Blood Reviews (2006) 20, 173–178

www.elsevierhealth.com/journals/blre

REVIEW

Duration of anticoagulation: decision makingbased on absolute risk

David Keeling *

Oxford Haemophilia Centre and Thrombosis Unit, Churchill Hospital, Oxford. OX3 7LJ, United Kingdom

Summary We are often faced with the question as to the optimum duration of sec-ondary prophylaxis with oral anticoagulants after an episode of venous thromboem-bolism. Theoretically if we know the recurrence rate, the case-fatality, theeffectiveness of oral anticoagulant therapy, and the rate of fatal haemorrhage ontreatment, we can calculate whether being on or off treatment is safest. Usingthese data and considering only the risk of death we would treat idiopathic deepvein thrombosis for six months. For those with DVT associated with a transient riskfactor it would be reasonable to stop treatment after 3 months in those over 50years old and we should certainly stop after 3 months in those over 70 years old.There are data to suggest that pulmonary embolism may have a higher case-fatalitythan deep vein thrombosis if there is a recurrence. If these data were accepted mostpatients with idiopathic pulmonary embolism would get long-term treatment. Wecan use these models to modify our assessment if other factors such as antiphospho-lipid antibodies or cancer are present.

�c 2005 Elsevier Ltd. All rights reserved.

KEYWORDSVenousthromboembolism;Anticoagulation;Duration

0d

I will neither give a deadly drug to anybody whoasked for it, nor will I make a suggestion to thiseffect.

The above is taken from the Hippocratic Oath(translated from the Greek by Ludwig Edelstein).Perhaps the modern version should be - I will notgive a drug which causes more deaths than it pre-vents. This gets to the crux of the problem with the

268-960X/$ - see front matter �c 2005 Elsevier Ltd. All rights reseoi:10.1016/j.blre.2005.09.001

* Tel.: +44 1865 225318; fax: +44 1865 225608.E-mail address: david.keeling@ndm.ox.ac.uk.

use of oral vitamin K antagonists (VKA) in the sec-ondary prevention of venous thromboembolism(VTE, either deep vein thrombosis (DVT) or pul-monary embolism (PE)). At what point does the riskof fatal recurrence off treatment become less thanthe risk of a fatal haemorrhage due to continuationof treatment? We need to consider absolute risk tomake this decision. We can also calculate the num-ber of patients we need to treat (NNT) to prevent(or cause) one death this is a useful figure to helpus to decide if treatment is worthwhile. See Box1 for definitions.

rved.

Box 1. Definitions

Absolute risk reduction (ARR) is the difference inthe event rate between control group (CER) andtreated group (TER)

ARR ¼ CER� TER

Number needed to treat (NNT) is the number ofpatients who need to be treated to prevent onebad outcome

NNT ¼ 1=ARR

174 D. Keeling

We need the answers to some key questions tomake these calculations:

� What is the recurrence rate after a venousthromboembolic event both on and offtreatment?� What is the mortality of a recurrent event?� What is the risk of fatal haemorrhage due to thetreatment?� Can we predict who is more likely to bleed andwho is more likely to get a recurrence?

I think we also need to consider how we commu-nicate the risks to our patients and involve them inthe decision. The group in Amsterdam have pro-duced two useful articles which assess these risks.1,2

Table 1 Risk of recurrence off treatment followingidiopathic DVT.

Risk ofrecurrencein period (%)

Annualisedrecurrencerate (%)

6 weeks to 3 months 5 403 months to 6 months 5 206 months to 2 years 10 7After 2 years 4

See Prins et al.1

What is the recurrence rate after avenous thromboembolic event both onand off treatment?

The dramatic trial by Barrit and Jordan in 19603

showed that an initial period of anticoagulationsaves lives. They randomised 35 patients with PEto have no treatment or to be treated with heparinand acenocoumarol. In the 16 treated patientsthere were no recurrences and no deaths due toPE. In the 19 untreated patients there were 10recurrences and 5 of these were fatal. This trialhas not been repeated! There have however beenmany trials attempting to assess how long weshould treat for. I will briefly mention three ofthem.

