DVT—deep-vein thrombosis

IPG—impedance plethysmography

Objective tests are used to diagnose deep-vein thrombosis (DVT) because its signs and symptoms are nonspecific and may be mimicked by various nonthrombotic disorders [1, 2]. Although venography is the reference standard for the diagnosis of DVT, its use has been limited because of its invasive nature and associated side effects [3-5]. For this reason there has been considerable interest in developing noninvasive diagnostic tests for DVT to replace venography [6]. Of these, occlusive cuff impedance plethysmography (IPG) was the first to be evaluated thoroughly and became one of the most widely adopted methods. Studies in the mid 1970s and early 1980s compared IPG with venography in patients with suspected DVT and reported that the sensitivity and specificity of IPG for proximal-vein thrombosis was approximately 95% [7-11]. Although IPG was found to be insensitive to DVT isolated to the calf veins, this was not considered to be an important shortcoming for three reasons: First, calf-vein thrombi do not appear to be dangerous provided that they remain confined to the calf [8, 12]; second, isolated calf-vein thrombi are uncommon in patients with symptomatic DVT [13]; and third, it was hypothesized that any calf-vein thrombi extending into the proximal veins could be readily detected by repeating the IPG test over a 7- to 10-day period.

Several studies subsequently demonstrated that 14% to 26% of symptomatic patients with proven proximal-vein thrombosis had a normal IPG result at presentation but developed an abnormal test during the serial IPG testing [14-16]. These patients were presumed to have had isolated calf-vein thrombosis at the time of their initial IPG test, which was detected after it extended to the proximal veins. These studies also showed that it was relatively safe to withhold anticoagulant treatment from patients with serial normal IPG results because only about 2% of these patients developed symptomatic venous thromboembolic complications during long-term follow-up. Consequently, serial IPG testing became the method of choice for the diagnostic management of patients with clinically suspected DVT in many centers in the world.

A recent report called into question the safety of withholding anticoagulants from patients with normal IPG results [17]. Prandoni and associates performed serial IPG testing in 311 patients with clinically suspected DVT whose initial IPG result was normal. Four (1.3%) of these patients subsequently developed fatal pulmonary emboli despite having normal IPG tests. At our thrombosis center, serial IPG testing has been used for many years to manage patients with clinically suspected DVT. Recently, we became increasingly aware that IPG was failing to identify an unexpectedly high number of symptomatic patients with proximal-vein thrombosis. We therefore decided to re-evaluate the accuracy of IPG for the detection of proximal-vein thrombosis in symptomatic outpatients.

Methods

Top Methods Results Discussion Author & Article Info References

On 1 January 1990, a comprehensive computerized data base was created that prospectively recorded pertinent information on all patients evaluated at the Thrombosis Clinic of the Henderson General Hospital in Hamilton, Ontario. This data base contained information about patients' presenting symptoms, other relevant clinical and demographic data, and results of objective tests. These data were analyzed retrospectively (prompted by our impression that IPG was failing to detect some proximal-vein thrombi) in November 1991. During this 22-month period, consecutive patients with clinically suspected DVT were evaluated using a standardized protocol. All patients were assessed by a physician and underwent IPG. (Patients with previously documented DVT or pulmonary embolism were excluded from this analysis.) If the IPG result was abnormal, patients were referred for venography to confirm the presence of DVT. Patients for whom a venogram could not be obtained or whose venogram was inadequate for interpretation had compression ultrasound. Patients with a normal IPG result were managed in one of two ways: Those in whom the clinical suspicion for DVT was low had two additional IPG tests over a 7-day period, whereas those in whom the clinical suspicion of venous thrombosis was high were referred for venography. The distinction between patients at high and low clinical suspicion for DVT was made at the discretion of the individual clinicians. In general, patients were considered at high risk if their symptoms were consistent with DVT and they had one or more recognized risk factors for venous thrombosis (such as cancer or recent immobilization) and no alternative explanation for their symptoms. Patients who were unable to return to the clinic for serial IPG testing were also evaluated with venography.

