PROPATEN heparin-bonded vascular graft for peripheral arterial disease

Clinical and technical evidence

Costs and resource consequences

PROPATEN grafts are currently used in the NHS. No published UK evidence on resource consequences was identified from literature searches.

The graft can be used as an alternative to standard ePTFE grafts and no additional resources would be needed before or during the bypass procedure.

There is some evidence that the duration of the bypass procedure may vary depending on the type of graft used. The Bellosta et al. (2013) study reported significantly longer operating times with PROPATEN grafts (with a distal vein patch) than with a pre‑cuffed ePTFE graft (215 minutes compared with 138 minutes; 95% CI 45 to 118 minutes, p<0.001). However, the Lindholt et al. (2011) trial observed no differences in operation time between PROPATEN grafts and standard PTFE grafts, though no supporting data were provided.

In their discussion, Daenens et al. (2009) note that ASV graft procedures are more complex than synthetic graft procedures, with more incisions (of larger size) and longer operating times. However, in the absence of references or supporting data, these observations should be considered with the same caution that applies to any anecdotal evidence.

Dorigo et al. (2012) noted significant differences in the mean post‑operative length of stay for patients who had PROPATEN grafts (13.1 days) compared with patients who had ASV grafts (10.2 days, p<0.001). A comparison of the PROPATEN graft (with a distal vein patch) and a pre‑cuffed ePTFE graft reported lengths of stay as 11.8 days and 11 days respectively (Bellosta et al. 2013). A statistical calculation was not reported. PROPATEN grafts are used in the same way as other prosthetic grafts so additional training is not essential, although the manufacturer does offer optional training if desired.

Strengths and limitations of the evidence

The quality of the included studies was generally weak. The evidence consisted of 1 randomised controlled trial (which was both conducted and reported poorly), and retrospective cohort studies (which are subject to selection bias). Three of the studies are limited by relatively small sample sizes at single sites, with the number of patients totalling less than 100 in each study. Three other studies, 2 of which were multicentre studies, had stronger sample sizes (total numbers of 350 to 950 patients).

The primary outcome was not specified by the authors of the cohort studies, and primary patency was evaluated as the primary effect in the clinical trial. Patency as a surrogate endpoint may not be as relevant as direct clinical outcomes such as limb salvage or survival. These clinical outcomes were inconsistently reported.

The literature search only identified 1 randomised controlled trial (Lindholt et al. 2011), which compared PROPATEN grafts with standard PTFE grafts. Neither the conduct nor the reporting of this study was robust. Randomisation was not done correctly at some sites; it is not clear how many patients were affected. patients were not stratified by disease severity (critical ischaemia or claudication) or type of bypass (femoral‑femoral crossover or femoropopliteal), which the authors acknowledged as a limitation of the design. Similarly the authors did not differentiate between those bypasses that terminated above or below the knee. This is of relevance, because it is widely recognised that patency rates for below‑knee grafts are considerably worse than those for below‑knee grafts (Twine and McLain 2010). No differences were found between baseline variables, but it is not clear whether unknown confounders could have resulted in a selection bias.

Although the authors report application of an intention‑to‑treat methodology, some patients were excluded. It is not clear how many patients from each group were lost after randomisation. The data presented by the authors are inconsistent and there are some miscalculations. After having treatment, it is reported that 7 patients were lost to follow‑up and 4 died before follow‑up, but there is no indication of which study arm they were in. A further 14 patients had been excluded, for reasons such as 'use of wrong graft' and 'technical errors', again with no record of the treatment arm. A flow diagram suggested that these exclusions were made after randomisation and before the operation, but it would seem more likely that these were post‑surgery effects.

There were slight differences in the appearance of the 2 grafts, meaning that the attempted blinding of surgeons was ineffective in at least some of the cases, although assessors of outcomes were blinded. It is not known if all vascular grafts of the 'standard PTFE' type were the same, because the device and manufacturer is not reported.

The described lengths of follow‑up add further confusion. Average lengths of follow‑up were reported as 9.75±3.79 months (PROPATEN) and 10.30±3.35 months (standard PTFE), but the authors also assert that all 546 included patients had follow‑up data for the assessment of 1‑year primary patency. This suggests there may have been some additional attrition bias that has not been acknowledged. The authors also note that 1 year may not be a sufficient length of follow‑up to determine whether the heparin‑bonding maintains its effects over time.

The authors noted that a 'surprisingly high' number of silent occlusions were detected at the 1‑year follow‑up appointment. Because of this, planned Cox regression analyses were changed to logistical regression analyses, and supplemental analyses were introduced to adjust for differences in indication and bypass type (though not above‑/below‑knee results). The reported significant difference in primary patency between the 2 graft types was lost once these adjustments had been made.

