- Email: [email protected]
LGM Vena Cava Filter: Objective Evaluation of Early Results
Venous Thromboembolic Disease
LGM Vena Cava Filter: Objective Evaluation of Early Results1 Timothy P. Murphy, M D Gary S. Dorfman, M D Joseph W . Yedlicka, M D Timothy C. McCowan, M D Robert L. Vogelzang, M D David W . Hunter, M D Danna K. Carver, RN Robert Pinsk, M D Wilfrido Castaneda-Zuniga, M D Ernest J. Ferris, M D Kurt Amplatz, M D
Index terms: Embolism, pulmonary, 60.72 Venae cavae, filters, 982.1299 Venae cavae, thrombosis, 982.442 JVIR 1991; 2107-115
Abbreviations: IVC = inferior vena cava, LGM = LG-Medical, VIQ = ventilation1 perfusion.
One hundred one LG-Medical (LGM) vena cava filters were placed in 97 patients a t four institutions. Placement was a complete technical success in 90% (91 of 101). In 6% of attempts, LGM filter insertion was complicated by incomplete opening of the filter. Pulmonary embolism after filter placement was not definitely demonstrated in any patient. The probability of inferior vena cava patency was 92% a t 6 months after filter insertion. Thrombosis a t the insertion site was seen in eight of 35 patients (23%) evaluated with duplex ultrasound or venography. Thrombus was observed in 37% of filters a t follow-up examination, with cephalic extension of thrombus above the filter in 20% of all patients examined. Filter migration (greater than l cm) was seen in 12%;significant angulation was observed in only one patient (2%). In vitro experimentation demonstrated that incomplete opening of the LGM filter during placement can be avoided, in part, by brisk retraction of the insertion cannula. The low-profile introducer system of the LGM filter allows increased alternatives in selecting the site for filter insertion. The low-profile system also makes outpatient filter placement a possibility. No significant difference in the prevalence of thrombosis a t the insertion site following LGM filter insertion was noted compared with previous results reported for percutaneous transfemoral placement of the Greenfield filter. The nonopaque sheath does not permit careful localization prior to filter deposition. Modification of the LGM filter to include a radiopaque sheath is suggested. T H E LG-Medical
' From the Department of Diagnostic Imaging, Rhode Island Hospital, 593 Eddy St, Providence, RI 02903 (T.P.M., G.S.D.); the Department of Radiology, University of Minnesota Medical School, Minneapolis, Minn (J.W.Y., D.W.H., W.C.Z., K.A.); the Department of Radiology, University of Arkansas for Medical Sciences, Little Rock, Ark (T.C.M., D.K.C., E.J.F.); and the Department of Radiology, Northwestern Memorial Hospital, Chicago (R.L.V., R.P.). From the 1990 RSNA scientific assembly. Received July 31, 1990; revision requested September 10; revision received November 2, accepted November 5. Address reprint requests to G.S.D. O SCVIR, 1991
(LGM) filter (Vena Tech, ~vanstbn,111) is an inferior vena cava (IVC) filter with recent Food and Drug Administration approval for use in the United States. This device, used for the prevention of pulmonary emboli, is introduced with a 12-F outer-diameter system, thereby necessitating a smaller venotomy site and tract than those needed with the stainless steel Greenfield filter (Medi-tech/Boston Scientific, Watertown, Mass), which is inserted through a 24-F sheath (29.5-F outer diameter). Differences between the pyramidal LGM filter and the Greenfield filter include (a) the presence of six mural side rails in the LGM filter that are attached to six radiating diagonal struts and are arranged to look like a cone, analogous to the shape of the Greenfield filter; (b) the metal
struts in the LGM device are broad and flat; and (c) tines or barbs are located on the entire length of the mural side rails of the LGM filter as opposed to the hooks that are limited to the base of the Greenfield filter (Fig 1). The purported advantages of the LGM filter over the stainless steel Greenfield filter are ease of insertion and decreased prevalence of thrombosis of the delivery site vein. These advantages are believed to be obtained with filtration capability and caval patency comparable to those of the standard of reference, the stainless steel Greenfield filter. Herein we describe a series of 97 patients in whom 101 LGM filters were placed. Prospective and objective imaging studies were performed to assess for (a) pulmonary embolism following fil-
108
Journal of Vascular and Interventional Radiology
February 1991
ter insertion, ( b ) patency of the IVC, (c) the presence of nonocclusive thrombus in and extending above the filter, ( d ) insertion site thrombosis following placement, and ( e ) changes in filter position and morphology (angulation and closure). We assessed whether the purported advantages of the LGM filter could be proved in clinical practice and whether the rates of pulmonary embolism and caval patency after filter insertion are comparable to those obtained with the Greenfield filter to justify continued use of the LGM filter. We also performed an in vitro experiment to investigate the influence of the speed of filter release on the occurrence of incomplete filter opening.
PATIENTS AND METHODS The study population included 49 female and 48 male patients with an average age of 61.8 years (range, 1694 years; standard deviation, 15.3 years). Indications for filter insertion included contraindications to anticoagulation (n = 75), progression of deep venous thrombosis or recurrent pulmonary embolism while receiving anticoagulants (n = 12), complications of anticoagulation (n = 6) (including gastrointestinal hemorrhage, vaginal bleeding, and soft-tissue hematoma), and prophylaxis for extensive free-floating or IVC thrombus (n = 4). All patients had documented venous thromboembolic disease. Seventy-two patients had deep venous thrombosis, and 25 patients had pulmonary embolism. Filters were placed by means of the right common femoral vein (n = 65), the right internal jugular vein (n = la), the left common femoral vein (n = 9), the right external jugular vein (n = 4), and the left internal jugular vein (n = 1). The filter was intentionally placed in a suprarenal location in three patients; 94 underwent infrarenal filter placement. Three patients underwent placement of two LGM filters each for complications encountered during insertion. One additional patient had an inferior vena cava diameter of 35
Figure 1. LGM IVC filter. (a) Side view. (b) View from above. Design differs from that of the Greenfield filter by the presence of side rails (arrows),which are in contact with the wall of the IVC along their entire length when properly positioned. Barbs on the outer aspect of the side rails are meant to minimize movement of the filter within the vena cava. Arrowheads indicate diagonal struts. mm (maximum diameter suitable for LGM filter placement is 28 mm), and bilateral common iliac filters were placed. These patients underwent placement of LGM filters at four institutions (Rhode Island Hospital, Providence; University of Minnesota Medical School, Minneapolis; University of Arkansas for Medical Sciences, Little Rock; and Northwestern Memorial Hospital, Chicago) between April 4, 1989, and February 21,1990. All LGM filters inserted at each institution were included in this study. Most filters placed during this time were LGM filters. Factors that may have negatively influenced operator decision to select an LGM filter include lack of familiarity with this filter and, in some cases, reluctance to place this filter in young patients due to lack of long-term experience with this filter. Evaluation of patients in whom these filters were placed was initiated separately at each institution. Collaboration between institutions began during the follow-up period. Therefore, follow-up methods and intervals were variable. Some patients refused to undergo certain examinations, most notably inferior vena cavography, but other studies-transabdominal ultrasound (US) for example-were acceptable to patients, resulting in variability in follow-up methods. However, results obtained with use of different methods that were applica-
ble to a single criterion were combined. For example, results of inferior vena cavography, transabdominal US, computed tomography (CT) with contrast material enhancement, and intravascular US were used to assess patency of the inferior vena cava. Follow-up studies were obtained up to 10 months after placement. Most radiologic studies were obtained prospectively. The technique of LGM filter insertion has been previously described (1). Patency of the IVC following LGM filter insertion was evaluated by means of cavography, transabdominal duplex US and intravascular US, CT, and autopsy (2). The presence of nonocclusive clot in the filter was ascertained with cavography, endovascular US, CT, and autopsy. Transabdominal US was not used to evaluate for nonocclusive thrombus trapped in the filter as there is no evidence to support the sensitivity of this method in the detection of a small thrombus in the vena cava. Pulmonary embolism following filter placement was assessed with ventilationlperfusion (VIQ) scanning, autopsy, and clinical follow-up. In those patients who died during the followup period, efforts were made to determine the possibility of death due to pulmonary embolism. VIQ scans were compared with prior studies when available. Insertion site thrombosis was as-
Murphy et a1
109
Volume 2 Number 1
sessed with lower extremity duplex US, venography, autopsy, and clinical follow-up. Again, comparisons were made with previous studies and clinical history when available. Filter migration, angulation, and closure were evaluated with plain radiography, and the resultant images were compared with radiographs obtained immediately after filter insertion. Filter migration greater than 1 cm was considered significant. Migration less than 1cm was not deemed significant because of the inability to confidently distinguish this degree of movement from perceived change in position due to breathing or parallax error. Filter angulation was considered significant if the measured tilt was 15' or more off the long axis of the IVC, which presumes similarity to the Greenfield filter in this regard (3). Crowding of the filter struts was established if a decrease in the diameter of the filter base greater than 7 mm had occurred since filter placement (4).
Results for each criterion evaluated were analyzed with use of objective imaging findings alone as well as imaging and clinical follow-up combined. Probability of maintaining patency of the IVC over time was projected by using the life-table estimate method of Kaplan and Meier. In vitro evaluation of filter deposition mechanics was performed as follows: Two femoral and two jugular delivery systems were discharged into an empty plastic tube with a 28-mm diameter. Since only opening characteristics were evaluated, it was considered unnecessary to use a fluidfilled tube. The systems were discharged and reloaded for a total of 10 depositions each. The speed of retraction of the insertion sheath was alternated on each delivery between fast (as fast as possible; delivery time, much less than 1sec) and slow (approximately 0.5 cmlsec; total delivery time, approximately 9 sec). The initial delivery for one femoral and one jugular system was fast, while the slow rate was employed in the other two devices. Forty deliveries were performed with the four systems. The
devices used were provided by the manufacturer: all deliveries and reloadings were performed by a single investigator (G.S.D.). Filters were reloaded by hand into the introducing syringe and then discharged from the syringe into the introducing sheath, similar to the method used during manufacture of the filter. The occurrence of complete and incomplete filter opening was tabulated for each device. The significance of difference in the prevalence of incomplete opening based on rate of delivery and type of delivery system was assessed with x2analysis.
