Regional Anesthesia

Section Editor: Terese T. Horlocker

A Comparison of Neuraxial Block Versus General

Anesthesia for Elective Total Hip Replacement:
A Meta-Analysis
William J. Mauermann, MD BACKGROUND: A recent meta-analysis showed that compared with general anesthe-
sia (GA), neuraxial block reduced many serious complications in patients under-
Ashley M. Shilling, MD going various types of surgeries. It is not known whether this finding from
studying heterogeneous patient groups is applicable to a particular surgical patient
population. We performed the present meta-analysis to determine whether anes-
Zhiyi Zuo, MD, PhD
Downloaded from https://journals.lww.com/anesthesia-analgesia by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3Gamkn0m7hy6Mhb8DYgpUeidgnZnFSyxNu2Qty0rsHh0= on 07/12/2018

thesia choice affected the outcome after elective total hip replacement (THR).
METHODS: Medline (1966 to August 2005), MD Consult (1966 to August 2005),
BIOSIS (1969 to August 2005), and EMBASE (1969 to August 2005) databases were
searched. Randomized and quasirandomized studies comparing GA and neuraxial
(spinal or epidural) block for elective THR were included in this analysis.
RESULTS: Ten independent trials, involving 330 patients under GA and 348 patients
under neuraxial block, were identified and analyzed. Pooled results from five trials
showed that neuraxial block significantly decreased the incidence of radiographi-
cally diagnosed deep venous thrombosis or pulmonary embolism. The odds ratio
(OR) for deep venous thrombosis was 0.27 with 95% confidence interval (CI)
0.17– 0.42. The OR for pulmonary embolism was 0.26 with 95% CI 0.12– 0.56.
Neuraxial block also decreased the operative time by 7.1 min/case (95% CI 2.3–11.9
min) and intraoperative blood loss by 275 mL/case (95% CI 180 –371 mL). Data
from three trials showed that patients under neuraxial block for THR were less
likely to require blood transfusion than were patients under GA (21/177 ⫽ 12% vs
62/188 ⫽ 33% of patients transfused, P ⬍ 0.001 by z-test). The OR for this
comparison was 0.26. However, the CIs were wide and compatible with both no
effect and a nine-tenths reduction (95% CI 0.06 –1.05).
CONCLUSIONS: Patients undergoing elective THR under neuraxial anesthesia seem to
have better outcomes than those under GA.
(Anesth Analg 2006;103:1018 –25)

H ip replacement is a common orthopedic proce-

dure, generally performed in elderly patients. In 2002,
surgical procedures resulted in a decrease in mortal-
ity, venous thromboembolism, myocardial infarction,
343,000 patients underwent 345,000 hip replacement and several other complications. However, they were
procedures in the United States (1). Despite the com- unable to draw conclusions regarding the validity of
mon occurrence of this procedure, there is controversy these findings to specific surgical procedures or pa-
as to whether total hip replacement (THR) is best tient population. Previous work has shown that THR
performed under neuraxial block, including epidural under neuraxial blockade may be associated with less
and spinal block, or general anesthesia (GA). In 2000, deep venous thrombosis (DVT) (3–7) and pulmonary
Rodgers et al. (2) published a meta-analysis showing embolism (PE) (3–5), and a reduced intraoperative
that the use of neuraxial techniques for a variety of estimated blood loss and transfusion requirement
(3,7–11) when compared with those under GA. How-
ever, many of these studies are now approximately 20
yr old. Neuraxial block may also decrease the time
From the Department of Anesthesiology, University of Virginia
Health System, Charlottesville, Virginia. needed to discharge the patient from the postanesthe-
Accepted for publication June 5, 2006. sia care unit (12) and provide stable intraoperative
Supported by Department of Anesthesiology, University of hemodynamics (13), but these benefits of neuraxial
Virginia; and National Institute of Health Grants R01 GM065211 block are not consistently shown in other studies
and R01 NS045983.
(7,13,14). In addition, many of these studies have been
Address correspondence and reprint requests to Dr. Zhiyi Zuo,
Department of Anesthesiology, University of Virginia Health Sys- hindered by relatively small cohorts of patients and, in
tem, 1 Hospital Drive, PO Box 800710, Charlottesville, VA 22908- some cases, the relatively rare occurrence of clinically
0710. Address e-mail to [email protected]. significant morbidity. Thus, it has been difficult to
Copyright © 2006 International Anesthesia Research Society draw conclusions regarding the effects of anesthesia
DOI: 10.1213/01.ane.0000237267.75543.59
choice on the outcomes for THR.