The first DURAC study4 compared six weeks withsix months of oral anticoagulant treatment in pa-tients who had a first episode of VTE. Anticoagulanttherapy consisted of warfarin or dicumarol. 443 pa-tients were randomly assigned to receive six weekstreatment and 454 patients to six months treat-ment. 790 patients had DVT and 107 PE. Aftertwo years of follow-up, there had been 123 recur-

rences, 80 in the six-week group (18.1 percent;95 percent confidence interval, 14.5 to 21.6) and43 in the six-month group (9.5 percent; 95 percentconfidence interval, 6.8 to 12.2). The differencebetween the two groups occurred between 6 weeksand 6 months after the start of treatment, and therates of recurrence remained nearly parallelthereafter.

The second DURAC study5 enrolled patients whohad had a second episode of VTE. 111 patients wererandomly assigned to six months of anticoagulationand 116 were assigned to receive anticoagulanttherapy indefinitely. 193 patients had DVT and 34PE. After four years of follow-up, there were 23recurrences in the group assigned to six monthsof therapy (20.7 percent) and 3 in the group as-signed to continuing therapy (2.6 percent). Therewere 13 major hemorrhages, 3 in the six-monthgroup, (2.7 percent) and 10 in the infinite-treat-ment group (8.6 percent).

Kearon and colleagues6 randomised patients whohad completed three months of anticoagulation fora first idiopathic venous thromboembolism to con-tinue receiving warfarin or to receive placebo fora further 24 months. A pre-specified interim analy-sis of efficacy led to the early termination of thetrial after 162 patients had been enrolled and fol-lowed for an average of 10 months. Of 83 patientsassigned to continue to receive placebo, 17 had arecurrent episode of venous thromboembolism(27.4 percent per patient-year), as compared with1 of 79 patients assigned to receive warfarin (1.3percent per patient-year, P < 0.001). Three pa-tients assigned to the warfarin group had majorbleeding as compared with none of those assignedto the placebo group.

From these three studies we can estimate thatwhilst a patient is taking warfarin the risk ofrecurrence is reduced by approximately 90%. Prinsand colleagues1 also used this data to estimate theabsolute risk of recurrence in different time peri-ods from the event (Table 1). Much effort hasbeen directed to trying to predict those patients

Duration of anticoagulation: decision making based on absolute risk 175

most likely to have a recurrence. The role ofthrombophilia testing has been much debated7

and looking for residual thrombosis8,9 or measuringD-dimers10,11 has been advocated. Whilst testingfor hereditary thrombophilia may not be particu-larly helpful in predicting recurrence7, the pres-ence of antiphospholipid antibodies might be.Vink et al.2, quoting four references, suggestedan odds ratio of 2.5 for recurrence in the presenceof antiphospholipid antibodies. Galli and col-leagues12 analysed 25 studies and found that anti-cardiolipin antibodies were not associated with anincreased risk of DVT but that LA increased therisk six fold; we cannot assume however that thisdifference applies to recurrences. The most usefulfactor in predicting recurrence remains the clini-cal history. Many studies have demonstrated thatif patients with transient risk factors are com-pared with patients with idiopathic thrombosisthe recurrence rate if approximately halved.4,7

This distinction will therefore be an importantconsideration when we consider balancing therisks against the benefits of treatment. Patientswith cancer are at an increased risk of recurrence,and this has been estimated to be 2–4 fold.2 Wewill also have to consider whether the initial eventwas a DVT or a PE (see below).

Table 2 Risk of bleeding on oral anticoagulants.

Age Major bleeding(% p.a.)

Fatal bleeding(% p.a.)

<40 0.6 0.1240-49 1.0 0.250-59 1.5 0.360-69 2.2 0.44>69 3.2 0.64

See Prins et al.1

What is the mortality of a recurrentevent?

The case fatality of VTE is difficult to assess. Textbooks often quote a figure of 1–2%. Randomisedtrials comparing various treatment regimens inPE have suggested a case fatality of 0.6–1.4 %once treatment has started.13–15 Population basedstudies16,17 and registries18 of PE have suggestedgreater case-fatality rates ranging from 3–23%.For our analysis we need a figure which may besubtly different, not the case-fatality of VTE perse but the case-fatality of a recurrence. Douketisand colleagues19 analysed 25 prospective studiesfor the case fatality of recurrences. In patientsinitially presenting with DVT (n = 4,221) 21out of258 recurrences were fatal PE (8%, 95% CI 5–12%). In patients initially presenting with PE(n = 1,302) 19 out of 72 recurrences were fatalPE (26%, 95% CI 17–38%). All fatalities in the pa-tients originally presenting with PE and 17 of the21 fatalities in patients originally presenting withDVT were on treatment. After discontinuation oftreatment 4 out of 79 recurrences in patients orig-inally presenting with DVT were fatal PE giving acase fatality of 5% (1.4–12.5%). For patients ini-tially presenting with DVT 21% of recurrences

