Interposition vein cuff for infragenicular prosthetic bypass graft

Types of studies

Randomised controlled trials of interposition vein cuffs for prosthetic infragenicular bypasses are included. Adequacy of randomisation was independently assessed by two authors (AK and GG). Blinding was not considered as it is impossible to achieve during surgery and at follow up.

Types of participants

All patients who underwent a femoro‐distal bypass (femoral to below knee popliteal bypass, femoro‐tibial, femoro‐peroneal and femoro‐pedal) for critical limb ischaemia. Patients were included whatever the cause of their ischaemia and regardless of their diabetes control, smoking or use of aspirin or anticoagulants in the follow‐up period.

Types of interventions

Bypass procedures to below knee vessels using a prosthetic graft with a vein cuff at the distal anastomotic site were considered and compared with prosthetic bypasses without cuffs, autologous bypasses and pre‐cuffed hooded prosthetic grafts. Modification to the technique of Miller’s cuff was included when authors adequately described the formation of a vein cuff at the distal anastomosis site. Additional procedures to the primary operation, such as the formation of an arterio‐venous fistula (AVF), were also considered.

Types of outcome measures

The outcome measures used in this review were the primary and secondary (after further intervention) patency rates and limb salvage rates. The patency rate following specific intervals of time from the intervention was reported in all trials included and was directly measured at the follow‐up examination for each patient. The limb salvage rate is the ultimate outcome measure, being the principle aim of the intervention.

Electronic searches

The Cochrane Peripheral Vascular Diseases Group Trials Search Co‐ordinator (TSC) searched the Specialised Register (last searched May 2012) and the Cochrane Central Register of Controlled Trials (CENTRAL) (2012, Issue 5), part of The Cochrane Library (www.thecochranelibrary.com). See Appendix 1 for details of the search strategy used to search CENTRAL. The Specialised Register is maintained by the TSC and is constructed from weekly electronic searches of MEDLINE, EMBASE, CINAHL, AMED, and through handsearching relevant journals. The full list of the databases, journals and conference proceedings which have been searched, as well as the search strategies used, are described in the Specialised Register section of the Cochrane Peripheral Vascular Diseases (PVD) Group module in The Cochrane Library (www.thecochranelibrary.com).

The following trial database was searched by the TSC for details of ongoing and unpublished studies:

World Health Organization International Clinical Trials Registry Platform (http://apps.who.int/trialsearch/).

Searching other resources

We searched for further reports from the studies identified from the electronic searches.

Selection of studies

Two authors, Ahmed Khalil (AK) and Gareth Griffiths (GG), independently selected trials for inclusion in the review and assessed the quality of the trials using quality assessment checklists provided by the Cochrane PVD Group. These are included in the 'Characteristics of included trials' table. Trials which meet most of the essential criteria were discussed between authors.

The assessment of the quality of included studies was based on adequacy of the randomisation process and concealment of allocation as reported by each study, as well as the 'Risk of bias' tool (Higgins 2011). Blinding was not considered because the nature of the problem, the intervention and the follow up make it impossible to blind the investigator.

Data extraction and management

AK and GG extracted data according to a fixed protocol, recorded on data extraction forms provided by the Cochrane PVD Group. Any disagreement was resolved by discussion and, where necessary, AB was consulted. AK entered the data into RevMan 5.1 for analysis.

Assessment of risk of bias in included studies

We rated the quality of the included trials according to the rating system used by the Cochrane Peripheral Vascular Diseases Group. We assessed the following domains: adequacy of the randomisation process, concealment of allocation, intention‐to‐treat analysis (ITT) and completeness of follow up, by allocating a score of A (clearly yes), B (unclear) or C (clearly no). If possible, a funnel plot (trial effect against trial size) was used to investigate publication bias. We contacted authors if evidence of ITT analysis was not clearly mentioned in the report.

