Standard EVAR

The 2014 Cochrane systematic review (Paravastu et al, 2014) included 4 RCTs (reported across multiple publications) comparing EVAR with open surgical repair of infrarenal AAA. The update literature search performed by Cochrane Vascular Group yielded 354 abstracts, of which 4 full manuscripts were ordered. Of the 4 articles reviewed, an additional publication reporting an RCT (EVAR-1 trial) that was already included in the Cochrane review was identified. Thus, a total of 4 RCTs, published across multiple publications, was considered relevant for comparisons between standard EVAR and open surgical repair of unruptured AAAs. The 2014 Cochrane systematic review included 1 RCT (EVAR-2 trial) comparing EVAR with non-surgical management, in patients for whom open surgical repair was considered unsuitable. The update literature search performed by Cochrane Vascular Group yielded 1 publication reporting long-term follow-up of the EVAR-2 trial.

In December 2017, Cochrane performed another literature search to identify studies which were published during guideline development. The search yielded a total of 296 abstracts; of which, 4 full manuscripts were ordered. Upon review of these 4 articles, a publication of another RCT (DREAM trial) already included in the Cochrane review was identified.

Clinical evidence

• Four RCTs provided moderate to high-quality evidence on all-cause mortality in people with unruptured AAAs in whom surgery was considered appropriate. The studies reported that:

  • Perioperative mortality (30-day or in-hospital) was lower with EVAR than with open surgical repair (high-quality evidence; from 4 RCTs, including 2,783 people).
  • 0–6-month mortality was higher with open surgical repair than with EVAR (high-quality evidence from 4 RCTs, including 2,783 people).
  • 6-month–4-year mortality could not be differentiated between EVAR and open surgical repair (moderate-quality evidence from 4 RCTs, including 2,664 people).
  • 4–8-year mortality could not be differentiated between EVAR and open surgical repair (moderate-quality evidence from 2 RCTs, including 1,665 people).
  • Above 8-year mortality could not be differentiated between EVAR and open surgical repair (moderate-quality evidence from 2 RCTs, including 1,230 people).
  • There was no difference in 0–15-year mortality between EVAR and open surgical repair (high-quality evidence from 3 RCTs, including 2,484 people).
• Four RCTs provided very low- to high-quality evidence on AAA-specific mortality in people with unruptured AAAs in whom surgery was considered appropriate. The studies reported that:

  • 0–4-year AAA-specific mortality could not be differentiated between EVAR and open surgical repair (very low-quality evidence from 4 RCTs, including 2,783 people).
  • 0–8- year AAA-specific mortality could not be differentiated between EVAR and open surgical repair (moderate-quality evidence from 3 RCTs, including 2,484 people).
  • 0–15-year AAA-specific mortality could not be differentiated between EVAR and open surgical repair (moderate-quality evidence from 3 RCTs, including 2,484 people).
  • 8–15-year AAA-specific mortality was higher with EVAR than with open surgical repair (high-quality evidence from 1 RCT including 1,252 people).
• High-quality evidence from 4 RCTs, including 2,747 people with unruptured AAAs, reported shorter length of hospital stay in patients treated by EVAR compared with those treated by open repair.
• Moderate-quality evidence from 4 RCTs, including 2,783 people, reported lower rates of pulmonary-related mortality in patients treated by EVAR than those treated by open surgery. Low- to moderate-quality evidence from 4 RCTs, including 2,783 people with unruptured AAAs, could not differentiate cardiac- and stroke-related mortality rates between patients treated by EVAR and those treated by open repair (follow-up not reported).
• High-quality evidence from 2 RCTs, including 2,432 people with unruptured AAAs, reported lower pulmonary complication rates in patients treated by EVAR compared with those treated by open repair (follow-up not reported). Low-quality evidence from 3 RCTs, including up to 2,432 people with unruptured AAAs, could not differentiate non-fatal stroke, sexual dysfunction and renal complication rates between patients treated by EVAR and those treated by open repair (follow-up not reported).
• Very low-quality evidence from 3 RCTs, including 2,484 people with unruptured AAAs, reported higher rates of any type of reintervention in patients treated by EVAR compared with those treated by open repair at 4-year and 8-year follow-up. Moderate-quality evidence from 1 RCT, including 546 people with unruptured AAA, could not differentiate rates of any type of reintervention between patients treated by EVAR and those treated by open repair between 8- and 15-year follow-up. When considering total follow-up periods, high-quality evidence from 2 RCTs including 1,603 people reported higher rates of any type of reintervention in patients treated by EVAR than those treated by open repair at follow-up of up to 15 years.
• High-quality evidence from 1 RCT, including 351 people with unruptured AAA reported higher rates of AAA-related reintervention in patients treated by EVAR compared with those treated by open repair at follow-up of up to 15 years. High-quality evidence from another RCT including up to 1,252 people with unruptured AAAs, reported higher rates of life-threatening reintervention in patients treated by EVAR compared with those treated by open repair at follow-up of up to 15 years.
• Moderate-quality evidence from 1 RCT, including 1,341 people with unruptured AAAs, could not differentiate quality of life measures (SF-36, and EQ-5D scores) between patients treated by EVAR and those treated by open repair at 2-year follow-up.

The results of a review of casemix-adjusted observational studies comparing EVAR and open repair are presented in Evidence review K2.

Economic evidence

Clinical evidence

No randomised trials of complex EVAR compared with open repair for people with complex AAAs were identified. The results of a review of casemix-adjusted observational studies are presented in Evidence review K2.

