Summary. Background: Low-molecular-weight heparins (LMWH) have an antitumor effect in vitro and in experimental animal models of malignancy. Retrospective data suggest that it might improve survival in cancer patients.
Objectives: To evaluate the effect of LMWH compared to placebo or no anticoagulant intervention on the survival of cancer patients.
Methods: We conducted a systematic review of randomized trials specifically evaluating the impact of LMWH on the survival of cancer patients. Data sources were: MEDLINE, EMBASE, HealthSTAR, Cochrane library, gray literature and cross-referencing from reference lists. Data extraction was performed by one reviewer, and accuracy was independently verified by a second reviewer. Meta-analysis was conducted using: (i) odds ratio (OR) and relative risk (RR); (ii) survival rates using censored endpoints; and (iii) hazard ratios (HR).
Results: The pooled HR in all patients was 0.83 (95% CI 0.70–0.99; P = 0.03), and in patients with advanced disease it was 0.86 (95% CI 0.74–0.99; P = 0.04), both in favor of the LMWH group. The results of the OR, RR and survival meta-analysis consistently favored the LMWH group. Sensitivity analyses according to tumor type were not conducted, because of a lack of information.
Conclusions: LMWH improves overall survival in cancer patients, even in those with advanced disease. Additional trials are required to define the tumor types, disease stages and dosing schedules most likely to provide the greatest survival benefit.
Since the description by Trousseau , the association between cancer and thrombosis has been well documented. Numerous studies have shown an intimate relationship between components of the coagulation system and the development, dissemination and progression of cancer. Tissue factor, fibrin and thrombin promote angiogenesis by upregulating the production of vascular endothelial growth factor, among other effects, and, together with plasmin, they induce a dysregulation of the matrix metalloproteinases, promoting cell migration and metastasis [2–7]. Fibrin and platelets impede the clearing of tumor cells by natural killer (NK) cells , and it has been demonstrated that the MET oncogene, involved in several types of tumors, directly induces the upregulation of the plasminogen activator inhibitor type 1 gene , and that the activation of the mutant K-ras oncogene, together with the inactivation of the p53 tumor suppressor gene, upregulate tissue factor expression .
The coagulation system thus appears to be a target of oncogenic events and necessary for the survival and spread of malignant cells. Therefore, it is likely that anticoagulant agents will have a positive impact on patient survival by multiple mechanisms, probably involving inhibition of the proliferation of malignant and other cell types , blockage of angiogenesis by decreasing the activity of proangiogenic factors , activation of the NK cells , and blockage of the metastatic process . In addition, low-molecular-weight heparin (LMWH) might decrease thrombotic complications, thus improving survival.
In 1931, it was reported for the first time that heparin inhibits the growth of transplanted tumor tissue . Since then, several studies have supported the existence of an antitumor effect of unfractionated heparin [15–20] and warfarin [21–26], and several meta-analyses of studies comparing unfractionated heparin and LMWH for treatment of venous thromboembolism have suggested that the risk of mortality in the subgroup of cancer patients might be reduced in those receiving LMWH [27–30]; however, other authors have reported contrasting findings . Because of this conflicting evidence, it cannot be concluded that the use of LMWH confers a survival benefit in cancer patients; thus, we aimed to conduct a systematic review to identify randomized trials specifically addressing the influence of LMWH on overall survival in cancer patients.
We included randomized trials comparing LMWH with placebo or no anticoagulant intervention in cancer patients with survival as the main outcome.
The search was conducted using the OVID interface in MEDLINE, EMBASE, HealthSTAR, The Cochrane Database of Systematic Reviews, The Database of Abstracts of Reviews of Effects, and The Cochrane Central Register of Controlled Trials in August 2005 and updated in July 2006, using the terms (cancer or tumor or malignancy).tw, (heparin or low-molecular-weight heparin or enoxaparin or dalteparin or reviparin or certoparin or tinzaparin or bemiparin or nadroparin).tw, (survival or mortality).tw, and random$.tw, where. tw indicates a textword search. No language or date restrictions were considered. The reference lists of the retrieved articles were reviewed for cross-referencing. We also conducted a search of the electronic versions of the proceedings of the meetings of the American Society of Hematology, the International Society on Thrombosis and Haemostasis, and the American Society of Clinical Oncology, which were considered to be gray literature (i.e. studies that are unpublished, have limited distribution, and/or are not included in bibliographic retrieval systems).
