Content uploaded by Daniel J Quinlan
Author content
All content in this area was uploaded by Daniel J Quinlan on Jan 21, 2021
Content may be subject to copyright.
For personal use. Only reproduce with permission from The Lancet Publishing Group.
ARTICLES
THE LANCET • Vol 358 • July 7, 2001 9
Summary
Background The optimum duration of prophylaxis against
venous thromboembolism after total hip or knee replacement
is uncertain. Our primary objective was to establish the
efficacy of extended-duration prophylaxis on symptomatic
venous thromboembolic events.
Methods We identified randomised trials comparing extended-
duration prophylaxis using heparin or warfarin with placebo or
untreated control in patients undergoing elective total hip
or knee replacement by searching electronic databases
(MEDLINE, EMBASE), references from retrieved articles, and
abstracts from conference proceedings, and by contact with
pharmaceutical companies and investigators. Two reviewers
independently extracted data on study design, symptomatic
and symptomless venographic venous thromboembolism,
death, and bleeding outcomes. Results from individual trials
were combined with the Mantel-Haenszel method.
Findings Nine studies met our inclusion criteria (3999
patients), eight with low molecular weight heparin, and one
with unfractionated heparin. Extended-duration prophylaxis for
30–42 days significantly reduced the frequency of
symptomatic venous thromboembolism (1·3% vs 3·3%,
OR 0·38; 95% CI 0·24–0·61, numbers needed to treat
[NNT]=50), with no statistical evidence of heterogeneity
(2test, p=0·69). There was a greater risk reduction in
patients undergoing hip replacement (1·4% vs 4·3%, 0·33;
0·19–0·56, 34) compared with knee replacement (1·0% vs
1·4%, 0·74; 0·26–2·15, 250). A significant reduction in
symptomless venographic deep vein thrombosis was also
observed (9·6% vs 19·6%, 0·48; 0·36–0·63, 10). There was
no increase in major bleeding but extended-duration
prophylaxis was associated with excess minor bleeding (3·7%
vs 2·5%, 1·56; 1·08–2·26, numbers needed to harm
[NNH]=83).
Interpretation Among patients undergoing total hip or knee
replacement, extended-duration prophylaxis significantly
reduces the frequency of symptomatic venous thrombo-
embolism. The reduction in risk is equivalent to about 20
symptomatic events per 1000 patients treated.
Lancet 2001; 358: 9–15
Thrombosis Unit, Department of Haematology, Royal Perth
Hospital, Perth, Australia (J W Eikelboom MB); Department of
Radiology, King’s College Hospital, London SE5 9RS, UK
(D J Quinlan MB); Department of Medicine, McMaster University,
and St Joseph’s Hospital, Hamilton, Ontario, Canada
(J D Douketis MD)
Correspondence to: Dr Daniel J Quinlan
(e-mail: dan.quinlan@kcl.ac.uk)
Introduction
The optimum duration of prophylaxis against venous
thromboembolism after total hip or knee replacement
surgery remains uncertain.1It is common practice to
administer prophylaxis until discharge from hospital,
usually 7 to 14 days after surgery. However, in patients
receiving in-hospital prophylaxis, the frequency of
venographic deep vein thrombosis is still 15–30% at the
time of hospital discharge, and an additional 10–25% of
patients develop symptomless new deep vein thrombosis
during the next 3 to 4 weeks.2-5 Randomised trials have
shown that extending prophylaxis beyond the time of
hospital discharge substantially reduces the risk of
developing new symptomless thrombi at 30 to 45 days,3-5
which has led these investigators to recommend that longer
duration prophylaxis should be used in all patients
undergoing hip replacement.
More recently, however, two prospective studies
conducted in patients without known proximal deep vein
thrombosis at the time of discharge from hospital
demonstrated that the frequency of new out-of-hospital
symptomatic deep vein thrombosis or pulmonary
embolism without extended-duration prophylaxis was only
about 2% after 3 months of follow-up.6,7 These data
suggest that most symptomless thrombi remain clinically
silent irrespective of whether extended-duration prop-
hylaxis is given. Meanwhile, the effect of extended-
duration prophylaxis on symptomatic venous thrombo-
embolism remains to be clarified. In the only study, to our
knowledge, specifically designed to examine the effect of
extended-duration prophylaxis on clinical outcomes,8a
significant reduction in symptomatic venous thrombo-
embolism could not be shown. This led these investigators
to conclude that prophylaxis confined to the in-hospital
phase is adequate in most patients. Existing international
guidelines for prevention of venous thromboembolism also
recommend only 7–10 days of therapy with warfarin or low
molecular weight heparin after total hip or knee
replacement.1,9
To further clarify the role of extended-duration
prophylaxis, we conducted a meta-analysis of all
randomised trials assessing the efficacy and safety of
extended-duration prophylaxis in patients undergoing
elective hip or knee replacement surgery. The primary
outcome was symptomatic venous thromboembolism,
including deep vein thrombosis and pulmonary embolism.
Secondary outcomes were symptomless proximal and
distal deep vein thrombosis detected by venography, all-
cause mortality, and adverse events including major and
minor bleeding.
Methods
We prospectively developed a protocol detailing the
specific objectives, criteria for study selection, the
Extended-duration prophylaxis against venous thromboembolism
after total hip or knee replacement: a meta-analysis of the
randomised trials
John W Eikelboom, Daniel J Quinlan, James D Douketis
Articles
For personal use. Only reproduce with permission from The Lancet Publishing Group.
approach to assessing study quality, primary and
secondary outcomes, and statistical methods.