Economic evidence

Clinical evidence

• Low- to moderate-quality evidence from 1 RCT, including 404 people with unruptured AAAs that were considered unsuitable for open repair, could not differentiate all-cause mortality rates between patients treated by EVAR and those who received no intervention at 6-month, 4-year, 8-year and 12-year follow-up.
• Low-quality evidence from 1 RCT, including 404 people with unruptured AAAs that were considered unsuitable for open repair, could not differentiate AAA-related mortality rates between patients treated by EVAR and those who received no intervention at 6-month follow-up. Conversely, moderate-quality evidence from the same study reported lower AAA-related mortality rates in patients treated by EVAR compared with those who received no intervention at 4- and 8-year follow-up.
• Very low-quality evidence from 1 RCT, including 404 people with unruptured AAAs that were considered unsuitable for open repair, could not differentiate rates of fatal myocardial infarction and stroke-related mortality between patients treated by EVAR and those who received no intervention at 4-year follow-up.
• Low-quality evidence from 1 RCT, including 404 people with unruptured AAAs that were considered unsuitable for open repair, reported higher rates of non-fatal myocardial infarction in patients treated by EVAR than those who received no intervention at 4-year follow-up. Low-quality-evidence from the same trial could not differentiate rates of non-fatal stroke in patients treated by EVAR compared with those who received no intervention at 4-year follow-up.
• Very low-quality evidence from 1 RCT, including 404 people with unruptured AAAs that were considered unsuitable for open repair, could not differentiate quality of life measures (SF-36, and EQ-5D scores) between patients treated by EVAR and those who received no intervention at 2-year follow-up.

Economic evidence

Other evidence sources

The key randomised controlled trials (RCTs) in this area are relatively old. The committee looked at more recent observational evidence, to see if changes in surgical techniques and technology have led to different outcomes. The observational studies are at high risk of bias, but their findings are broadly in line with the RCTs. They show that, while outcomes from EVAR have improved over the last 15 years, outcomes from open surgical repair have also improved by roughly the same amount. This means the difference in outcomes between the two has remained fairly constant. See evidence review K2.

Registries like the National Vascular Registry can provide a useful snapshot of current practice, and the analyses that informed the committee’s decision-making made use of data from them. However, they are not designed to evaluate the comparative benefits and harms of different surgical approaches, such as EVAR and open surgical repair. Therefore, they cannot be considered a reasonable alternative to RCT data. In addition, an analysis using the registry data showed that EVAR still did not provide greater long-term benefits than open surgical repair, and that it still has higher net costs.

The outcomes that matter most

The committee agreed that decision-making needs to balance the short- and long-term impacts of AAA repair. Naturally, the risk of perioperative mortality is a critical consideration that weighs heavily in the minds of people undergoing repair. However, long-term survival and the need for reintervention are also vital determinants of the overall value provided by the different approaches. This is because committee members believed that the fundamental goal of AAA repair is to ensure that people live as long as possible and have the best quality of life possible following intervention.

The quality of the evidence

Benefits and harms

Unruptured infrarenal AAA in people for whom open repair would be suitable

The committee discussed the cost-effectiveness evidence for the repair of unruptured infrarenal AAA. The committee were aware that this population, for whom open surgical repair is a suitable option, comprised the majority of both clinical and published economic evidence for this review question. The committee agreed that the published UK economic evidence could only reasonably be interpreted as evidence that EVAR was not likely to be an effective use of NHS resources, though it was noted that none of the studies included the longest-term follow-up that is currently available, namely 15 years of data from the EVAR-1 trial. The committee therefore considered evidence from the new economic model developed for this guideline.

The committee were satisfied with the modelling approach of: (1) using National Vascular Registry data to inform baseline perioperative mortality, and the results of a Cochrane meta-analysis to inform relative effects; (2) estimating long-term survival by calibrating general population mortality to the EVAR-1 open surgical repair data conditional on surviving for 30 days after the intervention; and (3) estimating relative long-term survival using a hazard ratio from a meta-analysis of long-term data from 3 RCTs (DREAM, EVAR-1 and OVER). The committee agreed that the new economic model provides compelling evidence that EVAR is not a cost-effective option for infrarenal AAA compared with open surgical repair. The base-case model results indicate that EVAR produces fewer QALYs than open surgery at a higher total cost to the NHS and PSS, and this is reflected in the probabilistic results, with a low probability of its ICER being £20,000 per QALY gained or better. Results were also robust to scenario and one-way sensitivity analyses, including using only EVAR-1 study data.

The committee discussed the cost results from the new model, and agreed that the high acquisition cost of EVAR was likely to be the key cost difference between EVAR and open surgery in practice. It advised that the modelled cost of complications following EVAR appeared low compared with clinical experience. However, it was agreed that any increase in EVAR complication costs would strengthen the cost-effectiveness results in favour of open surgical repair, and would therefore not affect interpretation of the evidence. The committee also discussed the cost of routine monitoring following EVAR and advised that, in practice, adherence to scheduled monitoring following EVAR is less than 100%. The committee discussed the implications of this on the cost-effectiveness evidence. It agreed that, although the expected cost of ongoing monitoring per patient may be lower than the model predicts, this would be counteracted to some degree because people who fail to attend scheduled scans may be more likely to experience complications that require reintervention. The committee also saw that the model conclusion did not change when assumptions were applied that were favourable to EVAR, but highly implausible, such as assuming monitoring appointments following EVAR incur no cost, or that no post-EVAR complications occur. It was therefore agreed that, while the effect of non-adherence to follow-up appointments on EVAR cost-effectiveness results is unclear, it cannot plausibly be sufficient to change conclusions drawn from the new economic model.