RESULTS Filter Insertion: Immediate Technical Success and Complications Complete technical success was achieved in 91 (90%)of 101 of attempts at LGM filter placement. Incomplete opening of the caudal struts was noted in six patients; a second filter was placed above the first in four patients. (Two were LGM and two were Greenfield filters.) In two patients, incomplete opening was attributed to placement of the filter base in the cephalic portion of an IVC thrombus. Two patients with incomplete opening of the filter on insertion did not undergo placement of additional filters due to the operator's belief that while filter opening was suboptimal, it was acceptable and the filter would probably function effectively. Further complications did not occur in these patients. Five of six filters that did not open completely were introduced via the jugular route (four via internal and one via external jugular vein). Three mechanisms of incomplete opening were postulated. These include nonopening due to placement of the base of the filter in the proximal extent of IVC thrombus, hooking of the diagonal strut on the barbs of the side rail, and uneven contact of the side rails with the caval wall with only the proximal end of the side rails touching the wall. The last mecha-
nism is believed to occur when the filter is delivered by means of the jugular approach. Retraction of the sheath after the side rails have exited the sheath but before the apex of the filter has exited causes the cephalic ends of the side rails to become embedded within the wall of the IVC, thereby preventing full expansion of the filter cone. We have no pathologic support for this last mechanism. One patient had a second LGM filter placed due to angulation of the filter on deposition of 20" off the long axis of the IVC. Two filters changed position immediately following placement in the IVC. Of these two, one patient had 1-cm cephalic migration and increased angulation after filter deposition. In the second patient, the filter was placed in a suprarenal location. The filter then migrated 1.5 cm caudad to the level of the renal veins. Finally, resistance to passage of the delivery sheath at the level of the confluence of the iliac veins was encountered in one patient when insertion was performed from the left common femoral vein. The filter was, however, placed successfully with this approach. There were no difficulties in eight other patients in whom filters were introduced through the left femoral vein. Other problems that occurred during filter insertion included inadvertent filter deposition at the iliac confluence in one patient and placement of a filter with a jugular system via the femoral approach due to operator error, resulting in reversed cephalocaudal filter orientation in another. This patient did not have prior pulmonary embolism and had limited lower extremity deep vein thrombosis. The operator believed that placement of an additional filter was not necessary. This inverted filter might function similar to the Amplatz filter, which is an effective filtration device, although it is associated with an increased rate of caval occlusion. Complications of LGM filter placement were encountered in four (4%) of 101 procedures. These included one periprocedural death. The patient was a 60-year-old woman who
110
Journal of Vascular and Interventional Radiology
February 1991
had undergone a total hysterectomy for uterine prolapse 1month previously. She was admitted with shortness of breath and suspected pulmonary embolism, as indicated by means of a high-probability V/Q scan. Filter placement was requested after the development of heparin-induced platelet aggregation syndrome, with the platelet count dropping to 23,000lmL (23 X 106/L).The diameter of the IVC was 35 mm; therefore, bilateral common iliac filters were placed. The patient became acutely dyspneic, bradycardiac, and hypotensive the following Resuscitative efforts were Autopsy showed bilatera1 iliac vein thrombosis and thromboembO1usin the right pulmonary artery. I t is uncertain whether pulmonary embolism occurred immediately prior to placement during the diagnostic Or after placement (Fig 2). Other complications of LGM placement were transient acute renal failure in one patient, thrombosis of the insertion site (right external jugular vein) during placement in One patient, and One groin hematoma in a patient who was not receiving anticoagulants.
val, 4 days) were reviewed. Clinical follow-up was obtained in 22 patients who underwent no other objective imaging studies or who underwent studies only during a limited time following filter insertion. The average interval of clinical follow-up was 17.8 weeks after placement. The method and duration of follow-up and the number of patients undergoing the examinations are shown in the Table.
,
Figure 2. Pulmonary embolism with LGM filter placement. (a)Inferior vena cavogram obtained prior to filter placement demonstrates absence of clot in the iliac veins after passage of a 5-F pigtail catheter and a large-caliber IVC (diameter, 35 mm). Hand injection of material immediately following puncture of the right common femoral vein demonstrated extensive thrombus. LGM filters were subsequently placed in both common iliac veins because of the large vena cava diameter. (b) Plain radiograph obtained immediately after filter placement demonstrates crowding of the base of the right common iliac vein filter with apparent crossing of one side rail and diagonal strut, which was confirmed at autopsy (arrow).Autopsy Follow-up performed 1day after filter placement Eighteen patients died during the also showedclot in both common iliac period' One patient lefused veins and fresh thrombus in the right One patient pulmonary artery. In this it is could not be contacted. Follow-up im- uncertain whether pulmonary embolism aging studies included inferior vena occurred before or after filter placecavography at an average interval of b. ment. 13.8 weeks following filter placement (n = 33), IVC US at an average of 7.8 Summary of Follow-up Studies weeks (n = 66 [including seven intravascular US examinations performed No. No. of Performed Patients Time after at the University of Arkansas]), du(n = 97) Filter Insertion (wk) Follow-up Study (n = 246) plex US of the insertion site a t an average of 5.7 weeks (n = 53), plain ra33 33 13.8 f 10.8 Vena cavography 66 47 7.8 f 10.8 IVC US diography at an average of 5.7 weeks 53 32 5.7 f 8.9 Insertion site US (n = 68), V/Q scanning at an average 68 48 5.7 f 10.0 Plain radiography of 13.9 weeks (n = 15), CT (four with 15 15 13.9 f 11.7 VIQ scanning intravenous administration of con5 5 10.7 f 9.9 CT trast material and one without) a t an 2 2 5.5 f 0.1 Lower extremity venography average of 10.7 weeks (n = 5), and 4 4 0.5 f 0.3 Autopsy lower extremity venography a t an avNote.-Time after filter insertion is given as an average f standard deviation. Overerage of 5.5 weeks (n = 2). A total of all average time to performance of the follow-upstudy was 9.0 weeks (standarderror, 242 objective imaging studies and re1.8 weeks). sults of four autopsies (average inter-
Murphy et a1
111
Volume 2 Number 1
Inferior Vena Caval Patency Thirty-three cavograms, 66 duplex US scans, five CT scans, and results from four autopsies were obtained in 60 patients to evaluate patency of the inferior vena cava, at an average of 9.5 weeks after filter insertion (range, 1day to 39 weeks 1day; standard deviation, 11.1weeks). Evidence for occlusion of the IVC or both iliac veins by thrombus was present in five patients, including one patient with bilateral iliac vein filters who experienced bilateral iliac vein thrombosis rather than IVC thrombosis. The 6month patency rate was estimated at 92% in this group with the KaplanMeier method (Fig 3). Two of five patients with IVC thrombosis were symptomatic. One patient was dys-
pneic, hypotensive, and bradycardiac the day after filter placement and eventually died. Bilateral iliac vein thrombosis was demonstrated a t autopsy. Another patient experienced transient hypotension, which responded to fluid administration, 1 day after filter insertion. An abdominal contrast-enhanced CT scan obtained that day demonstrated no opacification of the IVC below the filter and collapse of the vena cava above the filter. Cavography demonstrated thrombosis of the IVC below the filter 4 days after filter placement in this patient. Because of the inability of clinical assessment to enable diagnosis of IVC thrombosis with a high degree of sensitivity, clinical follow-up obtained in
0
a
2
a
86
10 20 30 Time following LGM filter insertion in weeks Figure 3. Life-table estimates of probability of maintaining IVC patency. The average percentage probability and standard error of IVC patency are plotted for intervals following LGM filter placement by using the method of Kaplan and Meier. Probability of caval patency at 6 months is 92% (standard error, 4.1%). The number of patients at the beginning of intervals is noted in parentheses. 0
22 patients was not included in assessment of IVC patency. Clinical data from this group revealed only one patient with symptoms possibly attributable to IVC occlusion. The new onset of bilateral lower extremity edema and ascites was noted in this patient 2 weeks after filter insertion. This patient had a history of colon carcinoma, and the significance of the clinical findings in regard to IVC occlusion is unclear. Presence of Thrombus i n Filters Among 42 objective examinations of 35 patients, clot was demonstrated in 13 filters (37%) on 33 cavograms, five CT scans, seven intravascular US scans, and at four autopsies. The average interval for follow-up evaluation was 12.2 weeks (range, 1day to 40 weeks 5 days; standard deviation, 10.9 weeks). Thrombus was occlusive in four patients and nonocclusive in nine. Seven (20%)of 35 patients demonstrated extension of clot in the IVC from 2 mm to 13 cm above the filter (Fig 4). These seven patients had no symptoms of pulmonary embolism. Pulmonary Embolism after Filter Placement The occurrence of pulmonary embolism after LGM filter placement was objectively assessed in 19 patients with results from 15 V/Q scanning procedures and four autopsies. Clinical follow-up alone was obtained in an additional 36 patients. The average follow-up interval was 12.7 weeks (standard deviation, 12.5 weeks). Two patients had abnormal V/Q scans. One was obtained of a 57year-old woman 14.9 weeks after filter placement and was interpreted to indicate a moderate probability of pulmonary embolus. The second was obtained of a 78-year-old woman 26.3 weeks following filter placement and was interpreted as indicating a high probability of pulmonary embolus. V/Q scans were not obtained of these patients prior to filter placement. Chest radiographs of both patients were unremarkable. Neither had experienced signs or symptoms suggestive of pulmonary embolism since fil-
112
Journal of Vascular and Interventional Radiology
February 1991
ter placement. As stated previously, one patient died in the periprocedural period. The cause of death in this patient may have been either pulmonary embolism (before or after filter placement) or inadequate cardiac output caused by thrombosis of the iliac veins bilaterally, as both were demonstrated a t autopsy. A second patient, a 70-year-old woman with a long history of cardiac disease, died suddenly 6 weeks following LGM filter placement. No autopsy was performed. Pulmonary embolism cannot be excluded as a cause of death in this patient, although no evidence of pulmonary embolism was noted in the clinical record. Of the 16 other patients who died during the follow-up period, the causes of death were malignancy (n = lo), sepsis (n = I), massive gastrointestinal hemorrhage (n = I), ischemic necrosis of bowel (n = I), cor pulmonale secondary to pulmonary arterial hypertension (n = 1) (VIQ scan demonstrating one subsegmental defect at the left base and patchy subsegmental defects at the right base, unchanged from findings obtained 3 months previously), irreversible hypotension during surgery (n = I), and aspiration (n = 1). There is therefore no definite case of pulmonary embolism following LGM filter insertion. Of the 19 patients who were evaluated objectively, one patient may have experienced pulmonary embolism immediately prior to or following filter insertion. However, the actual percentage of patients who experienced pulmonary embolism after LGM filter placement cannot definitely be determined due to the small number of patients in this group who underwent objective evaluation. Among the 36 patients who underwent only clinical follow-up, pulmonary embolism may have been responsible for sudden death in one. Insertion Site Thrombosis Insertion sites were evaluated objectively with 53 duplex US scans and three venograms obtained of 35 patients. Clinical follow-up alone was obtained in 20 additional patients.
