Systematic review flowchart. LMWH indicates low-molecular-weight heparin; RCT, randomized controlled trial; UFH, unfractionated heparin; and VTE, venous thromboembolism.
Inverted funnel plots for trials comparing the effect on mortality (A) and deep venous thrombosis (DVT) (B) of low-molecular-weight heparin and unfractionated heparin for perioperative thromboprophylaxis in patients with cancer. RR indicates relative risk.
Deaths in patients with cancer receiving perioperative thromboprophylaxis with low-molecular-weight heparin (LMWH) vs unfractionated heparin (UFH). CI indicates confidence interval; RR, relative risk.
Deep venous thrombosis (DVT) (any diagnostic strategy) in patients with cancer receiving perioperative thromboprophylaxis with low-molecular-weight heparin (LMWH) vs unfractionated heparin (UFH). CI indicates confidence interval; RR, relative risk.
Akl EA, Terrenato I, Barba M, Sperati F, Sempos EV, Muti P, Cook DJ, Schünemann HJ. Low-Molecular-Weight Heparin vs Unfractionated Heparin for Perioperative Thromboprophylaxis in Patients With CancerA Systematic Review and Meta-analysis. Arch Intern Med. 2008;168(12):1261-1269. doi:10.1001/archinte.168.12.1261
The relative benefits and harms of low-molecular-weight heparin (LMWH) and unfractionated heparin (UFH) are required for judgments regarding the appropriate perioperative thromboprophylaxis in patients with cancer. We systematically reviewed the literature to quantify these effects.
The comprehensive searches included (1) an electronic search of MEDLINE, EMBASE, ISI the Web of Science, and CENTRAL (The Cochrane Central Register of Controlled Trials); (2) a hand search of relevant conference proceedings; (3) a reference check of included trials; and (4) use of the PubMed “Related Articles” feature. Outcomes of interest included mortality, deep venous thrombosis, pulmonary embolism, bleeding complications, and thrombocytopenia.
Of 3986 identified citations, we included 14 randomized clinical trials in the meta-analysis (all using preoperative prophylactic anticoagulation). The overall methodological quality was moderate. The meta-analysis showed no differences in mortality in patients receiving LMWH compared with UFH (relative risk [RR], 0.89; 95% confidence interval [CI], 0.61-1.28) or in clinically suspected deep venous thrombosis (RR, 0.73; 95% CI, 0.23-2.28). In a post hoc analysis including all studies assessing deep venous thrombosis, irrespective of the diagnostic strategy used, LMWH was superior to UFH (RR, 0.72; 95% CI, 0.55-0.94). There were no differences in rates of pulmonary embolism (RR, 0.60; 95% CI, 0.22-1.64), minor bleeding (RR, 0.88; 95% CI, 0.47-1.66), or major bleeding (RR, 0.95; 95% CI, 0.51-1.77).
We found no differences in mortality in patients with cancer receiving perioperative thromboprophylaxis with LMWH vs UFH. Further trials are needed to more carefully evaluate the benefits and harms of different heparin thromboprophylaxis strategies in this population.
Patients with cancer undergoing surgical procedures have a higher risk of venous thromboembolism than patients without cancer undergoing similar procedures.1- 3 Patients with cancer and venous thromboembolism also have a higher risk of death than patients with cancer alone or with venous thromboembolism alone.4,5 Moreover, thromboprophylaxis might be less effective in patients with cancer owing to the prothrombotic state associated with malignant neoplasms.2,6
The American College of Chest Physicians recommends that patients with cancer undergoing surgery receive prophylaxis “that is appropriate for their current risk state,” including the type of surgery.7(p372S) Two systematic reviews8,9 have shown that heparins are superior to no anticoagulation in the prevention of deep venous thrombosis (DVT) and pulmonary embolism (PE) in patients undergoing colorectal and general surgery, respectively. Mismetti et al9 showed that in general surgery, the efficacy and safety of low-molecular-weight heparin (LMWH) relative to unfractionated heparin (UFH) were similar in patients with vs those without cancer. However, estimates of the relative effects of the 2 medications in patients with cancer were not provided. Relative effects for benefits and harms are required to make judgments regarding the appropriate use of anticoagulants in this setting. The objective of this systematic review is to compare the efficacy and safety of LMWH with the efficacy and safety of UFH for perioperative thromboprophylaxis in patients with cancer.
