The Wells Rule May Not Be Reliable for All Patients
By Jennifer A. Best, MD
Assistant Professor, University of Washington School of Medicine, Seattle, WA
Dr. Best reports no financial relationships in this field of study.
SOURCE: Geersing GJ, et al. Exclusion of deep vein thrombosis using the Wells rule in clinically important subgroups: individual patient data meta analysis. BMJ 2014 Mar 10;348:g1340. doi: 10.1136/bmj.g1340.
In a patient presenting with leg symptoms, however non-specific, it is appropriate that a physician consider the diagnosis of deep venous thrombosis (DVT), as failure to make this diagnosis can have life-threatening consequences (e.g., pulmonary embolism). Many physicians utilize clinical decision rules such as the Wells score in deciding whether to proceed with additional diagnostic testing and/or treatment. The Wells score assigns an estimated likelihood of DVT, based on the presence or absence of specific clinical factors, including active cancer, paralysis/paresis/immobilization of leg, bedrest >3 days prior, major surgery <4 weeks prior, localized tenderness of the venous system, leg/calf swelling, pitting edema, collateral veins, history of DVT and likelihood of an alternative diagnosis. Scores are categorized as low (<0), moderate (1-2) and high (>3). The Wells rule is often utilized with the D-dimer; studies have shown that patients with a low Wells score and a negative D-dimer need not be anticoagulated empirically prior to additional diagnostic testing. However, some studies evaluating the Wells score have suggested troubling uncertainties. These include a primary care study showing that a combination of Wells and D-dimer resulted in a high number of missed DVT cases and additional concern that the Wells score is inadequate in patients of male sex, with active cancer or with DVT recurrence, as in all these groups the prevalence of DVT is higher than in a general population. Furthermore, patients with active malignancy may have an elevated D-dimer level, whether or not DVT is present. So although the score is well validated in unselected populations, it has been felt that these particular subgroups warrant additional study.
To address these uncertainties, Geersing and colleagues in the Netherlands and Canada pooled individual patient data from 13 studies evaluating patients with suspected DVT. Seven of these studies evaluated the utility of the Wells rule in management decisions, four sought to validate the Wells rule, and two studies aimed to determine whether a positive D-dimer plus an initially normal ultrasound or venogram warranted a serial follow-up study. All selected studies recruited primary care and hospital-based outpatients with suspected DVT; categorized patients by Wells rule (and included data to substantiate this categorization); included D-dimer testing; and documented presence or absence of DVT by compression ultrasonography or venography, or absence of adverse event within 3 months of presentation. These investigators also documented patient age, sex and presence or absence of previous DVT. Where important data were missing (between <1 and 5%, varying by data set), data values were imputed to minimize bias, and discrete data sets were then combined. Several statistical analyses were performed: logistic regression to determine the impact of each patient subgroup on the Wells score, the extent to which a negative D-dimer result contributed to Wells score in excluding DVT, and additional analyses evaluating the heterogeneity of the included studies.
10,002 subjects and 1,864 cases were included in the composite data set; the median age was 59 years, and 62% of patients were female. Nineteen percent of these subjects had proximal DVT. Increasing Wells scores did correlate with higher probability of DVT. However, even in patients with a Wells score of -2, the actual probability of DVT approached 5%, suggesting that the Wells rule alone is not adequate to exclude DVT in these patients. Additionally, with Wells scores up to 1, the probabilities of DVT were nearly doubled in patients with prior DVT and malignancy. Increased DVT probability was also seen in males, but to a lesser degree; overall, no significant differences in Wells performance between men and women were noted. When considering the probably of documented DVT in the setting of negative D-dimer and low Wells score (<1), the failure rate was 1.2%, which was felt to be within limits of acceptability for all groups except of those with malignancy and prior DVT. The combination of low Wells score and negative D-dimer (the "rule-out strategy") was efficient in excluding DVT in all subgroups except patients with cancer.
This study has a number of strengths. It is large, including more than 10,000 patients and a high number of DVT cases, but also some drawbacks. Few studies documented adequate blinding between reference test and D-dimer/Wells results. Additionally, many of these studies utilized compression ultrasonography as a reference standard, although it has been shown that this test is less reliable for recurrent events. Citing these considerations, the authors render the following clinical recommendations: 1) The Wells score is useful in estimating the pretest probability of DVT. The 9th ACCP guidelines recommend following a high Wells score with compression ultrasonography and a low Wells score with a D-dimer test. 2) There exists a modified version of the Wells rule, which assigns an extra point to the overall score in the setting of prior DVT. This model has never been validated, but in this population would have improved the failure rate to an acceptable level and is recommended for this unique population. In this modified version, a Wells score of 1 or less, with a negative D-dimer test, would exclude DVT in 1 of 3 patients. This holds regardless of sex, diagnostic setting, or of type of D-dimer assay. 3) These findings hold for all patient subgroups, with the exception of those with active cancer or recurrent thrombosis.