Two IPG instruments were used in this study (IPG 200, Codman and Shurtleff Inc., Randolph, Massachusetts, and IPG 800, Electrodiagnostic Instruments Inc., Burbank California). Although the IPG 200 is no longer in production, it was the model used in most of the early studies evaluating IPG. The IPG 800 is the updated version of the IPG 200. Both machines use identical means to measure impedance. Impedance plethysmography was done using the occlusive cuff technique as described previously [7]. Briefly, the patient was placed in the supine position with the examined leg flexed at the knee, externally rotated at the hip and elevated 20 to 30 degrees. A pneumatic cuff was applied around the thigh and circumferential electrodes were placed around the calf. The cuff was inflated to a pressure that produced venous occlusion (45 cm H_2O for the IPG 200 and 68 cm H_2O for the IPG 800) for periods of 45 seconds and 2 minutes and rapidly deflated. The test was done on a maximum of five occasions and the rise in impedance occurring during cuff inflation was plotted against the fall on a nomogram. The stop line is a line on the IPG nomogram that is above and parallel to the discriminate line. If a result fell above the stop line, IPG was discontinued. In a previous study, it was shown that the risk for proximal-vein thrombosis was very low if any point on the five-test sequence fell above this line [18]. If, after the five-test sequence was completed, none of the tests was above the stop line and the test with the largest rise did not give the largest fall, then the test was considered inadequate and further tests were done. The interpretation of the IPG result was done as described previously. The test was read as normal if: 1) the test with the highest rise and fall was above the discriminant line; or 2) a test in the sequence fell above the stop line.

Contrast venography was done with the patient tilted in the semi-upright position and the examined leg was nonweight bearing. Approximately 60 to 120 mL of nonionic contrast (iodine, 300 mg/mL) was injected into a dorsal foot vein. Spot films of the calf, knee, thigh, and pelvis were obtained after maximal filling with contrast material as determined by fluoroscopy. Deep-vein thrombosis was diagnosed by the presence of a constant intraluminal filling defect present in at least two projections or by nonfilling of a vein or venous segment despite repeated injections with contrast material. A thrombus was regarded as nonocclusive if contrast material was seen between the vessel wall and the thrombus along its entire course. Deep-vein thrombosis was excluded if the peroneal, posterior tibial, popliteal, superficial femoral, common femoral, and iliac veins were adequately visualized, and no filling defects were observed. Visualization of the anterior tibial veins was not required for a venogram to be considered adequate for interpretation. Venograms were considered inadequate for interpretation if opacification of the deep veins was insufficient. Proximal-vein thrombi were defined as those involving the popliteal or more proximal leg veins. The extent of proximal-vein thrombosis was categorized according to the number of vein segments involved. On a scale from 1 to 4, 1 point was given for involvement of each of the popliteal, superficial femoral, common femoral, or iliac vein segments by thrombosis. Compression ultrasound was done with a high-resolution duplex scanner (Acuson 128, Acuson Corporation, Mountain View, California) equipped with electronically focused linear array transducers (5 to 7.5 Mhz) [19]. The deep venous system of the thigh (including the entire popliteal vein) was evaluated for compressibility. Deep-vein thrombosis was diagnosed if a vein or a venous segment was not fully compressible. In patients whose DVT was diagnosed by compression ultrasound, a thrombus was determined to be occlusive if venous flow was absent by Doppler ultrasound after augmentation.

Venography (or compression ultrasonography) was done in all patients with abnormal IPG results (to assess the positive predictive value) and in the subset of patients with a normal IPG result who had either a high clinical suspicion for venous thrombosis or in whom follow-up testing was inconvenient (to assess the sensitivity). Sensitivity of IPG was calculated only for proximal-vein thrombosis. The 95% confidence interval (CI) for sensitivity was determined using the binomial distribution. The size of venous thrombi in patients with false-negative IPG findings was compared with that in patients with true-positive IPG tests. To assess whether one of the IPG machines contributed disproportionately to the observed results, we calculated the sensitivity rate for each of the IPG machines separately. The Student t-test and the chi-square test were used when appropriate.

