Symptomatic recurrent carotid stenosis and aneurysmal degeneration after endarterectomy.

BACKGROUND: Aneurysmal degeneration of a carotid reconstruction was not recognized until the patient, who was known to have recurrent carotid artery stenosis, had a thromboembolic stroke. This sequelae of carotid endarterectomy is a serious complication, associated with a high morbidity and mortality rate. This review was conducted to establish the risk of transient ischemic attack and stroke for patients found to have recurrent carotid stenosis associated with aneurysmal degeneration of the carotid artery after endarterectomy. METHODS: A case is reported, and 100 literature references of aneurysmal degeneration of the carotid artery after endarterectomy were reviewed. RESULTS: False aneurysm from anastomotic disruption was the most common presentation identified in the cases reviewed. Nineteen of the patients had a significant neurologic event; however, three (50%) of six patients with aneurysm and recurrent carotid artery stenosis had a transient ischemic attack or stroke. CONCLUSIONS: The incidence of neurologic symptoms is markedly increased when recurrent carotid artery stenosis is associated with carotid aneurysm. During postoperative surveillance after endarterectomy, the identification of recurrent carotid artery stenosis requires evaluation for aneurysmal degeneration of the carotid artery with duplex scanning. These patients are at significant risk for transient ischemic attack and stroke. This rare complication merits operative repair.

Operative strategies for management of abdominal aortic gunshot wounds.

BACKGROUND: Although management of penetrating abdominal trauma has greatly improved, abdominal aortic gunshot wounds (AAGSWs) remain a highly lethal injury. Our experience with AAGSWs was reviewed to define operative strategies that may improve survival. METHODS: Forty-one patients with AAGSWs were treated between 1976 and 1996. Preliminary thoractomy was performed in seven patients. Thirty-nine patients had at least one major associated injury (average, 3.2). RESULTS: Twenty-one patients died. Six of seven patients who underwent preliminary thoracotomy died; all developed coagulopathy, which appeared to contribute to death. Four patients had missed vascular lesions, two of which contributed to their death. Associated injuries are currently managed by "damage control" strategy, in which some injuries are left untreated to focus on hemorrhage control. CONCLUSIONS: We have identified seven operative principles and procedures that we believe may improve survival: (1) thorough knowledge of supraceliac exposure; (2) rapid aortic control at the hiatus rather than by a preliminary thoracotomy; (3) use damage control or abbreviated laparotomy; (4) use packing and mesh closure when coagulopathy and hypothermia are present; (5) primary concern should be cessation of hemorrhage rather than the maintenance of flow; (6) delayed reconstruction using extraanatomic bypass can restore flow; and (7) use angiography to detect missed vascular lesions or problems with vascular repair.

Hemodynamics of in situ saphenous vein arterial bypass.

Doppler-derived blood flow velocity and limb blood pressure measurements were used to characterize the hemodynamics of 128 in situ saphenous vein arterial bypasses. The magnitude and configuration of the graft velocity waveform was the best predictor of clinical outcome. Successful bypasses had antegrade flow throughout the pulse cycle and a blood flow velocity above 40 cm/s. A low graft blood flow velocity (less than 40 cm/s) was associated with technical error or early graft failure due to poor runoff. The return of normal limb blood pressure correlated with a technically satisfactory bypass but was measured in only 50% of limbs on the first day after surgery. In 28 bypasses with high blood flow velocity (mean +/- SD) in systole (102 +/- 20 cm/s) and diastole (35 +/- 11 cm/s), postoperative limb blood pressure was initially low (mean ankle-brachial pressure index = 0.68) due to restriction of blood flow through small-diameter (less than 4-mm) venous conduits. As revascularization hyperemia abated, diastolic blood flow velocity decreased and limb blood pressure normalized. Despite the high blood flow velocity and pressure gradient associated with flow-restrictive venous conduits, limb ischemic symptoms resolved, and graft patency was not decreased. An understanding of graft and limb hemodynamics after in situ bypass grafting is critical when noninvasive diagnostic techniques are used to document technical adequacy and for postoperative surveillance.

Biochemical adaptation of venous patches placed in the carotid circulation.