The committee discussed the use of the RCTs to inform much of the new model, noting that a potential criticism of the model is its use of the relatively old evidence. They noted that more recent casemix-adjusted observational evidence has closely comparable results, with no evidence that the relative difference between the approaches has changed over time (though both have got better). They agreed that this strongly validates the model’s base case approach.

When feedback was sought from stakeholders during consultation on draft guidance, a common suggestion was that, instead of using RCT data for perioperative mortality, the model should make use of current data from the National Vascular Registry (NVR). The committee discussed and firmly rejected this idea. They agreed that the NVR is probably a relatively faithful snapshot of prevailing practice; however, this means that it faithfully reflects deeply ingrained selection habits, and no NVR data available to the committee make any attempt to adjust for these. The committee also expressed concern that there is a very high risk of reporting bias in the data that get submitted to the registry (for example, which AAAs get classified as ‘infrarenal’ is very likely to vary depending on the type of repair attempted). They also noted that their concerns were validated by the review of casemix-adjusted observational evidence undertaken for this guideline. Among the 38 studies reporting perioperative mortality that attempt to provide balanced cohorts (either by randomisation or by risk-adjustment), only 1 small study has ever found that EVAR is associated with a perioperative mortality benefit of the magnitude implied by unadjusted NVR data. Therefore, the committee were convinced that it would be inappropriate to use these data for their base-case health economic model. Despite these misgivings, the committee were interested to see a sensitivity analysis in which the NVR mortality data were used. This showed that EVAR would be associated with an ICER of over £55,000/QALY.

However, there were some areas in which the committee received feedback from stakeholders during consultation that they found more persuasive, and they agreed that it was appropriate to revise the model to take advantage of more recent data. In particular, they accepted that the rate of reintervention following EVAR procedures has fallen over time. This may be, as many people claim, because endograft technology has become more durable over time. However, the committee also considered it important to note that, over this period, knowledge has developed regarding which graft complications demand revision and which can be left alone. In order to reflect this change, the model was revised to adjust the RCT data using evidence from an Italian before–after study cited by multiple stakeholders, which compares results with older-generation stent-grafts (as used in the RCTs) with more recent technology (Verzini et al., 2014).

The committee also acknowledged a common stakeholder argument that, compared with when the RCTs were undertaken, people undergoing EVAR now spend much less time in hospital in general and in critical care in particular. The committee agreed that this corresponds with their experience, too. They acknowledged that the HE modelling supporting their decision-making should ideally reflect the resource use that would be expected if the decision being simulated were adopted in present-day NHS practice. Therefore, the committee accepted a suggestion from several stakeholders that, instead of the RCTs, the model should rely on the most recent resource-use data from the NVR. The committee expressed significant misgivings about this: there is no reason to suspect that the selection biases that made them unwilling to rely on the NVR for perioperative mortality data would pose any less of a risk to resource-use data, even though the attraction of a current data source with good coverage of UK NHS activity is obvious.

In the event, this discussion was moot for infrarenal AAAs, as the NVR data are very closely comparable with the results from the EVAR-1 RCT on which the consultation draft placed reliance. The NVR data suggest that the average person undergoing EVAR spends 2.95 fewer days in critical care than the typical OSR candidate, and 6.19 fewer days in hospital overall. The equivalent differences in the EVAR-1 RCT were 2.93 days and 6.00 days. From this, the committee understood that, although perioperative length of stay has certainly decreased for people undergoing EVAR, it has decreased an almost identical amount for OSR.

The committee discussed the QALY outcomes of the model, recognising that incremental QALYs were fairly small in absolute terms (equivalent to around 3 weeks of perfect health), and the point estimate was more uncertain than for incremental costs. However, the unequivocal high incremental cost associated with EVAR led the committee to agree that the ‘true’ QALY gain for EVAR would need to be implausibly high for EVAR to be cost effective compared with open surgery (via, for example, superior long-term survival in EVAR patients, counter to the available long-term evidence). To achieve an ICER of £20,000 per QALY gained, EVAR would need to generate 0.146 QALYs more than open surgery per patient, compared with a base-case estimate of 0.056 QALYs less than open surgery. The committee were aware that modelled and empirical survival curves crossed over, with a longer-term survival benefit associated with open surgical repair offsetting its worse perioperative outcomes. The committee saw that the model suggests open surgical repair is increasingly cost-effective in younger patients, and agreed that this was consistent with its expectations, as younger people will typically be more likely to survive the open surgery procedure and experience the long-term survival benefit.

The committee discussed whether the cost-effectiveness evidence suggested that there may be differences in the balance of benefits and harms in people with different characteristics. They reviewed a series of subgroup analyses in which cohorts with different age, sex and AAA diameters were simulated. None of the preferred ICERs were qualitatively sensitive to these factors. The committee therefore determined that there was no identifiable subgroup for whom EVAR represents a reasonable use of NHS resources, so its recommendations were appropriate to the relevant population as a whole.

