Chapter  22:  Prevention of Venous Thromboembolism After Injury: Evidence Report/Technology Assessment Number 22

Objectives

This project's goals are to evaluate the existing literature, summarize the evidence, and perform meta-analysis and cost-effectiveness analysis on data relevant to prevention of venous thromboembolism after injury. Venous thromboembolism occurs frequently after trauma and causes significant mortality and long-term disability. At the same time, methods to prevent and diagnose it are highly controversial and physicians' practices vary widely. With this evidence report, we intend to examine these controversial areas by analyzing the existing scientific literature. An equally important objective is to identify areas in which evidence is lacking in order to direct future research.

Search Strategy

Three databases were searched: MEDLINE (1966--99), EMBASE (1980--99), and the Cochrane Controlled Trials Register (1980--99). The following medical subject headings were used: Thrombophlebitis, Thrombosis, Thromboembolism, Pulmonary embolism, Wounds and injuries; the subheadings: pc (prevention and control), in (injuries); and the text words: prevent$, thromboprophyla$, prophylac$, trauma$, posttrauma$, post-trauma$.

Selection Criteria

Studies were selected if they specifically reported on methods of venous thromboembolism prevention and screening in trauma patients. Studies including only nontrauma patients were rejected. A panel of technical experts assisted in identifying four key questions:

• 1

  What is the best method of venous thromboembolism prophylaxis?

• 2

  What groups of patients are at high risk of developing venous thromboembolism?

• 3

  What is the best method of screening for venous thromboembolism?

• 4

  What is the role of vena cava filters in preventing pulmonary embolism?

Studies were selected if they addressed any of these four questions.

Data Collection and Analysis

Screening of 4,093 relevant titles by two independent reviewers resulted in acceptance of 2,437 of them for abstract review; 227 of these were accepted for further review. Finally, 73 studies were analyzed. Meta-analysis and supplemental analyses were performed on the available data.

Main Results

The reported incidence of deep venous thrombosis in trauma patients in the selected studies is 12 percent and varies from 3 percent to 23 percent according to study design, type of trauma population, and method of deep venous thrombosis prophylaxis and diagnosis. The reported incidence of pulmonary embolism in these studies is 1.5 percent and varies from 0.1 percent to 15 percent. Few randomized controlled trials provided data that could be combined for meta-analysis. From the limited data available, there is no evidence that mechanical prophylaxis or low-dose heparin is superior to no prophylaxis or to each other for prevention of deep venous thrombosis. The role of low-molecular-weight heparin in trauma patients is unclear because the few relevant studies are heterogeneous. Spinal fractures and spinal-cord injuries increase the risk of venous thromboembolism. No relevant data are available for drawing conclusions about the best method of screening for venous thromboembolism. Although vena cava filter placement in selected trauma patients may decrease the incidence of pulmonary embolism and fatal pulmonary embolism, the designs of the studies reporting these results do not allow definitive conclusions to be drawn.

Conclusions

The evidence on prevention of venous thromboembolism after injury is scanty. Many practices are based on extrapolations from data on nontrauma patients. The risk of venous thromboembolism increases in the presence of spinal trauma with or without injury to the spinal cord. Currently, the most frequently used methods of venous thromboembolism prophylaxis do not offer a proven benefit over no prophylaxis. There is a pressing need for well-designed studies that will identify the best method of prevention of venous thromboembolism in trauma patients.

This document is in the public domain and may be used and reprinted without permission except those copyrighted materials noted for which further reproduction is prohibited without the specific permission of copyright holders.

Suggested Citation:
Velmahos GC, Kern J, Chan L, et al. Prevention of Venous Thromboembolism After Injury. Evidence Report/ Technology Assessment No. 22. (Prepared by Southern California Evidence-based Practice Center/RAND under Contract No. 290-97-0001.) AHRQ Publication No. 01-E004. Rockville, MD: Agency for Healthcare Research and Quality. November 2000.

Overview

Venous thromboembolism (VT) is major national health problem, claiming 50,000 lives and resulting in 300,000 to 600,000 hospitalizations annually in the United States. VT presents in two forms: deep venous thrombosis (DVT) and pulmonary embolism (PE). Injured patients are at high risk for VT because of changes in coagulation and thrombolysis mechanisms that are induced by trauma.

Methods of prevention of VT include, among others, sequential compression devices (SCDs), low-dose heparin (LDH), low-molecular-weight heparin (LMWH), vena cava filters (VCFs), and combinations of these. All these methods are associated with contraindications and morbidity. Therefore, selecting the appropriate method for the appropriate trauma patient is important. The difficulty of selecting the appropriate prophylaxis is in part a result of the inconclusiveness of the relevant trauma literature. This allows wide variability among physicians' practices and prevents consistency in quality of care.

With this report, we evaluate and meta-analyze the existing data in the literature to produce scientific answers in controversial areas related to this topic. We also identify research gaps in areas in which the scientific evidence is absent or minimal, and we hope to assist interested organizations in producing relevant guidelines and in directing future research.

Reporting the Evidence

A panel of 17 technical experts, consisting of national authorities in the field and representing the academic, private, and managed care sectors, was formed to assist in the design and execution of the project. Important questions on the topic were distributed to the experts, who ranked them in order of importance. After two conference calls, four refined key questions were developed:

• 1

  What is the best method of VT prophylaxis?

• 2

  What groups of patients are at high risk of developing VT?

• 3

  What is the best method of screening for VT?

• 4

  What is the role of VCFs in preventing PE?

The panel decided to use data restricted to trauma patients only and to avoid extrapolations of conclusions from nontrauma patients to the trauma population. Defining "the trauma patient" was difficult. The panel decided to exclude elderly patients with injuries following low-energy trauma (such as hip fractures after ground-level falls) from consideration. We subsequently developed causal pathways for each key question. We felt it was important to report on the rates of DVT and PE from combined literature data because these rates varied widely among studies.

We summarized the existing evidence on all trauma patients included in the available literature as well as that on individual trauma patient groups (orthopedic trauma, neurosurgical trauma, minor trauma) when data were available. We evaluated the quality of studies included in our analysis using previously published methods of determining quality scores. We entered all data in a computerized database specifically designed for this project.

Methodology

We searched three literature databases: MEDLINE (1966--January 31, 1999), EMBASE (1980--January 31, 1999), and the Cochrane Controlled Trials Register (1980--January 1999). After a broad initial search, we performed multiple literature searches tailored to each question. Finally, we identified a total of 4,093 titles, which were screened according to specific inclusion and exclusion criteria by two independent medical reviewers. A third reviewer assisted in case of disagreements. After screening, 2,437 titles were accepted for abstract review. All three reviewers screened all abstracts against specific criteria; 227 of these were accepted for complete review. Of 225 articles retrieved, 73 were accepted for meta-analysis. We designed forms to extract relevant data on study design and quality, methods used, risk factors, and outcomes. Two reviewers extracted data, which were re-examined by a third reviewer. Discrepancies were resolved in meetings among all three reviewers. A random-effects model was used for all pooled results.

We first evaluated the reported incidence of DVT and PE in trauma patients. We extracted these rates from all studies as well as from studies grouped together by study design randomized, nonrandomized comparative cohorts, single cohort), method of VT diagnosis (routine screening or based on clinical suspicion), use of VT prophylaxis (yes or no), and type of trauma patients (all trauma, orthopedic trauma, neurosurgical trauma, minor trauma).

We addressed the question of the best method of VT prophylaxis in three ways:

• We examined the incidence of DVT and PE after combining groups of patients from different randomized trials who received LDH or LMWH or mechanical prophylaxis (MP) or no prophylaxis.

• We performed a meta-analysis of randomized controlled trials (RCTs) evaluating the same methods of prophylaxis.

• We performed a meta-analysis of RCTs and non-RCTs evaluating the same methods of prophylaxis.

This last meta-analysis, although methodologically weak, was performed, because the number of RCTs available for the first meta-analysis was limited.

We addressed the question of risk factors for developing VT by performing meta-analysis on studies (RCT and non-RCT) that used risk factors as either dichotomous variables (e.g., age greater or lower than 55) or continuous variables (e.g., age, without specifying a particular age cutoff point). We evaluated six dichotomous risk factors (gender, head injury, long-bone fracture, pelvic fracture, spinal fracture, and spinal-cord injury) and three continuous risk factors (age, Injury Severity Score [ISS], and units of blood transfused).

We were unable to address the question about methods of screening for VT using the current literature data. Only three studies addressed this issue in trauma patients, and each compared different methods of screening. The data could not be combined for analysis.

We addressed the question about VCFs by combining studies that included patients treated with VCF and patients without VCF and estimating the rates of PE in the two groups. None of these studies was an RCT. Other outcome parameters relevant to VCF placement, such as related complications, long-term outcome, or appropriate population to be treated with this modality, could not be extracted from the limited data available.

We also performed supplemental analyses on the two most frequent complications related to prophylactic heparin administration-bleeding and thrombocytopenia-as well as on the incidence of fatal PE and the length of hospital stay in patients who develop VT. Finally, we developed a cost-effectiveness model.

Results

• The reported incidence of DVT and PE varies widely among different studies depending on study design, type of trauma patients included, and methods of screening and prophylaxis. The pooled rates of DVT and PE across all studies are 11.8 percent (95 percent confidence interval [CI]: 0.104, 0.131) and 1.5 percent (95 percent CI: 0.011, 0.018), respectively.

• Only a few RCTs address methods of VT prophylaxis in trauma patients. Most of these studies use different methods. Combining the limited data from studies using the same methods produces small sample sizes.

• LDH is not statistically superior to no prophylaxis in preventing VT after injury (odds ratio [OR]: 0.965, 95 percent CI: 0.353, 2.636). This conclusion is based on meta-analysis of four RCTs with a total of 219 patients.

• MP is not statistically superior to no prophylaxis in preventing VT after injury (OR: 0.769, 95 percent CI: 0.265, 2.236). This conclusion is based on meta-analysis of three RCTs with a total of 234 patients.

• The addition of non-RCTs to the meta-analyses of studies examining DVT rates in trauma patients receiving LDH vs. no prophylaxis or MP vs. no prophylaxis does not change the above conclusions.

• Comparison of LMWH vs. LDH shows no statistically significant difference between the two methods in preventing PE (OR: 3.010, 95 percent CI: 0.585, 15.485). This conclusion is based on meta-analysis of three studies reporting on the incidence of PE (two RCTs and one non-RCT, total number of patients: 355). Although the difference in PE rates is not statistically significant, the limits of the 95 percent confidence interval for this result are very wide.

• Three RCTs (one of them with the highest possible quality score) showed separately statistical superiority of LMWH against LDH or SCD in preventing DVT. The reported DVT rates vary widely among these studies (38 percent, 7 percent, and 2 percent). Because the method of prophylaxis used to compare against LMWH was not the same, meta-analysis was not done.

• Comparison of LDH vs. MP after meta-analysis of seven studies (four RCTs and three non-RCTs, total number of patients: 620) shows no statistically significant difference between the two methods in preventing DVT (OR: 1.161, 95 percent CI: 0.495, 2.723).

• Spinal fractures and spinal-cord injury are risk factors for DVT. Other frequently reported risk factors, such as head injury, pelvic fractures, or long-bone fractures, were not shown in the meta-analysis to increase the risk for DVT. It is possible that the studies reporting on these factors included severely injured patients who were already at high risk regardless of the presence of the individual risk factor.

• Trauma patients who develop DVT are older and have more severe injuries than patients who do not develop DVT. However, a specific age or ISS threshold could not be extracted from the available data.

• The reported incidence of PE in patients who undergo VCF placement is 0.2 percent, which is lower than the incidence observed in concurrently managed patients without VCF (1.5 percent) and historical controls without VCF (5.8 percent). The observational design of these studies does not allow firm conclusions to be drawn.

• LDH or LMWH administration for VT prophylaxis produces a low and similar incidence of adverse events: on average, 3 percent for bleeding and 1 percent for thrombocytopenia. These low rates may occur because patients at high risk for bleeding were not given heparin.

• Fatal PE has been reported in one-third of trauma patients who develop PE, based on data from 16 studies that reported on both rates (PE and fatal PE).

• The length of hospital stay in patients who develop DVT is significantly longer (by 15 days) relative to patients without DVT. Although a cause-effect relationship between DVT and length of hospital stay cannot be established, DVT is associated with increased costs and use of health care resources.

• There are significant gaps in the literature regarding the prevention of VT after trauma.

Future Research

Future research should be directed to two areas: identifying the appropriate groups of trauma patients in need of VT prophylaxis and evaluating different methods of prophylaxis with regard to their safety and efficacy in trauma patients. Although evaluating different methods of screening for DVT would be useful, we do not feel that this should be a priority for future research. Duplex ultrasonography is the most convenient, noninvasive, and inexpensive method of screening severely injured patients. Even if other methods of screening prove to be more sensitive, associated technical and logistical difficulties make them impractical.

To address the two above areas, we propose a large multicenter trial. This trial should have a randomized controlled design, compare the most commonly used methods of prophylaxis (LDH, LMWH, SCD), identify DVT by routine screening, and evaluate multiple risk factors. Based on the findings of this evidence report, a no-prophylaxis group should be included.

Equally important future research should be directed towards evaluating the role of VCF in trauma patients. This question could be incorporated in the multicenter trial proposed above or become the sole objective of a separate randomized trial. Both designs should have a predetermined protocol for diagnosing PE, an aggressive autopsy policy to identify the cause-effect relationship of PE to death, and careful, long-term followup to detect VCF-related complications.

Purpose

The purpose of this evidence report is to review the scientific literature to establish the incidence of venous thromboembolism (VT) after injury, evaluate the role of different methods of prophylaxis, and identify trauma patients who are at high risk for developing VT. This information will be made available to health care providers and organizations to assist them in treating these patients. Professional trauma societies may develop guidelines based on the existing evidence and use the guidelines to improve the care of patients with VT.

Scope of Work

Initially, we found general information on all aspects of VT after injury, including the following:

• Incidence of VT in different trauma populations.

• Prevention and treatment of VT.

• Groups at risk.

• Methods of prevention and screening.

• Short- and long-term effects of VT.

• Adverse reactions to methods of prevention.

• Diagnostic tests.

• Cost-effectiveness of different methods of management.

Discussions with technical experts identified four key questions:

• 1

  What is the best method of VT prophylaxis?

• 2

  What groups of patients are at high risk of developing VT?

• 3

  What is the best method of screening for VT?

• 4

  What is the role of vena cava filters (VCFs) in preventing pulmonary embolism (PE)?

Background and Description of the Problem

Scope of Work

Shortly after project assignment to the Southern California Evidence-based Practice Center, and following two preliminary meetings (October 1 and 5, 1998) among the core project staff, the University of Southern California (USC)/RAND team met (October 9, 1998) to define the scope of work, discuss pertinent issues, form a general plan, and solve problems. Project staff outlined the scope of work, which included the following:

• Identification of technical experts to be primary advisors throughout the project.

• Selection of important questions relevant to the topic.

• Design of an exhaustive literature search for English and foreign-language articles, as well as abstracts or possibly unpublished work.

• Extraction of data from selected articles as the search progressed to increasingly relevant materials, and continuous data entry into a special computer database.

• Synthesis of evidence and supplemental analysis, if needed.

• Development and dissemination of an evidence-based report.