The influence of venous patching on luminal prostacyclin (PGI2) and thromboxane A2 (TBX) production of endarterectomized arteries was studied in a canine model. Fifteen dogs underwent bilateral common carotid endarterectomies. In each dog, one artery was closed primarily and the contralateral artery was closed by vein-patch angioplasty. At six and 12 weeks after operation, luminal prostanoid production from the venous patch, adjacent endarterectomized artery, and control artery and vein was measured by radioimmunoassay for 6-keto-prostaglandin F1 alpha and thromboxane B2. Venous patches underwent "incomplete" biochemical adaptation with decreased luminal production of PGI2 compared with control and endarterectomized arteries. Thromboxane production from venous patches was increased compared with normal vein but similar to control and endarterectomized arteries. At six weeks, PGI2 production of vein-patched arterial segments was increased compared with arteries closed primarily. These data indicate that vein-patch angioplasty favorably influences surface thrombogenicity of healing, endarterectomized arteries and support its use in reparative carotid surgery.

Monitoring functional patency of percutaneous transluminal angioplasty.

Duplex scanning and Doppler-derived blood pressure measurements were used to serially monitor lower limb hemodynamics in 73 patients who underwent percutaneous transluminal angioplasty. Ninety percutaneous transluminal angioplasty sites judged technically satisfactory by arteriography were evaluated. Significant hemodynamic improvement was seen in 81 (90%) of the 90 limbs, although both hemodynamic and clinical improvement were achieved in only 77 (86%) limbs. Duplex scanning within 1 week of successful angioplasty identified moderate (20% to 49% diameter reduction) or severe (greater than 50% diameter reduction) residual stenosis in 49 (63%) of 77 balloon-dilated arterial segments. The presence of a greater than 50% diameter reduction residual stenosis predicted further restenosis and late clinical failure (11% success rate at 1 year). When the degree of residual stenosis at the percutaneous transluminal angioplasty site was less than 50% diameter reduction by duplex scanning, the procedure was durable (80% success rate at 2 years), even in patients with critical ischemia, poor runoff, or diabetes mellitus.

Anastomotic tensile strength following in situ replacement of an infected abdominal aortic graft.

The tensile strength and histologic features of anastomotic bonding were studied prior to and following in situ replacement of aortic vascular prostheses infected by Staphylococcus epidermidis. Sterile (n = 6) and infected (n = 19) Dacron grafts were used to replace the abdominal aorta of 25 dogs. After five weeks, grafts were explanted, and peak tensile force (measured in kilograms) required for anastomotic disruption was measured using a linear gain tensiometer. Anastomotic tensile strength (mean +/- SEM) of infected grafts (5.4 +/- 0.5 kg) was decreased when compared with that of sterile, control grafts (9.0 +/- 0.9 kg). The decreased anastomotic tensile strength of infected grafts was the result of an inflammatory aortitis adjacent to the suture line. Only grafts infected with the study strain of bacteria demonstrated signs of infection. In 19 dogs, the graft infection was treated by graft excision, antibiotic administration, and in situ graft replacement (Dacron or polytetrafluoroethylene prostheses). After five weeks and 12 weeks, anastomotic tensile strength of polytetrafluoroethylene (10.6 +/- 0.6 kg) and Dacron (10.8 +/- 0.5 kg) replacement grafts was similar to that of uninfected control grafts. In situ replacement of vascular prostheses infected by S epidermidis can result in graft healing with normal anastomotic bonding.

Staphylococcus epidermidis graft infection is associated with locally suppressed major histocompatibility complex class II and elevated MAC-1 expression.

OBJECTIVES: To determine the local cellular immune response in a series of human patients with Staphylococcus epidermidis prosthetic graft infection and to use a murine model to investigate the response in polytef (PTFE) and in a nonslime-producing S epidermidis variant. METHODS: Externally supported Dacron and PTFE grafts, either sterile or colonized with slime (RP-62A)- or nonslime (RP-62NA)-producing S epidermidis (10(7) colony forming units/cm2) were implanted in a dorsal subcutaneous pocket of Swiss Webster mice (Taconic, Germantown, NY). The grafts were harvested at 7, 10, 14, and 28 days with local bacterial and leukocyte counts obtained. Perigraft and blood monocyte major histocompatibility complex class II (MHC-II) (immune antigen) and membrane attack complex type 1 (MAC-1) (glycoprotein) expression were analyzed by flow cytometry in the murine model and in 3 patients representing 4 Dacron graft infections. RESULTS: The human infected Dacron perigraft monocytes revealed a suppressed MHC-II and elevated MAC-1 expression, and early correlation with the murine model was seen. No notable perigraft monocyte MHC-II suppression occurred in the infected PTFE graft. The reciprocal relationship in Dacron between monocyte MAC-1 and MHC-II expression was exaggerated with the lack of slime production. Bacterial clearance was variable. Supranormal expression was observed at 1 month in sterile Dacron but not in PTFE grafts. CONCLUSIONS: Staphylococcus epidermidis infection is associated with local cellular immune suppression in Dacron but not PTFE grafts. Slime-producing S epidermidis induced a lesser cytotoxic-phagocytic response than the nonslime variant. The reduced immunologic response to slime-producing S epidermidis may explain, in part, its indolent nature and resistance to eradication.