Follow-up was obtained an average of 9.3 weeks (range, one day to 40 weeks 5 days; standard deviation, 11.9 weeks) following filter insertion. Insertion site thrombosis occurred in eight of 35 patients (23%)evaluated with objective imaging techniques (six duplex US scans and two venograms) an average of 2.1 weeks (standard deviation, 2.3 weeks) after filter la cement. Thrombosis occurred in three right common femoral veins, one left common femoral vein, two right internal jugular veins, one right external jugular vein, and one left internal jugular vein. Another patient, a 57-vear-old woman who was a multiple trauma victim, underwent duplex US examination of the insertion site 4 weeks after LGM filter placement for recurrent pulmonary embolus despite receiving anticoagulants. At duplex US, chronic intimal changes in the proximal (central) femoral vein with occlusive thrombosis present in the distal (peripheral) femoral vein were demonstrated. This patient had not undergone previous evaluation for the presence of deep venous thrombosis at the insertion site. Therefore, it is not known whether thrombosis antedated filter placement. Only one of 20 additional patients evaluated clinically experienced a new onset of symptoms (bilateral lower extremity edema) that could be referable to insertion site thrombosis during the follow-up period. Mechanical Stability of Filter Filter location, angulation, and closure were evaluated with 101 plain radiographs in 65 patients. Examinations were performed an average of 7.8 weeks (range, one day to 40 weeks 7 days; standard deviation, 9.9 weeks) after LGM filter placement. Comparison was made between these images and plain radiographs obtained during filter placement. Caudal migration (greater than 1cm) was present in eight patients (12%).The maximum caudal migration observed was 3 cm (average, 1.7 cm; standard deviation, 0.7 cm). Cephalic migration of 1cm occurred in one patient and was discovered the day after filter inser-
Figure 4. Thrombus trapped in filter with proximal extension. Large thrombus (arrowheads) is trapped within the filter cone. However, a significant extension of thrombus above the filter is present (arrow) and is a potential source of pulmonary embolism.
tion. Significant (greater than 1 cm) filter migration occurred in nine (14%)of 65 patients in whom this was evaluated. Significant (15' or greater) angulation of a filter that did not demonstrate angulation immediately after deposition in the IVC was found in only one patient (2%). This filter was angulated 15' relative to the long axis of the vena cava during cavography performed 12.9 weeks after filter insertion; a second filter was placed cephalad to the first. Significant crowding of the filter struts (greater than a 7-mm decrease in diameter of the filter base relative to the postinsertion diameter) was seen in five patients (8%);four of these had thrombus in the filter or
Murphy et a1
113
Volume 2 Number 1
Figure 5. Incomplete filter opening. Image obtained during in vitro experimentation. The tip of the side rail is indented where the barb is punched out; if a diagonal strut catches on this (arrow), incomplete filter opening results. A recent design modification (use of longer side rails) should help reduce the prevalence of this. thrombus in the IVC. This finding was observed an average of 24.4 weeks following filter placement. I n Vitro Evaluation of Filter Opening Incomplete opening of the LGM filter in the laboratory occurred in five of 40 attempts. All five of these occurred during slow deposition of the filter. This difference was statistically significant (P < .05, x2 analysis). Two of the filters that did not open completely were with femoral systems, and three were with jugular systems. When filters did not open completely, the barbs were caught on the diagonal luminal struts (Fig 5).
DISCUSSION Interruption of the IVC is an accepted adjunct or alternative to anticoagulant therapy to prevent pulmonary embolism in patients with deep venous thrombosis for whom anticoagulation is contraindicated, results in bleeding complications, or is unable to prevent pulmonary embolism
or extension of thrombosis despite administration of an adequate dose. Ideally, IVC filter devices should maintain caval patency while trapping significant emboli (5). The Greenfield vena cava filter is the most widely used of the currently available filters (5). This filter can be introduced into the IVC percutaneously (6-8). The device has been reported to have an acceptable caval patency rate of approximately 95% and a 5% frequency of pulmonary embolism after filter placement with experience extending from 1973 to date (4,9-13). However, clinical and in vitro experimental studies have demonstrated disadvantages of the Greenfield filter that have led to the development of several new IVC filters. One disadvantage of the Greenfield filter is the need to create a large-bore tract to accommodate the 24-F (29.5F outer diameter) sheath used for percutaneous filter introduction. This is presumed to be responsible for the prevalence of thrombosis at the insertion site following filter placement. A second disadvantage of the Greenfield filter is the occurrence of filter tilt after wlacement. which has been shown experimentally to reduce filtering capability (3). Injury to the aorta or right ureter and small bowel perforation have also been observed when angled filter struts penetrate the wall of the IVC (5,14,15).Third, filter migration, reported to occur in 00/0-35% of cases (4,10), can occasionally result in filter embolization to the heart or pulmonary artery, although this has been surprisingly well-tolerated by the patient (16). Finally, the in vitro demonstration of the inability of IVC filters to stop all emboli, especially when improperly situated (3), has led to design modifications to improve filtration of emboli. It has recently been stated that the ideal vena cava filter should result in a low prevalence of caval thrombosis, divide the cava into multiple small channels to prevent emboli, be easily introduced, readily assume an appropriate orientation in the IVC, and maintain stability of this position (5,14). Attempts to achieve these
ideals should not result in increased rates of postinsertion pulmonary embolism, placement vein thrombosis, or mortality compared with those for previously used devices or surgical placement techniques (5,14). The LGM filter is easily introduced, and the procedure is well tolerated. The low-profile introducer system of the LGM filter makes outpatient placement a possibility and increases insertion site alternatives. Insertion was accomwlished via the left common femoral vein in nine patients. Minor resistance was noted in only one patient in whom the filter was eventually successfully placed. The right external jugular vein was used for filter insertion in four patients. The left internal jugular vein was used for introduction in one patient (17). A primary recurrent problem encountered in six patients during filter insertion was incomplete opening. In the in vitro experiment, we found that incomplete opening of the LGM filter occurred during five of 20 attempts when retraction of the introducer was slow. When retraction was fast, incomplete opening was not seen in any of 20 attempts. The nonradiopaque sheath complicates filter insertion in this regard, as it is impossible to verify sheath position during filter placement because of the need to rapidly retract the introducer. Modification of the LGM filter to include a radiopaque sheath would increase operator confidence. For five of six times that incomplete opening of the filter occurred in this series, the filter was introduced via the jugular vein. This may indicate that the mechanics of filter opening after retraction of the introducer are better suited when the filter apex, rather than the base, presents at the tip of the introducer. This difference was not clearly seen in the in vitro experiment but was seen clinically in two patients. This probably results from premature opening of the side rails prior to complete expulsion of the filter apex from the introducer sheath. Presumably, expansion of the filter cone is prevented by partial im-
114
Journal of Vascular and Interventional Radiology
February 1991
plantation of the side rails in the wall of the IVC at that time, although this mechanism was not proved. In the two patients in whom this was observed, filter opening was delayed, occurring after 2 months. Crossing of the legs of the LGM filter, which occurs occasionally with the Greenfield filter. was not observed and would seem unlikely considering the LGM filter design (1). The LGM filter design has been modified to address both the problem of incomplete opening with hooking of the side rail barbs on the diagonal struts and the problem of premature expansion of the side rails. The new design includes longer side rails that extend to the apex of the filter cone. These initial results with the LGM filter indicate that it compares favorably with the Greenfield filter in preventing pulmonary emboli and in maintaining caval patency. The occurrence of pulmonary embolus after LGM filter placement was not well documented in any patient. Although only a small number of patients underwent objective evaluation for pulmonary embolism after filter placement in this series, the results compare favorably with those obtained for the Greenfield filter (11,13) as well as previous data for the LGM filter accumulated in Europe (1).The 92% probability of caval patency at 6 months after filter insertion observed with the LGM filter in this study also is comparable to results with the Greenfield filter and use of objective imaging techniques, with IVC patency rates of up to 96% (11,13). The LGM filter patency rate at 6 months also compares favorably with the 88% patency rate reported in a prospective study of caval patency a t 3 or 6 months after Greenfield filter placement in 216 cases evaluated with venography (18). Only two of five patients with IVC thrombosis diagnosed with use of imaging techniques experienced symptoms. Comparison with other studies in which follow-up was primarily clinical (9-11) is therefore imprudent. Most cases of caval thrombosis occurred early (four of five occurred by