The committee also discussed whether the cost-effectiveness results for EVAR might be influenced by a person’s underlying life expectancy. In particular, if it were possible to identify individuals who are less likely to live to experience the long-term survival benefit associated with open surgical repair, might EVAR be a cost-effective intervention for those people? A threshold analysis was conducted in which the hazard ratio used to calibrate general population survival to ‘match’ the EVAR-1 population was varied between 1 and 15. These values indicated a relatively healthy population with a mortality hazard equal to the general population of the same age, and an extremely unhealthy population with mortality hazard of 15-times the general population, respectively. Across this range of underlying life expectancies, EVAR remained associated with ICERs substantially worse than £20,000 per QALY gained, when compared with open surgical repair.

Unruptured infrarenal AAA in people for whom open repair is unsuitable because of medical comorbidities

The committee then considered the cost-effectiveness evidence for infrarenal AAA repair in people for whom open surgical repair is not a suitable option due to medical comorbidity. This evidence comprised 1 published, UK cost–utility analysis, and modelling conducted for this guideline. The committee were aware of the extensive trial crossover that occurred in EVAR-2, from the ‘no intervention’ control arm to EVAR, and that its per-protocol analysis breaks trial randomisation in a way that is likely to bias in favour of EVAR (as it can be expected that participants who ‘crossed over’ to receive EVAR were the fittest members of the cohort, with the longest life expectancy). The committee therefore placed greater emphasis on the economic model, which adjusted for crossover using a well established statistical method (RPSFT). These data did not show any long-term survival benefit for EVAR compared with no intervention. The committee explained that many people with AAAs die with – rather than from – their aneurysms, and this would be particularly true in a population which is defined by the presence of comorbidities that are invariably life-limiting.

The committee advised that, since the population for which open surgical repair is unsuitable is defined by substantial medical comorbidity, the most appropriate analysis uses calibrated general population life tables at 1999–2001 levels; not inflating the analysis to 2015–16 lifetables, which would reflect a general increase in the health of the UK population. The committee then discussed its preferred base-case ICER for EVAR, which exceeded £430,000 per QALY gained compared with ‘no intervention’, and agreed that this indicates EVAR for this population is not an effective use of NHS resources. The committee also understood that variation of parameters to extreme values – for example, assuming no survival differences beyond 5 years, and assuming there are no EVAR graft complications at any time – do not cause the ICER to fall to a level that would be considered to represent good value for money. To achieve an ICER of £20,000 per QALY gained in this population, compared with providing no intervention, EVAR would need to generate 0.651 incremental QALYs per patient, compared with a base case estimate of 0.030 QALYs.

In consultation feedback, some stakeholders expressed uneasiness about the possible effects of crossover in EVAR-2, and raised reasonable objections about the use of the RPSFT method to adjust for it. In response to this, an extreme-case sensitivity analysis was performed, assuming that everyone who crossed over from the no intervention arm to the treatment arm of EVAR-2 would have died immediately had they not done so. This found that EVAR would be associated with a QALY gain of 0.691 and an ICER of £18,314/QALY under this extreme assumption. The committee agreed that the fact that this totally implausible scenario produces an ICER that is only just below £20,000/QALY is a very strong indication that the ‘true’ ICER must be very much higher, with no realistic prospect of representing an effective use of NHS resources.

The committee discussed whether the cost-effectiveness evidence suggests that there may be differences in the balance of benefits and harms between men and women, and older and younger people. None of the preferred ICERs were sensitive to the sex of the cohort; nor were they sensitive to differences in age or AAA size. The committee therefore determined that there was no identifiable subgroup for whom EVAR represents a reasonable use of NHS resources, so its recommendations were appropriate to the relevant population as a whole.

The committee discussed whether living with an unrepaired AAA may cause psychological morbidity for people who are not offered repair. They noted that there is no evidence of this: the EVAR-2 RCT found no significant differences in EQ-5D between people receiving EVAR and those randomised to no intervention; nor was there a detectable effect on the SF-36 mental component summary score.

Committee members reported that some patients are relieved to learn that they do not have to undergo intervention. They agreed that the information patients are given is critical. Some clinicians inform patients that they have a ‘ticking time-bomb’ inside them. However, as the people for whom surgical repair is inappropriate are subject to comorbidities that are inevitably life-limiting, it is important to provide a realistic appraisal of the competing hazards they face. The committee agreed that it would be good practice to advise people in this situation that the EVAR-2 trial showed no overall survival benefit for people receiving EVAR.

For all these reasons, a negative impact of living with an untreated AAA was not included in the base-case economic model. However, the impact of a large lifetime utility decrement of 0.1 was explored in sensitivity analysis. The ICER for EVAR compared with no intervention remained worse than £30,000/QALY. The committee agreed that this demonstrates that no plausible level of disutility could be enough to counterbalance the harms and costs associated with EVAR.

Unruptured complex AAA in people for whom open repair would be suitable

The committee discussed the cost-effectiveness evidence for the repair of unruptured complex AAA. The committee agreed that here the term ‘complex’ has a broad meaning, generally referring to non-standard AAA repairs. Typically, a complex AAA is one for which a standard EVAR device cannot be used within the terms of its instructions for use (IFU), and a complex device is one that is custom made, requiring bespoke adaptations, such as fenestrations and branches. The committee agreed that optimal decision-making for this population would be based on detailed analysis of reliable data subdividing people according to types of complex aneurysm and repair. However, with the possible exception of fenestrated EVAR (fEVAR; see below), there is a critical dearth of specific evidence in this area. Therefore, in the absence of data enabling focused analysis on different types of complex AAA, the committee agreed that it would be of value to explore more general evidence which combines experience with various types of complex AAA repair.

