A clinical outcome study has shown that it is safe to withhold anticoagulation in symptomatic outpatients in whom the results of simplified compression ultrasonography are normal on initial testing and two repeated tests [17]. More than 500 000 patients are referred for testing each year in the United States [22]; 80% of these patients (400 000) will have a normal result on the first test and will need repeated testing [17]. Cost analysis shows that a single repeated test compared with two repeated tests results in substantial savings-$260 per patient (in 1990 U.S. dollars) [23]. Therefore, more than $100 million could be saved yearly in the United States if the use of two repeated tests were replaced by the use of one repeated test. However, the safety of this approach is uncertain because it has not been evaluated by clinical trials.

Of the patients with normal initial ultrasonography results who then have abnormal results on repeated testing, half have abnormal results the day after presentation and half have abnormal results on day 7 [17]. Patients whose results are abnormal the day after presentation may have detectable thrombosis on the initial test if the popliteal vein is imaged beyond the popliteal fossa to its most distal point (that is, to the trifurcation of the calf veins) because compression ultrasonography is sensitive for thrombi that barely extend out of the calf veins into the popliteal vein [24]. Repeated testing could then be limited to a single test done 5 to 7 days after the first test [16, 23].

We performed a prospective cohort study of outpatients suspected of having first-episode deep venous thrombosis. This was done to test the safety of withholding anticoagulation in patients who 1) have normal results on simplified compression ultrasonography that was done at presentation and that completely imaged the popliteal vein to its most distal point and 2) have a normal result on a single repeated test done 5 to 7 days after the first test. We used long-term follow-up to test the validity of this approach because inadequate management of proximal venous thrombosis results in clinically evident venous thromboembolic events that can be measured objectively [19-21].

Methods

Top Methods Results Discussion Author & Article Info References

Patients and Study Protocol

The study sample consisted of consecutive outpatients who were suspected of having first-episode deep venous thrombosis and were referred by their physicians to the noninvasive vascular laboratory of University Hospital or Veterans Administration Medical Center in Oklahoma City, Oklahoma, between December 1993 and 31 December 1995. Each patient was seen by a consultant physician who obtained a clinical history, performed a physical examination, and evaluated the patient's eligibility for the study. Eligible patients who gave informed consent were then managed according to the study protocol (Figure 1). Patients were ineligible if compression ultrasonography could not be done because of physical or technical limitations, if the patients were unable to return for repeated testing in 5 to 7 days, if long-term follow-up was not possible because of geographic inaccessibility, if the patients had received therapeutic doses of heparin for more than 24 hours before their referral, or if the patients were pregnant. The institutional review board approved the study protocol.

View larger version (31K): [in this window] [in a new window]   Figure 1. Study protocol and outcomes. *Anticoagulation was given unless it was contraindicated or the venogram was normal.

Objective Testing for Venous Thrombosis at Study Entry

Real-time B-mode ultrasonography was performed immediately after the clinical assessment. The simplified compression technique described by Lensing and colleagues [14] was used, with a minor modification as described below. Ultrasonography was performed by using an Acuson 128 scanner (Acuson Corp., Mountain View, California) equipped with a 7.5-MHz linear-array transducer. Both the common femoral and popliteal veins were imaged in gray scale and were assessed for compressibility [14]. The common femoral vein was imaged from the inguinal line to its bifurcation into the superficial femoral vein and profunda femoris. The popliteal vein was imaged from the proximal popliteal fossa to a point 10 cm distal from the mid-patella. This point was chosen to provide a reproducible method with which to approximate the most distal popliteal vein because the calf-vein trifurcation is often difficult to identify by ultrasonography. Vein anatomy in the popliteal fossa, as well as the formation of the popliteal vein itself, greatly varies; the "classic" trifurcation pattern is found in only a minority of patients [25-27].

Compressibility of the veins was assessed only in the transverse plane [14]. The results were categorized as normal if all imaged venous segments were fully compressible, as abnormal if a noncompressible segment was identified, or as inadequate for interpretation.

If the result of initial compression ultrasonography was normal, anticoagulation was withheld and testing was repeated 5 to 7 days later. Anticoagulation was withheld from all patients whose results remained normal on compression ultrasonography (the normal cohort), regardless of their symptoms.

