithrombotic Therapy and Prevention constraints. Value Health. 2015;18(1):116-
University Health Network and University of of Thrombosis, 9th ed: American College 126.
Toronto Interdepartmental Division of of Chest Physicians Evidence-Based 22. Brignardello-Petersen R, Bonner A,
Critical Care Medicine. P. S. W. reports Clinical Practice Guidelines. Chest. Alexander PE, et al. Advances in the
research funding from Pfizer and Bristol- 2012;141(2 suppl):e419S-e496S. GRADE approach to rate the certainty in
Myers Squibb and honoraria from Pfizer, 10. Qaseem A, Chou R, Humphrey LL, estimates from a network meta-analysis.
Bayer, Bristol-Myers Squibb, Sanofi, and Starkey M, Shekelle P. Venous J Clin Epidemiol. 2018;93:36-44.
Servier. M. C. is supported by a Tier 2 thromboembolism prophylaxis in 23. Arabi YM, Al-Hameed F, Burns KEA,
Research Chair in Venous Thrombosis and hospitalized patients: a clinical practice et al. Adjunctive intermittent pneumatic
Cancer from the Department of Medicine at guideline from the American College of compression for venous
the University of Ottawa; reports research Physicians. Ann Intern Med. 2011;155(9): thromboprophylaxis. N Engl J Med.
funding from Leo Pharma, Pfizer, and 625-632. 2019;380(14):1305-1315.
Bristol-Myers Squibb; and reports honoraria 11. Schünemann HJ, Cushman M,
from Pfizer, Bayer, Bristol-Myers Squibb, 24. Cohn SM, Moller BA, Feinstein AJ,
Burnett AE, et al. American Society of Burns GA, Ginzburg E, Hammers LW.
Sanofi, Leo Pharma, and Servier. M. A. C. has Hematology 2018 guidelines for Prospective trial of low-molecular-weight
performed data safety and monitoring board management of venous heparin versus unfractionated heparin in
work for Bayer and advisory board work for thromboembolism: prophylaxis for moderately injured patients. Vasc
Servier, Asahi Kasei, and Precision hospitalized and nonhospitalized medical Endovasc Surg. 1999;33(2):219-233.
Biologicals and holds individual stocks in patients. Blood Adv. 2018;2(22):3198-
Alnylam. M. E. K. is supported by a Canada 3225. 25. Cook D, Meade M, Guyatt G, et al.
Research Chair in Critical Care Dalteparin versus unfractionated heparin
12. Ho KM, Chavan S, Pilcher D. Omission of in critically ill patients. N Engl J Med.
Rehabilitation and Knowledge Translation. B. early thromboprophylaxis and mortality 2011;364(14):1305-1314.
R. is supported by a Hamilton Health in critically ill patients: a multicenter
Sciences New Investigator Grant. None registry study. Chest. 2011;140(6):1436- 26. De A, Roy P, Garg VK, Pandey NK. Low-
declared (S. M. F., A. T., W. C., Y. M. A., M. 1446. molecular-weight heparin and
H. M., S. M., M. C., D. C. S., S. W. E., K. K., S. unfractionated heparin in prophylaxis
13. Lauzier F, Muscedere J, Deland E, et al. against deep vein thrombosis in critically
K.). Thromboprophylaxis patterns and ill patients undergoing major surgery.
Additional information: The e-Figures and determinants in critically ill patients: a Blood Coagul Fibrinolysis. 2010;21(1):57-
e-Tables can be found in the Supplemental multicenter audit. Crit Care. 2014;18(2): 61.
R82.
Materials section of the online article. 27. Fraisse F, Holzapfel L, Couland JM, et al.
14. Page MJ, McKenzie JE, Bossuyt PM, et al. Nadroparin in the prevention of deep vein
The PRISMA 2020 statement: an updated thrombosis in acute decompensated
References guideline for reporting systematic reviews. COPD. The Association of Non-
1. Cook D, Crowther M, Meade M, et al. BMJ. 2021;372:n71. University Affiliated Intensive Care
Deep venous thrombosis in medical- 15. Hutton B, Salanti G, Caldwell DM, et al. Specialist Physicians of France. Am J
surgical critically ill patients: prevalence, The PRISMA extension statement for Respir Crit Care Med. 2000;161(4 pt 1):
incidence, and risk factors. Crit Care Med. reporting of systematic reviews 1109-1114.
2005;33(7):1565-1571. incorporating network meta-analyses of 28. Geerts WH, Jay RM, Code KI, et al.