In addition, we used the 'Risk of bias' tool from The Cochrane Collaboration to assess risk of bias of the following domains: selection bias, attrition bias, reporting bias and other bias. The domains were judged to be at low risk of bias, high risk of bias or unclear risk of bias according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Blinding was not considered as it is impossible to achieve during surgery and follow up.

Measures of treatment effect

Primary and secondary patency rates and limb salvage rates were the measures of treatment effect. They are included in the tables and reported in the text of the review.

Unit of analysis issues

Primary and secondary patency rates and limb salvage rates reported as percentage per year were compared between studies. We used bypasses as the unit of analysis.

Dealing with missing data

We contacted the corresponding authors of included studies asking for explanations of missing data and for any new data which may be available. Additional techniques were used to study the effects of any missing data (see assessment of reporting bias section).

Assessment of heterogeneity

Where appropriate, and depending on data availability, we planned to use a Chi² test of heterogeneity to investigate the heterogeneity between studies. When significance was encountered, we planned to perform a sensitivity analysis.

Assessment of reporting biases

Reporting bias was addressed when dealing with missing data. ITT was expected in the report and, if not mentioned, attempts were made to clarify this issue with the corresponding author. When missing data became available, the worst‐case scenario or sensitivity analysis techniques were followed and reported. Other forms of reporting bias such as selective reporting of data or stressing the desired outcome were reported when encountered. We contacted the authors for any unpublished data and looked for unpublished studies.

Data synthesis

We planned to pool data using a random‐effects model analysis. However, because sufficient data could not be compiled, we were unable to pool the data. All patency and limb salvage rates from different studies were organised at specific time intervals as reported by individual studies. Analyses were grouped in these reported intervals, usually by the number of months following the intervention.

Subgroup analysis and investigation of heterogeneity

Subgroup analyses were not performed in this review, but subgroup analyses from the individual studies are reported in the text.

Sensitivity analysis

In general, this analysis is performed when heterogeneity is encountered between the studies. It is also one of the methods used when missing data that cannot be replaced are encountered (see dealing with missing data section). Sensitivity analysis is also performed if significance is encountered to determine whether exclusion of individual studies would change the results of the analysis. We planned to perform sensitivity analyses if sufficient data were available.

Results of the search

See Figure 1. Eleven reports of six studies which met the inclusion criteria were identified, the remainder were not relevant. Multiple reports from the same study were reviewed to include the most up to date data. Only studies reporting on PTFE grafts were identified. Studies using other prosthetic grafts such as Dacron grafts were not identified.

Included studies

See: Characteristics of included studies.

Excluded studies

No randomised trials were excluded.

Allocation

Only two trials have a clear description of allocation concealment within the text. Stonebridge 1997 randomised patients at operation by a telephone call to the Scottish Cancer Trial Office, University of Edinburgh, and patient were stratified for each level of distal reconstruction. Laurila 2004 utilised a central block stratified randomisation by the study centre.

We were unable to assess allocation bias in three trials (Hamsho 1999; Kreienberg 2002; Panneton 2004) due to lack of text describing the method used. We have written to the contact authors of these trials and a reply is awaited. SCAMICOS 2010 used blocks of sealed envelopes with equal numbers of randomisation allocation. This method is prone to selection bias in some circumstances and many modern techniques avoid this risk.

Blinding

Blinding was not considered as it is impossible to achieve during surgery and at follow up.

Incomplete outcome data

All authors clearly reported the number of patients randomised and included the basic characteristics of each group in the study. Hamsho 1999 included all the randomised patients in the analysis with none missing or withdrawn from the study. Kreienberg 2002 reported loss to follow up of one patient, but the analysis included all randomised patients. Laurila 2004 accounted for all randomised patients but six were lost to follow up and none were excluded from the analysis. Panneton 2004 excluded 13 patients from the analysis with a clear explanation of the reasons (protocol violation, claudication in five patients, pedal vessels for distal anastomosis in four patients, vein patch used concomitantly with the pre‐cuffed graft in three patients and "others" in one patient). All included patients were accounted for. Stonebridge 1997 accounted for all patients with ninety patients out of original ninety five available for extended follow‐up (Griffiths 2004). A clear description was given for data censoring during the extended follow up analysis; however the effect of this is unclear. SCAMICOS 2010 excluded one patient from randomisation due to unknown group and seven patients in total were reported to have missing data.