We conducted electronic searches (MEDLINE from 1966, EMBASE from 1980, ISI the Web of Science, and CENTRAL [The Cochrane Central Register of Controlled Trials]) in January 2007 without language restrictions. The search strategies combined terms for the anticoagulants of interest with terms for cancer with search filters for randomized controlled trials (RCTs) (eTable).10
We hand searched the conference proceedings of the American Society of Clinical Oncology (from 1982) and the American Society of Hematology (from 2003). We reviewed the reference lists of included articles, relevant articles, and related systematic reviews2,8,9 and used the “Related Articles” feature in PubMed to identify additional citations.
Two reviewers (teams composed of many of us) independently screened the titles and abstracts for eligibility. We included studies that were RCTs, that enrolled patients with cancer, that compared LMWH with UFH, and that assessed the outcomes of interest (see the following subsection). We retrieved the full text of citations judged to be potentially eligible by at least 1 reviewer. Two reviewers independently screened the full-text articles for eligibility and resolved disagreements by discussion. We included abstracts only if the investigators supplied the necessary data on methods and results.
Two reviewers independently assessed the methodological quality of and abstracted data from each included trial using a pilot-tested and standardized form. They resolved disagreements by discussion, if necessary with help from an arbiter. We wrote to authors of included trials for incompletely reported data. The methodological criteria included allocation concealment, patient blinding, physician blinding, outcome assessor blinding, analyst blinding, percentage of follow-up, adherence to the principle of intention-to-treat analysis, a priori sample size calculation, and avoidance of early stoppage of the trial.
We aimed to collect data relating to the intervention and its control (type, dosing schedule, and duration), the cointerventions (including elastic stockings, pneumatic compressions, and early ambulation), participant characteristics (including type of cancer and surgery site), and outcome assessment (including type of screening and diagnostic tests).
We extracted the outcome data necessary to conduct intention-to-treat analyses, and we collected data for all-cause mortality, DVT (detected by means of screening or diagnosed secondary to clinical suspicion), PE, major bleeding, minor bleeding, wound hematoma, subsequent surgery for bleeding, thrombocytopenia, and heparin-induced thrombocytopenia (HIT). For the evaluation of bleeding complications and thrombocytopenia, we accepted the authors' definitions as long as they were standardized within studies. We also collected data for intraoperative and postoperative blood loss, intraoperative and postoperative blood product transfusion, and surgical tube drainage.
We calculated the agreement between the 2 reviewers for the assessment of eligibility using the κ statistic, and we checked for possible publication bias using inverted funnel plots.11 For categorical variables, we extracted the number of participants and the number of events by treatment arm, and we calculated a pooled relative risk (RR). For continuous variables, we extracted the mean and standard deviation by allocation arm, and we calculated the standardized mean difference (SMD). We measured homogeneity across trial results using the I2 statistic,12 and we considered the following classification of heterogeneity based on the value of I2: 0 to 30 indicates low; greater than 30 to 60, moderate and worthy of investigation; greater than 60 to 90, severe and worthy of understanding; and greater than 90 to 100, allowing aggregation only with major caution. We pooled the results using random effects. To explain heterogeneity, we attempted, whenever possible, to conduct subgroup analyses based on the characteristics of the participants, the interventions, and the outcomes. We conducted subgroup analyses for UFH administration 2 and 3 times daily and for abdominal surgery.
We defined a priori that we would conduct separate analyses for each diagnostic strategy (ie, studies using a diagnostic workup triggered by clinical suspicion were distinct from studies using a diagnostic workup triggered by a positive venographic result found by screening). Owing to the small number of trials using each of the specific diagnostic strategies and assuming that the relative effect is constant regardless of detection method, we conducted a post hoc analysis of the effect on DVT by pooling data from all the trials irrespective of whether they used the same strategy (“any diagnostic strategy”). For trials using more than 1 diagnostic strategy, we chose the data from 1 strategy using the following hierarchy: diagnostic workup triggered by clinical suspicion, diagnostic workup triggered by positive venography findings, diagnostic workup triggered by positive venous Doppler ultrasonographic findings, diagnostic workup triggered by a positive iodine 125–labeled fibrinogen uptake test result, and diagnostic workup triggered by positive impedance plethysmography findings.
The search strategy identified 3986 citations, including 322 duplicates. The title and abstract screening of the 3664 unique citations identified 166 as being potentially eligible for this review. The full-text screening of the 166 citations identified 26 eligible RCTs (Figure 1). We excluded 140 trials, as described in Figure 1. Agreement between reviewers for trial eligibility was excellent (κ = 0.94).
Of the 26 eligible RCTs, 11 included patients with cancer exclusively (n = 4006).13- 23 The remaining 15 trials included patients with cancer as subgroups: 2 trials24,25 reported the cancer subgroup data (n = 1341), and we obtained cancer subgroup data from the authors26 of 1 trial (n = 475). We were not able to obtain cancer subgroup data for the remaining 12 trials27- 38 (n = 3185); thus, we did not include them in the meta-analysis. The inverted funnel plots for the outcomes of death at 1 year and DVT did not suggest publication bias (Figure 2).