or before 4 weeks 2 days), when deep venous thrombus is poorly adherent to the vein wall (19) and therefore more likely to embolize (20). It is likely that patients are at greatest risk for caval thrombosis after filter placement in this early period. Spontaneous thrombosis of the IVC later in the course of deep venous thrombosis after organization of clot would seem improbable barring additional insult such as recurrent acute deep venous thrombosis. The prevalence of captured thrombus in filters in those patients for whom this could be assessed was high (37%, 13 of 35 patients). In 20% (seven of 35 patients), thrombus extended proximal to the filter. The observed high prevalence of thrombus in LGM filters and the proximal extension above the filters raises the question as to the origin of pulmonary embolism after placement of IVC filters. That is, do these pulmonary emboli represent primary caval, pelvic, or lower extremity thromboemboli that bypass the IVC filter, or are they the result of secondary proximal (central) propagation and embolization of previously trapped thrombi? Experimental studies have demonstrated that most IVC filter devices are effective in removing clots with a diameter greater than 3 mm from the blood stream (3). This, along with observation of large free-floating thrombi extending proximally from LGM filters in this series (Fig 4), suggests that in most cases pulmonary emboli occurring after filter placement probably do not bypass the IVC filter. Instead, they may arise from above the filter, either by proximal propagation of deep venous thrombosis within the cava and through the filter (with or without caval occlusion) or by embolization of proximal extensions of thrombi that are trapped within the IVC filter (21). Therefore, in patients who have IVC filters placed for thromboembolic disease and have transient contraindications to anticoagulation (eg, acute trauma, impending surgery), reinstitution of anticoagulant therapy after an adequate interval so that the risk of bleeding complications is minimal
may be beneficial, especially for patients who are immobile and/or hypercoagulable. As such, the IVC filter would seem to add little overall morbidity to long-term anticoagulation (5). Another desirable feature of IVC filters in this regard would be the exposure of maximal surface area of trapped emboli per unit volume to flowing blood. This would promote rapid thrombolysis of clot trapped within the filter, thereby reducing the potential for trapped thrombus to serve as a nidus for proximal propagation and embolization. Insertion site thrombosis was observed in eight of 35 patients (23%) following LGM filter placement as determined with objective imaging techniques (duplex US and venography). These results are comparable to previous insertion site thrombosis rates of 14.3%-41% (weighted average, 22%; 34 of 158 insertion sites) seen for percutaneous transfemoral placement of the Greenfield filter in patients for whom objective imaging was performed (8,22-24). The presumed superiority of the low-profile LGM filter over the Greenfield filter in this regard was not observed. Similar evaluation of other low-profile filters must be carried out. The LGM filter has demonstrated mechanical stability comparable to that of the Greenfield filter. Greenfield filter migration of more than 1 cm has been observed in 12%of patients with filters in place from l to 9 years (4). Another series demonstrated migration of the Greenfield filter between 5 and 68 mm in 53% of 53 patients (10). There is therefore a wide range of reported results, which is due in part to the criteria used to establish migration. We are confident that migration of more than 1cm is real and not attributable to parallax. Also, in most cases, migration in either direction of less than 1cm is not likely to be clinically significant. Our finding of significant migration greater than 1cm in 14%of patients is similar to observations made by others (1). This is probably at least comparable to findings with the Greenfield filter. The mural stabilizer bars seem to
Murphy et a1
115
Volume 2 Number 1
minimize the risk of significant filter sient for some criteria, such as pulmonary embolism, and the patients angulation, which was observed in evaluated may be asymptomatic, the only one of our patients. This was reported in 8% of patients in previously validity of conclusions derived from reported experience with the LGM small numbers of patients, many with only clinical follow-up, will be limited filter (1).Caval wall penetration seems unlikely to occur without angu- for these criteria. The dubitable nature of these conclusions should be lation of the filter, and this was not observed in our series. acknowledged outright, as was done Crowding of LGM filter struts that with results reported herein for pulmonary embolism after filter placewere open immediately following placement was seen in five of 65 pament. tients (8%). Four of these patients had documented thrombus within the filter. The mechanism of filter closure CONCLUSION in these four cases may be related to We observed an acceptable patency retraction of trapped thrombus or reof the IVC and occurrence of pulmotraction of the entire vena cava following IVC thrombosis. nary embolism after placement of 101 Much of the data obtained for vena LGM filters in 97 patients to justify continued use of this filter. In addicava filter devices combines clinical tion, the LGM filter offers potential and objective imaging results. The significant frequency of asymptomat- advantages over the Greenfield filter, including a decreased prevalence of ic recurrent pulmonary embolism, filter angulation and increased flexiIVC occlusion, and insertion vein thrombosis mandates the use of obbility in selection of the insertion site. There was no significant reduction in jective imaging studies in assessing filters. This can be performed in a rig- the prevalence of thrombosis at the id prospective manner by establishing insertion site compared with previa protocol for acquisition of follow-up ously reported data for the Greenfield filter. We observed incomplete opencavograms, V/Q scans, US scans, and plain radiographs; only patients who ing of the filter in six patients, which, comply with the entire protocol would in part, could be avoided by rapid rebe included. However, due to lack of traction of the insertion catheter over patient cooperation or patient death the implantation cannula, and we during follow-up, this study design hope this problem will be minimized would be impractical. A second meth- by a recent design modification. The od can be used, as in the present sehigh prevalence (20%)of patients ries, in which the duration and type demonstrating clot with cephalad exof patient follow-up is variable, and tension within the filter after placeresults and durations of follow-up for ment remains a concern, and this may those patients undergoing objective contribute to the development of pulimaging evaluation are reported sepa- monary embolism after filter placerately from results for those undergo- ment. Continued evaluation of all ing clinical follow-up only. This may available IVC filter devices with oballow results obtained for different jective imaging criteria is suggested. filters to be compared, assuming similar rates and reasons for patient drop- References 1. Ricco JB, Crochet D, Sebilotte P, e t al. out. A problem with this approach is Percutaneous transvenous caval interrupthat accurate assessment of some crition with the "LGM" filter: early results of teria, such as recurrent pulmonary a multicenter trial. Ann Vasc Surg 1988; embolism, is difficult without multi3:242-247. 2. McCowan TC, Ferris EJ, Carver DK. Inple frequent follow-up examinations ferior vena caval filter thrombi: evaluation for each patient, a large number of with intravascular US. Radiology 1990; patients, and excellent patient com177:783-788. pliance. Since radiologic demonstra3. Katsamouris AA, Waltman AC, Delichattion of abnormalities may be transios MA, Athanasoulis CA. Inferior vena
4.
5. 6.
7.
8. 9.
10.
11.
12. 13.
14. 15. 16.
17.
18.
19. 20.
cava filters: in vitro comparison of clot trapping and flow dynamics. Radiology 1988; 166:361-366. Messmer JM, Greenfield LJ. Greenfield caval filters: long-term radiographic follow-up study. Radiology 1985; 156:613618. Dorfman GS. Percutaneous inferior vena caval filters. Radiology 1990; 174:987-992. Tadavarthy SM, Castaneda-Zuniga W, Salomonowitz E, e t al. Kimray-Greenfield vena cava filter: percutaneous introduction. Radiology 1984; 151:525-526. Denny DF, Cronan J J , Dorfman GS, Esplin C. Percutaneous Kimray-Greenfield filter placement by femoral vein puncture. AJR 1985; 145827-829. Pais SO, Tobin KD. Percutaneous insertion of the Greenfield filter. AJR 1989; 152:933-938. Pais SO, Tobin KD, Austin CB, Qureral L. Percutaneous insertion of the Greenfield inferior vena cava filter: experience with ninety-six patients. J Vasc Surg 1988; 8:460-464. Rose BS, Simon DC, Hess ML, Van Aman ME. Percutaneous transfemoral placement of the Kimray-Greenfield vena cava filter. Radiology 1987; 165373-376. Greenfield LJ, Peyton R, Crute S, Barnes R. Greenfield vena caval filter experience. Arch Surg 1981; 116:1451-1456. Greenfield LJ. Current indications for and results of Greenfield filter placement. J Vasc Surg 1984; 1:502-504. Greenfield LJ, Michna BA. Twelve-year clinical experience with the Greenfield vena cava filter. Surgery 1988; 104:706712. Yune HY. Inferior vena cava filter: search for an ideal device. Radiology 1989; 172: 15-16. Sidawy AN, Menzoian JO. Distal migration and deformation of the Greenfield vena cava filter. Surgery 1986; 99:369-372. Friedell ML, Goldenkranz RJ, Parsonnet V, et al. Migration of a Greenfield filter to the pulmonary artery: a case report. J Vasc Surg 1986; 3:929-931. McCowan TC, Ferris EJ, Carver DK, Harshfield DL. Use of the external jugular vein as a route for percutaneous inferior vena caval filter placement. Radiology 1990; 176:527-530. Clement C, Nicaise H. La retraction des branches du filtre intra-cave de Greenfield: signe indirect de thrombose de la veine cave inferieure. Presse Med 1985; 14:17-56. Flanc C. An experimental study of the recanalization of arterial and venous thrombi. Br J Surg 1968; 55:519-524. Norris CS, Greenfield LJ, Herrmann JB. Free-floating iliofemoral thrombus: a risk of pulmonary embolism. Arch Surg 1985; 120:806-808.