Study identification
We attempted to identify all relevant published and
unpublished unconfounded randomised trials that
compared extended-duration anticoagulant thrombo-
prophylaxis with placebo or untreated control in patients
undergoing hip or knee arthroplasty. We searched the
MEDLINE and EMBASE electronic databases from
January, 1980, to July, 2000, using the terms thrombosis,
thromboembolism, pulmonary embolism, randomised
controlled trial, controlled clinical trial, random, placebo,
hip arthroplasty, and knee arthroplasty, in combination
with generic and trade names of individual low molecular
weight heparin preparations. Bibliographies of journal
articles were hand-searched to locate additional studies,
and we contacted manufacturers of low molecular weight
heparin agents for knowledge of unpublished studies. We
assessed relevance using a hierarchical approach based on
title, abstract, and the full manuscript.
Study selection
Two investigators (JWE, DJQ) independently assessed
studies for possible inclusion and any disagreements were
resolved by discussion. To be included, studies had to be
properly randomised; include patients undergoing elective
total hip or total knee replacement; compare extended-
duration prophylaxis with low molecular weight heparin,
unfractionated heparin, or warfarin with placebo or
untreated control; and use objective methods to confirm
the diagnosis of symptomatic venous thromboembolism.
We included data on symptomless deep vein thrombosis
from these studies only if screening was done with
ascending lower-limb contrast venography.
Assessment of study quality and data extraction
We adopted the study-quality criteria developed by
Schultz and colleagues10 to assess the studies included in
our meta-analysis. These criteria include proper generation
of the treatment allocation sequence; proper concealment
of the allocation sequence; masking of patient and the
investigator assessing clinical outcomes to treatment
allocation; and completeness of follow-up.
Two investigators (JWE, DJQ) independently extracted
data on study design, study quality, and the following
efficacy and safety outcomes during the extended-duration
phase: symptomatic venous thromboembolism, which
include deep vein thrombosis and pulmonary embolism;
symptomless deep vein thrombosis documented by
ascending lower limb contrast venography; major bleeding;
minor bleeding; and all-cause mortality. We accepted the
primary study investigators’ definitions for deep vein
thrombosis, pulmonary embolism, and major and minor
bleeding. The data abstracted for each trial were
confirmed by reviewer consensus, and then sent to the
primary investigator for verification. We requested missing
data from the primary investigator at this time.
Statistical analysis
We assessed agreement for study selection with a weighted
kappa statistic and by calculating the observed percentage
agreement. We used a fixed-effects model based on the
Mantel-Haenszel method for combining results from the
individual trials.11 This model is also known as an
assumption-free model since it does not assume that
included studies are a random sample of the universe of
studies, and provides a pooled estimate of treatment effect
that is conditional on the trial data that are available. We
calculated the odds ratio (OR), 95% CI, and the number
needed to treat (NNT) or harm (NNH). The test of
heterogeneity was calculated with the Mantel-Haenszel
method. All statistical calculations were done with
Comprehensive Meta Analysis (version 1.0.5).
Subgroup analyses
We prespecified the following subgroup analyses for
the primary outcome: hip replacement versus knee
replacement; agent (low molecular weight heparin,
unfractionated heparin, warfarin); duration of in-hospital
prophylaxis (up to 10 days, 10–14 days, 15 days or more);
and in trials with mandatory discharge venography versus
trials in which mandatory discharge venography was not
performed.
Sensitivity analyses
We did sensitivity analyses to further explore the
robustness of our results. To identify any study that may
have exerted a disproportionate influence on the summary
treatment effect, we deleted studies one at a time. We
examined the effect of excluding lower quality (open-label
studies and studies with incomplete follow-up) from the
analysis. An inverted funnel plot of treatment effect versus
study precision was created for the primary outcome to
look for possible publication bias,12 a technique that might
help to determine whether additional small studies might
have been undertaken but not published because of
unfavourable or negative results. Finally, results obtained
with a fixed effects model were compared with those
obtained using a random effects model.
Results
Figure 1 shows the process of study selection. Our search
identified 304 potentially eligible citations. After scanning
their titles and abstracts, 289 citations were excluded and
15 were retained for further assessment. Two studies
originally reported only in abstract form were subsequently
published as full reports5,8 and we obtained the completed
manuscript13 for a third abstract14 from the investigators.
These three studies were included. Of the 12 remaining
citations, one was excluded because it was not a properly
ARTICLES
10 THE LANCET • Vol 358 • July 7, 2001
Potentially eligible RCTs identified
and screened for retrieval (n=304)
Potentially appropriate RCTs to be
included in the meta-analysis (n=9)
RCTs included in the
meta-analysis (n=9)
RCTs retrieved for more
detailed assessment (n=15)
RCTs excluded
(not properly randomised n=1;
data previously reported, n=5)
RCTs excluded (n=0)
RCTs excluded (n=289)
Figure 1: Process of study selection
For personal use. Only reproduce with permission from The Lancet Publishing Group.
randomised study,15 and five were excluded because they
reported previously published data.16–20 Thus we had a total
of nine studies that met the inclusion criteria.3–5,8,13,21-24
Inter-rater agreement for study selection was 93%
(weighted kappa 0·88).
