Randomized Controlled Trial

There is only a single RCT comparing surgical to endovascular treatment in CLI, which also included a cost-effectiveness analysis.² This high-quality RCT is nonetheless limited in that the endovascular treatment was nearly all balloon angioplasty, which has now been superseded by the use of stents, initially bare metal and now drug-eluting stents. Nevertheless, as the only randomized trial on the question of interest it is worth discussing in detail.

The trial was conducted at 27 hospitals in the United Kingdom. Between 1999 and 2004, 452 patients presenting with CLI (defined as rest pain or tissue loss) and who were judged to be candidates for either procedure were randomized to a surgery-first or angioplasty-first² treatment and strategy. Between 65% and 70% of patients were over age 70, about 60% of patients were male, only 20% had never smoked, 42% of patients had diabetes, about 60% of patients were on drug treatment for hypertension, one-third of patients were on statin therapy, a little over half of patients were on anti-platelet drugs, and 90% of patients had rest pain or night pain, while 74% had tissue loss. Almost all patients (99%) completed a one-year follow-up, with 74% completing a 2-year and 48% completing 3-year follow-up, at the time of the initial report.² A later report, with longer follow-up, included all but 4 patients at 3-year follow-up and 54% of patients at 5 or more years of follow-up. The primary outcomes were amputation-free survival, and secondary outcomes included all-cause mortality, health-related quality of life, and costs. Results were reported in a 2005 “interim” publication and a 2008 “final” analysis (published in 2010).¹² In both analyses, there were no statistically significant differences in amputation-free survival, all-cause mortality, or health-related quality of life. Patients receiving surgery had a lower immediate failure rate (3% vs 20%), higher 30-day morbidity (57% vs 41%), and lower 12-month reintervention rate (18% vs 26%). One-year costs were higher in patients initially treated with surgery (by about a third) than patients treated with angioplasty first, but by years 2 and 3 differences in costs were no longer statistically significant. The primary outcomes from the original 2005 publication are summarized in Tables 1, 2, and 3.

Table 1

All-cause mortality after bypass surgery and balloon angioplasty (entire follow-up).

Table 2

Comparison of Health-Related Quality of Life (HRQL) by intention-to-treat analysis at different time points from randomization.

Table 3

Comparison of use of hospital resources by intention-to-treat during first 12 months from randomization.

In the 2010 report of the 2008 “final” results, the primary difference was in the all-cause mortality outcome. Whereas prior to 2 years of follow-up there was a non-statistically significant disadvantage for surgery compared to angioplasty, after 2 years of follow-up there was a statistically significant benefit favoring surgery (adjusted hazard ratio 0.61, 95% confidence interval 0.60, .075)

In a formal cost-effectiveness analysis²⁰ the incremental cost-effectiveness ratio of the surgery-first management option was $184,492 per quality-adjusted life year (2006 dollars).

Cost-effectiveness Analysis Models

We identified 3 publications of cost-effectiveness studies evaluating surgical vs endovascular interventions in a general CLI patient population.^16,19,21 Two additional cost-effectiveness modeling studies evaluated subpopulations of CLI patients with ESRD and marginal functional status, and are discussed under Key Question 2 below. The publications varied in the treatment strategies assessed, data sources utilized, length of modeled follow-up period, and findings. The 3 studies, 2 from the US and 1 from the UK, are summarized below. ICER represents incremental cost-effectiveness ratio.

Table 4

Cost-effectiveness analysis models.

The first study, Barshes et al 2012,¹⁶ used a probabilistic Markov model to simulate the cost-effectiveness of 6 management strategies for CLI with tissue loss: (1) wound care alone (reference), (2) primary amputation, (3) bypass first with subsequent endovascular interventions, (4) bypass first with subsequent surgical interventions, (5) endovascular first with subsequent surgical interventions, and (6) endovascular first with subsequent endovascular interventions. ICERs were calculated for the latter 5 strategies relative to the most conservative strategy of local wound care with amputation as needed. Outcomes included clinical events, wound healing, functional outcomes, and QALYs, and were estimated based on a review of the published literature on patients with CLI. In reviewing the literature to determine the outcome estimates for their model, the authors do include the findings of the BASIL trial, but also include outcomes data from other trials and meta-analyses. They also explicitly acknowledge some of the factors that limit the applicability of the BASIL data, such as a very low (2%) stent use rate and lower rates of diabetes, ESRD, and infrapopliteal occlusive disease in the UK population.²² The literature on inpatient costs was deemed to be of inadequate quality, so inpatient cost estimates in the model were based on a patient-level transaction cost-accounting system using data from the authors’ institution. These were combined with estimates of outpatient costs (prostheses, nursing home care, etc) based on a review of the literature. Both clinical outcomes and costs were modeled over a 10-year period. The authors do not provide an explicit justification for their 10-year time horizon, which is the longest of the studies we reviewed, but may have been motivated by a desire to present a longer-term assessment of cost-effectiveness than existing studies in the literature.

