Description of studies
We found five prospective randomised controlled studies comparing initial thrombolysis with initial surgery in the management of acute limb ischaemia (Nilsson 1992; Ouriel 1994; Ouriel 1996; Ouriel 1998a; STILE 1994). Major differences in participant demographics, including severity of ischaemia, site of occlusion, prosthetic or native vessel, thrombolytic regime and agent, requires caution in interpreting meta-analysis of the data. We have also included two further publications from the STILE trial looking at native vessel occlusions (Weaver 1996), and bypass graft occlusions (Comerota 1996) and one further publication from the Ouriel 1998a trial (Ouriel 1998b).
Follow-up was variable, from one month (Nilsson 1992) to one year (Comerota 1996; Ouriel 1994; Ouriel 1996; Ouriel 1998a; STILE 1994; Weaver 1996). See the Characteristics of included studies for further details.
Note: The STILE trial (Comerota 1996; STILE 1994; Weaver 1996) was stopped early because of an excess of complications in the thrombolysis group.
There were two studies which were excluded (Patel 1999; Tiek 2009), one of which was added in the 2013 update (Tiek 2009).
Risk of bias in included studies
The Nilsson trial (Nilsson 1992) was awarded a B grade for methodological quality, but all the other trials (Ouriel 1994; Ouriel 1996; Ouriel 1998a; STILE 1994), and additional papers from those trials (Comerota 1996; Ouriel 1998b; Weaver 1996) were awarded A grades.
All studies were randomised using telephone computerised randomisation with the exception of Nilsson 1992 (not actually stated). Due to the nature of the study, neither participants nor observers were blinded.
All studies were reported as intention-to-treat (ITT). The STILE trial (Comerota 1996; STILE 1994; Weaver 1996) also reported as per-protocol. Intention-to-treat includes analysis of all participants according to the treatment allocated by the randomisation procedure, regardless of whether they actually received that particular treatment. Per-protocol analysis was performed on the two groups of participants that actually received the two different treatments.
Duration of ischaemia requiring intervention was less than 14 days in Nilsson 1992, Ouriel 1996 and Ouriel 1998a. However, Ouriel 1994 used less than seven days of ischaemia as criteria for intervention. The STILE trial (Comerota 1996; STILE 1994; Weaver 1996), admitted participants with "signs or symptoms of worsening limb ischaemia within the past six months who required intervention" or who had a bypass graft occlusion.
Effects of interventions
The concept of a reduced level of procedure required was used in the STILE and Ouriel trials. This consists of a list of interventional procedures in order of increasing severity of intervention. Hence, if thrombolytic treatment were successful, it could be argued that the reduced need for further intervention would be less hazardous to the participant.
In this study, a total of 20 participants were randomised to thrombolysis (n = 11) with recombinant tissue plasminogen activator (rt-PA) infused over a three hour period, or to surgical thrombectomy (n = 9). There were no strokes or major haemorrhages with either thrombolysis or surgery at 30 days. Furthermore, there were no deaths or amputations in the thrombolytic group at 30 days, though there was one death and one amputation in the surgery group. Only four participants had successful thrombolysis with continued patency at one month. Three participants required a femoro-distal graft and two required thromboembolectomy. However, such small numbers hinder any firm conclusions from being drawn from this study.
In this trial, 114 participants were randomised to surgery (revascularisation or amputation, as required; n = 57) or to thrombolysis with urokinase (n = 57). At the 12 month follow-up (Ouriel 1994) there was a significant survival advantage of initial thrombolysis (84%) compared to initial surgery (58%; P = 0.01) (Peto odds ratio (OR) 0.28; 95% CI 0.13 to 0.63). This was thought to be due principally to the difference in in-hospital cardio-pulmonary complications between the two groups (thrombolysis group 16%; surgery group 41%; P = 0.001). There was no difference in limb salvage. At 30 days the stroke rate was 1.8% (n = 1) and major haemorrhage rate was 10.5% (n = 6) in the thrombolysis group, compared to no strokes (n = 0) and a rate of major haemorrhage of 1.8% (n = 1) in the surgical group. Thirty-six percent of participants in the thrombolytic therapy group had continued patency at 12 months.