LGM Vena Cava Filter: Objective Evaluation of Early Results1 Timothy P. Murphy, M D Gary S. Dorfman, M D Joseph W . Yedlicka, M D Timothy C. McCowan, M D Robert L. Vogelzang, M D David W . Hunter, M D Danna K. Carver, RN Robert Pinsk, M D Wilfrido Castaneda-Zuniga, M D Ernest J. Ferris, M D Kurt Amplatz, M D
Index terms: Embolism, pulmonary, 60.72 Venae cavae, filters, 982.1299 Venae cavae, thrombosis, 982.442 JVIR 1991; 2107-115
Abbreviations: IVC = inferior vena cava, LGM = LG-Medical, VIQ = ventilation1 perfusion.
One hundred one LG-Medical (LGM) vena cava filters were placed in 97 patients a t four institutions. Placement was a complete technical success in 90% (91 of 101). In 6% of attempts, LGM filter insertion was complicated by incomplete opening of the filter. Pulmonary embolism after filter placement was not definitely demonstrated in any patient. The probability of inferior vena cava patency was 92% a t 6 months after filter insertion. Thrombosis a t the insertion site was seen in eight of 35 patients (23%) evaluated with duplex ultrasound or venography. Thrombus was observed in 37% of filters a t follow-up examination, with cephalic extension of thrombus above the filter in 20% of all patients examined. Filter migration (greater than l cm) was seen in 12%;significant angulation was observed in only one patient (2%). In vitro experimentation demonstrated that incomplete opening of the LGM filter during placement can be avoided, in part, by brisk retraction of the insertion cannula. The low-profile introducer system of the LGM filter allows increased alternatives in selecting the site for filter insertion. The low-profile system also makes outpatient filter placement a possibility. No significant difference in the prevalence of thrombosis a t the insertion site following LGM filter insertion was noted compared with previous results reported for percutaneous transfemoral placement of the Greenfield filter. The nonopaque sheath does not permit careful localization prior to filter deposition. Modification of the LGM filter to include a radiopaque sheath is suggested. T H E LG-Medical
' From the Department of Diagnostic Imaging, Rhode Island Hospital, 593 Eddy St, Providence, RI 02903 (T.P.M., G.S.D.); the Department of Radiology, University of Minnesota Medical School, Minneapolis, Minn (J.W.Y., D.W.H., W.C.Z., K.A.); the Department of Radiology, University of Arkansas for Medical Sciences, Little Rock, Ark (T.C.M., D.K.C., E.J.F.); and the Department of Radiology, Northwestern Memorial Hospital, Chicago (R.L.V., R.P.). From the 1990 RSNA scientific assembly. Received July 31, 1990; revision requested September 10; revision received November 2, accepted November 5. Address reprint requests to G.S.D. O SCVIR, 1991
(LGM) filter (Vena Tech, ~vanstbn,111) is an inferior vena cava (IVC) filter with recent Food and Drug Administration approval for use in the United States. This device, used for the prevention of pulmonary emboli, is introduced with a 12-F outer-diameter system, thereby necessitating a smaller venotomy site and tract than those needed with the stainless steel Greenfield filter (Medi-tech/Boston Scientific, Watertown, Mass), which is inserted through a 24-F sheath (29.5-F outer diameter). Differences between the pyramidal LGM filter and the Greenfield filter include (a) the presence of six mural side rails in the LGM filter that are attached to six radiating diagonal struts and are arranged to look like a cone, analogous to the shape of the Greenfield filter; (b) the metal
struts in the LGM device are broad and flat; and (c) tines or barbs are located on the entire length of the mural side rails of the LGM filter as opposed to the hooks that are limited to the base of the Greenfield filter (Fig 1). The purported advantages of the LGM filter over the stainless steel Greenfield filter are ease of insertion and decreased prevalence of thrombosis of the delivery site vein. These advantages are believed to be obtained with filtration capability and caval patency comparable to those of the standard of reference, the stainless steel Greenfield filter. Herein we describe a series of 97 patients in whom 101 LGM filters were placed. Prospective and objective imaging studies were performed to assess for (a) pulmonary embolism following fil-
108
Journal of Vascular and Interventional Radiology
February 1991
ter insertion, ( b ) patency of the IVC, (c) the presence of nonocclusive thrombus in and extending above the filter, ( d ) insertion site thrombosis following placement, and ( e ) changes in filter position and morphology (angulation and closure). We assessed whether the purported advantages of the LGM filter could be proved in clinical practice and whether the rates of pulmonary embolism and caval patency after filter insertion are comparable to those obtained with the Greenfield filter to justify continued use of the LGM filter. We also performed an in vitro experiment to investigate the influence of the speed of filter release on the occurrence of incomplete filter opening.
PATIENTS AND METHODS The study population included 49 female and 48 male patients with an average age of 61.8 years (range, 1694 years; standard deviation, 15.3 years). Indications for filter insertion included contraindications to anticoagulation (n = 75), progression of deep venous thrombosis or recurrent pulmonary embolism while receiving anticoagulants (n = 12), complications of anticoagulation (n = 6) (including gastrointestinal hemorrhage, vaginal bleeding, and soft-tissue hematoma), and prophylaxis for extensive free-floating or IVC thrombus (n = 4). All patients had documented venous thromboembolic disease. Seventy-two patients had deep venous thrombosis, and 25 patients had pulmonary embolism. Filters were placed by means of the right common femoral vein (n = 65), the right internal jugular vein (n = la), the left common femoral vein (n = 9), the right external jugular vein (n = 4), and the left internal jugular vein (n = 1). The filter was intentionally placed in a suprarenal location in three patients; 94 underwent infrarenal filter placement. Three patients underwent placement of two LGM filters each for complications encountered during insertion. One additional patient had an inferior vena cava diameter of 35
Figure 1. LGM IVC filter. (a) Side view. (b) View from above. Design differs from that of the Greenfield filter by the presence of side rails (arrows),which are in contact with the wall of the IVC along their entire length when properly positioned. Barbs on the outer aspect of the side rails are meant to minimize movement of the filter within the vena cava. Arrowheads indicate diagonal struts. mm (maximum diameter suitable for LGM filter placement is 28 mm), and bilateral common iliac filters were placed. These patients underwent placement of LGM filters at four institutions (Rhode Island Hospital, Providence; University of Minnesota Medical School, Minneapolis; University of Arkansas for Medical Sciences, Little Rock; and Northwestern Memorial Hospital, Chicago) between April 4, 1989, and February 21,1990. All LGM filters inserted at each institution were included in this study. Most filters placed during this time were LGM filters. Factors that may have negatively influenced operator decision to select an LGM filter include lack of familiarity with this filter and, in some cases, reluctance to place this filter in young patients due to lack of long-term experience with this filter. Evaluation of patients in whom these filters were placed was initiated separately at each institution. Collaboration between institutions began during the follow-up period. Therefore, follow-up methods and intervals were variable. Some patients refused to undergo certain examinations, most notably inferior vena cavography, but other studies-transabdominal ultrasound (US) for example-were acceptable to patients, resulting in variability in follow-up methods. However, results obtained with use of different methods that were applica-
ble to a single criterion were combined. For example, results of inferior vena cavography, transabdominal US, computed tomography (CT) with contrast material enhancement, and intravascular US were used to assess patency of the inferior vena cava. Follow-up studies were obtained up to 10 months after placement. Most radiologic studies were obtained prospectively. The technique of LGM filter insertion has been previously described (1). Patency of the IVC following LGM filter insertion was evaluated by means of cavography, transabdominal duplex US and intravascular US, CT, and autopsy (2). The presence of nonocclusive clot in the filter was ascertained with cavography, endovascular US, CT, and autopsy. Transabdominal US was not used to evaluate for nonocclusive thrombus trapped in the filter as there is no evidence to support the sensitivity of this method in the detection of a small thrombus in the vena cava. Pulmonary embolism following filter placement was assessed with ventilationlperfusion (VIQ) scanning, autopsy, and clinical follow-up. In those patients who died during the followup period, efforts were made to determine the possibility of death due to pulmonary embolism. VIQ scans were compared with prior studies when available. Insertion site thrombosis was as-
Murphy et a1
109
Volume 2 Number 1
sessed with lower extremity duplex US, venography, autopsy, and clinical follow-up. Again, comparisons were made with previous studies and clinical history when available. Filter migration, angulation, and closure were evaluated with plain radiography, and the resultant images were compared with radiographs obtained immediately after filter insertion. Filter migration greater than 1 cm was considered significant. Migration less than 1cm was not deemed significant because of the inability to confidently distinguish this degree of movement from perceived change in position due to breathing or parallax error. Filter angulation was considered significant if the measured tilt was 15' or more off the long axis of the IVC, which presumes similarity to the Greenfield filter in this regard (3). Crowding of the filter struts was established if a decrease in the diameter of the filter base greater than 7 mm had occurred since filter placement (4).
Results for each criterion evaluated were analyzed with use of objective imaging findings alone as well as imaging and clinical follow-up combined. Probability of maintaining patency of the IVC over time was projected by using the life-table estimate method of Kaplan and Meier. In vitro evaluation of filter deposition mechanics was performed as follows: Two femoral and two jugular delivery systems were discharged into an empty plastic tube with a 28-mm diameter. Since only opening characteristics were evaluated, it was considered unnecessary to use a fluidfilled tube. The systems were discharged and reloaded for a total of 10 depositions each. The speed of retraction of the insertion sheath was alternated on each delivery between fast (as fast as possible; delivery time, much less than 1sec) and slow (approximately 0.5 cmlsec; total delivery time, approximately 9 sec). The initial delivery for one femoral and one jugular system was fast, while the slow rate was employed in the other two devices. Forty deliveries were performed with the four systems. The
devices used were provided by the manufacturer: all deliveries and reloadings were performed by a single investigator (G.S.D.). Filters were reloaded by hand into the introducing syringe and then discharged from the syringe into the introducing sheath, similar to the method used during manufacture of the filter. The occurrence of complete and incomplete filter opening was tabulated for each device. The significance of difference in the prevalence of incomplete opening based on rate of delivery and type of delivery system was assessed with x2analysis.
