Evidence-Based Management of Anticoagulant Therapy

Abstract

Background:

High-quality anticoagulation management is required to keep these narrow therapeutic index medications as effective and safe as possible. This article focuses on the common important management questions for which, at a minimum, low-quality published evidence is available to guide best practices.

Methods:

The methods of this guideline follow those described in Methodology for the Development of Antithrombotic Therapy and Prevention of Thrombosis Guidelines: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines in this supplement.

Results:

Most practical clinical questions regarding the management of anticoagulation, both oral and parenteral, have not been adequately addressed by randomized trials. We found sufficient evidence for summaries of recommendations for 23 questions, of which only two are strong rather than weak recommendations. Strong recommendations include targeting an international normalized ratio of 2.0 to 3.0 for patients on vitamin K antagonist therapy (Grade 1B) and not routinely using pharmacogenetic testing for guiding doses of vitamin K antagonist (Grade 1B). Weak recommendations deal with such issues as loading doses, initiation overlap, monitoring frequency, vitamin K supplementation, patient self-management, weight and renal function adjustment of doses, dosing decision support, drug interactions to avoid, and prevention and management of bleeding complications. We also address anticoagulation management services and intensive patient education.

Conclusions:

We offer guidance for many common anticoagulation-related management problems. Most anticoagulation management questions have not been adequately studied.

Summary of Recommendations

Note on Shaded Text: Throughout this guideline, shading is used within the summary of recommendations sections to indicate recommendations that are newly added or have been changed since the publication of Antithrombotic and Thrombolytic Therapy: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines (8th Edition). Recommendations that remain unchanged are not shaded.

2.1. For patients sufficiently healthy to be treated as outpatients, we suggest initiating vitamin K antagonist (VKA) therapy with warfarin 10 mg daily for the first 2 days followed by dosing based on international normalized ratio (INR) measurements rather than starting with the estimated maintenance dose (Grade 2C).

2.2. For patients initiating VKA therapy, we recommend against the routine use of pharmacogenetic testing for guiding doses of VKA (Grade 1B).

2.3. For patients with acute VTE, we suggest that VKA therapy be started on day 1 or 2 of LMWH or UFH therapy rather than waiting for several days to start (Grade 2C).

3.1. For patients taking VKA therapy with consistently stable INRs, we suggest an INR testing frequency of up to 12 weeks rather than every 4 weeks (Grade 2B).

3.2. For patients taking VKAs with previously stable therapeutic INRs who present with a single out-of-range INR of ≤ 0.5 below or above therapeutic, we suggest continuing the current dose and testing the INR within 1 to 2 weeks (Grade 2C).

3.3. For patients with stable therapeutic INRs presenting with a single subtherapeutic INR value, we suggest against routinely administering bridging with heparin (Grade 2C).

3.4. For patients taking VKAs, we suggest against routine use of vitamin K supplementation (Grade 2C).

3.5. (Best Practices Statement) We suggest that health-care providers who manage oral anticoagulation therapy should do so in a systematic and coordinated fashion, incorporating patient education, systematic INR testing, tracking, follow-up, and good patient communication of results and dosing decisions.

3.6. For patients treated with VKAs who are motivated and can demonstrate competency in self-management strategies, including the self-testing equipment, we suggest patient self-management (PSM) rather than usual outpatient INR monitoring (Grade 2B). For all other patients, we suggest monitoring that includes the safeguards in our best practice statement 3.5.

3.7. For dosing decisions during maintenance VKA therapy, we suggest using validated decision support tools (paper nomograms or computerized dosing programs) rather than no decision support (Grade 2C).

Remarks: Inexperienced prescribers may be more likely to improve prescribing with use of decision support tools than experienced prescribers.

3.8. For patients taking VKAs, we suggest avoiding concomitant treatment with nonsteroidal antiinflammatory drugs (NSAIDs), including cyclooxygenase (COX)-2-selective NSAIDs, and certain antibiotics (see Table 8) (Grade 2C).

Table 8.

—[Section 3.8] Drug Interactions With VKAs^a: Drug Families Associated With Increased Risk of Bleeding

Interacting Drug	Summary Effect on Bleeding (95% CI)b	Study
NSAIDS
NSNSAIDs	OR 1.9 (1.4-3.7)	Battistella et al100
	HR 3.6 (2.3-5.6)	Cheetham et al101
	RR 1.33 (0.78-2.25)	Delaney et al102
	OR 2.6 (1.6-4.2)	Hauta-Ato et al103
	OR 3.01 (1.42-6.37)	Knijff-Dutmer et al104,a
	RR 2.6-6.5c	Penning-van Beest et al105,a
	OR 4.6 (3.3-6.5)d	Schalekamp et al106
	NSNSAID vs COX-2 OR 3.07 (1.18-8.03)	Knijff-Dutmer et al104,a
	NSNSAIDs vs COX-2 HR 3.7 (1.4-9.6)	Cheetham et al101
COX-2-selective NSAIDs	OR 1.7-2.4e	Battistella et al100
	HR 1.7 (0.6-4.8)	Cheetham et al101
	RR 1.37 (0.44-4.30)	Delaney et al102
	OR 3.1 (1.4-6.7)	Hauta-Aho et al103
Antiplatelet agents
Aspirin	OR 1.43 (1.00-2.02)f	Dentali et al99
	RR 2.23 (1.46-3.41)	Delaney et al102
	IR 0.08/patient-y vs 0.06 for warfarin alone	Buresly et al107
	HR 1.83 (1.72-1.96)	Hansen et al108
	RR 3.0 (1.0-9.4)	Penning-van Beest et al105,a
Clopidogrel	HR 3.08 (2.3-3.9)	Hansen et al108
Aspirin plus clopidogrel	HR 3.70 (2.89-4.76)	Hansen et al108
Antiplatelet agents (any antiplatelet)	OR 2.06 (1.01-4.36)	Johnson et al109
	OR 1.53 (1.05-2.22)	Shireman et al110
	RR 1.76 (1.05-2.95)	Toyoda et al111
Antibiotics
Cephalexin	OR 1.38 (1.10-1.73)	Schelleman et al112
Cefradine	RR 43.0 (10.7-172.4)	Penning-van Beest et al113,a
Cephalosporins	OR 1.16 (1.04-1.29)	Zhang et al114
Metronidazole	OR 1.58 (1.32-1.89)	Zhang et al114
Cotrimoxazole	OR 3.84 (2.33-6.33)	Fischer et al115
	OR 2.54 (2.08-3.10)	Schelleman et al112
	RR 5.1 (2.1-12.3)	Penning-van Beest et al113,a,g
	Cotrimox vs cephalexin OR 1.68 (1.21-2.33)	Schelleman et al112
Ciprofloxacin	OR 1.94 (1.28-2.95)	Fischer et al115
	OR 1.62 (1.31-1.99)	Schelleman et al112
	RR 3.2 (1.3-7.7)	Penning-van Beest et al113,a
Levofloxacin	OR 1.55 (1.30-1.86)	Schelleman et al112
Norfloxacin	RR 5.9 (1.9-18.6)	Penning-van Beest et al105,a
Amoxycillin	OR 1.28 (1.03-1.58)	Schelleman et al112
	RR 3.1 (1.6-6.3)	Penning-van Beest et al113,a,g
Amoxycillin/clavulanic acid	RR 4.4 (2.5-7.8)	Penning-van Beest et al113,a,g
Doxycycline	RR 2.6 (1.2-4.8)	Penning-van Beest et al113,a
Fluconazole	OR 1.89 (1.35-2.64)	Schelleman et al112
	Fluconazole vs cephalexin OR 2.09 (1.34-3.26)	Schelleman et al112
Other
SSRIs	OR 2.6 (1.5-4.3)	Hauta-Aho et al103
	OR 1.7 (1.1-2.5)h	Schalekamp et al106,a
	OR 1.1 (0.9-1.4) to 1.2 (0.8-1.7); NS	Kurdyak et al116
Tramadol	RR 3.3 (1.1-10.4)	Penning-van Beest et al105,a
Complementary medicines	Coenzyme Q10 (OR 3.69, 95% CI 1.88-7.24) and ginger (OR 3.20, 95% CI 2.42-4.24).	Shalansky et al117

COX = cyclooxygenase; IR = incidence rate; NSNSAID = nonselective nonsteroidal antiinflammatory drug; SSRI = selective serotonin reuptake inhibitor. See Table 1 and 3 legends for expansion of other abbreviations.

^aStudy VKAs were warfarin, phenprocoumon, and acenocoumarol.

^bUnless stated, refers to drug plus VKA vs VKA alone.

^cDiclofenac (RR, 2.6) and naproxen (RR, 6.5) studied separately.

^dOR is for GI bleeding, whereas OR for non-GI bleeding is 1.7 (95% CI, 1.3-2.2).

^eSeparate OR given for celecoxib (1.7) and rofecoxib (2.4); both statistically significant.

^fDentali et al⁹⁹ meta-analysis of randomized clinical trials.

^gData duplication between two study publications; therefore, more conservative estimate used.

^hOnly statistically significant for non-GI bleeding; not significant for GI bleeding or intracranial bleeding.

For patients taking VKAs, we suggest avoiding concomitant treatment with antiplatelet agents except in situations where benefit is known or is highly likely to be greater than harm from bleeding, such as patients with mechanical valves, patients with acute coronary syndrome, or patients with recent coronary stents or bypass surgery (Grade 2C).

4.1. For patients treated with VKAs, we recommend a therapeutic INR range of 2.0 to 3.0 (target INR of 2.5) rather than a lower (INR < 2) or higher (INR 3.0-5.0) range (Grade 1B).

4.2. For patients with antiphospholipid syndrome with previous arterial or venous thromboembolism, we suggest VKA therapy titrated to a moderate-intensity INR range (INR 2.0-3.0) rather than higher intensity (INR 3.0-4.5) (Grade 2B).

5.0. For patients eligible to discontinue treatment with VKA, we suggest abrupt discontinuation rather than gradual tapering of the dose to discontinuation (Grade 2C).

6.1. For patients starting IV unfractionated heparin (UFH), we suggest that the initial bolus and the initial rate of the continuous infusion be weight adjusted (bolus 80 units/kg followed by 18 units/kg per h for VTE; bolus 70 units/kg followed by 15 units/kg per h for cardiac or stroke patients) or use of a fixed dose (bolus 5,000 units followed by 1,000 units/h) rather than alternative regimens (Grade 2C).

6.2. For outpatients with VTE treated with subcutaneous (SC) UFH, we suggest weight-adjusted dosing (first dose 333 units/kg, then 250 units/kg) without monitoring rather than fixed or weight-adjusted dosing with monitoring (Grade 2C).

7.1. For patients receiving therapeutic LMWH who have severe renal insufficiency (calculated creatinine clearance < 30 mL/min), we suggest a reduction of the dose rather than using standard doses (Grade 2C).

8.1. For patients with VTE and body weight over 100 kg, we suggest that the treatment dose of fondaparinux be increased from the usual 7.5 mg to 10 mg daily SC (Grade 2C).

9.1.

(a) For patients taking VKAs with INRs between 4.5 and 10 and with no evidence of bleeding, we suggest against the routine use of vitamin K (Grade 2B).

(b) For patients taking VKAs with INRs > 10.0 and with no evidence of bleeding, we suggest that oral vitamin K be administered (Grade 2C).

9.2. For patients initiating VKA therapy, we suggest against the routine use of clinical prediction rules for bleeding as the sole criterion to withhold VKA therapy (Grade 2C).

9.3. For patients with VKA-associated major bleeding, we suggest rapid reversal of anticoagulation with four-factor prothrombin complex concentrate (PCC) rather than with plasma. (Grade 2C).

We suggest the additional use of vitamin K 5 to 10 mg administered by slow IV injection rather than reversal with coagulation factors alone (Grade 2C).

This article deals with the evidence regarding managing anticoagulant therapy, that is, oral vitamin K antagonists (VKAs), heparins, and fondaparinux. Separate articles address the pharmacology of these drugs.¹ The questions that we address reflect those commonly posed in clinical practice.

1.0 Methods

The methods for the development of this article’s recommendations follow those developed for the Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines.² Although we aimed to summarize and use randomized controlled trial (RCT) evidence to inform recommendations for clinicians, we found only lower-quality evidence to address most of our questions. At the onset of our review process, our panel decided to limit the recommendations to questions in which evidence met a minimum threshold for quality: at least one comparative study with ≥ 50 patients per group with contemporaneous or historical controls reporting on patient-important outcomes or closely related surrogates. Despite this low threshold, evidence was unavailable for several important clinical management questions. When randomized trials were available, confidence in estimates often decreased because of indirectness (surrogate outcomes) and imprecision (wide CIs).

This article does not address anticoagulation management issues specific to pregnancy or to children. Issues believed to be specific to a particular diagnosis, such as VTE or atrial fibrillation, are dealt with in those specific articles of this supplement. Table 1 presents the questions for which we found evidence that met our quality threshold, including the relevant populations, interventions, comparators, and outcomes.