Design of Technical Experts Panel

Table 1. Technical experts panel

Table 1. Technical experts panel

Thomas V. Berne, MD	Professor of Surgery and Vice-Chairman University of Southern California, Los Angeles, CA AAST
Edward E. Cornwell III, MD	Assoc. Professor of Surgery and Director of Trauma Service Johns Hopkins University, Baltimore, MD AAST, EAST
Demetrios Demetriades, MD, PhD	Professor of Surgery and Director of Division of Trauma and Critical Care University of Southern California, Los Angeles, CA AAST
Richard Dorazio, MD	Chief of Vascular Surgery Kaiser Permanente Los Angeles, Los Angeles, CA
Timothy Fabian, MD	Professor of Surgery and Chief of Trauma Service University of Tennessee Center for the Health Sciences, Memphis, TN AAST, EAST (President)
Lazar Greenfield, MD	Professor and Chairman of Surgery University of Michigan, Ann Arbor, MI AAST
M. Margaret Knudson, MD	Associate Professor of Surgery University of California, San Francisco, CA AAST, WTA
Kenneth Mattox, MD	Professor of Surgery and Vice-Chairman, Chief of Trauma Surgery Baylor College of Medicine, Houston, TX AAST
William McGehee, MD	Associate Professor of Medicine, Department of Hematology Chief, Anticoagulation Division University of Southern California, Los Angeles, CA
Michael Pasquale, MD	Assistant Professor of Surgery Pennsylvania State University, Allentown, PA AAST; Chairman, Ad Hoc Committee on Practice Management Guideline Development, EAST
J. David Richardson, MD	Professor of Surgery and Vice-Chairman University of Louisville, Louisville, KY AAST (President)
Frederick Rogers, MD	Associate Professor of Surgery University of Vermont College of Medicine, Burlington, VT AAST; EAST: Chairman, Task Force on Deep Vein Thrombosis (1996)
C. William Schwab, MD	Professor of Surgery and Chief, Division of Traumatology and Surgical Critical Care University of Pennsylvania, Philadelphia, PA AAST, EAST
Steven R. Shackford, MD	Professor and Chairman of Surgery University of Vermont, Burlington, VT AAST, EAST, WTA (Vice President)
Kenneth Waxman, MD	Trauma Director Santa Barbara Cottage Hospital, Santa Barbara, CA AAST
Albert Yellin, MD	Professor of Surgery USC School of Medicine, Los Angeles, CA AAST, American Venous Forum, Society for Vascular Surgery

AAST, American Association for the Surgery of Trauma; EAST, Eastern Association for the Surgery of Trauma; WTA, Western Trauma Association; USC, University of Southern California

Table 2. Peer reviewers

Table 2. Peer reviewers

I. Elaine Allen, Ph.D.	Associate Professor of Statistics Babson College, Wellesley, MA
Howard Belzberg, MD	Assistant Professor of Surgery USC School of Medicine, Los Angeles, CA
H. Gill Cryer, MD	Professor of Surgery University of California, Los Angeles, CA
Donald Gaspard, MD	Director of Trauma Services Huntington Memorial Hospital, Pasadena, CA
Prof. William J. Gillespie	Dean, Dunedin School of Medicine University of Otago, Dunedin, New Zealand
Lazar Greenfield, MD	Professor and Chairman of Surgery University of Michigan, Ann Arbor, MI
David Hoyt, MD	Professor and Vice Chairman of Surgery University of California, San Diego, CA
John T. Owings, MD	Assistant Professor of Surgery University of California, Davis, Sacramento, CA
Basil A. Pruitt, Jr., MD	Professor of Surgery University of Texas Health Sciences Center, San Antonio, TX
William C. Shoemaker, MD	Professor of Surgery and Anesthesiology USC School of Medicine, Los Angeles, CA
All technical experts

USC, University of Southern California

After a preliminary phone call, formal letters of invitation to participate in the panel were sent to all the technical experts. All accepted the invitation, and a conference call was scheduled to outline the scope, goals, and methodology of the project. A preparatory face-to-face meeting was arranged with several attendees at the Annual Congress of the American College of Surgeons in Orlando, Florida. Most of those who did not attend this meeting were briefed in advance by telephone and in person.

We did not find any patient advocacy groups because of the nature of the population studied. Finding representative patients for the panel would be extremely difficult, because followup on trauma patients is notoriously unreliable once they leave the hospital. Identifying patients who could represent the entire trauma population that suffered from VT was not deemed feasible or essential.

Design of the Study

The time and resources of the project dictated that we could address three to five questions. We decided that a preliminary literature search would sample the existing evidence and suggest important questions that the evidence could answer. After defining these questions, we would submit them to the technical experts for ranking in order of importance. From the rankings we would select the three to five questions to be used for this project.

We also discussed the design of the database and the literature search and the allocation of responsibilities to the various members of the team. We decided to use separate quality screens to evaluate randomized and nonrandomized trials. A project member designed a novel, user-friendly computerized database that could easily be customized to different project needs. The foregoing issues were also discussed in a subsequent conference call (October 13) among the Task Order Officer (TOO), Center Director, and Task Order Manager. The TOO approved the plan of action developed during the meeting of the USC/RAND group.

Process of Selecting Variables to Be Studied

The staff held its first conference call to establish the role of the technical expert panel, discuss the methodology to be followed, identify the most important questions, and define clinically relevant outcomes. After the Task Order Manager and the Center Director gave a short history and background of the project, we discussed the following issues: defining "trauma patients" and selecting the questions to be asked, the studies to be used, and the relevant outcomes.

Selecting the Questions

Table 3. Initial set of questions sent to technical experts (more...)

Table 3. Initial set of questions sent to technical experts

Are all injured patients at the same risk for developing venous thromboembolism? How can subgroups at high risk be identified? Do deep venous thrombotic events at various sites pose similar risks for pulmonary embolism (above the knee vs. below the knee, upper extremities vs. lower extremities)? Is surveillance for deep venous thrombosis justified? If yes, in which subgroups of trauma patients? Which is the best (safest, most accurate, most cost-effective) method of screening for deep venous thrombosis? Which is the best (safest, most accurate, most cost-effective) method for prevention of venous thromboembolism? Are combination methods better than isolated methods? Which injured patients should receive an inferior vena cava filter prophylactically? What are the contraindications for the use of prophylactic doses of standard or low-molecular-weight heparin?

Table 4. Outcome of ranking of initial seven questions (more...)

Table 4. Outcome of ranking of initial seven questions

Expert	Question A	Question B	Question C	Question D	Question E	Question F	Question G
A	2		4	3	5	1
B	4	2		3	5		1
C	4		2		5	3	1
D	1	2	3	4	5
E	5	2	1		4	3
F	4	2		1	5	3
G			3	4	1	2	5
H	1	5		4	2		3
I	1		3	4	5		2
J	1	2	3	4		5
K	2			1	5	4	3
L	5	3	2	1	4
M	1		3		5	2	4
N	5	3	2	1	4
O	1		2		5	3	4
P	5		3	4	1		2
Q	4			3	5	2	1
TOTAL	46	21	31	37	66	28	26

Note: The technical experts were asked to rank the five most important questions with a score of 5 (most important) to 1 (least important).

Following the preliminary literature review, the core project staff developed seven possible questions to be studied (Table 3). These questions were circulated and approved by the entire USC/RAND team and then were faxed to the technical experts. We asked the experts to select the five most important questions and rank them in order of importance. Table 4 presents the outcome of the ranking process. The Kendall's tau coefficient of concordance was used to measure how closely the experts agreed on the ranking of questions (0.21). Friedman's test statistic was used to measure the significance of the agreement vs. no agreement (p=0.002), which indicates that there was satisfactory agreement among the experts on the relative importance of the different questions.

All the participants had already received a copy of the first ranking of the questions. In the first round, the experts agreed that the question, "hich is the best method for prevention of VT? Are combination methods better than isolated methods?" (see Table 3) was the most important one and that the question, "Are all injured patients at the same risk for developing VT? How can subgroups at high risk be identified?" (see Table 3) was the second most important one.

The preliminary literature search, however, had yielded very few well-designed studies that addressed these questions. For example, only four randomized controlled trials with trauma patients were initially found to address the most important question, a comparison of the different methods of VT prophylaxis. A proposal was offered to include studies from the orthopedic literature and extrapolate the results to the trauma population. The panel decided to reject this idea and study only trauma patients to ensure clinical homogeneity and avoid arbitrary extrapolations resulting from analysis of mixed populations.

The question on identifying the best methods of screening (see Table 3) was also debated. Some members of the panel thought that the question was obsolete. Even if venography proved to be more sensitive and specific than Duplex ultrasound, the latter would still be the preferred method of screening of critically injured patients for reasons of convenience and safety. However, other experts believed that evaluating studies comparing Duplex ultrasound with other methods would be important. Finally, the panel decided to investigate the literature on this question, which was ranked third in order of importance, and to decide whether to report on the evidence based on the amount and quality of the existing literature.

Refining the Questions: Defining Causal Pathways for the Selected Questions

Table 5. Four refined key questions for ranking with (more...)

Table 5. Four refined key questions for ranking with issues involved

1 What is the role of different chemical or mechanical methods in preventing venous thromboembolism? Which is the best (most efficient, safe and cost-effective) method to prevent venous thromboembolism? Are combination methods better than isolated methods? What are the contraindications to using each method, and what are the best alternatives when contraindications exist? 2 What are the factors placing trauma patients at high risk for venous thromboembolism? Are all injured patients at the same risk for venous thromboembolism? What is an acceptably low rate of venous thromboembolism to recommend that there is no need for screening and thromboprophylaxis? How can subgroups at high risk for venous thromboembolism be identified? 3 Which is the optimal method to screen for deep venous thrombosis? Which is the best (safest, most accurate, most cost-effective) method for deep venous thrombosis surveillance? Which groups should be screened? Should different groups be screened by different methods? How soon, how often, and for how long should screening be done? 4 What is the role of vena cava filters in preventing pulmonary embolism? Which groups need a prophylactic vena cava filter? What is the risk for pulmonary embolism and death from deep venous thrombosis in these groups, and to what extent do vena cava filters decrease this risk? How safe are vena cava filters?

Table 6. Outcome of ranking of the four refined key (more...)

Table 6. Outcome of ranking of the four refined key questions

	Question 1	Question 2	Question 3	Question 4
Expert
A
B	4	3	2	1
C	4	3	1	2
D	4	3	1	2
E	4	3	1	2
F
G
H
I	4	1	3	2
J	3	4	2	1
K	4	2	3	1
L
M	3	4	1	2
N	4	3	2	0
O	3	4	1	2
P
Q	4	3	2	1
TOTAL	41	33	19	16

Notes: The technical experts were asked to rank the questions on a scale of 4 (most important) to 1 (least important). Blanks indicate no response received.

Question on Methods of Prevention (Table 5, Question 1)

Table 7. Causal pathway for question 1: Methods of (more...)

Table 7. Causal pathway for question 1: Methods of prevention of venous thromboembolism

All the participants agreed that LDH, low-molecular-weight heparin (LMWH), and SCDs are the most commonly used methods of prophylaxis and should be the focus of our literature search. Data on other methods, such as dextran, warfarin, or dihydroergotamine, should be also collected and analyzed if possible. The causal pathway for this question (Table 7) indicated that efficacy and safety would be the primary focus of evaluation. Cost-effectiveness of specific methods would also be examined.

Question on Identification of High-Risk Groups (Table 5, Question 2)

Table 8. Causal pathway for question 2: Groups at high (more...)

Table 8. Causal pathway for question 2: Groups at high risk for venous thromboembolism

DVT, deep venous thrombosis; MI, myocardial infarction; CHF, congestive heart failure; PEEP, positive end-expiratory pressure.

The technical experts agreed that there were no uniformly accepted criteria for identifying trauma patients at high risk for developing VT. Although some proposals for high-risk criteria have been published, they have not been extensively validated. The panel did not reach consensus on this question. It was suggested that we should investigate all the published criteria and analyze those most frequently used. The causal pathway developed for this question (Table 8) specified a variety of risk factors, which were grouped into the following categories: preexisting conditions, type of injuries, physiologic criteria, and iatrogenic factors. This grouping would be maintained during literature review and data synthesis only if possible.

Question on Methods of Screening (Table 5, Question 3)

Table 9. Causal pathway for question 3: Best method (more...)

Table 9. Causal pathway for question 3: Best method of screening

ICU, intensive care unit.

The panel agreed that the term "screening" pertains only to DVT. No routine screening is done for PE, which is diagnosed on the basis of clinical suspicion. The experts suggested that clinical examination should not be included as a method of screening because it does not reliably detect DVT in severely injured patients. Some experts recommended that impedance plethysmography should also be excluded because it is not widely used for trauma patients. Because everyone agreed that the two most important screening methods are Duplex scanning and venography, it was again proposed that a comparison of these two might be obsolete: Venography could be more sensitive in detecting DVT, but the Duplex scan is more convenient and safe for injured patients in the intensive care unit. However, the panel decided that the literature should be assessed because information on the timing, frequency, and cost-effectiveness of screening methods could be extracted. Other methods, such as impedance plethysmography or D-dimers, would be analyzed if adequate data were available. The major characteristics that would determine the superiority of a test (indicated by thick arrows in Table 9) were sensitivity/specificity, safety, applicability, and cost-effectiveness. Other characteristics (indicated by thin arrows in Table 9) were thought to be secondary, though important. Data on the timing of screening methods would also be extracted.

Question on the Role of Vena Cava Filters (Table 5, Question 4)

Table 10. Causal pathway for question 4: Role of vena (more...)

Table 10. Causal pathway for question 4: Role of vena cava filters for prevention of venous thromboembolism

1 Reasons for VCF insertion Contraindications to heparin Permanent paralysis Long-term immobilization Very high risk for PE Breakthrough PE (despite heparin administration) Recurrent PE 2 Efficacy Decrease in PE rate Decrease in rate of death from PE 3 Safety Incidence of DVT/postphlebitic syndrome Complications malplacement dislodgment and migration perforation of vital structures (heart, veins) breakthrough PE

VCF, vena cava filter; PE, pulmonary embolism; DVT, deep venous thrombosis.

The panel was informed that our preliminary literature review identified several studies addressing this question, but none were RCTs. Many experts were concerned because several devices are available, which might make study results difficult to compare. Other experts believed that the evidence was not sufficient and might require too many extrapolations to analyze its cost-effectiveness. However, a suggestion was made to synthesize the existing data, if only to determine that the evidence is not sufficient for solid conclusions. All but three technical experts agreed to retain this topic as one of the key questions, but it was given the lowest priority. The causal pathway developed for this question (see Table 10) lists possible reasons for VCF insertion and issues relating to its safety and efficacy. The experts agreed that these issues should be researched in the literature.

Literature Search

A librarian with special training in literature retrieval for evidence-based projects conducted the literature searches. Core project staff selected the terms used, with input from the entire project staff and the technical experts.

Strategy and Execution

Three major databases (MEDLINE, EMBASE, and the Cochrane Controlled Trials Register) were searched according to the following strategy:

• Prior to the conference calls, a preliminary search was performed on MEDLINE (English-language articles only) to estimate the amount and type of evidence available.

• Following the suggestions by the technical experts, the formal computer-aided literature search was performed on all three databases using OVID: MEDLINE was searched from 1966 through January 1999, EMBASE from 1980 through January 1999, and Cochrane from May 1991 through January 1999. All languages were included. Terms used included the following medical subject headings:
  -Thrombophlebitis
  -Thrombosis
  -Thromboembolism
  -Pulmonary embolism
  -Wounds and injuries
  the subheadings:
  -pc (prevention and control)
  -in (injuries)
  and the text words:
  -prevent$
  -thromboprophyla$
  -prophylac$
  -trauma$
  -posttrauma$
  -post-trauma$.

Table 11. Prevention of venous thromboembolism after (more...)

Table 11. Prevention of venous thromboembolism after injury: Literature search in MEDLINE (1966--January 31, 1999) using OVID

1 *thrombosis/ or *thrombophlebitis/ or *pulmonary embolism/ or *thromboembolism/ 2 (trauma$ or posttrauma$ or post-trauma$).mp. 3 in.fs. 4 exp "wounds and injuries"/ 5 2 or 3 or 4 6 1 and 5 7 thrombosis/pc or thrombophlebitis/pc or pulmonary embolism/pc or thromboembolism/pc 8 6 and 7 9 (prevent$ or prophyla$ or thromboprophyla$).mp. 10 6 and 9 11 8 or 10 12 limit 11 to animal 13 limit 11 to human 14 12 and 13 15 11 not 12 16 15 or 14 17 16 not letter.pt. 18 17 not editorial.pt. 19 18 not case report/ 20 case report/ and clinical trial.pt. 21 18 and 20 22 19 or 21

Table 12. Prevention of venous thromboembolism after (more...)

Table 12. Prevention of venous thromboembolism after injury: Literature search in EMBASE (1980--January 31, 1999) using OVID

1 *Deep vein thrombosis/ 2 *Leg thrombosis/ 3 *Thrombosis/ or *Thrombosis prevention/ or *Vein thrombosis/ 4 *Thromboembolism/ 5 *Lung embolism/ 6 (venous and thrombosis).ti. 7 1 or 2 or 3 or 4 or 5 or 6 8 *Thrombophlebitis/ 9 7 or 8 10 (trauma$3 or postrauma$3).mp. 11 exp Injury/ 12 (injur$3 or fracture$1).mp. 13 10 or 11 or 12 14 9 and 13 15 limit 14 to human 16 Case control study/ or Case study/ or Clinical article/ or Clinical study/ or Clinical trial/ or Major clinical study/ or Prospective study/ or Retrospective study/ 17 Cohort analysis/ 18 exp Clinical trial/ 19 Meta analysis/ 20 exp Practice guideline/ 21 Randomized controlled trial/ 22 16 or 17 or 18 or 19 or 20 or 21 23 15 and 22

Table 13. Prevention of venous thromboembolism after (more...)

Table 13. Prevention of venous thromboembolism after injury: Literature search in Cochrane controlled trials register (1980--January 31, 1999) using OVID

(VEIN next THROMBOSIS or DEEP next VEIN next THROMBOSIS or DVT or THROMBOPHLEBITIS or PULMONARY next EMBOLISM or LUNG next EMBOLISM or THROMBOEMBOLSM or VENOUS next THROMBOSIS) not MEDLINE not EMBASE

Table 14. Adverse effects of prophylaxis for venous thromboembolism: (more...)