Selective reporting

All outcomes as stated in the methods have been addressed in all included studies, with no concerns about selective reporting. SCAMICOS 2010 declared another independent study which randomised patients to externally supported PTFE graft versus non‐supported grafts, with no data reported.

Other potential sources of bias

We did not identify any other issues or sources of bias in included studies except in three studies (Panneton 2004; SCAMICOS 2010, Stonebridge 1997). The Panneton 2004 and SCAMICOS 2010 studies were industry sponsored. Panneton 2004 has been contacted for further clarification and a reply is awaited. SCAMICOS 2010 reported that the sponsoring body had no access to data and no influence in preparing the manuscript. SCAMICOS 2010 recruited more patients to FemPopBK group than required according to the power analysis but failed to recruit an adequate number to the FemDist group leaving this group underpowered. Both Stonebridge 1997 and SCAMICOS 2010 studies had high numbers of participating surgeons (18 and 33 respectively). It is unclear what effect this may have on the outcome. In personal communication, SCAMICOS trialists stated that they investigated whether study centre size effected outcome and found none. While both study groups in SCAMICO 2010 were comparable, the whole study population had a high number of females and non‐smokers in the study.

PTFE with vein cuff versus PTFE without vein cuff

We identified two trials, Stonebridge 1997 (with extended follow up reported by Griffiths 2004) and SCAMICOS 2010, which compared PTFE (both externally supported and not externally supported) with and without a vein collar at the distal anastomosis.

Stonebridge 1997 included distal anastomoses to the above knee popliteal artery (150 grafts), below knee popliteal artery (96 grafts) and tibial vessels (15 grafts). For the below knee bypass group, the primary patency rate was higher in the vein cuff group (80.3% versus 65.3% and 51.8% versus 29.1% at 12 and 24 months respectively, P = 0.03). The difference in secondary patency rate was not statistically significant (82.9% versus 72.5% and 58.6% versus 34.9% at 12 and 24 months respectively, P = 0.14). The limb salvage rate was also not statistically different (86.3% versus 71.8% and 82.6% versus 62.2% at 12 and 24 months respectively, P = 0.08). Another 15 bypasses were to the tibial vessels, and if they are added to the 95 bypasses to the below knee popliteal artery, then the 24 months patency rate becomes 58% in the vein cuff group versus 27.6% in the non‐cuff group (P = 0.022). Data for 90 below knee bypasses were available in the extended follow‐up report. The data were censored when the grafts at risk number had fallen to 30% and 10% of the original total number of procedures, to give cumulative secondary patency rates of 63% versus 35% (P = 0.026) and 45% versus 19% (P = 0.018) at 24 and 36 months respectively. The difference in limb salvage rate at 36 months was not statistically significant (78% versus 61%, P = 0.08). See Table 1 and Table 2.

The SCAMICOS 2010 study included below knee femoro‐popliteal (FemPopBK) bypasses (202 grafts) and femoro‐distal (FemDist) bypasses (150 grafts). In the FemPopBK group, the differences in primary patency rate, secondary patency rate and limb salvage rate were not statistically significant at 36 months (primary patency rate 26% (95% CI 18 to 38) in the collar group and 43% (95% CI 33 to 58) in the no collar group, secondary patency rate 32% (95% CI 23 to 44) in the collar group and 42% (95% CI 31 to 56) in the no collar group, and limb salvage rate 64% (95% CI 54 to 75) in the collar group and 61% (95% CI 50 to 74) in the no collar group). Survival rates were not significantly different at 36 months (60%, 95% CI 51 to 73 in the collar group and 67%, 95% CI 56 to 81 in the no collar group).