Table 1.

—Structured Clinical Questions

Section	Informal Question	PICO				Comment
		Population	Intervention	Comparator	Outcome
2.0 VKAs—initiation of therapy
2.1. Loading doses of VKA	Is a loading dose of VKA superior to no loading dose?	Patients taking VKA	Loading dose	No loading dose	Hemorrhage, thromboembolic events, time to therapeutic range, rates of supratherapeutic or subtherapeutic INR
2.2. Dose by Pharmacogenetics	Should the initial dose of VKA be based on pharmacogenetic testing?	Patients taking VKA	Analysis of CYP2C9, VKORC1, and other polymorphisms	No pharmacogenetic testing	Hemorrhage, thromboembolic events, time to therapeutic range, rates of supratherapeutic or subtherapeutic INR
2.3. Initiation overlap	Should VKA be started simultaneously with heparin rather than delayed a few days?	Patients treated for acute thromboembolism (or other high-risk situation requiring long-term VKA)	Simultaneous start	Initial heparin followed by overlap with VKA	Hemorrhage, thromboembolic events, time to therapeutic range, rates of supratherapeutic or subtherapeutic INR, resource utilization (hospital stay)
3.0 VKAs—maintenance treatment
3.1. INR monitoring frequency	How frequently should treatment be monitored initially and once dose and INR have been stable for months?	Patients taking VKA	Higher frequency	Lower frequency	Hemorrhage, thromboembolic events, time to therapeutic range
3.2. Single out-of-range INR—dose adjustment	Should the VKA dose change for a single deviating INR in otherwise stable patients?	Patients taking VKA	Dose adjustment	Continue as usual	Hemorrhage, thromboembolic events, time in therapeutic range
3.3. Bridging for subtherapeutic INR	Does bridging anticoagulant therapy improve outcomes for low INR?	Patients taking VKA with subtherapeutic INR	Dose management and overlapping with heparin	Only dose management	Hemorrhage, thromboembolic events, time in therapeutic range
3.4. Vitamin K supplementation	Can outcomes be improved with low-dose vitamin K supplementation or dietary manipulation?	Patients taking VKA with variability of INR	Cotherapy with small-dose vitamin K or with dietary modification	No vitamin K	Hemorrhage, thromboembolic events, time in therapeutic range
3.5. Dose management services	Dose management services: does a specialized AMS improve outcomes?	Patients taking VKA	AMS care	Usual care (primary care or regular hospital physicians)	Hemorrhage, thromboembolic events, time in therapeutic range, resource utilization
3.6. Patient self-testing and self-monitoring	Does self-monitoring of anticoagulation improve outcomes	Patients taking VKA	Use of point-of-care monitor at home to measure INR and to adjust VKA dose	Usual care or AMS care	Hemorrhage, thromboembolic events, time in therapeutic range, resource utilization
3.7. Dosing decision support	Does dosing decision support improve outcomes	Patients taking VKA	Computer software, manual algorithms	Usual care	Hemorrhage, thromboembolic events, time in therapeutic range, resource utilization
3.8. Drug interactions to avoid	What anticoagulant drug or food interactions are important enough to avoid the interacting drug while patients take anticoagulants	Patients taking anticoagulants	Patients starting or stopping potentially interacting drugs	Patients not taking potentially interacting drugs	Hemorrhage, thromboembolic events, time in therapeutic range	Limited to randomized trials of clinical outcomes or large observational studies
4.0 VKAs—monitoring
4.1. Optimal INR range	What is the optimal INR range for best clinical outcomes?	Patients taking VKA	Optimal INR range	Wider INR range	Hemorrhage, thromboembolic events
4.2. Optimal INR range for high-risk groups	Should high-risk groups (such as APS, cancer) be treated more intensively?	Patients with APS (or other high-risk feature) and taking VKA	More intensive INR therapeutic range or alternative assay	Standard INR therapeutic range	Hemorrhage, thromboembolic events, time in therapeutic range
5.0 VKAs—discontinuing therapy
5.1 Tapering vs abrupt discontinuation	How should VKA be discontinued?	Patients discontinuing VKA	Tapered discontinuation	Abrupt discontinuation	Hemorrhage, thromboembolic events, time to normal anticoagulation status
6.0 Parenteral anticoagulants—UFH
6.1. UFH—dose adjustment by weight	Should the initial bolus dose or maintenance dose be weight adjusted?	Patients treated with IV UFH	Weight-adjusted dose	Fixed dose	Hemorrhage, thromboembolic events, time in therapeutic range
6.2. SC UFH dose adjustment and monitoring	Should doses of SC UFH be adjusted for weight and monitored by aPTT?	Patients treated with SC UFH	Weight-adjusted dose with and without aPTT monitoring	Fixed dose with or without aPTT monitoring	Hemorrhage, thromboembolic events, time in therapeutic range
7.0 Parenteral anticoagulants—LMWH
7.1. LMWH—dose modification by renal function	Should doses be modified for renal function?	Patients with mild to moderate renal failure treated with LMWH	Dose adjustment according to renal function	Dose adjustment only by body weight or no dose adjustment	Hemorrhage, thromboembolic events
7.2. LMWH—dose Frequency	Can doses be administered daily instead of twice daily?	Patients treated with LMWH	Doses administered daily	Doses administered bid	Hemorrhage, thromboembolic events	Moved to Kearon et al3 in this supplement
7.3. LMWH dose modification by weight for prophylaxis	Should the dose be weight adjusted?	Obese or significantly underweight patients receiving prophylaxis with LMWH	Dose adjustment according to body weight	Standard dose	Hemorrhage, thromboembolic events	Gould et al4 in this supplement
8.0 Parenteral anticoagulants—fondaparinux
8.1. Fondaparinux dose management	Should the dose be weight adjusted?	Obese or significantly underweight patients receiving fondaparinux	Dose adjustment according to body weight	Standard dose	Hemorrhage, thromboembolic events
9.0 Prevention and management of anticoagulant complications
9.1. Vitamin K for high INR without bleeding	Does vitamin K improve outcomes for high INRs without bleeding?	Patients taking VKA with high INR (> 4.5) without bleeding	KA dose management plus use of vitamin K	Only dose management (= holding the VKA until therapeutic)	Hemorrhage, thromboembolic events, time to therapeutic range, rates of overcorrection of INR
9.2. Predicting anticoagulant-associated bleeding	Does a bleeding clinical prediction rule improve outcomes? Which prediction rule should be used?	Patients taking anticoagulant therapy or considering therapy	Use of a bleeding clinical prediction rule to guide therapy (dose and whether to give)	No clinical prediction rule or alternate prediction rule	Hemorrhage, thromboembolic events, choice of therapy
9.3. Treatment of anticoagulant-related bleeding	What is the most effective and safe urgent treatment of anticoagulant-related bleeding?	Patients actively bleeding from excessive anticoagulation who need to have the bleeding stopped urgently	Vitamin K, FFP, PCC, recombinant factor VIIa	One of the other treatments or vitamin K alone	Time to resolution of bleeding, bleeding complications, thromboembolism rates, resource utilization
9.4. Investigating anticoagulant-associated bleeding	When is it appropriate to investigate anticoagulant-associated bleeding?	Patients taking VKAs with therapeutic INRs and major bleeding episodes	Patients who bleed	Patients who do not bleed	Incidence of malignancy, ulcer disease, other serious or treatable outcome
10.0 Other
10.1. Intensive patient education	Does additional structured patient education improve outcomes related to anticoagulation?	Patients who are to take or are taking VKAs or parenteral anticoagulants	Patient education on benefits, harms, and use of anticoagulants	Usual care	Hemorrhage, thromboembolic events, time in therapeutic range, compliance

AMS = anticoagulant management service; APS = antiphospholipid syndrome; aPTT = activated partial thromboplastin time; LMWH = low-molecular-weight heparin; FFP = fresh frozen plasma; INR = international normalized ratio; PCC = prothrombin complex concentrate; PICO = population, intervention, comparator, and outcome; UFH = unfractionated heparin; VKA = vitamin K antagonist; VKORC1 = vitamin K epoxide reductase complex 1.

2.0 VKA—Initiation of Therapy

2.1 Initial Dose Selection—Loading Dose

Loading doses of VKA may be worth considering where rapid attainment of therapeutic international normalized ratio (INR) is required and considered safe, primarily for patients with VTE. Predictable and timely achievement of therapeutic INRs without increased risk of bleeding or recurrent thromboembolic events avoids the inconvenience and pain of prolonged administration of subcutaneous (SC) low-molecular-weight heparin (LMWH) and facilitates early patient discharge and eligibility for outpatient dosing nomograms. Two large case series^5,6 involving a total of 1,054 outpatients suggest that a nomogram specifying a 10-mg loading dose is safe, with a recurrent VTE rate of 1.9% and a major bleeding rate of 1.0% at 3 months follow-up.⁵ However, pooling across both studies suggests that only 49.3% of participants followed the nomogram completely.

Table 2 and Table S1 (tables that contain an “S” before the number denote supplementary tables not contained in the body of the article and available instead in an online data supplement; see the “Acknowledgments” for more information) summarizes our confidence in effect estimates and main findings from a meta-analysis of five RCTs of loading dose vs no loading dose of warfarin.^7‐11 The table shows that clinical outcomes, where documented, were similar between the groups. The studies typically measured time to therapeutic range of anticoagulation as the primary outcome and the patients were mainly those starting treatment (not prophylaxis) for VTE. Many of those treated as inpatients at the time of the study would, in current practice, be treated as outpatients.

Table 2.

—[Section 2.1] Warfarin 10 mg Loading Dose Nomogram Compared With Warfarin 5 mg Loading Dose Nomogram for Warfarin Initiation^7,8,10,11

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effects
				Risk With Warfarin 5 mg Loading Dose Nomogram	Risk Difference With Warfarin 10 mg Loading Dose Nomogram (95% CI)
Bleeding events	420 (3 studiesa-c), 5-90 dd	Very lowe-g due to indirectness, imprecision	OR 1.90 (0.17-21.1)	5 per 1,000	0 more per 1,000 (from 10 fewer to 20 more)h
Recurrent VTE	420 (3 studiesa-c)	Very lowe-g due to indirectness, imprecision	Not estimable	0 per 1,000	10 more per 1,000 (from 30 more to 0 more)i

GRADE = Grades of Recommendations, Assessment, Development, and Evaluation. See Table 1 legend for expansion of other abbreviation.

^aAll pooled studies included only patients with acute VTE. Studies from which data could be pooled are Kovacs et al,⁹ Quiroz et al,¹⁰ and Schulman et al.¹¹

^bMinimal loss to follow-up; adherence to intention-to-treat principle in two of three studies; follow-up period short but adequate for this outcome; any lack of blinding should not impact objective outcome (laboratory value, INR); adequate allocation concealment; sample size calculations reported for two of three studies.

^cResults based on only three studies; one study shows no difference; one shows statistically significant reduction in time to therapeutic INR; and one had two parts to it, where one showed statistically significant reduction and the other did not.

^dMean follow-up period of 5 d for patients in the loading dose warfarin group from Schulman et al¹¹ (this was the shortest period, only mean is available).

^eData collectors unblinded.

^fIndirect given application aimed at outpatients with VTE; follow-up period is very short in two of three studies (5 d-2 wk).

^gNo studies were powered to detect differences in bleeding events between groups. Number of events is too sparse to draw any conclusions.

^hVery small number of events; risk difference calculated.

ⁱOR not estimable; absolute risk difference calculated.

Two studies by a single group^7,8 compared a 10-mg loading dose to 5 mg daily for the first 2 days. Both included primarily inpatients, and one did not report recurrent VTE.⁸ The concentrations of protein C and factor VII, but not those of factor II or X, decreased faster in the 10-mg group than in the 5-mg group⁸; an increased risk of recurrent thromboembolism, however, has not been demonstrated in any of the studies presumably because initiation overlaps with heparin or LMWH. Quiroz et al¹⁰ compared 5 vs 10 mg initial warfarin dosing in 50 inpatients and reported no difference in median time to two consecutive therapeutic INRs. This study had only a 2-week follow-up and excluded 322 of the 372 patients screened. Another study compared loading dose vs standard warfarin initiation for patients with VTE and showed a shorter time to a therapeutic INR (3.3 vs 4.3 days).¹¹ Finally, Kovacs et al⁹ found that the use of a 10- vs 5-mg initiation nomogram with 210 outpatients resulted in shorter mean time to therapeutic INR of 4.2 vs 5.6 days. The proportion therapeutic by day 5 was also significantly better at 86% vs 45% in the 10- vs 5-mg group, respectively. All studies followed the initiation period with INR-based dose adjustment.

Recommendation

2.1. For patients sufficiently healthy to be treated as outpatients, we suggest initiating VKA therapy with warfarin 10 mg daily for the first 2 days followed by dosing based on INR measurements rather than starting with the estimated maintenance dose (Grade 2C).