Table 14. Adverse effects of prophylaxis for venous thromboembolism: Supplemental literature search in MEDLINE (1966--January 31, 1999) using OVID

1 exp heparin/ae,to,ct 2 exp dextrans/ae,to,ct 3 ancrod/ae,to,ct 4 hirudin/ae,to,ct 5 1 or 2 or 3 or 4 6 Vena cava filters/ae,ct [Adverse Effects, Contraindications] 7 (((foot adj3 pump$1) or calf) adj3 pump$1).mp. [mp=title, abstract, registry number word, mesh subject heading] 8 7 and (ae or co).fs. 9 stockings.mp. 10 9 and (ae or co).fs. 11 (((((sequential adj4 compression) or intermittent) adj4 compression) or pneumatic) adj4 compression).mp. [mp=title, abstract, registry number word, mesh subject heading] 12 11 and (ae or co).fs. 13 Ultrasonography, doppler/ae,ct [Adverse Effects, Contraindications] 14 exp Ultrasonography, doppler, Duplex/ae,ct [Adverse Effects, Contraindications] 15 Ultrasonography, doppler, pulsed/ae [Adverse Effects] 16 Phlebography/ae,ct [Adverse Effects, Contraindications] 17 Thrombelastography/ae [Adverse Effects] 18 exp Plethysmography/ae [Adverse Effects] 19 dimer$1.mp. and (to or ae or co or ct).fs. [mp=title, abstract, registry number word, mesh subject heading] 20 thrombosis/ or thromboembolism/ or thrombophlebitis/ or pulmonary embolism/ 21 19 and 20 22 5 or 6 or 7 or 8 or 10 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 21 23 22 and su.fs. 24 22 and surgical.mp. 25 22 and surgery.mp. 26 22 and trauma$3.mp. 27 22 and posttrauma$3.mp. 28 exp "Wounds and injuries"/ 29 22 and 28 30 22 and in.fs. 31 23 or 24 or 25 or 26 or 27 or 29 or 30 32 31 not letter.pt. 33 32 not editorial.pt. 34 33 not review.pt. 35 case report/ and clinical trial.pt. 36 34 and 35 37 34 not case report/ 38 36 or 37 39 limit 38 to animal 40 limit 39 to human 41 38 not 39 42 40 or 41 43 meta-analysis/ 44 meta-analysis.pt. 45 medline.ti,ab. 46 ((metaanaly$ or meta) adj analy$5).ti,ab. 47 overview$.ti,ab. 48 systematic review$.ti,ab. 49 43 or 44 or 45 or 46 or 47 or 48 50 33 and 49 51 50 or 42

Table 15. Screening for venous thromboembolism: Supplemental literature (more...)

Table 15. Screening for venous thromboembolism: Supplemental literature search in MEDLINE (1966--January 31, 1999) using OVID

1 thrombosis/ra,ri,us 2 thrombophlebitis/ra,ri,us 3 thromboembolism/ra,ri,us 4 pulmonary embolism/ra,ri,us 5 1 or 2 or 3 or 4 6 thrombosis/ or thromboembolism/ or thrombophlebitis/ or pulmonary embolism/ 7 exp ultrasonography, doppler/ 8 phlebography/ 9 thrombelastography/ 10 exp plethysmography/ 11 (doppler or Duplex or venogram$ or venography or ipg).mp. 12 (ultrasound or ultrasonography).mp. 13 dimer$1.mp. 14 screen$3.mp. 15 or/7-14 16 6 and 15 17 5 or 16 18 (trauma$3 or posttrauma$3 or post-trauma$3).mp. 19 in.fs. 20 exp "wounds and injuries"/ 21 18 or 19 or 20 22 17 and 21 23 limit 22 to animal 24 limit 23 to human 25 22 not 23 26 24 or 25 27 26 not letter.pt. 28 27 not editorial.pt. 29 28 not review.pt. 30 case report/ and clinical trial.pt. 31 29 not case report/ 32 29 and 30 33 31 or 32 34 meta-analysis/ 35 meta-analysis.pt. 36 medline.ti,ab. 37 ((metaanaly$ or meta) adj analy$5).ti,ab. 38 overview$.ti,ab. 39 systematic review$.ti,ab. 40 34 or 35 or 36 or 37 or 38 or 39 41 40 and 26 42 33 or 41

Table 16. Cost of prevention for venous thromboembolism: Supplemental (more...)

Table 16. Cost of prevention for venous thromboembolism: Supplemental literature search in MEDLINE (1966--January 31, 1999) using OVID

1 thrombosis/ec or thrombophlebitis/ec or thromboembolism/ec or pulmonary embolism/ec 2 *thrombosis/ or *thromboembolism/ or *pulmonary embolism/ or *thrombophlebitis/ 3 exp cost allocation/ or exp cost control/ or exp cost of illness/ or exp cost-benefit analysis/ or exp health care costs/ or "Costs and cost analysis"/ 4 2 and 3 5 1 or 3 6 5 not letter.pt.

Table 17. Prevention of venous thromboembolism in all (more...)

Table 17. Prevention of venous thromboembolism in all types of patients: Inclusive literature search in MEDLINE (1966--January 31, 1999) using OVID

1 *thrombosis/ or *thrombophlebitis/ or *thromboembolism/ or *pulmonary embolism/ 2 thrombosis/pc or thrombophlebitis/pc or thromboembolism/pc or pulmonary embolism/pc 3 (prevent$ or thromboprophyla$ or prophyla$).mp. 4 1 and 3 5 1 and 2 6 4 or 5 7 exp heparin/ or ancrod/ or hirudin/ or exp dextrans/ 8 stockings.mp. 9 (vena cava filter$ or vena cava filter$).mp. 10 ((sequential adj4 compression) or (intermittent adj4 compression) or (pneumatic adj4 compression)).mp. 11 (vena cava filter$ or vena cava filter$).mp. 12 (greenfield adj4 filter$).mp. 13 ((foot adj3 pump$) or (calf adj3 pump$)).mp. 14 ((suprarenal adj3 filter$) or (umbrella adj3 filter$) or mobin$).mp. 15 (cava interruption$ or caval interruption$ or ivc interruption$).mp. 16 thrombosis/ or thromboembolism/ or thrombophlebitis/ or pulmonary embolism/ 17 3 and 7 and 16 18 or/8-15 19 16 and 18 20 6 or 19 or 17 21 20 not letter.pt. 22 21 not editorial.pt. 23 22 not case report/ 24 case report/ and clinical trial.pt. 25 22 and 24 26 23 or 25 27 meta-analysis/ 28 meta-analysis.pt. 29 medline.ti,ab. 30 ((metaanaly$ or meta) adj analy$5).ti,ab. 31 overview$.ti,ab. 32 systematic review$.ti,ab. 33 27 or 28 or 29 or 30 or 31 or 32 34 26 and 33 35 26 not review.pt. 36 34 or 35 37 exp research design/ 38 exp clinical trials/ 39 exp epidemiologic research design/ 40 exp epidemiologic study characteristics/ 41 comparative study/ or placebos/ 42 multicenter study.pt. 43 random$.ti,ab. 44 (double blind$ or triple blind$).mp. 45 (clinical adj trial$).mp. 46 placebo$.ti,ab. 47 practice guideline$.mp,pt. 48 feasibility studies/ or clinical protocols/ 49 exp treatment outcome/ 50 (volunteers or control$3).mp. 51 clinical trial$.pt. 52 (controlled clinical trial or randomized controlled trial).pt. 53 or/37-52 54 35 and 53 55 34 or 54 56 limit 55 to human 57 limit 55 to animal 58 55 not 57 59 56 or 58

Tables 11 through 13 list the complete set of search terms. Supplemental searches were performed to identify studies that addressed specific topics referring to the key questions and topics that might be missed by the broad search. Supplemental searches were performed on the adverse effects of the drugs and devices used to prevent VT (Table 14), on screening for VT (Table 15), and on the cost of prophylaxis for VT (Table 16). As the final step of the formal literature search, we included a broad search on the prevention of VT in all types of patients (not specifically in trauma patients) to ensure that no relevant studies were missed (Table 17).

In addition to the computer-aided literature searches, we checked the bibliographies of relevant studies (particularly review articles) and of pertinent chapters of the textbook, Venous Thromboembolism: An Evidence-Based Atlas (Hull, Raskob, Pineo, eds., 1996). None of the relevant references found in these bibliographies had been missed by our computer searches.

Table 18. Prevention of venous thromboembolism after (more...)

Table 18. Prevention of venous thromboembolism after injury: Final repeat search on only randomized controlled trials in MEDLINE (1966--January 31, 1999) using OVID

1 thrombosis/pc or thromboembolism/pc or thrombophlebitis/pc or pulmonary embolism/pc 2 thrombosis/ or thromboembolism/ or pulmonary embolism/ or thrombophlebitis/ 3 (thrombosis or thromboembolism or pulmonary embolism or thrombophlebitis or dvt or deep vein or venous thrombosis).mp. 4 (prevent$ or prophyla$ or thromboprophyla$).mp. 5 2 or 3 6 4 and 5 7 1 or 6 8 7 and randomized controlled trial.pt. 9 7 and clinical trial.pt. and random$.mp. 10 (trauma$ or posttrauma$ or post-trauma$ or injury or injuries or nonsurgical or non-surgical or fracture$).mp. 11 in.fs. 12 exp "wounds and injuries"/ 13 10 or 11 or 12 14 8 or 9 15 13 and 14

After we had evaluated all the studies and identified all the important articles, we performed a final search only on RCTs on the prevention of VT in trauma patients (Table 18). No new relevant articles were found in this search.

Design of the Database

To assemble and keep track of information from the literature reviews, one of the members of the core project staff designed a database in Microsoft Access^® with the assistance of professionals from Microsoft^®. All literature searches were imported using electronic mail from the librarian's desk directly into the database. The acceptance or rejection of titles, abstracts, or papers was also done directly on the computers by assigning categories from menus that included the predefined screening criteria. Electronic forms were developed to document the process of screening titles, abstracts, and full papers. Additional electronic forms were created to organize and tabulate the quality of studies, risk factors, treatment groups, and outcomes. The database design provided multiple flexible methods to create electronic queries to manage the data, generate evidence tables, and compile the desired information.

Title Screening Stage

Table 19. Inclusion and exclusion criteria: Screening (more...)

Table 19. Inclusion and exclusion criteria: Screening for titles

Inclusion Criteria

Exclusion Criteria

1 Trauma patients 2 Orthopedic and neurosurgical trauma patients Orthopedic injuries Not inpatient therapy Spinal or head injuries 3 General surgery (could not always exclude because it was unclear that it was only "elective") 4 Methods of prevention Complications of prevention Compression stockings Heparin Vena cava filters Methods of prevention 5 Methods of screening 6 Reviews and meta-analyses Meta-analysis Review Cost-efficiency analysis 7 Other VT etiology and incidence Diagnosis of DVT or PE Guidelines Other Background Clinical outcomes

1 Non-surgical patients Atrial fibrillation Cardiac: not valve, not myocardial infarction Medicine patients Myocardial infarction Pediatrics Stroke Thrombolytic therapy Non-drug, non-compression treatment Outpatients Cardiac valve replacement 2 Elective surgery patients Cardiac valve replacement Vascular surgery Gynecologic surgery Not trauma-related Elective general surgery 3 Animal and laboratory studies Animal In vitro clotting Drug mechanism 4 Other Cancer Ethics focus Treatment Arterial thrombosis Irrelevant topic Elderly patients

VT, venous thromboembolism; DVT, deep venous thrombosis; PE, pulmonary embolism.

A total of 4,093 titles were identified. Two nonblinded medical reviewers (a trauma surgeon with expertise in critical care and a clinical pharmacologist) independently screened all titles against inclusion and exclusion criteria (Table 19). Whenever the two reviewers disagreed, the Task Order Manager reviewed the titles, and then all three individuals met for a joint decision. If the title did not clearly merit exclusion, the article was accepted to the abstract review phase.

Table 20. Outcome of independent title screen of 4,093 (more...)

Table 20. Outcome of independent title screen of 4,093 titles from the two medical reviewers, and reconciliation of disagreements

Reviewer No. 1	Reviewer No. 2	Total
Accept	Accept	2,017 (49.3%)
Reject	Reject	1,523 (37.2%)
	Subtotal	3,540 (86.5%)
Accept	Reject	117 (2.8%)
Reject	Accept	436 (10.7%)
	Subtotal	553 (13.5%)
Following Reconciliation
Accept	Accept	2,437 (59.5%)
Reject	Reject	1,656 (40.5%)

Table 21. Reasons for inclusion and exclusion in title (more...)

Table 21. Reasons for inclusion and exclusion in title screening of 4,093 article titles

Inclusion/Exclusion Criteria	Accept/Reject	N
Background	Accept	66
Clinical outcomes	Accept	118
Complications of prevention	Accept	79
Compression stockings	Accept	55
Cost-effectiveness issues	Accept	44
Diagnosis of DVT or PE	Accept	239
Duplicates of other articles (found retrospectively)	Accept	43
General surgery	Accept	156
Guidelines	Accept	13
Meta-analysis	Accept	16
Methods of prevention	Accept	361
Orthopedic injuries	Accept	115
Orthopedic operations	Accept	342
Other	Accept	32
Reviews	Accept	46
Screening methods	Accept	135
Spine or head injuries	Accept	100
Trauma patients	Accept	94
Vascular/trauma patients	Accept	153
Vena cava filters	Accept	184
Venous thromboembolism etiology and/or incidence	Accept	46
TOTAL INCLUDED		2,437
Animal studies	Reject	9
Arterial thrombosis	Reject	54
Atrial fibrillation	Reject	37
Cancer	Reject	93
Cardiac valve replacement	Reject	86
Cardiac: not valve, not MI	Reject	115
Drug mechanism	Reject	10
Elderly patients	Reject	13
Ethnic focus	Reject	11
Gynecologic surgery	Reject	124
In vitro clotting studies	Reject	52
Irrelevant topic	Reject	762
Medicine patients	Reject	25
Myocardial infarction	Reject	69
Nondrug/noncompression treatment	Reject	7
Outpatients	Reject	25
Pediatrics	Reject	65
Stroke	Reject	33
Thrombolytic therapy	Reject	25
Treatment	Reject	41
TOTAL EXCLUDED		1,656

DVT, deep venous thrombosis; PE, pulmonary embolism; MI, myocardial infarction.

Table 20 shows the results of the two reviewers' independent title screen. After reconciliation of disagreements between the reviewers, 2,437 (59.5 percent) of these titles were accepted and passed to the abstract review phase. Table 21 shows the outcomes of the title screen process.

While the MEDLINE/EMBASE screen was in process, the Task Order Manager reviewed in parallel the literature search from the Cochrane database. This search was restricted to extract titles that had not been retrieved by the MEDLINE or EMBASE searches; therefore, the yield was low. Only two titles were added from the "Database of Abstracts of Reviews of Effectiveness" component of the Cochrane database. Both referred to unpublished conference presentations.

Abstract Screening Stage

Table 22. Inclusion and exclusion criteria: Screening (more...)

Table 22. Inclusion and exclusion criteria: Screening for abstracts

Inclusion Criteria Trauma patients Methods of prevention High-risk group Screening methods Vena cava filters in trauma

Exclusion Criteria Diagnosis/treatment of venous thromboembolism Methods of diagnosis of venous thromboembolism Nontrauma: orthopedic or neurosurgical patients Irrelevant topic Elective operations Nonclinical studies Inappropriate study design (panel discussion, comments, letters, etc.) Burns

The selected abstracts were screened against predetermined inclusion and exclusion criteria (Table 22). The inclusion criteria were related directly to the four key questions. Abstracts were rejected if their material related to any of the following:

• Treatment of VT (instead of prevention).

• Diagnosis of VT (instead of screening).

• Elective surgery patients (instead of trauma).

• Burns (after recommendation from the technical experts).

• Study did not include humans (animals, experimental).

• Study was not designed as a research project (panel discussions, editorials, letters, etc.).

• Study addressed an irrelevant topic.

The screening of abstracts was completed in a nonblinded fashion at meetings between the two medical reviewers and the Task Order Manager. Each abstract was discussed, and a consensus was reached on its acceptance or rejection. Titles without an abstract were accepted for retrieval of the full article. During the screening process, the inclusion and exclusion criteria were slightly revised to better reflect the populations studied.

Table 23. Results of abstract screening of 2,437 abstracts (more...)

Table 23. Results of abstract screening of 2,437 abstracts

Inclusion/Exclusion Criterion	Accept/Reject	N
Trauma	Accept	79
Head/spinal-cord injury	Accept	40
Trauma/orthopedic injuries	Accept	81
Possibly trauma	Accept	12
Screening methods	Accept	4
Vena cava filters	Accept	11
Economic 1	Accept	25
Meta-analysis 1	Accept	19
Total Articles Retrieved		271
Diagnosis/treatment of venous thromboembolism	Reject	127
Elective surgery	Reject	254
Healthy volunteers	Reject	8
Nontrauma orthopedic or neurosurgical patients	Reject	450
Irrelevant topic/inappropriate study design	Reject	1,327
Total Articles Rejected		2,166
Total		2,437

1

Articles retrieved for useful background material only; not intended for analysis.

Table 23 shows the final outcome of the abstract screening phase. A total of 271 abstracts passed to the article retrieval phase. Even if they did not strictly refer to trauma patients, abstracts on meta-analysis and cost-effectiveness analysis (44 studies) were accepted for retrieval in order to assess the methods and references used in similar studies. Two hundred twenty-six abstracts of original research studies were accepted for complete review.