The retrieved references were assessed for possible inclusion on the basis of evaluation of the title and the abstract, or in full if no abstract was available. Letters to the editor, review articles, editorials and commentaries were excluded.
Assessment of study quality and data extraction
Data extraction and quality assessment was performed by one reviewer and independently verified by a second reviewer. Discrepancies were resolved by consensus. The main outcome measures were: (i) number of deaths at 1 year; and (ii) number of major bleeding episodes defined according to the criteria of the International Society on Thrombosis and Haemostasis . The secondary outcome measure was number of deaths at 2 years.
The quality of the studies was assessed using the criteria proposed by Jadad et al. , and we defined allocation concealment as appropriate or inappropriate according to the criteria proposed by Schulz and Grimes . The possibility of publication bias was explored using inverted funnel plots of odds ratio (OR) vs. precision.
We planned a meta-analysis using a random-effects model according to the method of DerSimonian and Laird . We used the number of deaths at 1 and 2 years as the outcome of interest, and because of a lack of consensus regarding the best summary statistic for evaluation of pooled effect estimates in meta-analysis [36–39], we present the results as OR and relative risk (RR). Differences between effects were tested using a z-test, and P-values ≤ 0.05 were considered significant. Statistical heterogeneity was calculated using Cochran's Q statistic , considering a P-value < 0.1 for χ2 as indicative of heterogeneity, and the Higgins I2 statistic , for which heterogeneity was defined as low if < 25%, moderate if between 25% and 50%, or high if > 50%. Sensitivity analyses according to cancer site and disease stage were planned a priori.
Because it has been suggested that the use of OR and RR to analyze survival data might be inappropriate , we used two meta-analytical techniques designed specifically to deal with these type of data. The first technique was a survival meta-analysis as described by Messori and Rampazzo . We combined the survival rates of different studies using censored data; then, a ‘log-rank OR’ (which can be interpreted as a traditional OR) was calculated, and statistical significance was tested by a z statistic comparing it with a table of normal distribution. To perform the survival meta-analysis, we used a program developed by the same authors  (meta, Release 4.52; available at: http://www.wwwlabsifo.org). Information on the number of patients at risk was obtained directly from the studies, and the number of censored patients and number of deaths were subsequently calculated using the reported survival probabilities. If information on patients at risk was not available, we extracted it from the published survival curves using the method described by Parmar et al. , assuming a constant rate of censoring. The second technique was a meta-analysis of effects using hazard ratios (HRs). We used the methods described by Parmar et al. to indirectly estimate the log HR and its SE from the published survival curves if they were not explicitly reported in the studies . HRs were pooled using the generic inverse variance method under the assumption of a random-effects model.
Finally, although not planned a priori, we conducted an influence analysis by eliminating individual studies to analyze their influence on the pooled estimates. Unless otherwise stated, calculations were performed using microsoft excel 2002 (Microsoft Corp., Seattle, WA, USA) and Review Manager release 4.2.8 (The Cochrane Collaboration, Oxford, UK).
Literature search results
The search yielded 283 references, of which 276 were excluded (Fig. 1). Seven references were fully evaluated [44–50]; three were excluded because they did not include a placebo/no anticoagulant intervention control group [48–50], and in one, survival was not the primary outcome . Four studies were included in the final review.
Characteristics of included studies and methodological quality
The four included studies evaluated 898 patients with solid tumors randomized to receive LMWH vs. placebo or no intervention (Table 1). One study was an open-label randomized trial including patients with small cell lung carcinoma, nearly half of whom had extensive disease . Two were double-blind randomized placebo-controlled trials including patients with different solid tumors in advanced stages [45,46]. One study was initially planned as a double-blind placebo-controlled trial, but, because of slow accrual, it was concluded as an open-label study . All studies but one  included previously treated patients. All studies were analyzed on an intention-to-treat basis. We did not detect publication bias, but this cannot be completely ruled out, given the small number of studies included.