2.2 Initial Dose Selection and Pharmacogenetic Testing

Selection of the initial and maintenance doses of VKA therapy usually has been based on subjective estimates of patient age, size, nutritional status, and organ function. In section 2.1, we suggest a standard short loading dose for outpatients. Theoretically, individual patient pharmacogenetic testing of CYP2C9 (cytochrome P450 2C9), which is involved with VKA metabolism and VKORC1 (vitamin K epoxide reductase complex 1, the VKA target), might improve VKA therapy through more-accurate dose selection. There are four RCTs of pharmacogenetic testing-based dosing vs standard dosing; all addressed warfarin initiation.^12‐15 The studies included patients with artificial heart valves, atrial fibrillation, or acute VTE. All studies were small (total n = 544). None showed any difference in thrombotic events, major bleeding, or survival (Table S2).

Hillman et al¹² conducted a pilot study of 38 patients. Caraco et al¹³ randomized 283 patients but excluded 92 for reasons such as failure to follow warfarin dosing instructions. Huang et al¹⁵ included 121 valve inpatients and showed improvement in time to therapeutic range; the control group, however, used a substandard 2.5-mg daily regimen. Anderson et al,¹⁴ who had the highest methodologic quality, studied inpatients in which the control group experienced close INR monitoring following a loading-dose strategy. The investigators found no difference in time in therapeutic range or time to therapeutic range. A systematic review also concluded that there is a lack of evidence to support using pharmacogenetic testing to guide VKA dosing.¹⁶

Several recent economic evaluations have assessed the cost-effectiveness of pharmacogenetic testing to guide VKA (warfarin) initiation.^17‐19 The results of these studies estimated the incremental cost at ∼$50,000 to $170,000 per quality-adjusted life year gained, but in sensitivity analyses, the incremental cost-effectiveness ratios were as high as $200,000 to $300,000 per quality-adjusted life year and included scenarios in which pharmacogenetic testing led to poorer patient outcomes. These results would be judged as not cost-effective by most drug policy experts.

Recommendation

2.2. For patients initiating VKA therapy, we recommend against the routine use of pharmacogenetic testing for guiding doses of VKA (Grade 1B).

2.3 Initiation Overlap for Heparin and VKA

Historically, clinicians administered IV unfractionated heparin (UFH) to inpatients for 5 to 7 days with subsequent initiation of a VKA, leading to a total duration of IV UFH of 10 to 14 days. More recently, VKA therapy has been initiated on the first or second day of heparin therapy, leading to shorter durations of heparin and earlier discharge from the hospital.

Table 3 (and Table S3) summarizes the evidence from a meta-analysis of 807 patients in four RCTs addressing this issue (F. Qayyum, unpublished data, 2011). These trials compared early start (day 1 or 2 of heparin) vs late start (days 3-10 of heparin) for the VKA therapy together with UFH or LMWH therapy. Two studies^20,21 enrolled patients with DVT only, one enrolled patients with DVT or pulmonary embolism (PE),²² and the fourth included patients with left ventricular mural thrombosis.²² There were no differences between early vs late initiation of VKA for the outcomes of recurrent VTE, major bleeding, or death. Patients assigned to early initiation of VKA spent a mean of 4 fewer days in the hospital than patients assigned to late initiation of VKA. No studies have assessed early vs late initiation of VKA therapy in the outpatient setting, but we consider the results of the meta-analysis to be applicable to outpatients.

Table 3.

—[Section 2.3] VKA Started Early vs Late With Heparin in Patients With Acute Thromboembolism

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effects
				Risk With Late	Risk Difference With VKA Started Early (95% CI)
Death	807 (4 studies), 3-6 mo	Lowa-c due to inconsistency and imprecision	RR 1.28 (0.43-3.85)	58 per 1,000	16 more per 1,000 (from 33 fewer to 166 more)
Recurrent thromboembolism DVT: venography, Doppler ultrasonography or impedance plethysmography. PE: lung scanning Left ventricle thrombus: 2-dimensional transthoracic echocardiography	807 (4 studies), 3-6 mo	Lowc,d due to risk of bias and imprecision	RR 0.92 (0.46-1.82)	41 per 1,000	3 fewer per 1,000 (from 22 fewer to 33 more)
Major bleeding-required blood transfusion, bleeding in body cavity, bleeding that required anticoagulation withdrawal or intracranial or retroperitoneal, or bleeding that led to a hemoglobin level decrease of ≥ 2 g/dL or to death	807 (4 studies), 0.5-6 mo	Lowc,d due to risk of bias and imprecision	RR 1.22 (0.58-2.56)	33 per 1,000	7 more per 1,000 (from 14 fewer to 51 more)
Hospital utilization	536 (3 studies)	High		The mean hospital utilization in the control groups was 14 d	The mean hospital utilization in the intervention groups was 4.07 lower (4.76 to 3.37 lower)

PE = pulmonary embolism; RR = risk ratio. See Table 2 legend for expansion of other abbreviation.

^aFor three out of four studies, concealment of allocation was unclear. Lack of blinding of health-care professionals in some studies.

^bThe value for I2 test for death was 55%, and therefore, it was rated down for inconsistency.

^cThe 95% confidence intervals around the absolute risk values were very wide for this outcome.

^dPotential limitations in design for this outcome: allocation sequence concealment was not reported in three out of four studies; health-care professionals blinded in only one study (Hull et al²⁰) (outcome assessors were blinded in three of four studies).

Recommendation

2.3. For patients with acute VTE, we suggest that VKA therapy be started on day 1 or 2 of LMWH or UFH therapy rather than waiting for several days to start (Grade 2C).

3.0 Maintenance Treatment With VKAs

3.1 Monitoring Frequency for VKAs

The frequency of long-term INR monitoring is influenced by patient compliance, changes in health status, the addition or discontinuation of interacting medications, changes in diet, the quality of dose-adjustment decisions, and whether the patient has demonstrated stable INRs.^23‐25 We define stable INRs as at least 3 months of consistent results with no need to adjust VKA dosing.²⁶ Recall intervals for various clinical situations have not been extensively studied; rather, they evolved from routine clinical practice and expert opinion and differ substantially from one country to another.²⁷ For example, in North America, stable patients usually are tested every 4 weeks,²⁴ whereas in the United Kingdom, INR recall intervals of up to 90 days are routine.²⁸ This discussion does not apply to patients engaging in INR self-testing using portable finger-stick monitors in whom only weekly INR recall intervals have been adequately evaluated.

For patients receiving traditional laboratory-based INR monitoring, retrospective studies have found increasing INR recall intervals associated with both increased²⁹ and decreased^26,30 time in therapeutic range (TTR). Other observational studies have suggested that for patients who demonstrate a consistent pattern of stable therapeutic INRs, allowing INR recall intervals of up to 8 weeks would not result in increased risk for bleeding or thromboembolism.^31‐33

Three RCTs have evaluated the effectiveness of INR recall intervals exceeding the traditional North American standard of 4 weeks.^23,34,35 One study compared 6- to 4-week recall intervals,³⁴ whereas another evaluated a flexible approach that allowed recall intervals of up to 12 weeks based on several factors, including the number of prior INRs, longitudinal INR variability, and the risk of adverse events expressed as a function of the INR.²³ The third study compared 4- to 12-week recall intervals using a blinded design.³⁵ None of the studies found a difference in rates of thromboembolism, bleeding, or INR control (Table 4, Table S4). The appropriate length of the recall interval depends on the duration of prior stability and foreseeable future changes in medications or disorders that affect the INR. Whatever maintenance dose interval is chosen, when adjustments to the VKA dose are required, a cycle of more-frequent INR monitoring should be completed until a consistent pattern of stable therapeutic INRs can be reestablished.³⁶

Table 4.

—[Section 3.1] Prolonged INR Recall Intervals Compared With 4-Week Recall Intervals for Patients With a Stable INR^23,34,35

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With 4-wk Recall Intervals	Risk Difference With Prolonged INR Recall Intervals (95% CI)
Thromboembolism variously defined	994 (3 studies), 313 patient-y	Moderateb due to imprecision	OR 1.05 (0.28-3.97)	12 per 1,000	1 more per 1,000 (from 8 fewer to 33 more)
Major bleeding variously defined	994 (3 studies), 313 patient-y	Moderateb due to imprecision	RR 1.12 (0.57-2.23)	33 per 1,000	4 more per 1,000 (from 14 fewer to 41 more)

See Table1-3 legends for expansion of abbreviations.

^aTime frame in months.

^bWide CIs around the estimate of effect.

Recommendation

3.1. For patients taking VKA therapy with consistently stable INRs, we suggest an INR testing frequency of up to 12 weeks rather than every 4 weeks (Grade 2B).

3.2 Management of the Single Out-of-Range INR

A common dilemma encountered in clinical management of patients taking VKAs is what to do with an INR slightly outside the therapeutic range when INRs were previously in the therapeutic range.²⁵ The question is whether the dose should be adjusted or left unchanged until the next INR is obtained.

This issue has been evaluated in two studies. An open-label RCT compared a one-time dose increase or hold vs continuing as is when the INR was slightly below or above the therapeutic range.³⁷ Randomized patients had been taking a stable warfarin dose for at least 3 months, the out-of-range INR was between 1.5 and 4.4, and the target ranges were 2.0 to 3.0 or 2.5 to 3.5. Reduced or boosted doses were usually 50% lower or higher, respectively, than the regularly scheduled dose. Results were similar at follow-up ∼2 weeks later, with 44% outside the therapeutic range among patients randomized to a one-time dose change compared with 40% of those randomized to no dose change (OR, 1.17; 95% CI, 0.59-2.30; P = .75).

The other study evaluated the safety of not changing the usual warfarin maintenance dose in response to isolated, asymptomatic INRs of 3.2 to 3.4 in patients who had been taking warfarin for at least 30 days and had a targeted INR range of 2.0 to 3.0.³⁸ This was an observational study nested within an RCT evaluating anticoagulation management services (AMS) vs primary care management. The response to an isolated INR between 3.2 and 3.4 was to continue the same dose 78% of the time in AMS vs 47% in primary care. The proportion of patients with a therapeutic follow-up INR was not significantly different between the two groups (AMS, 63%; control, 54%). No major bleeding or thromboembolic events were observed during the 14 to 30 days follow-up in either of these studies.

The evidence from both studies suffers from relatively small sample sizes; lack of blinding; and in the second study, lack of randomization and a lack of uniformity in INR management between groups. Both studies were consistent with a dosing model developed from an observational study of 3,961 patients that suggested that warfarin doses did not need to be changed for INRs between 1.7 and 3.3.²⁵ It is reasonable to follow up with an INR after 1 to 2 weeks to exclude a progressive deviation from the therapeutic range.^36,37

Recommendation

3.2. For patients taking VKAs with previously stable therapeutic INRs who present with a single out-of-range INR of ≤ 0.5 below or above therapeutic, we suggest continuing the current dose and testing the INR within 1 to 2 weeks (Grade 2C).

3.3 Bridging for Low INRs

When the INR becomes subtherapeutic, there may be an increased risk of thrombosis. A 2008 retrospective study of 2,597 adult patients receiving warfarin mainly for atrial fibrillation or VTE matched 1,080 patients in the low-INR cohort with 1,517 patients in the therapeutic-INR cohort based on index INR date, indication for warfarin, and age.³⁹ All patients in the low-INR cohort had a subtherapeutic INR following two therapeutic INR measurements. There was no significant difference in thromboembolic events between the two groups, including the small number (99) of patients with artificial heart valves.

A second retrospective study addressed the same scenario in 294 patients with mechanical heart valves.⁴⁰ Bridging with LMWH was prescribed in 14 cases. The incidence of thromboembolic events was found to be 0.3% (95% CI, 0%-1.9%) for all patients included in the study and 0.4% (95% CI, 0%-2.0%) for all patients who did not receive bridging therapy. Both studies are limited by the observational study design and its potential for confounding. Unfortunately, this evidence only addresses the single low INR, not several consecutive low INRs.

Recommendation

3.3. For patients with stable therapeutic INRs presenting with a single subtherapeutic INR value, we suggest against routinely administering bridging with heparin (Grade 2C).

3.4 Vitamin K Supplementation

A low TTR as well as highly variable INR results are independent predictors of bleeding and thromboembolic complications during VKA therapy. One observational study using food diaries to quantify daily vitamin K intake showed that patients in the highest tertile of vitamin K intake had the most stable INR control over time, suggesting the possibility that daily vitamin K supplementation might improve anticoagulation control.⁴¹

Three randomized, placebo-controlled trials using pharmaceutically prepared vitamin K have addressed this issue.^42‐44 There are important differences among these RCTs, including the daily dose of vitamin K studied (100 μg,^42,44 150 μg,^43,44 or 200 μg⁴⁴), the study participants (general anticoagulation clinic patients^42,44 or patients with unstable INR control⁴³), the width of targeted INR range (1.5^42,44 or 1.0), and type of VKA (phenprocoumon or warfarin). Table 5 (and Table S5) shows the quality of evidence and main findings of our meta-analysis of the three RCTs. The absolute difference in TTR was a modest 3.54% (95% CI, 1.13%-5.96%). No difference in major bleeding or thromboembolic complications was seen.