Data Collection Forms

The core project staff designed data-collection forms and tested data entry from 20 studies into the forms. After revisions, the forms were approved by the TOO and the entire group (Appendix A). The final form had three sections: a General Information front page, a Quality Screen, and an Evaluation Form.

Article Screen

All articles but one (a Turkish-language article not found after two requests through the library to the national retrieval network) were retrieved (225 articles total). Two medical reviewers and the Task Order Manager reviewed the studies in a nonblinded fashion and completed the data-collection forms. Two independent reviewers screened every study: each medical reviewer screened half of the studies, and the Task Order Manager screened all the studies. Disagreements and inconsistencies were resolved in a meeting with all three reviewers. Translators with a medical background were used for a number of foreign-language articles.

Table 24. Reasons for exclusion of 152 of 225 retrieved (more...)

Table 24. Reasons for exclusion of 152 of 225 retrieved full studies after revisiting the rejection criteria

Rejection Criterion	No. of Articles
Irrelevant topic (laboratory, animals, therapy, etc.)	15
Irrelevant population (elderly, medical, etc.)	79
Method of prophylaxis after elective surgery	11
Inappropriate study category (review, editorial, etc.)	17
Other
Same patients as another accepted study	7
Old study or old drug no longer used	5
Patients with established DVT or PE	3
Cost-effectiveness in different populations	2
Inadequate or unreliable data	9
Causes of thrombosis	1
Does not address any of the four questions	3
Total	152

DVT, deep venous thrombosis; PE, pulmonary embolism.

Table 25. The 73 accepted studies by country, language (more...)

Table 25. The 73 accepted studies by country, language of publication, and target population

Country	No. of Accepted Studies
Australia	1
Belgium	1
Canada	6
England	1
Finland	1
France	1
Germany	4
Poland	1
Spain	1
Switzerland	1
USA	55
Language of Publication	No. of Accepted Studies
English	67
French	2
German	3
Polish	1
Target Population	No. of Accepted Studies
General trauma	33
Orthopedic trauma	10
Neurosurgical trauma	26
Minor trauma	4

Only studies that passed the Front Page "revisiting the rejection criteria" screen were evaluated. Table 24 presents the number of studies excluded and the reason for their rejection during the article-screening phase. A total of 73 studies were ultimately accepted and reviewed and had the Quality Screen and Evaluation Form sections completed. Table 25 lists these studies and their country, language of publication, and target population.

Quality of Studies

Among 73studies, there were the following:

• Nineteen randomized controlled trials.

• Seventeen comparative-cohort studies (8 comparing two prospective study cohorts, 7 comparing a prospective cohort with a retrospective cohort, and 2 comparing two retrospective cohorts).

• Thirty-seven single-cohort, observational prospective, or retrospective studies.

Table 26. Quality scores of the 73 accepted studies (more...)

Table 26. Quality scores of the 73 accepted studies by study design

Design	Range of Quality Scores (median quality score)	Number of Studies	Mean Quality Score ± SD	Number of Studies with Quality Score > Median Quality Score
Randomized controlled trial	0 to 5 (3)	19	2.1 ± 1.6	6 (32%)
Single-cohort study	0 to 5 (3)	37	1.9 ± 1.3	11 (30%)
Comparative-cohort study	0 to 8 (4)	17	4.0 ± 1.9	13 (76%)

SD, standard deviation.

Figure 1. Quality scores of the 73 accepted studies (more...)

   Figure 1. Quality scores of the 73 accepted studies according to study design

Table 26 shows the quality scores according to study design. Approximately two-thirds of the RCTs, one-third of the single-cohort studies, and three-fourths of the comparative-cohort studies received quality scores that were lower than the median of the highest possible score for the category (e.g., 3 on a scale of 0 to 5). Figure 1 shows the variation of quality scores among studies.

Synthesis of Existing Data

From the accepted studies, we produced the following:

• Evidence tables for each study.

• Calculations of the incidence of DVT and PE.

• Meta-analysis and supplemental analyses for each of the four questions.

Supplemental Analyses

We calculated the incidence of secondary outcomes such as fatal PE or complications of methods of prophylaxis. Because the data were very limited, different groups could not be compared.

Cost-Effectiveness Methodology

Figures 2 and 3. Decision tree for diagnosis and treatment (more...)

   Figures 2 and 3. Decision tree for diagnosis and treatment of venous thromboembolism

The cost-effectiveness (C-E) analysis evaluated different approaches for preventing VT in major trauma patients. We did not include patients with minor trauma in this analysis because the available literature was limited. We used the TreeAge software program (TreeAge Software, Inc., Williamstown, MA) to develop an initial decision-tree model based on different outcome probabilities with regard to VT. We derived the incidences of most of the various outcomes from our meta-analysis. In areas not evaluated in the meta-analysis, we derived outcome rates from review of the literature. Outcomes with very low incidence (<1 percent) were eliminated to simplify the final decision-tree model (Figures 2 and 3).

We made the following assumptions when we designed the decision-tree model:

• All patients' bilateral lower extremities were screened by Duplex ultrasonography once per week for as long as they were nonambulatory.

• The mean nonambulatory period was 2 weeks (Knudson, Collins, Goodman, et al., 1992).

• DVT found by Duplex was not confirmed by venography.

• DVT treatment consisted of intravenous heparin administered to achieve therapeutic levels (partial thromboplastin time [PTT] and international normalized ratio [INR] of at least twice normal).

• No routine screening for PE was performed, and the diagnosis was based on clinical symptoms.

• Diagnosis of PE consisted of ventilation-perfusion (V/Q) scan initially (Gottschalk, Bisesi, and Stein, 1996). If the V/Q scan indicated high probability for PE, the patient was treated. If the V/Q scan indicated intermediate probability, a pulmonary angiogram was performed and the patient was treated only if the angiogram was positive. If the V/Q scan indicated low probability or was normal, the patient was not treated.

• Each type of V/Q scan (high, intermediate, or low probability) was found in one-third of patients evaluated for PE.

• Intravenous heparin was used to treat PE. A vena cava filter was used only if there was contraindication to full heparinization (estimated to occur in 20 percent of patients). Our model did not include the prophylactic use of vena cava filters in high-risk patients.

• "Breakthrough" PEs that occurred despite adequate treatment with intravenous heparin were given a probability of less than 1 percent and were eliminated from the refined model.

The following indices were based on our meta-analysis or individual literature data:

• The incidence of adverse reactions after administration of VT prophylaxis was the same for LDH and LMWH (3 percent; range 2 percent to 4 percent), as shown by our analysis (see Chapter 4. Supplemental Analysis:Analysis of Complications of Methods of VT Prevention). SCDs were associated with no adverse reactions.

• The incidence of DVT in patients with or without prophylaxis was 12 percent (range 10 percent to 13 percent), as shown by our analysis (see Chapter 3. Results: Incidence of DVT and PE in Trauma Patients).

• The sensitivity of Duplex ultrasonography in detecting lower-extremity DVT was 80 percent (Lensing, Davidson, Prins, et al., 1996). The incidence of false negatives was 20 percent.

• The incidence of adverse reactions to full heparinization for established DVT was 11 percent (Hull, Raskob, Rosenbloom, et al., 1990).

• In the absence of reliable data, the incidence of death related directly to adverse reactions following full heparinization was arbitrarily placed at 5 percent of the patients who developed adverse reactions.

• The incidence of patients who developed PE after a falsely negative Duplex scan was 3 percent, as shown by our analysis (see Chapter 3. Results: Incidence of DVT and PE in Trauma Patients). Although our evidence report found the overall incidence of PE to be 1.5 percent, we used the highest end of the 95 percent confidence interval (1 percent to 3 percent) because we believe that the existing literature data show artificially low rates of PE because of insufficient index of suspicion and diagnostic evaluation. Other studies in surgical nontrauma patients show incidences of PE that range between 20 percent and 50 percent in patients who have unrecognized, and therefore untreated, DVT (Brandjes, Heijboer, Buller, et al., 1992; Hull, Raskob, Hirsh et al., 1986).

• The incidence of fatal PE in those with PE was 30 percent, as shown by our analysis (see Chapter 4. Supplemental Analysis: Incidence of Fatal PE).

• The incidence of false-negative V/Q scan (normal or low probability) was 14 percent (PIOPED Investigators, 1990).

• The incidence of death in patients with missed diagnosis of PE (false-negative V/Q scan) was 25 percent (Barritt and Jordan, 1960).

• The incidence of adverse reactions in patients who received heparin for treatment of PE was the same as the incidence in those receiving heparin for treatment of DVT (11 percent). The incidence of adverse reactions in those who had contraindication to heparin (20 percent of patients diagnosed with PE) and received a vena cava filter as treatment of PE was 0.

The cost estimates used average wholesale prices for the cost of drugs. We derived the cost estimates for medical services associated with each outcome branch from the resource-based relative-value scale (RBRVS) (St. Anthony's Complete RBRVS, 1997) of the Health Care Financing Administration. When multiple RBRVS prices were listed, we used the lowest price. The analysis was restricted to prices in the United States.

Initially, we compared the cost-effectiveness of VT prophylaxis with no prophylaxis. VT prophylaxis could be provided by one of the three most commonly used methods: LDH, LMWH, or SCDs. Because the meta-analysis did not identify any difference in the incidence of VT between prophylaxis and no prophylaxis (see Chapter 3. Results), the two arms of the decision tree differed only with regard to the possibility of adverse reactions related to the method of prophylaxis.

Following the C-E analysis, we estimated the reduction in the incidence of VT that must be achieved by each of the three prophylactic methods relative to no prophylaxis in order to become cost-effective. The results of this sensitivity analysis will help future investigators to design clinical trials of different methods of VT prophylaxis with adequate power to detect the magnitude of differences in outcomes required for cost-effectiveness.

Incidence of DVT and PE in Trauma Patients

Classification of Studies According to Design, Screening, and Prophylaxis

We classified the studies according to three variables:

• RCT or non-RCT.

• Use of routine screening or no routine screening.

• Type of prophylaxis.

Table 27. Methods of prophylaxis employed in accepted studies (more...)

Table 27. Methods of prophylaxis employed in accepted studies

Low-dose heparin

Low-molecular-weight heparin

Low-dose heparin or coumadin + sequential compression device

Low-dose heparin + dihydroergotamine

Heparin + electric stimulation

Polysulfate of pentosane

Suloctidil

Aspirin

Nadroparin

Sequential compression device

Sequential compression device + aspirin + dipyridamole

Sequential compression device + organon

Arteriovenous foot pump

Sequential compression device + arteriovenous foot pump

Vena cava filter

Despite the long list of pharmaceutical and mechanical methods of prophylaxis used in the comparative studies (Table 27), the three most frequently used methods were LDH, LMWH, and mechanical devices (SCDs or AFPs). We included only these three methods in our calculations. If a study included both patients receiving prophylaxis and patients not receiving any prophylaxis, we entered data from each group separately into the appropriate category of the prophylaxis/no prophylaxis field for the calculation of DVT and PE rates. Therefore, the same study could be used in two opposite fields.

Incidence

Of 73 studies, 67 reported on DVT, 49 on PE, and 43 on both. The incidence of DVT in a total of 12,527 patients was 11.8 percent (95 percent CI: 0.104, 0.131) and the incidence of PE in 22,336 patients was 1.5 percent (95 percent CI: 0.011, 0.018). The number of patients included in the PE studies was higher than the number of patients included in the DVT studies, primarily because of one study (Winchel, Hoyt, Walsh, et al., 1994) that reported on the incidence of PE in 9,721 trauma patients discharged from one trauma center over a period of 8 years. When this study was excluded, the incidence of PE remained essentially unchanged, increasing from 1.5 percent to 1.7 percent (95 percent CI: 0.013, 0.022). For patients who received some type of prophylaxis, the rates of DVT and PE were 6.8 percent (95 percent CI: 0.053, 0.083) and 1.8 percent (95 percent CI: 0.010, 0.020), respectively. For patients who received no prophylaxis, the rates were 10.1 percent (95 percent CI: 0.062, 0.140) and 1 percent (95 percent CI: -0.008, 0.028), respectively.

Table 28. Incidence of DVT and PE in different types (more...)

Table 28. Incidence of DVT and PE in different types of trauma patients who received LDH for VT prophylaxis

Type of Patient	Randomized Routine Screening	Number of Studies	Number of Patients	Number of Patients with Outcome	Incidence (random effects estimate)	95% Confidence Interval	Heterogeneity Test
							Q Test	p Value
DVT
All Patients	Randomized	10	549	89	0.145	0.061, 0.229	120.60	<0.001
	Not Randomized	3	351	27	0.175	-0.022, 0.373	20.22	<0.001
	Routine Screening	12	870	113	0.144	0.086, 0.201	139.53	<0.001
	No Routine Screening	1	30	3	(0.100) 1
General Trauma	Randomized	5	438	68	0.119	0.004, 0.235	97.95	<0.001
	Not Randomized	3	351	27	0.175	-0.022, 0.373	20.22	<0.001
	Routine Screening	8	789	95	0.127	0.063, 0.191	118.30	<0.001
Ortho Trauma	Randomized	0
	Not Randomized	0
Neuro Surgical	Randomized	4	81	18	0.202	0.057, 0.347	9.91	0.019
	Not Randomized	0
	Routine Screening	4	81	18	0.202	0.057, 0.347	9.91	0.019
Minor Trauma	Randomized	1	30	3	(0.100) 1
	Not Randomized	0
	No Routine Screening	1	30	3	(0.100) 1
PE
All Patients	Randomized	4	223	6	0.028	0.014, 0.071	6.36	0.095
	Not Randomized	2	303	9	0.147	-0.156, 0.450	9.79	0.002
	No Routine Screening	6	526	15	0.016	-0.005, 0.038	16.76	0.005
General Trauma	Randomized	2	173	1	0.001	-0.009, 0.011	0.99	0.320
	Not Randomized	1	281	2	(0.007) 1
	No Routine Screening	3	454	3	0.004	-0.003, 0.011	1.73	0.420
Ortho Trauma	Randomized	0
	Not Randomized	0
Neuro Surgical	Randomized	2	50	5	0.100	0.017, 0.183	0.01	0.923
	Not Randomized	1	22	7	(0.320) 1
	No Routine Screening	3	72	12	0.0149	0.035, 0.264	4.10	0.129
Minor Trauma	Randomized	0
	Not Randomized	0

1

A crude incidence and not a random-effects estimate was calculated when only one study was available. A row for Routine Screening or No Routine Screening was not included when no studies were identified in these categories.

DVT, deep venous thrombosis; PE, pulmonary embolism; LDH, low-dose heparin; VT, venous thromboembolism.

Table 29. Incidence of DVT and PE in different types (more...)

Table 29. Incidence of DVT and PE in different types of trauma patients who received LMWH for VT prophylaxis

Type of Patient	Randomized Routine Screening	Number of Studies	Number of Patients	Number of Patients with Outcome	Incidence (random effects estimate)	95% Confidence Interval	Heterogeneity Test
							Q Test	p Value
DVT
All Patients	Randomized	6	714	56	0.056	0.016, 0.095	70.71	<0.001
	Not Randomized	0
	Routine Screening	6	714	56	0.056	0.016, 0.095	70.71	<0.001
General Trauma	Randomized	2	249	41	0.157	-0.139, 0.452	52.71	<0.001
	Not Randomized	0
	Routine Screening	2	249	41	0.157	-0.139, 0.452	52.71	<0.001
Ortho Trauma	Randomized	0
	Not Randomized	0
Neuro Surgical	Randomized	1	20	0	(0) 1
	Not Randomized	0
	Routine Screening	1	20	0	(0) 1
Minor Trauma	Randomized	3	445	15	0.034	-0.011, 0.078	14.71	0.001
	Not Randomized	0
	No Routine Screening	3	445	15	0.034	-0.011, 0.078	14.71	0.001
PE
All Patients	Randomized	3	275	1	0.003	-0.006, 0.011	0.67	0.715
	Not Randomized	2	377	7	0.023	-0.019, 0.065	1.39	0.239
	No Routine Screening	5	652	8	0.007	-0.001, 0.016	4.73	0.316
General Trauma	Randomized	1	129	1	(0.008) 1
	Not Randomized	0
Ortho Trauma	Randomized	0
	Not Randomized	0
Neuro Surgical	Randomized	1	20	0	(0) 1
	Not Randomized	2	377	7	0.023	-0.019, 0.065	1.39	0.239
	No Routine Screening	3	397	7	0.015	0.003, 0.027	1.59	0.452
Minor Trauma	Randomized	1	126	0	(0) 1
	Not Randomized	0

1

A crude incidence and not a random effects estimate was calculated when only one study was available. A row for Routine Screening or No Routine Screening was not included when no studies were identified in these categories.

DVT, deep venous thrombosis; PE, pulmonary embolism; LMWH, low-molecular-weight heparin; VT, venous thromboembolism.

Table 30. Incidence of DVT and PE in different types (more...)

