A simplified, robust, clinical assessment, used in conjunction with a d-dimer assay, has been developed to identify patients with a low risk of deep vein thrombosis. These patients did not undergo radiological examination. The protocol has been used by junior doctors in a district general hospital with a predominantly elderly population. Four hundred and thirty-one patients with potential deep vein thrombosis of the lower limb were assessed using this method and 98 (22·7%) did not require further investigation. Of these, one was admitted 5 months later with a pulmonary embolus. None of the other patients subsequently required admission for a thromboembolic event or died during a minimum of 3 months follow-up.
The workload in most radiology departments has increased considerably over the past few years. In particular, there has been a large increase in requests for radiological tests to exclude lower limb deep vein thrombosis (DVT) and Doppler ultrasound has replaced venography to a great extent in straightforward cases. However, ultrasound is time-consuming and operator dependent, so alternative methods of excluding DVT are under investigation (Bounameaux et al, 1997; Freyburger et al, 1998).
The clinical diagnosis of DVT is notoriously inaccurate, but a clinical scoring system can help to assess the probability of a DVT in an individual patient (Wells et al, 1995). If the clinical risk of DVT is low, the absence of thrombosis is supported by a negative d-dimer assay (Wells et al, 1998).
There are several methods available for assaying d-dimer levels that are thought to be indicative of active clot formation and breakdown. The most sensitive methods are based on an enzyme-linked immunosorbent assay (ELISA), but these are also the most time-consuming and not appropriate when a rapid turnround time is required. Methods based on the visible agglutination of latex particles (Nycocard) or red cells (SimpliRED) are quick and simple to perform, but their interpretation is subjective, they are less sensitive and have a low specificity. d-dimer assays may be positive in the absence of venous thromboembolism in many situations, including pregnancy, post-operatively and in healthy elderly people.
The SimpliRED assay is probably not sensitive enough to be used alone in the exclusion of DVT, but can be useful in conjunction with a clinical score designed to identify patients in whom there is a low risk of DVT. The assay has been reported to have a sensitivity of 94% and a specificity of 71% in the diagnosis of lower limb DVT (Wells et al, 1998).
Patients and methods
Patients with suspected DVT were referred by their general practitioner to the admitting physicians or presented in the Accident and Emergency department. Inpatients with possible DVT, e.g. post-operatively, were not eligible for this assessment because of the exclusion criteria used in the clinical score; no other patients were excluded.
A simple clinical checklist, based on the ‘pretest probability’ score of Wells et al (1995), was devised for use by the junior medical staff (Fig 1). All the answers had to be negative if the d-dimer was to be used as an exclusion test and the patient avoid radiological imaging. If the d-dimer test was also negative, the patient could be discharged without imaging.
The SimpliRED d-dimer assay (Agen Biomedical) was introduced into our laboratory in July 1998. The assay was performed on EDTA samples by Biomedical Scientists in the Haematology laboratory, all of whom were trained in blood transfusion and were familiar with the assessment of red cell agglutination, in order to reduce any interobserver variation. The result was recorded as either positive (agglutination present) or negative (agglutination absent). In order to check the sensitivity of the assay in both DVT and pulmonary emboli (PE), it was also performed in all cases of suspected DVT, including high-risk cases, in which radiological diagnosis was mandatory, and in suspected pulmonary embolism; but in PE, the results were not released to the physicians responsible for the patients. Results were appended with the rider ‘A negative d-dimer result may only be used in conjunction with a negative clinical score for the exclusion of DVT’.
Ultrasound was performed by a Consultant Radiologist or by a supervised senior, experienced radiographer, both of whom were unaware of the d-dimer result. Colour duplex scanning was used to study the deep leg veins from the femoral vein to the proximal calf veins, looking for both venous compression and colour flow.
In all cases in which the radiology was positive but the d-dimer assay was negative, the patients' notes were examined and the radiology reviewed, if required, by the authors. As the population served by St Richard's Hospital is fairly static, follow-up was carried out by searching computer records for hospital-related events and patient deaths.
Between July 1998 and December 1999, 557 patients with suspected thromboembolic disease had a d-dimer assay performed: 431 with a putative diagnosis of DVT and 126 with pulmonary embolism.
Three hundred and six patients with suspected DVT had an ultrasound (or in a few cases, a venogram) and, of these, 93 (30·4%) were positive (21·6% of the total presenting with possible DVT). One hundred and twenty-five patients did not have an ultrasound, 98 because they fulfilled the exclusion criteria and 27 for clinical reasons (Fig 2).
Two hundred and thirteen patients with suspected DVT had a positive d-dimer assay: 27 had no further investigation and, of the 186 who had an ultrasound, 78 (42%) had a DVT. Two hundred and eighteen patients had a negative d-dimer assay, but 120 of these had a positive clinical risk score and so proceeded to ultrasound: 15 had a positive ultrasound. Therefore, 93 patients had a DVT diagnosed by ultrasound and, of these, 78 had a positive d-dimer test, giving a sensitivity of 84% [95% confidence intervals (CI) 78–90%]. This number includes two patients with previous DVT and three patients with calf DVT, both situations in which confirmation of a DVT by ultrasound can be unreliable. Review of the notes suggested that 10 patients had good evidence of a DVT on clinical and radiological grounds, which would give a revised sensitivity of the test of 89% (95% CI 83–94%). The negative predictive value (NPV) of the assay was 95%; the specificity of the d-dimer assay was only 50%. These figures assume that patients who did not re-present within 3 months did not have a clinically significant DVT. No patient was receiving heparin when the d-dimer assay was performed.
Ninety-eight patients with a negative clinical score for DVT and negative d-dimer assay were discharged without further investigation. One patient was admitted 5 months later with a pulmonary embolus; no others were admitted or died within a minimum follow-up period of 3 months from presentation.
