| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Vascular and Interventional Radiology |
1 From the Department of Radiology, GRB 290A, Massachusetts General Hospital, 32 Fruit St, Boston, MA 02114, and Harvard Medical School. Received May 25, 1999; revision requested July 16; final revision received September 29; accepted October 20. Address correspondence to C.A.A. (e-mail: athanasoulis.christos@mgh.harvard.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADDENDUMREFERENCES |
|---|
MATERIALS AND METHODS: During 1973–1998, 1,765 filters were implanted in 1,731 patients. Hospital files were reviewed, and data were collected about the indications, safety, effectiveness, numbers, and types of caval filters. Fatal postfilter pulmonary embolism (PE) was considered the primary outcome. Morbidity and mortality were determined as secondary outcomes. Survival and morbidity-free survival curves were calculated.
RESULTS: The prevalence of observed postfilter PE was 5.6%. It was fatal in 3.7% of patients. In most patients, fatal PE occurred soon after filter insertion (median, 4.0 days; 95% CI: 2.2, 5.8 days). Major complications occurred in 0.3% of procedures. The prevalence of observed postfilter caval thrombosis was 2.7%. The 30-day mortality rate was 17.0% overall, higher among patients with neoplasms (19.5%) as compared with those without neoplasms (14.3%; P = .004). Filter efficacy and associated morbidity were not different in 46 patients with suprarenal filters. The rate of filters placed for prophylaxis was 4.7% overall and increased to 16.4% in 1998. From 1980 to 1996, there was a fivefold increase in the number of caval filter implants. In recent years, more filters were implanted in younger patients.
CONCLUSION: Inferior vena caval filters provide protection from life-threatening PE, with minimal morbidity.
Index terms: Embolism, pulmonary, 60.72, 944.77 • Interventional procedures, complications, 982.77 • Venae cavae, filters, 982.1267 • Venae cavae, interventional procedures, 982.1267
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADDENDUMREFERENCES |
|---|
An early umbrella type of filter was introduced in the late 1960s, and relevant articles appeared with experimental data in 1967 (2) and clinical data in 1969 (3). An unacceptably high prevalence of caval thrombosis led to the demise of this umbrella filter. A method based on complete and permanent balloon obstruction of the inferior vena cava was proposed soon thereafter (4). It did not receive wide clinical acceptance. In 1973, "a new intracaval filter permitting continued flow and resolution of emboli" was introduced (5). It was the Greenfield stainless steel device, which became the standard caval filter.
The new technology was received with reservations at first and with mounting enthusiasm afterward. By the late 1970s, surgical interruption of the cava for the prevention of pulmonary embolism (PE) was virtually abandoned. However, the Greenfield filter required venotomy for insertion. Percutaneous introduction was desirable. Laboratory research continued for the development of new small-profile devices suitable for percutaneous insertion (6,7). The Bird's Nest device was the first true percutaneous filter. It became available for limited clinical use on an investigational basis in 1983, and initial results were published in 1984 (8). Another 4 years passed before it was released for wide clinical use.
In the interim, interventionalists became impatient. Physicians, mostly radiologists skillful in percutaneous vascular catheterization techniques, devised ways to insert the Greenfield filter percutaneously through large-bore femoral venous sheaths (9). This approach eliminated the need for a venotomy, but it was associated with a high prevalence of venous thrombosis at insertion sites (10). The transitional period lasted through the late 1980s, and the method was abandoned when additional small-profile percutaneous caval filters became available (11–14).
Investigators published articles declaring the effectiveness and safety of caval filters (15–18). Others, pointing to the lack of randomized clinical trials and the scarcity of long-term follow-up data, called for caution (19–22). But the percutaneously inserted caval filter was gaining ground. Ease and simplicity of insertion led to increasing use, while calls for relaxing and expanding the indications were getting more pronounced (23,24). Today, the placement of caval filters for prophylaxis is considered appropriate, efforts have intensified for the development of temporary retrievable devices, and informed debate continues about using caval filters in lieu of anticoagulant therapy (25–27).