1018 Vol. 103, No. 4, October 2006

Table 1. Characteristics of the Studies Contributing Data to this Meta-Analysis
Patient numbers
Outcome DVT Neuraxial
Source Design Neuraxial General measures prophylaxis technique
Keith (1977) (11) Randomized 10 9 Blood loss Dextran given Single-injection
prospective postoperatively epidural

Hole (1980) (14) Randomized 29 31 Operative time, 5000 units sc Single-injection

prospective PE, transfusion heparin from epidural
volume, DVT Day 1–7

Thorburn (1980) (6) Quasirandomized 47 38 Blood loss, Not noted Spinal

prospective transfusion
volume, DVT

Modig (1981) (3) Randomized 15 15 DVT, PE, blood None Continuous

prospective loss, transfusion epidural for
volume, 24 h
operative time

Modig (1983) (4) Quasirandomized 30 30 Blood loss, DVT, None Continuous

prospective PE, operative epidural for
time 24 h

Modig (1986) (5) Prospective 48 46 Operative time, None Continuous

randomized blood loss, epidural for
DVT, PE 24 h

Davis (1987, Randomized 69 71 DVT, PE, Stockings Spinal

1989)a(7,16) prospective hemostatic
markers,
operative time,
blood loss,
patients
transfused

Modig (1987) (17) Randomized 14 10 Operative time, None Continuous

prospective blood loss epidural for
24 h

Borghi (2002,2005)b Randomized 70 70 Hypotension, Not noted Continuous

(13,15) prospective bradycardia, epidural
operative time,
intraoperative
blood losses,
patients
transfused

Brueckner Randomized 16 10 Hemostatic Stockings and Spinal

(2003) (18) prospective markers, low molecular
operative time, heparin
transfusion (madroparin)
volume given
preoperatively
a
Davis et al. presented their final results from the same groups of patients in two publications in 1987 and 1989 (7,16).
b
Borghi et al. reported their findings from the same groups of patients in two papers published in 2002 and 2005 (13,15).

We performed this meta-analysis to test the hypoth- and intra- and postoperative outcome measurements
esis that elective THR under neuraxial block was asso- such as number of patients with DVT, PE, and mortality.
ciated with improved outcomes compared with the
surgery under GA. We focused our analysis on elective
THR to reduce many confounding factors, such as blood METHODS
loss before the procedure, in patients with hip fracture Medline (1966 to August 2005), MD Consult (1966
and trauma. We chose to analyze intraoperative out- to August 2005), BIOSIS (1969 to August 2005), and
come measurements including operative time, estimated EMBASE (1969 to August 2005) databases were inde-
intraoperative blood loss, and transfusion requirements pendently searched by two authors (WJM and AMS)

Vol. 103, No. 4, October 2006 © 2006 International Anesthesia Research Society 1019
Figure 1. Comparison of operative time for
elective total hip replacement under
neuraxial block versus under general anes-
thesia (GA). N ⫽ number of patients.

using the following keywords: total hip replacement, estimated intraoperative blood loss, number of pa-
epidural anesthesia, spinal anesthesia, general anes- tients requiring blood transfusion and the transfusion
thesia, hip fracture, deep venous thrombosis, regional volume, operative time, number of patients with DVT
anesthesia, elective hip surgery, and pulmonary em- or PE who were diagnosed radiographically, and the
bolism. The terms “epidural anesthesia,” “spinal an- associated mortality. The decision on the suitability of
esthesia,” and “general anesthesia” were linked with a study for our analysis and the extracted data by the
“or” and combined using “and” with each subsequent two reviewers/authors were compared. Discrepancy
term. No language limits were used. Bibliographies among them was resolved by discussion and recon-
were also searched for relevant publications. firming the data in the original paper. We contacted
All publications found during the search were the authors if multiple publications on the subject
manually and independently reviewed by the same were from the same authors to verify that the data in
two authors. Randomized and quasirandomized stud- each of the multiple publications were from indepen-
ies comparing the outcomes of elective THR under dent patient groups. Data of continuous parameters
neuraxial block and GA were included in the analysis. must have been presented in numerical format in the
Quasirandomized studies are studies in which pa- study to have been included in our analysis, whereas
tients are assigned into study groups by alteration the data in nontabular format (i.e., bar or line graphs)
based on variables such as surgical dates. Study were not included, as accurate numbers could not be
inclusion was limited to patient groups that under- assured.
went THR under either neuraxial block or GA. We did Meta-analysis was performed with the MedCalc
not include patients who had THR under combined software (Mariakerke, Belgium). Patients who had GA
techniques, nor did we include studies that compared were treated as control groups, and patients with
controlled hypotension patients under GA with pa- neuraxial block were treated as intervention groups.
tients under neuraxial block. The following outcome Odds ratio (OR) and 95% confidence intervals (CI)
data were extracted from each study if reported: were reported for dichotomous outcome parameters.

