Venous thromboembolism prophylaxis: do trial results enable clinicians and patients to evaluate whether the benefits justify the risk? proceedings of an ad hoc working group meeting

Abstract

Physicians and patients consider the balance between benefits and risks of treatment when making decisions about the use of anticoagulants for the prevention of venous thromboembolism (VTE). The results of early trials demonstrating the efficacy of heparin compared with placebo or no thromboprophylaxis for the prevention of a fatal pulmonary embolism (PE) led to the adoption of routine anticoagulant prophylaxis in patients considered to be at an increased risk of VTE. More recent trials comparing new anticoagulants with heparin have most commonly used the composite outcome, asymptomatic (or ‘silent’) deep vein thrombosis (DVT), detected by screening venography, and symptomatic (or ‘patient-important’) VTE, as the primary measure of efficacy [1-3]. The advantage of including an asymptomatic DVT detected by screening venography in the primary efficacy outcome of post-surgical prophylaxis trials is that it is much more common than symptomatic VTE, thus allowing much smaller sample sizes than would be required if the primary outcome was based on symptomatic VTE. The disadvantage is that physicians and patients do not know how to trade off an asymptomatic DVT against bleeding because asymptomatic events are of uncertain importance for patients [4]. A working group comprising clinicians, methodologists, regulators from the European Medicine Agency (EMA) and the Food and Drug Administration (FDA), and representatives from the pharmaceutical industry (see Appendix for list of delegates) met in Washington in 2009 to consider the balance between the benefits and risks of anticoagulants for the prevention of VTE. The objectives of this 1-day meeting were to (i) consider the validity of an asymptomatic DVT as a surrogate for symptomatic VTE and (ii) evaluate the trade-off between the prevention of VTE and risk of bleeding in trials of anticoagulant prophylaxis. This paper summarizes the meeting proceedings. A silent DVT detected by screening venography has the potential to be a valuable surrogate in thromboprophylaxis trials because it can be measured more easily and occurs much more frequently than the symptomatic VTE [5]. The disadvantage of including an asymptomatic DVT in the assessment of efficacy is that its relationship to outcomes of importance to patients (VTE that requires treatment with anticoagulants, a fatal PE and post-thrombotic syndrome) is difficult to quantify [6, 7]. The acceptance of an asymptomatic DVT as a valid surrogate measure for symptomatic VTE is based on: (i) the biological link between an asymptomatic and symptomatic VTE; (ii) the accuracy of venography for the diagnosis of an asymptomatic DVT; (iii) the similarity between pooled estimates of the relative effect of anticoagulant prophylaxis for an asymptomatic DVT and symptomatic VTE; and (iv) an assumption that we know the quantitative relationship between an asymptomatic DVT and symptomatic VTE. The biological link between an asymptomatic DVT and symptomatic VTE is supported by indirect comparisons of efficacy assessments based on DVT detected by screening radioactive fibrinogen leg scanning (FLS) and screening venography, and symptomatic VTE diagnosed by objective testing. Studies using FLS indicate that the incidence of a DVT in bed-ridden patients is high, that most DVT start in the calf [8] and, if that untreated, these asymptomatic DVTs may either propagate or undergo spontaneous lysis. The incidence of a DVT on screening venography performed 1 week after major surgery is lower than that detected by FLS, presumably because venography only detects the larger DVTs that have not undergone spontaneous lysis. The incidence of symptomatic VTE is much lower than that of asymptomatic DVTs, and most symptomatic thrombi declare themselves after the first week and are larger than asymptomatic DVTs [8]. PEs can be derived from either asymptomatic or symptomatic DVTs, but symptomatic (patient important) PEs tends to be associated with larger DVTs. Based on data derived from the above studies, it is reasonable to accept the view that symptomatic VTE originates from an asymptomatic DVT. Venography is regarded as the gold standard for the detection of a DVT. However, this may not necessarily be the case in all settings. While inter-observer agreement for a venographically-detected DVT is good within individual centers [9], rates reported in patients of similar age, undergoing the same type of operation and receiving the same thromboprophylaxis regimens differ by as much as two-fold between