Continuous long-term experience with caval filters at one health care facility mirrors their development. The purpose of this study was to present data we have collected from a 26-year clinical experience with caval filters and to submit observations relevant to their introduction, adoption, application, effectiveness, and safety.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADDENDUMREFERENCES |
|---|
Caval filters were placed by vascular radiologists, in an angiography procedure room, with the use of imaging guidance. In four patients who had two filters each, the first devices had been implanted at other hospitals where the insertion procedure may have been different. If surgical exposure of a vein was necessary for venous access, a surgeon performed the access venotomy in the angiography room, and the vascular radiologist inserted and deployed the filter. When percutaneous filters became available, the entire procedure was performed by vascular radiologists. The methodology was consistent during the 26-year study period. Inferior venacavography was performed prior to and following filter insertion. The position of the renal veins and other pertinent anatomic venous findings were recorded. The aim was to place the filter just below the renal veins, unless there were reasons dictating otherwise.
Various types of permanent and two types of temporary (retrievable) filters were implanted in the course of 26 years. Table 1 lists the types of devices and the number of each type implanted. We note that in the early literature, the standard Greenfield filter was also referred to as the "Kimray Greenfield filter." The Greenfield filter was designed for introduction by means of venotomy. However, in the 1980s, large-bore introducer sheaths were used for percutaneous insertion of this device. We make the distinction because percutaneous introduction was associated with different procedural complications. However, the reader should keep in mind that the "Greenfield 24-F filter" and "Greenfield 24-F filter introduced percutaneously" refer to the same device, which is to be distinguished from the true percutaneous Greenfield filters developed later.
|
Fatal postfilter PE was considered the primary outcome. Event-free survival and time from procedure-to-event curves were calculated according to the Kaplan-Meier method. Overall survival curves based on the risk factor of presence of underlying neoplasm similarly were calculated. Morbidity, including caval thrombosis, and mortality were considered secondary outcomes. Computations were performed with the use of statistical software (SPSS; SPSS, Chicago, Ill). Where appropriate, multiple comparisons of means were performed by means of the one-way analysis of variance procedure and post hoc analysis with the Bonferroni method. A P value less than .05 was considered to indicate a statistically significant difference. The log-rank test was used for comparisons of survival curves between groups of patients.
Computations required the use of different denominators in different sections of data analysis. The total number of patients (n = 1,739) who underwent a procedure for filter placement was used to analyze demographic data. The number of patients (n = 1,731) who had a device (one or two filters) implanted was the basis for calculating follow-up results. Where appropriate, reference was to the number of filter insertion procedures (n = 1,753), the number of thromboembolic events (n = 1,753), or the number of devices inserted (n = 1,765). A summary listing appears in Table 2.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADDENDUMREFERENCES |
|---|
Insertion was attempted but no filter was implanted.—In eight patients, no filter was implanted, despite attempted insertion. In one patient in 1979, a Greenfield 24-F filter was dislodged from the delivery capsule during insertion, and it became entangled in the right atrium. Cardiac surgeons removed the filter by means of cardiotomy through the atrial appendage. The cava was ligated. In three patients (one in 1980 and two in 1982), the delivery capsule of a standard Greenfield 24-F filter could not be advanced beyond the clavicle. A caval clip was placed in two patients, and the other was treated with heparin sodium. In two patients in 1986, attempted percutaneous insertion of a Greenfield 24-F filter via the femoral vein was not successful. Tortuosity of the iliac arteries distorted the course of the iliac veins and prevented the passage of the filter capsule into the cava. Both patients were treated with heparin. Extensive chronic caval thrombosis was found in two patients (one in 1985 and one in 1995), and the insertion of a filter was aborted. Both patients were treated with anticoagulants.