In the FemDist group, the differences in primary patency rate, secondary patency rate and limb salvage rate were not statistically significant at 36 months (primary patency rate 20% (95% CI 11 to 38) in the collar group and 17% (95% CI 9 to 33) in the no collar group, secondary patency rate 22% (95% CI 12 to 39) in the collar group and 20% (95% CI 11 to 35) in the no collar group, and limb salvage rate 59% (95% CI 46 to 76) in the collar group and 44% (95% CI 32 to 61) in the no collar group). Survival rates were not significantly different at 36 months (66% (95% CI 54 to 82) in the collar group and 53% (95% CI 40 to 72) in the no collar group).

See Table 3, Table 4, Table 5 and Table 6.

Pre‐cuffed PTFE versus vein cuff PTFE

One study was included (Panneton 2004). The difference in primary patency rate between the two groups was not statistically significant (62% pre‐cuffed PTFE versus 52% vein cuff PTFE at 12 months, and 49% versus 44% at 24 months, P = 0.53). The secondary patency rate also showed no statistically significant difference between the groups (66% pre‐cuffed PTFE versus 53% vein cuff PTFE at 12 months, and 55% versus 50% at 24 months, P = 0.30). Limb salvage rates were also similar in the two groups (75% pre‐cuffed PTFE versus 72% vein cuff PTFE at 12 months, and 62% versus 65% at 24 months, P = 0.88). Panneton 2004 reported a trend towards a shorter operative procedure in the precuffed group compared with the vein cuffed group (210 mins versus 242 mins respectively) but this was not statistically significant (P = 0.11). See Table 7.

Spliced vein bypass graft versus PTFE graft with a distal vein cuff

One trial was included (Kreienberg 2002). The primary patency rate at 24 months was 44% in the spliced vein group and 50% in the vein cuff group, which was not statistically significant. The difference in secondary patency rate, however, was statistically significant between the two groups at 24 months (86% in the spliced vein and 52% in the vein cuff group, P < 0.05). The limb salvage rate was not significantly different at the same time point, being 94% in the spliced vein group and 85% in the vein cuff group. Kreienberg 2002 also reported a statistically significantly shorter operative time for the vein cuff group compared with the spliced vein group (177 ± 9 mins versus 263 ± 21 mins respectively, P < 0.05), less blood loss (P < 0.05), shorter hospital stay in hospital (P < 0.05) and less need for revision or reoperation for wound complications in the vein cuff group compared with the spliced vein group. See Table 8.

PTFE graft with interposition vein cuff at the distal anastomosis with adjuvant AVF versus without adjuvant AVF

Two trials were included (Laurila 2004; Hamsho 1999). The trial by Laurila 2004 showed a primary patency rate at 24 months of 29% in the AVF group versus 36% in the non‐AVF group (P = 0.77). The secondary patency rate was 40% in both groups at 24 months (P = 0.89), and the limb salvage rate in the same period was 65% in the AVF group and 70% in the non AVF group (P = 0.97). The trial by Hamsho 1999 also showed similar statistically not significant results at 24 months with a primary patency rate of 32% versus 28% in the AVF group and non‐AVF group respectively (P = 0.20). The secondary patency rate was 28% in the AVF group and 24% in the non‐AVF group (P = 0.20) and the limb salvage rate was 62% and 71% respectively (P = 0.30). Hamsho 1999 reported that the AVF group had an increased average operative time of approximately 30 minutes compared with the non‐AVF group. Laurila 2004 also concluded that the operative time was increased in the AVF group but did not back this up with data. See Table 9 and Table 10.

PTFE with vein cuff versus PTFE without vein cuff

There is contradictory evidence from the two trials. One randomised trial (Stonebridge 1997) suggests that a vein cuff confers a benefit in the primary patency rate at 12 and 24 months. The secondary patency rate and limb salvage rate were not significantly different in the same period. The cumulative secondary patency rate at 24 and 36 months remains statistically significantly improved when a vein cuff is used, as reported by the extended follow‐up report, however the limb salvage rate at 36 months was not statistically different.