adjudication centers [10]. Pooled data from randomized trials indicate that on average, anticoagulant prophylaxis produces proportionately similar risk reductions in asymptomatic DVTs and symptomatic VTE [11]. It is possible, however, for average effects to be the same but for true underlying effects in individual trials or sets of trials to show important discrepancies. Indeed, individual examples of large differences in the effect of post-surgical thromboprophylaxis on asymptomatic and symptomatic events exist [12, 13]. More powerful than a demonstration of average effects being similar would be a very high level of agreement in individual trials or sets of trials. Moreover, estimates based on trials using screening venography are subject to two competing biases. On the one hand, patients known to have an asymptomatic DVT may be more likely to be diagnosed with symptomatic VTE than those with a normal venogram. On the other hand, patients diagnosed with an asymptomatic DVT may be less likely to develop symptomatic VTE because they receive anticoagulant treatment for the asymptomatic DVT. These potential biases reduce our confidence in the estimates of treatment effect from trials that include an asymptomatic DVT in the primary outcome. The incidence of an asymptomatic venographically detected DVT is much higher than the incidence of symptomatic VTE. Although various attempts have been made to quantify the numerical relationship between these two outcomes, the true relationship is uncertain, and without such knowledge, estimates of the trade-off between risks and benefits of anticoagulant prophylaxis are imprecise. Perspectives from guideline groups and agencies reflect uncertainty about the value of an asymptomatic DVT detected by screening venography as a surrogate for symptomatic VTE (Table 1) [1, 4, 14-16]. The 2012 American College of Chest Physicians (ACCP) guidelines make strong recommendations for anticoagulant prophylaxis in hospitalized surgical and high-risk medical patients [16, 17] but rate down the quality of the evidence if the efficacy outcome from randomized controlled trials is based on an asymptomatic DVT. The UK National Institute for Health and Clinical Excellence (NICE) [1] also rate down the quality of evidence if the efficacy outcome from randomized trials is based on an asymptomatic DVT. The 2011 American Association of Orthopedic Surgeons (AAOS) guidelines for the prevention of VTE in patients undergoing orthopedic surgery largely reject evidence from trials that use screening venography [4]. The EMA has stated that it considers an asymptomatic proximal DVT but not asymptomatic distal DVT as an appropriate outcome for drug approval [14]. The Food and Drugs Administration grants new drug approvals largely based on the results of randomized trials that report an asymptomatic DVT diagnosed by screening venography. They consider the totality of venographic data, including the clinical consequences associated with the location of the DVT. Physicians and patients need to take into account the incidence and consequences of VTE and bleeding when making decisions about the use of anticoagulant thromboprophylaxis. When symptomatic, a DVT causes leg pain and swelling. Both an asymptomatic and symptomatic DVT can result in a pulmonary embolism (PE). Patients who develop a DVT or PE are generally prescribed anticoagulant treatment for at least 3–6 months [18]. The long-term consequences of an asymptomatic DVT are uncertain but up to 40% of symptomatic DVTs are associated with the development of post-thrombotic syndrome, which is debilitating in about 15% of affected patients [19, 20]. The most important consequence of a PE is death, and in the long term ∼3% of PE patients develop thromboembolic pulmonary hypertension [21]. Bleeding at the operative wound site post-surgery is generally considered to be the most important adverse event associated with the use of anticoagulant prophylaxis. Wound bleeding can cause local pain and swelling and increase the risk of an infection, which may lead to the need for blood transfusion, often prolongs hospitalization and on rare occasions requires further surgery. Orthopedic surgeons have expressed concerns that anticoagulants cause bleeding into prosthetic joints and thereby impair functional outcomes but there are no reliable data on the frequency of this problem. In spite of making treatment recommendations based on the results of venographic trials, the NICE guidelines do not provide guidance on the assessment of the trade-off between an asymptomatic DVT and wound bleeding [1]. Instead, the NICE guidelines take into account patient bleeding risk and stratify recommendations for thromboprophylaxis