Table 30. Incidence of DVT and PE in different types of trauma patients who received mechanical prophylaxis for VT

Type of Patient	Randomized Routine Screening	Number of Studies	Number of Patients	Number of Patients with Outcome	Incidence (random effects estimate)	95% Confidence Interval	Heterogeneity Test
							Q Test	p Value
DVT
All Patients	Randomized	6	266	18	0.049	0.005, 0.092	14.12	0.015
	Not Randomized	3	170	25	0.121	- 0.05, 0.246	9.87	0.007
	Routine Screening	9	436	43	0.073	0.026, 0.120	32.15	<0.001
	No Routine Screening	0
General Trauma	Randomized	4	216	11	0.037	0.000, 0.075	5.66	0.129
	Not Randomized	2	156	25	0.178	0.060, 0.296	2.02	0.156
	Routine Screening	6	372	36	0.079	0.024, 0.134	22.80	<0.001
Ortho Trauma	Randomized	1	35	1	(0.029) 1
	Not Randomized	0
	Routine Screening	1	35	1	(0.029) 1
Neuro Surgical	Randomized	1	15	6	(0.40) 1
	Not Randomized	1	14	0	(0) 1
	Routine Screening	2	29	6	0.183	-0.208, 0.573	8.82	0.003
Minor Trauma	Randomized	0
	Not Randomized	0
PE
All Patients	Randomized	2	111	7	0.052	0.003, 0.101	1.45	0.228
	Not Randomized	2	40	6	0.153	-0.044, 0.350	2.52	0.113
	No Routine Screening	4	150	13	0.079	0.014, 0.025	5.22	0.156
General Trauma	Randomized	1	76	6	(0.079) 1
	Not Randomized	1	26	2	(0.077) 1
	No Routine Screening	2	102	8	0.078	0.026, 0.131	0.00	0.973
Ortho Trauma	Randomized	1	35	1	(0.029) 1
	Not Randomized	0
Neuro Surgical	Randomized	0
	Not Randomized	1	14	4	(0.29) 1
Minor Trauma	Randomized	0
	Not Randomized	0

1

A crude incidence and not a random-effects estimate was calculated when only one study was available. A row for Routine Screening or No Routine Screening was not included when no studies were identified in these categories.

DVT, deep venous thrombosis; PE, pulmonary embolism; VT, venous thromboembolism.

Table 31. Incidence of DVT and PE in different types (more...)

Table 31. Incidence of DVT and PE in different types of trauma patients who received no VT prophylaxis

Type of Patient	Randomized Routine Screening	Number of Studies	Number of Patients	Number of Patients with Outcome	Incidence (random effects estimate)	95% Confidence Interval	Heterogeneity Test
							Q Test	p Value
DVT
All Patients	Randomized	9	511	53	0.100	0.051, 0.149	28.78	<0.001
	Not Randomized	5	223	25	0.117	0.035, 0.200	16.65	0.002
	Routine Screening	14	734	78	0.101	0.062, 0.140	45.60	<0.001
General Trauma	Randomized	3	130	9	0.061	0.006, 0.116	3.10	0.212
	Not Randomized	3	188	17	0.083	0.000, 0.166	9.44	0.009
	Routine Screening	6	318	26	0.068	0.025, 0.111	12.55	0.028
Ortho Trauma	Randomized	1	38	3	(0.079) 1
	Not Randomized	0
	Routine Screening	1	38	3	(0.079) 1
Neuro Surgical	Randomized	3	53	13	0.227	0.004, 0.450	9.98	0.007
	Not Randomized	2	35	8	0.216	-0.019, 0.450	3.09	0.079
	Routine Screening	5	88	21	0.215	0.076, 0.355	13.22	0.010
Minor Trauma	Randomized	2	290	28	0.101	-0.019, 0.221	11.19	0.001
	Not Randomized	0
	Routine Screening	2	290	28	0.101	-0.019, 0.221	11.19	0.001
PE
All Patients	Randomized	2	165	1	0.001	-0.009, 0.012	0.98	0.322
	Not Randomized	2	132	4	0.036	-0.037, 0.109	1.55	0.213
	No Routine Screening	4	297	5	0.010	-0.008, 0.028	4.57	0.206
General Trauma	Randomized	0
	Not Randomized	1	114	2	(0.018) 1
Ortho Trauma	Randomized	1	38	1	(0.026) 1
	Not Randomized	0
Neuro Surgical	Randomized	0
	Not Randomized	1	18	2	(0.111) 1
Minor Trauma	Randomized	1	127	0	(0) 1
	Not Randomized	0

1

A crude incidence and not a random-effects estimate was calculated when only one study was available. A row for Routine Screening or No Routine Screening was not included when no studies were identified in these categories.

DVT, deep venous thrombosis; PE, pulmonary embolism; VT, venous thromboembolism.

Tables 28 to 31 show the different pooled rates of DVT and PE according to the method of prophylaxis. According to results from RCTs, the incidences of DVT and PE are 14.5 percent and 2.8 percent, respectively, for patients who received LDH; 5.6 percent and 0.3 percent for patients who received LMWH; 4.9 percent and 5.2 percent for patients who received mechanical prophylaxis; and 10 percent and 0.1 percent for patients who received no prophylaxis. Most of these results were extracted from heterogeneous studies and are associated with wide confidence intervals.

Table 32. Incidence of DVT and PE in all trauma patients, (more...)

Table 32. Incidence of DVT and PE in all trauma patients, included only in RCTs, who received DH or LMWH or mechanical prophylaxis or no prophylaxis for VT

Intervention	Outcome	Number of Studies	Number of Patients in Studies	Number of Patients with Outcome	Incidence (random effects estimate)
No Prophylaxis	DVT	9	511	53	0.100
	PE	2	165	1	0.001
LDH	DVT	10	549	89	0.145
	PE	4	223	6	0.028
LMWH	DVT	6	714	56	0.056
	PE	3	275	1	0.003
Mechanical Prophylaxis	DVT	6	266	18	0.049
	PE	2	111	7	0.052

DVT, deep venous thrombosis; PE, pulmonary embolism; RCT, randomized controlled trial; LDH, low-dose heparin; LMWH, low-molecular-weight heparin; VT, venous thromboembolism.

Table 33. Incidence of DVT and PE in general trauma (more...)

Table 33. Incidence of DVT and PE in general trauma patients, included only in RCTs, who received LDH or LMWH or mechanical prophylaxis or no prophylaxis for VT

Intervention	Outcome	Number of Studies	Number of Patients in Studies	Number of Patients with Outcome	Incidence (Random Effects Estimate)
No Prophylaxis	DVT	3	130	9
	PE	0
LDH	DVT	5	438	68	0.119
	PE	2	173	1	0.001
LMWH	DVT	2	249	41	0.157
	PE	1	129	1	(0.008) 1
Mechanical Prophylaxis	DVT	4	216	11	0.037
	PE	1	76	6	(0.079) 1

1

A crude incidence and not a random-effects estimate was calculated when only one study was available.

DVT, deep venous thrombosis; PE, pulmonary embolism; RCT, randomized controlled trial; LDH, low-dose heparin; LMWH, low-molecular-weight heparin; VT, venous thromboembolism.

Table 34. Incidence of DVT and PE in orthopedic trauma (more...)

Table 34. Incidence of DVT and PE in orthopedic trauma patients, included only in RCTs, who received LDH or LMWH or mechanical prophylaxis or no prophylaxis for VT

Intervention	Outcome	Number of Studies	Number of Patients in Studies	Number of Patients with Outcome	Incidence (random effects estimate)
No Prophylaxis	DVT	1	38	3	(0.079) 1
	PE	1	38	1	(0.026) 1
LDH	DVT	0
	PE	0
LMWH	DVT	0
	PE	0
Mechanical Prophylaxis	DVT	1	35	1	(0.029) 1
	PE	1	35	1	(0.029) 1

1

A crude incidence and not a random effects estimate was calculated when only one study was available.

DVT, deep venous thrombosis; PE, pulmonary embolism; RCT, randomized controlled trial; LDH, low-dose heparin; LMWH, low-molecular-weight heparin; VT, venous thromboembolism.

Table 35. Incidence of DVT and PE in neurosurgical (more...)

Table 35. Incidence of DVT and PE in neurosurgical trauma patients, included only in RCTs, who received LDH or LMWH or mechanical prophylaxis or no prophylaxis for VT

Intervention	Outcome	Number of Studies	Number of Patients in Studies	Number of Patients with Outcome	Incidence (random effects estimate)
No Prophylaxis	DVT	3	53	13	0.227
	PE	0
LDH	DVT	4	81	18	0.0202
	PE	2	50	5	0.100
LMWH	DVT	1	20	0	(0) 1
	PE	1	20	0	(0) 1
Mechanical Prophylaxis	DVT	1	15	6	(0.40) 1
	PE	0

1

A crude incidence and not a random-effects estimate was calculated when only one study was available.

DVT, deep venous thrombosis; PE, pulmonary embolism; RCT, randomized controlled trial; LDH, low-dose heparin; LMWH, low-molecular-weight heparin; VT, venous thromboembolism

Table 36. Incidence of DVT and PE in minor trauma patients, (more...)

Table 36. Incidence of DVT and PE in minor trauma patients, included only in RCTs, who received LDH or LMWH or mechanical prophylaxis or no prophylaxis for VT

Intervention	Outcome	Number of Studies	Number of Patients in Studies	Number of Patients with Outcome	Incidence (random effects estimate)
No Prophylaxis	DVT	2	290	28	0.101
	PE	1	127	0	(0)1
LDH	DVT	1	30	3	(0.100)1
	PE	0
LMWH	DVT	3	445	15	0.034
	PE	1	126	0	(0)1
Mechanical Prophylaxis	DVT	0
	PE	0

1

A crude incidence and not a random-effects estimate was calculated when only one study was available.

DVT, deep venous thrombosis; PE, pulmonary embolism; RCT, randomized controlled trial; LDH, low-dose heparin; LMWH, low-molecular-weight heparin; VT, venous thromboembolism.

Tables 32 to 36 show the different rates of DVT and PE according to the method of prophylaxis in the four predetermined types of trauma patients (general trauma, orthopedic trauma, neurosurgical trauma, and minor trauma), as well as in all trauma patients receiving any of the three methods of prophylaxis or no prophylaxis. Only RCTs were included in the calculation of these rates in order to produce more internally valid data. The incidence of DVT and PE for patients receiving no prophylaxis was calculated from data on the placebo group in the RCTs.

It is important to recognize that the incidences calculated in Tables 28 to 36 were derived by grouping together patients of different studies who received the same methods of prophylaxis. This analysis provides only an approximate estimate of the incidence of DVT and PE associated with each method of prophylaxis. It does not allow direct comparison among different methods of prophylaxis. Heterogeneity-as demonstrated by the results of the chi-squared heterogeneity test-among studies used in the various calculations indicates the high degree of variability in these rates. Some fields also had a limited sample of patients and studies (e.g., the "minor trauma" category). Therefore, these results should be interpreted with caution (see also Chapter 5. Conclusions).

Question 1. What Is the Best Method of VT Prophylaxis?

This question was considered to be the most important by the panel of the technical experts. We meta-analyzed RCTs that used the same methods of prophylaxis. Because the number of RCTs was limited, in a second step we also meta-analyzed studies of any design (RCT and non-RCT) that included the same methods of prophylaxis.

Meta-Analysis of RCTs

Table 37. Methods of VT prophylaxis in the 13 RCTs (more...)

Table 37. Methods of VT prophylaxis in the 13 RCTs that were not used in the meta-analyses

RCT with LMWH LMWH vs. no prophylaxis (two studies) LMWH vs. SCD LMWH vs. aspirin LMWH fixed dose vs. LMWH adjusted dose LMWH vs. LWMH (different doses)

RCT with LDH LDH vs. dihydroergotamine LDH vs. polysulfate of pentosane LDH vs. SCD LDH fixed dose vs. LDH adjusted dose

RCT with SCD SCD vs. AFP SCD vs. aspirin/dipyridamole

Other RCT Suloctidil vs. no prophylaxis

LMWH, low-molecular-weight heparin; LDH, low-dose heparin; SCD, sequential compression device; AFP, arteriovenous foot pump; VT, venous thromboembolism; RCT, randomized controlled trial.

Our meta-analysis of RCTs was limited by the number of studies that were sufficiently clinically homogenous to pool. Of 19 studies with a randomized design, we were able to use only six to help answer question 1. The remaining 13 studies could not be pooled for analysis (Table 37) because they all compared different treatments. These 13 randomized studies are described in Evidence Table 1. Because we chose to require at least three studies to perform each meta-analysis, there were sufficient data to make only two comparisons: LDH vs. no prophylaxis, and mechanical prophylaxis vs. no prophylaxis.

LDH vs. No Prophylaxis

Four RCTs compared LDH with no prophylaxis, two each of neurosurgical trauma (spinal patients) and general trauma patients. Two RCTs were included in the same original study (Knudson, Lewis, Clinton, et al., 1994) but were evaluated separately (see Chapter 2. Methodology: Article Screen).

Merli, Herbison, Ditunno, et al. (1988) found no difference in the incidence of DVT between 16 spinal cord injured patients who received LDH and 17 similar patients who received placebo. Routine screening for DVT was performed by 125-I fibrinogen scan every week. Venography was performed if the 125-I scan was positive or at the end of the 1-month period during which each patient was studied. Patients were admitted to the study hospital within 2 weeks of injury and were excluded if they had established DVT on admission. Eight patients developed venographically proven DVT in each group (50 percent LDH vs. 46 percent placebo).

Frisbie and Sasahara (1981) included spinal-cord-injured patients received from the acute care facility within 1 week of their trauma. For 60 days, the patients were surveyed weekly by impedance plethysmography. Positive scans were confirmed by venography. The incidence of DVT was very low and equal in both groups (1/17 or 6 percent LDH vs. /15 or 7 percent no prophylaxis). This low incidence is different from that found in other reports on similar patients and suggests an unidentified prophylactic factor or difference in sensitivity in detecting DVT.

Knudson, Lewis, Clinton, et al. (1994) examined the incidence of DVT in trauma patients who had any of a number of high-risk criteria, including laparotomy, thoracotomy, ventilation greater than 24 hours, spine fracture, pelvic fracture, and femur fracture. Evaluation for DVT was performed by Duplex ultrasonography at 5- to 7-day intervals until discharge or for at least 3 consecutive weeks. Patients who could receive LDH or SCD were randomized to LDH, SCD, or no-prophylaxis (control) groups. Patients who could not receive SCD because of lower extremity fractures were randomized to LDH or control groups. Both analyses showed no difference in the incidence of DVT between patients receiving LDH and patients with no prophylaxis. In the first analysis, the incidence of DVT was 2 percent (1/44) in LDH patients and 3 percent (2/64) in controls. In the second analysis, DVT was found in 5 percent (1/19) of LDH patients and 7 percent (2/27) of controls.

Table 38. Rates of DVT reported in four RCTs of trauma patients, (more...)

Table 38. Rates of DVT reported in four RCTs of trauma patients, comparing LDH with no prophylaxis

Study(first author-year)	Number of patients with LDH	Number of patients with NO PROPH	DVT Incidence In LDH Patients	DVT Incidence In NO PROPH Patients	Odds Ratio	95% CI of Odds Ratio
Merli-1988	16	17	50%	47%	1.1	0.3, 4.4
Frisbie-1981	15	17	7%	6%	1.1	0.1, 20.0
Knudson-1994	44	64	2%	3%	0.7	0.1, 8.2
Knudson-1994	19	27	5%	7%	0.7	0.1, 8.3
Random-effects estimates	94	125	0.108	0.106	0.965	0.353, 2.636
Heterogeneity chi-squared test p values			0.002	0.004	0.980

DVT, deep venous thrombosis; RCT, randomized controlled trial; LDH, low-dose heparin; NO PROPH, no prophylaxis; CI, confidence interval.

Figure 4. Shrinkage and funnel plots of RCTs comparing (more...)

   Figure 4. Shrinkage and funnel plots of RCTs comparing the rates of DVT in patients who received LDH or no prophylaxis

The combined analysis of the pooled data from these studies (Table 38, Figure 4, Evidence Table 2) shows no difference in the incidence of DVT between patients receiving LDH and patients receiving no prophylaxis for DVT (OR: 0.965, 95 percent CI: 0.353, 2.636). The chi-squared test is not significant for heterogeneity among studies (p = 0.980).

Mechanical Prophylaxis vs. No Prophylaxis

Three RCTs, two of general trauma patients and one of orthopedic trauma patients, compared mechanical prophylaxis vs. no prophylaxis. Mechanical prophylaxis was provided by SCD in all three RCTs.