In the Table, we present the characteristics of the 14 included trials. Of these trials, 6 concealed allocation, 8 reported blinding patients and physicians, 5 reported blinding outcome assessors, 4 reported blinding data analysts, 7 used intention-to-treat analysis, and 13 had follow-up greater than 80%. Five trials had a formal sample size calculation, but in 2 trials, patients with cancer constituted subgroups. One trial was published only as an abstract and did not meet most of the methodological quality criteria, probably owing to limited reporting secondary to space constraints. Otherwise, the overall methodological quality of the trials was moderate.
Seven trials13,14,17,19,21,23,24 reported death, but 2 of these reported no events in either group.13,14 The pooled analysis of the remaining 5 trials showed no significant survival benefit with LMWH compared with UFH (RR, 0.89; 95% confidence interval [CI], 0.61-1.28; I2 = 20%) (Figure 3). There were no differences in mortality in the subgroup of trials comparing LWMH with UFH administered 2 times daily (RR, 1.11; 95% CI, 0.55-2.25; I2 = 30%) or 3 times daily (RR, 0.74; 95% CI, 0.51-1.08; I2 = 0%). The difference between the RRs for the 2 subgroups was not significant (P = .20).
Six trials reported data on DVT events that were initially clinically suspected and subsequently objectively confirmed (using venography in 4 studies,13,16- 18 Doppler ultrasonography in 1 study,15 and either test in 1 study14); however, 4 trials reported no events.13- 16 The pooled analysis of the remaining 2 trials,17,18 both administering UFH 3 times daily, showed no benefit with LMWH (RR, 0.73; 95% CI, 0.23-2.28; I2 = 0%).
In 2 trials,17,26 DVT was diagnosed based on venography screening. The pooled analysis showed no difference using LMWH compared with UFH (RR, 0.78; 95% CI, 0.57-1.06; I2 = 0%). In 6 trials,13,16,18,19,24,25 a DVT diagnostic workup was triggered by a positive isotopic screening (iodine 125–labeled fibrinogen uptake test) result and was subsequently confirmed by venography in all but 1 trial.24 The pooled analysis of the 5 trials reporting events showed a significantly lower DVT rate with LMWH compared with UFH (RR, 0.72; 95% CI, 0.52-0.99; I2 = 0%). The reduction was not significant in either of the 2 subgroups comparing LMWH with UFH administered 2 times daily (RR, 0.70; 95% CI, 0.48-1.01; I2 = 0%) or 3 times daily (RR, 0.79; 95% CI, 0.41-1.50; I2 = 0%). The difference between the RRs for the 2 subgroups was not significant (P = .39).
Twelve trials13- 20,22,24- 26 assessed DVT outcome using any diagnostic strategy. In the post hoc analysis pooling data from 8 of these trials17- 20,22,24- 26 that reported events, LMWH was superior to UFH (RR, 0.72; 95% CI, 0.55-0.94; I2 = 0%) (Figure 4). The benefit was significant in the subgroup of trials comparing LMWH with UFH administered 2 times daily (RR, 0.66; 95% CI, 0.44-0.99) but not in the subgroup comparing LMWH with UFH administered 3 times daily (RR, 0.78; 95% CI, 0.53-1.15; I2 = 0%). The difference between the RRs for the 2 subgroups was not significant (P = .28).
Twelve trials assessed PE,13- 21,24- 26 but 5 of them reported no events.13- 16,25 The pooled analysis of the remaining 7 trials showed no difference comparing LMWH with UFH (RR, 0.60; 95% CI, 0.22-1.64; I2 = 26%). The difference between LMWH and UFH was not significant in the 2 subgroups administering UFH 2 times daily (RR, 0.41; 95% CI, 0.11-1.55; I2 = 0%) or 3 times daily (RR, 0.70; 95% CI, 0.14-3.57; I2 = 45%). The difference between the RRs for the 2 subgroups was not significant (P = .13).
Only 3 trials17,21,26 reported minor bleeding; the pooled analysis showed no difference in minor bleeding (RR, 0.88; 95% CI, 0.47-1.66). Heterogeneity was severe (I2 = 75%). Similarly, there were no differences in the 6 trials assessing major bleeding (RR, 0.95; 95% CI, 0.51-1.77). Heterogeneity was moderate (I2 = 42%).