Table 1 shows the key features of study design. Three
studies were undertaken in Canada or the USA5,8,13 and six
were undertaken in one or more European countries.3,4,21–24
There was considerable variation in trial design. Seven
studies included only patients undergoing total hip
replacement,3-5,21–24 and two included both total hip and
knee replacements.8,13 Four low molecular weight heparin
preparations were assessed (ardeparin, dalteparin,
enoxaparin, nadroparin), and one study used fixed-dose
unfractionated heparin.23 No study assessed the use of
extended-duration warfarin. Prophylaxis was started
preoperatively in six studies,3,4,21–24 and postoperatively in
two.8,13 In one study,5low molecular weight heparin
prophylaxis was begun either preoperatively or postoper-
atively and patients underwent two randomisations, the
first to either low molecular weight heparin or warfarin
during the in-hospital phase, and the second to low
molecular weight heparin or placebo during the extended-
duration phase. There were differences in the duration of
in-hospital thromboprophylaxis, reflecting both variations
in clinical practice as well a tendency to earlier hospital
discharge in more recent studies. The older studies
continued in-hospital prophylaxis for 10–15 days,3,21 but
the two most recent studies used between 4 and 10 days of
in-hospital prophylaxis.5,8 However, the total duration of
prophylaxis was 4–6 weeks in all studies. Three studies
used bilateral venography,3–5 and two used duplex
ultrasound24 or impedance plethysmography,23 respectively,
to screen patients for the presence of symptomless deep
vein thrombosis before hospital discharge.
Proper methods were used to generate the randomised
treatment allocation in all studies, there seemed to be
adequate concealment of treatment allocation in all trials,
and both the patient and the investigator were unaware of
treatment allocation in seven of the nine studies.3-5,8,13,21,22
There was 100% clinical follow-up in six studies,3,4,8,21,23 and
greater than 90% follow-up in the remaining three
studies.5,13,24
Table 2 shows the data on symptomatic venous
thromboembolism. All studies showed a reduction in
symptomatic venous thromboembolism, during extended-
duration prophylaxis, but a statistically significant
reduction was seen in only two of the nine trials.13,21 The
pooled estimate from all the trials revealed a highly
significant reduction in symptomatic events with extended-
duration prophylaxis compared with placebo or untreated
control, with no statistical evidence of heterogeneity
among the studies (figure 2). When individual components
of this outcome were considered separately (table 2), there
was a similar relative risk reduction for deep vein
thrombosis and pulmonary embolism, but the absolute
event rate for pulmonary embolism was low (total number
of events=14) and the relative risk reduction not
significant.
The reduction in symptomatic venous thrombo-
embolism seen with extended-duration prophylaxis was
greatest in patients undergoing total hip replacement
(1·4% vs 4·3%, 0·33; 0·19–0·56, 34), and was similar for
deep vein thrombosis (1·4% vs 3·6%, 0·39; 0·23–0·68, 45)
and for pulmonary embolism (0% vs 0·7%, 0·32;
0·11–0·96, 143). In the knee replacement group there was
a more modest reduction in venous thromboembolism
with extended-duration prophylaxis (1·0% vs 1·4%, 0·74;
0·26–2·15, 250), but this estimate was based on aggregate
data only from two studies with seven events and was not
ARTICLES
THE LANCET • Vol 358 • July 7, 2001 11
Trial Eligibility In-hospital thromboprophylaxis Out-of-hospital thromboprophylaxis Total dura-
n Agent Duration (days) VTE screening‡n Active agent§tion (days)¶
Bergqvist et al, 199621 THR 288 Enoxaparin 40 mg 10–11 Nil 262 Enoxaparin 40 mg 30 (2)
once a day once a day
Planes et al, 19963THR 210 Enoxaparin 40 mg 13–15 Bilateral venography 179 Enoxaparin 40 mg 35 (2)
once a day once a day
Dahl et al, 19974THR 308 Dalteparin* 7 (2) Bilateral venography 265 Dalteparin 35 (2)
5000 IU once a day 5000 IU once a day
Lassen et al, 199822 THR 300 Dalteparin 7 Nil 281 Dalteparin 35
5000 IU once a day 5000 IU once a day
Manganelli et al, 199823 THR 80 UFH 5000 IU 15 IPG 79 UFH 5000 IU 30
three times a day three times a day
NPHDO, 199824 THR 346 Nadroparin 16–17 Duplex ultrasound 296 Nadroparin 37–38
weight-adjusted weight-adjusted
Heit et al, 20008THR 1320 Ardeparin 4–10 Nil 1195 Ardeparin 100 IU/kg 42
or TKR 50 IU/kg twice a day once a day
Hull et al, 20005THR 991 Dalteparin 5000 IU 6 (2) Bilateral venography 569 Dalteparin 35 (2)
once a day 5000 IU once a day
or warfarin†
Comp et al, 200113 THR 968 Enoxaparin 7–10 Nil 873 Enoxaparin 27–29
or TKR 30 mg twice a day 40 mg once a day
DVT=deep vein thrombosis; IPG=impedance plethysmography; THR=total hip replacement; TKR=total knee replacement; UFH=unfractionated heparin; VTE=venous
thromboembolism. *All patients also received dextran –70 (day 0 and day 1). †Patients were randomised to either dalteparin for 35 (2) days or warfarin 6 (2) days
followed by out-of-hospital placebo until 35 (2) days. ‡All trials excluded patients with objectively confirmed symptomatic VTE during the in-hospital phase from the out-
of-hospital phase; Planes et al and Hull et al excluded patients with any symptomless venographic DVT while Dahl et al excluded only symptomless proximal DVT. §In
each trial the randomised comparator was either placebo or untreated control. ¶Screening bilateral venography done at completion of randomised treatment in all trials
except Manganelli et al (who did unilateral venography on the operative leg on day 45) and Heit et al.
