Le Maître, Aurélie MSc*; Ding, Keyue PhD*; Shepherd, Frances A. MD, FRCPC†; Leighl, Natasha MD, MMSc, FRCPC†; Arnold, Andrew MD, ECFMG, MRCP, FRCPC‡; Seymour, Lesley MD, PhD*
Cancer and its treatments are well-recognized risk factors for venous thromboembolism (VTE). A large population based, case control study1 showed that the risk of VTE was increased sevenfold in patients with malignancy, and that lung cancer had one of the highest incidence rates. The National Hospital Discharge Survey2 revealed a VTE incidence of 2.1% among lung cancer patients, compared with 1% in hospitalized patients without cancer. More recently, Chew et al.3 found that the 1-year incidence of VTE in lung cancer patients was significantly increased compared with the general population (Standardized Incidence Ratio = 21.2, 95% CI 20.4–22.0). The risk of VTE seems to be higher still in patients with lung cancer receiving chemotherapy.4
Long-term anticoagulation, usually with warfarin has potential problems in this population. Both recurrent VTE and bleeding during oral anticoagulant therapy are more frequent in patients with cancer than in patients without malignancy5,6 and drug-drug interactions frequently have been described and may be more common with orally administered molecularly targeted agents. Low-molecular weight heparin (LMWH) may be a safer alternative to oral anticoagulant therapy. A meta-analysis7 including noncancer trials between 1994 and 2001 showed a reduction in the rate of VTE recurrences and major bleeding complications in favor of LMWH compared with oral anticoagulants although the difference was not significant. A more recent meta-analysis suggests that LMWH may be superior in terms of recurrence of VTE.8 In addition, Meyer et al.9 showed that warfarin was associated with a high bleeding rate in patients with VTE and cancer. In 2003, the CLOT study (Randomized Comparison of Low-Molecular-Weight Heparin versus Oral Anticoagulant Therapy for the Prevention of Recurrent Venous Thromboembolism in Patients with Cancer)10 revealed that dalteparin was more effective than an oral anticoagulant in reducing the risk of recurrent thromboembolism without increasing the risk of bleeding in patients with cancer.
We report here the results of a retrospective pooled analysis of three lung cancer trials from the NCIC Clinical Trials Group that investigates whether the prevalence of bleeding was different according to anticoagulant treatment received. In addition, the impact of nonsteroidal antiinflammatory drugs and aspirin was examined.
Three randomized controlled trials of the NCIC Clinical Trials Group completed between 2000 and 2006 were analyzed to assess the risk of bleeding according to anticoagulant received. The treatment arms for each of the trials are described in Table 1. All randomized patients were including, irrespective of random assignment.
The BR.18 trial11 compared the matrix metalloproteinase inhibitor (MMPI) BMS-275291 to placebo in 774 non-small cell lung cancer (NSCLC) patients receiving first-line paclitaxel plus carboplatin chemotherapy. The BR.20 trial12 included 107 patients and compared vandetanib to placebo in small cell lung cancer patients who had achieved complete or partial response to induction chemotherapy with or without radiation therapy. The BR.21 trial13 compared the epidermal growth factor receptor inhibitor erlotinib to placebo (2:1 randomization) in 731 NSCLC patients with cancer treated after the failure of first-line or second-line chemotherapy. All randomized patients (1612) are included in this study. All three trials used Common Toxicity Criteria version two (CTC) to categorize and grade toxicity, including laboratory values. There were no limitations on concomitant anticoagulants in BR.21 and BR.18, although BR.21 required close monitoring for patients on warfarin, and suggested that LMWH should be used preferentially. In BR.18, VTE was reported in 7% in patients receiving chemotherapy alone and in 9% of patients on the MMPI arm, although in BR.20 1 patient receiving placebo had VTE compared with three vandetanib-treated patients. In BR.21, 3% of patients were reported to have had VTE on study (same in both arms).
Concomitant medications taken at entry to the study and during the study were recorded with onset and stop dates. Reasons for usage were recorded in BR.18 only. LMWH, warfarin, and nonsteroidal antiinflammatory drugs including aspirin (NSAID) usage was summarized. As expected, some patients used more than one category of these agents concomitantly.