The other trial (SCAMICOS 2010) included a high number of female and non‐smoking patients. The primary patency rate, secondary patency rate and limb salvage rate showed no statistically significant difference at 36 months in either below knee femoro‐popliteal or femoro‐distal bypasses.

Pre‐cuffed PTFE versus vein cuff PTFE

The evidence from a randomised trial of moderate quality (Panneton 2004) suggests no significant difference in primary patency rate, secondary patency rate and limb salvage rate between the two techniques. The operative time was shorter in the pre‐cuffed group, but this was not statistically significant. The authors felt that the vein cuff PTFE graft is more technically demanding and so the pre‐cuffed PTFE grafts can probably offer a technically less demanding alternative with similar benefits.

Spliced vein bypass graft versus PTFE graft with a distal vein cuff

One intermediate quality randomised trial (Kreienberg 2002) showed no statistically significant difference in the primary patency rate and limb salvage rate between the two groups. The secondary patency rate was significantly improved in the spliced vein group at 24 months. Spliced veins required significantly longer operative time, resulted in more blood loss, longer stays in hospital and needed revision or reoperation more often due to wound complications.

PTFE graft with interposition vein cuff at the distal anastomosis with adjuvant AVF versus without adjuvant AVF

One good quality study (Laurila 2004) and one intermediate quality study (Hamsho 1999) reported no effect of adding an AVF to the distal anastomosis on any of the primary outcomes. Laurila 2004 reported no significant difference between the two groups throughout the study period in the primary and secondary patency rates and limb salvage rates. There were no perioperative deaths. Immediate reoperative and graft occlusion rates were similar between the groups. Hamsho 1999 observed a trend toward improved patency in the first six months in the AVF group, which diminished with time resulting in no significant differences in the primary patency rate, secondary patency rate and limb salvage rate over the observation period of 24 months. The construction of an AVF increased the operative time by about 30 minutes. No conclusion can be drawn from the relationship between AVF occlusion and graft failure.

PTFE with vein cuff versus PTFE without vein cuff

Stonebridge 1997 reported only 95 bypasses out of 261 that were to the below knee popliteal artery. As such, this is an underpowered study to look at the differences in this subgroup, and the authors acknowledged this point.The assumption used in the power calculations before the study was that the benefit from below knee vein cuffed PTFE bypass would be higher than in above knee bypasses. The randomisation process and allocation concealment is clear. The study reports no demographic differences between the two groups. The complication rate is not reported separately for below knee bypasses in this trial. In the discussion section, the authors looked further to the effect of a history of previous graft failure and concluded that previous graft failure did not affect the outcome. The risk of bias was low in this study.

SCAMICOS 2010 reported 202 below knee femoro‐popliteal bypasses and 150 femoro‐distal bypasses. The randomisation process is described clearly. Power calculations are discussed but the study failed to recruit the proposed number for the FemDist group. Sufficient numbers were recruited for the FemPopBk group. The two study groups are comparable but the whole study population has a high number of females and non‐smokers. All patients are accounted for, including missing data. The number of participating centres and surgeons is high, which may create a lot of heterogeneity in the procedure performed, and no adjustment has been made for this. The study is industry sponsored. The overall risk of bias is low in this study but some limitations are noted.

Pre‐cuffed PTFE versus vein cuff PTFE

Panneton 2004 organised a randomised study comparing a pre‐cuffed PTFE graft (Impra Distaflo^TM PTFE graft, Bart Peripheral Vascular Inc., Tempe, AZ) to a vein cuff PTFE graft for below knee bypasses. The patients were randomised before surgery but details of the method used are not described and there is no allocation concealment. Because of the overall quality of the study, we felt that the randomisation probably followed an acceptable method. Power calculations were not mentioned and the number of patients included in the study was low (n = 91). However, the study was well organised and followed a strict protocol. The two groups were comparable in their basic characteristics but the pre‐cuffed group (n = 47) had more patients with acute limb ischaemia than the vein cuff group (n = 44), mainly following graft failure or acute thrombosis (19% versus 4.5%, P = 0.03). The author performed a subgroup analysis by excluding the acute ischaemic patients and found no significant difference in the results. The risk of bias in this study is moderate.