Not all outcomes were adequately reported. Perioperative bleeding was only mentioned in the abstract. There was a reference to the influence of 'prosthetic infections' on statistical analyses, but no details of how many there were or which patients were affected. It is not known whether other adverse events occurred. Finally, throughout the text and the tables there were many reporting errors and inconsistencies. It is probable that these problems will have affected the validity of the results, which should be interpreted with a high level of caution.

All retrospective studies may, by their nature, have been affected by performance bias and detection bias due to the lack of blinding. Selection bias is another common flaw of these non‑randomised studies. This is less of a risk in those instances where the analyses have taken all important confounders into account, but there remains some potential for unknown confounders to have influenced results.

Bellosta et al. (2013) compared the PROPATEN graft (with a distal vein patch, n=40) with a pre‑cuffed PTFE graft (Distaflo, n=39) in a retrospective cohort analysis at a single site. With a total of 79 patients included, this was a relatively small sample compared with the studies with an autologous graft comparator. The use of a distal vein patch may have influenced the results and may not be directly comparable to other outcomes where adjunctive techniques were not performed, affecting external validity.

Treatment decisions depended upon the surgeon's preference, so there may have been some selection or performance bias. In the methods, the authors describe how they used propensity scoring to adjust for baseline differences between groups. However, it is not clear that this was applied and the authors suggest that age differences are the most likely explanation for a discrepancy in survival between groups.

Attrition bias may have been present. Mean follow‑up for all patients was 17 months, but ranged from 3 to 82 months, and was not reported by intervention group. Results are reported up to 24 months, so there must have been some degree of extrapolation/estimation in the time‑to‑event analyses. Safety outcomes were not reported separately for the 2 treatment groups.

Bechara (2014) did not distinguish between above‑ and below‑knee results for the patients who had femoropopliteal bypasses in this single‑site study. Sample sizes were relatively small (PROPATEN n=39, SLFG n=20). Methods and inclusion criteria were not explicitly stated, and there is a high risk of selection bias in this retrospective review. There was no reporting of any baseline measures or attempts to account for potential confounders, except that results for femoropopliteal and femorotibial bypasses were reported separately. No explanation is given for the absence of 24‑month results for the femorotibial bypasses. The apparent increase in primary patency between 12 and 18 months for the PROPATEN group suggests that comparisons did not take into account loss to follow‑up. Safety outcomes were not reported. Given the paucity of information reported, caution is advised in interpretation of the results.

One strength of the multicentre study by Dorigo et al. (2012) was the large sample sizes for both PROPATEN grafts (n=556) and ASV grafts (n=394). However, there is a high risk of selection bias. Data were obtained from a register that did not have specific inclusion or exclusion criteria, and treatment choices were made by each participating surgeon. The influence of some confounders was determined to be significant following univariate and multivariate analyses, but these findings were not used to statistically adjust key outcome measures. The study only included results for below‑knee bypasses, and so would not have been affected by any above‑knee results.

The average length of follow‑up was not reported by graft type, and full data were not available for all patients. Of the 950 patients, 97% (921) had at least 1 post‑operative clinical and duplex ultrasound investigation, but only 50% had at least 2 years of follow‑up data available. Some of the longer‑term (2 ‑ear) follow‑up results did not differentiate between treatment groups. It is not clear how many patients were lost to follow‑up from each of the 2 treatment groups, possibly resulting in attrition bias.

The single‑site study by Daenens et al. (2009) included 240 heparin‑bonded ePTFE grafts (presumed to be PROPATEN) and 110 ASV grafts. The PROPATEN graft was named in the introduction and in another study by the same authors, but because it was not specifically described in the methods there is a small chance that a different graft was used. Results were analysed separately depending on bypass type, which is a strength of the study. The authors recognised that patients who had the heparin‑bonded ePTFE grafts were more likely to have had a secondary intervention, and that a lower proportion of these patients was represented in the below‑knee femoropopliteal group. Other potential confounders were similarly taken into account in the analyses, and supported by data. The risk of selection bias was therefore relatively low for a non‑randomised study.

Data were available from all patients after the same length of follow‑up, although 4% of patients (15/350) may not have had duplex ultrasonography at the end of the study period to confirm patency, leading to a slight risk of attrition bias.

The study by Dorigo et al. (2005) was the only study to report comparisons of PROPATEN grafts with both other (non‑heparin‑bonded) PTFE grafts and ASV grafts. Another strength was that it focused on below‑knee bypasses only.

Small sample sizes (21–25 patients in each group) were a limitation, and it appears that the study was done at a single vascular centre. There may also have been some performance bias, because the results of the PROPATEN group were prospectively collected and the surgeons may have been aware that outcomes for this group would be scrutinised. There is a high likelihood of selection bias, particularly with the retrospective choice of comparators. The authors state that these control groups were 'randomly selected', but provide no further detail about how this was achieved.

The mean follow‑up was 19±11 months, indicating that there was a high degree of variation between patients; differences between groups were not reported. Primary patency at 18 months was estimated, and results may have been affected by differential attrition. There also appeared to be a few errors in the reported data.