If the result of initial or repeated testing was abnormal (the abnormal cohort), venography was done to confirm the extent of thrombosis. The venographic results were categorized as normal, positive for proximal venous thrombosis (thrombosis of the popliteal, femoral, or iliac vein with or without thrombosis of the calf vein), positive for isolated thrombosis of the calf vein, or inadequate for interpretation. The diagnostic criterion for the presence of deep venous thrombosis was an intraluminal filling defect that was constant on all films [28]. If the venogram was abnormal or inadequate for interpretation, anticoagulation was given unless contraindicated. If the venogram was normal, anticoagulation was not given and the abnormal ultrasonography result was considered falsely abnormal.

Long-Term Follow-up

All patients were instructed to immediately return to our clinic or emergency department if they had symptoms or signs of venous thrombosis or pulmonary embolism. They were also assessed routinely at 3 months either in the clinic or by telephone. At this follow-up assessment, an interval history was taken that addressed general health, specific symptoms (including leg pain, tenderness and swelling, chest pain, dyspnea, hemoptysis, and syncope), hospitalization, and use of anticoagulants. For all patients who died, the cause of death was determined either from autopsy or by independent clinical review.

The primary outcome measure was a diagnosis of venous thrombosis or pulmonary embolism during follow-up confirmed by objective testing. The 3-month follow-up period was chosen because inadequate management of proximal venous thrombosis results in a high rate of recurrent venous thromboembolism during the subsequent 3 months [19-21]. All patients in either cohort who returned during 3-month follow-up with clinically suspected deep venous thrombosis underwent objective testing with impedance plethysmography, which was performed serially according to published protocols [7-929, 30]. Serial impedance plethysmography is highly sensitive and specific for proximal venous thrombosis in symptomatic patients [7-9]. Venography was indicated in patients with abnormal results on impedance plethysmography. If venography could not be done or the results of venography were inadequate, deep venous thrombosis was diagnosed if impedance plethysmography yielded abnormal results in the absence of conditions known to produce false-positive results [7-9, 29-31]. Patients suspected of having pulmonary embolism underwent objective testing with lung scanning and, if indicated, pulmonary angiography, according to published protocols and diagnostic criteria [32-34].

Methodologic Issues and Avoidance of Bias

Care was taken to avoid bias. Selection bias was avoided by entering consecutive patients into the study. Bias during the initial testing period was avoided by defining criteria for normal and abnormal ultrasonography results a priori; by prohibiting venography or other objective leg testing in patients with normal ultrasonography results; and by withholding anticoagulation from all patients with normal ultrasonography results, regardless of their symptoms. Diagnostic suspicion bias [35] was avoided by objectively testing all patients who returned during follow-up with symptoms or signs suggestive of deep venous thrombosis or pulmonary embolism. Incorporation bias [35] was avoided by using serial impedance plethysmography (rather than ultrasonography) as the primary test to evaluate patients with symptoms or signs of deep venous thrombosis during long-term follow-up. A normal ultrasonography result was not used to exclude deep venous thrombosis during follow-up. Interpretation bias [35] was avoided by independently interpreting the objective test results obtained during follow-up, without knowledge of other test results, the patient's condition, or the patient's ultrasonography findings at entry. Results of all objective tests performed during follow-up were reviewed independently by two readers; disagreements were resolved through adjudication by a third independent reader. For each patient who died, the case was reviewed independently by two internists who had not been involved in the patient's care; disagreements were resolved through adjudication by a third independent physician. The cause of death was determined without knowledge of the patient's ultrasonography findings at study entry.

Statistical Analysis

Before the study, we hypothesized from previous studies of noninvasive testing [7-9, 17] that the incidence of venous thromboembolism on follow-up in patients with normal ultrasonography results would be 2% or less. Using this incidence, we estimated that the normal cohort would require approximately 300 patients, based on a 95% CI, to exclude a true incidence of venous thromboembolism on follow-up of 5% or more; this was consistent with the approach used in previous studies [7-9, 17]. We estimated that the prevalence of abnormal ultrasonography results during initial testing would be 20% [17]. Therefore, we evaluated 400 patients to provide 300 or more patients for the normal cohort.

The incidences of venous thromboembolism on follow-up for the normal and abnormal cohorts were compared by using the Fisher exact test. The exact 95% CIs for the true incidences of venous thromboembolism occurring during follow-up were calculated from the binomial distribution.

Role of Funding Sources

Our funding sources had no role in gathering, analyzing, or interpreting the data and had no role in the decision to publish the study.