Two RCTs were derived from one study (Knudson, Lewis, Clinton, et al., 1994). In trauma patients with predefined risk factors for VT, patients who could not receive LDH were randomized to either SCD or no prophylaxis. Another group of patients who did not have any contraindication to receiving LDH or SCD was randomized to LDH or SCD or no prophylaxis. Of those patients, only the ones who were randomized to SCD or no prophylaxis were included in this analysis. Duplex screening was performed weekly until discharge or for at least 3 consecutive weeks. The incidence of DVT in patients with contraindications to LDH was 0 percent for those randomized to SCD and 13 percent for those randomized to no prophylaxis (p=0.057). In patients without contraindications to LDH, the incidence was 12.5 percent for those randomized to receive SCD and 3 percent for those receiving no prophylaxis. This comparison was not statistically significant. The patients in the group who had contraindications to LDH were predominantly neurotrauma patients, whereas most patients in the group without contraindications did not have significant head or spinal-cord injuries. This study did not provide any additional data comparing patient characteristics between the two RCTs.

The dissimilar magnitudes of the differences in DVT rates between patients receiving SCD and patients receiving no prophylaxis across the two RCTs (0 percent vs. 13 percent in one RCT, but 12.5 percent vs. 3 percent in the other RCT) resulted from either or both of the following: (1) the first RCT included mostly neurotrauma patients and the other did not, and SCD offers good prophylaxis against DVT only in neurotrauma patients; or (2) the difference in DVT rates in the first RCT was almost significant, and in the other it was not; with adequate numbers to achieve statistical significance, both RCTs would show similar results.

Fisher, Blachut, Salvian, et al. (1995) reported on patients with hip and pelvic fractures. According to our predetermined criteria (see Chapter 1. Introduction: Defining Trauma Patients), only the 73 patients with pelvic fractures were included in our analysis. Duplex screening was performed every 5 days until the patient was ambulating. Thirty-five patients were randomized to the SCD group and one developed DVT, whereas 38 were randomized to the no-prophylaxis group and three developed DVT (3 percent vs. 8 percent, p=0.24). It should be noted that patients with severe trauma were not managed primarily by the authors, who belonged to the orthopedic service; therefore, those patients were excluded from the study.

Table 39. Rates of DVT reported in three RCTs of trauma patients, (more...)

Table 39. Rates of DVT reported in three RCTs of trauma patients, comparing mechanical prophylaxis (MP) with no prophylaxis

Study (first author-year)	Number of patients with MP	Number of patients with NO PROPH	DVT Incidence In MP Patients	DVT Incidence In NO PROPH Patients	Odds Ratio	95% CI of Odds Ratio
Fisher-1995	35	38	3%	8%	0.3	0.0, 3.5
Knudson-1994	32	64	12%	3%	4.4	0.8, 25.6
Knudson-1994	26	39	0%	13%	0.1	0.0, 2.2
Random-effects estimates	93	141	0.024	0.051	0.769	0.265, 2.236
Heterogeneity chi-squared test p values			0.145	0.198	0.061

DVT, deep venous thrombosis; RCT, randomized controlled trial; MP, mechanical prophylaxis; NO PROPH, no prophylaxis; CI, confidence interval.

Figure 5. Shrinkage and funnel plots of RCTs comparing (more...)

   Figure 5. Shrinkage and funnel plots of RCTs comparing the rates of DVT in patients who received mechanical prophylaxis or no prophylaxis

Overall, the analysis of the pooled data from these three RCTs (Table 39, Figure 5, Evidence Table 3) showed no difference in DVT rates between patients who received treatment with SCDs and those who did not (OR: 0.769, 95 percent CI: 0.265, 2.236). Because the confidence interval is very wide, a significant effect cannot be excluded. The chi-squared test approaches but does not reach statistical significance for heterogeneity among these studies (p=0.061).

Publication Bias Evaluation of Both Analyses

Funnel plots were created for both analyses (LDH vs. no prophylaxis and mechanical prophylaxis vs. no prophylaxis). They did not show any evidence of publication bias, but the small number of studies makes it difficult to draw firm conclusions from the funnel plots.

Meta-Analysis of RCTs and Non-RCTs

Because the number of RCTs was limited, we decided to perform a meta-analysis of the data using non-RCTs. Pooled analysis including observational data is inherently more prone to bias than is analysis of RCT-only data, and these results should be viewed with caution. However, this analysis may provide additional information about the roles of different method of prophylaxis.

We were able to make four comparisons: LDH vs. LMWH, LDH vs. mechanical prophylaxis, mechanical prophylaxis vs. no treatment, and LDH vs. no treatment. In the LDH vs. LMWH comparison, the outcome was PE because one of the three studies did not report DVT; DVT was the measured outcome in the other comparisons.

LDH vs. LMWH (for PE)

Three studies-two RCTs and one non-RCT (retrospective review)-were included in this analysis. One of these studies (Geerts, Jay, Code, et al., 1996) was the highest quality RCT in the trauma literature (Quality Score of 5, the highest possible score for RCTs). The authors of this study included major trauma patients (Injury Severity Score >9) and screened all patients by venography. Evaluation for PE was based on clinical suspicion and consisted mainly of ventilation/perfusion scan and, if needed, pulmonary angiography. One hundred thirty-four patients received LDH and 129 received LMWH. The incidence of DVT was 44 percent (60/134) in the LDH group and 31 percent (40/129) in the LMWH group, whereas the incidence of proximal DVT was 15 percent (20/134) and 6 percent (8/129) in the LDH and LMWH groups, respectively. The incidence of PE was 0 percent in the LDH group and 0.7 percent (1/129) in the LMWH group. The differences in the incidence of DVT and proximal DVT, but not PE, were statistically significant and favored LMWH. There were no fatal PEs in this study.

Green, Lee, Lim, et al. (1990) evaluated 41 spinal-cord-injured patients for thrombotic events following random administration of two prophylactic regimens: LDH (21 patients) and LWMH (20 patients). The patients were screened by impedance plethysmography and Duplex ultrasonography for 8 weeks. Of LDH patients, three (15 percent) developed proximal DVT and two (10 percent) developed fatal PE. No LMWH patients developed DVT or PE.

The non-RCT (Green, Twardowski, Wei, et al., 1994) included 51 patients with spinal-cord injuries. Nine of them suffered fatal PE. The incidence of PE in 22 patients who received LDH was 32 percent (7 patients) and in 27 patients who received LMWH was 7 percent (2 patients).

Table 40. Rates of PE reported in three studies (two (more...)

Table 40. Rates of PE reported in three studies (two RCT, one non-RCT) of trauma patients, comparing LMWH to LDH

Study (first author-year)	Number of Patients with LDH	Number of Patients with LMWH	PE Incidence In LDH Patients	PE Incidence In LMWH Patients	Odds Ratio	95% CI of Odds Ratio
Geerts-1996	136	129	0%	1%	0.3	0.0, 7.6
Green-1990	21	20	10%	0%	5.3	0.2, 116.6
Green-1994	22	27	32%	7%	5.8	1.1, 31.8
Random-effects estimates	179	176	0.113	0.009	3.010	0.585, 15.485
Heterogeneity chi-squared test p values			0.002	0.415	0.275

PE, pulmonary embolism; RCT, randomized controlled trial; LMWH, low-molecular-weight heparin; LDH, low-dose heparin; CI, confidence interval.

Figure 6. Shrinkage plot of RCTs and non-RCTs comparing (more...)

   Figure 6. Shrinkage plot of RCTs and non-RCTs comparing PE rates in patients who received LDH or LMWH

The meta-analysis of the pooled data from these three studies (Table 40, Figure 6, Evidence Table 4) shows no difference in PE rates in patients receiving LDH and those receiving LMWH (OR: 3.010, 95 percent CI: 0.585, 15.485). The confidence interval is very wide; therefore, a significant effect cannot be excluded. The chi-squared test indicates that the studies are not heterogeneous (p=0.275).

Two of the three studies evaluated DVT as an outcome (Geerts, Jay, Code, et al., 1996; Green, Lee, Lim, et al., 1990). In both studies, LMWH was associated with lower DVT rates relative to LDH. However, meta-analysis was not performed because we required at least three studies to perform meta-analysis.

LDH vs. Mechanical Prophylaxis

We identified four studies (two RCTs and two non-RCTs) comparing these two methods. Knudson, Collins, Goodman, et al. (1992) randomized 113 trauma patients to receive either LDH or SCD. The patients were screened by venous Doppler ultrasonography every 5 days for 3 weeks or until discharge. DVT was found in 3 of 37 LDH patients (8 percent) and 5 of 76 SCD patients (7 percent). Obviously, the discrepancy between the number of patients included in each of the two randomization groups is problematic.

In another study including three separate RCTs (as described in previous sections of this document), Knudson, Lewis, Clinton, et al. (1994) examined the incidence of DVT in trauma patients at risk for DVT. In one of these RCTs, the patients were randomized to LDH, mechanical prophylaxis, or no prophylaxis. The incidence of DVT in LDH patients was 2 percent (1 of 44 patients) and in SCD patients, 12.5 percent (4 of 32 patients). Two of 64 patients assigned to the no-prophylaxis group developed DVT (3 percent).

Dennis, Menawat, Von Thron, et al. (1993) attempted a randomized study on prophylaxis vs. no prophylaxis. Patients received prophylaxis by LDH or SCD. However, many patients were switched from their initial random assignment to the no-prophylaxis group to the SCD group at the discretion of the attending surgeon. Therefore, three nonrandom groups were created: LDH (92 patients), SCD (189 patients), and no prophylaxis (114 patients). DVT screening was done by venous Doppler or Duplex ultrasonography every 5 days. The incidence of DVT was 3 percent for LDH and SCD patients and 9 percent for the no-prophylaxis group.

Headrick, Barker, and Pate (1997) prospectively evaluated a cohort of 228 trauma patients who required bed rest of more than 3 days or had a lower extremity, pelvic, or spinal fracture with paralysis. In an nonrandom manner, patients received LDH (20 patients), SCD (130), both (54), or none (24). They were screened by venous Duplex ultrasonography on a weekly basis. The incidence of DVT for the four groups was 25 percent (5 patients), 14 percent (18 patients), 19 percent (10 patients), and 25 percent (6 patients), respectively.

Table 41. Rates of DVT reported in four studies (two (more...)

Table 41. Rates of DVT reported in four studies (two RCT and two non-RCT) of trauma patients, comparing LDH to mechanical prophylaxis

Study (first author-year)	Number of Patients with LDH	Number of Patients with MP	DVT Incidence In LDH Patients	DVT Incidence In MP Patients	Odds Ratio	95% CI of Odds Ratio
Knudson-1992	37	76	4%	7%	1.3	0.3, 5.6
Knudson-1994	44	32	4%	12%	0.2	0.0, 1.5
Dennis- 1993	92	189	3%	3%	1.2	0.3, 5.3
Headrick-1997	20	130	25%	14%	2.1	0.7, 6.4
Random-effects estimates	193	427	0.049	0.080	1.161	0.495, 2.723
Heterogeneity chi-squared test p values			0.100	0.002	0.267

DVT, deep venous thrombosis; RCT, randomized controlled trial; LDH, low-dose heparin; MP, mechanical prophylaxis.

Figure 7. Shrinkage plot of RCTs and non-RCTs comparing (more...)

   Figure 7. Shrinkage plot of RCTs and non-RCTs comparing the rates of DVT in patients who received LDH or mechanical prophylaxis

Cumulatively, the meta-analysis (Table 41, Figure 7, Evidence Table 5) shows no difference in the incidence of DVT between patients receiving LDH and those receiving mechanical prophylaxis (OR: 1.161, 95 percent CI: 0.495, 2.723). The chi-squared test is not significant for heterogeneity among studies (p=0.267).

Mechanical Prophylaxis vs. No Prophylaxis

In addition to the three RCTs described above in the initial meta-analysis of RCTs only, we found two more non-RCT studies comparing mechanical prophylaxis with no prophylaxis. Both of these studies (Dennis, Menawat, Von Thron, et al., 1993; Headrick, Barker, and Pate, 1997) were described in the LDH vs. mechanical prophylaxis comparison above. These two studies had LDH, SCD, and no-prophylaxis groups. The two latter groups from each study were included in this analysis.

Table 42. Rates of DVT reported in five studies (three (more...)

Table 42. Rates of DVT reported in five studies (three RCT and two non-RCT) of trauma patients, comparing mechanical prophylaxis to no prophylaxis

Study (first author-year)	Number of Patients with MP	Number of Patients with NO PROPH	DVT Incidence In MP Patients	DVT Incidence In NO PROPH Patients	Odds Ratio	95% CI of Odds Ratio
Fisher-1995	35	38	3%	8%	0.3	0.0, 3.5
Knudson-1994	32	64	12%	3%	4.4	0.8, 25.6
Knudson-1994	26	39	0%	13%	0.1	0.0, 2.2
Headrick-1997	130	24	14%	25%	0.5	0.2, 1.4
Dennis-1993	189	114	3%	9%	0.3	0.1, 0.8
Random-effects estimates	412	279	0.054	0.086	0.527	0.190, 1.460
Heterogeneity chi-squared test p values			0.003	0.063	0.092

DVT, deep venous thrombosis; RCT, randomized controlled trial; MP, mechanical prophylaxis; NO PROPH, no prophylaxis; CI, confidence interval.

Figure 8. Shrinkage plot of RCTs and non-RCTs comparing (more...)

   Figure 8. Shrinkage plot of RCTs and non-RCTs comparing the rates of DVT in patients who received mechanical prophylaxis or no prophylaxis

The meta-analysis of the pooled data from these four studies (Table 42, Figure 8, Evidence Table 6) shows no difference in the incidence of DVT between patients using SCD and those not receiving any DVT prophylaxis (OR: 0.527, 95 percent CI: 0.190, 1.460). The chi-squared test is not significant for heterogeneity among studies (p=0.092).

LDH vs. No Prophylaxis

In this analysis, we included the four RCTs described in the previous section of meta-analysis dealing with RCTs only and three additional non-RCTs. Two of these non-RCTs are described above (Dennis, Menawat, Von Thron, et al., 1993; Headrick, Barker, and Pate, 1997). Of the three nonrandomized groups of patients found in these studies (LDH, SCD, no prophylaxis), the LDH and no prophylaxis groups were used for this analysis.

In the third non-RCT, Ruiz, Hill, and Berry (1991) prospectively evaluated 100 multiple trauma patients with an Injury Severity Score greater than or equal to 10 and followed up these patients by venous Duplex ultrasonography on days 1, 3, 6, 10, and 21 or until discharge. At the discretion of the surgeon, 50 patients received LDH and 14 of them developed DVT (28 percent); 50 patients received no prophylaxis and one of them developed DVT (2 percent). Obviously, the patients who were selected to receive prophylaxis were at greater risk for the development of DVT.

Table 43. Rates of DVT reported in seven studies (four (more...)

Table 43. Rates of DVT reported in seven studies (four RCT and three non-RCT) of trauma patients, comparing LDH to no prophylaxis

Study (first author-year)	Number of Patients with LDH	Number of Patients with NO PROPH	DVT Incidence In LDH Patients	DVT Incidence In NO PROPH Patients	Odds Ratio	95% CI of Odds Ratio
Merli-1988	16	17	50%	47%	1.1	0.3, 4.4
Frisbie-1981	15	17	7%	6%	1.1	0.1, 20.0
Knudson-1994	44	64	2%	3%	0.7	0.1, 8.2
Knudson-1994	19	27	5%	7%	0.7	0.1, 8.3
Dennis-1993	281	114	2%	9%	0.2	0.1, 0.6
Headrick-1997	20	24	25%	25%	1.0	0.3, 3.9
Ruiz-1991	50	50	28%	2%	19.1	2.4, 151.6
Random-effects estimates	445	313	0.116	0.084	1.033	0.360, 2.965
Heterogeneity chi-squared test p values			0.000	0.001	0.020

DVT, deep venous thrombosis; RCT, randomized controlled trial; LDH, low-dose heparin; NO PROPH, no prophylaxis; CI, confidence interval.

Figure 9. Shrinkage plot of RCTs and non-RCTs, comparing rates (more...)

   Figure 9. Shrinkage plot of RCTs and non-RCTs, comparing rates of DVT in patients who received LDH or no prophylaxis

This meta-analysis (Table 43, Figure 9, Evidence Table 7) shows that there is no difference in the incidence of DVT between patients receiving LDH and those receiving no prophylaxis (OR: 1.033, 95 percent CI: 0.360, 2.965). The chi-squared test for heterogeneity shows that these studies are heterogeneous (p=0.020).

Publication Bias

We produced funnel plots for all four comparisons described in the RCT and non-RCT meta-analysis, which indicated that there may have been publication bias in comparisons of LDH versus no prophylaxis. However, reliable conclusions cannot be drawn because of the limited number of studies. These funnel plots are not reported here.

Question 2: What Groups of Patients Are at High Risk of Developing VT?

We considered all studies including risk factors for analysis, regardless of study design. Different authors reported numerous risk factors, but we only analyzed risk factors reported in at least three studies.

We treated the risk factors as either dichotomous or continuous variables according to the data provided. For instance, if three or more studies provided data on VT incidence for patients who were younger or older than 55 years old, then the risk factor of "age >55" was considered a dichotomous value. Other studies provided only the age (mean and standard deviation) of patients with and without DVT but did not use a specific age cutoff point. We combined these data and examined the risk factor "age" as a continuous variable.

Risk Factors as Dichotomous Variables

We analyzed the following variables:

• Gender (four studies, Evidence Table 8).