RESULTS Filter Insertion: Immediate Technical Success and Complications Complete technical success was achieved in 91 (90%)of 101 of attempts at LGM filter placement. Incomplete opening of the caudal struts was noted in six patients; a second filter was placed above the first in four patients. (Two were LGM and two were Greenfield filters.) In two patients, incomplete opening was attributed to placement of the filter base in the cephalic portion of an IVC thrombus. Two patients with incomplete opening of the filter on insertion did not undergo placement of additional filters due to the operator's belief that while filter opening was suboptimal, it was acceptable and the filter would probably function effectively. Further complications did not occur in these patients. Five of six filters that did not open completely were introduced via the jugular route (four via internal and one via external jugular vein). Three mechanisms of incomplete opening were postulated. These include nonopening due to placement of the base of the filter in the proximal extent of IVC thrombus, hooking of the diagonal strut on the barbs of the side rail, and uneven contact of the side rails with the caval wall with only the proximal end of the side rails touching the wall. The last mecha-
nism is believed to occur when the filter is delivered by means of the jugular approach. Retraction of the sheath after the side rails have exited the sheath but before the apex of the filter has exited causes the cephalic ends of the side rails to become embedded within the wall of the IVC, thereby preventing full expansion of the filter cone. We have no pathologic support for this last mechanism. One patient had a second LGM filter placed due to angulation of the filter on deposition of 20" off the long axis of the IVC. Two filters changed position immediately following placement in the IVC. Of these two, one patient had 1-cm cephalic migration and increased angulation after filter deposition. In the second patient, the filter was placed in a suprarenal location. The filter then migrated 1.5 cm caudad to the level of the renal veins. Finally, resistance to passage of the delivery sheath at the level of the confluence of the iliac veins was encountered in one patient when insertion was performed from the left common femoral vein. The filter was, however, placed successfully with this approach. There were no difficulties in eight other patients in whom filters were introduced through the left femoral vein. Other problems that occurred during filter insertion included inadvertent filter deposition at the iliac confluence in one patient and placement of a filter with a jugular system via the femoral approach due to operator error, resulting in reversed cephalocaudal filter orientation in another. This patient did not have prior pulmonary embolism and had limited lower extremity deep vein thrombosis. The operator believed that placement of an additional filter was not necessary. This inverted filter might function similar to the Amplatz filter, which is an effective filtration device, although it is associated with an increased rate of caval occlusion. Complications of LGM filter placement were encountered in four (4%) of 101 procedures. These included one periprocedural death. The patient was a 60-year-old woman who
110
Journal of Vascular and Interventional Radiology
February 1991
had undergone a total hysterectomy for uterine prolapse 1month previously. She was admitted with shortness of breath and suspected pulmonary embolism, as indicated by means of a high-probability V/Q scan. Filter placement was requested after the development of heparin-induced platelet aggregation syndrome, with the platelet count dropping to 23,000lmL (23 X 106/L).The diameter of the IVC was 35 mm; therefore, bilateral common iliac filters were placed. The patient became acutely dyspneic, bradycardiac, and hypotensive the following Resuscitative efforts were Autopsy showed bilatera1 iliac vein thrombosis and thromboembO1usin the right pulmonary artery. I t is uncertain whether pulmonary embolism occurred immediately prior to placement during the diagnostic Or after placement (Fig 2). Other complications of LGM placement were transient acute renal failure in one patient, thrombosis of the insertion site (right external jugular vein) during placement in One patient, and One groin hematoma in a patient who was not receiving anticoagulants.
val, 4 days) were reviewed. Clinical follow-up was obtained in 22 patients who underwent no other objective imaging studies or who underwent studies only during a limited time following filter insertion. The average interval of clinical follow-up was 17.8 weeks after placement. The method and duration of follow-up and the number of patients undergoing the examinations are shown in the Table.
,
Figure 2. Pulmonary embolism with LGM filter placement. (a)Inferior vena cavogram obtained prior to filter placement demonstrates absence of clot in the iliac veins after passage of a 5-F pigtail catheter and a large-caliber IVC (diameter, 35 mm). Hand injection of material immediately following puncture of the right common femoral vein demonstrated extensive thrombus. LGM filters were subsequently placed in both common iliac veins because of the large vena cava diameter. (b) Plain radiograph obtained immediately after filter placement demonstrates crowding of the base of the right common iliac vein filter with apparent crossing of one side rail and diagonal strut, which was confirmed at autopsy (arrow).Autopsy Follow-up performed 1day after filter placement Eighteen patients died during the also showedclot in both common iliac period' One patient lefused veins and fresh thrombus in the right One patient pulmonary artery. In this it is could not be contacted. Follow-up im- uncertain whether pulmonary embolism aging studies included inferior vena occurred before or after filter placecavography at an average interval of b. ment. 13.8 weeks following filter placement (n = 33), IVC US at an average of 7.8 Summary of Follow-up Studies weeks (n = 66 [including seven intravascular US examinations performed No. No. of Performed Patients Time after at the University of Arkansas]), du(n = 97) Filter Insertion (wk) Follow-up Study (n = 246) plex US of the insertion site a t an average of 5.7 weeks (n = 53), plain ra33 33 13.8 f 10.8 Vena cavography 66 47 7.8 f 10.8 IVC US diography at an average of 5.7 weeks 53 32 5.7 f 8.9 Insertion site US (n = 68), V/Q scanning at an average 68 48 5.7 f 10.0 Plain radiography of 13.9 weeks (n = 15), CT (four with 15 15 13.9 f 11.7 VIQ scanning intravenous administration of con5 5 10.7 f 9.9 CT trast material and one without) a t an 2 2 5.5 f 0.1 Lower extremity venography average of 10.7 weeks (n = 5), and 4 4 0.5 f 0.3 Autopsy lower extremity venography a t an avNote.-Time after filter insertion is given as an average f standard deviation. Overerage of 5.5 weeks (n = 2). A total of all average time to performance of the follow-upstudy was 9.0 weeks (standarderror, 242 objective imaging studies and re1.8 weeks). sults of four autopsies (average inter-
Murphy et a1
111
Volume 2 Number 1
Inferior Vena Caval Patency Thirty-three cavograms, 66 duplex US scans, five CT scans, and results from four autopsies were obtained in 60 patients to evaluate patency of the inferior vena cava, at an average of 9.5 weeks after filter insertion (range, 1day to 39 weeks 1day; standard deviation, 11.1weeks). Evidence for occlusion of the IVC or both iliac veins by thrombus was present in five patients, including one patient with bilateral iliac vein filters who experienced bilateral iliac vein thrombosis rather than IVC thrombosis. The 6month patency rate was estimated at 92% in this group with the KaplanMeier method (Fig 3). Two of five patients with IVC thrombosis were symptomatic. One patient was dys-
pneic, hypotensive, and bradycardiac the day after filter placement and eventually died. Bilateral iliac vein thrombosis was demonstrated a t autopsy. Another patient experienced transient hypotension, which responded to fluid administration, 1 day after filter insertion. An abdominal contrast-enhanced CT scan obtained that day demonstrated no opacification of the IVC below the filter and collapse of the vena cava above the filter. Cavography demonstrated thrombosis of the IVC below the filter 4 days after filter placement in this patient. Because of the inability of clinical assessment to enable diagnosis of IVC thrombosis with a high degree of sensitivity, clinical follow-up obtained in
0
a
2
a
86
10 20 30 Time following LGM filter insertion in weeks Figure 3. Life-table estimates of probability of maintaining IVC patency. The average percentage probability and standard error of IVC patency are plotted for intervals following LGM filter placement by using the method of Kaplan and Meier. Probability of caval patency at 6 months is 92% (standard error, 4.1%). The number of patients at the beginning of intervals is noted in parentheses. 0
22 patients was not included in assessment of IVC patency. Clinical data from this group revealed only one patient with symptoms possibly attributable to IVC occlusion. The new onset of bilateral lower extremity edema and ascites was noted in this patient 2 weeks after filter insertion. This patient had a history of colon carcinoma, and the significance of the clinical findings in regard to IVC occlusion is unclear. Presence of Thrombus i n Filters Among 42 objective examinations of 35 patients, clot was demonstrated in 13 filters (37%) on 33 cavograms, five CT scans, seven intravascular US scans, and at four autopsies. The average interval for follow-up evaluation was 12.2 weeks (range, 1day to 40 weeks 5 days; standard deviation, 10.9 weeks). Thrombus was occlusive in four patients and nonocclusive in nine. Seven (20%)of 35 patients demonstrated extension of clot in the IVC from 2 mm to 13 cm above the filter (Fig 4). These seven patients had no symptoms of pulmonary embolism. Pulmonary Embolism after Filter Placement The occurrence of pulmonary embolism after LGM filter placement was objectively assessed in 19 patients with results from 15 V/Q scanning procedures and four autopsies. Clinical follow-up alone was obtained in an additional 36 patients. The average follow-up interval was 12.7 weeks (standard deviation, 12.5 weeks). Two patients had abnormal V/Q scans. One was obtained of a 57year-old woman 14.9 weeks after filter placement and was interpreted to indicate a moderate probability of pulmonary embolus. The second was obtained of a 78-year-old woman 26.3 weeks following filter placement and was interpreted as indicating a high probability of pulmonary embolus. V/Q scans were not obtained of these patients prior to filter placement. Chest radiographs of both patients were unremarkable. Neither had experienced signs or symptoms suggestive of pulmonary embolism since fil-
112
Journal of Vascular and Interventional Radiology
February 1991
ter placement. As stated previously, one patient died in the periprocedural period. The cause of death in this patient may have been either pulmonary embolism (before or after filter placement) or inadequate cardiac output caused by thrombosis of the iliac veins bilaterally, as both were demonstrated a t autopsy. A second patient, a 70-year-old woman with a long history of cardiac disease, died suddenly 6 weeks following LGM filter placement. No autopsy was performed. Pulmonary embolism cannot be excluded as a cause of death in this patient, although no evidence of pulmonary embolism was noted in the clinical record. Of the 16 other patients who died during the follow-up period, the causes of death were malignancy (n = lo), sepsis (n = I), massive gastrointestinal hemorrhage (n = I), ischemic necrosis of bowel (n = I), cor pulmonale secondary to pulmonary arterial hypertension (n = 1) (VIQ scan demonstrating one subsegmental defect at the left base and patchy subsegmental defects at the right base, unchanged from findings obtained 3 months previously), irreversible hypotension during surgery (n = I), and aspiration (n = 1). There is therefore no definite case of pulmonary embolism following LGM filter insertion. Of the 19 patients who were evaluated objectively, one patient may have experienced pulmonary embolism immediately prior to or following filter insertion. However, the actual percentage of patients who experienced pulmonary embolism after LGM filter placement cannot definitely be determined due to the small number of patients in this group who underwent objective evaluation. Among the 36 patients who underwent only clinical follow-up, pulmonary embolism may have been responsible for sudden death in one. Insertion Site Thrombosis Insertion sites were evaluated objectively with 53 duplex US scans and three venograms obtained of 35 patients. Clinical follow-up alone was obtained in 20 additional patients.