One patient with a positive clinical score and d-dimer test and a negative ultrasound was readmitted 1 week later with a confirmed pulmonary embolus; four others were anticoagulated despite a negative ultrasound on the basis of risk factors and clinical findings: all these had positive d-dimer assays.
One hundred and twenty-six patients with possible pulmonary embolus had a d-dimer assay (no clinical scoring system was employed in the diagnosis of PE, and d-dimer results were not issued when PE was stated as the primary diagnosis). Ninety-one of these patients had a ventilation-perfusion (VQ) scan and, of these, 27 (29·7%) showed a moderate to high probability of pulmonary embolus (21·4% of the total). Twenty patients had a positive and seven patients a negative d-dimer test, giving a sensitivity of 74%.
In most of the 27 patients with a positive d-dimer assay who did not have an ultrasound or VQ scan, the junior doctor's original differential diagnosis had been changed by a more senior physician, for example to cellulitis or chest infection. These patients were treated for their clinical diagnosis and not for venous thromboembolism under the direction of the physician in charge of their care. They were excluded from the analysis.
We have developed a simple clinical scoring system that has been used by junior doctors and casualty officers, in conjunction with a SimpliRED assay for d-dimer, to identify patients at low risk of DVT. This protocol could also be used by general practitioners with access to a rapid d-dimer assay and, therefore, fewer patients need attend hospital.
Our figures, which show that approximately only one in five presenting patients actually have a DVT, are similar to other studies (Wells et al, 1995). Because of this low positive rate, elimination of the large low-risk group is very useful in reducing the need for further investigation.
Patients who have no identifiable risk factors for DVT and whose clinical signs are not classical have been shown to have a low (approximately 5%) risk of actually having a DVT when they present (Wells et al, 1995). By making the exclusion criteria stricter, such that if the patient was positive for any of the criteria they were excluded from the low-risk group, we hoped to reduce the risk of DVT in this group to a minimum. This was important because, in our experience, the sensitivity of the d-dimer assay has been 84–90%. The combination should therefore reduce the risk of missing a DVT to less than 1%, which is comparable to the accuracy achieved using a pretest probability and ultrasound (Wells et al, 1998). We have therefore reduced the risk of false negative results, but at the expense of performing more ultrasound examinations. There is the potential to reduce the radiology workload further with an assay with increased sensitivity and specificity. Our criteria could also be revised to exclude family history, which has not been shown to be a risk factor (Wells et al, 1997).
We have found the SimpliRED assay to have a lower sensitivity than has sometimes been quoted. Some small studies have found the assay to be 100% sensitive (Mayer et al, 1996; Turkstra et al, 1996). Wells et al (1998) used pretest probability and SimpliRED to define a low-risk group and reduce the number of serial ultrasounds required in patients in whom the initial ultrasound was negative. They quote a sensitivity of 94% and a NPV of 98%. The same group used pretest probability, d-dimer and impedance plethysmography to exclude DVT in 398 patients. Four patients with proximal vein thrombosis had a negative d-dimer assay, giving a NPV for the test of 97% (Ginsberg et al, 1997). A rapid ELISA d-dimer test with a cut-off of 500 μg/l was used to exclude DVT and PE, irrespective of the clinical probability, in 31% of 918 patients (Perrier et al, 1999). The 3 month thromboembolic risk in patients who were not anticoagulated was 1·8%, although ultrasound examinations were also performed on patients with negative d-dimer and suspected DVT. In an overview, Bounameaux et al (1997) quote an overall sensitivity for d-dimer assays of 82%, with 95% confidence intervals of 77–87%.
One group investigated 47 patients with false negative d-dimer assays and found that the risk of false negative result was increased if the symptoms have been present for a particularly long or short time, if the patients are receiving heparin or if the thrombus is small (unpublished observations). Other authors have found d-dimer assays to be more sensitive in the diagnosis of PE. In our study, five of the patients with a diagnosis of PE and a negative d-dimer had a probable diagnosis based on clinical findings and a moderate to high probability VQ scan, with no alternative diagnosis. One patient had an axillary vein thrombosis and a chest infection was probable in one further patient. On the basis of these figures, we do not use d-dimer in the exclusion of PE.
The introduction of a standardized clinical score has also improved the detection of high-risk patients (two or more risk factors, excluding d-dimer result) in whom a negative ultrasound is unexpected. Venography or serial ultrasound studies can be focused on these patients.
d-dimer assays are often said to be positive in healthy elderly people. Our hospital serves a predominantly elderly population and we identified 77 patients aged over 80 years who had a d-dimer assay at presentation with a possible DVT. Eleven of these patients (14%) had a negative clinical score and negative d-dimer and were discharged. Thirty-two (42%) had a ‘false positive’d-dimer assay, but many of these patients, as would be expected in an elderly population, had other underlying pathologies such as malignancy, which are often associated with increased fibrinolysis and positive assays. The incidence of false positive assays in the whole patient population was 25%. Therefore, elderly patients probably require more radiology, but a significant number can still be discharged directly.
We have used this diagnostic protocol and developed the outpatient management of DVT to substantially reduce the number of patients requiring admission when they present with a possible DVT. Most patients do not have a DVT when they present and, using the clinical score and d-dimer, we have identified a significant proportion of these, thus reducing the number of patients being treated with low-molecular-weight heparin while awaiting an ultrasound examination and also reducing the possibility of bleeding complications connected with heparin administration. Limited radiological resources can be focused on high-risk cases. Patients requiring radiology after their clinical assessment can usually be discharged on low-molecular-weight heparin and scanned on the next ultrasound list. Beds are not being occupied by patients waiting for confirmatory tests, a development which helps the hospital and is generally popular with patients.