Thromboembolic and bleeding risk of periprocedural bridging anticoagulation: A systematic review and meta-analysis.

Abstract The risk and benefit of periprocedural heparin bridging is not completely clarified. We aimed to assess the safety and efficacy of bridging anticoagulation prior to invasive procedures or surgery. Heparin bridging was associated with lower risks of thromboembolism and bleeding compared to non‐bridging. PubMed, Ovid and Elsevier, and Cochrane Library (2000‐2016) were searched for English‐language studies. Studies comparing interrupted anticoagulation with or without bridging and continuous oral anticoagulation in patients at moderate‐to‐high thromboembolic risk before invasive procedures were included. Primary outcomes were thromboembolic events and bleeding events. Mantel‐Haenszel method and random‐effects models were used to analyze the pooled risk ratio (RR) and 95% confidence interval (CI) for thromboembolic and bleeding risks. Eighteen studies (six randomized controlled trials and 12 cohort studies) were included (N = 23 364). There was no difference in thromboembolic risk between bridged and non‐bridged patients (RR: 1.26, 95% CI: 0.61‐2.58; RCTs: RR: 0.71, 95% CI: 0.23‐2.24; cohorts: RR: 1.45, 95% CI: 0.63‐3.37). However, bridging anticoagulation was associated with higher risk of overall bleeding (RR: 2.83, 95% CI: 2.00‐4.01; RCTs: RR: 2.24, 95% CI: 0.99‐5.09; cohorts: RR: 3.09, 95% CI: 2.07‐4.62) and major bleeding (RR: 3.00, 95% CI: 1.78‐5.06; RCTs: RR: 2.48, 95% CI: 1.29‐4.76; cohorts: RR: 3.22, 95% CI: 1.65‐6.32). Bridging anticoagulation was associated with increased bleeding risk compared to non‐bridging. Thromboembolism risk was similar between two strategies. Our results do not support routine use of bridging during anticoagulation interruption.


| INTRODUCTION
An estimated 2.5 million patients use oral anticoagulants for the prevention of arterial thromboembolic events in North America, and onetenth of them require temporary interruption in preparation for an elective procedure or surgery. 1,2 However, the safety and efficacy of bridging anticoagulation is not completely clarified for patients who need an anticoagulation interruption before invasive procedures. Two main concerns remain unsolved, the risk of thromboembolism, and the risk of bleeding. 1,2 To reduce the bleeding risk for patients undergoing invasive procedures, oral anticoagulant is typically interrupted prior to the procedure, and then continued when hemostasis is achieved postprocedurally. 1,2 Because the interruption of anticoagulation may expose patients to the risk of thromboembolism, heparin bridging (unfractionated heparin [UFH] or low-molecular-weight heparin [LMWH]) is administered to minimize the period of inadequate level of anticoagulation. 1,2 It is of great importance to confirm whether bridging therapy reduces thromboembolic risk and to ascertain the safety of bridging therapy in relation to bleeding risk. 3 There have been many published articles related to bridging anticoagulation, 4,5 but the quality of evidence with best practices is uneven across studies. Current guidelines from the 2019 American College of Cardiology/American Heart Association suggest bridging anticoagulation used in patients with a high thrombosis risk, such as certain mechanical valve prostheses or recent pulmonary embolism during interruption of vitamin K antagonist (VKA) therapy. 3 However, these recommendations are primarily based on observational studies and experts' opinions. 3 Although Siegal and colleagues concluded that bridging anticoagulation increases bleeding risk and produces similar thromboembolic risk, their review included only one underpowered randomized trial together with some observational studies with no control arm to assess the safety and efficacy of bridging therapy. 6 To better clarify the risk and benefit of bridging therapy, we updated the current published data and conducted a meta-analysis to compare the periprocedural thromboembolic and hemorrhagic risks between patients receiving interrupted anticoagulation with or without bridging therapy and continuous oral anticoagulation.