Immunosuppression augments growth of graft-adherent Staphylococcus epidermidis.

OBJECTIVE: To determine if systemic suppression of host defenses during graft implantation alters the initial adherence and subsequent growth of Staphylococcus epidermidis on vascular prostheses. DESIGN: Dacron grafts 1 cm2 were implanted in the back subcutaneous tissue of Swiss-Webster mice (n = 247), followed by topical inoculation with 2 x 10(7), 2 x 10(5), 2 x 10(3), or 2 x 10(1) colony-forming units of S epidermidis. Half of the mice were immunosuppressed with cyclophosphamide (150 mg/kg intraperitoneally), to achieve a consistent, significant decrease in the white blood cell count and major histocompatibility complex class II (Ia) expression. Control mice received an equal volume of saline solution. Graft bacterial biofilm concentrations were determined at 1 day for adherence and within 2 weeks for bacterial growth, by using sonication and quantitative agar culture. RESULTS: Immunosuppression did not significantly alter the initial adherence of bacteria to vascular grafts. Immunosuppressed animals that were inoculated with 2 x 10(7) and 2 x 10(5) colony-forming units of S epidermidis had significantly higher bacterial biofilm concentrations as compared with those in control animals. Graft infection persisted at 14 days in all animals, with and without immunosuppression. CONCLUSIONS: Suppression of immune function during graft implantation augmented growth of adherent bacteria. The effect of short-term perioperative immunosuppression on late-appearing S epidermidis graft infection needs further study.

Effect of multilevel sequential stenosis on lower extremity arterial duplex scanning.

BACKGROUND: The sensitivity of lower extremity arterial duplex scanning in detecting a > 50% stenosis may be decreased in extremities with multilevel sequential stenosis. PATIENTS AND METHODS: The results of lower extremity arterial duplex scanning of the common femoral, upper and lower superficial femoral, above- and below-knee popliteal arteries, and tibioperoneal trunk in 80 extremities of 44 patients were compared to those of arteriography. Thirty-one arterial segments in 27 extremities had a > 50% stenosis. The stenoses were categorized as first order (first or only stenotic segment in the extremity, n = 23) and second order (stenosis occurring distal to a > 50% stenosis, n = 4, or occlusion, n = 4). RESULTS: The sensitivity, specificity, and positive predictive values of duplex scan detection of a > 50% arterial stenosis or occlusion were 86%, 96%, and 67% for the common femoral artery; 95%, 98%, and 95% for the upper superficial femoral artery; 97%, 90%, and 88%, for the lower superficial femoral artery; 84%, 90%, and 87% for the above-knee popliteal artery; 47%, 98%, and 90% for the below-knee popliteal artery; and 25%, 100%, and 100% for the tibioperoneal trunk. Duplex scanning detected 18 (78%) of the 23 first-order stenoses compared to only 1 (13%) of the 8 second-order stenoses in limbs with multilevel sequential disease (P < 0.01). The peak systolic velocity at the stenotic site was significantly higher for first-order (mean +/- SD 168 +/- 54 cm/s) compared to second-order (38 +/- 13 cm/s) stenoses (P < 0.00002). CONCLUSIONS: Duplex scanning was highly sensitive in detecting lower extremity first-order stenoses. Low peak systolic velocities at second-order stenoses of limbs with multilevel sequential disease significantly decreased the sensitivity of duplex scanning.

Controversies in the management of pancreatic pseudocysts.

Review of the diagnosis and treatment of patients with pancreatic pseudocysts over the past 8 years has led us to three conclusions regarding controversial aspects of their treatment. We found that patients who present with chronic pseudocysts can be identified with the help of computerized axial tomography and promptly undergo successful internal drainage, whereas patients with acute peripancreatic fluid secondary to pancreatitis can be observed expectantly with a 43 percent frequency of spontaneous resolution. Patients with infected pancreatic pseudocysts can be safely drained internally. The most common cause of extrahepatic biliary obstruction in this group of patients with pancreatic pseudocysts was stricture due to pancreatitis and fibrosis, not extrinsic compression.

