Wells Score for deep vein thrombosis

Clinical background

Suspected deep vein thrombosis (DVT) of the lower limb is a common problem in which the prevalence among patients referred for investigation is only around 10 to 15 per cent. This means that the great majority of ultrasound examinations performed are negative. At the same time the consequences of missing a DVT are serious: pulmonary embolism and post-thrombotic syndrome. The decision whether a patient should proceed to objective diagnostic testing, or whether testing can safely be withheld, is therefore both common and difficult. The Wells score is intended to structure the clinical assessment into a pre-test probability, which is then combined with the D-dimer to determine whether ultrasound can be avoided.

Calculating the Wells score for deep vein thrombosis

The score is the sum of nine clinical variables, where each positive criterion scores 1 point and an alternative diagnosis at least as likely as DVT subtracts 2 points:

$$\text{Wells score} = \sum_{i=1}^{9} x_i - 2 \cdot x_{\text{alternative diagnosis}}$$

where each $x_i \in {0, 1}$ corresponds to one of the nine criteria: active cancer, paresis/paralysis/recent immobilisation of the leg, bedridden >3 days or major surgery within 12 weeks, localised tenderness along the deep venous system, entire leg swollen, calf circumference >3 cm greater than the other leg, pitting oedema confined to the symptomatic leg, collateral superficial (non-varicose) veins, and previously documented DVT. The variable $x_{\text{alternative diagnosis}}$ is set to 1 if an alternative diagnosis is judged at least as likely, otherwise 0.

The derivation cohort consisted of outpatients with suspected lower-limb DVT recruited in Canada. The study that established the two-level interpretation (DVT likely or unlikely) included 1,096 patients, of whom 530 were randomised to a control group undergoing ultrasound and 566 to a D-dimer group in which ultrasound could be avoided when the clinical probability was low and the D-dimer negative [1]. The prevalence of DVT was 15.7 per cent. The outcome measure was confirmed venous thromboembolism at baseline or during three months of follow-up.

Interpretation in practice

The Wells score is used in a two-level model with the threshold at 2 points:

In the original study, patients with a score ≤1 and a negative D-dimer had a rate of venous thromboembolism of 0.4 per cent over three months, compared with 1.4 per cent in the control group who underwent ultrasound regardless of clinical probability [1]. Of the patients in the D-dimer group, 39 per cent did not need to undergo ultrasound.

An older three-level model (low 0, moderate 1 to 2, high ≥3) exists but has largely been abandoned in favour of the two-level model, which is simpler to combine with the D-dimer in a binary algorithm.

Validation and performance

The most extensive external validation is an individual patient data meta-analysis of 13 diagnostic studies comprising 10,002 outpatients in Canada, the Netherlands, the USA and Sweden [2]. Of these, 19 per cent had proximal DVT. The study confirmed that a rising Wells score was associated with a rising probability of DVT, and that the combination of an unlikely Wells score (≤1) and a negative D-dimer gave a failure rate of 1.2 per cent (95% CI 0.7 to 1.8). This combination occurred in 29 per cent of patients, meaning that just under a third could avoid ultrasound. The results were consistent regardless of sex, care setting (inpatient or outpatient) and type of D-dimer assay.

In primary care, however, performance has been questioned. A Dutch validation study in primary care comprised 1,295 consecutive patients and found that 12 per cent of patients in the low-risk group had DVT, compared with 3 per cent in the original cohort [3]. With a negative D-dimer the prevalence fell to 2.9 per cent, compared with 0.9 per cent in the original study. The authors concluded that the Wells score alone or combined with the D-dimer does not guarantee a reliable risk estimate in primary care populations, where the prevalence and case mix differ from secondary care.

For patients with active cancer, the Wells score combined with the D-dimer is neither safe nor efficient. In an individual patient data meta-analysis of 834 cancer patients among 10,002 cases of suspected DVT, the prevalence of DVT was 37.5 per cent compared with 15.1 per cent in patients without cancer [4]. The failure rate with an unlikely Wells score and a negative D-dimer was 2.2 per cent, and the combination occurred in only 9 per cent of cancer patients. A recently developed five-variable prediction model for cancer patients achieved a c-statistic of 0.80 but could still identify only 4.3 per cent with a post-test probability below 2 per cent. The conclusion is that cancer patients with suspected DVT should proceed directly to compression ultrasound.

In suspected recurrent ipsilateral DVT the Wells score performs less well. In a European multicentre study (the Theia study) of 231 patients not receiving anticoagulation and with a DVT prevalence of 45 per cent, the failure rate was 6.1 per cent (95% CI 1.3 to 18) with an unlikely Wells score and a negative D-dimer [5]. Sensitivity was 97 per cent but specificity only 36 per cent. A modified Wells score, adding an extra point for previous DVT, had an even higher failure rate in this cohort. The authors recommend that all patients with suspected recurrent ipsilateral DVT undergo ultrasound regardless of the Wells score and D-dimer.

Limitations

The Wells score was derived for and validated in outpatients. For inpatients there is insufficient validation evidence, and the score should be used with caution in this population.

The following patient groups are important exceptions in which the score is not reliable:

• Active cancer: the prevalence of DVT is more than twice as high, the D-dimer is often raised for other reasons, and the combination of Wells score + D-dimer safely excludes DVT in only a fraction of patients [2, 4]. Direct referral to compression ultrasound is recommended.
• Suspected recurrent ipsilateral DVT: the failure rate is too high for the Wells score + D-dimer to replace ultrasound [5]. The modified Wells score (extra point for previous DVT) has produced contradictory results and cannot be recommended routinely.
• Primary care patients: although the IPD meta-analysis showed consistent results across care settings, a dedicated primary care study has shown a higher failure rate than in the original cohort [3]. The clinician should be aware that the prevalence, and therefore the positive predictive value, may differ.

A further limitation is the subjectivity of the variable "alternative diagnosis at least as likely", which can give −2 points and thereby move a patient from the likely to the unlikely category. This judgement varies between assessors and affects the result considerably. In the IPD meta-analysis, data on this variable were missing in up to 5 per cent of patients [2].

The score applies only to suspected DVT of the lower limb and should not be used for upper-extremity venous thrombosis or suspected pulmonary embolism, for which other instruments exist (the Wells score for pulmonary embolism, the Geneva score).

References

1. Wells PS et al. Evaluation of D-dimer in the diagnosis of suspected deep-vein thrombosis. N Engl J Med. 2003. PMID: 14507948
2. Geersing GJ et al. Exclusion of deep vein thrombosis using the Wells rule in clinically important subgroups: individual patient data meta-analysis. BMJ. 2014. PMID: 24615063
3. Oudega R et al. The Wells rule does not adequately rule out deep venous thrombosis in primary care patients. Ann Intern Med. 2005. PMID: 16027451
4. Takada T et al. Diagnosing deep vein thrombosis in cancer patients with suspected symptoms: an individual participant data meta-analysis. J Thromb Haemost. 2020. PMID: 32433797
5. van Dam LF et al. Safety of using the combination of the Wells rule and D-dimer test for excluding acute recurrent ipsilateral deep vein thrombosis. J Thromb Haemost. 2020. PMID: 32613731