Advanced stage (III–IV) solid malignancy, no restriction for previous treatment
Advanced non-curable solid malignancy
Advanced incurable solid malignancy with or without failure of first-line treatment
Dalteparin 5000 U OD for 18 weeks
Dalteparin 5000 U OD for 1 year or until death
Weight-based nadroparin (3800–7600 U) BID for 2 weeks and OD for 4 weeks
Dalteparin 5000 U OD for 2 years or until death
All patients received six cycles of cyclophosphamide, epirubicin, and vincristine. Patients with limited disease received local RT if they responded after CT
Usual CT, RT or surgery
Usual CT, RT, hormone and other therapeutic modalities. Excluded patients receiving CT or RT potentially leading to thrombocytopenia < 50 000
Standard CT or RT (not investigational). 29 dalteparin and 31 control patients were not receiving any treatment at inclusion
Three studies used prophylactic doses of dalteparin (5000 U daily) for 18 weeks , 1 year , 2 years , or until time of death. One study used weight-adjusted nadroparin for 6 weeks, with a high dose during the first 2 weeks . Three studies allowed the administration of concomitant treatment as judged necessary by the treating oncologists [45–47], whereas one study prospectively randomized patients to a standard chemotherapy protocol with or without dalteparin .
Patient characteristics are shown in Table 2. The median age at inclusion was similar in all studies, and one study included predominantly male patients . Three studies included predominantly patients with good performance status (Eastern Cooperative Oncology Group performance status score < 3 or World Health Organization performance scale < 3) [44,46,47] and one did not report on this issue .
Table 2. Characteristics of patients included in randomized trials evaluating the impact of low molecular weight heparin on cancer survival
Altinbas et al. 
Kakkar et al. 
Klerk et al. 
Sideras et al. 
LMWH low molecular weight heparin; ECOG Eastern Cooperative Oncology Group; WHO World Health Organization; NS not specified.
*The figures in brackets indicate the data for the open-label patients.†Lung cancer patients were included in the ‘other’ types group. ‡Includes non-small cell lung carcinoma and prostate cancer. §Figures represent the treatment that the patients were receiving at the time of inclusion in the study.
Median age at entry (years)
Male gender (%)
(stage I–II) (%)
Median overall survival (months)
Performance status at inclusion (%)
Type of cancer (%)
Lung (Small Cell)
Concomitant therapies (%)
One- and 2-year mortality The number of events and patients included are shown in supplementary Table S1. The pooled results of all the patients included in the four studies showed that at 1 year, the OR of death was 0.70 (95% CI 0.49–1.00; P = 0.05) (Fig. 2A), and the RR of death was 0.87 (95% CI 0.77–0.99; P = 0.04) (Fig. 2B), both in favor of patients receiving LMWH. As it has been suggested that the beneficial effect of LMWH might be limited to patients with less advanced disease , we conducted further analyses after excluding the patients in stages I and II included in the study by Altinbas et al. . These analyses showed a reduction in the OR of death (OR 0.75; 95% CI 0.57–0.99; P = 0.04) (Fig. 2A) and the RR of death (RR 0.89; 95% CI 0.80–0.99; P = 0.03) (Fig. 2B), both in favor of the patients receiving LMWH.
The analysis of the 2-year mortality favored the LMWH group in all patients (OR 0.57; 95% CI 0.34–0.96; P = 0.03) and in patients with advanced disease (OR 0.59; 95% CI 0.42–0.84; P = 0.004) (Fig. 2A). The analysis of the 2-year mortality using RR was consistent with these results: for all patients, the RR of death was 0.90 (95% CI 0.84–0.97; P = 0.007); for patients with advanced disease, it was 0.92 (95% CI 0.86–0.98; P = 0.02) (Fig. 2B). For all the analyses, the χ2 test for statistical heterogeneity was non-significant, and low to moderate heterogeneity was found by the Higgins I2 test.
Meta-analysis of survival data The meta-analysis of survival based on censored endpoints showed that for all patients there was a statistically significant difference between the pooled proportions of survival at 12 and 24 months favoring the LMWH group. The difference remained significant when we analyzed only patients with advanced disease (Table 3).