Results

Top Methods Results Discussion Author & Article Info References

Patients

Of the 428 consecutive patients evaluated, 408 patients (95%) were eligible and 20 were ineligible. The reasons for ineligibility were pregnancy (14 patients), geographic inaccessibility (5 patients), and inability to undergo compression ultrasonography because of a groin wound (1 patient). Of the 408 eligible patients, 405 (99%) gave informed consent to participate. The clinical characteristics of the patients at study entry are shown in the (Table 1).

Objective Testing for Venous Thrombosis at Study Entry

Results of ultrasonography at presentation were normal in 342 (84%) and abnormal in 63 (16%) of the 405 patients. None of the 342 patients with normal ultrasonography results died or had symptomatic pulmonary embolism during the initial 5- to 7-day testing period (95% CI, 0% to 1.07%). Repeated testing was successfully completed in 311 (91%) of the 342 patients and yielded abnormal results in 7 patients; 31 patients did not return for repeated testing. Thus, the normal cohort comprised 335 patients (83%) (Figure 1): 304 patients whose compression ultrasonography results were normal both at presentation and on the repeated test done 5 to 7 days after the first and 31 patients who had normal test results at presentation but did not undergo the repeated test. Anticoagulation was withheld in all 335 patients in the normal cohort.

The abnormal cohort comprised 70 patients (17%): 63 patients whose test results were abnormal at presentation and 7 patients whose test results were normal at presentation but abnormal on the repeated test done 5 to 7 days after the first (Figure 1). Venography was performed in 37 (53%) of the 70 patients in the abnormal cohort. The reasons for not performing venography were a contraindication (5 patients), leg amputation (2 patients), the patient's inability to cooperate (6 patients), the physician's inability to obtain intravenous access (5 patients), the patient's refusal (13 patients), and the referring physician's refusal (2 patients). Venography provided adequate visualization of the proximal and calf veins in 28 patients; 23 of these patients had deep venous thrombosis, and 5 had normal venograms. Anticoagulation was given to 65 of the 70 patients in the abnormal cohort.

Outcome during Long-Term Follow-up

All 405 patients in the study were successfully followed for 3 months. No patient was lost to follow-up. Of the 335 patients in the normal cohort, none died of pulmonary embolism (CI, 0% to 1.09%). One patient (0.3% [CI, 0.01% to 1.65%]) returned at 10 weeks with symptomatic deep venous thrombosis that was then confirmed by objective testing. This patient did not have repeated ultrasonography at 5 to 7 days because the attending physician refused. The patient had a malignant lesion in the pelvis that was unresponsive to radiation or chemotherapy. Impedance plethysmography could not be performed because of the inability to pneumatically occlude the thigh as a result of tumor invasion. The result of compression ultrasonography was abnormal at the groin and in the popliteal vein. Venography was not done. During the eighth week of follow-up, one additional patient returned to the hospital emergency department with dyspnea and pleuritic chest pain. The patient had a history of breast cancer and had received radiation to the chest 2 weeks earlier. A perfusion lung scan documented several small subsegmental defects. No further objective testing for venous thromboembolism was done. The patient returned again during the 10th week of follow-up with the same symptoms. Repeated perfusion lung scanning showed the same results. Further objective testing for venous thromboembolism was not performed, and anticoagulation was not given. This patient completed follow-up and was alive at 6 months with no additional symptomatic presentations. With this patient considered to have pulmonary embolism, the incidence of venous thromboembolism on follow-up was 2 of 335 patients (0.6% [CI, 0.07% to 2.14%]). Thus, the negative predictive value of normal ultrasonography results was 99.4% (CI, 97.87% to 99.96%; 333 of 335 patients).

Of the 70 patients in the abnormal cohort, 4 patients (5.7% [CI, 1.58% to 13.99%]) returned with recurrent symptomatic deep venous thrombosis confirmed by objective testing (the patients returned at 5, 10, 11, and 12 weeks, respectively). All 4 patients had received anticoagulation. The diagnosis was confirmed by impedance plethysmography in 2 patients, by impedance plethysmography and ultrasonography in 1 patient (ultrasonography documented a new noncompressible venous segment), and during surgery in 1 patient. This latter patient presented in the 10th week of follow-up with phlegmasia cerulea dolens and venous gangrene and immediately underwent surgery, during which fresh venous thrombosis was documented.

The difference in the incidence of objectively documented venous thromboembolism during follow-up between the normal cohort (2 of 335 patients [0.6%]) and the abnormal cohort (4 of 70 patients [5.7%]) was statistically significant (P = 0.009).