Spliced vein bypass graft versus PTFE graft with a distal vein cuff

Kreienberg 2002 randomised only 39 procedures. Power calculations were not mentioned. The randomisation method and allocation concealment is not clear. The study was well designed and approved by the Committee on Research Involving Human Subjects. For this reason we felt that randomisation was probably adequate and assigned a low risk of bias. The two groups were comparable except in cardiac status and the author suggested that this may have contributed to a higher perioperative mortality and morbidity in the spliced vein group.

PTFE graft with interposition vein cuff at the distal anastomosis with adjuvant AVF versus without adjuvant AVF

Laurila 2004 randomised 59 procedures. The power of the study was probably inadequate and no power calculation was reported. The randomisation process was clearly described and allocation concealment was adequate. Therefore, we allocated this study to a very low risk of bias category. The baseline characteristics were comparable for the two groups. Hamsho 1999 randomised 89 bypasses over six years. There were no power calculations but the author acknowledges that the power of the study was questionable. Overall, the quality of the study is good and we felt that despite no clear description of the randomisation process or allocation concealment this was probably carried out correctly with a low risk of bias.

Summary

We have some critical remarks concerning this Cochrane review. We argue that the evaluation of the three articles concerning the use of vein cuff for infragenicular prosthetic bypass graft does not represent "trustworthy, independent and relevant information".

The three articles:

1. Stonebridge PA, Prescott RJ, Ruckley CV. Randomized trial comparing infrainguinal polytetrafluoroethylene bypass grafting with and without vein interposition cuff at the distal anastomosis. The Joint Vascular Research Group. Journal of Vascular Surgery 1997;26(4):543–50.

2. Griffiths GD, Nagy J, Black D, Stonebridge PA. Randomized clinical trial of distal anastomotic interposition vein cuff in infrainguinal polytetrafluoroethylene bypass grafting. British Journal of Surgery 2004;91(5):560–2.

3. SCAMICOS. PTFE bypass to below‐knee arteries: distal vein collar or not? A prospective randomised multicentre study. European Journal of Vascular and Endovascular Surgery 2010;39(6):747–54.

Two studies (a: JVRG study – Stonebridge et al. & Griffiths et al. and b: SCAMICOS) report conflicting results with respect to patency after femoro‐popliteal below‐knee by‐pass with and without a vein cuff at the distal anastomosis when a PTFE‐graft is used. JVRG found a better patency with a vein cuff while the larger SCAMICOS could not show any such effect.

1. In the second report (Griffiths et al.) with longer follow‐up from the JVRG's study only 90 patients are presented at risk for occlusion among patients with femoro‐popliteal below‐knee bypass while in the first report (Stonebridge et al.) 95 patients are presented at risk for occlusion. No explanation of why these five patients have been excluded is given in the second report. We think this must be presented and discussed in the review as a potential source of bias of the JVRG's study.

2. In the second report (Griffiths et al.) patients are censored when 30 or 10% of the original number of patients are at risk – meaning that emphasize is laid on the early follow‐up period and late occlusions are withhold from the statistical analysis, thus the statistical analysis is not made on all available data. The reasons for this procedure is not explained or motivated in the report. We think this must be presented and discussed in the review as a potential source of bias of the JVRG's study.

3. In the SCAMICOS "Patients were recruited to the FemPopBK group in excess of that which was required according to the power analysis, but the study failed to recruit the stipulated number of patients to the FemDist group." In the review this is described as "Study stopped recruiting before proposed number of recruits was achieved leaving the study with unconfirmed power to detect a difference". Where conflicting results exist – femoro‐popliteal bypass below knee – our study recruited enough patients according to the power analysis. We think this should be presented in fair way in the review, which in its present form question the general power of our study.