Results

Top Methods Results Discussion Author & Article Info References

From January 1990 to October 1991, 413 consecutive patients were evaluated for suspected symptomatic DVT. Twenty-three patients with previously documented venous thromboembolic disease were excluded from the study. The remaining 390 eligible patients were interviewed and examined by a physician and underwent IPG testing. Adequate IPG tests were obtained in 386 patients. The four patients with technically inadequate IPG tests were excluded from the study cohort. Therefore, a total of 386 patients were eligible for this analysis. The mean age of the cohort was 60.7 years; 161 (41.7%) were men and 225 (58.3%), women.

The initial IPG result was abnormal in 53 (13.7%) of the patients and normal in the remaining 333 patients (Figure 1). Patients with normal IPG results were subdivided into two groups based on the clinical suspicion of DVT. One group of 73 patients was referred for venography despite the normal IPG result because of a high clinical suspicion of DVT or because serial IPG testing could not be done. The other 260 patients had a low or moderate clinical suspicion of venous thrombosis and underwent serial IPG testing over a 7-day period.

View larger version (17K): [in this window] [in a new window]   Figure 1. Flow chart of the diagnostic workup of patients eligible for study. The asterisks indicate two patients (one in the impedance plethysmography [IPG] abnormal group and one in the IPG normal group) who had inadequate venograms and were not evaluated with compression ultrasound. These two patients were excluded from the analysis.

Of those patients undergoing serial IPG testing, five developed an abnormal IPG result and were referred for confirmatory venography. Another 13 patients were referred for venography because of persistent symptoms and no obvious alternative diagnosis. No further tests were done in the remaining 242 patients with serial normal IPG tests.

Of the 58 patients with abnormal IPG results, confirmatory tests showed proximal-venous thrombosis in only 37 patients (by venography in 30 patients and by compression ultrasound in 7 patients). One patient with a positive IPG had an inadequate venogram and did not undergo compression ultrasound. This patient was excluded from the analysis. Of the 20 patients with falsely abnormal IPG results, proximal-vein thrombosis was excluded by venography in 18 patients and by compression ultrasound in 2 patients.

Patients with Normal Impedance Plethysmography Who Were Referred for Venography

Seventy-three patients with a normal IPG result on the day of referral, but who had clinical signs and symptoms highly suspicious of DVT or who could not be followed by serial IPG, were investigated with venography (or compression ultrasound). Proximal-vein thrombosis was diagnosed in 12 (16%) patients (by venography in 10 and by compression ultrasound in 2). In addition, two patients had thrombi confined to the calf veins diagnosed by venography. Venous thrombosis was excluded in 58 patients. One patient had an inadequate venogram and did not undergo compression ultrasound; she was excluded from the analysis.

Venography was done in 13 (5%) of the 255 patients with normal serial IPG tests because symptoms persisted and no alternative diagnosis was evident. Proximal-vein thrombosis was found in seven of these patients and venous thrombosis was excluded in the remaining six. In total, 19 (22%) of the 85 analyzed patients in whom there remained a high clinical suspicion for venous thrombosis despite normal IPG studies had proximal-vein thrombosis.

Sensitivity and Positive Predictive Value of Impedance Plethysmography for Proximal-Vein Thrombosis

Of the 384 patients analyzed in this cohort, 142 were referred for venography. The reasons for referral for venography are listed in Table 1. Fifty-six patients were diagnosed as having proximal-vein thrombosis (by venography in 47 and by compression ultrasound in 9) and 2 were found to have DVT confined to the calf veins. The two patients with isolated calf-vein thrombosis, both of whom had normal IPG, were not included in the analysis of sensitivity of IPG. Thirty-seven of the 56 patients with proximal-vein thrombosis were correctly identified by IPG, whereas the test failed to detect proximal DVT in the remaining 19 patients. Thus, the sensitivity of IPG for proximal-vein thrombosis was 66% (95% CI, 52% to 78%) (Figure 2). Impedance plethysmography was performed using the IPG 200 in 268 patients and with the IPG 800 in 116 patients. The sensitivity for proximal-vein thrombosis did not differ between the IPG 200 and the IPG 800 machines (63% and 71%, respectively; P > 0.2). In 14 (74%) of the 19 patients in whom IPG failed to detect proximal-vein thrombosis, the IPG result was above the stop line, whereas in the other five, the point with the highest rise and fall in the five-test sequence fell between the discriminate line and stop line.