Table 3. Results of the survival meta-analysis based on censored endpoints
Survival estimates for patients with limited and advanced disease
Survival estimates for patients with advanced disease
The meta-analysis of HR (Fig. 3) showed that the pooled HR for all patients was 0.83 (95% CI 0.70–0.99; P = 0.03), and for patients with advanced disease was 0.86 (95% CI 0.74–0.99; P = 0.04), both in favor of the LMWH group. In all patients, the Cochran's Qχ2 had a P-value of 0.08 and the Higgins I2 was 55.4%, both indicating the presence of heterogeneity. For the analysis of patients with advanced disease, the χ2P-value was 0.23 and the Higgins I2 was 30.5%. Because of these findings, we conducted an influence analysis (not defined a priori), whose results are shown in Fig. 4. This analysis showed that the exclusion of the study by Altinbas et al.  resulted in a pooled HR of 0.88 (95% CI 0.76–1.08; P = 0.11), the exclusion of the study by Sideras et al.  resulted in a pooled HR of 0.78 (95% CI 0.66–0.93; P = 0.005), and the exclusion of both studies resulted in a pooled HR of 0.84 (95% CI 0.73–0.96; P = 0.01), with homogeneous results (χ2P = 0.33; Higgins’I2 = 0%).
Bleeding episodes There was no increase in the risk of major bleeding (OR 1.51; 95% CI 0.25–9.20; P = 0.65), with no statistical heterogeneity assessed by the Cochran's Q statistic (χ2P = 0.15) and moderate heterogeneity assessed by the Higgins I2 statistic (48%).
The results of the present study showed that the use of LMWH in addition to standard cancer treatment is associated with an improvement in the survival of patients with solid malignancies, with a reduction in the odds (30% at 1 year and 43% at 2 years) and the risk (13% at 1 year and 10% at 2 years) of death without an increase in major bleeding episodes. The analyses of OR and RR were consistent; however, as it has been suggested that a meta-analysis based on death counts ignores follow-up, reducing its statistical power , we conducted a meta-analysis using two methods designed to deal with survival data. Both methods showed a significant survival advantage favoring the use of LMWH, which resulted in a 17% reduction in the HR compared to placebo or no anticoagulant intervention. The results did not change when only patients with advanced disease were considered. We were not able to determine whether different tumor types might benefit differently from LMWH, and whether different drugs, dosing schedules or treatment durations have different effects.
Some limitations should be noted. Firstly, the number of patients included in the four analyzed studies was relatively small, so it is conceivable that the inclusion of future trials might change the results; however, we are confident that, on the basis of currently available data, our conclusions are robust. Secondly, we were unable to conduct analyses according to tumor type, as no information on outcomes for each tumor type was available, although it is likely that, because of the small number of patients included for each individual type of cancer, this analysis would have been methodologically inadequate. Thirdly, although three of the four trials included in our meta-analysis reported improved survival in patients with ‘better prognosis’, the evaluation and definition of prognosis in the studies varied greatly, and the analysis was post hoc in one of them . As performance status is a well-known prognostic indicator in cancer patients, it would have been important to analyze its effect on the results of the trials, and although this was not reported, three of the four studies stated that they only included patients with good performance status, so our results cannot be generalized to all cancer patients.
The finding of statistical heterogeneity deserves a separate comment. It is pertinent to mention that the Higgins I2 statistic is probably the most adequate way of assessing statistical heterogeneity in a meta-analysis including few studies; it describes the percentage of variability in the estimates that is due to heterogeneity rather than sampling error . In the present study, whereas there was no significant heterogeneity for most analyses, the analysis of HR for all patients did present heterogeneity (Higgins I2 = 55.4%). In order to explore this finding, we performed an influence analysis whose results showed that the study by Altinbas et al. influenced the pooled estimate towards a beneficial effect of LMWH, whereas the study by Sideras et al. had the opposite effect. A possible explanation of heterogeneity might be differences in study design or in the population included. The study by Altinbas et al. included a large proportion of patients with limited disease, and only included previously untreated patients, which might be associated with a better survival . On the other hand, the study of Sideras et al. included a large proportion of patients who were not receiving any more treatment at the time of inclusion, and survival might be poorer if patients are no longer receiving chemotherapy . However, the exclusion of both studies – which also had the lowest Jadad scores – still resulted in a pooled estimate favoring the LMWH group with low statistical heterogeneity (Higgins I2 = 0%), suggesting that the pooled estimates are robust. A particularly important consideration that might explain the conflicting results is that our data suggest that the actual survival benefit conferred by LMWH might be less than was supposed at the inception of the included trials; thus, it is likely that the studies were underpowered to detect the true difference between arms, and that might have resulted in a type II error. This could also explain why studies with a longer duration of treatment yielded negative results.