according to whether or not patients are considered to be at an elevated risk for bleeding. By largely disregarding the results of venographic trials, the 2011 AAOS guidelines avoid the need to trade-off a reduction in an asymptomatic DVT against bleeding. The 2012 ACCP guidelines offer explicit strategies for estimating the absolute risk reduction in symptomatic VTE when trials have focused on asymptomatic events and explicit tradeoffs between symptomatic VTE and bleeding [3]. Based on the totality of the evidence, and particularly the evidence from trials demonstrating that anticoagulant prophylaxis prevents fatal PE, this class of agents has been widely accepted by regulators, guidelines, clinicians and patients for the prevention of VTE. When new anticoagulants are compared with accepted agents, it is reasonable to approve them if they show equivalent (or increased) efficacy, based only on screening venographic outcomes, as long as the new agent is not associated with increased bleeding. When new anticoagulants compared with accepted agents show increased efficacy based only on screening venography outcomes, but also increased bleeding, the trade-off between risks and benefits of anticoagulant prophylaxis is uncertain. Uncertainty about the consequences of local bleeding events further complicates evaluation of the trade-off between reduced VTE and increased bleeding. In order to clarify this issue, investigators in future orthopedic prophylaxis trials should try to objectively quantify post-surgical wound bleeding and follow patients long term (ideally for 1 year) to determine the effect, if any, of wound bleeding on the risk of infection and long-term joint function. The efficacy and safety of aspirin relative to anticoagulants remains an important unresolved question. Aspirin was shown to be more effective than placebo for VTE prevention in the Pulmonary Embolism Prevention (PEP) [22] study. A similar large trial focusing on symptomatic VTE and avoiding screening tests is required to assess the relative effectiveness and safety of aspirin (an inexpensive drug) with an established anticoagulant. Meeting Participants J.W. Eikelboom and J. Hirsh (Co-Chairs); L. Anderson, R. Becker, S. Berkowitz, H. Büller, R. Califf, N. Cater, M. Crowther, W.L. Daly, C. Delagrange, B. Donohue, D. Garcia, M. Geraldes, E. Gupta, G. Guyatt, R. Harrington, J. Heit, G. Karthikeyan, R. Knabb, J.H. Lawrence, J. Lieberman, P. Mendys, M. Molbert, J. Paolini, F. Plat, D.J. Quinlan, I. Richard-Lordereau, D. Rieves, C. Rusconi, C.M. Samama, W. Sapirstein, L.R. Schwocho, M. Sobieraj-Teague, M. Stein, N. Stockbridge, R. Temple, T. Treasure and S. Zelenkofske. University of Pennsylvania, Philadelphia Pennsylvania USA (JB); McMaster University, Hamilton, Ontario, Canada (J.W.E., G.G., M.S., J.H. and M.C.); Academic Medical Center, Amsterdam, The Netherlands (H.B.); Duke Clinical Research Institute, Durham, North Carolina, USA (R.H., R.B., B.D. and M.M.); GlaxoSmithKline R&D, Harlow, Essex, England (L.A.); Bayer HealthCare Pharmaceuticals, Montville, New Jersey, USA (S.B. and P.J.); Duke University Medical School, Durham, North Carolina USA (R.C.); Pfizer, Inc. New York, New York, USA (N.C.); Sanofi-Aventis, Bridgewater, New Jersey, USA (W.L.D.); Daiichi Sankyo, Parsippany, New Jersey, USA (C.D. and M.S.); University of New Mexico, Albuquerque, New Mexico, USA (D.G.); Bristol-Myers Squibb, Pennington, New Jersey, USA (M.G.); Bristol-Myers Squibb, Princeton, New Jersey, USA (E.G., R.K. and J.H.L.); Mayo Clinic, Rochester, Minnesota, USA (J.H.); All India Institute of Medical Sciences, New Delhi (K.G.); University of Connecticut Health Center, Farmington, Connecticut, USA (J.L.); Pfizer, Inc. Chapel Hill, North Carolina USA (P.M.); Daiichi Sankyo Pharma Development, Edison, New Jersey, USA (F.P. and L.R.S.); Kings College Hospital, London, United Kingdom (D.J.Q.); Bristol Myers Squibb, France (I.R.L.); FDA, Center for Drug Evaluation & Research, Silver Spring, Maryland, USA (D.R., N.S. and R.T.); Regado Biosciences, Durham, North Carolina, USA (C.R. and S,Z.); Hotel-Dieu University Hospital, Paris, France (C.M.S.); Office of Device Evaluation, Rockville, Maryland, USA (S.W.); Clinical Operational Research Unit, UCL, London, United Kingdom (T.T.). All authors participated in the preparation of the manuscript and agreed to the submitted version of the paper. The role and contribution of each author was as follows: Drafting of the manuscript: Gordon Guyatt, John Eikelboom. Critical revision of the manuscript for important intellectual content: all authors. We thank Marelle Molbert for organizing the meeting. The authors state that they have no conflict of interest.