| Data sources and searches
We used the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines 7 and performed a search of PubMed, Ovid and Elsevier, and Cochrane Library for published randomized and observational studies in English from January 1, 2000 to August 30, 2016. We searched with keywords "long-term oral anticoagulant," "chronic oral anticoagulant," "periprocedural anticoagulant," "perioperative anticoagulant," "uninterrupted anticoagulant," "continued oral anticoagulant," "interrupted anticoagulant," "unfractionated heparin bridging," and "low-molecular-weight heparin bridging." References of articles and previous meta-analyses were also reviewed to confirm that no studies were missed.

| Data retrieval and quality evaluation
Included studies met all of the following criteria: patients ≥18 years of age, long-term use of oral anticoagulant before a procedure, elective invasive procedure or surgery, comparison of periprocedural bridging anticoagulation and non-bridging management (continuous oral anticoagulation or interrupted oral anticoagulation without bridging therapy), reporting of both thromboembolic and bleeding events, and articles published in peer-reviewed journals. Studies with unclear reporting of thromboembolic or bleeding events, case reports, or letters to the editor were excluded. Studies with no control arm or designed specific to patients using novel oral anticoagu- Guidelines. 8 Disagreements on data acquisition were resolved by consensus with the third reviewer.

| Data synthesis and analysis
Data were pooled by using the Mantel-Haenszel method, and the random-effects model was performed to generate risk ratios (RRs) by using the RevMan software, version 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration; 2014). 9 The I 2 statistics was used to check for quantitative heterogeneity of results; it defines low heterogeneity with I 2 < 25%, moderate heterogeneity with I 2 between 25% and 50%, and high heterogeneity with I 2 more than 50%. A funnel plot was used to assess a potential publication bias. Visual estimation was performed for the asymmetry of the funnel plot. Descriptive statistics were generated with the SAS software, version 9.4 (SAS Institute Inc., Cary, North Carolina).

| Risk of bleeding
The overall bleeding risk in the bridged group was significantly higher compared to non-bridged group regardless of study design with high heterogeneity of results, RR: 2.83 (95% CI: 2.00-4.01, P < .0001; I 2 = 71%; Figure 3). Funnel plot manufactured for the outcome of bleeding events showed no obvious publication bias.
The criteria for major bleeding were revealed in nine studies (two randomized trials, seven cohort studies). 10 The analysis showed a higher risk of major bleeding associated with bridging anticoagulation compared to non-bridging strategy with a high level of heterogeneity, RR: 3.00 (95% CI: 1.78-5.06, P < .0001; I 2 = 57%; Figure 4).