The primary study outcome was the occurrence of bleeding after the usage (if used at or prior to randomization) or initiation of the concomitant anticoagulant or NSAID. Bleeding adverse events were included irrespective of causality to protocol therapy and included central nervous system bleeding, melena or other gastro-intestinal bleeding, epistaxis, bleeding associated with surgery, hemoptysis, rectal bleeding, petechiae/purpura, hemorrhage/bleeding, hematuria, vaginal bleeding, hematemesis, and disseminated intravascular coagulation.
Comparisons of baseline and on-treatment characteristics of patients in the three trials were performed to determine whether the patient populations varied among the three trials. The Mantel–Haenszel test stratified by treatment arm was used to test the association between baseline and on-treatment characteristics in the pooled database.
The χ2 test was used to compare the occurrence of bleeding according to baseline and on-treatment characteristics (including anticoagulant and NSAID usage) of patients in each trial and the Mantel–Haenszel test stratified by treatment arm was used in the pooled database. The same analysis was made to explore grade 3 or higher bleeding toxicities.
A logistic regression analysis was performed to explore the influence of baseline and on-treatment characteristics on the risk of bleeding in the pooled database. Interactions between warfarin, LMWH and NSAIDs were introduced into the model. The stepwise model-selection was used to select the appropriate variables in the model.
Description of Baseline and on Treatment Characteristics by Trial
Pretreatment and on-treatment patient characteristics are presented in Table 2. As expected due to the differing eligibility criteria of the three trials, there were more patients with performance status 2 or 3 in the BR.21 trial, more patients with abnormal creatinine in BR.20 and more patients with thrombocytopenia during treatment in BR.18, the only chemotherapy trial. More patients on BR.18 received NSAIDs than in the two others trials. The incidence of hemorrhage of any grade or causality was lowest in the placebo (BR21/20 and chemotherapy ± MMPI arms, BR.18, 2–3%), intermediate with BR.21 (EGFRI, 6%), and highest on the active arm of BR20 (vandetanib, 19%). The majority of the events were mild.
Relationship Between Baseline Characteristics and Concomitant Drugs
Table 3 describes the relation between anticoagulants and NSAIDs taken during study treatment and baseline patient characteristics. One hundred fifty two patients (9.4%) had more than one drug (14% in BR.18, 3% in BR.20 and 6% in BR.21). Patients with a poorer performance status (2, 3) were more likely to be treated with LMWH. Few patients had platelet counts under 150 × 109/liter and the observation that warfarin or NSAID usage seemed more frequent in these patients is difficult to interpret. Patients with an elevated creatinine were more likely to receive warfarin compared with those with a normal creatinine. The mean duration of treatment with anticoagulants, (although on protocol only, as data on concomitant medications were not collected thereafter), was shorter than treatment with NSAIDs (up to 4 months versus 7 months). These differences in duration of therapy are likely due to protocol requirements for discontinuation of protocol therapy with severe adverse effects such as VTE. The most common reason for usage of anticoagulants was for the treatment, prevention and or prophylaxis of venous or arterial thromboembolism, and usage for central line prophylaxis was not reported.
Univariate Analysis of Bleeding
Overall, 25% of patients had bleeding adverse events reported at any time in the three trials: 23% in BR.18, 23% in BR.20, and 29% in BR.21 (Table 4). Older patients (≥65 years) were more likely to have bleeding of all grades (p = 0.054), but there was no apparent increase in severe bleeding (≥grade 3), although a history of bleeding predicted for an increased risk of severe bleeding (p = 0.003). There was a trend (p = 0.075) for an increased risk of severe bleeding with platelet counts ≤150 × 109/liter at entry. Gender, performance status, and creatinine were not predictive.
There was a significantly increased risk of bleeding of all grades and severe bleeding with on-treatment thrombocytopenia (≤100 × 109/liter p = 0.013 and 0.001, respectively) and warfarin usage (p = 0.005 and 0.0002, respectively; Table 5). LMWH usage was associated with increased severe bleeding (p = 0.003) although NSAID usage did not result in an increased risk of bleeding.