4. In the review the SCAMICOS is criticised for heterogeneity due to the large number of participating centres. It is difficult to see why the 18 contributing surgeons of the JVRG's study should represent a substantially different problem with heterogeneity than in the SCAMICOS study. In addition SCAMICOS tested for any interaction between study centre size and use of vein collar with respect to patency and no interaction was found. We think that this issue, if actually relevant, should be presented in the review as a potential problem with both studies with the addition that SCAMICOS but not JVRG evaluated if study centre size influenced the outcome.

5. The review questions the large proportion of female patients and non‐smokers in the SCAMICOS. It is true that the SCAMICOS patients were a decade older than the patients of the JVRG's study and it is probably the reason for the female dominance in the study. Recently we have shown that there is no interaction between age & vein collar on one hand or sex & vein collar on the other hand with respect to patency (SCAMICOS, F, Lundgren. Does patency after a vein collar and PTFE‐bypass depend on sex and age? Re‐analysis of a randomised trial. International Angioliology 2012;31:156‐62).  Current smoking was 29% in the JVRG's study and regular smoking during the last five years was 32% in SCAMICOS – this does not appear to be much of a difference. These aspects should be presented in the review.

Several of the above issues have been published in the discussion in connection to the publication of SCAMICOS in EJVES.
  
 Finally in the plain language summary of the review: "The results show that if a cuff of vein is inserted at the lower end of the synthetic graft before attaching it to the artery below the knee, then the bypass graft remains functional for a longer period of time.", without any reservations and in spite of the fact that two randomised studies have presented conflicting results with respect to the effect of a distal vein collar in a femoro‐politeal bypass below‐knee. For bypasses to the crural arteries no evidence at all exists to prove any superiority with an added vein collar.   

Reply

The authors have made the following responses to the points raised by the SCAMICOS collaborators:

1. The authors had clearly stated within the ‘Effects of intervention’ section that ‘data for 90 patients were available for extended follow up’. We acknowledge that no explanation of the change from 95 patients in the original study is given. A Khalil and G Griffiths discussed this issue and we could not track why these five patients were lost to follow up.To clarify this we have now added ‘Ninety patients were available for extended follow‐up’ to the Incomplete outcome data (attrition bias) section. We have amended the review text (incomplete outcome data) to account for the fact that these five patients have been lost to follow up in the extended follow up report and we have changed the Risk of bias table judgement from 'low' to 'unclear' to reflect this issue.

2. Although clearly described, we are aware that the censoring of data may be perceived as a potential bias. We have now amended the sentence ‐ 'the description was clear for data censoring during the extended follow up analysis’ to 'A clear description was given for data censoring during the extended follow up analysis; however the effect of this is unclear' ‐ within the Incomplete outcome data (attrition bias) section and also within the Risk of bias table where the judgement of ‘low’ attrition bias has been changed to ‘unclear’.

3. The authors agree their description of SCAMICOS patient recruitment is not precise. We have amended the review text from ‐ ‘SCAMICOS 2010 stopped recruiting before the proposed number of recruits was achieved leaving the study with unconfirmed power to detect a difference’ to ‘SCAMICOS 2010 recruited more patients to FemPopBK group than required according to the power analysis but failed to recruit an adequate number to the FemDist group leaving this group underpowered’. This change is also reflected in the Risk of bias table text.

4. The authors agree that as both Stonebridge 1997 and SCAMICOS 2010 have used multiple surgeons any potential problem caused by this may apply to both studies. Our review did include the text ‘surgical technique not standardised’ within Stonebridge 1997 included study characteristics to acknowledge this. However, to further clarify we have added the text ‐ ‘Both Stonebridge 1997 and SCAMICOS 2010 studies had high numbers of participating surgeons (18 and 33 respectively). It is unclear what effect this may have on the outcome. In personal communication, SCAMICOS 2010 stated that they investigated whether study centre size effected outcome and found none’ ‐ to the Other potential sources of bias section. The Risk of bias tables text of both studies have also been amended. Stonebridge: other bias – ‘18 surgeons participated. This is not controlled for and the effect of this on the outcome is not known (judgement changed from ‘low’ to ‘unclear’). SCAMICOS: other bias – ‘33 surgeons participated. This is not controlled for and the effect of this on the outcome is not known (domain judgement remains ‘high’).

