Low-molecular-weight heparins are superior to vitamin K antagonists for the long term treatment of venous thromboembolism in patients with cancer: a cochrane systematic review

Background Cancer and its therapies increase the risk of venous thromboembolism. Compared to patients without cancer, patients with cancer anticoagulated for venous thromboembolism are more likely to develop recurrent thrombotic events and major bleeding. Addressing all important outcomes including harm is of great importance to make evidence based health care decisions. The objective of this study was to compare low molecular weight heparin (LMWH) and oral anticoagulants (vitamin K antagonist (VKA) and ximelagatran) for the long term treatment of venous thromboembolism in patients with cancer. Methods A systematic review of the medical literature. We followed the Cochrane Collaboration methodology for conducting systematic reviews. We assessed methodological quality for each outcome by grading the quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology. Results Eight randomized controlled trials (RCTs) were eligible and reported data for patients with cancer. The quality of evidence was low for death and moderate for recurrent venous thromboembolism. LMWH, compared to VKA provided no statistically significant survival benefit (Hazard ratio (HR) = 0.96; 95% CI 0.81 to 1.14) but a statistically significant reduction in venous thromboembolism (HR = 0.47; 95% (Confidence Interval (CI) = 0.32 to 0.71). There was no statistically significant difference between LMWH and VKA in bleeding outcomes (RR = 0.91; 95% CI = 0.64 to 1.31) or thrombocytopenia (RR = 1.02; 95% CI = 0.60 to 1.74). Conclusion For the long term treatment of venous thromboembolism in patients with cancer, LMWH compared to VKA reduces venous thromboembolism but not death.


Background
The presence of cancer increases the risk of venous thromboembolism four to six fold [1]. Cancer related interventions such as chemotherapy, hormonal therapy and indwelling central venous catheters also increase the risk of venous thromboembolism [1]. Similarly, patients undergoing surgery for cancer have a higher risk of venous thromboembolism than those undergoing surgery for benign diseases [2,3]. Furthermore, patients with cancer and venous thromboembolism have a higher risk of death than patients with cancer alone or with venous thromboembolism alone [4,5].
Cancer patients also have different benefits and risks from anticoagulant treatment than those without cancer. For instance, during oral anticoagulation therapy for venous thromboembolism, patients with cancer, compared to those without cancer, have higher incidence of recurrent venous thromboembolism (27.1 versus 9.0 events per 100 patient years, p = 0.003) and of major bleeding (13.3 versus 2.2 events per 100 patient years, p = 0.002) [6].
Three systematic reviews have compared low molecular weight heparin (LMWH) and vitamin K antagonists (VKA) in the long treatment of venous thromboembolism, but in populations not restricted to patients with cancer [7][8][9] The review by van der Heijden et al. did not complete a preplanned subgroup analysis in patients with cancer as the required data was not specifically reported [7] The review by Conti et al. did not conduct a meta-analysis in the subgroup of patients with cancer [8] In the review by Ioro et al. a meta-analysis in the subgroup of patients with cancer found no statistically significant difference in mortality (OR = 1.13; 95% CI 0.54, 2.38).
No systematic review has focused on the long term treatment of venous thromboembolism in patients with cancer. The above mentioned subgroup analysis did not report on the comparative safety of LMWH and VKA [9] The Cochrane Collaboration has recognized that addressing all important outcomes including harm is of great importance to make evidence based health care decisions [10]. In addition, an analysis that includes an evaluation of direct comparative trials and direct subgroup comparison could prevent the potential pitfalls of indirect subgroup analysis [11].
The objective of this study was to conduct a systematic review to compare the efficacy and safety of LMWH and oral anticoagulants for the long term treatment of venous thromboembolism in patients with cancer.

Eligibility criteria
We included RCTs including patients with cancer with a confirmed diagnosis of venous thromboembolism (deep venous thromboembolism (DVT) or pulmonary embolism). The venous thromboembolic event should have been diagnosed using an objective diagnostic test. RCTs should have compared long term treatment with LMWH versus oral anticoagulants (VKA or ximelagatran) and should have treated patient groups similarly apart from the intervention of interest.

Outcomes of interest
Outcomes of interest included: survival, symptomatic recurrent DVT, symptomatic recurrent pulmonary embolism, major bleeding, minor bleeding, thrombocytopenia, and postphlebitic syndrome. We accepted the definitions of major bleeding, minor bleeding, thrombocytopenia and postphlebitic syndrome of the authors of the original studies as long as they were standardized.

Data Sources and Searches
The search was part of a comprehensive search for studies of anticoagulation in patients with cancer. We electronically searched in January 2007 the following databases from the date of their inception: The Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE and ISI the Web of Science (Additional file 1). We also hand searched the conference proceedings of the American Society of Clinical Oncology and of the American Society of Hematology. We reviewed the reference lists of included papers and used the related article feature in PubMed. We applied no language restrictions.

Study Selection
Two reviewers independently screened the titles and abstracts for eligibility. We retrieved the full texts of articles judged as potentially eligible by at least one reviewer. Two reviewers then independently screened the full texts articles for eligibility and resolved their disagreements by discussion. We included studies published as abstracts only if authors supplied us with the necessary information about their methods and results.