| DISCUSSION
Our analyses showed bridging anticoagulation was associated with a nearly threefold increased risk of overall and major bleeding compared to non-bridging management and there was no significant difference in the risk of thromboembolism between the two strategies. Findings were broadly consistent across subgroups irrespective of study design.
In this study, we critically selected relevant studies and conducted a rigorously scientific synthesis of research results. Prior metaanalyses evaluating the risk and benefit of bridging anticoagulation are flawed with inclusions of few randomized trials 6 and studies with no control arm, 6 no separate analysis of three strategies (interrupted anticoagulation with bridging therapy, interrupted anticoagulation without bridging therapy and continuous oral anticoagulation), 6,28 and no detailed assessment of study quality. 28 We found that bridging anticoagulation was associated with higher bleeding risk and similar thromboembolic risk compared to non-bridging strategy. Our subgroup analyses further showed the higher bleeding risk of bridging therapy seemed to be augmented in patients receiving interrupted anticoagulation without bridging therapy compared to those with continuous anticoagulation.
Our analysis demonstrated the risk of periprocedural thromboembolic events was 0.89% and 0.46% for the bridged and non-bridged patients, respectively, and the difference did not reach statistical significance. Some possible reasons may explain this finding. First, most patients were classified as having a moderate risk of thromboembolism in individual studies; however, the risk-stratification criteria were inconsistent across studies, which may particularly bias the results of observational studies, for example, indication bias (patients in a higher thromboembolic risk may be given a bridging anticoagulation). Second, the overall number of thromboembolic events was small, which may yield underpowered statistics. Third, interruption and reinitiation of warfarin may deplete protein C and protein S and thereafter contribute to a hyper-coagulable status. Protein C and S are two vitamin Kdependent plasma proteins that work together as a natural anticoagulant system, and deficiency in proteins C and S is associated with thrombotic tendency. 29 Fourth, it takes at least 5 days for the international normalized ratio to normalize after stopping warfarin 29 and NOACs are eliminated in 48 to 72 hours after discontinuation 30 ; therefore, in patients who discontinued warfarin for <5 days or NOACs <48 hours, the risk of thromboembolism may remain low due to residual anticoagulation.
Although thromboembolism may cause severe morbidity and mortality, it should be weighed against the bleeding risk of bridging anticoagulation. 31 Considering the higher risk of bleeding related to bridging anticoagulation, our results suggested non-bridging management seems to have a favorable risk-benefit profile in terms of thromboembolic and bleeding complications. There are some explanations for the bleeding risk of bridging anticoagulation. First, residual anti-Xa effect or heparin-induced thrombocytopenia may contribute to postoperative bleeding. 32 Second, due to the interindividual variability in the sensitivity of aPTT test (the most common laboratory measurement to monitor UFH), control of aPTT range may not correlate well with the activity of bridging anticoagulation. 33 NOACs have been increasingly used for the prevention of thromboembolic events in patients with moderate-to-high risk. 32 NOACs are non-inferior for prevention of stroke in patients with atrial fibrillation and associated with less bleeding compared to warfarin. 34 NOACs have advantages of short half-lives, fast onset of action, predictable pharmacokinetic properties (concentration-dependent), and few drug-drug interactions. 35 Although the experience of periprocedural use of NOACs is accumulating, there are limited data available pertaining to the bridging anticoagulation for patients on NOAC therapy in terms of perioperative bleeding risk. 36 The Dresden NOAC registry revealed major bleeding rate of 1.2% and clinically relevant nonmajor bleeding rate of 3.4% in patients using NOACs during invasive procedures. 37 A subgroup analysis of the RE-LY trial showed no significant difference in the risk of periprocedural major bleeding between patients using dabigatran 110 mg (3.8%), dabigatran 150 mg (5.1%), or warfarin (4.6%). 38 However, these studies did not compare the bleeding risk between bridging and non-bridging therapy. 37,38 Whether it is better for patients using NOACs to receive bridging anticoagulation is currently unclear. Two specific reversal agents for NOACs have been approved in the United States: idarucizumab for dabigatran reversal and andexanet alfa for apixaban and rivaroxaban reversal. 39,40 Tailoring periprocedural management of NOACs to the type of invasive procedure may reduce the risk of bleeding.
Attention to some limitations of this study is needed. First, 12 out of the 18 studies were observational, and the number of patients enrolled in the randomized controlled trials was relatively small. Second, the heterogeneity of results among studies was high, which may relate to the variations in the preoperative thromboembolic risks, types of procedure and definitions of outcomes. Third, few patients were classified as high thromboembolic risk, 13,20,22,25,26 and the safety and benefit of bridging anticoagulation among these patients remain uncertain. Fourth, most of the included studies were relevant to warfarin-treated patients; therefore, the results cannot be generalized to patients receiving NOACs. Fifth, types and doses of bridging regimens (UFH or LWMH; prophylactic dose or therapeutic dose), and the timing of periprocedural initiation of bridging were different across studies. Finally, we are unable to conduct subgroup analyses of high-risk vs low-risk procedures and outcomes of fatal bleeding due to unavailability of individual data of the included studies.
In conclusion, bridging anticoagulation was associated with increased bleeding risk compared to non-bridging management.
Besides, thromboembolism risk was similar between these two strategies. Our results do not support the use of routine bridging during the periprocedural interruption of oral anticoagulation.

CONFLICT OF INTEREST
The authors declare no potential conflict of interest.

AUTHOR CONTRIBUTIONS
H.C.K contributed to data acquisition, formal analysis, and manuscript writing. F.L.L. contributed to data acquisition, statistical consultation, and review. J.T.C., Y.G.C., and K.W.T. contributed to manuscript revision. Y.H.T. contributed to data verification and manuscript revision.