The Rehabilitation of Clinical Assessment in the Diagnosis of Deep Vein Thrombosis

CLINICAL SCORE FOR PROXIMAL DEEP VEIN THROMBOSIS Strategies for using accurate testing to exclude or diagnose acute proximal deep vein thrombosis (DVT) should always start with a clinical suspicion of DVT. Landefeld et al.[ 1 ] provided a sound basis for quantifying clinical judgment for the diagnosis of acute proximal DVT affecting the popliteal, femoral, and iliac veins. Combinations of clinical findings are often more useful than individual signs and symptoms in estimating the probability of a particular diagnosis. Landefeld applied this technique to the diagnosis of proximal DVT by comparing systematically the clinical signs and physical symptoms in patients with phlebographically proven proximal DVT and in patients with normal venograms. In his study, symptoms and swelling of the lower legs were the chief complaints in nearly two thirds of patients with suspected DVT. Five clinical findings could be identified to be independently related to positive venograms as shown by linear discriminant analysis: cancer, immobility, swelling above the knee, swelling below the knee, and measurement of calf circumference of both legs (Table [ 1 ]). Increase of calf swelling in the affected lower leg was a very sensitive indicator of acute proximal DVT. The striking differences in calf circumferences between the affected and the normal legs in patients with DVT (p-value 0.0001, Table [ 1 ]) strongly suggest that such measurements are very helpful and should be included in the diagnostic work-up in patients with suspected venous thromboembolism. As shown in Figure [ 1 ], the number of clinical findings present was correlated directly with the probability of acute proximal DVT. In patients with none, one, two, and three or more of these clinical findings, venographically proven DVT was present in 8%, 14%, 35%, and 55%, respectively. Subsequently, Landefeld et al.[ 1 ] tested the predictive accuracy of the number of clinical findings in an independent testing group of 119 patients with suspected DVT. In this testing group, the probability of proximal DVT was also predicted by the number of clinical symptoms present (Fig. [ 1 ]). In the groups of patients with none, one, and two or more of these clinical findings, phlebographically proven DVT was present in 5%, 15%, and 42%. Landefeld et al. recommended that the diagnostic value of the measurements in estimating the pretest and post-test clinical probability of proximal DVT should routinely be examined systematically in all patients with suspected DVT.[ 1 ] Estimates of the probability of proximal DVT should be very useful in choosing appropriate diagnostic tests to evaluate the possibility of DVT in symptomatic patients and in interpreting their results. Landefeld introduced the concept that a reasonable clinician might defer further testing for proximal DVT when the probability of proximal DVT is 1% or less and institute long-term anticoagulant therapy when the probability of proximal DVT is 90% or greater.[ 1 ] Wells et al.[ 2 ] [ 3 ] extended the concept of Landefeld and developed a modified clinical model to stratify pretest clinical probability (PCP) for DVT into low-, moderate-, and high-risk groups for having proximal DVT. Items included in this clinical model were derived and assembled from information obtained from consensus of the participating investigators. The items they included were signs and symptoms of DVT, risk factors for DVT, and potential alternative diagnosis. In Wells's first study, clinical probability for proximal DVT was estimated as low, moderate, and high based on a checklist of major and minor criteria.[ 2 ] In the second study of Wells et al.,[ 3 ] the checklist was simplified to a score list of clinical features (Table [ 2 ]), which indeed appeared to be very similar to the predictive factors as defined by Landefeld et al.[ 1 ] (Table [ 1 ], Fig. [ 1 ]). The presence of an alternative diagnosis as likely or greater than that of DVT was scored as minus-2. When the PCP model is used by inexperienced observers such as physicians in training for internal medicine or surgery, the assumption of no alternative diagnosis is very likely to ensure the highest level of safety in predicting the clinical probability for proximal DVT.[ 4 ] [ 5 ] Therefore, we propose not to use ``alternative diagnosis minus-2 scores'' in the modified clinical model of predicting the probability of proximal DVT (Table [ 2 ], Fig. [ 2 ]).[ 4 ] [ 5 ]