Deep vein thrombosis: a clinical prediction rule and investigation strategy can accurately diagnose DVT.

Clinical bottom line (level 1a)

  1. A clinical prediction rule can accurately rank patients as high, moderate or low risk of deep vein thrombosis.
  2. Using this management strategy, only 0.6% of deep vein thromboses are missed.
Wells et al: Lancet 1997; 350: 1795-1798
Expires May 2003

The study

Setting: two teaching hospitals, Canada

593 patients (aged mean 57 years, 58% female) outpatients with suspected deep venous thrombosis

Excluded if
  • <18 years old
  • previous objectively documented DVT or pulmonary embolism
  • signs or symptoms of current pulmonary embolism
  • patients requiring long term anticoagulation
  • patient could not be followed up
  • imminent death



    • Independent blinded reference standard, applied in all patients from a consecutive appropriate spectrum.
    Reference standard:
    • ultrasound followed by venogram if required (lack of vein compressibility used as sole diagnostic criteria for positive ultrasound); follow-up for three months
    Diagnostic test: clinical prediction rule used to rank patients into high, moderate or low risk for deep venous thrombosis:
    • active cancer (on-going treatment or diagnosed within 6 months or palliative care)- score 1
    • paresis, paralysis or recent plaster cast immobilisation of lower extremity- score 1
    • recently bedridden for more than 3 days and/or major surgery within 4 weeks- score 1
    • localised tenderness over distribution of deep veins- score1
    • entire leg swollen- score1
    • calf swelling more than 3 cm compared with asymptomatic side, measured at 10 cm below tibial tubercle- score 1
    • pitting oedema (greater in symptomatic leg)- score 1
    • collateral superficial veins (non-varicose)- score 1
    • alternative diagnosis as likely or greater than that of DVT- score -2

    • In patients with symptoms in both legs, the most symptomatic leg is used.
      • score 0 or less- low risk
      • score 1 or 2- moderate risk
      • score 3 or more- high risk
      • high risk: USS positive- DVT: USS negative- venogram, positive- DVT; negative- no DVT
      • moderate risk: USS positive- DVT: USS negative- repeat USS in one week; if positive- DVT; if negative- no DVT
      • low risk: USS positive- venogram, if positive- DVT; negative-no DVT: USS negative- no DVT

    The evidence

    pre-test probability of DVT: 16%, (95% CI: 13% to 19%)

    diagnostic test DVT no DVT LR+
    (95% CI)    		 post-test probability LR-
    (95% CI)    		 post-test probability
    clinical prediction rule- high 53 18 15
    (9.5 to 25)     		75% 0.46
    (0.37 to 0.58)     		8%
    clinical prediction rule- moderate 32 161 1.04
    (0.76 to 1.42)     		17% 0.98
    (0.84 to 1.15)     		16%
    clinical prediction rule- low 10 319 0.16
    (0.091 to 0.30)     		3% 2.49
    (2.17 to 2.85)     		32%
    total 95 498


    diagnostic test DVT no DVT LR+
    (95% CI)    		 post-test probability LR-
    (95% CI)    		 post-test probability
    ultrasound in low risk patients 9 2 140
    (36 to 580)     		82% 0.10
    (0.016 to 0.65)     		0%
    total 10 319


    diagnostic test DVT no DVT LR+
    (95% CI)    		 post-test probability LR-
    (95% CI)    		 post-test probability
    ultrasound in moderate risk 29 0 inf
    (49 to inf)     		100% 0.094
    (0.032 to 0.28)     		2%
    total 32 161


    diagnostic test DVT no DVT LR+
    (95% CI)    		 post-test probability LR-
    (95% CI)    		 post-test probability
    ultrasound in high risk 49 4 4.2
    (1.8 to 9.9)     		92% 0.092
    (0.035 to 0.26)     		22%
    total 53 18

    • Clinical prediction rule high (12% of patients)- DVT prevalence 75% (95% CI: 63%-84%).
    • Clinical prediction rule moderate (33% of patients)- DVT prevalence 17% (95% CI: 12% to 23%).
    • Clinical prediction rule low (56% of patients)- DVT prevalence 3.0% (95% CI: 1.7% to 5.9%).
    • 3/481 (0.6%, 95% CI: 0.1% to 1.8%) patients with low or moderate DVT risk and negative investigations developed DVT over the next three months.

    Comments

    1. Clinical prediction rule derived from Wells PS, Hirsch J, Anderson DR et al. Accuracy of clinical assessment of deep-vein thrombosis. Lancet 1995; 345: 1326-1330. Stepwise logistic regression analysis was performed on data to identify significant features. Coefficient rounded off to give score.
    2. Considered and rejected for prediction rule: age, symptom duration, sex, recent trauma, family history, erythema, hospital admission.
    3. K interobserver for prediction rule =0.75 (two nurses and two doctors).
    4. 5.6% of all patients required venogram; 28% required serial ultrasound scan testing, ie. 0.38 extra hospital visits per patient.
    5. 3/166 patients in the moderate risk group were found to have a DVT on serial ultrasound scan.
    6. The likelihood ratio calculated for ultrasound scanning of various risk groups, should be treated with caution since gold standard uses ultrasound- makes tests appear better than it should.

    Citation

    1. Wells PS, Anderson DR, Bormanis J, et al: Value of assessment of pretest probability of deep-vein thrombosis in clinical management. Lancet 1997; 350: 1795-1798
    Search Terms: hand search
    Contributor: Chris Ball and Clare Wotton, May 2000
    Reviewer: Daniel Sontheimer

    Clinical Question.
    Patient suspected deep venous thrombosis
    Intervention or Exposure clinical prediction rule
    Outcome diagnosis