Combined topical and systemic antibiotic prophylaxis in experimental wound infection.

The effectiveness of a combined topical and systemic antibiotic regimen was studied in an animal model previously shown to simulate clinical surgical wound infection. At a high level of bacterial contamination, the combination regimen produced a lower infection rate than either a placebo (p less than 0.01), a topical antibiotic administered alone (p less than 0.01), or a systemic antibiotic administered alone. At a lower level of bacterial contamination, no additional reduction in infection rates was produced by the combination regimen when compared with systemic antibiotic administered alone. These experimental results suggest that when wound contamination is great, a combination of topical and systemic antibiotics is the more effective regimen. Where wound contamination is less severe, systemic antibiotic prophylaxis is all that is required; no further benefit is obtained by the additional administration of topical antibiotics. Clinical trials appear justified to confirm or refute this hypothesis.

Carotid endarterectomy results from a state vascular society. Kentucky Vascular Surgery Society Study Group.

PURPOSE: This study analyzes the results of carotid endarterectomy (CEA) performed statewide by members of the Kentucky Vascular Surgery Society (KVSS). METHODS: Between September 1, 1991 and September 1, 1993, 22 vascular surgeons in the KVSS submitted 1490 CEAs to the vascular registry. Follow-up data were obtained on 986 (66%) CEAs performed on 889 patients (average age, 68 years). RESULTS: Carotid endarterectomy was performed on 505 men and 384 women. Indications for operation were asymptomatic carotid stenosis (43%), transient ischemic attack (TIA; 27%), amaurosis fugax (13%), stroke (11%) and nonhemispheric symptoms (6%). A total of 384 cases had primary closure, and 602 had patch reconstruction following CEA. The combined stroke-mortality rate was 2.3% (10 strokes and 13 deaths). The combined stroke-mortality rate of CEA was 2.1% for patients treated by academic surgeons and 2.3% for those treated by community surgeons. Deaths were due to stroke (4), sepsis (5), cardiac complications (2), intracerebral hemorrhage (1) and cancer (1). Five patients had postoperative TIAs. After CEA, duplex scan surveillance was performed in 629 (64%) patients, with 23 (3.6%) residual/recurrent stenosis (10, 50%-75%; 13, 75%-99%) detected; 5 undergoing reoperation. CONCLUSIONS: These data support the efficacy and safety of CEA performed by a large number of vascular surgeons in both community and academic practice.

Long-term venous access: indications and choice of site and catheter.

Advances in treatment of malignancy, infection, renal failure, and malnutrition have expanded the need for long-term venous access. The placement of central venous catheters (CVC) requires preoperative evaluation to identify associated risk factors such as coagulopathy, thrombocytopenia, leukopenia, and variations in the vascular anatomy. When selecting a venous access catheter, it is important to consider its purpose, duration, and who will maintain it. The technique of inserting a catheter into the central venous system has become standardized by Seldinger. Catheters now are prepackaged in kits equipped for percutaneous placement. Knowledge of the technical aspects of placing a CVC can prevent early occlusion and dislodgment. With the increasing numbers of CVCs being inserted, it is possible to overuse the common sites for placement. We describe techniques that have been developed to use alternate routes for venous access.

Descending thoracic aortobifemoral bypass: an alternative approach for difficult aortic revascularization.

Descending thoracic aortobifemoral bypass is an alternative inflow operation in cases in which standard aortobifemoral or axillobifemoral bypass is not an option. We performed descending thoracic aortobifemoral bypass for failed inflow operations in four patients, prior abdominal/pelvic radiation in two patients, poor quality distal aorta (extensive atherosclerotic disease or poor tissue quality) in two patients, and abdominal sepsis in two patients. Eight have had excellent results with patency at a mean follow-up of 38 months. There was no limb loss. One patient died of organ failure, and one patient with hypercoagulability developed a graft clot. A literature review disclosed that a descending thoracic aortobifemoral bypass was reported to have been performed in 203 patients, with most cases reported in the last decade. This procedure was the primary inflow operation in 42 per cent of cases. Indications for the operation included failed aortic grafts (38%), "hostile" abdomen (21%), infected aortic grafts (18%), and other (23%). The patency rate was 95 per cent at 6 months. Few long-term results are known, but the results appear to be durable. Descending thoracic aortobifemoral bypass is a useful operation in highly selected circumstances in which conventional methods of aortic reconstruction are not available.