Patients with two filters.—Thirty-four patients had two filters implanted. In twenty of these patients, two devices were inserted during one session, whereas in the remaining fourteen, implantation of a second filter took place during another later procedure. The mean time between the first and the second procedure was 416 days (range, 1–2,694 days). Reasons why a second filter was considered necessary included postfilter PE (12 patients), double inferior vena cava (three patients), circumaortic left renal vein (one patient), and technical issues at the time of filter insertion (18 patients). Technical issues included the following: low placement of the first device, with the body of the filter positioned mostly in an iliac vein (seven patients); incomplete filter expansion (five patients); and extreme filter tilting (three patients). Also included were a filter with a broken strut (one patient) and one instance of mistakenly introducing from a femoral vein a device designed for a jugular venous approach (one patient). In the last patient, a second filter became necessary when the previously implanted device was dislodged accidentally from the cava during insertion of a central venous catheter.
Number of Patients per Year and Demographic Changes over Time
Figure 1 shows the annual distribution of 1,739 patients who underwent caval filter placement. For fourteen patients who had two filters implanted during two different sessions, reference is to the date of the first procedure. Also shown, as a point of reference, is the number of patients per year who, during the same time period, underwent surgical interruption of the vena cava (n = 238). The number of patients who received caval filters each year has been increasing. This rising trend is illustrated in Figure 1. Thirty-five patients had a caval filter in 1980. There was a twofold increase by 1990 and a fivefold increase by 1996.
|
2). The trend is illustrated graphically in Figure 2.
|
|
|
|
|
2, goodness of fit with one degree of freedom, no continuity correction, based on 839 cases [467 left + 372 right]; left, 55.7% [467 of 839]; right, 44.3% [372 of 839]). In 47 patients, thrombosis was limited to the veins of the calf below the knee. In 40 of these 47 patients, the diagnosis was established by means of conventional venography; US was the imaging test in only seven. When US replaced venography, the calf veins were largely ignored during the imaging study. The diagnosis of calf venous thrombosis consequently was made only sporadically when US became the main imaging modality for the diagnosis of deep venous thrombosis.
There was no particular pattern observed when the numbers of filters implanted were analyzed by month or by season. Most procedures were performed on a Friday (331 [18.9%] of 1,753). Only 6.0% (106 of 1,753) of the procedures were performed on Sundays and 8.7% (152 of 1,753) on Saturdays (Fig 6). The day-to-day difference was significant for Sunday and for Saturday when compared with all weekdays (P < .05). The differences among other weekdays, excluding weekends, were not significant (P > .05, 2, goodness of fit test, Bonferroni post hoc analysis).
|
|
Thrombus in the vena cava or its tributaries was found in 105 cases. It was an isolated caval thrombus in 44 (2.5%) of 1,753 procedures, an extension of iliac venous thrombosis into the cava in 40 (2.3%), or thrombosis of a common iliac vein in 21 (1.2%). Isolated caval thrombus was above, below, or both above and below the renal veins in three, 37, and four cases, respectively.
Problems encountered during filter insertion.—Intraprocedural problems were encountered in 6.0% (106 of 1,753) of procedures. In 43 procedures, technical problems related to venous access. In 61 of the 106 cases, technical problems were related to filter deployment. Deployment problems were observed most frequently with the Greenfield slim device (two [9.1%] of 22) and least frequently with the Greenfield standard filter (10 [2.2%] of 458). Full expansion was achieved in most cases by pulling or pushing on the filter with the aid of a catheter or guide wire. In two cases, there were unusual problems. In one patient, a Simon nitinol filter was deployed mistakenly in an ascending lumbar vein and retrieved, followed by correct placement in the vena cava. In another case, a Simon nitinol device intended, packaged, and labeled for insertion via a jugular vein was introduced mistakenly by means of a femoral route. The filter was inverted. A second device was implanted above the first.