The STILE trial (STILE 1994) recruited 393 participants from 31 centres (follow up information was available for 392 participants) to compare surgery against thrombolysis in the initial treatment of limb ischaemia. It should be noted that two different thrombolytic regimes were used (see the Characteristics of included studies table). The trial included participants with a duration of ischaemia of up to six months. It was originally anticipated that interim analyses would be performed at 300 and 600 participants. However, the data monitoring committee stopped the trial at 393 participants due to a higher incidence of adverse events (Composite Clinical Outcome P = 0.011) in the thrombolytic group at one month. This was similar in both native artery and bypass graft subgroups at one month.
Composite Clinical Outcome was a combination of outcomes including: ongoing or recurrent ischaemia, death or major amputation (above or below-knee), life-threatening haemorrhage (hypotension, requiring resuscitation), or stroke, peri-intervention complications such as myocardial infarction, pulmonary oedema, renal failure requiring dialysis, serious anaesthetic complications, vascular complications and post-operative wound complications.
There was a significantly reduced level of the procedure required in the thrombolysis group compared to the surgical group (55.8% versus 5.5%; P = 0.001). At six months, those in the thrombolysis group with less than 14 days ischaemia had a significantly reduced rate of death or amputation, or both (15.3% versus 37.5%; P = 0.01) (OR 0.29; 95% CI 0.12 to 0.72). For those with more than 14 days duration, a reversed trend was seen (death/amputation rate 9.9% in the surgical group versus 17.8% in the thrombolysis group), a difference which failed to reach significance (P = 0.08). Stroke occurred in 1.2% of participants receiving thrombolysis, while major haemorrhage occurred in 5.6%. In the surgical group haemorrhage occurred in 0.7%. Catheter placement failed in 28% of participants randomised to receive thrombolysis - a high technical failure rate, which potentially limits its widespread application. However, the study group may be unrepresentative.
Subsequent subgroup analysis by the trialists of those participants less than 14 days and those more than 14 days duration of ischaemia suggested alternative interpretations, although there had been no initial stratification in the randomisation protocol to allow for this. Participants with more than 14 days duration of history fared better (Composite Clinical Outcome) with surgery than with initial thrombolysis (P < 0.001). For participants with less than 14 days duration of ischaemia there was no difference in overall Composite Clinical Outcome (P = 0.439), but there were more amputations in the surgical group (17.9% versus 5.7%; P = 0.061).
The 12 month follow-up data on bypass grafts were reported by Comerota (Comerota 1996), and the native artery results by Weaver (Weaver 1996).
Further subgroup analysis resulted in low numbers of participants in each sub-group. For example, analysis by duration of history less than 14 days reduced numbers to 35 in the thrombolysis group and 23 in the surgical group. A similar advantage of a reduced amputation rate demonstrated in the original STILE study was again seen in the less than 14 day thrombolysis group. Overall however, thrombolysis was associated with a higher level of continued ischaemia, claudication or critical limb ischaemia (73% versus 50%; P = 0.01) (OR 2.72; 95% CI 1.2 to 5.80). Those participants with ischaemia for more than 14 days were associated with a significantly increased risk of ongoing or recurrent ischaemia. Significantly increased morbidity was encountered with prosthetic grafts compared to autogenous grafts (P = 0.038). While no difference was found between supra-inguinal and infra-inguinal grafts, it should be noted that there were only 22 (38%) grafts in the supra-inguinal group.
The native artery 12 month results confirmed a reduction in the surgical procedure required in approximately 50% of those participants randomised to thrombolysis (i.e. 50% of those participants receiving thrombolysis were able to undergo a more limited procedure than would have been expected prior to thrombolysis). However, at one year both recurrent ischaemia and major amputation were significantly higher in the thrombolysis group (recurrent ischaemia 64% versus 35%; P < 0.0001) (OR 3.26; 95% CI 1.92 to 5.52), major amputation 10% versus 0% (P = 0.0024). Stratification looking at ilio-femoral/femoro-popliteal and less than 14 days or more than 14 days duration of history all served to reduce the numbers in each group to as low as n = 4 in one instance.