Liver abscess. Review of a 12-year experience.

Experience with liver abscess was reviewed to determine whether new trends in treatment have altered clinical outcome. Thirty-one cases of liver abscess were identified from 1973 to 1985. In this group there were ten deaths for a mortality rate of 32 per cent; the primary disease accounted for seven deaths. Significant predictors of liver abscess mortality were multiple abscesses, elevated bilirubin levels, and underlying disease. Percutaneous drainage was effective in three of four patients and should be attempted before operative intervention in selected patients.

Blunt popliteal artery trauma: a challenging injury.

Blunt popliteal artery trauma is a challenging injury, particularly when associated with major soft tissue damage. We reviewed our experience with this injury to determine 1) the incidence of vascular injury associated with fractures and/or dislocations about the knee, 2) the incidence of limb loss, and 3) factors associated with amputation. We treated 37 patients with 38 blunt popliteal artery injuries and either fractures about the knee or posterior knee dislocations. Patients who underwent primary amputations were excluded. The incidence of popliteal artery injuries with fractures about the knee was 3 per cent, whereas 16 per cent of patients with posterior knee dislocations had vascular injuries (P < 0.05). Amputations were required in 14 of the 38 injured limbs (36%). None of these patients had a pulse or Doppler signal on admission, and 13 had major soft tissue injury. No patient with a pulse or Doppler signal lost a limb (P < 0.05). Limb loss was primarily related to limited venous outflow and/or severe infection in damaged tissue. Failure of the arterial repair rarely led to amputation, particularly in recent years. Two patients with angiographically proven arterial injuries were treated nonoperatively without complications. The incidence of vascular injuries associated with fractures about the knee is low, but somewhat higher with posterior knee dislocations. The overall 9 per cent rate of positive angiograms suggests that a selective approach may be indicated. The amputation rate remains high, but it has improved with an integrated, multidisciplinary team approach. In patients without a pulse or Doppler signal and with severe soft tissue injuries, primary amputation may be appropriate.

Intraoperative duplex scanning reduces both residual stenosis and postoperative morbidity of carotid endarterectomy.

To evaluate the effect of intraoperative duplex scanning (IDS) on the incidence of perioperative and postoperative strokes as well as residual and recurrent stenosis, we reviewed 141 patients who underwent 152 consecutive carotid endarterectomies (CEAs) between July, 1990 and June, 1995. Follow-up of 129 cases, with a mean follow-up of two years, revealed no perioperative deaths and three strokes for a combined perioperative stroke-death rate of 2.3 per cent. In 50% (64 of 129) of the CEAs, intraoperative duplex scans were obtained based on the attending surgeon's preference. We noted that the incidence of residual stenosis (>50% stenosis on the first duplex after CEA) was significantly lower in those undergoing IDS (3/64) versus those without IDS (13/65) (P < 0.05; risk ratio 0.31; 95% confidence interval 0.11, 0.91). IDS resulted in a modification of the internal carotid reconstruction in 9 per cent (6 of 64) of the cases with no resulting postoperative strokes or residual/recurrent stenosis. There was no significant difference in the frequency of recurrent stenosis (>50% stenosis after a normal duplex) in the two groups (3 of 64 with vs 2 of 65 without). Of patients not undergoing intraoperative scanning, four underwent redo CEA for symptomatic residual stenosis due to a retained intimal flap in the internal carotid artery. There were three strokes observed within 30 days of the initial CEA, all of which occurred in patients who did not undergo IDS at their initial operation. We conclude that IDS can identify technical defects following internal carotid reconstruction, thereby reducing the incidence of both residual stenosis and postoperative morbidity in patients undergoing CEA.

Surgical implications of hypercoagulable syndromes.