were PE but for patients initially presenting withPE 81% of recurrences were PE. That patientswho present with PE are more likely to have a fu-ture PE than patients who present with DVT is notwidely appreciated. This is supported by an anal-ysis by Murin and colleagues.20 Using linked hospi-tal discharge records they analyzed 71,250patients hospitalized with a principal diagnosis ofDVT alone (51,233) or PE (21,625). Patients ini-tially diagnosed with DVT alone were 2.7 (2.3–3.1) times as likely to be rehospitalised withDVT in the following six months than patients pre-senting with PE. Conversely, patients with PEwere 4.2 (3.8–4.7) times as likely to be rehospita-lised with PE.

When we come to our risk benefit calculations Iwill use a figure of 5% for the case-fatality of arecurrence in a patient whose initial event was aDVT. The figure for PE may be as high as 25% whichas we will see would have significant implicationsfor management.

What is the risk of fatal haemorrhagedue to the treatment?

In a prospective cohort from thirty-four Italiananticoagulation clinics, 2,745 consecutive pa-tients were studied from the start of their oralanticoagulation.21 Major bleeds occurred at therate of 1.4 % p.a. and fatal bleeds at 0.25%p.a. The rate was higher in older patients the rel-ative risk being 1.75 in those over 70 years. Inthe same year it was shown that the risk ofbleeding increased by a factor of 1.5 per dec-ade.22 A meta-analysis of 33 studies involving4374 patient-years of oral anticoagulant therapyshowed that, for patients who received anticoag-ulant therapy for more than 3 months, the rateof major bleeding was 2.7% p.a. and fatal bleed-ing 0.25% p.a. after the initial 3 months of anti-coagulation.23 From these data it is possible toestimate the risk of major and fatal bleeding by

176 D. Keeling

decade1 as shown in Table 2. The risk of bleedingin patients with cancer is increased and has beenestimated to be doubled.2

For how long should we anticoagulatepatients after a venous thromboembolicevent?

With the information gathered it is easy to calcu-late if it is safer to be on warfarin or to be off itand if on it how many patient years of treatmentare required to save one life. Below is a workedexample. Suppose a 55 year old has an idiopathicDVT - is it safer to be on warfarin between 3 and6 months after the event?

Assumptions:Annualised recurrence rate is 20%Case-fatality 5%Effectiveness of warfarin 90%

Table 3 Number of years of treatment needed to prevent a

(a) Idiopathic DVT

Period after event

<40 y 40-49 y 50-59 y 60-69 y >69 y

6w-3m 60 63 67 74 86

3-6m 128 143 167 217 385

6m-2y 556 1000 H H H

>2y 1667 H H H H

(b) DVT with transient risk factors

Period after event

<40 y 40-49 y 50-59 y 60-69 y >69 y

6w-3m 128 143 167 217 385

3-6m 303 400 667 10000 H

6m-2y 3333 H H H H

>2y H H H H H

(c

Paf

6

3

6

>

(

Paf

6

3

6

>

H = treatment results in more deaths than lives saved. Reproduced f

Annualised fatal haemorrhage on treatment0.3%

Calculation:Annualised fatality due to recurrent thrombo-sis = 0.2 x 0.05Annualised fatal thrombotic deaths prevented bywarfarin = 0.9 x 0.2 x 0.05Thrombotic deaths minus bleeding deaths = (0.9 x0.2 x 0.05) – 0.003 = 0.006Years of treatment need to prevent one death =1/0.006 = 167

Prins et al.1 performed this calculation for pa-tients of different ages at various intervals afteran idiopathic DVT (Table 3a). Assuming a tran-sient risk factor at the time of a VTE halves therisk of recurrence it is simple to calculate thecorresponding figures for this situation (Table3b). Assuming that the case fatality of a recur-rence is five times greater in those presenting

fatal event by patient age.

) Idiopathic PE

eriod ter event

<40 y 40-49 y 50-59 y 60-69 y >69 y

w-3m 11 11 11 12 12

-6m 23 23 24 25 26

m-2y 72 77 83 94 116

2y 128 143 167 217 385

d) PE with transient risk factors

eriod ter event

<40 40-49 50-59 60-69 >69

w-3m 232 23 24 25 26

-6m 47 49 51 55 62

m-2y 159 182 222 323 909

2y 303 400 667 10000 H

rom Prins et al.1 with permission.