View larger version (12K): [in this window] [in a new window]   Figure 2. Impedance plethysmographic results in analyzed patients confirmed to have proximal-vein thrombosis. Sensitivity, 37/56; 66% (95% CI, 52% to 78%); positive predictive value, 37/57; 65%. * Seven patients were diagnosed with proximal-vein thrombosis by compression ultrasound. ** Two patients had proximal deep-vein thrombosis excluded with compression ultrasound. *** Two patients had proximal-vein thrombosis diagnosed by compression ultrasound. **** 242 patients did not undergo venography (or compression ultrasound). Thus, a true figure for the total number of patients in whom proximal-vein thrombosis was absent cannot be determined. IPG = impedance plethysmography.

Fifty-seven of the 384 analyzed patients had abnormal IPG results. Confirmatory tests demonstrated proximal-vein thrombosis in 37 patients for a positive predictive value of 65% (Figure 2). Two hundred forty-two patients did not undergo confirmatory testing after serial IPG (see Figure 1). We therefore cannot accurately determine specificity of IPG or prevalence of DVT in this cohort of patients.

Occlusiveness and Extent of Proximal-Vein Thrombosis

Two of the 37 (5%) patients with proximal-vein thrombosis and an abnormal IPG result had nonocclusive thrombosis (Table 2). In contrast, nonocclusive thrombosis was observed in 7 (37%) of the 19 patients with proximal-vein thrombosis and normal IPG findings (P < 0.01). The extent of proximal-vein thrombi was less in patients with false-negative IPG results (P < 0.01). Nevertheless, most of the 19 proximal thrombi missed by IPG were occlusive and large; eight involved the popliteal vein only; six involved the popliteal and superficial femoral veins; three extended from the popliteal to the common femoral veins; and two involved the entire proximal venous system.

Clinical Characteristics of Patients with Deep-Vein Thrombosis

Baseline characteristics of patients with venous thrombosis are listed in Table 3. Patients with abnormal IPG findings tended to be older than those with normal IPG tests. No statistical differences were observed between gender, duration of symptoms, presence of cancer, or recent surgery between patients with normal IPG and those with abnormal IPG who were confirmed to have DVT.

Discussion

Top Methods Results Discussion Author & Article Info References

We conducted this study because of an increasing concern that IPG was failing to detect proximal-vein thrombi in symptomatic patients at our center. The results of our study confirmed that the sensitivity of IPG for proximal-vein thrombosis was too low for us to continue to use it as the sole diagnostic test for patients with suspected DVT. Although venography was only done in the subgroup of patients with a normal IPG result in whom the clinical suspicion for DVT was high or who were inaccessible for serial testing, this method of selection could not have produced a falsely low sensitivity. Performing venography in any or all of the remaining 242 patients with a normal IPG test could only lower the sensitivity of IPG by detecting proximal-vein thrombosis in additional patients.

The observed sensitivity of IPG of less than 70% for proximal-vein thrombosis in symptomatic patients is in distinct contrast to a sensitivity of approximately 95% reported in earlier studies from our institution and other centers [7-11]. We could not detect any bias to account for the observed low sensitivity. Although our analysis was done retrospectively, it was performed on a cohort of consecutive symptomatic patients in whom complete clinical and demographic data were collected prospectively over 2 years. In addition, all objective tests were interpreted by experts who were unaware of the results of any other tests or of the patients' clinical status. Seventeen (89%) of the 19 patients with normal IPG and proximal-vein thrombosis had the diagnosis made by venography, and 2 patients with inadequate venograms were diagnosed by compression ultrasound, a test with a sensitivity and specificity for symptomatic proximal-vein thrombosis of over 95%.