Table 1: Design of trials included in the meta-analysis.
Outcome Low molecular weight heparin Placebo or untreated control OR (95% CI) NNT
or unfractionated heparin
Deep vein thrombosis 22/1964 (1·1%) 47/1742 (2·7%) 0·41 (0·24–0·68)* 62
Pulmonary embolism 3/1961 (0·2%) 11/1744 (0·6%) 0·43 (0·17–1·06)†250
Any venous thromboembolism 25/1964 (1·3%) 58/1744 (3·3%) 0·38 (0·24–0·61)‡50
*Heterogeneity: 2=4·46 (df=8), p=0·81. †Heterogeneity: 2=3·69 (df=8), p=0·88. ‡Heterogeneity: 2=5·65 (df = 8), p=0·69
Table 2: Out-of-hospital symptomatic venous thromboembolism
For personal use. Only reproduce with permission from The Lancet Publishing Group.
significant. Nevertheless, the direction and magnitude of
the treatment effect during knee replacement was similar
for the overall outcome and for the two components of this
outcome, deep vein thrombosis (0·5% vs 0·7%, 0·75;
0·17–3·33, 500) and pulmonary embolism (0·5% vs 0·7%,
0·77; 0·19–3·13, 500). There was no evidence of
heterogeneity between the hip and knee replacement trials
for the outcome of symptomatic venous thromboembolism
(2value 1·87, p=0·98), with clear overlap of the respective
95% CI.
One open-label trial of 79 patients,23 used unfractionated
heparin, whereas eight studies used low molecular weight
heparin during the extended-duration phase. Although
indirect comparisons suggest that there was a greater
treatment effect with unfractionated heparin (0% vs
15·8%, 0·06; 0·01–1·11, 6) compared with low molecular
weight heparin (1·3% vs 3·0%, 0·42; 0·26–0·68, 59), there
was complete overlap of the 95% CI and no evidence of
heterogeneity between unfractionated heparin and low
molecular weight heparin trial results (2 value 2·48,
p=0·96).
The duration of in-hospital thromboprophylaxis could
affect the frequency of venous thromboembolism during
the extended treatment phase and, thereby, also the
magnitude of the observed treatment effect of extended-
duration prophylaxis. However, on the basis of an indirect
comparison of placebo or untreated control groups across
the trials, the frequency of symptomatic venous
thromboembolism was no higher in patients receiving
prophylaxis for 5–10 days of in-hospital prophylaxis (2·4%;
1·6–3·3%) compared with 10–15 days (8·3%; 4·9–13·0%)
or 15 or more days (5·0%; 2·3–9·3%), though the
confidence intervals were wide. The overall treatment
effect of extended-duration prophylaxis was also similar in
these three groups of trials, with no statistical evidence of
heterogeneity (data not presented).
Mechanical injury to the veins of the lower limbs
induced during bilateral ascending venography at hospital
discharge could cause deep vein thrombosis and thereby
inflate the observed incidence of deep vein thrombosis,
although a low incidence of postdischarge symptomatic
venous thromboembolism (1·5%) was reported in an
overview of 13 trials of 2361 patients with a normal
venogram at time of hospital discharge.25 On the other
hand, exclusion of patients with subclinical deep vein
thrombosis detected by venography at the time of hospital
discharge might reduce the subsequent frequency of
symptomatic venous thromboembolism.
Our data showed a 4·9% (2·8–7·8%) rate of out-of-
hospital symptomatic deep vein thrombosis among
placebo-treated or untreated controls in trials in which
mandatory bilateral venography was done at hospital
discharge compared with 3·0% (2·2–4·0%) in trials not
undertaking discharge venography. This difference of
1·9% was not significant (95% CI 0·5% to 4·4%), and
there was overlap of the respective 95% CI for these
estimates.
Extended-duration prophylaxis was associated was a
similar relative risk reduction in symptomatic venous
thromboembolism in trials where discharge venography
was not undertaken (0·9% vs 3·0%, 0·32; 0·18–0·58, 48)
compared with trials in which mandatory discharge
venography was done (2·2% vs 4·9%, 0·53; 0·24–1·18, 37)
with overlap of the 95% CI and no statistical evidence of
heterogeneity (2 value0·23, p=0·99).
Table 3 shows data on symptomless venous thrombo-
embolism. Extended-duration prophylaxis was associated
with a statistically significant reduction in deep vein
thrombosis in patients undergoing hip or knee
replacement, with no statistical evidence of heterogeneity
(2value 5·81, p=0·44; figure 3). Both proximal and distal
deep vein thrombosis were significantly reduced.