Table 5 Idiopathic DVT associated with cancer.

<40 40-49 50-59 60-69 >69

6w-3m 14 15 15 16 17

3-6m 30 31 33 37 43

6m-2y 104 125 167 313 H

(a) Cancers which increase risk of recurrence four-fold eg pancreatic, ovarian, lung, mucin-secreting gastro-intestinal. See Vink et al. [2].

Duration of anticoagulation: decision making based on absolute risk 177

with an initial PE they calculated similar tablesfor patients presenting with PE (Tables 3c andd). I have put an ‘‘H’’ in the boxes and colouredthem red when continuing warfarin would casemore deaths than it prevents. I have distinguishedthe boxes where less than 500 years of treatmentwould prevent a death (coloured green) fromthose requiring greater than 500 years of treat-ment (coloured orange), in the later case it mightbe felt that the marginal benefit is not worth theinconvenience.

>2y 208 313 833 H H

(b) Other cancers assumed to increase risk of recurrence two-fold. See Vink et al. [2].

<40 40-49 50-59 60-69 >69

6w-3m 30 31 233 37 43

3-6m 64 71 83 109 192

6m-2y 278 500 H H H

>2y 833 H H H H

Haemorrhagic risk assumed to be doubled. See Vink et al.2.Number of years of treatment needed to prevent a fatal event bypatient age. H = treatment results in more deaths than livessaved.

Treatment after a DVT

From Table 3a we can conclude that for patientswith idiopathic DVT treatment should be given forsix months. For those with a transient risk factor(Table 3b) it would be reasonable to stop treat-ment after 3 months in those over 50 years oldand we should certainly stop after 3 months inthose over 70 years old.

It is also simple to adapt the tables if otherfactors are present. Table 4 shows how a factorwhich increases the risk of recurrence 2.5 fold al-ters the assessment. This might apply if the pa-tient had antiphospholipid antibodies, and in thefuture we may apply a similar weighting to raisedD-dimers or residual vein thrombosis. On thisanalysis we might treat patients with idiopathicDVT and antiphospholipid antibodies continuouslyuntil they are 50 years old or treat them for twoyears if they are between 50 and 70 years of age.We would not put them on life-long treatment if

Table 4 Idiopathic DVT with a risk factor such asantiphospholipid antibodies which increases the risk ofrecurrence 2.5 fold.

<40 40-49 50-59 60-69 >69

6w-3m 23 23 24 25 26

3-6m 47 49 51 55 62

6m-2y 159 182 222 323 909

>2y 303 400 667 10000 H

Number of years of treatment needed to prevent a fatal eventby patient age. H = treatment results in more deaths than livessaved.

they were over 70 years old. Patients with cancerare represented in Table 5a and b. In this casewe have to factor in the increased risk of recur-rence and the increased risk of bleeding. Weshould probably consider treatment for longerthan six months in some cancers although wemay wish to consider the alternative of prolongedLMWH.24

Treatment after a PE

If we accept the evidence discussed earlier for theincreased risk of a fatal recurrence if the originalpresentation was with PE19, then we will have tore-assess our practice . We would anticoagulateall patients presenting with idiopathic PE for life(see Table 3c). This therefore becomes an impor-tant area for future research.

178 D. Keeling

Practice points

� Theoretically if we know the recurrence rate,the case-fatality, the effectiveness of oralanticoagulant therapy, and the rate of fatalhaemorrhage on treatment, we can calculatewhether being on or off treatment is safest.Using these data and considering only the riskof death we would treat idiopathic deep veinthrombosis for six months.� For those with DVT associated with a transientrisk factor it would be reasonable to stop treat-ment after 3 months in those over 50 years oldand we should certainly stop after 3 months inthose over 70 years old.� There are data to suggest that pulmonaryembolism may have a higher case-fatality thandeep vein thrombosis if there is a recurrence. Ifthese data were accepted most patients withidiopathic pulmonary embolism would getlong-term treatment.

Research agenda

�We urgently need to assess the case-fatality ofrecurrences after an initial presentation withPE. We may be under treating these patients.

References

1. Prins M et al. Long-term treatment of venous thromboem-bolic disease. Thrombosis and Haemstasis 1999;82: 892–8.

2. Vink R et al. Individualized duration of oral anticoagulanttherapy for deep vein thrombosis based on a decisionmodel. J Thromb Haemost 2003;1(12):2523–30.

