Venous thromboembolism (VTE), which includes deep vein thrombosis (DVT) and pulmonary embolism (PE), is a common disorder in patients with cancer 1 . It is an important cause of morbidity and one of the leading causes of death in that population 2 .

Anticoagulants are used for treatment of VTE and for thromboprophylaxis in high risk conditions such as surgery or the presence of an indwelling central venous catheter 3 4 5 6 . A survival benefit from anticoagulants as a result of VTE prevention and a possible direct antitumor effect has also been explored 7 8 .

High-quality randomized controlled trials (RCTs) are the preferred method to establish the effects on efficacy and safety outcomes, since they minimize systematic error (bias). However, the reliability of their results depends on the extent to which potential sources of bias have been avoided 9 . Both the Cochrane Collaboration and the GRADE working group have advanced the methods to define which criteria should be evaluated in order to judge the risk of bias of a trial or a body of trials 9 10 . Poorly designed and/or conducted RCTs may lead to biased results, mislead clinical practice and adversely affect patients’ outcomes. Unfortunately, overwhelming evidence shows the quality of RCTs is not optimal 11 .

A few studies assessing the quality or the reporting of RCTs in patients with cancer have concluded that the quality of reporting is low, and some suggest that it is lower than in other areas 12 13 14 15 16 . However, we have identified no study focusing on trials of anticoagulation in patients with cancer.

The objective of this study is to systematically describe the characteristics, the risk of bias and their time trends in RCTs of anticoagulation in patients with cancer 3 4 5 6 7 8 .

The study sample consists of all trials included in a series of six Cochrane systematic reviews of anticoagulation in patients with cancer 3 4 5 6 7 8 . The series covered the majority of topics for which RCTs were conducted in this field: parenteral anticoagulation for survival benefit (VTE thromboprophylaxis trials in ambulatory patients with cancer), oral anticoagulation for survival benefit, central venous catheters thromboprophylaxis, perioperative thromboprophylaxis, initial anticoagulation treatment of VTE, and long term anticoagulation treatment of VTE. A common search was conducted for all the reviews.

Our study describes the characteristics, the risk of bias and the related time trends in randomized trials of anticoagulation in cancer. We used a comprehensive search strategy and a systematic methodology for selecting studies and abstracting data.

We found that the majority of RCTs of anticoagulation in patients with cancer appear to use insufficiently rigorous outcome assessment methods and to have deficiencies in key methodological features. Specifically, included studies had low rates of blinding of participants and caregivers, and intention to treat analysis. Our findings likely reflect a problem in the design, conduct and the reporting of these trials.

It has to be acknowledged it might be particularly challenging for trials of anticoagulation in the cancer setting, relative to other types of interventions in other populations, to adhere to some of the methods employed to reduce the risk of bias. For example, clinicians caring for patients with active cancer might be reluctant to accept to be blinded given their patients are already at increased risk of bleeding, frequently undergo invasive procedures, and might have chemotherapy-induce thrombocytopenia.

Our findings are consistent with those of several methodological studies in cancer 12 13 14 15 16 18 19 20 21 22 23 . When improvement over time has been explored, they have also found advances in some criteria, but at a rate slower than expected 16 18 . One study suggested that reporting is the main problem in trials of radiation therapy in oncology 13 .

Given standard treatments have been established by previous studies as effective and safe, 70% of included trials had an active comparator (as opposed to placebo or no intervention). In our assessment of risk of bias, we did not consider the use of an active comparator as inappropriate. Rather, we focused on whether those involved in the trial were blinded to allocation. While blinding of providers and participants is feasible in these cases, it is challenging in the situations where different routes of administration are used, and dosage adjustments are required. However, given the risk of bias associated with a lack of blinding, this could be overcome with double dummy designs, sham-adapted dosages and several other techniques 24 . Considering that lack of blinding can overestimate treatment effect estimates, it is important to make every effort to ensure blinding in studies 25 . Failure to report intention to treat analysis can incorporate systematic and unpredictable bias in a trial 26 . Other types of analysis (i.e. per protocol analysis) should not be presented in isolation. It is surprising that only the reporting of adequate allocation concealment and of intention to treat analysis significantly changed over time. However, and especially for the latter, there is ample room for improvement.

Only 31% of studies reported data on diagnosis of deep vein thrombosis triggered by clinical suspicion and the rest reported data on DVT screening. However, estimating absolute symptomatic events from studies that included screening may be misleading 27 . If screening is positive, patients will typically receive anticoagulation. If they had not received anticoagulation, some of them would likely have developed symptomatic VTE. This would underestimate the number of symptomatic VTE and the benefit of any intervention that reduces VTE. Bias in the opposite direction is also possible. A positive screening test may lead to interpreting any minor symptom, that otherwise might not have warranted investigation, as a symptomatic VTE. This would overestimate the number of symptomatic VTE, and the benefit of any intervention that reduces VTE. Others might argue that screening-identified asymptomatic may be more informative than misleading. Indeed, there is evidence that incidentally diagnosed PEs have similar prognostic value on survival as symptomatic PEs 28 29 . Moreover, false positive results should be minimized in the hands of experienced imaging centers thus avoiding some of the above concerns. More research is needed to shed more light on this controversial area.

Interestingly, 20% of trials do not report the source of funding. Of those who report, 3 of every 4 trials are completely or partially funded by private for profit sources. Industry funded trials were more likely to report intention to treat analysis, and less likely to report caregiver blinding. It is well known that industry sponsored trials favor products made by the company funding the research, but one of the more likely explanations is not because a higher risk of bias but through the selection of an inappropriate comparator to the product being investigated 30 31 .

We have found that the topic of the systematic review not to be associated with risk of bias. While this might reflect the absence of such association, it might be simply due to lack of statistical power given the relatively small number of studies included in the analysis.

Many RCTs of anticoagulation in patients with cancer appear to use insufficiently rigorous outcome assessment methods and to have deficiencies in key methodological features. It is not clear whether this reflects a problem in the design, conduct or the reporting of these trials, or both. Future trials should avoid the shortcomings described in this article.

The authors declare that they have no competing interests.

GR performed the statistical analysis and drafted the manuscript. HS participated in the design of the study and provided methodological expertise. NL, PE and VK extracted the data from individual studies. EA conceived the study, participated in its design and coordination, extracted the data from studies and helped to draft the manuscript. All authors read and approved the final manuscript.