Acute procedural complications.—Complications during the procedure of filter insertion occurred in 18 cases (Table 5); they were considered major in 0.3% (six of 1,753) of procedures. In one patient, a Greenfield 24-F filter inserted via the jugular vein was released prematurely and was entangled in the interstices of the right atrium. Cardiotomy was performed to extract the device. In two patients, the inserted filter migrated to the right atrium at deployment. One episode (in 1985) involved a Bird's Nest filter, of the so-called first generation, before the anchoring hooks were strengthened. The second episode (in 1995) involved a Simon nitinol device. Both of these filters were retrieved percutaneously. Inadvertent puncture of the carotid artery during percutaneous insertion of a Greenfield 24-F filter resulted in a mediastinal hematoma in one case. The patient recovered.
|
Technical Data
Types of filters implanted over time.—Eight types of caval filters were implanted in the course of 26 years. Different types of filters were relatively more popular at different times. Relevant trends are illustrated in Figure 7. The Greenfield 24-F filter was used until 1989. During the years 1989-1994, a variety of percutaneous filters were used. In 1995, the Simon nitinol became the device most frequently implanted. During 1995-1998, 80.4% (505 of 628) of the filters implanted were of the Simon nitinol type.
|
Filter position.—The standard position for filter deployment was in the infrarenal cava, close to the point of entry of the renal veins. This was the intended position in all cases. However, in 83 cases (4.7% of 1,765 filters), the device was deployed and implanted in a different location: above the renal veins in 33 cases, in a renal vein in two, adjacent to the renal veins in 11, and low in the distal cava in 37. Placement of a filter in a location other than the "typical" was unintentional in 20 (1.1%) and intentional in 63 (3.6%) of 1,765 cases. The reasons for alternate filter positioning were anatomic variations, presence of thrombus in the cava, or technical difficulties. Ten of the 33 filters placed in the suprarenal cava were second filters in patients who had two implants.
Follow-up
Follow-up information in the eight patients who had an attempt at filter placement but received no filter is as follows: Two patients died within 24 hours after the procedure. PE was the major cause of death and was confirmed at autopsy. Three patients died of unrelated causes at 16, 26, and 3,722 days after the procedure. Two patients were alive and well at 4,516 and 4,489 days. One patient with cerebral embolism was lost to follow-up. We will present later the follow-up data in 1,731 patients who had one or more filters implanted. The mean length of clinical follow-up was 479 days (range, 0–6,082 days).
Primary Outcomes
Postfilter PE.—PE after filter placement was observed in 5.6% (97 of 1,731) of patients. In 12 (0.7%) of these patients, PE was an incidental autopsy finding, unrelated to the cause of death. Clinically relevant postfilter PE was observed in 4.9% (85 of 1,731) of patients. It was fatal in 3.8% (65 of 1,731) of patients. The mean time between the placement of a filter and the time of postfilter PE was 135 days ± 433 (SD; range, 0–2,694 days; median, 8 days).
Postfilter PE was confirmed by means of autopsy in 52, pulmonary angiography in eight, and radionuclide lung scan in three patients. In another three patients, the clinical suspicion of postfilter PE was supported by the finding of extensive caval thrombosis noted during cavography. In the remaining 31 patients (1.8%), the diagnosis of postfilter PE was made on clinical grounds only. There was no correlation between the nature of the original thromboembolic event (PE or leg venous thrombosis) and the development of PE after filter placement. Similarly, there was no correlation between the presence or absence of an underlying neoplasm and the development of postfilter PE. Management of postfilter PE included the placement of a second filter in 12 patients.
Fatal postfilter PE.—PE in the presence of a caval filter was considered the cause or main factor contributing to death in 3.8% (65 of 1,731) of patients. Fatal PE was confirmed by means of autopsy in 40 (2.3%) of 1,731 patients. In 12 of these 40 patients, the terminal event occurred within 24 hours after filter placement. In these patients, it could not be determined with certainty whether death was due to recurrent PE or whether it was the result of embolism preceding the placement of a filter. If these 12 patients were excluded, the prevalence of postfilter, autopsy-proved fatal PE was 1.6% (28 of 1,731 patients). Fatal PE occurred soon after filter placement in most patients. Figure 8 shows the Kaplan-Meier survival curve for the 65 patients who died as a result of PE. The median survival time was 4.0 days (95% CI: 2.2, 5.8 days). Figure 9 shows the Kaplan-Meier event-free survival curve for all 1,731 patients, where the event is defined as fatal postfilter PE; the median could not be estimated because of the large numbers of patients alive at the end of data collection.