When trials of hip replacement were examined
separately, extended-duration prophylaxis was associated
with a substantial reduction in the frequency of symptom-
free deep vein thrombosis (7·7% vs 18·6%, 0·39;
ARTICLES
12 THE LANCET • Vol 358 • July 7, 2001
Study
Planes et al, 1996
Bergqvist et al, 1996
Dahl et al, 1997
NPHDO, 1998
Manganelli et al, 1998
Lassen et al, 1998
Hull et al, 2000
Hull et al, 2000
Comp et al, 2001
Heparin
3/85 (3·5%)
2/117 (1·7%)
4/114 (3·5%)
1/115 (0·6%)
0/41 (0%)
2/113 (1·8%)
4/291 (1·4%)
7/607 (1·2%)
2/441 (0·4%)
Control
7/88 (8·0%)
10/116 (8·6%)
6/106 (5·7%)
3/141 (2·1%)
6/38 (1·6%)
3/102 (2·9%)
3/133 (2·3%)
10/588 (1·7%)
10/432 (2·3%)
OR
0·42
0·18
0·61
0·30
0·06
0·59
0·60
0·67
0·19
95% CI
0·11–1·69
0·04–0·86
0·17–2·21
0·03–2·90
0·01–1·11
0·10–3·63
0·13–2·74
0·26–1·78
0·04–0·88
0·01 0·1 1 10 100
Log odds ratio
Favours heparin Favours control
Total 25/1964 (1·3%) 58/1744 (3·3%) 0·38 0·24–0·61
Figure 2: Symptomatic venous thromboembolism in trials comparing extended-duration low molecular weight heparin or unfractionated
heparin with placebo or untreated control after hip or knee arthroplasty
Outcome Low molecular weight heparin Placebo or untreated control OR (95% CI) NNT
or unfractionated hepanin
Distal deep vein thrombosis 71/1047 (6·8%) 89/854 (10·4%) 0·69 (0·49–0·96)* 28
Proximal deep vein thrombosis 30/1047 (2·9%) 78/854 (9·1%) 0·33 (0·21–0·51)†16
Any deep vein thrombosis 101/1047 (9·6%) 167/854 (19·6%) 0·48 (0·36–0·63)‡10
*Heterogeneity: 2=5·89 (df=6), p=0·43. †Heterogeneity: 2=4·51 (df=6), p=0·61. ‡Heterogeneity: 2=5·81 (df=6), p=0·44
Table 3: Out-of-hospital symptomless venographic deep vein thrombosis
For personal use. Only reproduce with permission from The Lancet Publishing Group.
0·28–0·54, 9), but the risk reduction in knee replacement
patients was more modest (20·5% vs 24·0%, 0·82;
0·49–1·40, 29) and was not significant. This pattern is
similar to that seen for symptomatic events, and supports
the notion that venous thromboembolism might be more
difficult to prevent in patients undergoing knee
replacement.
Table 4 shows the data on bleeding. Among a total of
more than 2000 patients randomly assigned to extended-
duration prophylaxis, the frequency of major bleeding was
no higher than in those receiving placebo and untreated
controls. There was, however, a significant excess of minor
bleeding with extended-duration prophylaxis with no
statistical evidence of heterogeneity.
Table 4 shows all-cause mortality data. A total of eight
deaths were reported, 3/2115 (0·1%; 0–0·4%) in patients
receiving extended-duration prophylaxis and 5/1855
(0·3%; 0·1–0·6%) in patients receiving placebo or
untreated controls.
Deleting individual studies did not significantly alter our
primary outcome. Likewise, when we removed open-label
trials,23,24 our results were not altered much. A funnel plot
of effect size versus study precision was fairly symmetrical
with a similar number of studies on either side of the
summary treatment effect for symptomatic venous
thromboembolism (figure not shown). This finding is
consistent with no major publication bias. There were no
important differences between results obtained using the
fixed versus a random-effects model.
Discussion
We have shown that extended-duration prophylaxis with
low molecular weight heparin or unfractionated heparin in
patients undergoing major hip or knee replacement surgery
significantly reduces the risk of symptomatic venous
thromboembolism. This benefit is achieved with no excess
major bleeding but with increased minor bleeding. About
50 patients need to be treated to prevent one episode of
symptomatic venous thromboembolism, whereas a minor
bleed occurs in every 80 patients treated. A significant
reduction in pulmonary embolism or mortality could not
be shown but the direction of the treatment effect for these
outcomes was consistent with that seen for the primary
outcome, suggesting that a similar reduction for these
outcomes might be expected.
Our findings of a significant treatment benefit of
extended-duration prophylaxis for prevention of symptom-
atic venous thromboembolism are based on a pooled
analysis of both hip and knee replacement trials. However,
our meta-analysis suggests that extended-duration
prophylaxis might be less effective in knee replacement
then in hip replacement. This possibility is also biologically
plausible since more extensive disruption of soft tissue and
bone occurs during knee replacement surgery, which is
likely to be associated with extensive local release of
prothrombotic tissue factor. Disruption of vascular
anatomy and the use of occlusive tourniquets can also
promote local thrombus formation,26 potentially further
reducing the efficacy of thromboprophylaxis in patients
undergoing knee replacement surgery. However, our study
was not sufficiently powered to show clear evidence of
heterogeneity of treatment effect on symptomatic
outcomes between hip and knee replacement surgery.
Results from prospective cohort studies have shown that
most symptomless thrombi in patients undergoing hip or
knee replacement remain clinically silent. This raises
questions about the clinical relevance of symptomless deep
vein thrombosis detected by screening venography.