Table 5.

—[Section 3.4] Low-Dose Vitamin K Supplementation Compared With Placebo for Patients Taking VKAs To Stabilize INR^42‐44

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With Placebo	Risk Difference With Low-Dose Vitamin K Supplementation (95% CI)
Thromboembolism	626 (3 studies), 6-12 mo	Very lowb-e due to inconsistency, imprecision	RR 1.65 (0.08-34.03)	0 per 1,000	Not estimablef
Major bleeding	626 (3 studies), 6-12 mo	Very lowb-e due to inconsistency, imprecision	RR 2.61 (0.34-20.28)	0 per 1,000	Not estimablef

See Table 1-3 legends for expansion of abbreviations.

^aTime frame in months.

^bAllocation concealment not reported; uncertain whether outcome adjudicators were blinded.

^cDefinition of thromboembolism and major bleeding different in each study.

^dStudies not powered to detect bleeding or thromboembolic events; total number of events is extremely low.

^eUnable to rule out publication bias because not enough studies exist to populate a funnel plot.

^fTotal of two thromboembolic and three major bleeding events in low-dose vitamin K groups.

The TTR observed in the control arms of these vitamin K RCTs indicates that studied patients had relatively stable INRs (TTR range, 78.0%-85.5%). It would be of greater interest to evaluate the effect of daily vitamin K supplementation in a population with unstable INRs that are not due to other correctable factors. In summary, current evidence does not support supplementation with vitamin K to increase TTR or to improve clinical outcomes.

Recommendation

3.4. For patients taking VKAs, we suggest against routine use of vitamin K supplementation (Grade 2C).

3.5 Anticoagulation Management Services for VKAs

In response to the recognized difficulty in coordinating oral anticoagulation therapy, AMS have evolved in both inpatient and outpatient settings. For the purposes of this review, an AMS was defined as having a designated, trained staff member responsible for patient INR monitoring and follow-up, the use of a standardized local procedure for VKA management (eg, dosing nomogram), and the management of regular INR testing. Further, usual care was defined as regular medical care that generally was provided by the patient’s personal physician in the absence of an AMS.

Four prospective RCTs comparing usual care with the care of an AMS failed to show a significant difference in major bleeding, thromboembolism, or anticoagulation therapy-related mortality.^45‐48 None of these RCTs were blinded, only two studies clearly specified an intention-to-treat analysis,^46‐48 one study allowed patients to switch between treatment arms,⁴⁵ and all patients in two studies were stabilized in an AMS prior to randomization.^45,48

In contrast, the results of many low-quality observational studies have reported higher TTR and better outcomes in patients when anticoagulant therapy is managed by an AMS compared with usual care.^49‐65 The absolute difference in TTR between AMS and community practices in a systematic review was 8.3% (95% CI, 4.4%-12.1%), favoring AMS.⁶⁶

Given the conflicting results between randomized and nonrandomized studies and the lack of economic analysis or compelling patient preference data, the Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines committee decided to make the following best practice statement on this question:

3.5. We suggest that health-care providers who manage oral anticoagulation therapy should do so in a systematic and coordinated fashion, incorporating patient education, systematic INR testing, tracking, follow-up, and good patient communication of results and dosing decisions.

3.6 Patient Self-Testing and Self-Management

Patients using long-term oral anticoagulation therapy usually are monitored by going to a hospital or laboratory to provide blood by venipuncture for INR testing. Point-of-care (POC) devices allow INR testing to be performed by patients in their homes with a drop of blood from the finger. This is defined as patient self-testing (PST). If the patients who perform their own INR testing also adjust their anticoagulant dose, this is called patient self-management (PSM).⁶⁷ Several systematic reviews have evaluated RCTs of PST/PSM to determine whether these approaches to oral anticoagulation therapy result in better clinical outcomes than traditional laboratory-based INR monitoring.^67‐71 A recent individual patient meta-analysis clarified several aspects; our recommendations are based primarily on this more-detailed analysis.⁷²

Pooled analyses show a significant reduction in the rate of thromboembolic complications with PST/PSM but not in the rate of major bleeding or overall mortality compared with usual laboratory-based INR monitoring (Table 6, Table S6). These benefits are seen most prominently in PSM rather than PST groups and possibly in patients with mechanical heart valves rather than other indications.⁷² The largest RCT of PST (n = 2,915), the Home INR Study (THINRS), demonstrated no advantage in clinical outcomes vs laboratory-based monitoring but did show modest, significant improvements in patient satisfaction with anticoagulant therapy and quality of life.⁷³ Data from a pooled analysis also show better patient satisfaction, quality of life, or both with PST/PSM, but these results are difficult to interpret because of the wide range and variable quality of the outcome measures used.⁷¹

Table 6.

—[Section 3.6] Patient Self-Testing/Self-Monitoring Compared With Usual Laboratory-Based Monitoring for VKA Therapy Management⁷²

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With Usual Laboratory-Based Monitoring	Risk Difference With Patient Self-Testing/Patient Self-Monitoring (95% CI)
Thromboembolism various methodsb	6,417 (11 studies), 3-36 mo	Lowc,d due to risk of bias, inconsistency	OR 0.51 (0.31-0.85)	48 per 1,000	23 fewer per 1,000 (from 7 fewer to 33 fewer)
Major bleeding various definitionsb	6,417 (11 studies), 3-36 mo	Moderatec due to risk of bias	OR 0.88 (0.74-1.06)	77 per 1,000	9 fewer per 1,000 (from 20 fewer to 4 more)
Mortality all-cause mortality	6,417 (11 studies), 3-36 mo	Moderatec due to risk of bias	OR 0.82 (0.62-1.09)	87 per 1,000	15 fewer per 1,000 (from 32 fewer to 7 more)

See Table 1-3 legends for expansion of abbreviations.

^aTime frame in months.

^bDefined by individual studies.

^cFlaws in study design, most commonly lack of blinding.

^dSignificant heterogeneity in pooled analysis (I² = 52.6%).

Pooled results from RCTs show only modest (weighted mean difference, 1.50%; 95% CI, −0.63%-3.63%), nonsignificant improvement in TTR with ST/PSM compared with usual laboratory-based monitoring.⁷¹ The frequency of INR testing was considerably higher with PST/PSM compared with usual laboratory-based monitoring, with a mean of 22 to 24 more INR tests annually compared with control groups.⁷²

Resource utilization is relevant when considering whether to recommend widespread use of PST/PSM. Some analyses have deemed PST/PSM to be cost-effective,^74‐76 whereas others have not.^68,69,77 Higher costs with PST/PSM are driven largely by the cost of test strips and increased testing frequency.⁷⁷ However, the increased convenience that PST/PSM offers, particularly to those who travel frequently or who live remotely from testing facilities, can result in lower personal costs for individual patients.⁷⁷

Successful PST/PSM requires well-trained, highly motivated patients. In most RCTs, more than one-half of patients were excluded because of physical limitations, inability to demonstrate competence with POC devices, apprehension about self-care, or patient refusal.⁷¹ Furthermore, up to 25% of patients randomized to PST/PSM withdrew prior to study completion.⁶⁷ THINRS was more promising in that ∼80% were able to pass a PST competency assessment, but 16% switched from PST to the clinic testing group during the study.⁷³

Recommendation

3.6. For patients treated with VKAs who are motivated and can demonstrate competency in self-management strategies, including the self-testing equipment, we suggest PSM rather than usual outpatient INR monitoring (Grade 2B).

For all other patients, we suggest monitoring that includes the safeguards in our best practice statement 3.5.

3.7 Dosing Decision Support

There have been many reports of experience with paper nomograms and computer programs used to assist with VKA dosing.^46,78‐97 These dosing adjuncts have been studied at the initiation of therapy (no prior VKA doses) and during the maintenance phase of therapy and were compared with dose decisions made without the use of decision support (manual dosing). Both nomogram/computer-assisted and manual dosing were performed by experienced anticoagulation providers in some studies^{78,86,87,90,91} and by providers without specialized training (eg, trainee physicians, house staff, regular physician, nurses) in others.^{46,79‐85,88,89,92‐95}

Decision support-guided dosing (paper nomograms or computer programs) performed no better than manual dosing during initiation of VKA therapy in pooled analyses of available RCTs (Table 7, Table S7). Pooled analyses of RCTs evaluating decision support-guided dosing during maintenance therapy (all were computer-assisted dosing programs) revealed a mean TTR improvement of 4.5% (95% CI, 2.4%-6.7%) compared with no decision support. Although statistically significant, this did not result in improvements in thromboembolism, major bleeding, or mortality outcomes (Table 7). The magnitude of TTR improvement with decision support-guided dosing was smaller when manual dosing in control groups was performed by experienced anticoagulation providers vs providers without specialized training (2.04% vs 8.22%, respectively; no P value provided). Higher TTR also has been associated with a paper nomogram in an observational study.⁹⁶

Table 7.

—[Section 3.7] Dosing Decision Support Compared With Manual Dosing for VKA Therapy^{78‐80,82,86‐88,90‐92,94}

Outcomes	No. of Participants (studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With Manual Dosing	Risk Difference With Dosing Decision Support (95% CI)
Thromboembolism, initiation variously defined	503 (4 studies), 3 mo	Lowb,c due to risk of bias, imprecision	RR 0.61 (0.27-1.37)	63 per 1,000	63 fewer per 1,000 (from  46 fewer to 23 more)
Major bleeding, initiation variously defined	926 (7 studiesd), 1-3 mo	Lowb,c due to risk of bias, imprecision	RR 0.43 (0.17-1.09)	30 per 1,000	17 fewer per 1,000 (from  25 fewer to 3 more)
Mortality, initiation all-cause mortality	748 (5 studiese), 1-3 mo	Lowb,c due to risk of bias, imprecision	RR 0.73 (0.36-1.46)	44 per 1,000	12 fewer per 1,000 (from 28 fewer to 20 more)
Thromboembolism, maintenance variously defined	14,213 (7 studiesf), 1-12 mo	Moderateb due to risk of bias	RR 0.9 (0.7-1.17)	17 per 1,000	2 fewer per 1,000 (from 5 fewer to 3 more)
Major bleeding, maintenance variously defined	14,035 (5 studiesg), 4.8-12 mo	Moderateb due to risk of bias	RR 0.92 (0.71-1.21)	15 per 1,000	1 fewer per 1,000 (from  4 fewer to 3 more)
Mortality, maintenance all cause mortality	14,044 (5 studiesh), 4.8-12 mo	Moderateb due to risk of bias	RR 1.07 (0.78-1.48)	10 per 1,000	1 more per 1,000 (from  2 fewer to 5 more)

See Table 1-3 legends for expansion of abbreviations.

^aTime frame in days to months.

^bMost studies were unblinded, including patients, health-care providers, and outcome adjudicators.

^cCI of relative effect encompasses wide range of benefit and harm.

^dAsnis et al,⁷⁹ Doecke et al,⁸² Kovacs et al,⁸⁵ Landefeld and Anderson,⁴⁶ Vadher et al,⁹² van den Bemt et al,⁹⁴ and White et al.⁹⁵

^eAsnis et al,⁷⁹ Doecke et al,⁸² Kovacs et al,⁸⁵ Landefeld and Anderson,⁴⁶ and Vadher et al.⁹²

^fClaes et al,⁸¹ Fitzmaurice et al,⁸³ Fitzmaurice et al,⁸⁴ Mitra et al,⁸⁸ Poller et al,⁹¹ Vadher et al,⁹² and Vadher et al.⁹³

^gClaes et al,⁸¹ Fitzmaurice et al,⁸³ Fitzmaurice et al,⁸⁴ Poller et al,⁹¹ and Vadher et al.⁹³

^hClaes et al,⁸¹ Fitzmaurice et al,⁸³ Fitzmaurice et al,⁸⁴ Poller et al,⁸⁹ and Poller et al.⁹¹

The use of computerized VKA dosing decision support reduces the time taken to dose each patient (mean time for computer-assisted dosing, 94 s; [95% CI, 66-123 s]; manual dosing, 149 s [95% CI, 102-196 s]).⁹⁷ This difference is unlikely to be clinically meaningful except in high-volume AMS locations.^84,92,93 Inexperienced anticoagulation providers have safely used decision support-guided dosing.^84,92,93 Although the computer-assisted dosing software is expensive, an economic analysis of the largest computer-assisted dosing RCT concluded that investment in computer-assisted dosing could represent good value if per-patient costs of dosing were reduced.⁹⁷

Recommendation

3.7. For dosing decisions during maintenance VKA therapy, we suggest using validated decision support tools (paper nomograms or computerized dosing programs) rather than no decision support (Grade 2C).