Data collection
Two reviewers independently extracted data using a standardized form and resolved their disagreements by discussion. Extracted data related to participant characteristics, the details of the interventions, the outcomes and methodological quality indicators. We contacted authors for incompletely reported data.
We assessed the following methodological criteria for each study: allocation concealment, blinding (patient, provider, outcome assessor, data analyst), whether the analysis followed the ITT principle, whether study was stopped early for benefit, and percentage of follow-up. We assessed the methodological quality for each outcome by grading the quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach [12]. The GRADE approach involves making separate ratings for quality of evidence for each patient important outcome and identifies five factors that can lower the quality of the evidence when considering RCTs: study limitations relating to the above methodological criteria (lack of allocation concealment; lack of blinding; failure to adhere to an intention to treat analysis; stopping early for benefit; and large losses to follow-up), inconsistency of results, indirectness of evidence, imprecision, and publication bias [13].
We extracted time to event data by abstracting the log(hazard ratio) and its variance from trial reports; if these were not reported, we digitised the published Kaplan-Meier survival curves and estimated the log(hazard ratio) and its variance using Parmar's methods [14]. We also noted the minimum and maximum duration of follow-up, which are required to make these estimates. We performed these calculations in Stata 9, using a specially written program, which yielded the reported log(HR) and variance when used on the data presented in Table V of Parmar 1998 [14].
We also extracted categorical data necessary to conduct intention-to-treat analyses. We collected outcome event rates whenever they were reported in each trial. When the authors did not report and could not provide the number of events at specific time points, two biostatisticians estimated these numbers independently and in duplicate from survival curves, if available.

Analysis
We calculated the agreement between the two reviewers for the assessment of trial eligibility using kappa statistic. We analyzed, when possible, both time to event data and binary data. For time to event data, we pooled the log(HR)s using a random-effects model and the generic inverse variance facility of RevMan 4.2. For binary data, for a specific outcome, and for each trial, we used the intention-to-treat principle to calculate the relative risk. We then pooled the results of trials with similar comparisons using a random-effects model.
We evaluated heterogeneity across trials using the I 2 statistics. I 2 describes the percentage of total variation across studies that is due to heterogeneity rather than chance [15]. The interpretation of I 2 depends on the magnitude and direction of effects as well as the strength of evidence for heterogeneity (e.g. P value from the chi-squared test, or a confidence interval for I 2 ) [10]. We used the following classification based on the value of I 2 [15]: 0-30 = low; 30-60 = moderate and worthy of investigation; 60-90 = severe and worthy of understanding; 90-100 = allowing aggregation only with major caution.
We created inverted funnel plots of individual study results plotted against sample size in order to evaluate possible publication bias. We conducted sensitivity analysis by excluding the study of lowest methodological quality [16] and then a study that used a different initial anticoagulant in the two study arms (post hoc analysis) [17]. Figure 1 shows the trial flow. The search identified 3986 citations, including 322 duplicates. The title and abstract screening of the 3664 unique citations identified 57 as potentially eligible for this review. The full text screening excluded 40 citations for the following reasons: The reasons for excluding the 40 citations are as follows: case series (1), review (15), retrospective study (4), protocol (2), observational study (6), trial but not randomized and controlled (4), no cancer patients included (3), only one patient with cancer was included (1), and no relevant outcome (2), comparison of two LMWH (tinzaparin and dalteparin) (1), 18 months extended treatment with Ximelagatran versus placebo (1).

Results of the search
Of the remaining 17 eligible RCTs, 11 included patients with cancer as a subgroup the data of which was not reported and not obtainable from the authors [18][19][20][21][22][23][24][25][26][27][28]. We thus report data from eight RCTs, five published in full text [17,[29][30][31][32][33] and one published as an abstract [16]. We also identified 2 publications related to RCTs we included in this review [34,35]. Agreement between reviewers for study eligibility was excellent (kappa = 0.94). Table 1 details the characteristics of the six included studies. Only one of these studies used a different initial anticoagulant in the two study arms (LMWH in the LMWH group and UFH in the vitamin K antagonist group) [17]. The inverted funnel plot for the outcome of all cause mortality did not suggest publication bias ( Figure 2).

Methodological quality of included studies
The concealment of allocation was adequate in three trials [17,30,32] and unclear in the other 3 [16,29,31]. None of the studies blinded patients or caregivers, four studies blinded outcome assessors [17,[30][31][32], and three studies blinded data analysts [17,30,32]. Five studies (all those published in full text) conducted ITT analysis [17,[29][30][31][32]; this was not clear in the study published as an abstract [16]. Two studies reported a priori sample size calculations [30,32]. The percentage follow-up ranged from 89% to 100%. None of the studies was stopped early for benefit. The methodological quality varied by outcome. It was low for mortality, moderate for recurrent VTE, low for major bleeding and very low for minor bleeding, low for major bleeding and (Table 2). Table 2 also provides the absolute reductions in the risks of the different outcomes for a number of illustrative baseline risks, including low and high baseline risks.