|
|
|
2, goodness of fit). Late morbidity.—Late complications, occurring beyond 30 days after filter insertion, developed in 108 (6.2%) of 1,731 patients. The complications were caval thrombosis in 30 patients, new leg venous thrombosis in 46, and new or worse leg edema in 23. Other infrequent complications included minimal caudal filter migration in two patients (one Vena Tech and one Simon nitinol filter) and broken filter struts in another three patients (one Greenfield standard and two Bird's Nest devices). In one patient, a strut of a Simon nitinol filter protruded through the caval wall toward the spine. All four patients (0.2% of 1,731) with broken or protruding struts have remained asymptomatic (mean follow-up, 2,834 days). Three miscellaneous late complications occurred: groin pseudoaneurysm in one patient, hypertrophic scar in the neck at the site of a Greenfield 24-F filter insertion in one, and guide wire entrapment with dislodgment of a Vena Tech filter in one.
Postfilter thrombosis of the inferior vena cava.—Vena caval thrombosis after filter placement was observed in 3.2% (55 of 1,731) of patients, if we add both the 30-day and the later cases (Table 6). The prevalence was highest for the Mobin-Uddin device (nine [32.1%] of 28). Multiple comparisons of the prevalence of observed postfilter caval thrombosis among all types of filters showed a significant difference for the Mobin-Uddin device (2 x 7 contingency tables, 2, P < .05). The differences among other filters were not significant. The 3.2% (55 of 1,731) prevalence of caval thrombosis drops to 2.7% (46 of 1,703) if the nine cases of caval thrombosis related to the Mobin-Uddin filter, which is not in current use, are excluded. Caval thrombosis was confirmed by means of imaging in 37 cases or by means of clinical impression without imaging confirmation in 18 patients.
Thrombosis of the vena cava was associated with leg edema in 48 cases. It was asymptomatic in seven cases. In these seven cases, thrombosis was discovered during imaging examinations conducted for unrelated reasons. Caval thrombosis was observed more frequently in patients with neoplasms. The prevalence was 4.4% (40 of 915) among patients with neoplasms and 1.8% (15 of 816) among those without underlying neoplasms (P = .002, 2, goodness of fit). The prevalence of postfilter caval thrombosis was unrelated to the nature of the initial thromboembolic event.
Figure 10 shows the Kaplan-Meier time-to-event curve for the 55 patients with observed caval thrombosis after filter placement. The median time to event was 57 days (95% CI: 23, 91 days). Figure 11 shows the Kaplan-Meier event-free survival curve for all 1,731 patients, where the event is defined as postfilter caval thrombosis; the median could not be estimated because of the large number of patients free of known caval thrombosis at the conclusion of data collection.
|
|
2, goodness of fit). Figure 12 shows the Kaplan-Meier survival analysis for patients without and those with neoplasms. For patients without neoplasms, the median survival time was 2,064 days, (95% CI: 1,637, 2,490 days). For patients with neoplasms, the median survival time was 168 days (95% CI: 138, 198 days). The difference in survival was significant (log-rank test, P < .05) between the group with neoplasms and the group without neoplasms.
|
Imaging Follow-up Data
Data from direct imaging of the filter were available in 43.9% (760 of 1,731) of the patients. Direct imaging included conventional abdominal radiography, CT, or venacavography. The mean time between filter placement and the latest pertinent direct imaging examination was 512 days ± 956 (range, 0–5,898 days). In 45 patients, imaging revealed an abnormality relative to the filter or the vena cava. These findings were described earlier. In an additional 467 patients, chest radiographs were available and were used to confirm that the filter had not migrated to the chest. The mean time between filter placement and the latest chest radiograph was 467 days ± 881 (range, 1–5,739 days).