Remarks: Inexperienced prescribers may be more likely to improve prescribing with use of decision support tools than experienced prescribers.

3.8 VKA Drug Interactions to Avoid

Previous systematic reviews addressing drug interactions with VKAs have examined INR results as outcomes and included case reports as evidence.⁹⁸ Through a literature review, we sought evidence generated from 1996 to early 2011, looking for randomized trials with > 50 patients per group or for large observational studies reporting on clinical outcomes (hemorrhage or VTE) related to drug interactions with VKAs. Our research identified 21 relevant studies. One meta-analysis of RCTs, one prospective cohort study, and many large health database studies were included. A meta-analysis of 10 RCTs (n = 4,180) compared VKA plus aspirin vs VKA alone and showed a reduced rate of arterial thromboembolism (OR, 0.66; 95% CI, 0.52-0.84). However, these benefits were limited to patients with a mechanical heart valve (OR, 0.27; 95% CI, 0.15-0.49), whereas the five studies that dealt with atrial fibrillation and cardiac disease showed no benefit with the combination.⁹⁹ Major bleeding was increased in the meta-analysis regardless of the indication for the combination of VKA plus aspirin vs VKA alone (OR, 1.43; 95% CI, 1.00-2.02).

The remaining nonexperimental studies, which varied in size from 53 bleeding events to > 13,000 events, measured hemorrhage as the clinical outcome. In general, the quality of evidence from these studies was low. The VKA studied in ∼70% of the reports was warfarin. There was sufficient consistency in statistically significant increased rates of bleeding to be concerned about three main therapeutic drug categories. As noted in Table 8, nonsteroidal antiinflammatory drugs (NSAIDs), both nonselective and cyclooxygenase (COX)-2 selective; antiplatelet agents; and some antibiotics are associated with an increased risk of bleeding in patients taking VKAs.

For nonselective NSAIDs, studies reported ORs or risk ratios (RRs) from 1.9 (95% CI, 1.4-3.7) to 4.6 (95% CI, 3.3-6.5).^{100‐103,105,118} In addition, two studies reported a higher risk of bleeding with nonselective NSAIDs compared with COX-2-selective NSAIDs.^101,104 There was less consistency in the relationship between COX-2-selective NSAIDs plus VKAs vs VKA alone and bleeding outcomes, varying from a nonsignificant RR of 1.4 (95% CI, 0.44-4.30) to a significant OR of 3.1 (95% CI, 1.4-6.7).^100‐103 Antiplatelet agents, either undifferentiated, aspirin alone, or clopidogrel alone, were associated with increased rates of bleeding, with estimates of risk from an OR of 1.5 (95% CI, 1.05-2.22) to a hazard ratio (HR) of 3.1 (95% CI, 2.3-3.9).^{102,105,105‐111} Aspirin plus clopidogrel plus VKA compared with VKA alone was associated with an HR of 3.70 (95% CI, 2.89-4.76).¹⁰⁸

Data addressing interactions of antibiotics from multiple large database studies present a somewhat confusing picture. However, there are sufficient studies to suggest a risk of increased bleeding with cotrimoxazole (OR, 2.54 [95% CI, 2.08-3.10]; RR, 5.1 [95% CI, 2.1-12.3])^112,113,115 and quinolones (OR, 1.55 [95% CI, 1.30-1.86]; RR, 5.9 [95% CI, 1.9-18.6]).^{105,112,113,115} There is a suggestion that cephalosporins (ignoring the anomalously high RR provided for cefradine), metronidazole, amoxicillin, amoxicillin/clavulanic acid, doxycycline, and fluconazole may have some impact on bleeding risk, but these drugs in general are insufficiently studied.^112‐114 Similarly, some studies suggest that selective serotonin reuptake inhibitors, tramadol, acetaminophen, coenzyme Q, and ginger may increase the risk of bleeding, but these also require confirmation.^{103,105,106,116,117}

Recommendations

3.8. For patients taking VKAs, we suggest avoiding concomitant treatment with NSAIDs, including COX-2-selective NSAIDs, and certain antibiotics (Grade 2C).

For patients taking VKAs, we suggest avoiding concomitant treatment with antiplatelet agents except in situations where benefit is known or is highly likely to be greater than harm from bleeding, such as patients with mechanical valves, patients with acute coronary syndrome, or patients with recent coronary stents or bypass surgery (Grade 2C).

4.0 VKA—Monitoring

4.1 Optimal Therapeutic INR Range

The desired effect of VKA on the prothrombin time, expressed as INR, can be provided as a therapeutic range (eg, INR 2.0-3.0) or a therapeutic target (eg, INR 2.5). The former provides information on INR values considered acceptable for the patient, whereas the latter is intended to induce those managing anticoagulant therapy to strive for an ideal level.

In a systematic review of 19 studies (one RCT, five with analysis of INR-specific outcomes from RCTs, and 13 observational studies) reporting clinical outcomes in at least three discrete INR ranges and including > 80,000 patients, the lowest rate of a composite outcome of major hemorrhage and symptomatic thromboembolism was seen with INR 2.0 to 3.0.¹¹⁹ Compared with INR 2.0 to 3.0, the RR for the composite outcome was 2.4 (95% CI, 1.9-3.1) for INR < 2 and 1.8 (95% CI, 1.2-2.6) for INR 3.0 to 5.0. For INR > 5, the RR was 11.9 (95% CI, 6.0-23.4) based on 13 studies for bleeding and only one study for thromboembolism. The evidence profiles are shown separately for comparisons of INR 2.0 to 3.0 vs INR 3.0 to 5.0 (Table 9, Table S8) vs INR < 2.0 (Table 10, Table S9). The definition of major bleeding differed among studies, and the type of thromboembolic events varied according to the studied indication for VKA. However, the pattern of relative risks was consistent among atrial fibrillation, valvular heart disease, and other indications taken together.

Table 9.

—[Section 4.1.1] Optimal Therapeutic INR Range: Higher Target vs 2 to 3¹¹⁹

Outcomes	No. of Participants, (Studies) Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With INR 2-3	Risk Difference With INR 3-5 (95% CI)
Major hemorrhage per 100 patient-y, various definitions	76,646 (17 studiesb), 1.8 y	Lowc,d due to risk of bias, dose-response gradient	RR 2.7 (1.8-3.9)	6 per 1,000	10 more per 1,000 (from 5 more to 17 more)
Thromboembolism per 100 patient-y, various definitions	835 (10 studiese)	Very lowf,g due to risk of bias, inconsistency	RR 0.9 (0.6-1.3)	Study population
				46 per 1,000	5 fewer per 1,000 (from 18 fewer to 14 more)
				Moderate
				50 per 1,000	5 fewer per 1,000 (from 20 fewer to 15 more)

See Table 1-3 legends for expansion of abbreviations.

^aTime frame in days to months.

^bSix studies had a randomized controlled trial design.

^cThe majority of studies (eight) were retrospective cohorts.

^dIt is biologically plausible that with increased intensity there will be more bleeding.

^eOne study had a randomized control design.

^fThree of four studies had a retrospective cohort design.

^gThromboembolic events were more frequent with an INR of 2 to 3 in two studies, less frequent in one study, and similar in one study.

Table 10.

—[Section 4.1.2] Optimal Therapeutic INR Range: Lower Target vs 2 to 3¹¹⁹

Outcomes	No. of Participants (Studies)	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effects
				Risk With INR 2-3	Risk Difference With INR < 2 (95% CI)
Major hemorrhage per 100 patient-y, various definitions	78,493 (17 studiesa)	Very lowa,b due to risk of bias, inconsistency	RR 1.1 (0.7-1.7)	Study population
				6 per 1,000	1 more per 1,000 (from 2 fewer to 4 more)
				Moderate
				23 per 1,000	2 more per 1,000 (from 7 fewer to 16 more)
Thromboembolism per 100 patient-y	827 (4 studiesc)	Moderated-f due to risk of bias, large effect, dose-response gradient	RR 3.5 (2.8-4.4)	Study population
				46 per 1,000	115 more per 1,000 (from 83 more to 157 more)
				Moderate
				40 per 1,000	100 more per 1,000 (from 72 more to 136 more)

See Table 1-3 legends for expansion of abbreviations.

^aEight of the studies were retrospective cohorts.

^bFour studies showed higher risk of bleeding, with INR < 2.

^cOnly one study had a randomized control design.

^dNo explanation was provided.

^eAt least 2.8 times more frequent thromboembolism.

^fIt is biologically plausible with more thromboembolism at a lower INR.

Patients with an increased risk of thromboembolic complications are those with (1) a mechanical mitral valve; (2) a mechanical aortic valve in combination with atrial fibrillation, anterior-apical ST-segment elevation myocardial infarction, left atrial enlargement, low ejection fraction, or hypercoagulable state; and (3) caged-ball or caged-disk valve or thromboembolic complications while in INR 2.0 to 3.0. These subsets of patients are traditionally, although with lack of evidence, treated at a higher-intensity INR 2.5 to 3.5 (see Whitlock et al¹²⁰ in this supplement).

4.1.1 Low-Intensity VKA for Patients With VTE:

Low-intensity treatment with VKA corresponds to INR 1.5 to 1.9/2.0 and is of interest because of the possibility that it might cause less bleeding than conventional intensity (INR 2.0-3.0). In addition, given a wider margin of safety from excessive anticoagulation, laboratory monitoring intervals could perhaps be increased to decrease the burden of therapy on the patient. Two RCTs, both blinded, investigated the efficacy and safety of low-intensity VKA in patients with unprovoked VTE.^121,122 Patients were recruited after having received initial conventional-intensity anticoagulation for months to years. Kearon et al¹²¹ compared low intensity with conventional intensity in 738 patients and found a higher risk of recurrent VTE without any reduction of bleeding events in patients treated with low-intensity VKA. Ridker et al¹²² compared low-intensity warfarin with placebo in 508 patients and observed a reduction of recurrent VTE with active treatment without any significant increase in bleeding.

In conclusion, the benefit of low-intensity VKA in terms of reduced risk of bleeding is uncertain because of these inconsistent results. The second benefit of reduced frequency of monitoring is attainable also with conventional-intensity VKA for patients with a stable INR, as reviewed in section 2.1. Thus, the proposed advantage of lower-intensity VKA therapy in the extended-treatment phase is questionable.

4.1.2 Low-Intensity VKA for Patients With Atrial Fibrillation:

For stroke prophylaxis in atrial fibrillation, two less-intensive alternatives to conventional-intensity VKA have been studied. Minidose or low-intensity fixed-dose VKA, usually corresponding to 1.25 mg (0.5-3 mg) warfarin daily, was given with the intention to minimize the need for laboratory monitoring. A meta-analysis of four randomized trials with 2,753 patients showed that minidose warfarin was inferior to conventional-intensity VKA with regard to thrombotic events (RR, 0.50; 95% CI, 0.25-0.97). Results were uncertain for major hemorrhage (RR, 1.23; 95% CI, 0.67-2.27) or fatal bleeding (RR, 0.97; 95% CI, 0.27-3.54).¹²³

Low-intensity VKA with a therapeutic range of INR 1.5 to 2.0 (or 2.1 in one study) has been compared head to head with conventional intensity, without the addition of aspirin, in two randomized trials.^124,125 One study from Japan was stopped prematurely after an excess of major hemorrhages in the conventional-intensity group.¹²⁴ A similar trend was seen in a separate study from Italy.¹²³ Neither study showed a difference in stroke or deaths. The mean age of the patients differed; 65 years in the Japanese study¹²⁴ and 80 years in the Italian trial.¹²⁵ The pooled results show that there is a significant reduction of nonfatal extracranial hemorrhages with low-intensity VKA (OR, 0.21; 95% CI, 0.06-0.6) without any appreciable increase in the rate of stroke or mortality.

A case-control study in patients with atrial fibrillation suggested that the risk of stroke increases at INR < 2.0.¹²⁶ Compared with an INR of 2.0, the OR for stroke was 2.0 (95% CI, 1.6-2.4) at an INR of 1.7 and 3.3 (95% CI, 2.4-4.6) at an INR of 1.5. There is a trade-off that pits a substantial relative risk reduction of stroke (∼80%) with INR 1.5 to 2.0 compared with INR < 1.2,^127,128 with a greater risk of thromboembolic events with INR 1.4 to 1.7 compared with INR 2.0 to 2.5 (OR, 3.72; 95% CI, 2.67-5.19) (Anticoagulation and Risk Factors in Atrial Fibrillation [ATRIA] cohort).¹²⁹ In this study, there was no evidence for a reduced risk for intracranial hemorrhage at INR < 2.0 compared with 2.0 to 3.5. The event rate of intracranial hemorrhage is low with long-term VKA therapy (0.3% per year),¹³⁰ thus very large numbers are required to detect a difference. There was a reduction of major, nonfatal extracranial hemorrhage with low- vs standard-intensity VKA in the two RCTs (OR, 0.21; 95% CI, 0.06-0.6), and this could be important for patients with a documented bleeding diathesis.