Three trials13,15,21 reported wound hematoma as an outcome. The pooled analysis showed no difference in LMWH- and UFH-treated patients (RR, 0.65; 95% CI, 0.39-1.09; I2 = 0%). Only 2 trials13,21 assessed subsequent surgery for bleeding as an outcome. The pooled analysis showed no difference in groups (RR, 0.70; 95% CI, 0.06-7.89). Heterogeneity was moderate (I2 = 40%).
For the outcomes of intraoperative blood loss (SMD, −0.06; 95% CI, −0.25 to 0.13), postoperative drain volume (SMD, 0.05; 95% CI, −0.08 to 0.19), and postoperative transfusion (SMD, 0.26; 95% CI, −0.18 to 0.70), there were no differences between LWMH and UFH. Based on 1 trial,16 intraoperative transfusion volume was higher with LMWH (SMD, 1.16; 95% CI, 0.69-1.62).
Three trials13,20,21 reported thrombocytopenia rates, but 1 reported no occurrence of thrombocytopenia.13 The pooled analysis from the 2 remaining trials showed no difference (RR, 1.18; 95% CI, 0.49-2.81). None of the studies reported HIT.
For patients undergoing abdominal surgery, subgroup data were available for some outcomes from 4 studies.13,24- 26 Results were similar to those of the primary analysis. There was no difference between LMWH and UFH in effect on DVT diagnosed after isotopic screening (RR, 0.73; 95% CI, 0.49-1.08), PE (RR, 0.74; 95% CI, 0.06-9.83), or death (RR, 0.66; 95% CI, 0.22-1.98). The LMWH was superior to UFH in effect on DVT using any diagnostic strategy (RR, 0.72; 95% CI, 0.52-0.99).
This systematic review showed no significant difference in the effect on survival between LMWH and UFH for perioperative thromboprophylaxis in patients with cancer. The main analysis showed no differences in DVT, PE, and bleeding rates. Although the absence of a statistically significant difference might reflect a true absence of effect, it could also be related to the lack of power to detect important differences.
In a subgroup analysis of trials comparing LMWH with UFH administered 2 times daily rather than 3 times daily, DVT rates were lower. This subgroup analysis should be interpreted cautiously because it fails 2 important criteria of the 7 criteria for a credible subgroup difference39: the effect is not suggested by comparisons within rather than between studies and the effect is not statistically significant. We found no trials directly comparing 2 times with 3 times daily UFH dosing regimens in this population. Indirect comparisons, even when adjusted, do not always agree with the results of head-to-head comparisons.40
There are several strengths to this systematic review, including the rigorous search strategy without language restrictions and the assessment of publication bias. We conducted study selection in duplicate to minimize the likelihood of missing relevant trials, and we had excellent agreement. We evaluated methodological quality and abstracted data in duplicate to minimize random and systematic error. The overall methodological quality of the included trials was moderate.
Although we used a comprehensive search strategy, a potential limitation of this review is the restriction of the electronic search strategy to patients with cancer; indeed, some of the data included in this review were from trials not restricted to patients with cancer. However, the search strategy identified all trials included in earlier systematic reviews on the same topic unrestricted to patients with cancer.2 We could not obtain data for 12 trials including subgroups of patients with cancer. These trials could have contributed 3185 additional participants to the meta-analyses, whereas 5822 are included in the present analysis. If the treatment effect estimated from those 12 trials was different from the true effect, these results would be biased.
Another limitation is that the included studies varied in the types of malignant neoplasms, types of surgical procedures, dosing of anticoagulant medications, follow-up periods, and measurement of end points. This might be a particular concern with older studies. Because of the limited number of studies, we could not explore in subgroup analyses the impact of all of these characteristics. In addition, this meta-analysis lacked sufficient power to detect a statistically significant and meaningful difference between the RR of LMWH and UFH 2 and 3 times daily, if one exists.
The event rate of PE was relatively low (0.6%). Thus, even now, individual trials, and even meta-analyses of these trials, remain underpowered to show statistically significant and meaningful differences in PE rates. On the other hand, the event rate of DVT in the control group (UFH) is lower when DVT is symptomatic (1.6%) rather than asymptomatic and detected by screening (8.1% for the iodine 125–labeled fibrinogen uptake test and 16.5% for venography). Consequently, the post hoc analysis pooling DVT data from all the trials (any diagnostic strategy) includes more events from studies incorporating screening (ie, DVTs that are asymptomatic and detected by screening), but it likely has relatively little effect on relative estimates of effect (ie, the RR), expressed by the low heterogeneity.