Autopsy Data
Autopsy was performed in 108 patients. The autopsy was limited to the brain in 12 patients. Thus, autopsy data pertinent to the lungs and the cava were available in 96 patients. The mean time from filter placement to autopsy was 137 days ± 415 (range, 0–3,409 days). PEs at various stages of organization and of varied sizes were found in 52 patients. In 12 of these patients, PEs were small and well organized. They were thought to be completely unrelated to the events leading to the patient's death. In the remaining 40 patients, it was the opinion of the pathologist that PE caused or contributed to the patient's death. These patients were included in the category of postfilter PE that was discussed earlier.
The cava and the filter were examined in 96 patients. Thrombus embedded in the filter was found in 20 patients, and caval thrombosis below the filter was observed in nine patients. In another nine patients, parts of the filters penetrated through the wall of the cava. No symptoms had been recorded related to this finding. In one of these nine patients, the prongs penetrating the wall had been embedded into the adjacent aorta. This patient died as a result of a cerebral vascular accident 934 days after filter insertion.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADDENDUMREFERENCES |
|---|
Thus, according to the data presented in this article, the prevalence of postfilter PE was 5.6%, 1.6%, or in-between. Acceptance of one or the other value depends on the criteria we set for the definition of postfilter PE. These findings are in general agreement with a 3.3% (11 of 324; 2.4% [eight of 324] fatal) rate of postfilter PE reported by Ferris et al (15) and a 4.4% (220 of 498 patients) rate reported by Greenfield and Proctor (17).
The rates of postfilter PE for each type of filter are listed in Table 6. The lowest failure rate was observed in patients with the Simon nitinol device. We note, however, that it would be unwise to attempt to rank devices in terms of efficacy. Without randomization and stratification for associated risk factors and the concomitant use of anticoagulants, comparisons are not feasible.
Complications
The overall prevalence of major complications was 0.3%. The prevalence of central migration was 0.1% (two of 1,731 patients). This is compared with a 1.2% (four of 320) prevalence of central (cephalic) filter migration reported by Ferris et al (15). We do not consider short caudal filter migrations to be major complications. Caudal migrations are clinically relevant only when the devices move to an iliac vein, in which instance they would not provide full protection. In seven cases of distal migration, a second filter was placed above the first.
Additional complications relate to the structural integrity of the metallic devices and perforation of the caval wall by the filter prongs. We observed a lower prevalence of filter fractures: 0.2% (four of 1,731) versus the 2.1% (five of 230 filters) prevalence reported by Ferris et al (15). This may be related to the fact that radiologic or pathologic follow-up information was available in 70.9% (227 of 320) of the patients reported on by Ferris et al (15) as compared with 43.9% in our series. Caval wall perforation was observed in one patient (0.1% of 1,731). It was not a source of clinical symptoms.
The prevalence of the observed postfilter caval thrombosis was 3.2% overall and 2.7% when patients with the Mobin-Uddin device were excluded. The prevalence of caval thrombosis according to each type of device is listed in Table 6. Historically, authors of major articles on inferior vena caval filters have referred to postfilter caval thrombosis as a complication of the device. This is because, by design, the first filter (Mobin-Uddin) impeded blood flow in the cava. Impedance of flow resulted in a high prevalence of caval thrombosis with serious clinical sequelae. Consequently, the Greenfield device was introduced as a filter that permitted continued flow and resolution of emboli (5). For the new device, as well as for others that followed, caval thrombosis was not and it should not be a major issue. Caval filters made of biocompatible metals are not thrombogenic in nature. In cases of postfilter caval thrombosis, it is more plausible that thrombosis is the result of entrapment of a major embolus by the filter rather than that the filter stimulates thrombus formation.