Recommendation

4.1. For patients treated with VKAs, we recommend a therapeutic INR range of 2.0 to 3.0 (target INR of 2.5) rather than a lower (INR < 2) or higher (INR 3.0-5.0) range (Grade 1B).

4.2 Therapeutic Range for High-Risk Groups

The most common therapeutic range for treatment with VKAs is INR 2.0 to 3.0, as discussed previously. Higher intensity for patients with a mechanical mitral valve or with a mechanical aortic valve in combination with other risk factors is discussed in Whitlock et al¹²⁰ in this supplement.

Patients with severe thrombophilia (antiphospholipid syndrome, deficiency of protein C, protein S, or antithrombin homozygous factor V Leiden) who have thromboembolic events have an increased risk of recurrent VTE compared with those without thrombophilia or with mild defects (eg, heterozygous factor V Leiden) in the absence of anticoagulant treatment. It is not clear to what extent this is true while taking VKAs. Case series of patients with deficiency of any of the natural inhibitors (protein C, protein S, antithrombin) or with the common factor V Leiden or prothrombin gene polymorphisms have not provided any indication that moderate intensity (INR 2.0-3.0) is inadequate for these conditions.

In retrospective studies, moderate-intensity anticoagulation often was insufficient to prevent arterial or venous thrombosis in patients with antiphospholipid antibodies. Many of the patients in these studies were recruited from specialized centers for patients with rheumatic disease,^131‐133 which may be a different population than those with primary antiphospholipid syndrome (ie, thromboembolism without identified underlying disease).

A systematic review compared the efficacy and safety of different approaches of secondary prophylaxis against thromboembolism in patients with antiphospholipid antibodies based on 16 studies (two RCTs, two subgroup analyses from RCTs, three prospective cohorts or subgroup analysis, and nine retrospective cohorts or subgroup analyses).¹³⁴ There were more fatal thromboembolic events than fatal hemorrhages (18 vs one), and the risk of thrombotic events was inversely related to the INR value in the observational studies but not in the RCTs. In many of the studies, only a single laboratory test had been used to confirm the syndrome, whereas according to current criteria (revised Sapporo criteria), at least two positive tests should be recorded with an interval of at least 12 weeks.¹³⁵

The results of the two RCTs^136,137 are shown in Table 10 (Table S10). Both studies were small, with 110 patients randomized to higher-intensity (INR 3.0-4.0 or INR 3.0-4.5) and 110 randomized to moderate-intensity (INR 2.0-3.0) warfarin therapy. Three patients with nonembolic arterial disease were assigned to aspirin alone (not included in Table 11¹³⁸). Because the CIs for the relative risk are wide and risk of bias is substantial, the quality of evidence is low.

Table 11.

—[Section 4.2] High-Intensity VKA Compared With Moderate-Intensity VKA for Patients With Antiphospholipid Syndrome^136,137

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effects
				Risk With Moderate-Intensity VKA	Risk Difference With High-Intensity VKA (95% CI)
Thromboembolism objective confirmation	220 (2 studiesa), 3 y	Very lowb,c due to risk of bias, indirectness, and imprecision	OR 2.33 (0.82-6.66)	Study populationd
				45 per 1,0002	54 more per 1,000 (from 8 fewer to 195 more)
				Lowd
				50 per 1,000a	59 more per 1,000 (from 9 fewer to 210 more)
				Highd
				700 per 1,000a	145 more per 1,000 (from 43 fewer to 240 more)
Major bleeding	220 (2 studiese), 3 y	Moderatef due to imprecision	OR 0.70 (0.23-2.16)	Study population
				64 per 1,000a	18 fewer per 1,000 (from 48 fewer to 64 more)
				Low
				25 per 1,000a	7 fewer per 1,000 (from 19 fewer to 27 more)
				High
				100 per 1,000a	28 fewer per 1,000 (from 75 fewer to 94 more)
Mortality all-cause mortality	220 (2 studies), 3 y	Moderatef due to imprecision	OR 1.51 (0.3-7.72)	18 per 1,000	9 more per 1,000 (from 13 fewer to 107 more)

See Table 1 and 2 legends for expansion of abbreviations.

^aIn the study by Finazzi et al,¹³⁷ three patients with nonembolic arterial thrombosis received, as planned, only aspirin. They had no events and have not been included here.

^bThe study by Finazzi et al¹³⁷ was open label.

^cBoth studies were designed to show superiority of the more intensive regimen, not equivalence. The 95% CI includes both benefit and significant harm.

^dLow of 5% from Schulman et al; high of 70% from Khamashta et al.¹³¹

^eThe types of major hemorrhage were not disclosed.

^fThe 95% CI includes both benefit and significant harm.

Patients with cancer and VTE have a higher risk of recurrent events during anticoagulant therapy than patients without cancer.^139,140 When such a breakthrough event occurs, an intensification of treatment sometimes is suggested.¹⁴¹ There are no published aggregate data on the effectiveness and safety of intensified treatment with VKA, only single-patient case reports. Dose escalation of LMWH appeared effective to prevent further recurrence in a retrospective review of 70 patients.¹⁴²

Recommendation

4.2. For patients with antiphospholipid syndrome with previous arterial or venous thromboembolism, we suggest VKA therapy titrated to a moderate-intensity INR range (INR 2.0-3.0) rather than higher intensity (INR 3.0-4.5) (Grade 2B).

5.0 VKA—Discontinuation of Therapy

There is a theoretical concern that abrupt VKA discontinuation may result in a temporary hypercoagulable state due to an imbalance in the rates of normalization of activity of the coagulation factors II, VII, IX, and X on the one hand and the natural inhibitors protein C and protein S on the other.¹⁴³ Five small controlled trials (total n = 217) have addressed this issue.^144‐147 The primary outcomes of four of the studies were laboratory results suggestive of a hypercoaguable state^{144,145,147,148} and produced inconsistent results. Elevations tended to persist for 8 to 9 weeks, regardless of discontinuation strategy, suggesting an unmasked prothrombotic state in the absence of anticoagulant protection rather than a rebound phenomenon associated with abrupt discontinuation.

The thromboembolism event rate appeared similar between groups across the five studies (Table 12, Table S11).^144‐148 The only major hemorrhage occurred in the gradual withdrawal group. Gradual discontinuation of VKA is likely to be more confusing and inconvenient for the patient.

Table 12.

—[Section 5.0] Gradual Withdrawal Compared With Abrupt Withdrawal for Patients Taking VKAs for at Least One Month^144‐148

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With Abrupt Withdrawal	Risk Difference With Gradual Withdrawal (95% CI)
Thromboembolism imaging diagnostics	217 (5 studies), 3 mo	Lowb,c due to risk of bias, imprecision	OR 0.96 (0.42-2.18)	126 per 1,000d	4 fewer per 1,000 (from 69 fewer to 113 more)
Mortality all-cause mortality	217 (5 studies), 1 mo	Very lowb,c due to risk of bias, imprecision	OR 0 (0.01-5.6)	9 per 1,000	9 fewer per 1,000 (from 9 fewer to 39 more)d
Major hemorrhage	217 (5 studies), 1 mo	Very lowb,c due to risk of bias, imprecision	OR 1 (0.1-5.6)	9 per 1,000	0 fewer per 1,000 (from 8 fewer to 39 more)d

See Table 1 and 2 legends for expansion of abbreviations.

^aTime frame is weeks.

^bUnclear whether allocation was adequate in Tardy et al,¹⁴⁸ de Groot et al,¹⁴⁵ and Ascani et al.¹⁴⁴ In Michaels and Beamish,¹⁴⁶ it was according to year of birth. Unclear whether allocation was concealed in Tardy, de Groot, and Ascani; it was not concealed in Michaels. Clinicians and patients were not blinded in de Groot, Michaels, Palareti et al,¹⁴⁷ or Ascani.

^cVery small patient groups and few events.

^dThere is no better source than these trials, so low or high estimates are not provided.

Recommendation

5.0. For patients eligible to discontinue treatment with VKA, we suggest abrupt discontinuation rather than gradual tapering of the dose to discontinuation (Grade 2C).

6.0 Parenteral Anticoagulants

6.1 UFH—Dose Adjustment by Weight

Five RCTs compared initial IV UFH dosing according to a weight-based nomogram with a fixed-dose approach.^149‐153 The study by Jaff et al¹⁵¹ was excluded because no weight-adjusted group for the initial bolus was included. The study by Toth and Voll¹⁵³ was excluded because the fixed dose varied by treating physician, and thromboembolic or bleeding complications were not specified. In the remaining three RCTs a total of 292 patients were randomized to either weight based or fixed dose initially. The fixed dose was a bolus of 70 to 80 units/kg followed by an infusion rate of 15 to 18 units/kg per h. Activated partial thromboplastin time (aPTT) values were monitored, and UFH dose titrated to the therapeutic range.^149,150,152 In one of the studies, a POC device for measuring aPTT was used.¹⁴⁹ Patients with acute coronary syndromes¹⁵⁰ or mixed diagnosed conditions, including VTE,^149,152 were recruited. Study follow-up periods ranged from 48 h^149,150 to 3 months.¹⁵² The weight-based and fixed-dose approaches achieved similar therapeutic aPTTs during the first 24 to 48 h. Patient-important adverse events, which were not well defined, were few; thromboembolism in eight vs two (OR, 0.22; 95% CI, 0.02-1.13) in the fixed-dose vs weight-adjusted group and only one major bleed (fixed-dose group) (Table 13, Table S12). These results suggest that weight-adjusted dosing and fixed dosing of IV UFH are similar in outcomes. Small numbers of clinical events and failure to specify the timing of thromboembolic complications are major limitations of available studies.

Table 13.

—[Section 6.1] UFH: Weight-Based Nomogram Compared With Fixed Initial Dose for Patients With Thromboembolic Disease^149,150,152

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With Fixed Initial Dose	Risk Difference With UFH-Weight-Based Nomogram (95% CI)
Thromboembolism	292 (3 studies), 2-90 db	Lowc,d due to risk of bias and imprecision	OR 0.22 (0.02-1.13)e	57 per 1,000f	44 fewer per 1,000 (from 56 fewer to 7 more)
Major hemorrhage	179 (2 studiesg), 1 wk	Very lowc,d due to risk of bias and imprecision	Not estimableh	11 per 1,000	10 fewer per 1,000 (from 30 fewer to 10 more)

See Table 1 and 2 legends for expansion of abbreviations.

^aTime frame is days to weeks.

^bOnly Raschke et al¹⁵² collected data over a 3-mo period.

^cThe studies did not use blinding.

^dNone of the studies was powered for clinical outcomes, which were few and poorly reported with regard to type and timing.

^eFisher exact test.

^fTwo of the eight events occurred after discontinuation of warfarin.

^gBecker et al¹⁴⁹ reported 2% bleeding without specifying allocation group or type of bleeding.

^hZero events in control group; 95% CI on OR not estimable.

Either regimen can be monitored with plasma heparin levels, but there is no evidence to suggest that monitoring improves clinical outcomes. The evidence linking plasma heparin levels of 0.3 to 0.7 International Units/mL anti-Xa activity by the amidolytic assay to the occurrence of either bleeding or thrombosis is also of low quality.¹⁵²

Recommendation

6.1. For patients starting IV UFH, we suggest that the initial bolus and the initial rate of the continuous infusion be weight adjusted (bolus 80 units/kg followed by 18 units/kg per h for VTE; bolus 70 units/kg followed by 15 units/kg per h for cardiac or stroke patients) or a fixed-dose (bolus 5,000 units followed by 1,000 units/h) rather than alternative regimens (Grade 2C).

6.2 UFH—Dose Management of SC UFH

Treatment with UFH has traditionally been monitored with aPTT plasma tests, whether administered by IV or SC. The SC treatment regimens for UFH generally were based on a fixed initial dose.¹⁵⁴ In contrast, short-term treatment with LMWH is given without any laboratory monitoring because the pharmacokinetic characteristics are believed to be more predictable than for UFH. Studies of SC UFH have not compared weight-based dosing vs fixed dosing with or without the use of aPTT monitoring. Weight-adjusted SC UFH monitored with aPTT has been compared with SC LMWH in three RCTs (n = 937) with similar clinical outcome results as follows: recurrent VTE (OR, 1.13; 95% CI, 0.52-2.46), major bleeding (OR, 1.28; 95% CI, 0.42-4.09), and death (OR, 1.34; (95% CI, 0.62-2.93).¹⁵⁵

One RCT in patients with VTE has compared the use of weight-adjusted dosing of SC UFH to weight-based dosing of LMWH without monitoring.¹⁵⁶ The SC UFH was administered at an initial dose of 333 units/kg followed by a dose of 250 units/kg bid; subsequent UFH dosing was kept constant. Clinical outcomes were similar between the SC UFH and LMWH groups (Table 14, Table S13).

Table 14.