However, in our study, the reduction in symptomatic
thrombi with extended-duration prophylaxis was
paralleled by a similar reduction in symptomless
venographic deep vein thrombosis. These findings support
the idea that symptomless thrombosis detected by
screening venography is a valid surrogate for symptomatic
events.
One widely recognised impediment to the more
widespread use of effective prophylaxis in patients
undergoing hip or knee replacement is concern about the
risk of bleeding.1In this regard, the data from our study are
reassuring in that, even with sustained use for up to
6 weeks, prophylactic-dose heparin is not associated with
an excess in major bleeding, and the excess in minor
ARTICLES
THE LANCET • Vol 358 • July 7, 2001 13
S
tudy
Planes et al, 1996
Bergqvist et al, 1996
Dahl et al, 1997
Manganelli et al, 1998
Lassen et al, 1998
Hull et al, 2000
C
omp et al, 2001
Heparin
3/85 (3·5%)
21/117 (17·9%)
11/93 (11·8%)
4/33 (12·1%)
3/113 (2·7%)
10/291 (3·4%)
49/315 (15·5%)
Control
10/88 (11·4%)
43/116 (37·1%)
23/89 (25·8%)
2/28 (7·1%)
9/102 (8·8%)
11/133 (8·3%)
69/298 (23·2%)
OR
0·28
0·37
0·38
1·79
0·28
0·39
0·61
95% CI
0·08–1·08
0·20–0·68
0·18–0·85
0·30–10·6
1
0·07–1·07
0·16–0·95
0·41–0·95
0·1110
Log odds ratio
Favours heparin Favours control
T
otal 101/1047 (9·6%) 167/854 (19·6%) 0·48 0·36–0·63
Outcomes Low molecular weight heparin Placebo or untreated control OR (95% CI) NNH
or unfractionated heparin
Minor bleeding 77/2073 (3·7%) 46/1834 (2·5%) 1·56 (1·08–2·26)* 83
Major bleeding 2/2114 (0·1%) 5/1872 (0·3%) 0·62 (0·22–1·75)†..
Death 3/2132 (0·1%) 5/1888 (0·3%) 0·68 (0·25–1·88)‡..
*Heterogeneity: 2=7·34 (df=7), p=0·39. †Heterogeneity: 2=0·58 (df=8), p=1·00. ‡Heterogeneity: 2=1·39 (df=8), p=0·99
Table 4: Out-of-hospital data: minor and major bleeding, and death
Figure 3: Symptomless venous thromboembolism in trials comparing extended-duration low molecular weight heparin or unfractionated
heparin with placebo or untreated control after hip or knee arthroplasty
For personal use. Only reproduce with permission from The Lancet Publishing Group.
bleeding is modest. Nevertheless, the excess in minor
bleeding cannot be discounted since this can represent a
source of discomfort and inconvenience for the patient,
particularly if bleeding occurs at the operative site.
The robustness of our primary outcome is shown by the
results of sensitivity analyses. Removal of individual
studies or studies of lower quality did not affect
significantly our primary outcome, results were similar
irrespective of the statistical approach that was used, and
there was a consistent directional treatment effect across
both primary and secondary efficacy outcomes.
Our study has several potential limitations. First, despite
examining the totality of the evidence by pooling results
from all the available properly randomised trials, the total
number of patients randomly assigned and the number of
outcome events was modest. Therefore, our meta-analysis
lacked statistical power to provide precise estimates of
frequency and treatment effect for clinically important
outcomes such as pulmonary embolism. However, deep
vein thrombosis and pulmonary embolism represent
clinical manifestations of the same underlying disease
process. Therefore, strategies that are effective for the
prevention of deep vein thrombosis, especially those
proximally located, are likely also to be effective for the
prevention of non-fatal and fatal pulmonary embolism.
Second, there was a lot of variation in the design of studies
included in our meta-analysis. However, differences
among trials are inevitable since individual trials look at
different populations with different treatment protocols,
and there is always some heterogeneity, even within
individual trials.27,28 Differences in trial design do not
necessarily preclude pooling of their results since, in a
meta-analysis, individual patients are directly compared
only with other patients within the same trial, and not
across the trials. The validity of our approach is further
supported by the external consistency of our findings with
the results of individual trials, and the lack of statistical
evidence of heterogeneity for any of the outcomes
examined in our meta-analysis. Third, although
compliance with extended-duration prophylaxis seemed to
be 90% or higher in most trials included in our meta-
analysis, the definition of compliance varied among the
studies and not all studies reported the level of compliance
with randomised treatment allocation that was achieved.
However, the probable impact of lack of compliance is to
reduce the power of a randomised trial to detect a
significant treatment benefit. Therefore an even greater
benefit of extended-duration prophylaxis might be realised
in populations in which higher levels of compliance are
achieved.
Fourth, there are several antithrombotic agents besides
heparin and low molecular weight heparin that might be
used for out-of-hospital prophylaxis, which were not
addressed in our meta-analysis. In particular, warfarin
seems to be effective for the long-term prevention of
venous thromboembolism after hip or knee replacement
surgery.29 However, to our knowledge, there are currently
no randomised trials that have compared warfarin with
placebo or untreated control for this indication.