All included studies comparing LMWH with UFH started anticoagulant drug treatment preoperatively. Thus, it is not certain how the results apply to settings in which anticoagulant drug treatment is started postoperatively. The results suggest that physicians should consider preoperative rather than postoperative use. Further support for preoperative use comes from studies that did not find statistically significant differences in the amount of blood loss when patients were randomized to a first dose of enoxaparin sodium 12 hours before surgery vs postoperatively.41
A systematic review8 of thromboprophylaxis in colorectal surgery (search date: 2003) showed that LMWH and UFH were similarly effective in preventing DVT and PE (odds ratio, 1.01; 95% CI, 0.67-1.52); however, DVT and PE were not analyzed separately. A systematic review42 of thromboprophylaxis in gynecologic surgery (search date: 2005) showed no statistically significant difference between LMWH and UFH on DVT.
Another systematic review43 compared the risk of thrombocytopenia and HIT with UFH and LMWH thromboprophylaxis. Most of the studies enrolled patients undergoing orthopedic surgery. Only 2 trials prospectively examined HIT, and they identified only 10 events overall (all in the UFH group); a meta-analysis of these 2 RCTs measuring HIT showed an odds ratio of 0.10 (95% CI, 0.01-0.82), and a meta-analysis of 15 studies measuring thrombocytopenia showed an odds ratio of 0.47 (95% CI, 0.22-1.02), favoring LMWH. Furthermore, a recent meta-analysis44 comparing therapeutic doses of UFH with LMWH found only 2 trials examining this outcome, and no differences in HIT rates were found (RR, 1.33; 95% CI, 0.77-2.30). Although none of the prophylaxis trials included in this review reported HIT, the thrombocytopenia rates were similar (RR, 1.18; 95% CI, 0.49-2.81).
Evidence of a survival benefit of anticoagulation in patients with cancer mediated through an antineoplastic effect is accumulating.45,46 However, it is unclear whether a short course of lower-dose perioperative heparin thromboprophylaxis can exert such as an effect. We postulate that longer-term treatment with heparin may be required to realize this potential benefit.
As the American College of Chest Physicians recommends, patients with cancer should receive prophylaxis that is appropriate for their current surgical risk.7 For example, Andtbacka et al47 showed that in patients undergoing breast cancer surgery and treated with mechanical antiembolic devices and early ambulation, venous thromboembolism is rare (0.16% per procedure in 60 days).
In conclusion, this systematic review suggests no survival benefit and no harm associated with LMWH compared with UFH for thromboprophylaxis in patients with cancer undergoing surgery. In choosing one or the other agent, physicians should consider factors such as cost, ease of administration, and patient preferences. Further randomized trials are needed to confirm or refute the hypothesis that if UFH is used, 3 times daily dosing may be more effective in DVT prevention than 2 times daily dosing.
Correspondence: Elie A. Akl, MD, MPH, Department of Medicine, State University of New York at Buffalo, Erie County Medical Center CC-142, 462 Grider St, Buffalo, NY 14215 (firstname.lastname@example.org).
Accepted for Publication: December 18, 2007.
Author Contributions:Study concept and design: Akl and Schünemann. Acquisition of data: Akl, Terrenato, Barba, Sperati, and Schünemann. Analysis and interpretation of data: Akl, Terrenato, Sempos, Muti, Cook, and Schünemann. Drafting of the manuscript: Akl. Critical revision of the manuscript for important intellectual content: Terrenato, Barba, Sperati, Sempos, Muti, Cook, and Schünemann. Statistical analysis: Akl, Terrenato, and Schünemann. Obtained funding: Muti and Schünemann. Administrative, technical, and material support: Muti and Schünemann.
Financial Disclosures: Dr Cook received a donation of LMWH from Pfizer to compare it with UFH in a peer review–funded trial of thromboprophylaxis for critically ill patients. Dr Schünemann has received research grants and honoraria that were deposited into research accounts or received by a research group that he belongs to from AstraZeneca, Amgen, Chiesi Foundation, Lily, Pfizer, Roche, and UnitedBioSource for development or consulting regarding quality-of-life instruments for chronic respiratory diseases and as lecture fees related to the method of evidence-based practice guidelines development and research methods. Institutions or organizations that he is affiliated with likely receive funding from for-profit sponsors that are supporting infrastructure and research that may serve his work. He also is an editor of the American College of Chest Physicians Antithrombotic and Thrombolytic Clinical Practice Guidelines.
Funding/Support: This study was supported by European Commission: The Human Factor, Mobility and Marie Curie Actions Scientist Reintegration Grant IGR 42192 (Dr Schünemann). Dr Cook holds a chair from the Canadian Institutes for Health Research.
Role of the Sponsor: The funding bodies had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.
Additional Contributions: Ann Grifasi, BS, provided administrative support. We thank all the authors who provided needed information.