Multivariate Analysis of Bleeding Toxicities
The risk of bleeding was assessed after adjusting for baseline characteristics, treatment arm, and anticoagulants or NSAIDs. The only significant covariate that affected the risk of bleeding was warfarin, with an odds ratio of 1.7 (95% CI 1.2–2.6) (Table 6a). Interaction between warfarin and LMWH or NSAID did not meet the statistical criterion for entry into the model.
As patients commonly were exposed to more than one drug of interest, we also performed three different models with only one concomitant drug as the covariate in each model. The model with only LMWH and the model with only an NSAID as the concomitant drug did not show an excess risk of bleeding with these drugs. The only significant covariate in these two models was age with an odds ratio of 1.3 (p value = 0.058) for patients more than 65 years of age. In contrast, in the model with only warfarin as the concomitant drug, there was a significant association between the risk of bleeding (odds ratio = 1.7, 95% CI 1.2–2.6, p value = 0.007). Once again, the only other covariate selected in the model was age (odds ratio = 1.2, p value = 0.07). We also performed an analysis to check potential drug and drug interaction on grade 3 or higher bleeding toxicities, and found that the use of LMWH only (OR = 8.1, 95% C.I. 3.3 – 20.3, p < 0.0001) or warfarin only (OR = 4.4, 95% C.I. 1.9 – 10.3, p = 0.0005) increased the risk of grade 3 or higher bleeding, while the alternative use of LMWH and warfarin decreased the risk (OR= 0.05, 95% C.I. 0.01 – 0.49, p = 0.01) (Table 6b).
In patients with cancer, the risk of VTE is higher than in the general population, and this risk seems to increase further with treatment, especially chemotherapy and hormonal therapy. Many patients included in clinical trials have smoking related cancers such as lung cancer, and thus are more likely to have comorbid conditions which may further increase the risk of both venous and arterial thomboembolism.
Anticoagulants are thus commonly required in patients with cancer, including lung cancer, but may increase the risk of bleeding, especially when used in combination with chemotherapeutic agents that may themselves increase the risk of hemorrhage due to thrombocytopenia. More recently, some molecularly targeted agents such as the angiogenesis inhibitors have been shown to be associated with an increased risk of bleeding and even fatal hemorrhage due to their direct vascular effects.14 The oral receptor tyrosine kinase inhibitors of the epidermal growth factor receptor may also increase the risk of bleeding when used concomitantly with warfarin, which is highly protein bound and thus commonly cited in drug-drug interactions. Finally, the concomitant use of anticoagulants and corticosteroids that are frequently prescribed for cancer patients for hypersensitivity or emesis prophylaxis may increase the risk of gastrointestinal bleeding.
Many clinical trial protocols, particularly those trials evaluating orally administered novel agents exclude patients taking anticoagulants agents, because of fears of bleeding and or drug-drug interactions. Even in clinical trials where anticoagulation is permissible at entry or on study, treatment may be restricted to LMWH.
Our study suggests that warfarin usage significantly increases the risk of bleeding, while NSAIDs and LMWH do not. These results are consistent with results from other studies.8,10 Interestingly, in univariate analyses, increased age, low platelet counts (at entry or on treatment), and a prior history of bleeding were also associated with an increased risk of bleeding.
Our study has a number of limitations. The three trials included a heterogeneous group of patients, including both NSCLC patients (two trials) and small cell lung cancer patients (one trial), as well as patients receiving therapy in different settings (first-line, second- and third-line therapy). However, it should be noted that all trials were palliative in nature and included patients with relapsed or refractory disease. Furthermore, the trials tested different therapeutic modalities (chemotherapy, EGFRI, and MMPIs). As expected, the incidences of VTE were also different among the trials, with the highest incidence in BR.18, suggesting that both the underlying cancer and type of anticancer therapy may increase the risk of VTE. However, there were no significant differences in the incidence of VTE between arms of each trial. Another limitation of our study is that many patients took combinations of anticoagulants and NSAIDs - in BR.18 for example, 49% of patients taking LMWH also took NSAIDs. However we performed additional analyses to correct for this. Finally, although we collected data on concomitant anticoagulants and NSAID usage, including the rationale for usage, we did not collect information on dosage, and cannot definitively define the number of patients who may have been receiving nontherapeutic doses of the agents for line or port prophylaxis. Despite these confounding factors, we believe the results provide information useful to researchers defining protocol parameters regarding anticoagulant and NSAID usage.