The surgical and endovascular intervention strategies all conferred clinical benefit over wound care alone, with the most cost-effective strategy being surgical bypass with endovascular revisions ($47,738/QALY) and the least cost-effective strategy being endovascular first with endovascular revisions ($121,010/QALY). Primary amputation was dominated – that is to say, it was less effective and more costly than local wound care. The authors also present sensitivity analyses in which they vary the assumptions of their Markov model to test whether their conclusions still hold. In these sensitivity analyses, endovascular-first management became cost-effective when the initial foot wound closure rate was >37% or when procedural costs were decreased by >42%. The primary limitation of this study, as with most cost-effectiveness modeling analyses, is the accuracy of their clinical and cost estimates, which is based on a review of published studies of varying quality as well as inpatient cost data from a single institution. Notably, while their model uses a 10-year time horizon, the literature on which their outcomes estimate is based includes studies with at most 5 years of follow-up, so their outcomes from 5-10 years post-intervention are modeled and not directly observed. Also, their model estimates an annual baseline mortality of 11.7%, meaning nearly one-quarter of patients die in the first 2 years following the index procedure, limiting long-term recoupment of costs and the meaningfulness of long-term avoidance of complications. The authors do not mention or address these limitations in the discussion of their article.

The second study, Holler et al 2006¹⁹ also utilized a probabilistic Markov model to simulate the cost-effectiveness of 16 treatment strategies, all possible pairwise combinations of (1) “no treatment,” or conventional medical management, (2) prostaglandin E1 (PGE1) infusion, (3) surgical bypass, and (4) percutaneous transluminal angioplasty. As PGE1 infusion is not within the scope of this review, and neither are any strategies beginning with no treatment followed by bypass or angioplasty in subsequent years, 7 relevant treatment strategies emerge: (1) conventional medical management (medications and wound care), (2) initial angioplasty alone, (3) initial bypass alone, (4) angioplasty with endovascular revisions, (5) angioplasty with surgical revisions, (6) bypass with endovascular revisions, and (7) bypass with surgical revisions. Costs and outcomes data were obtained from a study published in 2004 on German patients from 2001, which predates the BASIL trial, and the model was run over a hypothetical 5-year period.

Relative to conventional medical management, the most cost-effective treatment strategy was initial angioplasty alone (€3,431.60/QALY), and the least cost-effective was angioplasty with surgical revisions (€4,904.66/QALY). Full results are shown in the above table. A principal limitation of this study is the inclusion of the PGE1 infusion treatment strategies in the modeling algorithm, which may alter the estimated outcomes and costs of the reference conventional medical management group. Another important limitation is the fact that the endovascular intervention studied was angioplasty alone, as opposed to more contemporary strategies of stenting and/or atherectomy. Lastly, this study faces the same limitation as the first cost-effectiveness analysis; namely, it is constrained by the accuracy of its cost and outcomes estimates, which are based on studies of German patients from more than 15 years ago. Given these significant limitations and the wide discrepancy between the cost-effectiveness estimates from this study and those of Barshes 2012, we tend to favor the conclusions reached by the Barshes 2012 study, as it is based on more recent cost data from the US and reflects modern-day endovascular techniques and materials.

The third study, Stoner et al 2008,²¹ used actual outcomes and cost data from a cohort of 381 limbs undergoing open and endovascular revascularization at the authors’ institution to create an amortized cost model for determining cost per patient-day of patency for the 2 treatment groups. Results were reported separately for the 188 limbs undergoing revascularization for CLI, and all results presented in this review refer to the CLI subgroup. Clinical outcomes, including the primary outcome of primary assisted patency, were obtained from hospital records, and costs were obtained from hospital billing data. All results were presented at 12 months from index procedure. The initial cost of the index procedure was significantly higher for open bypass ($13,277±598) than endovascular revascularization ($7,176±309), with p<0.001 for difference. However, cost per patient-day of patency at 12 months from index procedure was not significantly different between the open bypass ($210±80) and endovascular revascularization ($359±143) groups. The principal limitation of this study is that it is based on clinical outcomes and costs at a single institution, which may limit its generalizability to other sites and populations. Also, in contrast to the first 2 cost-effectiveness studies, this study only models cost-effectiveness to 1 year after the index procedure, precluding a comparison of the long-term cost-effectiveness of surgical versus endovascular revascularization.