• Head injury (eight studies, Evidence Table 9).

• Long-bone fractures (12 studies, Evidence Table 10).

• Pelvic fractures (eight studies, Evidence Table 11).

• Spinal fractures (10 studies, Evidence Table 12).

• Spinal-cord injuries (five studies, Evidence Table 13).

Table 44. Rates of DVT reported in four studies comparing (more...)

Table 44. Rates of DVT reported in four studies comparing male vs. female patients (using gender as a risk factor)

Study (first author-year)	Number of Male Patients	Number of Female Patients	DVT Incidence In Male Patients	DVT Incidence In Female Patients	Odds Ratio	95% CI of Odds Ratio
Waring-1991	1,145	274	14%	10%	1.5	1.0, 2.3
Spannagel-1993	146	107	12%	8%	1.5	0.7, 3.6
Knudson-1994	200	51	5%	10%	0.5	0.2, 1.5
Abelseth-1996	70	32	24%	37%	0.5	0.2, 1.3
Random-effects estimates	1,561	464	0.127	0.123	0.983	0.544, 1.744
Heterogeneity chi-square test p values			0.000	0.014	0.076

DVT, deep venous thrombosis; CI, confidence interval.

Table 45. Rates of DVT reported in eight studies comparing (more...)

Table 45. Rates of DVT reported in eight studies comparing patients with head injury vs. patients without head injury (using head injury as a risk factor)

Study (first author-year)	Number of Patients with HI	Number of Patients without HI	DVT Incidence In Patients with HI	DVT Incidence In Patients without HI	Odds Ratio	95% CI of Odds Ratio
Kudsk-1989	9	29	44%	69%	0.4	0.1, 1.7
Knudson-1992	36	77	11%	10%	1.1	0.3, 3.8
Dennis-1993	92	303	4%	5%	0.9	0.3, 2.9
Knudson-1994	39	212	5%	6%	0.8	0.2, 3.8
Meyer-1995	6	177	17%	12%	1.5	0.2, 13.3
Piotrowski-1996	115	228	9%	4%	2.1	0.8, 5.1
Spain-1997	131	149	4%	5%	0.8	0.2, 2.6
Velmahos-1998	52	148	12%	14%	0.8	0.3, 2.2
Random-effects estimates	480	1,323	0.068	0.107	1.019	0.664, 1.564
Heterogeneity chi-squared test p values			0.108	0.000	0.701

DVT, deep venous thrombosis; HI, head injury; CI, confidence interval.

Table 46. Rates of DVT reported in 12 studies comparing (more...)

Table 46. Rates of DVT reported in 12 studies comparing patients with long-bone fractures vs. patients without long-bone fractures (using long-bone fractures as a risk factor)

Study (first author-year)	Number of Patients with LBF	Number of Patients without LBF	DVT Incidence In Patients with LBF	DVT Incidence In Patients without LBF	Odds Ratio	95% CI of Odds Ratio
Kudsk-1989	13	25	54%	68%	0.5	0.1, 2.2
Knudson-1992	38	75	13%	9%	1.5	0.4, 5.0
Hill-1994	70	30	13%	20%	0.6	0.2, 1.8
Geerts-1994	74	275	80%	52%	3.7	2.0, 6.8
Knudson-1995	65	186	9%	5%	2.0	0.7, 5.9
Napolitano-1995	212	230	7%	11%	0.6	0.3, 1.2
Abelseth-1996	20	82	40%	26%	1.9	0.7, 5.4
Geerts-1996	168	97	44%	27%	2.1	1.2, 3.7
Knudson-1996	101	271	1%	3%	0.3	0.0, 2.7
Piotrowski-1996	114	229	3%	7%	0.3	0.1, 1.2
Spain-1997	126	154	2%	6%	0.2	0.0, 1.1
Velmahos-1998	46	154	13%	13%	1.0	0.4, 2.7
Random-effects estimates	1,047	1,808	0.2808	0.189	1.034	0.622, 1.717
Heterogeneity chi-squared test p value			0.000	0.000	0.000

DVT, deep venous thrombosis; LBF, long-bone fracture; CI, confidence interval.

Table 47. Rates of DVT reported in eight studies comparing (more...)

Table 47. Rates of DVT reported in eight studies comparing patients with pelvic fractures vs. patients without pelvic fractures (using pelvic fractures as a risk factor)

Study (first author-year)	Number of Patients with PF	Number of Patients without PF	DVT Incidence In Patients with PF	DVT Incidence In Patients without PF	Odds Ratio	95% CI of Odds Ratio
Kudsk-1989	8	30	63	63	1.0	0.2, 4.8
Knudson-1992	109	171	2%	6%	0.3	0.1, 1.4
Geerts-1994	100	249	61%	56%	1.2	0.8, 2.0
Knudson-1994	43	208	9%	5%	1.8	0.6, 6.1
Napolitano-1995	52	390	6%	9%	0.6	0.2, 2.0
Knudson-1996	22	350	0%	3%	0.8	0.0, 14.2
Piotrowski-1996	74	269	5%	6%	0.9	0.7, 4.9
Velmahos-1998	31	169	19%	12%	1.8	0.7, 4.9
Random-effects estimates	439	1,836	0.171	0.177	1.109	0.788, 1.562
Heterogeneity chi-squared test p values			0.000	0.000	0.570

DVT, deep venous thrombosis; PF, pelvic fracture; CI, confidence interval.

Table 48. Rates of DVT reported in 10 studies comparing (more...)

Table 48. Rates of DVT reported in 10 studies comparing patients with spinal fractures vs. patients without spinal fracture (using spinal fracture as a risk factor)

Study (first author-year)	Number of Patients with SF	Number of Patients without SF	DVT Incidence In Patients with SF	DVT Incidence In Patients without SF	Odds Ratio	95% CI of Odds Ratio
Kudsk-1989	5	33	60%	64%	0.9	0.1, 5.9
Dennis-1993	50	345	14%	3%	4.9	1.8, 13.4
Geerts-1994	26	323	81%	56%	3.3	1.2, 9.1
Knudson-1994	17	234	12%	6%	2.3	0.5, 11.0
Meyer-1995	6	177	17%	12%	1.5	0.2, 13.3
Napolitano-1995	24	418	25%	8%	3.8	1.4, 10.1
Knudson-1996	41	331	5%	2%	2.4	0.5, 11.8
Piotrowski-1996	98	245	11%	4%	3.3	1.3, 8.3
Spain-1997	62	218	3%	5%	0.7	0.1, 3.3
Velmahos-1998	41	159	10%	14%	0.7	0.2, 2.1
Random-effects estimates	370	2,483	0.204	0.150	2.260	1.415, 3.610
Heterogeneity chi-squared test p values			0.00	0.000	0.193

DVT, deep venous thrombosis; SF, spinal fracture; CI, confidence interval.

Table 49. Rates of DVT reported in five studies comparing (more...)

Table 49. Rates of DVT reported in five studies comparing patients with spinal-cord injury vs. patients without spinal-cord injury (using spinal-cord injury as a risk factor)

Study (first author-year)	Number of Patients with SCI	Number of Patients without SCI	DVT Incidence In Patients with SCI	DVT Incidence In Patients without SCI	Odds Ratio	95% CI of Odds Ratio
Kudsk-1989	5	33	60%	64%	0.9	0.1, 5.9
Geerts-1994	26	323	81%	56%	3.3	1.2, 9.1
Napolitano-1995	24	418	25%	8%	3.8	1.4, 10.1
Knudson-1996	25	347	8%	2%	4.2	0.8, 21.5
Piotrowski-1996	43	237	9%	3%	2.9	0.8, 10.2
Random-effects estimates	123	1,358	0.347	0.244	3.107	1.794, 5.381
Heterogeneity chi-squared test p values			0.000	0.000	0.730

DVT, deep venous thrombosis; SCI, spinal-cord injury; CI, confidence interval.

Figure 10. Shrinkage and funnel plots of four studies (more...)

   Figure 10. Shrinkage and funnel plots of four studies comparing rates of DVT between male and female patients (using gender as a risk factor)

Figure 11. Shrinkage and funnel plots of eight (more...)

   Figure 11. Shrinkage and funnel plots of eight studies comparing rates of DVT between patients with and patients without head injuries (using head injuries as a risk factor)

Figure 12. Shrinkage and funnel plots of 12 (more...)

   Figure 12. Shrinkage and funnel plots of 12 studies comparing rates of DVT between patients with and patients without long-bone fractures (using long-bone fractures as a risk factor)

Figure 13. Shrinkage and funnel plots of eight (more...)

   Figure 13. Shrinkage and funnel plots of eight studies comparing rates of DVT between patients with and patients without pelvic fractures (using pelvic fractures as a risk factor)

Figure 14. Shrinkage and funnel plots of 10 (more...)

   Figure 14. Shrinkage and funnel plots of 10 studies comparing rates of DVT between patients with and patients without spinal fractures (using spinal fractures as a risk factor)

Figure 15. Shrinkage and funnel plots of five studies (more...)

   Figure 15. Shrinkage and funnel plots of five studies comparing rates of DVT between patients with and patients without spinal-cord injuries (using spinal cord injuries as a risk factor)

A number of studies included age as a risk factor, but the different cutoff points used in each study (age >30, 40, 50, 55, etc.) did not allow analysis of this variable. For the above six risk factors, Tables 44 to 49 report the DVT rates and Figures 10 to 15 provide the graphic representation of the meta-analysis of these studies with the corresponding funnel plots. The six corresponding evidence tables (Evidence Tables 8 to 13) describe these studies. The only risk factors found to place the patient at higher risk for development of DVT are spinal fracture (OR: 2.260, 95 percent CI: 1.415, 3.610) and spinal-cord injury (OR: 3.107, 95 percent CI: 1.794, 5.381). The chi-squared heterogeneity test indicates that only the studies used to evaluate long-bone fractures are heterogeneous (p=0.000). For all other comparisons, the chi-squared test is not significant for heterogeneity among studies.

Risk Factors as Continuous Variables

We examined three continuous variables: age, Injury Severity Score (ISS), and units of blood transfused.

Age

Figure 16. Shrinkage and funnel plots of seven studies comparing (more...)

   Figure 16. Shrinkage and funnel plots of seven studies comparing patients with and patients without DVT with regard to patient age

We examined the mean age in seven studies. The meta-analysis of the pooled data (Figure 16) shows that patients with DVT are significantly older than those without DVT by an average of 8.133±1.504 years (95 percent CI: 5.115, 11.141). The corresponding funnel plot does not show evidence of publication bias. The seven studies are described in Evidence Table 14. The Q-statistic of the heterogeneity test shows that the studies are not heterogeneous (p=0.323).

ISS (Injury Severity Score)

Figure 17. Shrinkage and funnel plots of six studies comparing (more...)

   Figure 17. Shrinkage and funnel plots of six studies comparing patients with and patients without DVT with regard to Injury Severity Score

Similarly, the meta-analysis of the data from six studies on ISS (Figure 17) shows that patients with DVT have a significantly higher ISS than patients without DVT by 1.430±0.747 (95 percent CI: 0.000, 2.924). Although this difference isstatistically significant, it does not carry clinical significance. The corresponding funnel plot shows possible publication bias. The studies are described in Evidence Table 15. The Q-statistic of the heterogeneity test shows that the studies are not heterogeneous (p=0.843).

Blood Transfusion

Figure 18. Shrinkage and funnel plots of three studies comparing (more...)

   Figure 18. Shrinkage and funnel plots of three studies comparing patients with and patients without DVT with regard to units of blood transfused

The difference in the amount of blood transfused for patients with or without DVT, as shown by the meta-analysis of the data from three studies (Figure 18), is not statistically significant (mean ± standard deviation [SD]: 1.882±2.815 units of blood, 95 percent CI: -3.637, 7.401). The funnel plot shows no publication bias, but the sample size is small. The studies are described in Evidence Table 16. The Q-statistic for the test of heterogeneity shows that the studies are not heterogeneous (p=0.935).

Question 3: What Is the Best Method of Screening for VT?

None of the studies compared Duplex ultrasonography against venography in all patients included in the study. The technical experts determined that the real question of interest is whether bedside ultrasonography can substitute for venography to reliably detect DVT in patients with severe trauma who are clinically unevaluable or asymptomatic. Although there are studies comparing these methods in other types of patients, the trauma literature provides no relevant evidence. For this reason, Question 3 cannot be answered in this Evidence Report.

Question 4: What Is the Role of VCFs in Preventing PE?

The evidence that we identified on VCFs derived entirely from nonrandomized, uncontrolled trials. Although many studies on VCFs have been published that include trauma patients along with other types of patients, we could not isolate the data referring to the trauma patients alone. For this reason, we restricted our literature analysis for this issue to studies that included only trauma patients.

The study designs frequently included historical controls and presented multiple outcomes, including PE, fatal PE, DVT, VCF-insertion-site DVT, and VCF-related complications. Different periods of time were used for followup. For these reasons, comparison of these studies was very difficult, and our results should be considered as generating, rather than proving, hypotheses related to the use of VCF.

Table 50. Incidence of pulmonary embolism in patients (more...)

Table 50. Incidence of pulmonary embolism in patients with and without vena cava filters

Study (first author, year)	Total Patients (N)	VCF (N)	Group (# PE)	Prosp (N)	No VCF (# PE)	Hist (N)	No VCF (# PE)
239Gosin 1997	499	99	0	151	4	249	12
393Rogers 1997	2,090	35	1	905	1	1,150	11
605Khansarinia 1995	324	108	0	-	-	216	13
821Rodriguez 1996	120	40	1	-	-	80	14
973Wilson 1994	126	15	0	-	-	111	7
1,146Webb 1992	51	24	0	27	2	-	-
Random-effects estimates		0.002		0.015		0.058
95% CI		(-0.007, 0.010)		(-0.011, 0.041)		(0.020, 0.096)
Heterogeneity chi-squared tests Q-statistic p value		1.93 0.86		5.83 0.054		35.76 0.000
TOTAL	3,210	321	2	1,083	7	1,806	57

VCF, vena cava filter; Prosp, prospective control group; Hist, historical control group; PE, pulmonary embolism; CI, confidence interval. Note: The treatment (VCF) group in each study was compared with a prospective control group, a historical control group, or both.

Table 51. Incidence of fatal pulmonary embolism in (more...)

Table 51. Incidence of fatal pulmonary embolism in patients with and without vena cava filters

Study (first author, year)	Total Patients	VCF Group		Prosp No VCF		Hist No VCF
	(N)	(N)	(# fatal PE)	(N)	(# fatal PE)	(N)	(# fatal PE)
393Rogers 1997	2,090	35	0	905	0	1,150	4
605Khansarinia 1995	324	108	0	-	-	216	9
821Rodriguez 1996	120	40	1	-	-	80	8
973 Wilson 1994	126	15	0	-	-	111	3
1,146Webb 1992	51	24	0	27	1	-	-
Random-effects estimates				0.001		0.033
95% CI				(-0.009, 0.011)		(0.002, 0.064)
Heterogeneity chi-squared tests Q-statistic p value				1.04 0.308		18.06 <0.0001
TOTAL	3,711	222	0	932	1	1,557	24

VCF, vena cava filter; Prosp, prospective control group; Hist, historical control group; PE, pulmonary embolism; CI, confidence interval. Note: The treatment (VCF) group in each study was compared with a prospective control group, a historical control group, or both.

Analysis of Complications of Methods of VT Prevention

Bleeding and thrombocytopenia are two complications frequently reported in the literature as being associated with the administration of heparin (LDH or LMWH) for VT prophylaxis. There were not enough data to perform meta-analysis on these two outcomes. Our calculation of the incidence of these variables is confounded by the different drugs and doses used, the lack of standard definitions of the outcomes, and the differences in study designs.

Table 52. Incidence of drug-related bleeding in studies (more...)

Table 52. Incidence of drug-related bleeding in studies describing trauma patients receiving different drugs for prophylaxis against venous thromboembolism

Method of Prophylaxis	Author, Year	Number in Group	N with Bleeding
LDH	Geerts, 1996	136	1
	Knudson, 1992	37	1
	Green, 1990	21	2
	Shackford, 1990	71	6
	Green, 1998	58	7
	Hachen, 1983	318	11
	Casas, 1977	21	0
	Total	662	28 (4.2%)

Method of Prophylaxis	Author, Year	Number in Group	N with Bleeding
LMWH	Green, 1994	48	1
	Haentjens, 1996	283	10
	Harris, 1996	105	3
	Geerts, 1996	129	5
	Green, 1990	20	0
	Total	585	19 (3.2%)

Method of Prophylaxis	Author, Year	Number in Group	N with Bleeding
Ancrod	Cole, 1995	30	0
Aspirin + Dipyridamole	Green, 1982	12	1
LDH+ Aspirin or Coumadin	Spain, 1997	84	3

Random-Effects Estimate: 0.036, 95% CI: (0.010, 0.061)

Random-Effects Estimate: 0.031, 95% CI: (0.017, 0.045)

LDH, low-dose heparin; LMWH, low-molecular-weight heparin; CI, confidence interval.

Table 53. Incidence of heparin-related thrombocytopenia (more...)