Other factors in our study that could have reduced the sensitivity of IPG for proximal-vein thrombosis seem unlikely. First, only technologists with extensive experience performed the IPG tests, which were always executed and interpreted according to the standardized procedure. We also carefully reviewed the IPG technique, the testing protocol, and the IPG tracings but could not find an explanation for the observed low sensitivity. Second, patient characteristics (including age, sex, period elapsed between the onset of symptoms and the day of referral, presence of malignancy, and postoperative status) were comparable between those patients whose proximal-vein thrombi were correctly identified by IPG and those whose proximal-vein thrombi were missed by IPG. In addition, these patient characteristics were similar to those described in previous studies of symptomatic patients [8, 14]. Third, equipment-related factors seem an improbable explanation for the poor sensitivity because similar low sensitivities (that is, 63% and 71%) were observed with two different IPG models. Nevertheless, most of the previous studies reporting higher sensitivities were performed using the older model (that is, the IPG 200 that is now out of production), and a deterioration in the reliability of our machine over time cannot be excluded. The new IPG model (IPG 800), which was equally insensitive, has not previously been evaluated formally in symptomatic outpatients. Finally, it is possible that the low sensitivity observed in our study was due, at least in part, to chance; however, the upper 95% confidence interval value for IPG sensitivity was only 78%, a disturbingly low figure.

Although our findings contrast with some of the earlier studies, they are not inconsistent with the results of four recent large studies in symptomatic patients that reported that 14% to 26% of all patients with confirmed proximal-vein thrombosis developed their abnormal IPG result during serial testing [14-17]. If venography had been done in all these patients on the day of presentation, the observed sensitivity of IPG could have been as low as 74% to 86%. Although some of the thrombi missed by the initial IPG test might have been isolated to the calf veins, two lines of evidence make it more likely that most of the thrombi had already involved the proximal veins at the time of initial presentation. First, recent studies in symptomatic patients have shown that isolated calf-vein thrombi account for only 10% to 20% of all thrombi, with the remainder being proximal-vein thrombi [13]. Data from cohorts of patients with untreated or inadequately treated calf-vein thrombosis suggest that only 20% to 30% of these subsequently extend into the proximal veins [12, 20]. On the basis of these findings, we would anticipate that, at most, 6% of the thrombi missed by the initial IPG test would have been isolated to the calf veins and that the remaining 8% to 20% would involve the proximal venous system. Second, recent studies using serial compression ultrasonography (of the proximal leg veins) to diagnose DVT in symptomatic patients have reported a frequency of conversion from a normal to abnormal test of 0% to 7%, which is much lower than that seen with serial IPG [21, 22]. Recent studies have also shown that IPG is very insensitive to proximal-vein thrombosis in asymptomatic patients after hip surgery (reported sensitivity, 20% to 40%) because most of these thrombi are nonocclusive [6, 23-25]. Compression ultrasound is much more sensitive than IPG for the diagnosis of proximal-vein thrombosis developing in asymptomatic postoperative patients [23]. Therefore, it is likely that the lower rate of conversion from a normal to an abnormal test during serial testing with ultrasound than IPG occurs, in part, because ultrasound detects a higher proportion of nonocclusive proximal-vein thrombi.

Nonocclusive thrombi occur less commonly in symptomatic than in asymptomatic patients. We found that 9 of the 56 proximal-vein thrombi in our study of symptomatic patients were nonocclusive and that IPG detected only 2 (22%) of these thrombi. Also, the thrombi missed by IPG were less extensive than those it correctly identified. Most of the proximal thrombi that IPG failed to detect, however, were large (involving at least the superficial femoral vein) and occlusive (see Table 2). These findings suggest that IPG is failing to detect proximal-vein thrombosis that can be diagnosed by compression ultrasonography in symptomatic patients [19].

Our observations should be interpreted with caution because they differ considerably from the findings of several well-designed studies. Furthermore, studies from different centers have demonstrated the relative safety of withholding treatment in patients with suspected DVT provided that their IPG remains normal on serial testing. It is likely that some of the normal IPG findings in the symptomatic patients with proximal-vein thrombosis in our study would have become abnormal if serial testing had been done and that many of these patients would not have developed clinically important complications had they been left untreated. The report of fatal pulmonary emboli by Prandoni and associates [17], however, in patients with a negative IPG using equipment with a reported sensitivity for proximal-vein thrombosis of approximately 86% is troublesome for physicians who rely on a negative IPG to exclude clinically important DVT.

In view of our findings and those of Prandoni and associates, we no longer rely on the safety of a normal IPG result in symptomatic patients and, like many centers, use compression ultrasound as our initial diagnostic test for DVT. For those centers that still use IPG testing, it would be prudent to check the sensitivity of their equipment by performing venography in a cohort of patients with normal IPG results.