The committee were aware that there is no randomised comparative evidence evaluating EVAR and open repair for complex AAA. They understood that there are 2 broad approaches that can be used to estimate cost–utility results in the economic model. The first approach, which constitutes the base-case model, relies on a degree of assumption regarding the transferability of data on infrarenal AAA. The committee advised that, once a person has survived to 30 days after their intervention, survival thereafter is expected to be relatively similar to people with repaired infrarenal AAA. On this basis, the use of data for infrarenal AAA to model long-term survival was agreed to be a reasonable approach. The second possible approach would be to rely on lower-quality data from the directly applicable population, as identified in the review of casemix-adjusted observational data. This approach was pursued as a sensitivity analysis.

The committee were aware that the bespoke nature of many complex EVAR devices has implications for obtaining reliable unit costs. However, they were satisfied that an average cost obtained from 3 NHS Trusts was likely to adequately reflect a typical UK cost, significantly in excess of the cost of a standard EVAR device. Only one ‘off-the-shelf’ complex endograft appears in the NHS Supply Chain Catalogue – a fenestrated anaconda device manufactured by Vascutek. The costs of this graft are similar to those estimated from the committee data (although cases with more than 2 fenestrations cost somewhat more).

The committee reviewed the ICERs predicted by the base-case economic model for the repair of unruptured, complex AAA. The committee noted that EVAR was associated with more net QALYs than open surgery in this population, as it is predicted – using a relative effect generalised from the infrarenal setting – to have a larger absolute perioperative survival benefit than in the infrarenal population, owing to the higher underlying risk of surgery, in these patients. This means fewer patients are expected to survive to experience any long-term survival benefits of open surgery. The committee agreed that these results were plausible, though less certain than in the unruptured infrarenal population, because of the lack of directly applicable clinical evidence. However, they agreed that the magnitude of these uncertain benefits were unlikely to be sufficient to outweigh the unambiguous additional costs associated with complex EVAR compared with open surgical repair, as reflected in a base-case ICER of over £34,000 per QALY gained and a very small probability of the true figure being £20,000 or better. To achieve an ICER of £20,000 per QALY gained, complex EVAR would need to generate 0.485 QALYs more than open surgery per patient, compared with a base case estimate of 0.284 QALYs.

The committee advised that the 30-day mortality rate reported in the NVR for open repair in this population (18.4%) is very high compared with what would be expected for the open repair of an ‘average’ complex AAA in practice. They thought that the estimate for EVAR (3.5%) is more representative of current practice. They agreed that this discrepancy reflects substantial selection bias: they suggested that, in many units, the only cases that are offered open repair are those that are so anatomically complex that an endovascular approach is impossible, and it is unsurprising that those cases would be subject to a very high risk of perioperative death. The committee noted that this view is shared by the authors of the NVR report, who comment that ‘direct comparison of [EVAR versus open repair] figures is difficult and the open procedures may represent a more complex anatomical AAA to repair.’

Furthermore, the committee agreed that the weaknesses of the NVR data are likely to be exacerbated by substantial reporting biases. They suggested that the kinds of anatomical features that constitute ‘complexity’ vary between the approaches and are very likely to have been reported differently in returns to the NVR. For example, there are cases that, in an open operation, can be accomplished with an infrarenal cross-clamp, whereas the same anatomy would require a fenestrated EVAR graft, rendering the case ‘complex’ for EVAR. Biases like this would lead to the infrarenal and complex NVR results looking worse for open repair than for EVAR.

In this way, the committee concluded with confidence that, due to the selection and reporting biases underlying the NVR data, a cost-effectiveness analysis using the reported complex repair perioperative mortality rates directly would not provide a meaningful comparison of EVAR and open surgical repair. Rather, the preferred approach was to take the EVAR NVR data as the baseline mortality rate – as it more closely reflects clinical experience than the open surgical repair value – and then apply a measure of relative effect to this, derived from (infrarenal) RCT evidence, to estimate the mortality rate for open surgical repair.

The committee discussed other assumptions applied in the model. They were mindful that the use of NVR data to estimate perioperative resource-use almost certainly biased the model in favour of EVAR (for the reasons referred to above: there is no reason to suspect that the selection biases that made the committee unwilling to rely on the NVR for perioperative mortality data would pose any less of a risk to resource-use data).

The committee also discussed complication rates used in the model. They agreed that complex AAA repairs are likely to be more susceptible to subsequent complications and reintervention than infrarenal aneurysm repairs. The committee noted that a scenario analysis had been included in the model that applied a complication rate double that of infrarenal repair, and that this led to a notable increase in the ICER for EVAR versus open repair.

The committee saw that, when the new model is configured to use best-available data that are specific to complex AAAs from casemix-adjusted observational evidence, EVAR becomes massively dominated by open repair (with a lifetime health loss of more than 1.6 QALYs).

The committee also reviewed the results of an analysis that had attempted to estimate cost–utility results for fEVAR, as a specific subtype of complex AAA. This suggested that fEVAR is dominated by OSR, with somewhat worse net health effects (0.095 QALYs worse than OSR), but substantially higher costs (slightly more than £10,000), even though fEVAR benefits from the likely biased estimate of perioperative resource use for complex AAA from the NVR. They understood that the model’s strongly negative conclusion has been shared by other authors who have attempted to analyse the cost effectiveness of fEVAR compared with open repair (Michel et al., 2015, 2018; Ciani et al., 2018).