Effects of interventions Survival
We used time to event data reported by two studies [30,32] and supplied by the author of a third study [17].
8 RCTs included in the review and in the meta-analysis 19 trials eligible for the review 11 RCTs not included as cancer subgroups data were not available We finally pooled data from all studies irrespectively of the timing of outcome assessment and using the 6 months data from the study by Meyer et al.

Recurrent venous thromboembolism
We used time to event data reported by two studies [30,32] and supplied by the author of a third study [17].
None of the studies reported postphlebitic syndrome as an outcome.

Discussion
For the long term treatment of venous thromboembolism in patients with cancer, LMWH compared to VKA provided no statistically significant survival benefit but a statistically and patient important reduction in venous thromboembolism. There was no statistically significant difference between LMWH and VKA in terms of bleeding outcomes or thrombocytopenia.
Our systematic approach to searching, study selection and data extraction should have minimized the likelihood of missing relevant studies. This increases the confidence in the internal validity of our findings. A major limitation of this review is our inability to include in the meta-analyses 11 eligible RCTs with subgroups of patients with cancer because relevant data was not reported and not obtainable from the authors. However, the inverted funnel plot for the outcome of all cause mortality did not suggest publication bias. This suggests that the treatment effect from those 11 RCTs should be similar to the one estimated from the included studies. One has to keep in mind that funnel plots have limited power to detect bias if the number of studies is small [10].
Inverted funnel plot for the mortality outcome in rand-omized controlled trials of long term anticoagulation in patients with cancer and venous thromboembolism Figure 2 Inverted funnel plot for the mortality outcome in randomized controlled trials of long term anticoagulation in patients with cancer and venous thromboembolism. The pooled results for all cause mortality and bleeding outcomes showed moderate to severe heterogeneity. Unfortunately, the number of pooled studies was relatively small to explore the causes of heterogeneity by conducting subgroup analyses. However, the findings suggest that the trial that used a different initial anticoagulant in the two study arms is the source of heterogeneity [17].
Three published systematic reviews compared LMWH and VKA in the long treatment of venous thromboembolism [7][8][9]. Two of these systematic reviews showed no statistically significant reduction of recurrent venous thromboembolism by LMWH compared to VKA when the metaanalysis is not restricted to patients with cancer [7,9]. However, our meta-analysis shows a significant reduction in recurrent venous thromboembolism in patients with cancer. The reason for this differential effect in patients with cancer is not clear. A similar differential effect of anticoagulants has been found in the initial treatment of venous thromboembolism where LMWH was superior to UFH in patients with cancer but not in patients without cancer [36].
Of the three published systematic reviews comparing LMWH and VKA in the long treatment of venous thromboembolism [7][8][9], only the study by Ioro et al. conducted a meta-analysis in the subgroup of patients with cancer and found no statistically significant difference in mortality (OR = 1.13; 95% CI 0.54, 2.38). This finding is consistent with the results of our meta-analysis. While the reduction in venous thromboembolic events with LMWH in patients with cancer is expected to reduce thrombosis related mortality, this did not translate into a reduction in all cause mortality. This finding is not apparently  Comparison of the effects of LMWHs and vitamin K anatomists on survival (time to event analysis) in patients with cancer and venous thromboembolism Figure 3 Comparison of the effects of LMWHs and vitamin K anatomists on survival (time to event analysis) in patients with cancer and venous thromboembolism.
explained by an increase in any specific-cause mortality (e.g. fatal bleeding), but might be due to the lack of power to detect a reduction in all cause mortality especially that the results suggest a trend in that direction.
We were not able to conduct subgroup analyses based on type of cancer because of the lack of data. Such analyses would be interesting because of the survival benefits of LMWH in patients with limited small cell lung cancer [37] and of VKA in patients with small cell lung cancer [38] that are independent of any antithrombotic effects.

Conclusion
The decision for a patient with cancer and venous thromboembolism to start long term LMWH versus oral anticoagulation should balance the benefits and downsides and integrate the patient's values and preferences for outcomes and management options [39]. While LMWH decreases the incidence of venous thromboembolism and possibly of death, we speculate that it might be more costly and less acceptable because of its subcutaneous route of administration.
Future research should compare LMWH to other anticoagulants such as ximelagatran and fondaparinux. There is also a need for research assessing patients' values and preferences regarding long term anticoagulant agents for treating venous thromboembolism. Researchers should consider making the raw data of RCTs available for individual patient data meta-analysis. Further RCTs including subgroups of patients with cancer should report separate results for these subgroups. Comparison of the effects of LMWHs and vitamin K anatomists on mortality (categorical analysis) in patients with cancer and venous thromboembolism Figure 4 Comparison of the effects of LMWHs and vitamin K anatomists on mortality (categorical analysis) in patients with cancer and venous thromboembolism.

List of abbreviations
Comparison of the effects of LMWHs and vitamin K anatomists on recurrent venous thromboembolism (survival analysis) in patients with cancer and venous thromboembolism Figure 5 Comparison of the effects of LMWHs and vitamin K anatomists on recurrent venous thromboembolism (survival analysis) in patients with cancer and venous thromboembolism.