5. Our review has discussed the difference in patient populations between the Stonebridge 1997 study and SCAMICOS 2010 study. We have now added text following the feedback comments from the SCAMICOS trialists which may explain the high number of female and non‐smokers in their patient population and included their recent publication where they have further investigated this ‐ ‘However, one of the trials (SCAMICOS 2010), which showed no significant benefits, included a population with a high percentage of females and non‐smokers, unusual for a population with arterial disease. By personal communication, SCAMICOS trialists have stated that this may be explained by the older population recruited in this trial. They have further commented on the interaction of sex and vein collar; and age and vein collar; in a recent publication (SCAMICOS 2012)'. We do not believe that the SCAMICOS 2010 data for regular smoking in the last five years (32%) and the Stonebridge 1997 current smoking data (29%) is comparable in this case. We believe a more comparable figure would be former+current smokers which was 83% of patients, as expected in a vascular disease patient population.

6. The authors agree that the plain language summary is not an accurate reflection of what we have reported in both the Summary of main results and Authors Conclusions. We have amended the text within the PLS in respect to this ‐ ‘Results from two trials which looked at the effect of inserting a cuff of vein at the lower end of the synthetic graft before attaching it to the artery below the knee are conflicting. With one study showing that the bypass graft remains functional for a longer period of time and in the other study no benefit was seen’.

Contributors

Feedback: Dr Fredrik Lundgren, principal investigator on behalf of SCAMICOS collaborators.

Reply: A Khalil and G Griffiths.

Summary

It's a good thing that some of the shortcomings of the JVRG’s study have been mentioned in the new version of the review and some misleading descriptions of SCAMICOS have been corrected. (CD007921. DOI: 10.1002/14651858.CD007921.pub2)
 
 However, our study is still criticised for industrial sponsoring, use of a factorial design and for investigating a different population of patients compared to the JVRG study. All are presented as sources of bias in the review. The companies sponsoring SCAMICOS had no access to the data and had no influence in preparation of the manuscript and this is also clearly stated in the article. Factorial designs have been used since the late 19th century and are usually regarded as very effective methods rather than as a source of bias. The design is also described in full detail in our SCAMICOS paper. Concerning the patient population, we studied the patients that were available to us. Different patient populations can of course generate different results but at least the influences of age and sex on the effect of a vein collar has been studied by us and appear not to change the outcome of a graft with/without a vein collar. There were, of course, previous smokers also in the SCAMICOS and usually you may assume that most patients with critical ischemia are current or previous smokers. Thus, in all likelihood the distributions of smoking habits are very similar in the two discussed studies – current smoking and regular smoking during the last 5 years being almost the same. The suggestion that different ways to describe the smoking habits of the population should be a source of bias with respect to the outcome of a graft with/without a vein collar appears somewhat farfetched. To sum up, we seriously question the reviewers’ classification of the above discussed circumstances as likely sources of bias concerning the outcome of a graft with/without a vein collar.
 
 Finally, after our criticism of the first version of the review and the JVRG's study our methods of randomisation and allocation concealment (which are accepted methods due to the Cochrane Handbook) has been "degraded" from 'Low risk of bias' to 'Unclear risk of bias' without any further argumentation or comment compared to the first version of the review.
 
 Cochrane Collaboration has been extremely important for the development of clinical practise and usually has an unsurpassed reputation. However, based on the above arguments we are still concerned that this review does not represent "trustworthy, independent and relevant information". Possibly, authors should not review their own studies to avoid any conflict of interests.

Reply

A reply from the review authors is awaited.

Contributors

Feedback: Dr Fredrik Lundgren, principal investigator on behalf of SCAMICOS collaborators.