Meanwhile, there are also emerging data that aspirin might
be effective for the prevention of venous thromboembolism
after hip fracture,30 but this possibility requires further
confirmation in randomised trials. Finally, meta-analysis
remains retrospective research that is subject to publication
bias and the methodological deficiencies of the included
studies. However, we kept the likelihood of bias to a
minimum by developing a detailed protocol before starting
this study, undertaking a meticulous and exhaustive search
for both published and unpublished studies, and using
explicit methods for study selection, data extraction, and
data analysis. Further, we considered the totality of the
randomised evidence by including all relevant properly
randomised trials, and we approached investigators from
each study to verify and, if necessary, update the data that
were extracted from their trial reports.
On the basis of an absolute risk reduction of about 2% in
symptomatic venous thromboembolism, we estimate that
the routine use of extended-duration prophylaxis will
prevent about an additional 20 symptomatic thrombi for
every 1000 elective hip or knee replacement patients
treated. Assuming a case-fatality rate of 5%,31 this figure is
equivalent to preventing one additional death for every
1000 patients treated. However, the cost-effectiveness of
implementing routine extended-duration prophylaxis
remains to be shown and is subject to the constraints of
different health-care systems. Estimates are that the
outpatient cost of the most widely used agents to prevent
venous thromboembolism is about US$4–US$7 in the
UK32 and US$24–US$28 per day in the USA.33 Although
attempts at further defining high-risk groups for targeted
prophylaxis may pose substantial challenges,33 the
increasing use of hip and knee arthroplasty in Europe,34 as
well as North America,35 highlights the potential magnitude
of the future burden of venous thromboembolic disease in
orthopaedic patients and further emphasises the need to
reliably predict risk through adequately powered studies.
Contributors
John Eikelboom, Daniel Quinlan, and James Douketis designed the
protocol; John Eikelboom and Daniel Quinlan conducted the literature
search, selected the studies, extracted the data, and contacted the authors;
John Eikelboom performed the statistical analyses; John Eikelboom and
Daniel Quinlan wrote the first draft of the manuscript; all investigators
contributed to the writing of the paper.
Acknowledgments
We thank the authors of primary studies included in our meta-analysis
(D Bergqvist, M R Lassen, O E Dahl, J A Heit, A Palla, G F Pineo,
A Planes) as well as Michèle De Knock from Sanofi-Synthelabo, Carin
Fellenius from Pharmacia Corp, and Theodore Spiro from Aventis Pharma
for their assistance in verifying the accuracy of the data and providing
missing data. We also thank Michael Borenstein for providing the statistical
software to conduct this meta-analysis.
This study received no external funding. James Douketis is the recipient
of a research scholarship from the Heart and Stroke Foundation of
Canada. Daniel Quinlan has received educational support as an
Investigator from companies making low molecular weight heparins.
References
1 Geerts WH, Heit JA, Clagett GP, et al. Prevention of venous
thromboembolism. Chest 2001; 119: 132S–75.
2 Hirsh J. Evidence for the needs of out-of-hospital thrombosis
prophylaxis: introduction. Chest 1998; 114: 1135–45.
3 Planes A, Vochelle N, Darmon JY, Fagola M, Bellaud M, Huet Y. Risk
of deep-venous thrombosis after hospital discharge in patients having
undergone total hip replacement: double-blind randomised comparison
of enoxaparin versus placebo. Lancet 1996; 348: 224–28.
4 Dahl OE, Andreassen G, Aspelin T, et al. Prolonged
thromboprophylaxis following total hip replacement surgery—results of
a double-blind, prospective, randomised, placebo-controlled study with
dalteparin (Fragmin). Thromb Haemost 1997; 77: 26–31.
5 Hull RD, Pineo GF, Francis C, et al. Low-molecular-weight heparin
prophylaxis using dalteparin extended out-of-hospital vs in-hospital
warfarin/out-of-hospital placebo in hip arthroplasty patients: a double-
blind, randomized comparison. North American Fragmin Trial
Investigators. Arch Intern Med 2000; 160: 2208–15.
6 Robinson KS, Anderson DR, Gross M, et al. Ultrasonic screening
before hospital discharge for deep vein thrombosis after arthroplasty:
the post-arthroplasty screening study—a randomised controlled trial.
Ann Intern Med 1997; 127: 439–45.
7 Leclerc JR, Gent M, Hirsh J, Ginsberg JS, Geerts WH. The incidence
of venographic venous thromboembolism during and after prophylaxis
with enoxaparin: a multi-institutional cohort study in patients who
underwent hip or knee arthroplasty. Arch Intern Med 1998;
158: 873–78.
ARTICLES
14 THE LANCET • Vol 358 • July 7, 2001
For personal use. Only reproduce with permission from The Lancet Publishing Group.
8 Heit JA, Elliott CG, Trowbridge AA, Morrey BF, Gent M, Hirsh J, for
the Ardeparin Arthroplasty Study Group. Ardeparin sodium for
extended out-of-hospital prophylaxis against venous thromboembolism
after total hip or knee replacement. A randomized, double-blind,
placebo-controlled trial. Ann Intern Med 2000; 132: 853–61.
9 Nicolaides AN. Prevention of venous thromboembolism: International
Consensus Statement—guidelines compiled in accordance with the
scientific evidence. Int Angiol 2001; 20: 1–37.
10 Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of
bias. Dimensions of methodological quality associated with estimates of
treatment effects in controlled trials. JAMA 1995; 273: 408–12.
11 Mantel N, Haenszel W. Statistical aspects of the analysis of data from
retrospective studies of disease. J Natl Cancer Inst 1959; 22: 719–48.