1020 Anesthesia Choice and Elective THR ANESTHESIA & ANALGESIA

 Figure 2. Comparison of intraoperative
blood loss volume for elective total hip
replacement under neuraxial block versus
under general anesthesia (GA). N ⫽ num-
ber of patients.

Standardized mean difference (SMD) and 95% CI and narcotic consumption that are not included in our
were presented for continuous outcome parameters. analysis (12). Two papers of Borghi et al. published in
Heterogeneity among studies was tested by ␹2 test. 2002 (13) and 2005 (15) reported findings from the same
The results for both the fixed effects model and the groups of patients. Although data of different outcome
random effects model were presented. The fixed effects variables were reported in these two papers and are
model assumes that all studies are from a common included in our analysis, we considered that these two
population and that the effect size is not significantly papers reported results from one study. Davis et al.
different among different trials. However, when there reported findings from a total of 140 patients in 1989 (7),
was significant heterogeneity among the studies (P ⬍ of which findings from the first 101 patients were
0.05), we read the original studies again to identify published in 1987 (9). However, the authors did not
possible differences in study design (inclusion criteria report the intraoperative blood loss from the 140 patients
and exclusion criteria) and in the patient characteristics in the latter study (7). Instead, this result was presented
(mean age and comorbidities) among the trials to deter- in the thesis submitted by F. Michael Davis for his MD
mine whether we could separate trials into homoge- degree (16). These three publications are considered as
neous groups. If this attempt failed to identify the cause reports for one study. Thus, only 10 independent studies
of the heterogeneity, results calculated by using the had the relevant data for our analysis. These 10 studies
random effects model are more appropriate because this had a total of 330 patients undergoing GA and 348
model incorporated both the random variation within patients undergoing neuraxial block. Characteristics of
the studies and the variation among the different studies. these trials are displayed in Table 1. Among them, no
study presented data on mortality.
RESULTS
Our search identified 144 publications. Among them, Operative Times
studies in 14 publications met the inclusion criteria. One Eight studies reported this outcome. Six of them
paper reported outcome measures such as pain scores showed no statistical difference in operative times

Vol. 103, No. 4, October 2006 © 2006 International Anesthesia Research Society 1021
Figure 3. Comparison of number of patients
with blood transfusion for elective total hip
replacement under neuraxial block versus
under general anesthesia (GA). n ⫽ number
of patients requiring blood transfusion;
N ⫽ total number of patients in the study
group.