Fifteen patients died during the study (8 in the normal cohort and 7 in the abnormal cohort). The deaths in the normal cohort were caused by insidious, anticipated causes in 5 patients (metastatic carcinoma in 3, Hodgkin lymphoma in 1, and sepsis in 1); aortic dissection in 1 patient (confirmed by autopsy); sudden cardiac death at 8 weeks in 1 patient (pulmonary embolism was excluded by autopsy); and sudden cardiac death at 7 weeks in 1 patient with recent, recurrent myocardial infarction and recurrent ventricular tachyarrhythmia, chronic severe left ventricular failure (ejection fraction, 20%), and previous mitral valve replacement and coronary bypass surgery. All deaths in the abnormal cohort were insidious and anticipated; death resulted from metastatic carcinoma in 4 patients, sepsis in 2 patients, and AIDS in 1 patient.

Discussion

Top Methods Results Discussion Author & Article Info References

The results indicate that it is safe to withhold anticoagulation in outpatients suspected of having first-episode deep venous thrombosis if compression ultrasonography testing yields normal results on presentation and on a single repeated test done 5 to 7 days after the first. The outcome for patients in our normal cohort (no deaths from pulmonary embolism and venous thromboembolism on follow-up in only 0.6% of patients) is similar to the outcome for patients with normal venograms [36] and to the outcomes seen with other noninvasive approaches [6-9, 17]. Patients in whom treatment was withheld on the basis of normal results on impedance plethysmography (with two to four repeated tests) have developed venous thromboembolism during follow-up at rates of 0.3% to 2.5% [7-9, 17]. The outcome in our study is also similar to the results of a previous study of compression ultrasonography that used two repeated tests [17]. The difference between our study and this previous study is that one repeated test improves patient convenience and reduces cost [23] without detracting from patient safety.

A limitation of our study is that too few patients had venography to allow a precise estimate of the positive predictive value of an abnormal result on compression ultrasonography. We are continuing to perform venography on patients with abnormal ultrasonography results in order to assess the positive predictive value of compression ultrasonography in our population. This limitation does not, however, detract from the primary focus of our study: the safety of withholding anticoagulation in patients with normal ultrasonography results.

Compression ultrasonography as used in our study is easy to perform, and the results are easy to interpret. Testing requires only basic ultrasonography equipment with gray-scale imaging capability [14, 16]. A complete examination of both legs takes about 15 minutes [14, 16, 37]. Defining the popliteal vein territory in terms of external body landmarks provides a standardized approach for this anatomically variable area. The use of vein compressibility as the single diagnostic criterion has high interobserver agreement [14]. The addition of Doppler imaging (duplex or color-flow) increases the cost of the equipment and the complexity of the test [13, 16]. The low rate of venous thromboembolism on follow-up in patients with normal results on the simplified compression approach indicates that the addition of Doppler or color-flow imaging is unlikely to produce a clinically important improvement in outcome for these patients.

In 10% of our patients with abnormal ultrasonography results (7 of 70 patients), the abnormality was detected on the repeated test. Although the rate of conversion from normal to abnormal results was low (7 of 311 patients [2%]), repeated testing is indicated because proximal venous thrombosis can be fatal and because ultrasonography is a noninvasive test that effectively identifies patients with proximal venous thrombosis.

Our results directly affect clinical practice because they provide a basis for standardizing the approach to ultrasonography for suspected deep venous thrombosis. No standard approach is available in the United States [38]. Our results support standardizing ultrasonography for deep venous thrombosis by using the simplified compression approach with imaging of the popliteal vein completely to its most distal point and a single repeated test done 5 to 7 days after the first. This approach provides useful information for clinicians by separating patients into two distinct groups: 1) those with abnormal results, in whom either treatment or venography is indicated, and 2) those with normal results, in whom further testing and anticoagulation can be safety avoided.

Mr. Raskob: Department of Biostatistics and Epidemiology, University of Oklahoma Health Sciences Center, PO Box 26901, Oklahoma City, OK 73190.

Drs. Whitsett, Durica, and George: Department of Medicine, University of Oklahoma Health Sciences Center, PO Box 26901, Oklahoma City, OK 73190.

Dr. Comp: University Hospital, EB 400, PO Box 26307, Oklahoma City, OK 73126.

Dr. Tytle: Department of Radiological Sciences, University of Oklahoma Health Sciences Center, PO Box 26901, Oklahoma City, OK 73190.

Dr. McKee: W.K. Warren Medical Research Institute, University of Oklahoma Health Sciences Center, PO Box 26901, Oklahoma City, OK 73190.