Observational Studies

We identified 15 studies,^21,23–36 reporting 14 distinct cohorts, meeting all the eligibility requirements (Appendix G). Six were multi-institutional studies and 9 were single-institution studies. Twelve studies were from US institutions, 2 studies were from Germany, and 1 was from Austria. Nine had more than 500 patients included while 6 had 500 or fewer patients. Two of the included studies focused on the VA population, both reporting from the Veterans Affairs Western New York Healthcare System.

Most defined CLI as lower extremity ischemic rest pain with or without tissue loss (Rutherford class 4-6). Surgical interventions evaluated were lower extremity bypass using vein or prosthetic graft, above or below the knee. Endovascular interventions generally included atherectomy and balloon angioplasty with or without stent placement.

Patient demographics generally reflected the expected distribution based on the underlying disease process – with an average age in the mid-70s, over half of patients were male (except in VA studies where most patients were male), with high rates of diabetes mellitus, coronary artery disease, and high rates of previous and current tobacco use.

The quality of studies was variable (see Appendix E and F). Two were prospective and 13 were retrospective studies. While most studies included all or a representative sample of eligible patients and used medical records to assess outcomes, there were significant concerns about the imbalance of patient characteristics between the endovascular and surgical arms. For example, in the Dosluoglu et al²⁸ VA study, patients in the endovascular group were older and more likely to have diabetes mellitus and renal insufficiency. Many of the included studies attempted to control for these differences through multivariate analysis or propensity matching, but the decision of who to offer endovascular or surgical therapy was left to the surgeon and likely included a variety of unmeasured covariates such as lesion complexity, anatomic factors, and patient frailty. Many studies also included in their cohorts patients treated over an extended time period, with more surgical interventions happening earlier in their study and more endovascular interventions occurring later, making it difficult to discern the causal effect of the intervention from underlying secular changes. Finally, missing data through loss of follow-up was a significant concern for studies looking at outcomes beyond 1 year. For example, Siracuse et al³⁶ analyzed outcomes up to 3 years, but even at 1 year follow-up was 45.8% for percutaneous vascular interventions and 53.5% for lower extremity bypass. Thus, all these observational studies were judged to be at high risk of bias.

None of these studies reported cost-effectiveness per se; rather, these studies reported an effectiveness outcome (such as mortality, amputation rate, patency rate) and an utilization outcome (length of stay, readmission rate, cost). Readers will need to interpret these data in the context of their own clinical circumstances (how much an extra day of hospitalization costs, or a readmission, etc) in comparison to any observed differences in efficacy outcomes. Figures 2 and 3 present some of the results from these observation studies, Figure 2 presenting the short-term outcomes (less than 1 year) and Figure 3 presenting the long-term outcomes. Each study is included along the horizontal axis, and the points plotted are the different outcomes, each in a separate color (amputation, death, length of stay, reintervention rate, etc). Each outcome is plotted as the difference between the value reported for patients treated with surgery versus those treated endovascularly. The middle of the vertical axis is the zero line, meaning outcomes were the same in the surgery and endovascular groups. Points above this had a difference in outcomes that favored the surgery group, while points below the zero line had a difference in outcomes that favored the endovascular group. The 95% confidence intervals are included for each difference, except for length of stay, as data were not available in the original articles to calculate the 95% confidence interval. In the short-term outcomes (Figure 2), most of the point estimates (for amputation, death, length of stay, and reintervention rate) favor the endovascular group, but very few are statistically significant. In the long-term outcomes (Figure 3), there is more variation in outcomes both within and across studies, with some outcomes in some studies being statistically significant (for example, death and reintervention rate favoring the surgical group in the study by Darling).²⁶ In the longer-term outcomes, with one exception the studies reporting mortality favored treatment with surgery. Concluding a cause-and-effect relationship is premature, though, since patients treated endovascularly tended to be older by a few years compared to patients treated with surgery, and therefore may have had a shorter life expectancy regardless of treatment choice. Across studies, and considering the caveat that these studies as a group are at high risk of bias due to potential unmeasured cofounders, some trends are apparent:

• The short-term outcome length of stay (LOS) is consistently shorter in patients treated with endovascular therapy.
• The short-term outcome mortality in general favors endovascular therapy, although not statistically significant in any individual study.
• In the long-term outcomes, there is a potential signal of mortality favoring surgical treatment.
• All other outcomes, including costs, are too sparse or too inconsistent to draw even tentative conclusions.