Several studies, some dating back more than 30 years, have reported a survival benefit associated with the use of unfractionated heparin in cancer patients [15–17]; unfortunately, these studies had serious methodological flaws. The first properly conducted trial evaluating heparin was published in 1994 by Lebeau et al. , who reported an improvement in the overall survival and complete response rate in patients with small cell lung cancer who received unfractionated heparin in addition to standardized chemotherapy. Subsequently, von Tempelhoff et al. randomized patients with a first diagnosis of gynecologic cancer to either 3000 anti-Xa units of certoparin daily or 5000 U of unfractionated heparin three times daily for 7 days after cancer surgery, and observed a reduction in mortality up to 650 days after surgery for patients with pelvic cancer receiving certoparin . More recently, Lee et al. published the results of a post hoc survival analysis of the CLOT trial, which compared dalteparin with warfarin for the prevention of recurrent venous thromboembolism in cancer patients [49,52]. They found an HR at 12 months of 0.50, favoring the dalteparin group, in patients with non-metastatic disease, whereas in the group with metastatic disease, there was no difference. A third study published in abstract form (and in full after the completion of our search) reported an analysis of patients undergoing surgery for abdominal malignancy randomized to 7 or 28 days of dalteparin 5000 U daily. The median survival rates were 42.2 vs. 60 months for the 7-day and 28-day arms, respectively [50,53]. Shortly after our review was completed, the results of the LITE study were published . This study randomized patients with malignancy-associated deep vein thrombosis to 3 months of either tinzaparin 175 U kg−1 day−1 or i.v. heparin followed by warfarin. There was no difference in mortality at 1 year. Because the results of these studies do not allow evaluation of the absolute effect of LMWH on cancer survival, they were excluded from the review. It is worth noting a report published in abstract form giving the results of the TOPIC studies, which compared 6 months of certoparin (3000 anti-Xa units daily) with placebo for the prevention of chemotherapy-associated venous thromboembolism in 900 patients with advanced breast carcinoma or non-small cell lung carcinoma . Survival was not the primary outcome of these studies, so they were excluded from the review. In addition, although the authors commented that no survival benefit was observed in the LMWH arm, the actual survival rates were not given, and their effect could not be explored.
It is also conceivable that the survival benefit might derive at least in part from a decrease in venous thromboembolic events. Given the small number of patients randomized in the studies included in the present review and the extremely low rate of thrombotic events, no conclusions can be drawn.
Whereas our results suggest that addition of LMWH to conventional treatment improves overall survival in cancer patients, even in those with advanced stage cancer, they also raise several questions that are yet to be answered. Firstly, it is not known whether LMWH has similar effects in different tumor types and disease stages. Future adequately powered studies should be conducted on specific tumors and consider possible differences in LMWH-related survival advantage according to disease stage. Our group is currently planning a large trial in patients with non-small cell lung carcinoma. Secondly, as a dose–response relationship is biologically plausible, different dosing schedules and durations of treatment should be evaluated. Thirdly, it is important to investigate whether different LMWH preparations are interchangeable, as there are data suggesting that some of the antitumor effects of LMWH are dependent on molecular weight . Fourthly, it is not yet possible to be sure whether the addition of LMWH to chemotherapy in newly diagnosed patients confers a greater survival benefit. Finally, in order to facilitate future comprehensive evaluations of the results across different trials, investigators should provide all pertinent information that could potentially impact on the results or require further analysis. In this sense, adherence to standards of reporting such as the CONSORT statement  might be of help.
This work was presented in part at the 48th Annual Meeting of the American Society of Hematology in Orlando, FL on 12 December 2006, and published in abstract form in Blood 2006; 108 (11): Abstract 717.
Disclosure of Conflict of Interests
A. Lazo-Langner is the recipient of a Graduate Scholarship from Consejo Nacional de Ciencia y Tecnología (CONACyT), México and is supported in part by an International Fellowship awarded by the University of Ottawa and by Program Grant PRG 5513 of the Heart and Stroke Foundation of Ontario. M. A. Rodger is supported by a New Investigator Research award from the Heart and Stroke Foundation of Ontario. Funding sources did not participate in the design, conduction or analysis of this work. M. A. Rodger has received grant funding from Pfizer, Sanofi, Leo Pharma, Bayer and AstraZeneca, and has received honoraria for educational sessions from Leo Pharma and AstraZeneca. G. D. Goss and J. N. Spaans state that they have no conflict of interest.