The committee were satisfied that the new model provides a reasonable prediction of the likely cost-effectiveness of EVAR in people with a complex unruptured AAA for whom surgical repair is a suitable option. However, they were cautious about the lack of directly applicable, randomised comparative evidence underlying the model, as this increases uncertainty regarding the true ICER for EVAR in this population. The committee were also mindful that the model had plausibly demonstrated that the benefits of complex EVAR may outweigh its harms, albeit at a cost that was very unlikely to be justified by any gains. The committee therefore made a recommendation that the use of EVAR in this population should be limited to the context of an RCT (that should include resource-use in its data collection), to ensure that any use of EVAR in this population provides direct, comparative clinical effectiveness and cost-effectiveness evidence. Mindful of the heterogeneity of the ‘complex’ AAA category, they added the stipulation that this research should be stratified in a way that will help to reveal any differences in the balance of benefits, harms and costs between EVAR and OSR according to AAA anatomy (at least distinguishing between juxtarenal, pararenal and suprarenal AAAs).

Unruptured complex AAA in people for whom open repair is unsuitable because of medical comorbidities

The committee then discussed complex AAA repair in people for whom open surgical repair is not a suitable option, because of concerns regarding medical comorbidity. The committee agreed that outcomes associated with complex EVAR would certainly be no better than infrarenal EVAR, and would probably be worse, whereas outcomes in complex AAA patients who receive no intervention are not likely to be different to infrarenal AAA patients who receive no intervention. The committee were also aware that bespoke EVAR devices for complex repair are more expensive than standard EVAR devices for infrarenal repair, and that the ICER for infrarenal AAA repair in this population was £430,602 per QALY gained. The committee therefore agreed that complex EVAR will be more expensive than standard EVAR and will provide health outcomes that are at best equivalent and at worst substantially less favourable, meaning there is no possibility that EVAR could be cost effective in this population compared with a strategy of ‘no intervention’. This result is clearer than in people with complex AAA for whom open surgery is a suitable option, where the base-case ICER for EVAR compared with open surgery was £34,288/QALY (described above). In this population, it is feasible that EVAR may be more likely to be cost-effective than in infrarenal cases, because AAA complexity also worsens the expected outcomes from open surgery.

The committee were aware that there is no published cost-effectiveness evidence in this population, and so the only evidence was from the economic model developed by NICE. The base-case model found EVAR to be dominated by a strategy of ‘no intervention’, though the committee recognised that the analysis had necessarily been informed by some assumptions, such as generalising long-term survival data from the EVAR-2 population, and low-quality data, namely estimating a ‘complexity effect’ from the National Vascular Registry. The estimated EVAR perioperative mortality rate of 42% was felt to be much higher than observed in clinical practice; therefore this analysis was deemed to be more speculative than the infrarenal AAA repair analyses conducted for this guideline. However, the unequivocal result of EVAR being dominated was seen to be supportive of the committee’s view that complex EVAR cannot be cost effective in this population. The committee therefore made a strong recommendation against the use of EVAR in people with a complex unruptured AAA for whom surgical repair is not a suitable option.

The committee considered whether the cost-effectiveness evidence suggests there may be differences in the balance of benefits and harms between men and women, for the elective repair of unruptured complex AAA. None of the preferred ICERs from the modelling were sensitive to the sex of the cohort; nor were they sensitive to differences in age or AAA size. The committee therefore determined that there was no identifiable subgroup for whom complex EVAR represents a reasonable use of NHS resources, so its recommendations were appropriate to the relevant population as a whole.

Equality considerations

The committee agreed that the effectiveness and cost-effectiveness evidence discussed above was compelling. However, they were also mindful of their responsibility to consider the broader context of their decision-making. NICE’s Social Value Judgements stipulate that ‘Decisions about whether to recommend interventions should not be based on evidence of their relative costs and benefits alone. NICE must consider other factors when developing its guidance, including the need to distribute health resources in the fairest way within society as a whole’ (Principle 3). Accordingly, the committee gave careful consideration to factors that might lead them to depart from decision-making that simply seeks to maximise population-level QALYs.

In particular, the committee explored whether limiting access to EVAR would result in any unfairness to identifiable groups of people with AAA.

For the comparison of EVAR and OSR, it is commonly asserted that access to EVAR is most vital for people with higher baseline risk of perioperative mortality (that is, the oldest and most comorbid). However, the committee were aware that some evidence suggests the opposite. In a subgroup analysis of the EVAR-1 cohort, Brown et al. (2007) found that EVAR only confers a perioperative survival benefit, compared with OSR, in people judged to benefit from ‘good’ fitness. In OVER, Lederle et al. (2012) found that younger participants had a significant benefit from EVAR whereas older people did not (indeed, the results were very nearly reversed).

The committee also discussed the comparison of EVAR and no intervention, in people for whom OSR is not a suitable option, in this context. It might be argued that, because the population in question is – by definition – sicker and likely to be older than the average person with AAA, it is unfair to deny such people repair, when younger, fitter people have access to OSR. The committee did not agree with this argument. They noted that their primary reason for not recommending EVAR, in this populationwas that there is no evidence that it results in better outcomes. Therefore, it cannot be said that anyone is being denied a meaningful benefit. On the contrary, the committee agreed that the many people with AAA and life-limiting comorbidities who are currently receiving EVAR are being inappropriately exposed to risk (and clinicians are imposing unnecessary opportunity costs on the NHS by doing so).