Figure 2

Outcomes of observational studies of CLI: difference between EV and surgery group short-term (with 95% CIs).

Figure 3

Outcomes of observational studies of CLI: difference between EV and surgery group long-term (with 95% CIs).

Summary of Findings

The cost-effectiveness of surgery compared to an endovascular approach for patients who could be treated with either is not known. The only randomized trial of this comparison, which resulted in an incremental cost-effectiveness ratio for surgery at or above the thresholds normally used to categorize an intervention as cost-effective, is too dated in terms of the endovascular intervention (balloon angioplasty) and general improvements in care (for example, length of stay) to be used as a basis for conclusion about contemporary CLI care. Cost-effectiveness models find a much lower incremental cost-effectiveness ratio than that found in the randomized trial, yet these models can only be as sound as their underlying data, for which no randomized comparisons of modern therapy have been published. Observational studies of effectiveness and utilization have in general a consistent finding that the initial hospital length-of-stay is shorter for patients treated with endovascular therapy, and similar (or even better) short-term outcome, such as 30-day mortality, but there are signals that longer-term outcomes like mortality and patency may favor surgical therapy.

With regard to length of stay, given that the 1 RCT found shorter LOS for patients treated endovascularly and it is a consistent finding in observational studies, and the finding is compatible with what we know about the need for in-hospital care for the 2 treatments, and that in CVD these differences between surgery and percutaneous coronary interventions in LOS also exist, we judge the certainty of evidence as high for the conclusion that endovascular therapy has a lower initial length of stay.

For short-term mortality, we judge the certainty of evidence as low that endovascular therapy has lower short-term mortality than surgical therapy: the RCT is too dated to be of much value, and the observational studies are consistent but at high risk of bias.

For the long-term outcome of mortality, we judge the certainty of evidence to be very low that surgical therapy has lower long-term mortality than endovascular therapy: there is a signal in the observational studies, and there is a statistically significant benefit in the 1 RCT, but these are subject to the same reservations about the indirectness of the RCT.

As the differences between groups have not been large (although they could still be very clinically important), without randomized data about the differences in effectiveness it is impossible to draw strong conclusions. It is likely that cost-effectiveness will vary by the time horizon, analogous to that seen for percutaneous coronary interventions compared to open revascularization, where initial outcomes and utilization tend to favor percutaneous interventions, but longer-term outcomes tend to favor open revascularization.

Certainty of Evidence for Key Question 1

We judged the certainty of evidence for the outcome of cost-effectiveness as low, meaning we expect that future research to substantially change the estimate of the effect. The 1 randomized trial comparing these therapies was judged go be at low risk of bias, but to have serious limitations in terms of its directness and applicability to modern care. The observational studies were judged to be much more applicable to modern care, but to have serious limitations in terms of their risk of bias. The cost-effectiveness models can only be as strong as the underlying evidence for effectiveness. Thus, we are unable to draw strong conclusions about the cost-effectiveness of these different therapies. For the other outcomes, we judged the certainty of evidence for LOS as high, for short-term mortality as low, and for all other outcomes as very low.

Table 5

Certainty of Evidence for Key Question 1.

Randomized Controlled Trial

A study published in 2017 evaluated the subset of patients in BASIL who underwent infrapopliteal interventions.¹⁵ This subset included 104 patients, of which 56 were randomized to vein bypass surgery and 48 to balloon angioplasty. Balloon angioplasty was associated with a 32% lower amputation-free survival and 40% lower overall survival over the 5 years post-intervention, although these differences were not statistically significant. For secondary outcomes, the investigators found higher 30-day morbidity in the vein bypass group (RR 2.86, p=0.01; driven by higher rates of wound infections, sepsis and cardiovascular complications), but no difference in 30-day mortality. Improvements in ankle brachial pressure indices were similar between the 2 groups, although patients receiving vein bypass had shorter time to heal, tissue loss, and more frequent relief of ischemic rest pain. The median LOS for the index hospitalization was longer in the vein bypass group (18 vs 10 days (p<0.001)), but there was no difference in total hospital days between randomization and the primary endpoint (43.5 days vs 42 days). The primary limitation of this study is that it is a subanalysis of a randomized trial and therefore is underpowered for both the primary and secondary outcomes. It continues to be limited by the same weaknesses as the original study – namely that the endovascular arm only included balloon angioplasty and not contemporary methods/materials.