Follow-up was obtained an average of 9.3 weeks (range, one day to 40 weeks 5 days; standard deviation, 11.9 weeks) following filter insertion. Insertion site thrombosis occurred in eight of 35 patients (23%)evaluated with objective imaging techniques (six duplex US scans and two venograms) an average of 2.1 weeks (standard deviation, 2.3 weeks) after filter la cement. Thrombosis occurred in three right common femoral veins, one left common femoral vein, two right internal jugular veins, one right external jugular vein, and one left internal jugular vein. Another patient, a 57-vear-old woman who was a multiple trauma victim, underwent duplex US examination of the insertion site 4 weeks after LGM filter placement for recurrent pulmonary embolus despite receiving anticoagulants. At duplex US, chronic intimal changes in the proximal (central) femoral vein with occlusive thrombosis present in the distal (peripheral) femoral vein were demonstrated. This patient had not undergone previous evaluation for the presence of deep venous thrombosis at the insertion site. Therefore, it is not known whether thrombosis antedated filter placement. Only one of 20 additional patients evaluated clinically experienced a new onset of symptoms (bilateral lower extremity edema) that could be referable to insertion site thrombosis during the follow-up period. Mechanical Stability of Filter Filter location, angulation, and closure were evaluated with 101 plain radiographs in 65 patients. Examinations were performed an average of 7.8 weeks (range, one day to 40 weeks 7 days; standard deviation, 9.9 weeks) after LGM filter placement. Comparison was made between these images and plain radiographs obtained during filter placement. Caudal migration (greater than 1cm) was present in eight patients (12%).The maximum caudal migration observed was 3 cm (average, 1.7 cm; standard deviation, 0.7 cm). Cephalic migration of 1cm occurred in one patient and was discovered the day after filter inser-
Figure 4. Thrombus trapped in filter with proximal extension. Large thrombus (arrowheads) is trapped within the filter cone. However, a significant extension of thrombus above the filter is present (arrow) and is a potential source of pulmonary embolism.
tion. Significant (greater than 1 cm) filter migration occurred in nine (14%)of 65 patients in whom this was evaluated. Significant (15' or greater) angulation of a filter that did not demonstrate angulation immediately after deposition in the IVC was found in only one patient (2%). This filter was angulated 15' relative to the long axis of the vena cava during cavography performed 12.9 weeks after filter insertion; a second filter was placed cephalad to the first. Significant crowding of the filter struts (greater than a 7-mm decrease in diameter of the filter base relative to the postinsertion diameter) was seen in five patients (8%);four of these had thrombus in the filter or
Murphy et a1
113
Volume 2 Number 1
Figure 5. Incomplete filter opening. Image obtained during in vitro experimentation. The tip of the side rail is indented where the barb is punched out; if a diagonal strut catches on this (arrow), incomplete filter opening results. A recent design modification (use of longer side rails) should help reduce the prevalence of this. thrombus in the IVC. This finding was observed an average of 24.4 weeks following filter placement. I n Vitro Evaluation of Filter Opening Incomplete opening of the LGM filter in the laboratory occurred in five of 40 attempts. All five of these occurred during slow deposition of the filter. This difference was statistically significant (P < .05, x2 analysis). Two of the filters that did not open completely were with femoral systems, and three were with jugular systems. When filters did not open completely, the barbs were caught on the diagonal luminal struts (Fig 5).
DISCUSSION Interruption of the IVC is an accepted adjunct or alternative to anticoagulant therapy to prevent pulmonary embolism in patients with deep venous thrombosis for whom anticoagulation is contraindicated, results in bleeding complications, or is unable to prevent pulmonary embolism
or extension of thrombosis despite administration of an adequate dose. Ideally, IVC filter devices should maintain caval patency while trapping significant emboli (5). The Greenfield vena cava filter is the most widely used of the currently available filters (5). This filter can be introduced into the IVC percutaneously (6-8). The device has been reported to have an acceptable caval patency rate of approximately 95% and a 5% frequency of pulmonary embolism after filter placement with experience extending from 1973 to date (4,9-13). However, clinical and in vitro experimental studies have demonstrated disadvantages of the Greenfield filter that have led to the development of several new IVC filters. One disadvantage of the Greenfield filter is the need to create a large-bore tract to accommodate the 24-F (29.5F outer diameter) sheath used for percutaneous filter introduction. This is presumed to be responsible for the prevalence of thrombosis at the insertion site following filter placement. A second disadvantage of the Greenfield filter is the occurrence of filter tilt after wlacement. which has been shown experimentally to reduce filtering capability (3). Injury to the aorta or right ureter and small bowel perforation have also been observed when angled filter struts penetrate the wall of the IVC (5,14,15).Third, filter migration, reported to occur in 00/0-35% of cases (4,10), can occasionally result in filter embolization to the heart or pulmonary artery, although this has been surprisingly well-tolerated by the patient (16). Finally, the in vitro demonstration of the inability of IVC filters to stop all emboli, especially when improperly situated (3), has led to design modifications to improve filtration of emboli. It has recently been stated that the ideal vena cava filter should result in a low prevalence of caval thrombosis, divide the cava into multiple small channels to prevent emboli, be easily introduced, readily assume an appropriate orientation in the IVC, and maintain stability of this position (5,14). Attempts to achieve these
ideals should not result in increased rates of postinsertion pulmonary embolism, placement vein thrombosis, or mortality compared with those for previously used devices or surgical placement techniques (5,14). The LGM filter is easily introduced, and the procedure is well tolerated. The low-profile introducer system of the LGM filter makes outpatient placement a possibility and increases insertion site alternatives. Insertion was accomwlished via the left common femoral vein in nine patients. Minor resistance was noted in only one patient in whom the filter was eventually successfully placed. The right external jugular vein was used for filter insertion in four patients. The left internal jugular vein was used for introduction in one patient (17). A primary recurrent problem encountered in six patients during filter insertion was incomplete opening. In the in vitro experiment, we found that incomplete opening of the LGM filter occurred during five of 20 attempts when retraction of the introducer was slow. When retraction was fast, incomplete opening was not seen in any of 20 attempts. The nonradiopaque sheath complicates filter insertion in this regard, as it is impossible to verify sheath position during filter placement because of the need to rapidly retract the introducer. Modification of the LGM filter to include a radiopaque sheath would increase operator confidence. For five of six times that incomplete opening of the filter occurred in this series, the filter was introduced via the jugular vein. This may indicate that the mechanics of filter opening after retraction of the introducer are better suited when the filter apex, rather than the base, presents at the tip of the introducer. This difference was not clearly seen in the in vitro experiment but was seen clinically in two patients. This probably results from premature opening of the side rails prior to complete expulsion of the filter apex from the introducer sheath. Presumably, expansion of the filter cone is prevented by partial im-
114
Journal of Vascular and Interventional Radiology
February 1991
plantation of the side rails in the wall of the IVC at that time, although this mechanism was not proved. In the two patients in whom this was observed, filter opening was delayed, occurring after 2 months. Crossing of the legs of the LGM filter, which occurs occasionally with the Greenfield filter. was not observed and would seem unlikely considering the LGM filter design (1). The LGM filter design has been modified to address both the problem of incomplete opening with hooking of the side rail barbs on the diagonal struts and the problem of premature expansion of the side rails. The new design includes longer side rails that extend to the apex of the filter cone. These initial results with the LGM filter indicate that it compares favorably with the Greenfield filter in preventing pulmonary emboli and in maintaining caval patency. The occurrence of pulmonary embolus after LGM filter placement was not well documented in any patient. Although only a small number of patients underwent objective evaluation for pulmonary embolism after filter placement in this series, the results compare favorably with those obtained for the Greenfield filter (11,13) as well as previous data for the LGM filter accumulated in Europe (1).The 92% probability of caval patency at 6 months after filter insertion observed with the LGM filter in this study also is comparable to results with the Greenfield filter and use of objective imaging techniques, with IVC patency rates of up to 96% (11,13). The LGM filter patency rate at 6 months also compares favorably with the 88% patency rate reported in a prospective study of caval patency a t 3 or 6 months after Greenfield filter placement in 216 cases evaluated with venography (18). Only two of five patients with IVC thrombosis diagnosed with use of imaging techniques experienced symptoms. Comparison with other studies in which follow-up was primarily clinical (9-11) is therefore imprudent. Most cases of caval thrombosis occurred early (four of five occurred by