3. Barritt DW, Jordan SC. Anticoagulant drugs in the treatmentof pulmonary embolism; a controlled trial. Lancet 1960;1:1309–12.

4. Schulman S et al. A comparison of six weeks with six monthsof oral anticoagulant therapy after a first episode of venousthromboembolism. Duration of Anticoagulation Trial StudyGroup see comments. N Engl J Med 1995; 332(25):1661–5.

5. Schulman S et al. The duration of oral anticoagulanttherapy after a second episode of venous thromboembo-lism. The Duration of Anticoagulation Trial Study Group. NEngl J Med 1997;336(6):393–8.

6. Kearon C et al. A comparison of three months of anticoag-ulation with extended anticoagulation for a first episode ofidiopathic venous thromboembolism. N Engl J Med 1999;340(12):901–7.

7. Baglin T et al. Incidence of recurrent venous thromboem-bolism in relation to clinical and thrombophilic risk factors:prospective cohort study. Lancet 2003;362(9383):523–6.

8. Prandoni P. Risk factors of recurrent venous thromboem-bolism: the role of residual vein thrombosis. PathophysiolHaemost Thromb 2003;33(5-6):351–3.

9. Prandoni P et al. Residual venous thrombosis as a predictivefactor of recurrent venous thromboembolism. Ann InternMed 2002;137(12):955–60.

10. Fattorini A et al. Risk of deep vein thrombosis recur-rence: high negative predictive value of D-dimer per-formed during oral anticoagulation. Thromb Haemost2002;88(1):162–3.

11. Palareti G et al. Risk of venous thromboembolism recur-rence: high negative predictive value of D-dimer performedafter oral anticoagulation is stopped. Thromb Haemost2002;87(1):7–12.

12. Galli M et al. Lupus anticoagulants are stronger risk factorsfor thrombosis than anticardiolipin antibodies in the anti-phospholipid syndrome: a systematic review of the litera-ture. Blood 2003;101(5):1827–32.

13. Low-molecular-weight heparin in the treatment of patientswith venous thromboembolism. The Columbus Investigators.it N Engl J Med, 1997. 337(10): p. 657–62.

14. Buller HR et al. Subcutaneous fondaparinux versus intrave-nous unfractionated heparin in the initial treatment ofpulmonary embolism.N Engl J Med 2003;349(18): 1695–702.

15. Simonneau G et al. A comparison of low-molecular-weightheparin with unfractionated heparin for acute pulmonaryembolism. The THESEE Study Group. Tinzaparine ou Hepar-ine Standard: Evaluations dans l’Embolie Pulmonaire. N EnglJ Med 1997;337(10):663–9.

16. Anderson FJ et al. A population-based perspective of thehospital incidence and case-fatality rates of deep veinthrombosis and pulmonary embolism. The Worcester DVTStudy. Arch Intern Med 1991;151(5):933–8.

17. Heit J et al. Predictors of survival after deep vein throm-bosis and pulmonary embolism: a population-based, cohortstudy. Arch Intern Med 1999;159(5):445–53.

18. Goldhaber SZ, Visani L, De Rosa M. Acute pulmonaryembolism: clinical outcomes in the International Coopera-tive Pulmonary Embolism Registry (ICOPER). Lancet1999;353(9162):1386–9.

19. Douketis J et al. Risk of fatal pulmonary embolism inpatients with treated venous thromboembolism. JAMA1998;279(6):458–62.

20. Murin S, Romano PS, White RH. Comparison of outcomesafter hospitalization for deep venous thrombosis or pulmon-ary embolism. Thromb Haemost 2002;88(3): 407–14.

21. Palareti G et al. Bleeding complications of oral anticoag-ulant treatment: an inception-cohort, prospective collab-orative study (ISCOAT). Italian Study on Complications ofOral Anticoagulant Therapy. Lancet 1996;348(9025):423–8.

22. van der Meer FJ et al. Assessment of a bleeding risk index intwo cohorts of patients treated with oral anticoagulants.Thromb Haemost 1996;76(1):12–6.

23. Linkins LA, Choi PT, Douketis JD. Clinical impact of bleedingin patients taking oral anticoagulant therapy for venousthromboembolism: a meta-analysis. Ann Intern Med 2003;139(11):893–900.

24. Lee AY et al. Low-molecular-weight heparin versus acoumarin for the prevention of recurrent venous thrombo-embolism in patients with cancer. N Engl J Med 2003;349(2):146–53.