Cost Effectiveness Analyses

We identified 2 cost-effectiveness analyses that focused on a subpopulation of patients with CLI.^18,37 Both used probabilistic Markov modeling strategies, relied on estimates from the existing literature, and were generated by the same group that published the cost-effectiveness analysis in the general CLI population (above). They compared the same 6 management strategies: (1) wound care alone, (2) primary amputation, (3) bypass first with subsequent endovascular interventions, (4) bypass first with subsequent surgical interventions, (5) endovascular first with subsequent surgical interventions, and (6) endovascular first with subsequent endovascular interventions. Brief summaries are included in the table below.

The first study evaluated ambulatory, independent patients with nonhealing ulcers and concomitant end stage renal disease (ESRD). The authors made numerous changes from their original model in Barshes 2012 based on a review of the existing literature on patients with CLI and ESRD, including an increased baseline and perioperative mortality, lower 1-year limb salvage and wound healing rates, as well as reduced probability of remaining ambulatory/independent. They did not evaluate quality of life, and instead used years of ambulation as their denominator. They continued to use internal costs from a single health system, but updated values to 2011 US dollars. Over a 10-year time horizon, total median costs were highest for primary amputation ($152k) and lowest for wound care alone ($118k), with all of the remaining strategies in a relatively narrow margin (between $121k and $128k). Compared to local wound care alone, endovascular-only therapies had an incremental cost-effectiveness ratio of $15,043/additional year of ambulation, followed by $40,594/year for the endovascular/surgery strategy, and >$70k for both surgery-first strategies; primary amputation was “dominated” (less effective and more costly than local wound care alone). The primary limitation of this study is the validity of the numeric estimates included in their model. None of their data points came from randomized data and several of their assumptions appeared anecdotal; for example, the “periprocedural mortality for endovascular intervention was estimated to be 75% of [that of surgical intervention]”.

The second study evaluated independent patients with nonhealing ulcers but with marginal functional status, defined as patients who were 81 years or older and/or had a prior major amputation of the contralateral limb. The authors again made numerous changes from their original model based on a review of the literature, including an increase in baseline and procedural mortality, increased risk of discharged to skilled nursing, and reduced probability of ambulation and independence. They did not evaluate quality of life, and instead used years of ambulation as their primary denominator. They maintained the same limb salvage rates for this analysis as in their original CLI model. They used internal costs updated to 2011 US dollars. Over a 10-year time horizon, total median costs were highest for primary amputation ($186k) and lowest for initial endovascular followed by subsequent endovascular interventions ($104k). Wound care alone fell in the middle ($129k), and the remaining interventional strategies ranged from $108k to $114k. They did not provide ICERs, likely because the cheapest strategy (endovascular first with subsequent endovascular interventions) also provided the highest median years of limb salvage and median ambulatory years, therefore dominating all other strategies. This finding persisted in a sensitivity analysis in which they assumed increased rates of limb loss in this marginal population. Limitations are similar to those above, namely that their analysis relies upon the validity of existing point estimates, none of which were generated from randomized trials.

Table 6

Cost-effectiveness analysis models.

Cohort Study

We identified 3 cohort studies of interest. One focused on patients with tibial (infrapopliteal) disease,³⁸ the second compared patients with and without ESRD undergoing intervention for CLI,³⁹ and the third compared patients with insulin dependent diabetes and non-insulin dependent diabetes to patients with no diabetes.⁴⁰

The first was a small Canadian single-center retrospective review of patients with Rutherford stage 4 or 5 CLI undergoing tibial angioplasty who were deemed “high-risk” for surgical intervention, defined as those with an ASA of 3 or greater. The study included patients from 2001 to 2007 and the study’s overall mean follow-up was 7.7 months. The study was primarily descriptive without a comparative component, but they did make a cost comparison to a “control cohort of surgical patients, considered high-risk candidates (ASA ≥ 3) undergoing elective femoral tibial bypass for the same indications at the same period in our institution.” Cost data were collected from an internal platform, excluding “medical fees”, and converted to USD. The year of cost analysis was not stated. Of 45 patients included in the endovascular arm of the study, they had cost information on 26 subjects, with an average hospital cost of $2,910.60 compared to 32 surgical patients with an average hospital cost of $17,703.50. They also noted a less than 1-day LOS for the endovascular group and a 9-day LOS for the surgical group. Limitations include the small sample, confounding from patient and time differences between the endovascular and surgical groups, sampling bias from missing data, lack of transparency with respect to what costs were included, and inadequate description of methods to deal with changing costs over time.