—[Section 6.2] UFH: Weight-Adjusted Nonmonitored UFH SC Compared With Weight-Adjusted Nonmonitored LMWH SC for Outpatients With Acute VTE¹⁵⁶

Outcomes	No. of Participants (Studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With Weight-Adjusted Nonmonitored LMWH SC	Risk Difference With Weight-Adjusted Nonmonitored UFH SC (95% CI)
Recurrent VTE objectively measured with same method as for index event	697 (1 study), 3 mo	Lowb,c due to indirectness and imprecision	OR 1.11 (0.49-2.52)	34 per 1,000	4 more per 1,000 (from 17 fewer to 48 more)
Major bleeding by ISTH criteria	697 (1 study), 3 mo	Lowb,c due to indirectness and imprecision	OR 0.5 (0.17-1.34)	34 per 1,000	17 fewer per 1,000 (from 28 fewer to 11 more)
Mortality	697 (1 study), 3 mo	Lowb,c due to indirectness and imprecision	OR 0.83 (0.43-1.57)	62 per 1,000	10 fewer per 1,000 (from 35 fewer to 32 more)

ISTH = International Society on Thrombosis and Haemostasis; SC = subcutaneous. See Table 1 and 2 legends for expansion of other abbreviations.

^aTime frame is days to weeks.

^bThe comparison should actually be vs fixed-dose UFH SC with monitoring, but weight-adjusted UFH SC has only been compared directly with weight-adjusted LMWH. Thus, the comparison is indirect.

^cBecause of premature discontinuation, the study was not powered to demonstrate equivalence.

Because all of the evidence for initial dosing and monitoring of SC UFH is indirect, the quality of evidence for any recommendation is very low. Outpatient use of SC UFH while transitioning to VKA treatment derives some benefit from the elimination of daily blood work. Treatment with UFH often is preferred for patients with severe renal insufficiency, where there is a risk for accumulation of LMWH or fondaparinux.

Recommendation

6.2. For outpatients with VTE treated with SC UFH, we suggest weight-adjusted dosing (first dose 333 units/kg, then 250 units/kg) without monitoring rather than fixed or weight-adjusted dosing with monitoring (Grade 2C).

7.0 LMWH—Dosing

7.1 Should the Therapeutic Dose of LMWH Be Modified for Decreased Renal Function?

LMWH, as opposed to UFH, is primarily eliminated through renal excretion. We found no RCTs comparing a standard, body-weight-adjusted dose to a reduced dose of LMWH in severe renal insufficiency, defined as creatinine clearance < 30 mL/min.

A meta-analysis of 18 observational studies or subgroup analyses of studies using therapeutic doses of LMWH provides some indirect evidence on this patient population.¹⁵⁷ On the basis of four of the studies, this review suggested that standard doses of LMWH led to higher peak levels of anti-factor Xa in patients with a creatinine clearance < 30 mL/min compared with those with a creatinine clearance > 30 mL/min. On the basis of three studies, when the dose of LMWH was reduced for severe renal failure, no such difference in peak level was observed. All of these seven studies used enoxaparin, so there are insufficient data to comment on other LMWHs. In addition, the relevance of anti-factor Xa levels is unclear; several studies have failed to show a relationship between the anti-Xa levels and bleeding.^158‐160

For patients treated with LMWH, the risk of bleeding was generally higher in patients with a creatinine clearance < 30 mL/min compared with patients with a creatinine clearance > 30 mL/min (5.0% vs 2.4%; OR, 2.25; 95% CI, 1.19-4.27; P = .013).¹⁵⁷ However, because the risk of bleeding is also increased when patients with severe renal failure are treated with UFH,¹⁶¹ the problem may be the renal function rather than the dosing regimen. Four observational studies in the review using enoxaparin suggested that lowering doses for severe renal impairment may reduce the incidence of bleeding (Table 15).¹⁵⁷ The dose adjustment was either empirical or to 0.5 vs the standard 1 mg/kg bid of enoxaparin. There are insufficient data on VTE outcomes. Overall, the evidence is indirect and from studies of low quality and provides no advice on how LMWH should be reduced if the decision is to reduce.

Table 15.

—[Section 7.1] Risk of Bleeding With Enoxaparin According to the Calculated Creatinine Clearance

	Bleeding Rate CalCrCl ≤ 30, n/N (%)	Bleeding Rate CalCrCl > 30, n/N (%)	OR (95% CI)
Studies where dose was unadjusted for CalCrCl	17/206 (8.3)	96/4,081 (2.4)	3.88 (1.78-8.45)
Studies where dose was adjusted for CalCrCl	1/106 (0.9)	5/265 (1.9)	0.58 (0.09-3.78)

CalCrCl = calculated creatinine clearance.

Recommendation

7.1. For patients receiving therapeutic LMWH who have severe renal insufficiency (calculated creatinine clearance < 30 mL/min), we suggest a reduction of the dose rather than using standard doses (Grade 2C).

8.0 Fondaparinux—Dosing

8.1 Fondaparinux Dose Management by Weight

Doses of heparins for the treatment of thrombosis often are administered according to patient body weight for both LMWH and UFH. Both total body weight and lean body weight have been used. In clinical trials, patients with morbid obesity (> 120-130 kg) often have been excluded. We did not identify any studies comparing weight-adjusted dosing of fondaparinux to standard doses not adjusted for weight. Two randomized trials for symptomatic venous thrombosis^162,163 used doses adjusted for the total body weight of the patient (5.0, 7.5, or 10 mg in patients weighing < 50, 50-100, or > 100 kg, respectively). These trials—one in DVT,¹⁶² one in PE¹⁶³—compared fondaparinux to enoxaparin and UFH, respectively. A separate study was a subgroup analysis comparing the 3-month incidence of recurrent VTE or major bleeding events in a subset of patients weighing < 100 kg and > 100 kg.¹⁶⁴ The incidences of recurrence and major bleeds appeared to be similar for each patient subset of weight and BMI for patients treated with fondaparinux; VTE occurred in 75 of 1,946 (3.9%) nonobese patients vs 10 of 251 (4%) obese patients, and major bleeds occurred in 25 of 1,993 (1.3%) nonobese patients vs in one of 248 (0.4%) in obese patients. This subgroup analysis has several limitations (no tests for interaction, small number of obese patients, unclear definitions of major bleeds) and provides only low-quality evidence. There are insufficient data on patients with low body weight to make any recommendation or suggestion regarding dose adjustment for these patients.

Recommendation

8.1. For patients with VTE and body weight over 100 kg, we suggest that the treatment dose of fondaparinux be increased from the usual 7.5 mg to 10 mg daily subcutaneously (Grade 2C).

9.0 Prevention and Management of Anticoagulant Complications

9.1 Vitamin K for Patients Taking VKAs With High INRs Without Bleeding

The risk of bleeding increases significantly when the INR exceeds 4.5.¹⁶⁵ In a retrospective review, patients with mechanical heart valves had a risk of adverse events that increased logarithmically from two per 100 patient-years at INR 2.5 to 4.9, to 4.8 per 100 patient-years for INR 5 to 5.5, then to 75 per 100 patient-years for INR ≥ 6.5.¹⁶⁶ Similarly, a case-control analysis of adults sustaining intracerebral bleeding while on warfarin noted a doubling of intracerebral bleeding for every 0.5-s increment in prothrombin time (approximately every 1-point increase in INR).¹⁶⁷

When the INR is supratherapeutic without evidence of bleeding, strategies used to lower the INR have included withholding VKA, adjusting the dose of VKA, and providing some dose of vitamin K. Vitamin K shortens the time to return to normal INR.^168‐170 A 2006 meta-analysis found that administration of vitamin K orally or by IV was more likely to reverse overanticoagulation (INR > 4) at 24 h compared with simply withholding VKA.¹⁷¹

INR 4.5 to 10 Without Bleeding:

Four RCTs compared vitamin K with placebo for patients with INR 4.5 to 10, and all reported on major bleeding as an outcome (Table 16, Table S14).^{168,169,172,173} Pooled analysis suggests that rates of major bleeding were similar over 1 to 3 months of follow-up (2% [10 of 452] of patients receiving vitamin K vs 0.8% [four of 4/471] in the placebo group). Thromboembolism as reported in three of the studies^168,169,172 and occurred in five of 423 patients in the vitamin K group vs four of 441 patients in the placebo group. In summary, although vitamin K use may reverse supratherapeutic INRs more rapidly, there is no evidence of benefit for patient-important outcomes.

Table 16.

—[Section 9.1] Vitamin K vs Only Withholding VKA for Patients Taking Warfarin With an Elevated INR (4.5-10) Without Evidence of Bleeding^{a,168,169,172,173}

Outcomes	No. of Participants (studies), Follow-up	Quality of the Evidence (GRADE)	Relative Effect (95% CI)	Anticipated Absolute Effectsa
				Risk With Only Holding VKA	Risk Difference With Vitamin K (95% CI)
Major bleeding	923 (4 studiesb), 1-3 moc	Moderated,e due to imprecision	OR 2.6 (0.8-9.8)	8 per 1,000	13 more per 1,000 (from 2 fewer to 69 more)
Thromboembolism	864 (3 studiesf), 1-3 moc	Moderated,e due to imprecision	OR 1.3 (0.3-6.6)	9 per 1,000	3 more per 1,000 (from 6 fewer to 48 more)
Mortality all-cause mortality	863 (3 studiesf), 1-3 moc	Moderated,e due to imprecision	OR 1.3 (0.6-2.9)	29 per 1,000	9 more per 1,000 (from 12 fewer to 51 more)

See Table 1 and 2 legends for expansion of other abbreviations. See Table 1 through 3 legends for expansion of other abbreviations.

^aTime frame is days.

INR 6.0-12.0 in Ageno et al.¹⁷³

^bNone of the studies specified whether any bleeding events were fatal or intracranial.

^cFollow-up was 3 mo in both studies by Crowther et al.^168,169

^dTwo small studies, Ageno et al¹⁷² and Ageno et al,¹⁷³ were open label.

^eWide CIs encompass both benefit and significant harm.

^fAgeno et al¹⁷³ did not report thromboembolism, and Ageno et al¹⁷² did not report deaths.

INR > 10 Without Bleeding:

We found no randomized trials that tested treatment strategies in this patient group. A prospective case series of 107 patients with INR > 10 and without evidence of bleeding showed that 2.5 mg of oral vitamin K resulted in a low rate of observed major bleeding by 90 days (3.9%; 95% CI, 1.1-9.7).¹⁷⁴ Another retrospective study of 89 patients found that such patients given oral vitamin K 2 mg were less likely to still have an INR > 5 by day 3 compared with those who only had warfarin withheld (11.1% vs 46.7%).¹⁷⁵ Patient preferences and clinical assessment of risks of thrombosis and bleeding are likely important factors in determining whether to give vitamin K. In summary, the benefit and harm of vitamin K administration for patients with an INR > 10 and no bleeding are unclear, although the risk of bleeding may be substantial.

Recommendations

9.1.

(a) For patients taking VKAs with INRs between 4.5 and 10 and with no evidence of bleeding, we suggest against the routine use of vitamin K (Grade 2B).

(b) For patients taking VKAs with INRs > 10.0 and with no evidence of bleeding, we suggest that oral vitamin K be administered (Grade 2C).

9.2 Clinical Prediction Rules for Bleeding While Taking VKA

The annual incidence of warfarin-associated major bleeding is estimated at 1% to 3%.¹⁷⁶ Clinicians continually struggle with estimating and weighing patient risk of thromboembolic events with risk of major bleeding. A clinical prediction rule for an individual’s risk of bleeding while taking warfarin or other VKAs would be very useful if prediction of low risk reassured patients sufficiently to start VKA therapy or, more importantly, if prediction of high risk of bleeding was sufficiently accurate to withhold VKA therapy.

A 2007 systematic review by Dahri and Loewen¹⁷⁷ examined studies developing clinical prediction rules for bleeding while taking warfarin for any indication. Seven studies were included, with the primary outcome being the ability of the clinical prediction rule to distinguish between patients at high vs low risk of experiencing major bleeding.^6,178‐183 The performance of a rule was considered moderate if the likelihood ratio for a high score to predict major bleeding was > 5.0 and strong if it was > 10.0.^184,185 Two variants of the same clinical prediction rule had a likelihood ratio of ∼9.^178,179 The independently validated mOBRI (modified Outpatient Bleeding Risk Index)¹⁷⁹ includes the following predictors: age ≥ 65 years, history of stroke, GI bleed in the past 2 weeks, and at least one of the following comorbidities: recent myocardial infarction, hematocrit level < 30%, creatinine level > 1.5 mg/dL, or diabetes mellitus. One point is given for each of the four risk factor categories, with high risk defined as ≥ 3 points.