The advent of improved diagnostic tests for primary hypercoagulability has led to increased recognition of this entity as a problem in surgical patients. We treated 20 patients with documented evidence of increased coagulability from 1975 to 1995. Clinical presentations included venous (16) and arterial (4) thrombosis. Symptoms usually occurred early in life (mean age, 38 years) and developed spontaneously without a secondary inciting event or factor. Deficiencies in naturally occurring anticoagulant proteins including antithrombin III (n = 7), protein C (n = 3), and protein S (n = 1) were seen, as were problems with lupus anticoagulant (n = 2) and anticardiolipin antibody (n = 4) deficiencies. Treatment of these patients is difficult, and results are often suboptimal. A total of 12 vascular reconstructions were required in 5 of the 20 patients; 11 eventually failed. Patients with primary venous thrombosis were often successfully treated with anticoagulant therapy in the short term but fared less well in the long term. There were three deaths directly related to thrombotic complications. Surgeons may encounter patients with primary hypercoagulable syndromes. The diagnosis should be expected in patients with unusual patterns of vascular disease or arterial or venous thrombosis without cause or at an early age, or in patients with recurrent or migratory clotting. Evaluation of this population, although expensive, is indicated. Treatment with chronic anticoagulation is also generally indicated. Arterial reconstruction in this subset of patients usually leads to a poor outcome.

Quality assessment of intraoperative blood salvage and autotransfusion.

Intraoperative blood salvage and autotransfusion are commonly used to minimize exposure to banked blood. Although this technique has been used widely for years, data vary regarding the quality of autotransfused blood. Salvaged blood may contain plasma, residual heparin, and free hemoglobin released from damaged cells. All of these factors may contribute to the adverse sequelae sometimes seen with autotransfusion. For these reasons, we have monitored autotransfused blood to assess its quality. Intraoperative blood salvage was used during most cardiac procedures and at the discretion of the surgeon in other specialties. Blood was collected through a double lumen catheter that was anticoagulated with heparin, filtered, centrifuged, and washed with saline. A sample of the blood was removed for analysis, which included hematocrit, heparin assay, fibrinogen, and free hemoglobin levels. Over a 6-year period, 1593 patients had intraoperative blood salvage with quality assessment. The majority of patients underwent cardiac operations (941 patients, 59%), whereas 243 had orthopedic (15%) and 208 had vascular (13%) procedures. Additionally, there were 127 pediatric patients (8%) and 74 miscellaneous procedures (5%). The highest average yield of salvaged blood was during vascular procedures (1073 +/- 76 mL), whereas orthopedic cases had the lowest yield (378 +/- 19 mL) and hematocrit (39%). There was minimal residual heparin activity, even in patients requiring systemic anticoagulation (0.3 to 0.5 units/mL). Patients undergoing pediatric procedures had the lowest concentration of free hemoglobin (476 mg/L), whereas all adult patients had higher free hemoglobin levels, especially vascular operations (990 mg/L). Intraoperative salvaged blood has minimal heparin activity, even in procedures requiring systemic anticoagulation. Fibrinogen, a marker of residual plasma, was undetectable in the majority of cases. These data indicate that intraoperative blood salvage generally results in a high-quality product (good hematocrit, low heparin, minimal plasma), although there are significant differences in free hemoglobin levels depending on the operative procedure.

Current options in prosthetic vascular graft infection.

Debate continues over which procedure is the best treatment for prosthetic graft infections. We retrospectively reviewed the medical records at our institution for all vascular graft infections that occurred from 1985 to 1995 to evaluate their occurrence, treatment, and outcome. Twenty-four patients had prosthetic graft infections. The average patient age was 62 years, and 67 per cent of the patients studied were men. The initial operation was for treatment of occlusive disease in 92 per cent of the patients, and aortofemoral bypasses were the most common procedures performed (15 of 24 patients, 63%). The average interval from graft implantation to presentation of infection was 29 months. In lower-extremity bypasses, the site of infection was most commonly in the groin (87%). Gram-positive organisms, including coagulase-negative Staphylococcus (32%) and Staphylococcus aureus (28%), were the most frequently isolated bacteria. Thirty procedures were performed for management of the graft infections. Extra-anatomic bypass was associated with no recurrent graft infections. Graft preservation was successful in two cases of early S. aureus infection (less than 1 year after original procedure), and in situ graft replacement was successful in all four cases of late-appearing coagulase-negative Staphylococcus infection (more than 1 year after original procedure). Both treatments failed in all five cases of Gram-negative infection (P = 0.008 by Fisher's exact test). The overall mortality and amputation rates were 17 per cent and 21 per cent, respectively, without significant differences between the treatment modalities. Extra-anatomic bypass remains the best treatment for prosthetic graft infection. In situ replacement and graft preservation treatments should be selective and based on presentation of the infection and the type of pathogenic organism.