between neuraxial block and GA. Two studies showed three studies that reported number of patients trans-
that the operative times of THR under neuraxial block fused, one showed that neuraxial block significantly
were shorter than those under GA (7,14). The pooled reduced the number of patients requiring blood trans-
data from the eight studies showed a statistically fusion (6). The pooled data from these three studies
significant decrease in operative time (Fig. 1). The demonstrated that fewer patients were transfused
THR procedure under neuraxial block was finished when THR was performed under neuraxial block
7.1 min (95% CI 2.3–11.9 min) sooner than the proce- (21/177 ⫽ 12% patients) than that under GA
dure performed under GA. (62/188 ⫽ 33%, P ⬍ 0.001 by z-test) (Fig. 3, OR 0.26).
However, the CIs were wide and compatible with
Intraoperative Blood Loss Volume both no effect and a nine-tenths reduction (95% CI
Eight studies reported intraoperative blood loss, 0.06 –1.05).
and six of them showed that neuraxial block signifi-
Deep Venous Thrombosis
cantly decreased blood loss compared with GA
Five studies included data on the number of pa-
(3– 6,9,16). The pooled data from the eight studies
tients who developed radiographically proven DVT.
showed a statistically significant decrease in blood loss
All of them showed that neuraxial block significantly
in patients under neuraxial block versus GA (Fig. 2,
decreased the incidence of DVT compared with GA
mean difference 275 mL/case and 95% CI 180 –371 mL).
(3–7). The pooled data showed that significantly fewer
patients developed DVT when the THR was per-
Number of Patients Requiring Blood Transfusions formed under neuraxial block (58/200 ⫽ 29% patients)
Six studies reported data on number of patients than under GA (116/209 ⫽ 56% patients) (Fig. 4, OR
transfused and/or the blood transfusion volume. 0.27, 95% CI 0.17– 0.42).
Among the four studies that reported blood transfu-
sion volume, two reported the volume in numerical Pulmonary Embolism
format. Meta-analysis was not performed with data Five studies presented data on the number of
from these two studies because of the concern for too patients who suffered from a PE evidenced by radio-
few studies. One of the studies noted that neuraxial graphic or nuclear medicine studies. Three of these
block reduced blood transfusion volume per trans- studies showed that neuraxial block significantly de-
fused patient when compared with GA (3). Among the creased the incidence of PE compared with GA (3–5).
1022 Anesthesia Choice and Elective THR ANESTHESIA & ANALGESIA
 Figure 4. Comparison of number of patients
with deep venous thrombosis for elective
total hip replacement under neuraxial block
versus under general anesthesia (GA). n ⫽
number of patients with deep venous
thrombosis; N ⫽ total number of patients in
the study group.

The other two studies did not show a significant funnel plots (plots are not shown) of sample size
difference in the number of patients who suffered versus OR or sample size versus smd for intraopera-
from PE after THR under neuraxial block versus GA. tive blood loss, operative time, number of patients
The pooled data showed that significantly fewer pa- transfused, and the incidence of DVT and PE did not
tients had PE when the THR was performed under show evidence for significant publication bias. It
neuraxial block (14/191 ⫽ 7% patients) than under GA should be noted that funnel plots derived from a small
(38/193 ⫽ 20% patients) (Fig. 5, OR 0.26, 95% CI number of studies may not be a sensitive tool to detect
0.12– 0.56). publication bias. It is also possible that our study
suffers from “informed censoring.” This refers to a
DISCUSSION situation in which the authors of original studies
Our meta-analysis showed statistically significant collected data on all our selected outcome variables
reductions in the operative time, intraoperative blood but failed to report on results that were not different
loss, and the incidence of DVT and PE when neuraxial between the groups or were not interesting to the
blockade was used in a specific patient population: authors. We could not use these data in our analysis.
patients undergoing elective THR. Among the 10 As a result, the estimated differences between patient
independent studies that contributed data to our groups by meta-analysis are likely to be more than the
analysis, three studies compared the outcomes be- actual differences. To reduce this possibility, we at-
tween spinal anesthesia and GA (6,7,18), and the tempted to contact the authors of all trials included in
others compared outcomes between epidural anesthe- this analysis to forward any outcome data they had on
sia and GA. In our analysis, we did not separate the record that were not reported in their original papers.
neuraxial block into spinal and epidural block sub- There may also have been selection bias. We included
groups because of the concern of small sample size for all identified studies that were prospective, random-
each subgroup. ized, or quasirandomized trials comparing neuraxial
Our analysis may have limitations. All the data block versus GA for elective THR. Thus, selection bias
included in our analysis are from published studies, in our analysis may be small. Lastly, our analysis is
which may have produced biased results. However, hindered by the datedness of the studies contributing

Vol. 103, No. 4, October 2006 © 2006 International Anesthesia Research Society 1023
Figure 5. Comparison of number of patients
with pulmonary embolism for elective total
hip replacement under neuraxial block ver-
sus under general anesthesia (GA). n ⫽
number of patients with pulmonary embo-
lism; N ⫽ total number of patients in the
study group.