Since the 2007 systematic review, two additional clinical prediction rules have been published.^186‐188 Table 17^{179-183,186-190} summarizes the clinical prediction rules according to (1) the proportion of patients classified as high risk, (2) the risk of major bleeding measured in that subset, and (3) the annual risk of stroke required to prefer an alternative therapy with a lower risk of bleeding for patients with atrial fibrillation. The column on stroke risk required is based on the assumption of a stronger preference for avoiding stroke compared with avoiding a major bleeding event by a factor of 3:1.² Using this metric, most of the rules would suggest a prohibitively high risk of major bleeding only for patients with a CHADS₂ (congestive heart failure, hypertension, age ≥ 75 years, diabetes mellitus, prior stroke or transient ischemic attack) score of 0, a group for whom VKA therapy might not be preferred anyway. However, for patients with a greater preference of avoiding bleeding events compared with stroke, use of CHADS₂ score along with a clinical prediction rule, such as mOBRI, may provide some prognostic guidance. Similarly, the studies involving a population treated for VTE do not identify a group with a risk of bleeding sufficiently high to preclude secondary prophylaxis with VKA. A clinical prediction rule that could predict an individual’s risk of both benefit and harm at the time of initiation of VKA therapy would be desirable, but none has been validated.¹⁹¹

Table 17.

—[Section 9.2] Clinical Prediction Rules for VKA-Associated Major Bleeding

Study Acronym or Authors	Sample, No.	Population	Follow-up Duration, Mean	Proportion With High Risk	Major Bleeding Events in High-Risk Group	Stroke Risk Required to Avoid VKAa
mOBRI179	Derivation: 565	VTE, valves, other	Derivation: 2 y	Derivation: 6.1%	Derivation: 3 m: 23% 12 m: 48%	N/Ab
	Validation: 264		Validation: 6-7 y	Validation: 6.9% (≥ 3 p)	Validation: 3 m: 6% 12 m: 30%
mOBRI—validation180	Validation 1,269	50% AF, 50% other diagnoses	1 y	15.4%	Patients with AF: 12.3%/y	< 4%/y
mOBRI—validation6	Validation: 222	VTE	1.5 y	1%	0%	N/Ab
Kuijer et al182	Derivation: 241	VTE	3 mo	21%	Derivation: 14%/3 mo	N/Ab
	Validation: 780			19% (> 3 p)	Validation: 7%/3 mo
HEMORR2HAGES181	1,604 discharged  on warfarin	AF	0.83 y	16.3% (3-4 p)	8.8%/y	< 3%/y
Shireman et al183	Derivation: 19,875	AF, warfarin naïve	3 mo	Validation: 3.4%  (score ≥ 2.19)	Validation: 5.4%/3 moc	< 1.8% first 3 mo
	Validation: 6,470
RIETE registry187	Derivation: 13,057	VTE	3 mo	Derivation: 5.8%	Derivation: 7.3%/3 mo	N/Ab
	Validation: 6,572			Validation: 5.2%	Validation: 6.2%/3 mo
HAS-BLED186	Derivation: 3,381	AF	1 y	Derivation: 1.7%	Derivation: 20%/y	Validation: < 1.7%/y
	Validation: 3,071			Validation: 7.9% (≥ 3 p)	Validation: 4.9%/y
HAS-BLED—separate validation188	Validation: 3,665	AF	499 d?	18.7%	6.7%/y	< 2%/y

AF = atrial fibrillation; HAS-BLED = hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile INR, elderly (> 65), drugs/alcohol concomitantly; HEMORR₂HAGES = hepatic or renal disease, ethanol use, malignancy, reduced platelet count, re-bleeding, hypertension, anemia, genetic factors, elevated risk of fall including neuropsychiatric disease, stroke; mOBRI = modified Outpatient Bleeding Risk Index; p = total points within the CPR; RIETE = Computerized Registry of Patients with Venous Thromboembolism. See Table 1 legend for expansion of abbreviation.

^aBased on assumption that the dysutility of a stroke is three times that of a major bleeding event, where most major bleeding is GI.

^bFor patients with VTE, the alternative would be no therapy, which can be estimated to result in a risk of recurrence of 22% to 29%^189,190 during the first 3 mo. With the assumption that the dysutility of a recurrence corresponds to the dysutility of a major bleeding event, where the majority consists of GI bleeding, the risk of major bleeding would have to be at least the same during the first 3 mo to avoid VKA.

^cThis risk normally will decrease after the first 3 mo of treatment.

Recommendation

9.2. For patients initiating VKA therapy, we suggest against the routine use of clinical prediction rules for bleeding as the sole criterion to withhold VKA therapy (Grade 2C).

9.3 Treatment of Anticoagulant-Related Bleeding

When patients present with major bleeding due to VKA use, rapid reversal of anticoagulation is desirable, particularly if the bleeding is life threatening. Several products are available to assist, with treatment, often combining vitamin K with one of prothrombin complex concentrate (PCC), fresh frozen plasma (FFP), or recombinant factor VIIa. FFP has the disadvantage of potential allergic reaction or transmission of infection, preparation time, and higher volume. PCC and recombinant factor VIIa are more rapidly concentrated with less infection transmission risk but have not been compared with FFP in adequately powered RCTs.

Vitamin K is given to sustain the effects of the other products because of the relatively short half-lives of the latter. In emergency situations, vitamin K 10 mg IV instead of given orally is recommended because of its more rapid onset.^24,171,192 IV injection of vitamin K is reported to cause anaphylaxis in three of 100,000 patients, resulting in advice to infuse slowly.¹⁹³ In one RCT of patients with INR 6 to 10 without bleeding, IV injection (0.5 mg) compared with po (2.5 mg) phytonadione more rapidly brought the INR back to therapeutic range (11 of 24 patients vs 0 of 23 patients at 6 h).¹⁹² However, by 24 h, the mean INR in both groups was similar. In a second RCT of patients with INR 6 to 10, vitamin K 0.5 mg IV led to faster resolution than vitamin K 3 mg SC, with an INR < 5 in 95% vs 45% of patients and a mean INR of 3.7 vs 5.4 at 24 h.¹⁹⁴ Accordingly, SC injection is not recommended.

Several studies have compared products in addition to vitamin K, three of which reported rates of intracranial hemorrhage. A small case series of 17 patients compared the use of FFP and three-factor PCC; all patients received vitamin K.¹⁹⁵ The mean INR decreased from 2.83 to 1.22 within 4.8 h in patients receiving PCC vs from 2.97 to 1.74 within 7.3 h for those receiving FFP (P < .001). The reaction level grade, used to assess symptoms and signs of intracerebral hemorrhage, suggested less progression in those receiving PCC (0.2 vs 1.9 grades on a scale of 1-8) (P < .05). Another small before-after study of 12 patients reported that the six patients receiving three-factor PCC compared with six age- and sex-matched historical controls given FFP had a mean INR correction time of 41 min for PCC vs 115 min for FFP.¹⁹⁶

Finally, a small RCT compared factor IX complex concentrate (four-factor PCC) plus FFP vs FFP alone in 13 patients (five in factor IX concentrate and eight in FFP).¹⁹⁷ Factor IX concentrate plus FFP corrected the INR more quickly than FFP alone (2.95 vs 8.9 h, P < .01). In addition, five of eight patients in the FFP-alone group experienced significant fluid overload complications, despite monitoring of central venous pressure and the use of furosemide, compared with no reported complications in the combination group.

FFP has also been compared with four-factor PCC in patients undergoing cardiopulmonary bypass surgery.¹⁹⁸ Forty patients admitted to the hospital for urgent or semiurgent cardiac surgery who were taking oral anticoagulants (INR 2.1-7.8) were randomized, 20 to each treatment. Seven PCC patients vs no FFP patients had an INR < 1.5 by 15 min (P = .007); an additional six PCC vs four FFP patients had this level an hour later (P = .70).

Three very small case series addressed the use of recombinant factor VIIa. In a series of 13 patients presenting with bleeding (four patients), requiring rapid reversal for interventions (five patients), or with an INR > 10 and not good candidates for FFP (four patients), all had a reduction in INR after administration but to variable degrees.¹⁹⁹ Use in four patients presenting with major bleeding (two with spinal cord hemorrhages and two with intracerebral hemorrhages) resulted in a normal INR within 2 h, with no complications reported.²⁰⁰ Finally, in a series of seven patients with acute intracranial hemorrhage while taking warfarin, the mean INR was reduced from 2.7 prerecombinant factor VIIa to 1.1 afterward. Several of the patients also received vitamin K and FFP. Five of the patients survived with severe disability, and two died.²⁰¹ Factor concentrates including PCC are expensive and, therefore, not available in some jurisdictions.

Recommendations

9.3. For patients with VKA-associated major bleeding, we suggest rapid reversal of anticoagulation with four-factor PCC rather than with plasma (Grade 2C).

We suggest the additional use of vitamin K 5 to 10 mg administered by slow IV injection rather than reversal with coagulation factors alone (Grade 2C).

9.4 Investigating Anticoagulant-Associated Bleeding

No randomized trials have addressed different strategies of investigating bleeding in patients taking anticoagulants. The topic is of great practical importance in patient management, but the evidence found was not of sufficient quality to make a recommendation. One small case-control study found the monthly incidence and prevalence of hematuria to be 0.05% and 3.2% in those taking anticoagulants vs 0.08% and 4.8% for those in the control group.²⁰² Subsequent diagnosis of cancer was also similar at two of 32 patients in the anticoagulation group compared with one of 11 patients in the control group. Two small case series of patients investigated for anticoagulant-associated hematuria found two of 30 and four of 24, respectively, had neoplastic disease.^203,204 A retrospective analysis of all patients presenting with gross hematuria over a 9-year period while taking anticoagulant or aspirin therapy found that 25% (six of 25) of those patients presenting with hematuria were found to have a tumor.²⁰⁵

Several studies addressed the question of GI bleeding. A retrospective series of 166 patients presenting with lower GI bleeding, with 100 of the patients taking an antiplatelet or anticoagulant and 66 not, found that nine of 88 (10.2%) patients taking anticoagulants had colon cancer compared with two of 62 (3.2%) not taking anticoagulants.²⁰⁶ Another analysis of 98 patients taking warfarin who presented to a Veterans Affairs hospital with acute GI bleeding found on endoscopy that 52 of the 71 had upper-GI lesions, whereas on colonoscopy, 26 of 41 had lesions, including five cancers.²⁰⁷ In summary, although the data are of low quality, they suggest that there might be sufficient incidence of pathologic causes for VKA-associated hematuria or GI bleeding to warrant investigation.

10.0 Other

10.1 Intensive Patient Education and Anticoagulation Outcomes

Intensive patient education (defined as dedicated patient education sessions beyond the usual VKA information distributed by pamphlet or the patient’s usual provider) has been proposed to reduce adverse events related to anticoagulation and to improve TTR. Although better patient knowledge of anticoagulation has been associated with improved INR control, these were no randomized trials, and INRs were surrogate outcomes.^208,209

Seven RCTs (n = 1,195) compared supplemental patient education with usual care and provided some data on clinical outcomes.^210‐216 Patient age varied widely (18-91 years), and the indications for VKA therapy included atrial fibrillation and VTE. Six of the studies were based in anticoagulation clinics. Educational interventions varied among studies. Several allowed for only one teaching session delivered in person by a health-care professional, by means of a video presentation of a physician-patient interaction, or by a patient-administered self-guided instruction booklet.^215,216 Others had repeated interaction with patients at daily intervals on a ward until discharge or at weekly or bimonthly intervals in outpatient clinics.^210,212,213 The curricula covered similar content, including indications for VKAs, benefits and risks, the importance of INR surveillance, drug interactions, the effect of diet, and information on dose management. The amount and type of education in the control arms were unclear. The length of follow-up ranged from 3 to 6 months.

The quality of evidence based on these studies is low primarily because of limitations in design and imprecision for the clinical outcomes. In pooling data from three of the studies that reported clinical outcomes in a similar manner, there was no significant difference between supplemental patient education and usual care (VTE RR, 0.61 [95% CI, 0.06-6.56]; hemorrhage RR, 0.92 [95% CI, 0.04-20.56]).^210,212,213 TTR was reported in four trials and was similar between groups (mean difference, 2.03%; 95% CI, −2.79-6.86).^{210‐212,214} In the single study where the difference in intensity of education was marked (described as minimal vs daily intensive education for mean of 8 days), there was no difference in outcomes, including TTR.²¹² Although we found no compelling evidence favoring intensive patient education over standard patient education practices, the panel believed that a specific recommendation could not be made at this time.

Abbreviations

AMS

anticoagulation management service

aPTT

activated partial thromboplastin time

COX

cyclooxygenase

FFP

fresh frozen plasma

HR

hazard ratio

INR

international normalized ratio

LMWH

low-molecular-weight heparin

NSAID

nonsteroidal antiinflammatory drug

PCC

prothrombin complex concentrate

PE

pulmonary embolism

POC

point-of-care

PSM

patient self-management

PST

patient self-testing

RCT

randomized controlled trial

RR

risk ratio

SC

subcutaneous

TTR

time in therapeutic range

UFH

unfractionated heparin

VKA

vitamin K antagonist