Survival
It has been suggested that in many cases the implantation of a caval filter is a preterminal event (30,31). In this regard, we found that 17.0% of the patients died within 30 days after filter placement and that the 30-day mortality rate was higher among patients with neoplasms. This observation is in agreement with a 30-day mortality rate of 18% (seven of 40 patients) reported by McLoughlin et al (30). Similarly, Rosen et al (31) found that death within 3 weeks after filter placement occurred in 26% (16 of 61) of patients with malignancies and in 46% (six of 13) of patients with malignancy and suprainguinal venous thrombus. A 30-day mortality rate of 17.0% is substantial. However, long-term survival analysis offers a different outlook. We found that the median survival of patients with cancer was 168 days, but for those without cancer median survival was 2,064 days.
Suprarenal Filters
A filter was deployed above or at the level of the renal veins in 2.6% of the patients. There were no migrations and no ill effects on renal function. The rate of postfilter PE in this group of patients was 6.5%, similar to that of patients with filters in the infrarenal cava. As far as the safety of suprarenal placement is concerned, our findings are in general agreement with those of Greenfield and Proctor (32) and Matchett et al (33). It may be of interest to note that a 2.6% rate of suprarenal placement in our series is comparable to a 2.9% (22 of 764) rate in the article by Matchett et al (33). However, both of these rates are lower than the 7.7% (148 of 1,932) rate of suprarenal placements listed in the article of Greenfield and Proctor (32).
Increasing Numbers of Filter Implantations: Changing Demographics and Expanding Indications
A relatively low rate of filter failure, a low rate of complications, and ease of insertion are likely explanations for the increasing application of caval filters. Our data show a continuous increase in the annual number of filter implants. Beginning in 1973, we implanted a few devices each year. The number of filter insertions increased during 1979–1980. This increase coincided with the introduction and adoption of the Greenfield filter (Figs 1, 7). A second substantial increase took place during 1984-1985, coincidental with the introduction of the percutaneous devices. There were 35 filter implantations in 1980. Subsequently, there was a twofold increase by 1990 and a fivefold increase by 1996 (Figs 1, 7). During the recent 5 years, we observed an increase in the number of outpatient visits. However, most patients receiving filters were inpatients. The number of hospital admissions has remained relatively stable over the years at approximately 33,000 per year. Therefore, the increase in the numbers of caval filters cannot be explained by increasing numbers of hospital admissions.
During the past 10 years (1989-1998), there has been a remarkable shift toward the application of caval filters in younger patients. Figure 2 illustrates the changing age trend on the basis of the most recent 4 years. The relative percentage of patients younger than 30 years of age has been rising. This should be viewed in the light of changing indications. The number of filters placed because of contraindications to anticoagulant therapy has decreased, while the percentage of filters placed for prophylaxis has increased. These changes are related. A more aggressive approach to the prevention of life-threatening PE in patients with serious multiple injuries is now the clinical doctrine. These patients tend to be young victims of automobile or motorcycle accidents. Extensive injuries, often including head trauma, render anticoagulation therapy unsafe. In addition, multiple fractures of the extremities render the performance of commonly used survey tests for the development of venous thrombosis difficult or impossible; hence the need for caval filters and therefore the rising numbers of filters placed for prophylaxis, mostly in younger patients.
Patients with injuries and the need for prophylaxis from venous thromboembolism would be ideal candidates for the placement of temporary caval filters. Temporary filters are devices that can be placed in the cava for a brief time, such as 5–15 days, for example. The aim would be to offer protection during a critical period at the end of which the devices could be retrieved and removed. Several designs of temporary retrievable filters are currently in different stages of development or of the approval process. Other patients who might benefit from the availability of retrievable filters are those who may need protection during and after surgery. We estimate that 27.0% of the patients included in this article would have qualified for temporary protection with retrievable filters. Included would be patients receiving a filter for prophylaxis and those who needed short-term protection perioperatively.