The committee discussed any potential differences between postoperative outcomes of EVAR between men and women. They agreed that, although the majority of the evidence presented was in men, the issue of whether a different balance of benefits, harms and costs could be expected in women was explored in the original economic model. These analyses found no evidence of any subgroup effects of a sufficient magnitude to overturn the results in the wider cohort. Similarly, regression analyses based on the Vascunet dataset suggest that female sex is a greater risk factor for people undergoing EVAR than it is for people having OSR (Mani et al., 2015; Budtz-Lilly et al., 2017).

Therefore, the committee were content that their recommendations did not induce any particular inequality with respect to sex, so no recommendations were made that were specific to women.

Patient choice

The committee discussed their responsibility to provide guidance that acknowledges individual patients’ differing preferences at length. The committee advised that patients often express a preference for EVAR compared with open surgical repair, typically due to the increased short-term risks associated with open surgery. In response to the consultation draft of this guideline, stakeholders made the committee aware of a small amount of research formally eliciting patient preferences regarding EVAR and open surgery (Winterborn et al., 2009; Reise et al., 2010; Faggioli et al., 2011). This evidence suggests that, when offered a choice, people tend to express a preference for EVAR over OSR. However, the committee noted that – as would be expected, given the state of available evidence at the time they were undertaken – none of these studies provided participants with information about long-term outcomes with EVAR and OSR, and certainly none referred to an excess hazard of mortality being associated with EVAR, for people who survive the initial operation. Since the committee found the evidence that there are differences in long-term survival expectation convincing, they felt this had to be a critical consideration in weighing up the benefits and harms of EVAR and OSR. Accordingly, they agreed that the results of these studies were of limited relevance to the present-day decision problem. The committee heard that, to the extent that a stated preference for EVAR reflects the priority people place on short-term benefits over long-term risks, this had been captured in the model by making use of evidence showing a larger quality of life decrement following open surgery, compared with EVAR, and by discounting health outcomes over time. The committee were also aware of evidence showing that AAA patients put much more weight on future outcomes than surgeons (Dion et al., 2017). The committee noted that, while individual choice is important in all care provided by the NHS, this did not compel them to recommend care that is not cost effective, as per Principle 5 of NICE’s Social Value Judgements. Given this, and based on its assessment of the evidence from the new economic model (and other published economic evaluations), the committee made strong recommendations that people with an unruptured infrarenal AAA for whom open surgical repair is a suitable option should be offered open surgical repair, and that EVAR should not be offered in such cases. They agreed that it would be to the benefit of the average candidate for elective AAA repair if the vascular community adopted a broader focus that puts appropriate weight on medium-to-long-term outcomes, rather than concentrating on the perioperative period.

Symptomatic AAA

The committee discussed whether it was necessary to specify AAA symptomatology in their recommendations. Although none of the RCTs included participants with symptomatic AAAs, it was noted that several of the studies identified in the review of casemix-adjusted non-randomised evidence include symptomatic (or ‘emergent’) cases. Among these, only 1 reports results for symptomatic cases though, helpfully, that is one of the few UK studies in the dataset. In univariable analysis across EVAR and OSR, Choke et al. (2012) found that symptomatic AAAs may be associated with a higher risk of perioperative death; however, at a 95% confidence level, the data are comfortably consistent with no difference (OR=1.94 [0.64 to 5.95]). The committee were not aware of any data exploring the possibility of interaction between symptomatic status and repair approach, which would be necessary to inform any specific recommendations regarding the relative benefit of EVAR and OSR, in these patients. However, as noted above, many of the studies included in the review of observational data included emergent cases, and the fact that pooled results from these studies are closely comparable to results from RCTs provides some validation for the committee’s view that the balance of benefits and harms is unlikely to be very different in such cases.

Feasibility of randomised research

As the committee had concluded that the best way to resolve uncertainty regarding the benefits, harms and costs of complex EVAR would be to limit such activity to a randomised trial, they were obliged to discuss the feasibility of such research in detail. In consultation, multiple stakeholders expressed the view that such an RCT would not be considered ethical by the vascular community, mostly because of a perceived lack of equipoise (with complex EVAR thought to be the superior option). This perception is predominantly based on unadjusted figures from the NVR, which the committee rejected as being completely unrepresentative of the results that would be expected if cases were selected at random (see above). Moreover the committee agreed that the uncertainty about the long-term effects of complex EVAR is substantial enough that, even if it could be shown that complex EVAR is associated with a large reduction in perioperative mortality, there should be real equipoise about whether any such effect translates into net health gain over a patient’s lifetime. The 1 study in the review of casemix-adjusted observational evidence that reported post-perioperative data reflected a substantial benefit for open repair over EVAR for complex AAAs. If estimates such as these were to prove accurate, open repair would be the superior approach even if the unadjusted, almost certainly biased NVR perioperative numbers were true. In this context, it is extremely difficult to see how a RCT could be considered unethical on the grounds that complex EVAR is unimpeachably superior.

The committee also noted that, if the community will tolerate a learning curve when introducing new technologies, it would be perverse not to see the benefit in empowering the workforce to provide a higher long-term standard of care when the ‘innovation’ that is required is for surgeons to refamiliarise themselves with older techniques.