to the analysis. Some aspects of these studies do not Intraoperative Blood Loss and Incidence of
reflect current practice patterns. For example, pharma- Blood Transfusion
cologic prophylaxis for DVT is currently used for The potential for decreasing intraoperative blood
patients after THR. Most of the patients in our analysis loss is an often quoted advantage for performing THR
did not receive this therapy. This issue will be dis- under neuraxial anesthesia. In this meta-analysis, we
cussed further in Thromboembolic Events. showed a statistically significant decrease in blood
loss in the neuraxial block group. Although the mean
decrease was only 275 mL, this amount may be
Operative Times clinically significant, as neuraxial blockade also de-
creased the number of patients requiring intraopera-
Concerns over the use of neuraxial block include a tive blood transfusion (12% patients under neuraxial
potentially delayed start time of surgery due to the blockade versus 33% patients under GA).
placement of the block, failure of the block with
subsequent conversion to GA, and potentially less
Thromboembolic Events (DVT and PE)
than optimal muscle relaxation, which some orthope-
PE remains a potentially catastrophic complication
dic surgeons believe will make the dissection and
of THR with a reported incidence of clinical PE in
placement of the prosthesis more difficult. Our data 0.2%–2.0% of patients (20). The incidence of DVT is
indicate a small reduction in the operative time for around 1%–10% now, but was as high as 40%– 60% in
elective THR using neuraxial block when compared some series where DVT prophylaxis was not used
with GA. Our data are consistent with a recent Co- (21). This meta-analysis shows a significant reduction in
chrane Report on hip fracture patients by Parker et al. the number of patients developing DVT (29% vs 56%)
(19) in which anesthesia choice had a minimal effect and PE (7% vs 20%) when neuraxial anesthesia is used
on operative times. Although we were able to show a for THR. The authors in these series actively searched for
statistically significant decrease in operative times PE and DVT using the combinations of phlebography,
when THR was performed under neuraxial blockade, plethysmography, venography, ventilation/perfusions
the average decrease in duration of 7.1 min/case is scans, and fibrinogen uptake tests. Our finding that
likely not clinically significant. neuraxial block decreases the incidence of DVT and PE is

