You are evaluating a 45-year-old male who is complaining of calf pain. He has a history of cancer however he has never had a clot in the past. The leg is neither swollen nor warm but he notes a cramping sensation in the posterior portion of his calf. You are concerned for a deep vein thrombosis (DVT) and consider the multiple means to reliable exclude the diagnosis: Wells score, D-dimers, ultrasound? What works?
Deep Vein Thrombosis: Background
The concern when approaching a patient with potential DVT is for the downstream embolic complications particularly pulmonary embolism (PE). In patients evaluated with a suspected DVT, the prevalence of thrombosis has been estimated to be 10-15% 1 demonstrating the low testing threshold of many physicians when encountering this diagnosis for fear of future morbidity. The role for peripheral testing, if it is to be useful particularly to exclude thrombosis, is to lower the 10-15% pre-test probability further to a reliable and predictable percentage conventionally considered to be safe. Currently a 2% miss rate is considered acceptable 2 and is therefore utilized as a safety metric by which to measure the usefulness of ancillary clinical and/or laboratory tests.
The Wells rule has been utilized as a clinical decision instrument to help physicians adjust patient risk based upon assessed clinical parameters. As a reminder, the criteria are as follows:
- Active cancer (+1 point)
- Bedridden recently >3 days or major surgery within four weeks (+1 point)
- Calf swelling >3 cm compared to the other leg (+1 point)
- Collateral (nonvaricose) superficial veins present (+1 point)
- Entire leg swollen (+1 point)
- Localized tenderness along the deep venous system (+1 point)
- Pitting edema, greater in the symptomatic leg (+1 point)
- Paralysis, paresis, or recent plaster immobilization of the lower extremity (+1 point)
- Previously documented DVT (+1 point)
- Alternative diagnosis to DVT as likely or more likely (-2 points)
Those with 0-1 points are deemed “low risk.” Combining the D-dimer with a Wells score seems to impart a low enough risk to not require further testing or the risk of empiric anticoagulation. However, there have been recent studies that question this strategy arguing that the risk is unacceptably high in this group and further that this strategy is useless in particular higher-risk subgroups (e.g. males, active malignancy, recurrent DVT).
Geersing GJ, Zuithoff NP, Kearon C, 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.
- Meta-analysis of individual patient data from 13 diagnostic studies of patients with suspected DVT to determine the diagnostic safety of utilizing Wells criteria and D-dimers, specifically as they apply to different patient subgroups
- Authors contacted principle investigators of included primary studies to obtain original patient data
- Inclusion criteria for studies
- Consecutive patients with suspected DVT
- Dataset with Wells parameters
- D-dimer results (if performed)
- Reference standard for DVT
- Compression ultrasonography or venography
- Uneventful follow-up in 3 months
- 13 included studies: Canada, the Netherlands, U.S., Sweden
- In the new constructed dataset: n=10,002 patients
- Missing data
- Five studies did not record D-dimer levels
- <1% presence/absence of DVT
- 5% “Alternative diagnosis as likely as or more possible than DVT”
- 864 (19%) had proximal DVT
- Median age: 59 years
- Female: 62%
- Missing data
- Lowest possible Wells score in isolation = -2 points
- DVT rate = 5% (range 2.0-5.9%)
- Wells score in isolation = 0 points
- DVT rate = 7.9% (range 6.1-10.2%)
- Low Wells score (≤1) + negative D-dimer test
- DVT rate (a.k.a failure rate) = 1.2% (0.7-1.8%)
- Worked for all subgroups EXCEPT those with a history of cancer
- Excluding cancer, this approach would exclude DVT in 1 in every 3 patients with suspected DVTs
The Bottom Line
The study demonstrates that the strategy of combining a Well score + D-dimer was superior to Wells alone and would allow the exclusion of DVT in 1/3 of patients who presented with this scenario. Importantly, it was found that patients with a history of active malignancy continued to have an unacceptably high risk of DVT and were not able to be safely excluded utilizing this strategy. It has been demonstrated 3 that a targeted D-dimer strategy (that is risk-based D-dimer testing) is useful in the work-up in DVT and may preclude the need for ultrasonography. In the pooled data above 1 out of every 3 individuals who presented with a low risk Wells score (≤1) combined with a negative D-dimer would not require further testing.
Words of Caution
A few important points regarding the interpretation of this study must be raised however when analyzing the clinical implications.
- An extraordinarily low risk (<2%) was defined as a baseline safety metric in order to deem the testing strategy efficacious. Though the Wells+ D-dimer approach allows the reduction of risk amongst the cohort to <2%, the range of probabilities for low risk Wells scores alone (-2 to 1 point) were 2.8% to 12.8% respectively. In spite of the fact that safety was a priori defined as <2%, it is not clear that the reduction of risk from say 5% pre-test probability to less than 2% has any real clinical benefit.
- Interpreting the results in reverse, it will require that 2/3 patients undergo further testing in order to reduce the probability of one patient to < 2%.
- Intrinsic to the Wells score is a clinical assessment of the patient— if a singular assessment of “alternative diagnosis more likely” by a physician (with no other Wells findings) is associated with an at best 5% risk of DVT, then one could add a D-dimer in the hopes of reducing risk to <2% or be comfortable with what should be considered an extraordinarily low risk of disease based upon a more specific test (history and physical) than D-dimer. Phrased in a different fashion, clinical judgement alone (alternative diagnosis is more likely) risk stratifies individuals to such a reasonably low level that the net gains of additional testing would seem to be offset by the amount of false positives generated by further testing to reduce 3 percentage points. Though the current medical fashion suggests that we need to drastically reduce the risk of all patients at the expense of specificity it is not clear that in the aggregate we are doing more good than harm.
The study is intriguing both from its findings and its reflections of the current medical diagnostic quandaries. It would seem that the culture of diagnostic medicine, at least in the U.S., continues to be one of extraordinary discomfort with uncertainty, necessitating testing strategies that require unrealistic levels of performance. The trend of diagnostics continues to be a movement towards perfection finding every disease no matter how small the risk or trivial the disease at the expense of unnecessary work-ups and a multitude of false positives. A study like this, while helpful in some respects, demonstrates quite well our generation’s approach to diagnostic uncertainty. Clinically this study is intriguing and should be reproduced prospectively in order to confirm the findings of a pooled group of disparate studies. The meta-analysis was extraordinarily well done but still subject to the statistical vicissitudes of lumped data from various studies pooled into one comprehensive dataset.
Updated 10/28/14: Corrected typo that Wells criteria of “alternative diagnosis to DVT as likely or more likely” was -2 points and NOT -1 point.