Additional Observations
Inferior venacavography performed prior to filter deployment provided information about anatomic variations and the presence of caval thrombosis. The main anatomic variants included double cava (0.2%), left cava (0.1%), circumaortic left renal vein (1.3%), retroaortic left renal vein (0.7%), and double renal veins (0.9%). A megacava (diameter larger than 30 mm) was observed in one patient. When this rare variation is observed, an appropriate device must be chosen. Currently, all devices except the Bird's Nest filter may be placed safely in a cava no larger than 28 mm. The Bird's Nest device may be implanted in a cava as large as 40 mm in diameter.
The number of the observed variations was lower than the number detected by means of MR imaging or spiral CT angiography. By means of MR angiography, Kaufman et al (34) found a double cava in 0.6% (one of 150), circumaortic left renal veins in 4.6% (seven of 150), and multiple renal veins in 8.0% (12 of 150) of patients. By means of spiral CT angiography, Trigaux et al (35) found a double cava in 0.3% (three of 1,014), circumaortic left renal veins in 6.3% (64 of 1,014), and retroaortic left renal veins in 3.7% (38 of 1,014) of patients. MR imaging and spiral CT cavography are more sensitive than catheter cavography in the detection of anatomic variants.
Selective renal venography may reveal even more variations. Hicks and colleagues (36) found an 11.1% (12 of 108) prevalence of anatomic variants by using catheter cavography, but the prevalence increased to 37.9% (41 of 108) of patients with the addition of selective renal venography. We believe that cavography prior to filter placement reveals the information necessary for safe filter deployment. Data from additional imaging studies offer diminishing returns. Inferior venacavography also may reveal the presence of thrombus in the cava, a finding that may play a role in choosing the precise site of filter deployment. In our series, the prevalence of isolated caval thrombosis was 2.5%.
There was no seasonal pattern observed in the requests for and the implantation of caval filters. There is a general perception that requests for pulmonary angiography, and consequently for filter placement, are more frequent on Fridays than on any other weekday. We have reported previously that for pulmonary angiography the perception is true that day-to-day differences are statistically significant (P < .001) for Friday (37). The data in this article show that although there was a certain predilection for filter placements on Fridays (18.9% of the procedures were performed on a Friday [Fig 6]), the difference was not significant.
We have presented data collected through an observational, retrospective review. Analysis of retrospective data may reveal possible associations, but it allows no reasonable conclusions about cause-and-effect relationships. In this study, we set the prevalence of the observed fatal postfilter PE as the primary outcome. We found differences in the proportions of the observed postfilter PE among the various devices. We reported the different proportions and refrained from performing statistical comparisons and generating P values. Such comparisons might be construed as an attempt to rank the devices in terms of their effectiveness. This was not our aim. Instead, we hope that the observed differences will stimulate interest in the performance of prospective randomized studies in the future.
On the basis of a 26-year single-center experience with 1,739 patients, we concluded that inferior vena caval filters provide protection from life-threatening PE, with minimal morbidity and few complications. Effectiveness, safety, and ease of insertion of currently available devices explain why the indications for filter placement are expanding and the number of filter implants is increasing.
|
ADDENDUM |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADDENDUMREFERENCES |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, C.A.A.; study concepts and design, C.A.A.; definition of intellectual content, C.A.A.; literature research, C.A.A.; clinical studies, C.A.A., J.A.K., A.C.W., S.C.G., C.M.F.; data acquisition, C.A.A., S.C.G., A.C.W., C.M.F.; data analysis, C.A.A., E.F.H., J.A.K.; statistical analysis, C.A.A., E.F.H.; manuscript preparation, C.A.A., A.C.W., S.C.G., C.M.F.; manuscript editing, C.A.A., J.A.K.; manuscript review, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADDENDUMREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. A. Binkert, M. C. Morash, and J. D. Gates Venographic Findings at Retrieval of Inferior Vena Cava Filters Am. J. Roentgenol., April 1, 2007; 188(4): 1039 - 1043. [Abstract] [Full Text] [PDF]
|
|
|
) or surgical caval interruption (
) over time.
= PE; 
= deep venous thrombosis; and 
= no event, filter for prophylaxis.
= conventional venography, and • = ventilation-perfusion lung scans.