1024 Anesthesia Choice and Elective THR ANESTHESIA & ANALGESIA

consistent with the data published by Rodgers et al. (2) 2. Rodgers A, Walker N, Schug S, et al. Reduction of postoperative
mortality and morbidity with epidural or spinal anaesthesia:
involving nearly 10,000 patients and showed that results from overview of randomised trials. BMJ 2000;321:1493–7.
neuraxial blocks for a variety of surgeries decreased 3. Modig J, Hjelmstedt A, Sahlstedt B, Maripuu E. Comparative
DVT by 44% and PE by 55%. influences of epidural and general anaesthesia on deep venous
thrombosis and pulmonary embolism after total hip replace-
Regardless of the causes for the decreased inci- ment. Acta Chir Scand 1981;147:125–30.
dences of DVT and PE by neuraxial block, the impli- 4. Modig J, Borg T, Karlstrom G, Maripuu E, Sahlstedt B. Throm-
cation of our findings must be viewed cautiously. boembolism after total hip replacement: role of epidural and
Only one of the studies reviewed here used pharma- general anesthesia. Anesth Analg 1983;62:174 – 80.
5. Modig J, Maripuu E, Sahlstedt B. Thromboembolism following
cologic DVT prophylaxis, and interestingly, that study total hip replacement: a prospective investigation of 94 patients
showed no significant difference in the rate of DVT with emphasis on the efficacy of lumbar epidural anesthesia in
between the two groups (14). For the last 15 yr, prophylaxis. Reg Anesth 1986;11:72–9.
6. Thorburn J, Louden JR, Vallance R. Spinal and general anaes-
pharmacologic DVT prophylaxis has been a compo- thesia in total hip replacement: frequency of deep vein throm-
nent of the standard of care for THR patients. Al- bosis. Br J Anaesth 1980;52:1117–21.
though neuraxial blockade apparently decreases the 7. Davis FM, Laurenson VG, Gillespie WJ, et al. Deep vein
thrombosis after total hip replacement. A comparison between
risk of DVT when no chemical prophylaxis is used, it spinal and general anaesthesia. J Bone Joint Surg Br
is not as effective as using low-molecular-weight 1989;71:181–5.
heparin (22). Four of the studies reviewed here used 8. Brinker MR, Reuben JD, Mull JR, et al. Comparison of general
and epidural anesthesia in patients undergoing primary unilat-
epidural catheters for postoperative analgesia (3–5,17). eral THR. Orthopedics 1997;20:109 –15.
The fact that most of these studies were performed 9. Davis FM, McDermott E, Hickton C, et al. Influence of spinal
before the use of DVT prophylaxis was the standard of and general anaesthesia on haemostasis during total hip arthro-
plasty. Br J Anaesth 1987;59:561–71.
care poses a significant limitation in the application of 10. Flordal PA, Neander G. Blood loss in total hip replacement. A
our findings to today’s practice. It remains to be seen retrospective study. Arch Orthop Trauma Surg 1991;111:34 – 8.
whether or not the effects of a single-injection 11. Keith I. Anaesthesia and blood loss in total hip replacement.
Anaesthesia 1977;32:444 –50.
neuraxial technique followed by postoperative phar- 12. Wulf H, Biscoping J, Beland B, et al.; Ropivacaine Hip Replace-
macologic anticoagulation would be additive in the ment Multicenter Study Group. Ropivacaine epidural anesthe-
prevention of DVT and PE. These studies are needed sia and analgesia versus general anesthesia and intravenous
before we can determine whether neuraxial block patient-controlled analgesia with morphine in the perioperative
management of hip replacement. Anesth Analg 1999;89:111–16.
indeed reduces the incidence of DVT and PE after 13. Borghi B, Casati A, Iuorio S, et al. Frequency of hypotension and
THR in our current practice. bradycardia during general anesthesia, epidural anesthesia, or
In summary, we analyzed the literature to deter- integrated epidural-general anesthesia for total hip replace-
ment. J Clin Anesth 2002;14:102– 6.
mine whether anesthesia choice will affect the out- 14. Hole A, Terjesen T, Breivik H. Epidural versus general anaes-
come of a specific surgical patient population: patients thesia for total hip arthroplasty in elderly patients. Acta Anaes-
undergoing elective THR. Our data indicate that thesiol Scand 1980;24:279 – 87.
15. Borghi B, Casati A, Iuorio S, et al. Effect of different anesthesia
neuraxial block is associated with a decrease in intra- techniques on red blood cell endogenous recovery in hip
operative blood loss and the number of patients arthroplasty. J Clin Anesth 2005;17:96 –101.
requiring blood transfusions. It is not known whether 16. Davis FM. Anesthesia for hip surgery. Comparative studies of
spinal anaesthesia and general anaesthesia for elective total hip
some of the beneficial effects such as reduced inci- arthroplasty and internal fixation of fractures of the femoral
dence of DVT and PE provided by neuraxial block are neck in the elderly. [M.D. Thesis]. Dunedin, New Zealand:
applicable to today’s practice when compared with University of Otago, 1987.
17. Modig J, Karlstrom G. Intra- and post-operative blood loss and
investigations performed 20 years ago. However, our haemodynamics in total hip replacement when performed
findings indicate that neuraxial block should be con- under lumbar epidural versus general anaesthesia. Eur J Anaes-
sidered as a valid and potentially beneficial technique thesiol 1987;4:345–55.
18. Brueckner S, Reinke U, Roth-Isigkeit A, et al. Comparison of
for elective THR. Our analysis also points out the need general and spinal anesthesia and their influence on hemostatic
for further studies designed to investigate the effects markers in patients undergoing total hip arthroplasty. J Clin
of anesthetic choice on outcomes for THR in the Anesth 2003;15:433– 40.
context of current clinical practice, as many improve- 19. Parker M, Hondoll H, Griffiths R. Anaesthesia for hip fracture
surgery in adults. Cochrane Libr 2005;2005:1– 62.
ments in surgical and anesthetic techniques and post- 20. Phillips CB, Barrett JA, Losina E, et al. Incidence rates of
operative care have evolved since these early studies dislocation, pulmonary embolism, and deep infection during
were performed. the first six months after elective total hip replacement. J Bone
Joint Surg Am 2003;85:20 – 6.
21. Lieberman JR, Geerts WH. Prevention of venous thromboem-
REFERENCES
bolism after total hip and knee arthroplasty. J Bone Joint Surg
1. Kozak L, Owings M, Hall M. National hospital discharge survey: Am 1994;76:1239 –50.
2002 annual summary with detailed diagnosis and procedure data. 22. Geerts WH, Pineo GF, Heit JA, et al. Prevention of venous
Hyattsville: US Department of Health and Human Services, 2005. thromboembolism: the seventh ACCP conference on antithrom-
Vital and health statistics; Series 13, No. 158. botic and thrombolytic therapy. Chest 2004;126:338S– 400S.

Vol. 103, No. 4, October 2006 © 2006 International Anesthesia Research Society 1025