or before 4 weeks 2 days), when deep venous thrombus is poorly adherent to the vein wall (19) and therefore more likely to embolize (20). It is likely that patients are at greatest risk for caval thrombosis after filter placement in this early period. Spontaneous thrombosis of the IVC later in the course of deep venous thrombosis after organization of clot would seem improbable barring additional insult such as recurrent acute deep venous thrombosis. The prevalence of captured thrombus in filters in those patients for whom this could be assessed was high (37%, 13 of 35 patients). In 20% (seven of 35 patients), thrombus extended proximal to the filter. The observed high prevalence of thrombus in LGM filters and the proximal extension above the filters raises the question as to the origin of pulmonary embolism after placement of IVC filters. That is, do these pulmonary emboli represent primary caval, pelvic, or lower extremity thromboemboli that bypass the IVC filter, or are they the result of secondary proximal (central) propagation and embolization of previously trapped thrombi? Experimental studies have demonstrated that most IVC filter devices are effective in removing clots with a diameter greater than 3 mm from the blood stream (3). This, along with observation of large free-floating thrombi extending proximally from LGM filters in this series (Fig 4), suggests that in most cases pulmonary emboli occurring after filter placement probably do not bypass the IVC filter. Instead, they may arise from above the filter, either by proximal propagation of deep venous thrombosis within the cava and through the filter (with or without caval occlusion) or by embolization of proximal extensions of thrombi that are trapped within the IVC filter (21). Therefore, in patients who have IVC filters placed for thromboembolic disease and have transient contraindications to anticoagulation (eg, acute trauma, impending surgery), reinstitution of anticoagulant therapy after an adequate interval so that the risk of bleeding complications is minimal
may be beneficial, especially for patients who are immobile and/or hypercoagulable. As such, the IVC filter would seem to add little overall morbidity to long-term anticoagulation (5). Another desirable feature of IVC filters in this regard would be the exposure of maximal surface area of trapped emboli per unit volume to flowing blood. This would promote rapid thrombolysis of clot trapped within the filter, thereby reducing the potential for trapped thrombus to serve as a nidus for proximal propagation and embolization. Insertion site thrombosis was observed in eight of 35 patients (23%) following LGM filter placement as determined with objective imaging techniques (duplex US and venography). These results are comparable to previous insertion site thrombosis rates of 14.3%-41% (weighted average, 22%; 34 of 158 insertion sites) seen for percutaneous transfemoral placement of the Greenfield filter in patients for whom objective imaging was performed (8,22-24). The presumed superiority of the low-profile LGM filter over the Greenfield filter in this regard was not observed. Similar evaluation of other low-profile filters must be carried out. The LGM filter has demonstrated mechanical stability comparable to that of the Greenfield filter. Greenfield filter migration of more than 1 cm has been observed in 12%of patients with filters in place from l to 9 years (4). Another series demonstrated migration of the Greenfield filter between 5 and 68 mm in 53% of 53 patients (10). There is therefore a wide range of reported results, which is due in part to the criteria used to establish migration. We are confident that migration of more than 1cm is real and not attributable to parallax. Also, in most cases, migration in either direction of less than 1cm is not likely to be clinically significant. Our finding of significant migration greater than 1cm in 14%of patients is similar to observations made by others (1). This is probably at least comparable to findings with the Greenfield filter. The mural stabilizer bars seem to
Murphy et a1
115
Volume 2 Number 1
minimize the risk of significant filter sient for some criteria, such as pulmonary embolism, and the patients angulation, which was observed in evaluated may be asymptomatic, the only one of our patients. This was reported in 8% of patients in previously validity of conclusions derived from reported experience with the LGM small numbers of patients, many with only clinical follow-up, will be limited filter (1).Caval wall penetration seems unlikely to occur without angu- for these criteria. The dubitable nature of these conclusions should be lation of the filter, and this was not observed in our series. acknowledged outright, as was done Crowding of LGM filter struts that with results reported herein for pulmonary embolism after filter placewere open immediately following placement was seen in five of 65 pament. tients (8%). Four of these patients had documented thrombus within the filter. The mechanism of filter closure CONCLUSION in these four cases may be related to We observed an acceptable patency retraction of trapped thrombus or reof the IVC and occurrence of pulmotraction of the entire vena cava following IVC thrombosis. nary embolism after placement of 101 Much of the data obtained for vena LGM filters in 97 patients to justify continued use of this filter. In addicava filter devices combines clinical tion, the LGM filter offers potential and objective imaging results. The significant frequency of asymptomat- advantages over the Greenfield filter, including a decreased prevalence of ic recurrent pulmonary embolism, filter angulation and increased flexiIVC occlusion, and insertion vein thrombosis mandates the use of obbility in selection of the insertion site. There was no significant reduction in jective imaging studies in assessing filters. This can be performed in a rig- the prevalence of thrombosis at the id prospective manner by establishing insertion site compared with previa protocol for acquisition of follow-up ously reported data for the Greenfield filter. We observed incomplete opencavograms, V/Q scans, US scans, and plain radiographs; only patients who ing of the filter in six patients, which, comply with the entire protocol would in part, could be avoided by rapid rebe included. However, due to lack of traction of the insertion catheter over patient cooperation or patient death the implantation cannula, and we during follow-up, this study design hope this problem will be minimized would be impractical. A second meth- by a recent design modification. The od can be used, as in the present sehigh prevalence (20%)of patients ries, in which the duration and type demonstrating clot with cephalad exof patient follow-up is variable, and tension within the filter after placeresults and durations of follow-up for ment remains a concern, and this may those patients undergoing objective contribute to the development of pulimaging evaluation are reported sepa- monary embolism after filter placerately from results for those undergo- ment. Continued evaluation of all ing clinical follow-up only. This may available IVC filter devices with oballow results obtained for different jective imaging criteria is suggested. filters to be compared, assuming similar rates and reasons for patient drop- References 1. Ricco JB, Crochet D, Sebilotte P, e t al. out. A problem with this approach is Percutaneous transvenous caval interrupthat accurate assessment of some crition with the "LGM" filter: early results of teria, such as recurrent pulmonary a multicenter trial. Ann Vasc Surg 1988; embolism, is difficult without multi3:242-247. 2. McCowan TC, Ferris EJ, Carver DK. Inple frequent follow-up examinations ferior vena caval filter thrombi: evaluation for each patient, a large number of with intravascular US. Radiology 1990; patients, and excellent patient com177:783-788. pliance. Since radiologic demonstra3. Katsamouris AA, Waltman AC, Delichattion of abnormalities may be transios MA, Athanasoulis CA. Inferior vena
4.
5. 6.
7.
8. 9.
10.
11.
12. 13.
14. 15. 16.
17.
18.
19. 20.
cava filters: in vitro comparison of clot trapping and flow dynamics. Radiology 1988; 166:361-366. Messmer JM, Greenfield LJ. Greenfield caval filters: long-term radiographic follow-up study. Radiology 1985; 156:613618. Dorfman GS. Percutaneous inferior vena caval filters. Radiology 1990; 174:987-992. Tadavarthy SM, Castaneda-Zuniga W, Salomonowitz E, e t al. Kimray-Greenfield vena cava filter: percutaneous introduction. Radiology 1984; 151:525-526. Denny DF, Cronan J J , Dorfman GS, Esplin C. Percutaneous Kimray-Greenfield filter placement by femoral vein puncture. AJR 1985; 145827-829. Pais SO, Tobin KD. Percutaneous insertion of the Greenfield filter. AJR 1989; 152:933-938. Pais SO, Tobin KD, Austin CB, Qureral L. Percutaneous insertion of the Greenfield inferior vena cava filter: experience with ninety-six patients. J Vasc Surg 1988; 8:460-464. Rose BS, Simon DC, Hess ML, Van Aman ME. Percutaneous transfemoral placement of the Kimray-Greenfield vena cava filter. Radiology 1987; 165373-376. Greenfield LJ, Peyton R, Crute S, Barnes R. Greenfield vena caval filter experience. Arch Surg 1981; 116:1451-1456. Greenfield LJ. Current indications for and results of Greenfield filter placement. J Vasc Surg 1984; 1:502-504. Greenfield LJ, Michna BA. Twelve-year clinical experience with the Greenfield vena cava filter. Surgery 1988; 104:706712. Yune HY. Inferior vena cava filter: search for an ideal device. Radiology 1989; 172: 15-16. Sidawy AN, Menzoian JO. Distal migration and deformation of the Greenfield vena cava filter. Surgery 1986; 99:369-372. Friedell ML, Goldenkranz RJ, Parsonnet V, et al. Migration of a Greenfield filter to the pulmonary artery: a case report. J Vasc Surg 1986; 3:929-931. McCowan TC, Ferris EJ, Carver DK, Harshfield DL. Use of the external jugular vein as a route for percutaneous inferior vena caval filter placement. Radiology 1990; 176:527-530. Clement C, Nicaise H. La retraction des branches du filtre intra-cave de Greenfield: signe indirect de thrombose de la veine cave inferieure. Presse Med 1985; 14:17-56. Flanc C. An experimental study of the recanalization of arterial and venous thrombi. Br J Surg 1968; 55:519-524. Norris CS, Greenfield LJ, Herrmann JB. Free-floating iliofemoral thrombus: a risk of pulmonary embolism. Arch Surg 1985; 120:806-808.