The STILE trial papers (Comerota 1996; STILE 1994; Weaver 1996) all reported ITT but also per protocol analyses. There was a high percentage of participants classified as 'thrombolysis', who never actually received thrombolysis, after failure to site the catheter.
One of the major criticisms of the STILE publications was that the less than 14 days or more than 14 days analysis was a post-hoc arbitrary division, not a stratified part of the original protocol. Over 80% of all participants had a more than 14 days duration of history.
This paper (Ouriel 1996) reports the early experience of the TOPAS group (The Study of Thrombolysis Or Peripheral Arterial Surgery). Seventy-nine participating centres enrolled 213 participants over approximately six months. Participants had Class II ischaemia for less than 14 days (Reporting Standards). Three different urokinase thrombolysis regimes were used - 2000 IU/min, 4000 IU/min or 6000 IU/min for the first two hours, after which all groups received 2000 IU/min. The comparison between surgery and thrombolysis was performed with the optimal thrombolysis group (i.e. 4000 IU/min). No overall thrombolysis figures were available. There was no significant difference in the one year mortality rate or amputation-free survival rate between thrombolysis with 4000 IU/min urokinase and the surgical group. All thrombolytic groups had a significant reduction in open surgery compared to the surgery group at 30 days (P = 0.01).
In this subsequent trial, the 'optimal dose' of urokinase was used (4000 IU/min) as described for Ouriel 1996 above. Five hundred and forty-eight participants from 113 centres throughout the USA and Northern Europe were recruited over a period of 17 months. Although the randomisation was done by telephone, there were significant differences between the treatment groups. There were significantly higher numbers of men, increased hepatic or renal insufficiency, or both, and increased rest pain at presentation in the thrombolysis group. Death at one year was similar between the two groups (thrombolysis = 20%; surgery = 17%) and there was no significant difference in limb salvage (thrombolysis = 65%; surgery = 69.9%). At 30 days four (1.6%) strokes had occurred in the thrombolytic group, but none in the surgery group. Major haemorrhage was significantly higher in the thrombolysis group at 30 days (12.5% versus 5.5%; P = 0.005) (OR 2.35; 95% CI 1.28 to 4.29). Open surgical procedures were significantly lower in the thrombolysis group compared to the surgical group at six months (54% versus 91% respectively; P < 0.001).
Using the TOPAS database, the authors found in a univariate analysis that amputation-free survival was associated with race (69.4% white versus 60.0% non-white); age (75.1% less than 65 years versus 60.2% greater than 65 years); body weight (60.5% less than 160 lb versus 73.5% greater than 160 lb); history of central nervous system disease (70.6% history versus 51.15 with no history); history of congestive heart failure (70.7% history versus 46.1% with no history); mottled or cyanotic skin (72.7% mottled or cyanotic versus 58.3% normal) and presence of rest pain (80.1% rest pain versus 62.9% no rest pain). All these were significant at the 5% level.
Meta-analysis of the included trials showed no significant difference in limb salvage or death at 30 days, six months or one year between initial surgery and initial thrombolysis. Stroke was significantly more likely to occur with initial thrombolysis at 30 days. There were eight strokes in the 640 participants receiving thrombolysis, compared to none in the 540 participants receiving initial surgery (OR 6.41; 95% CI 1.57 to 26.22). Major haemorrhage at 30 days was also more likely within the thrombolysis group, with 52/588 occurrences versus 16/482 occurrences in the surgery group (OR 2.80; 95% CI 1.70 to 4.60). Distal embolization occurred in 42/340 participants receiving thrombolysis, with no reported occurrences in 338 participants undergoing initial surgery (OR 8.35; 95% CI 4.47 to 15.58).
As mentioned above concerning the STILE and Ouriel trials participants treated by initial thrombolysis may undergo a less severe degree of intervention (OR 5.37; 95% CI 3.99 to 7.22), and can expect an equivalent overall survival compared to initial surgery (OR 0.87; 95% CI 0.61 to 1.25). However, they may have a higher incidence of recurrent or ongoing limb ischaemia, major haemorrhage or stroke.