Anticoagulation: computer-dosing of warfarin meant shorter time to a stable therapeutic INR.

Clinical bottom line (level 1b)

  1. Patients who were prescribed warfarin using a computer program rather than based on physician's prediction reached a therapeutic and stable INR faster, and left hospital sooner.
  2. Patients were more likely to have a therapeutic INR at 2 weeks (NNT = 2 at 2 weeks) .
White et al: Journal of General Internal Medicine 1987; 2: 141-148
Expires July 2003

The study

Unblinded concealed randomised trial with intention-to-treat
Setting: general hospital, USA

75 patients (aged mean 59 years, 88% male) started on warfarin therapy

Excluded if
  • <19 or >90 years old
  • pregnant
  • ongoing treatment with a warfarin competitor (eg. phenylbutazone)
  • on an inducer/ inhibitor of warfarin within the last month
  • recent vitamin K or FFP
  • bleeding disorder
  • cancer
  • uncontrolled heart failure
  • sepsis
  • malabsorption


    • Control Group: (n = 36, 36 analysed): intern predicted next day's dose
    Experimental Group: (n = 39, 39 analysed): computer-controlled dosing
    Patients had 10 mg of warfarin initially (5 mg if valve replacement) for 2 to 3 days, and adjusted so that PT ratio was 1.4 to 2.2 (INR ~3 to 4.5) after that.
    91% followed for 14 days

    The evidence

    Outcome Time to outcome CEREERRRR
    (95% CI)    		ARR
    (95% CI)    		NNT
    (95% CI)
    supratherapeutic PT 14 days 7
    (19.4%)    		 2
    (5.13%)    		 74%
    (-19% to 94%)    		 14.3%
    (-0.35% to 29.0%)    		 7
    (NNT = 3 to infinity;
    NNH = 290 to infinity)
    any bleed 14 days 3
    (8.33%)    		 0
    (0.00%)    		 100%
    (% to %)    		 8.33%
    (-0.70% to 17.4%)    		 12
    (NNT = 6 to infinity;
    NNH = 140 to infinity)
    no therapeutic PT 14 days 25
    (69.4%)    		 11
    (28.2%)    		 59%
    (30% to 76%)    		 41.2%
    (20.6% to 61.9%)    		 2
    (2 to 5)

    Outcome Control Group
    (SD)    		 Experimental Group
    (SD)    		 Mean Difference
    (95% CI)
    time to therapeutic dose (days) 4.5
    (3.4)     		3.2
    (1.6)     		1.3
    (0.09 to 2.5)
    time to stable PT (days) 9.4
    (5.2)     		5.7
    (1.7)     		3.7
    (2.0 to 5.5)
    length of hospital stay (days) 20
    (15)     		13
    (8.0)     		7.0
    (1.5 to 12)

  • One patient in the intern-adjusted group had a major bleed.

    • Comments

    1. The study was too small to show any effect on bleeding or patients with supratherapeutic PTs
    2. In order to enhance patient safety, warfarin should be initiated at a dose of 5 mg per day, not 10 mg per day as was the case when this study was performed, and dosing algorithms should be utilized. Whether these newer algorithms would make a difference to the study outcomes is uncertain, but unlikely.
    3. The computer program used pharmacokinetic variable estimates and pharmacodynamic population parameters with each patient's PT response to individual doses of drug to predict the maintenance dose.
    4. Need to consider the usefulness of the computer for patients with multiple comorbidities and for those on medications that interact with warfarin.

    Citation

    1. White RH, Hong R, Venook AP, et al: Initiation of warfarin therapy: comparison of physician dosing with computer-assisted dosing. Journal of General Internal Medicine 1987; 2: 141-148
    Contributor: Chris Ball and Clare Wotton, July 2000
    Reviewer: Deepak L Bhatt

    Clinical Question.
    Patient started on warfarin
    Intervention or Exposure computer-controlled dosing
    Comparison physician predicted next day's dose
    Outcome therapeutic INR, length of hospital stay