12 Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-
analysis detected by a simple, graphical test. BMJ 1997; 315: 629–34.
13 Comp PC, Spiro TE, Friedman RJ, et al. Prolonged enoxaparin
therapy to prevent venous thromboembolism after primary hip or knee
replacement. J Bone Joint Surg Am 2001; 83: 336–45.
14 Spiro TE. A double-blind multicenter clinical trial comparing long
term enoxaparin and placebo treatments in the prevention of venous
thromboembolic disease after hip and knee replacement surgery: the
enoxaparin clinical trials group. Blood 1997; 90: 295A.
15 Caprini JA, Arcelus JI, Motykie G, Kudrna JC, Mokhtee D, Reyna JJ.
The influence of oral anticoagulation therapy on deep vein thrombosis
rates four weeks after total hip replacement. J Vasc Surg 1999;
30: 813–20.
16 Planes A, Vochelle N, Darmon JY, et al. Efficacy and safety of post
discharge administration of enoxaparin in the prevention of deep
venous thrombosis after total hip replacement: a prospective
randomised double-blind placebo-controlled trial. Drugs 1996;
52: 47–54.
17 Planes A, Vochelle N, Darmon JY. Out-of-hospital prophylaxis with
low-molecular-weight heparin in hip surgery: the French study—
venographic outcome at 35 days. Chest 1998; 114: 125–129.
18 Planes A, Vochelle N. The post-hospital discharge venous thrombosis
risk of the orthopedic patient. Orthopedics 1997; 20: 18–21.
19 Nilsson PE, Bergqvist D, Benoni G, et al. The post-discharge
prophylactic management of the orthopedic patient with low-
molecular-weight heparin: enoxaparin. Orthopedics 1997; 20: 22–25.
20 Davidson BL. Out-of-hospital prophylaxis with low-molecular-weight
heparin in hip surgery: the Swedish study. Chest 1998; 114: 1305–25.
21 Bergqvist D, Benoni G, Bjorgell O, et al. Low-molecular-weight
heparin (enoxaparin) as prophylaxis against venous thromboembolism
after total hip replacement. N Engl J Med 1996; 335: 696–700.
22 Lassen MR, Borris LC, Anderson BS, et al. Efficacy and safety of
prolonged thromboprophylaxis with a low molecular weight heparin
(dalteparin) after total hip arthroplasty—the Danish Prolonged
Prophylaxis (DaPP) Study. Thromb Res 1998; 89: 281–87.
23 Manganelli D, Pazzagli M, Mazzantini D, et al. Prolonged prophylaxis
with unfractionated heparin is effective to reduce delayed deep vein
thrombosis in total hip replacement. Respiration 1998; 65: 369–74.
24 Haentjens P. Venous thromboembolism after total hip arthroplasty: a
review of incidence and prevention during hospitalization and after
hospital discharge. Acta Orthop Belg 2000; 66: 1–8.
25 Ricotta S, Iorio A, Parise P, Nenci GG, Agnelli G. Post-discharge
clinically overt venous thromboembolism in orthopaedic surgery
patients with negative venography—an overview analysis. Thromb
Haemost 1996; 76: 887–92.
26 Sharrock NE, Sculco TP, Ranawat CS, Maynard MJ, Harpel PC.
Changes in circulatory indices of thrombosis and fibrinolysis during
total knee arthroplasty performed under tourniquet. J Arthroplasty
1995; 10: 523–28.
27 Lau J, Ioannidis JPA, Schmid CH. Summing up the evidence: one
answer is not always enough. Lancet 1997; 351: 123–27.
28 Thompson SG. Why sources of heterogeneity in meta-analysis should
be investigated. BMJ 1994; 309: 1351–55.
29 White RH, Gettner S, Newman JM, Trauner KB, Romano PS.
Predictors of rehospitalization for symptomatic venous
thromboembolism after total hip arthroplasty. N Engl J Med 2000; 343:
1758–64.
30 Pulmonary Embolism Prevention (PEP) Trial Collaborative Group.
Prevention of pulmonary embolism and deep vein thrombosis with low
dose aspirin: Pulmonary Embolism Prevention (PEP) Trial. Lancet
2000; 355: 1295–302.
31 Douketis JD, Kearon C, Bates S, Duko EK, Ginsberg JS. Risk of fatal
pulmonary embolism in patients with treated venous
thromboembolism. JAMA 1998; 279: 458–62.
32 British National Formulary, 40th edn. London: BMJ Books; 2000.
33 Davidson BL, Lensing AWA. Prolonged prophylaxis after joint
replacement: another step sideways? Ann Intern Med 2000; 132:
914–15.
34 Rajaratnam G, Black NA, Dalziel M. Total hip replacements in the
National Health Service: is need being met? J Public Health Med 1990;
12: 56–59.
35 Madhok R, Lewallen DG, Wallrichs SL, Ilstrup DM, Kurland RL,
Melton LJ 3rd. Trends in the utilization of primary total hip
arthroplasty, 1969 through 1990: a population-based study in Olmsted
County, Minnesota. Mayo Clin Proc 1993; 68: 11–18.
ARTICLES
THE LANCET • Vol 358 • July 7, 2001 15
Further information on the design and results of studies included can be found on The Lancet’s website www.thelancet.com in WEBTables 1–4.
