Oral anticoagulation and risk of death: a medical record linkage study

Sir,

In their comprehensive review of the monitoring of anticoagulant
therapy in Sweden, Oden and Fahlen(1) emphasise the excess mortality
associated with high INR values(2), particularly in those whose dose of
anticoagulant had recently been increased. They do not, however, discuss
the implications of sub-therapeutic INR in those for whom oral
anticoagulant therapy has been recommended. We recently conducted a
mortality study over a 2 year period of 4686 patients registered for
monitoring of oral anticoagulant therapy in a central teaching hospital.
523 (11%) died during the 24 months. In 52 of these it was considered
likely that anticoagulant control may have influenced the outcome of the
terminal illness. In 13, a high INR was associated with bleeding (5 intra
cerebral haemorrhage, 6 intra abdominal bleeding and 2 “stroke”). A
haemorrhagic death occurred in a further 25 patients when the INR was
within the therapeutic range (9 intra cerebral haemorrhage, 11 acute
“stroke”, 3 gastrointestinal bleeding, 1haemothorax and 1 dissecting
aneurysm).

Of the 14 patients who died when their INR was subtherapeutic, 10 had
a thrombotic cause (3 pulmonary embolism and 7 cerebral thrombosis).
Amongst those who died there were also 20 whose anticoagulant therapy had
been discontinued some weeks or months before without notification to the
anticoagulant clinic and amongst these 11 (55%) of these succumbed to a
thrombotic death (8 ‘stroke’, 2 venous thrombembolism, 1 mesenteric
vascular occlusion). There were, therefore, a total of 21 patients out of
76 studied who died as a result of thrombosis when either anticoagulant
therapy had been discontinued or its dosage was subtherapeutic.

We would therefore, caution equally against sub optimal
anticoagulation in those for whom there is a clear indication for its
prescription. The morbidity after stopping Warfarin demonstrates that a
very careful assessment of risk versus benefit must be undertaken before
treatment is discontinued.

Great care is taken, quite properly, to avoid the severe consequences
of haemorrhagic complications of oral anticoagulant therapy but our study
highlights the equally serious outcome of under-coagulation in patients at
risk of stroke(3)

Yours faithfully

DR Vanessa J Martlew,
Consultant Haematologist,

Dr Ewan Wilkinson,
Consultant in Public Health Medicine,

Royal Liverpool University Hospital,
Prescot Street,
Liverpool L7 8XP

1. Oden A and Fahlen M
Oral anticoagulation and risk of death: a medical record
WH Kay Study
BMJ 2002 325: 1073-1075

2. Haemostasis & Thrombosis Task force of the British Committee
for Standards in Haematology –
Chair: ID Walker.
Guidelines on Oral Anticoagulant Therapy: 3rd Edition
Brit J Haematol. 1998 101: 374-374

3. Stroke Prevention in Atrial Fibrillation Investigations.
Adjusted does Warfarin versus low-intensity fixed dose Warfarin plus
Aspirin for high risk patients with atrial fibrillation.
Lancet: 1996: 633-638

Competing interests:  
None declared

Oden and Fahlen [1] present useful information regarding oral
anticoagulation and the risk of death and conclude that patients with an
international normalised ratio (INR) value in the range 2.2-2.3 are at the
lowest risk of death. They go on to suggest that anticoagulation treatment
should be less intensive and that the therapeutic window for all patients
receiving anticoagulation therapy should be narrower with "INR close to
2.2-2.3". We believe that this latter advice requires further
consideration.

We analysed 26122 INR measurements representing 1696 patients with
the therapeutic range 2.0-3.0. Just 13766 (52.7%) of the INR measurements
were in range, a result in broad concurrence with those of other studies
[2,3]. The proportion of patients in, above and below therapeutic range
changed little with duration of anticoagulation treatment (see figure).

This arguably poor quality of anticoagulation control is certainly
not for the want of trying. Rather, the complex pharmacology of warfarin
and uncertainty concerning the large number of drugs and medical
conditions which may effect warfarin dose-response [4] makes
anticoagulation control very challenging. A narrow therapeutic window
close to 2.2-2.3 is therefore likely to prove unrealistic and
unachievable.

Many clinicans would of course claim that they are used to being set
unrealistic and unachievable targets. However, it is plausible that
attempting to achieve such narrow target ranges could degrade the quality
of control delivered to patients. Anticoagulation control has strong
parallels with extensively studied control engineering methods [5] that
are used in non-medical applications. Such studies suggest that aiming to
achieve and maintain INRs within a narrow target range could result in
system instability, poorer control and an excessive amount of adjustment.
One analogy is the type of system behaviour many novice drivers experience
when they first try to steer a car and find that it snakes from side to
side despite their increasingly desperate attempts to keep it going in a
straight line. In the case of anticoagulation therapy, one is further
hampered since the "steering wheel" of dose modification has a patient-specific, non-linear and time-varying response and one's eyes are closed
to the results of the "steering" until the next outpatient appointment.

By attempting to achieve and maintain INR values within a narrow
target INR range, it is conceivable that patients may be subject to
excessive dose changes and monitoring. Oden and Fahlen [1] provide an
excellent starting point for further investigation of control engineering
analysis of anticoagulation therapy.

Yours sincerely,

Martin Utley, David Patterson, Steve Gallivan

References

[1] Oden A, Fahlen M, Oral anticoagulation and risk of
death: a medical record linkage study, BMJ 2002;325:1073-5

[2] Rose P., Audit of anticoagulant therapy, J Clin Pathol, 1996;48:5
-9

[3] Doble N., Baron J.H., Anticoagulation control with warfarin by
junior doctors, J R Soc Med 1987;80:627

[4] Wells PS, Holbrook AM, Crowther NR et al, Interactions of
warfarin with drugs and food, Ann Intern Med. 1994;121:676-683.

[5] Burghes D, Graham A, Introduction to control theory including
optimal control, Chichester, Ellis Horwood, 1980.
References

Competing interests: �
None declared