Table 53. Incidence of heparin-related thrombocytopenia in studies describing trauma patients receiving LDH or LMWH for prophylaxis against venous thromboembolism

Method of Prophylaxis	Author, Year	Number in Group	N with Thr/cyt
LDH	Knudson, 1992	37	0
	Hachen, 1983	318	10
	Geerts, 1996	136	2
	Total	491	12

Method of Prophylaxis	Author, Year	Number in Group	N with Thr/cyt
LMWH	Haentjens, 1996	283	2
	Geerts, 1996	129	0
	Total	412	2

Random-Effects Estimate: 0.019, 95% CI: 0.004, 0.035

Random-effects estimate: 0.004, 95% CI: -0.003, 0.011.

Chi-squared heterogeneity test: Q-statistic: 0.92, p value: 0.337.

LDH, low-dose heparin; LMWH, low-molecular-weight heparin;

Thr/cyt, thrombocytopenia; CI, confidence interval.

Length of Hospital Stay

Figure 19. Shrinkage and funnel plots of five studies (more...)

   Figure 19. Shrinkage and funnel plots of five studies comparing patients with DVT and patients without DVT for length of hospital stay

It is not clear whether length of stay (LOS) is a risk factor for, or a consequence of, DVT; therefore, it is not reported as a risk factor. A meta-analysis of five studies reporting on the hospital LOS between patients with and without DVT showed that this parameter was statistically different in the two populations (Figure 19). Patients with DVT stayed14.6±2.1 days longer (95 percent CI: 10.4, 18.8) in the hospital than patients without DVT.

Cost-Effectiveness Analysis

Figure 20. Cost of VT prophylaxis with LDH vs. no prophylaxis (more...)

   Figure 20. Cost of VT prophylaxis with LDH vs. no prophylaxis

Figure 21. Cost of VT prophylaxis with mechanical (more...)

   Figure 21. Cost of VT prophylaxis with mechanical vs. no prophylaxis

Figure 22. Cost of VT prophylaxis with LMWH vs. no prophylaxis (more...)

   Figure 22. Cost of VT prophylaxis with LMWH vs. no prophylaxis

20

22

Table 54. Cost per death avoided with different probabilities (more...)

Table 54. Cost per death avoided with different probabilities of DVT

Probability of DVT	Sequential Compression Device	Low-Dose Unfractionated Heparin	Low-Molecular-Weight Heparin
0.120	$∞	$∞	$∞
0.107	$1,465,201	$170,940	$3,101,343
0.100	$952,381	$103,175	$2,007,937
0.093	$699,588	$70,547	$1,481,481
0.080	$464,286	$39,683	$996,032
0.067	$347,409	$26,954	$745,732
0.053	$270,078	$16,584	$585,169
0.040	$222,222	$9,921	$488,095
0.027	$189,452	$6,827	$416,453
0.013	$161,697	$2,967	$358,997
0.000	$141,534	$0	$318,783

DVT, deep venous thrombosis.

Table 55. Life expectancy and discounted years of life (more...)

Table 55. Life expectancy and discounted years of life from 1996

Age	Life Expectancy	Discounted Years of Life
20 to 25	57	28.0
25 to 30	52	26.9
30 to 35	48	26.0
35 to 40	43	24.7
40 to 45	38	23.2
45 to 50	34	21.8
50 to 55	30	20.2
55 to 60	25	17.9
60 to 65	21	15.9
65 to 70	18	14.2
70 to 75	14	11.6
75 to 80	11	9.5
80 to 85	8	7.2
85 & over	6	5.6

If the trauma patient receiving prophylaxis with LDH were 60 years old, the cost per life saved ($103,175) would be divided by 15.9, resulting in an estimated cost per life-year of $6,489. Again, prophylaxis with LDH would be cost-effective in a 60-year-old trauma patient. Obviously, prophylaxis of trauma patients becomes less cost-effective with older age because of reduced life expectancy.

What Is the Incidence of DVT and PE?

What Is the Best Method of Preventing VT?

Firm conclusions as to the best method of preventing VT cannot be made based on the available literature data. The randomized clinical trials that have been published are usually unique (i.e., only one type of comparison has ever been published), frequently of low methodologic quality, and clinically heterogeneous. We assessed both clinical trial data and observational data, but our results remained consistent. We were unable to demonstrate any benefit of one method over another in preventing VT. However, we note that the confidence intervals in our pooled results are wide, in some instances very wide, and we therefore cannot exclude a clinically important effect. For example, many of our odds ratios have confidence intervals that include 2, which for these low-frequency events approximates the risk ratio. A prophylactic method with a risk ratio of 2 in its favor would be associated with one-half as many episodes of DVT or PE-a clinically important result. The existing limited data do not allow us to exclude a benefit of this magnitude for most of these prophylactic methods.

We compared three methods in our primary meta-analysis: LDH, SCDs, and no prophylaxis. The outcome parameter was DVT. There was no difference in DVT rates after administration of LDH or application of SCD. Even more surprising, there was no difference in DVT rates between LDH or SCD and no prophylaxis. This may be because of the small number of patients included in each study, the methodologic flaws of the studies, or the real absence of a difference. We compared a fourth method (LMWH) against LDH in a combined RCT/non-RCT meta-analysis with PE as the outcome parameter. This comparison showed no difference in PE prevention.

LMWH is a new and promising pharmacologic agent for VT prophylaxis. We hoped that we could compare its safety and efficacy against other methods of prophylaxis by meta-analysis. Unfortunately, we identified only two RCTs (Geerts, Jay, Code, et al., 1996; Knudson, Morabito, Paiement, et al., 1996) comparing LMWH against LDH or SCD. Both studies are of high quality and show an outcome benefit in favor of LMWH. However, they are heterogeneous: One study (Geerts, Jay, Code, et al., 1996) reported DVT rates of 31 percent and 44 percent according to the use of LMWH or LDH, respectively, whereas the other (Knudson, Morabito, Paiement, et al., 1996) reported DVT rates of 0.8 percent and 2.4 percent when LMWH or SCD are used respectively. The sample sizes were too small to make any conclusion on comparisons on these drugs' safety. These widely different DVT rates between the two studies could be because of different study methodologies (particularly in the methods of DVT diagnosis) or different populations examined, which may make the interpretation of their results difficult. In the absence of at least three RCTs, meta-analysis could not be performed on this topic.

Our analysis of the limited reliable literature data leads us to conclude that LDH or SCD offer no proven advantage over each other or over no prophylaxis for the prevention of DVT after trauma.

What Are the Risk Factors for VT?

Multiple risk factors have been reported for the development of VT. Our analysis focused on risk factors included in at least three studies. The risk factors were examined as continuous or categorical (dichotomous) values according to how they were reported in their respective articles.

Of the categorical risk factors examined (gender, head injuries, spinal fractures, spinal cord injuries, long-bone fractures, and pelvic fractures), only spinal fractures and spinal cord injuries were found on pooled analysis to affect the incidence of VT. The presence of spinal fractures or spinal cord injures increases the odds for development of DVT twofold and threefold, respectively, relative to patients without spinal trauma. We were unable to confirm that widely published risk factors, such as pelvic fractures, long-bone fractures, or head injuries, affect the incidence of VT. A possible explanation for this outcome is that the studies we analyzed included multiple trauma patients who were already at the highest risk of VT. In such patients, individual risk factors may not increase an already high risk any further.

Of the continuous variables we examined (age, Injury Severity Score [ISS], blood transfusion), age and ISS were significantly different between patients with and without VT. Patients with DVT were on average 9 years older and had an ISS that was 1.5 points higher than that of patients without DVT. The ISS difference, however, has limited clinical significance. There were insufficient data to identify threshold figures of age or ISS above which VT rates increased significantly.

We conclude that the presence of spinal fractures or spinal cord injuries is a significant risk factor for the development of VT. The likelihood of developing VT increases with age and ISS, but the threshold at which the rate of increase changes significantly cannot be determined.

What Is the Role of Vena Cava Filters in Preventing PE?

The data we analyzed on the role of vena cava filters (VCFs) were derived from methodologically poor studies: no RCT in the trauma literature addressed the role of VCF. From the existing literature, it seems that VCF decreases the risk of PE and fatal PE in severely injured patients. The absence of severe complications associated with VCF use in the studies that we examined suggests that this device is safe. However, firm conclusions on its short-term and long-term safety cannot be made from the available data.

Most authorities would agree that a mechanical interruption to the flow of blood clots from the systemic veins to the pulmonary circulation is an effective method for preventing this complication. However, defining the appropriate trauma population for VCF placement remains difficult. The physician needs to balance the risk of PE with the risk of placing a device that will remain in the bloodstream for life in a (typically) young trauma patient. We cannot draw conclusions as to the role of VCF in the prevention of VT after trauma on the basis of the existing data.

Cost-Effectiveness

Because the meta-analysis showed no differences in VT rates in patients treated by different methods of prophylaxis vs. no prophylaxis, the cost-effectiveness of prophylaxis cannot be established. All methods of prophylaxis will be associated with the cost of the drug or device used and the possible complications of its use. Therefore, because there is no proof that the risk of VT is not decreased by any method of prophylaxis, no prophylaxis will be more cost-effective than any other approach. However, the small sample sizes used in our meta-analysis and the wide 95 percent confidence intervals of most calculated odds ratios indicate that differences among different methods of prophylaxis may reveal themselves if appropriate numbers of patients are evaluated. For this reason, we proceeded to estimate the cost per year-life saved according to different probabilities of DVT occurrence. Our results showed that prophylaxis for DVT may be cost-effective (using accepted limitations) at rates of efficacy within the 95% confidence intervals of our estimates.

In this analysis, we estimated the cost of each of the three most commonly used methods of prophylaxis (LDH, LMWH, mechanical) that would save 1 year of a person's life over the estimated life span according to his/her age by decreasing the incidence of DVT (and eventual PE). This information will be very useful when future research on this topic attempts to use power analysis to identify the sample sizes required to prove that a certain method is cost-effective.

The figure of approximately $50,000 per life-year saved is used as the upper limit to determine cost-effectiveness for each method or drug investigated. The entire cost of a method used is divided by the estimated years that a person may live, yielding the cost per life-year saved. Therefore, the cost per life-year saved will be lower for a young person than for an older person for a therapy that has the same cost (entire cost of therapy will be divided by more years of projected life). So, a method becomes more cost-effective the younger a person is.

Because LDH is identified as the least expensive method for VT prevention of the three examined, it will be more cost-effective than the other methods because no method is superior in preventing VT. If future studies show that another method of prophylaxis is more effective than LDH, the cost-effectiveness results will change. We believe that the information provided in this analysis will help design future studies on this topic.

Limitations

The primary limitation of the current evidence report is the quantity and quality of original studies. Only a few RCTs were identified, and approximately one-half of thesecould not be combined for meta-analysis. Most of our meta-analyses addressing methods of prophylaxis used only three or four studies with a low total number of patients. For this reason, we added non-RCTs to RCTs in order to increase the sample size. Although such a meta-analytic design is inherently weaker, it was consistent with the results of the meta-analysis of RCTs. The absence of a proven benefit for different methods of prophylaxis of VT was found after meta-analysis of RCTs as well as RCTs and non-RCTs together.

The majority of controlled or noncontrolled trials scored one-half or lower of the highest possible score on our quality scoring scale. Because we had so few RCTs for meta-analysis, we could not exclude studies that were of low quality. We also made no attempt to give greater importance to studies that had better design and, therefore, presumably more valid results. This was because, in general, there is a lack of empirical evidence relating study design to bias.

The way we defined "trauma patients" could be viewed as an additional limitation. After careful consideration by the technical expert panel, the definition excluded burn patients and elderly patients with minimal mechanism of injury. However, we believe that this definition limits this report to the population intended to be studied.

Finally, the heterogeneity of studies, as shown by the respective statistical tests, may have influenced some of the results. Surprisingly, the tests did not indicate significant heterogeneity for most comparisons. However, the trauma populations that were combined to perform meta-analysis (e.g., spinal-cord injuries with general trauma) were not always similar. These groups of patients have individual characteristics and risk factors for VT. Combining them could be inappropriate in some cases from a pathophysiologic point of view. Again, the limited number of available studies provided little opportunity to isolate these groups for individual study.

Methods of Screening

The best method of screening is probably the most difficult issue to resolve. A study comparing the most convenient method for screening (Duplex ultrasonography) with the current gold standard for diagnosis of DVT (ascending venography) would yield valuable conclusions. Results from such a study could make physicians aware of the limitations of Duplex ultrasound in critically ill patients and alert them to the possibility of missing a DVT diagnosis, with catastrophic consequences, resulting from reliance on Duplex results. Alternatively, such results could indicate that Duplex is sensitive and specific, even for edematous, severely injured patients. Even with its limitations, Duplex ultrasonography may still remain the screening method of choice in most centers because it is convenient, noninvasive, inexpensive, and repeatable, and it can be done at the bedside. D-dimers is a promising new test that still needs to be explored.

Other methods of screening such as radiolabeled fibrinogen scanning or impedance plethysmography have not been adopted by most trauma centers and should probably not be studied further. We believe that further studies to identify the best method of screening should not be the highest priority at this time.

Selection of Appropriate Patients for VT Prophylaxis

Identification of the Best Method of Prophylaxis

Our evidence report shows that high-quality data for these two very important topics is scanty, and firm conclusions cannot be drawn, despite the common belief that the answers exist in the literature. Although numerous studies could be designed around these topics, ideally one large multicenter trial should answer the major relevant questions. This study must do the following:

• Involve multiple trauma centers with high volumes of trauma patients and expertise in this particular field.

• Have a randomized, controlled design to help insure equivalence among groups at baseline.

• Define appropriate inclusion criteria for "trauma" patients.

• Compare the most important (and most commonly used) methods of prophylaxis (i.e., LMWH, LDH, and SCDs).

• Adopt a strict screening protocol for DVT, regardless of symptoms, and a flexible protocol for investigating pulmonary embolism in the presence of symptoms.

• Accumulate numbers that can provide statistical significance based on careful power analysis.

• Analyze this large sample of patients to identify risk factors for VT.

Some physicians may consider the inclusion of a no-prophylaxis group to be unethical at this point. However, in this evidence report the limited amount of reliable data generated the conclusion that conventional methods of prophylaxis do not offer any proven advantage over no prophylaxis. A panel of experts should decide if a no-prophylaxis group should be included in the multicenter trial that we propose. Based on the available data, the efficacy of prophylactic therapy in preventing DVT in trauma patients is in sufficient doubt to justify the use of a no-prophylaxis group in clinical trials.

Table 56. Power analysis to estimate the number of (more...)

Table 56. Power analysis to estimate the number of patients required to achieve reduction in DVT rates¹

Test significance level, α	0.050	0.050	0.050	0.050	0.050	0.050
Group 1 proportion, π 1	0.118	0.118	0.118	0.118	0.118	0.118
Group 2 proportion, π 2	0.090	0.090	0.090	0.060	0.060	0.060
Odds ratio, ψ= π 2 (1-π 1)/[ π 1 (1-π 2)]	0.739	0.739	0.739	0.477	0.477	0.477
Power (%)	80	90	95	80	90	95
n per group	1,936	2,567	3,158	411	539	658

1Two-groups continuity-corrected chi-squared test of equal proportions.

DVT, deep venous thrombosis.

We estimated the sample sizes that would be needed for a large, randomized, controlled trial to demonstrate statistically significant reductions in DVT rates (Table 56). The power analysis assumed an -error of 0.05 and a power of 80 percent, 90 percent, or 95 percent. We estimated sample sizes according to two different scenarios:

• 1

  A reduction in the incidence of DVT from 11.8 percent (the general incidence in all trauma patients, as found in our analysis) to 9 percent (the incidence that would make the most expensive method of prophylaxis, LMWH, cost-effective for a person of the average age of the trauma population: 30--35 years old, see Tables 54-55).

• 2

  A reduction in the incidence of DVT from 11.8 percent to 6 percent (a 50 percent reduction, which is a clinically significant endpoint).

According to different power estimates and reduction rates, the sample sizes range from 411 to 3,158 patients per group. For example, to reduce the incidence of DVT with LMWH from 11.8 percent to 9 percent in a statistically significant way, assuming a power of 80 percent, approximately 4,000 patients should be included (2,000 in each group). This sample size alone strongly indicates the necessity of a multicenter trial.

The Role of Vena Cava Filters

The issue of VCFs is extremely important, and further research should be definitely done in this direction. We propose two study methodologies on this topic:

• Including this question in the multicenter trial: Patients should be stratified to those who could receive other methods of prophylaxis and those who had contraindications to other methods or had lifelong risk for VT. These latter patients should be randomized to receive or not receive a VCF.

• Designing a different study, specifically to answer this question: In such a study, all patients who were deemed to be at high risk for DVT according to predefined criteria should be randomized to receive or not receive a VCF, regardless of other methods of prophylaxis given simultaneously.

Both studies should have a predetermined protocol for evaluating PE according to symptoms, an aggressive autopsy policy to examine for PE and its possible association with death, and careful short-term and long-term followup to detect complications related to VCF.