Current issues of ACP Journal Club are published in Annals of Internal Medicine


Review: A prediction guide that includes clinical assessment and noninvasive tests is accurate for diagnosing deep venous thrombosis

ACP J Club. 1998 Sep-Oct;129:47. doi:10.7326/ACPJC-1998-129-2-047

Source Citation

Anand SS, Wells PS, Hunt D, et al. Does this patient have deep vein thrombosis? JAMA. 1998 Apr 8;279: 1094-9.



What is the validity of clinical assessment and noninvasive tests for the diagnosis of suspected deep venous thrombosis (DVT)?

Data sources

English-language studies were identified from a MEDLINE search (1966 to April 1997) using the terms thrombosis, physical examination, diagnostic tests, sensitivity and specificity, and phlebography and from the bibliographies of relevant papers.

Study selection

Studies were selected if they addressed the diagnosis of DVT.

Data extraction

Data were extracted on risk factors, symptoms, and physical signs.

Main results

68 articles were selected for review, of which only 5 provided data on the relation between clinical findings and venographic confirmation of DVT. A cohort study showed that about 50% of patients had a major risk factor for DVT (immobility, trauma, or recent surgery). Other factors, such as male sex (odds ratio [OR] 1.7), age > 60 y (OR 1.6), heart failure (OR 1.8), lower limb arteriopathy (OR 1.9), cancer (OR 2.4), and lupus (OR 4.4), were also identified. 3 prospective studies estimated the likelihood ratios (LRs) for a positive clinical assessment to be 1.25 to 2.10 and for a negative clinical assessment to be 0.40 to 0.64. A more recent retrospective study using multiple linear regression found overall positive and negative LRs of 1.20 and 0.21, respectively. One study reported on the development, validation, and refinement of a clinical prediction guide. Patients were initially stratified into low, moderate, or high risk groups on the basis of a clinical model (major points in the model included active cancer, bedridden > 3 d or major surgery within 4 wk, localized tenderness of deep venous system in calf or thigh, swelling of calf and thigh, calf swelling > 3 cm compared with asymptomatic leg, and strong family history of DVT). The positive LRs for high-, moderate-, and low-risk strata were {16.2 (95% CI 9.4 to 28.1), 1.4 (CI 1.1 to 1.9), and 0.2 (CI 0.1 to 0.3)}*, respectively. The LRs of compression ultrasonography (US) for the 3 strata are reported in the Table. Patients at high and moderate risk who have abnormal US results should be treated for DVT. DVT is excluded in patients at low risk who have normal US results. Patients with discordant results need further testing (venography or serial US).


A clinical prediction guide that includes patient risk factors, signs, and symptoms reliably predicts the probability of deep venous thrombosis and simplifies the management strategies for patients with suspected deep venous thrombosis when it is used in conjunction with noninvasive diagnostic tests.

Sources of funding: No external funding.

For correspondence: Dr. S.S. Anand, Hamilton General Hospital, 237 Barton Street East, Hamilton, Ontario L8L 2X2, Canada. FAX 905-527-9642.

*Calculated from data in Wells PS, Hirsh J, Anderson DR, et al. Lancet. 1995;345:1326-30.

Table. Likelihood ratios (95% CI) for ultrasonography results by clinical probability

Clinical probability Normal results (CI) Abnormal results (CI)
High 0.06 (0.03 to 0.16) ∞ (3 to ∞)=
Moderate 0.17 (0.07 to 0.34) 72 (13 to 412)
Low 0.20 (0.06 to 0.52) 34 (14 to 76)


Yogi Berra allegedly said, "If you don't know where you're going, you'll end up someplace else." Fortunately, Bayesian, or probabilistic, reasoning can lead clinicians through the diagnostic maze of suspected DVT and avoid the consequences of Berra's dictum. The systematic review by Anand and colleagues not only explains how the history, physical examination, and various objective tests help to diagnose DVT, it also clarifies how the predictive power of isolated, but relatively important, clinical findings can be collected and then linked to test results to yield accurate and robust predictions of disease. Positive or negative US results mean different things for different patients. High- or moderate-risk patients with positive US results should be treated for DVT. Low-risk patients with positive results need further evaluation (i.e., venography). High-risk patients with normal US findings remain at risk and need to be followed (serial testing) or receive definitive testing (venography). The clinical criteria used to establish levels of risk are simple and easy to apply.

Missing from the clinical model is the D-dimer test. In some circumstances, a negative D-dimer test result would have a high negative predictive value and in effect would rule out DVT regardless of other clinical findings. Many diseases that lead to DVT are also associated with fibrin generation (trauma, active cancer), and thus positive D-dimer test results can be nonspecific. Most institutions do not provide a rapid and sensitive D-dimer assay, such as the SimpliRed whole blood agglutination test, and therefore the role of D-dimer testing in a generalizable clinical prediction model awaits further dissemination (1).

Daniel M. Becker, MD
University of Virginia School of MedicineCharlottesville, Virginia, USA


1. Becker DM, Philbrick JT, Bachhuber TL, Humphries JE.D-dimer testing and acute venous thromboembolism. A shortcut to accurate diagnosis? Arch Intern Med. 1996;156:939-46.