2. Minet C, Potton L, Bonadona A, et al. health care interventions: checklist and A comparison of low-dose heparin with
Venous thromboembolism in the ICU: explanations. Ann Intern Med. low-molecular-weight heparin as
main characteristics, diagnosis and 2015;162(11):777-784. prophylaxis against venous
thromboprophylaxis. Crit Care. 16. Sampson M, Shojania KG, McGowan J, thromboembolism after major trauma.
2015;19(1):287. et al. Surveillance search techniques N Engl J Med. 1996;335(10):701-707.

chestjournal.org 427
29. Ginzburg E, Cohn SM, Lopez J, J Trauma Acute Care Surg. 2015;79(6): heparin thromboprophylaxis: a meta-
Jackowski J, Brown M, Hameed SM. 961-968; discussion 968-969. analysis. Blood. 2005;106(8):2710-2715.
Randomized clinical trial of intermittent
35. Vignon P, Dequin PF, Renault A, et al. 41. Fowler RA, Mittmann N, Geerts W, et al.
pneumatic compression and low
Intermittent pneumatic compression to Cost-effectiveness of dalteparin
molecular weight heparin in trauma. Br J
prevent venous thromboembolism in vs unfractionated heparin for the
Surg. 2003;90(11):1338-1344.
patients with high risk of bleeding prevention of venous thromboembolism
30. Goldhaber SZ, Hirsch DR, hospitalized in intensive care units: the in critically ill patients. JAMA.
MacDougall RC, Polak JF, Creager MA, CIREA1 randomized trial. Intensive Care 2014;312(20):2135-2145.
Cohn LH. Prevention of venous Med. 2013;39(5):872-880. 42. Douketis J, Cook D, Meade M, et al.
thrombosis after coronary artery bypass
36. Levy M, Levi M, Williams MD, Prophylaxis against deep vein
surgery (a randomized trial comparing
Antonelli M, Wang D, Mignini MA. thrombosis in critically ill patients with
two mechanical prophylaxis strategies).
Comprehensive safety analysis of severe renal insufficiency with the
Am J Cardiol. 1995;76(14):993-996.
concomitant drotrecogin alfa (activated) low-molecular-weight heparin
31. Knudson MM, Morabito D, Paiement GD, and prophylactic heparin use in patients dalteparin: an assessment of safety
Shackleford S. Use of low molecular with severe sepsis. Intensive Care Med. and pharmacodynamics: the DIRECT
weight heparin in preventing 2009;35(7):1196-1203. study. Arch Intern Med. 2008;168(16):
thromboembolism in trauma patients. 1805-1812.
J Trauma. 1996;41(3):446-459. 37. Shorr AF, Williams MD. Venous
thromboembolism in critically ill patients. 43. Lim W, Dentali F, Eikelboom JW,
32. Kurtoglu M, Yanar H, Bilsel Y, et al. Observations from a randomized trial in Crowther MA. Meta-analysis: low-
Venous thromboembolism prophylaxis sepsis. Thromb Haemost. 2009;101(1):139- molecular-weight heparin and bleeding in
after head and spinal trauma: intermittent 144. patients with severe renal insufficiency.
pneumatic compression devices versus Ann Intern Med. 2006;144(9):673-684.
low molecular weight heparin. World J 38. Sud S, Mittmann N, Cook DJ, et al.
Screening and prevention of venous 44. Middeldorp S, Coppens M, van Haaps TF,
Surg. 2004;28(8):807-811.
thromboembolism in critically ill patients: et al. Incidence of venous
33. Levi M, Levy M, Williams MD, et al. a decision analysis and economic thromboembolism in hospitalized patients
Prophylactic heparin in patients with evaluation. Am J Respir Crit Care Med. with COVID-19. J Thromb Haemost.
severe sepsis treated with drotrecogin alfa 2011;184(11):1289-1298. 2020;18(8):1995-2002.
(activated). Am J Respir Crit Care Med.
39. Weitz JI. Low-molecular-weight heparins. 45. Wichmann D, Sperhake JP,
2007;176(5):483-490.
N Engl J Med. 1997;337(10):688-698. Lütgehetmann M, et al. Autopsy findings
34. Olson EJ, Bandle J, Calvo RY, et al. and venous thromboembolism in patients
Heparin versus enoxaparin for prevention 40. Martel N, Lee J, Wells PS. Risk for with COVID-19: a prospective cohort
of venous thromboembolism after trauma: heparin-induced thrombocytopenia with study. Ann Intern Med. 2020;173(4):268-
a randomized noninferiority trial. unfractionated and low-molecular-weight 277.

428 Original Research [ 161#2 CHEST FEBRUARY 2022 ]