Table 7

Evidence table for “Tibial angioplasty for limb salvage in high-risk patients and cost analysis” (Werneck 2009).

The second study evaluated ESRD as part of the larger German Registry of First-line Treatments in Patients With Critical Limb Ischemia (CRITISCH) registry described in the Bisdas et al studies^23,24 in the previous section. The analysis was limited to in-hospital outcomes. They compared 102 patients with ESRD to 674 patients without ESRD. They omitted patients with reduced, but not end-stage, renal function. Patients with ESRD were more likely to be male and more likely to have coronary heart disease and diabetes, but less likely to have current tobacco use. ESRD patients were also more likely to have Rutherford 6 disease, but did not differ from non-ESRD patients in TASC lesion type. ESRD patients in this registry were more likely to undergo endovascular procedures compared to non-ESRD patients. Despite the imbalance of procedures, ESRD patients had similar median hospital stays and median ICU stays. On multivariate analysis, controlling for gender, Rutherford class, and treatment strategy, ESRD patients had an increased risk of amputation or death (as a composite) and hemodynamic failure compared to non-ESRD patients. Treatment strategy (ie, open vs endovascular) was not predictive of death or hemodynamic failure, although patients undergoing open surgery had slightly higher odds (OR 1.74, p=0.04) of reintervention. Limitations of the study include the small sample (especially among ESRD patients undergoing surgical procedures (n=13)), the focus on only in-hospital outcomes, and the inadequate adjustment for risk differences in ESRD and non-ESRD patients.

Table 8

Evidence table for “In-hospital outcomes in patients with critical limb ischemia and end-stage renal disease after revascularization” (Meyer 2016).

The third study was an assessment of all lower-extremity interventions for CLI at one institution between 2005 and 2014. Patients were classified as having no diabetes, non-insulin dependent diabetes (NIDDM), and insulin dependent diabetes (IDM). Among 1,294 patients assessed, more than half (703) had NIDDM, and the numbers of patients in the other 2 categories were 329 and 262. Patients with diabetes of either type were younger than patients without diabetes, less likely to have rest pain, more likely to have CVD, and less likely to be smokers. TASC classification in general did not differ, except patients without diabetes were more likely to have TASC D femoropopliteal disease. Patients with IDDM were more likely to have incomplete wound healing when treated with either surgery or endovascular therapy, while the composite outcome of reintervention, amputation or stenosis was more likely in patients with IDDM treated with endovascular therapy.

Table 9

Evidence table for “Outcomes after first-time lower extremity revascularization for chronic limb-threatening ischemia in insulin-dependent diabetic patients” (Darling 2018).

Summary of Evidence

There is insufficient evidence to assess whether surgery versus endovascular therapy may be preferred in certain populations or settings.

The only randomized data evaluated patients with infrapopliteal disease and found that endovascular therapy may have worse long-term outcomes, but the study was underpowered and did not include contemporary materials/methods. As with the larger trial, they found increased short-term utilization in the surgical group but similar utilization between groups over longer time horizons. The one cohort study similarly found increased utilization in the surgical group for the in-hospital period but did not provide long-term data.

Patients with ESRD undergoing treatment for CLI likely have worse overall outcomes than patients without ESRD, such as increased risk of amputation, death, and hemodynamic failure. However, the one observational study in this domain did not find an independent effect of treatment strategy on these outcomes. Diabetes also has a deleterious influence on all outcomes, and one observational cohort found that patients with insulin dependent diabetes had a higher risk of the composite outcome of reintervention, amputation, and stenosis. A cost-effectiveness model found lower costs per year of ambulation with endovascular-first approaches compared to surgery-first, but is again limited by the quality and quantity of data informing the underlying parameter estimates, none of which are derived from a randomized trial.

Finally, a cost-effectiveness model among patients with borderline functional status also favored endovascular-first approaches over surgery-first. However, differences in both the numerator (costs) and denominator (number of ambulatory years) among the various strategies were very small. As a result, even small changes to these point estimates may markedly alter conclusions in the future.

Certainty of Evidence for Key Question 2

We judged the certainty of evidence for the outcome of cost-effectiveness varying in certain populations as very low, meaning we cannot even estimate and effect, with one exception: we judge the certainty of evidence is low that endovascular therapy will be less cost-effective than surgery in infrapopliteal disease, based on the evidence from the one RCT suggesting possibly worse outcomes for endovascular therapy in such patients.