In a related way, consultation elicited stakeholder feedback suggesting that a well designed registry might provide an adequate alternative to an RCT. In several cases, the NIHR-funded UK-COMPASS study, which is currently collecting observational data on juxtarenal aneurysms, was highlighted. The committee’s view was that, as current practice is subject to strong prior beliefs about the relative benefits and harms of EVAR and OSR for complex AAA, randomisation is critical to provide an unbiased estimate of comparative effectiveness. Stakeholders assert that a randomised trial is unfeasible, because the vascular community has a strong consensus that EVAR is superior to OSR, which is reflected in a strong preference to offer EVAR wherever possible. However, if this is true across the vascular community, there can be no expectation that observational evidence of current practice will provide a valid basis on which to compare EVAR and OSR for complex AAAs, no matter how carefully it is collected or how rigorously it is analysed. It is only where there is a good degree of overlap between the types of people who receive different treatments that it is even theoretically possible to isolate the independent effect of treatment on outcomes (see Faria et al., 2015). If prevalent attitudes lead to OSR being offered to a small, highly selected group of people with complex AAAs, who have little in common with the people who receive complex EVAR, it will not be possible to estimate a counterfactual, and the beliefs that led to the selection cannot be validated or disproven.

Some stakeholders also suggested in consultation that the numbers of people requiring complex repair are too low to make recruitment to a trial possible. The committee noted that the NVR reports over 2,000 complex procedures in the last 3 years, and agreed that this suggests that any such concerns are overstated.

Implementation challenges

The committee recognised that the recommendations represent a substantial change to practice and some resistance to change may be encountered. The committee were under no illusion regarding the perioperative mortality risks associated with EVAR and OSR: it is inarguable that the latter has significantly greater odds of death, probably around a threefold increase. However, the committee felt certain that it should be possible to optimise systems so that OSR, as well as EVAR, is associated with a low absolute risk of mortality. They firmly disagreed with stakeholders who suggest that returning to an OSR-led approach to AAA repair will inevitably lead to perioperative mortality levels that were seen before EVAR became the predominant mode of repair. They noted the impact of the Vascular Society’s AAA Quality Improvement Programme, the provisions of which raised standards in EVAR and OSR alike. The introduction of the National AAA Screening Programme, starting in 2008, has also led to many AAAs being diagnosed at a smaller diameter and at a younger age than would be the case if they had been left to present symptomatically or incidentally; this will also have contributed to lower perioperative risk for both procedures. The committee also argued that many general improvements in patient care have had beneficial impacts for the perioperative survival of people undergoing both EVAR and OSR. Factors such as improvements in imaging technology, better cardiovascular risk management (including increasingly widespread use of statins), improved prevention and treatment of nosocomial infections would all contribute to reducing perioperative mortality across the board. The committee noted that the 2 most recent datapoints in the supplementary review of casemix-adjusted observational evidence report perioperative mortality rates of 0.6% and 0.5% for OSR (Sugimoto et al., 2017 and Symonides et al., 2018). They considered the latter figure an especially attractive target, as it comes from a recent, countrywide database of publicly funded practice in Europe (Poland). In view of these features, the committee saw no reason why the NHS should not aspire to a similarly low level of perioperative mortality.

The committee also thought about a number of practical implementation challenges that may be encountered.

First, the committee acknowledged that the predominance of endovascular techniques for most unruptured AAAs, in recent years, could mean that the skills-mix of the current vascular workforce is ill-equipped to deal with a rebalancing of activity in favour of open repair. The committee were unconvinced by this argument. They expressed the view that modern vascular units should be competent to provide the necessary volume of open repair of AAA, and cited a small amount of evidence that supports this view (Beiles et al., 2016; Modrall et al. 2011). They also noted that the Intercollegiate Surgical Curriculum Programme’s Vascular Surgery Curriculum places greater emphasis on open repair than endovascular techniques. In consultation, stakeholders express related concerns about the implications of the guidance for recruitment and training of vascular surgeons. However the Royal College of Surgeons report a ‘high competition ratio of 14:1’ for vascular surgical trainee positions. Relatedly, the committee noted that, while service models addressing volume–outcome dynamics were explicitly beyond the scope of this guideline, it remains possible for the NHS to give consideration to an appropriate level of centralisation, if that is deemed necessary to optimise results.

Second, the committee considered the argument that, regardless of the funds that might be made available from cost savings elsewhere, there is currently a lack of capacity in the NHS – in critical care beds in particular – to handle the additional demand that the committee’s recommendations would create. The committee considered this argument to be unduly nihilistic: given their confident interpretation of the evidence that OSR is associated with better net outcomes and lower total costs than EVAR, it should not be acceptable to continue to rely on the inferior approach because it forms the basis of current capacity planning. However, the committee accepted that – at least in the short term – one possible knock-on effect of increased use of OSR (and the critical care capacity it necessitates) would be a lengthier waiting list for AAA repair. To explore this possibility, a scenario analysis was undertaken in the infrarenal economic model. One extra month was added to the OSR waiting time, making it 3 months compared with 2 for EVAR. OSR remained the dominant option. In fact, further exploration shows that the waiting list for OSR would have to be over 3 months longer than that for EVAR before EVAR would generate more lifetime discounted QALYs than OSR and over 8 months longer than that for EVAR before OSR would be associated with an ICER worse than £20,000/QALY. The committee were reassured by this analysis, as they did not believe that there is any risk of waiting lists reaching extreme lengths such as these.

Third, the committee discussed whether it will be possible to retain skilled EVAR capacity (which may be necessary for the repair of ruptured AAAs), if the elective EVAR workload is likely to reduce to near-zero. Detailed considerations are discussed in Evidence review T.