In conclusion, based on this retrospective analysis of three randomized trials of lung cancer, concomitant warfarin usage should be avoided. For clinical trials, concomitant LMWH usage seems acceptable, although consideration may be given to exclusion of patients requiring LMWH with risk factors such as age ≥65 or a history of bleeding, from trials testing agents with a significant risk of bleeding (such as angiogenesis inhibitors).
This study was supported by the Canadian Cancer Society.
1. Blom JW, Doggen CJ, Osanto S, Rosendaal FR. Malignancies, prothrombotic mutations, and the risk of venous thrombosis. JAMA 2005;293:715–722.
2. Stein PD, Beemath A, Meyers FA, Skaf E, Sanchez J, Olson RE. Incidence of venous thromboembolism in patients hospitalized with cancer. Am J Med 2006;119:60–68.
3. Chew HK, Davies AM, Wun T, Harvey D, Zhou H, White RH. The incidence of venous thromboembolism among patients with primary lung cancer. J Thromb Haemost 2008;6:601–608.
4. Shepherd F, Hasan B, Hicks L, et al. Venous thromboembolism and non-small cell lung cancer: a pooled analysis of National Cancer Institute of Canada Clinical Trials Group trials. Eur J Caner 2007;5(Suppl):361.
5. Hutten BA, Prins MH, Gent M, Ginsberg J, Tijssen JG, Büller HR. Incidence of recurrent thromboembolic and bleeding complications among patients with venous thromboembolism in relation to both malignancy and achieved international normalized ratio: a retrospective analysis. J Clin Oncol 2000;18:3078–3083.
6. Prandoni P, Lensing A, Piccioli A, et al. Recurrent venous thromboembolism and bleeding complications during anticoagulant treatment in patients with cancer and venous thrombosis. Blood 2002;100:3484–3488.
7. Iorio A, Guercini F, Pini M. Low-molecular-weight heparin for the long-term treatment of symptomatic venous thromboembolism: meta analysis of the randomized comparisons with oral anticoagulants. J Thromb Haemost 2003;1:1906–1913.
8. Akl E, Barba M, Rohilla S, et al. 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. J Exp Clin Cancer Res
[serial online] 2008;27:21. Available from: www.jeccr.com
. Accessed August 15, 2008.
9. Meyer G, Marjanovic Z, Valcke J, et al. Comparison of low molecular-weight heparin and warfarin for the secondary prevention of venous thromboembolism in patients with cancer. Arch Int Med 2002;162:1729–1735.
10. Lee AYY, Levine MN, Baker RI, et al. Low-molecular-weight heparin versus a coumarin for the prevention of recurrent venous thromboembolism in patients with cancer. New Engl J Med 2003;349:146–153.
11. Leighl NB, Paz-Ares L, Douillard JY, et al. Randomized phase III study of matrix metalloproteinase inhibitor BMS-275291 in combination with paclitaxel and carboplatin in advanced non–small-cell lung cancer: National Cancer Institute of Canada–Clinical Trials Group Study BR. 18. J Clin Oncol 2005;23:2831–2839.
12. Arnold AM, Seymour L, Smylie M, et al. Phase II study of vandetanib or placebo in small-cell lung cancer patients after complete or partial response to induction chemotherapy with or without radiation therapy: National Cancer Institute of Canada Clinical Trials Group Study BR. 20. J Clin Oncol 2007;25:4278–4284.
13. Shepherd FA, Pereira JR, Ciuleanu T, et al. Erlotinib in previously treated non–small-cell lung cancer. N Engl J Med 2005;353:123–132.
14. Crabb SJ, Patsios D, Sauerbrei E, et al. Tumor cavitation: impact on objective response evaluation in trials of angiogenesis inhibitors in non-small cell lung cancer. J Clin Oncol 2009;27:404–410