Comparison of lower extremity bypass and peripheral vascular intervention for chronic limb-threatening ischemia in the Medicare-linked Vascular Quality Initiative

Abstract

Objective:

There is a relative lack of comparative effectiveness research on revascularization for patients with chronic limb-threatening ischemia (CLTI). We examined the association between lower extremity bypass (LEB) vs peripheral vascular intervention (PVI) for CLTI and 30-day and 5-year all-cause mortality and 30-day and 5-year amputation.

Methods:

Patients undergoing LEB and PVI of the below-the-knee popliteal and infrapopliteal arteries between 2014 and 2019 were queried from the Vascular Quality Initiative, and outcomes data were obtained from the Medicare claims-linked Vascular Implant Surveillance and Interventional Outcomes Network database. Propensity scores were calculated on 15 variables using a logistic regression model to control for imbalances between treatment groups. A 1:1 matching method was used. Kaplan-Meier survival curves and hierarchical Cox proportional hazards regression with a random intercept for site and operator nested in site to account for clustered data compared 30-day and 5-year all-cause mortality between groups. Thirty-day and 5-year amputation were subsequently compared using competing risk analysis to account for the competing risk of death.

Results:

There was a total of 2075 patients in each group. The overall mean age was 71 ± 11 years, 69% were male, and 76% were white, 18% were black, and 6% were of Hispanic ethnicity. Baseline clinical and demographic characteristics in the matched cohort were balanced between groups. There was no association between all-cause mortality over 30 days and LEB vs PVI (cumulative incidence, 2.3% vs 2.3% by Kaplan Meier; log-rank P-value = .906; hazard ratio [HR], 0.95; 95% confidence interval [CI], 0.62–1.44; P-value = .80). All-cause mortality over 5 years was lower for LEB vs PVI (cumulative incidence, 55.9% vs 60.1% by Kaplan Meier; log-rank P-value < .001; HR, 0.77; 95% CI, 0.70–0.86; P-value < .001). Accounting for competing risk of death, amputation over 30 days was also lower in LEB vs PVI (cumulative incidence function, 1.9% vs 3.0%; Fine and Gray P-value = .025; subHR, 0.63; 95% CI, 0.42–0.95; P-value = .025). There was no association between amputation over 5 years and LEB vs PVI (cumulative incidence function, 22.6% vs 23.4%; Fine and Gray P-value = .184; subHR, 0.91; 95% CI, 0.79–1.05; P-value = .184).

Conclusions:

In the Vascular Quality Initiative-linked Medicare registry, LEB vs PVI for CLTI was associated with a lower risk of 30-day amputation and 5-year all-cause mortality. These results will serve as a foundation to validate recently published randomized controlled trial data, and to broaden the comparative effectiveness evidence base for CLTI.

Chronic limb-threatening ischemia (CLTI), the most severe expressions of peripheral arterial disease (PAD), presents as resting ischemic leg pain or tissue loss. About 2 million individuals in the United States are affected by CLTI,^1–3 and admissions for PAD overall have been on the rise, particularly among younger populations under 65 years of age.^4,5 CLTI manifests with complex, multiple comorbidity profiles, making its management challenging, with high mortality and amputation risk associated with this disease. Following revascularization, long-term mortality rates range from 37.7% to 63.5%, and amputation rates are between 12.1% and 67.3%.^6,7

In addition to aggressive risk factor modification, revascularization remains a mainstay to realize symptom improvement and limb salvage. Risk-benefit ratios differ for lower extremity bypass (LEB) and peripheral vascular intervention (PVI) have been evolving. Prior comparative effectiveness work demonstrated no differences in outcomes when comparing LEB with PVI.^8–10 More recently, the Best Endovascular vs Best Surgical Therapy in Patients with Critical Limb Ischemia (BEST-CLI) trial demonstrated a lower risk of major adverse limb events for LEB compared with PVI.⁷ It remains unclear, however, if the benefit seen for the reduction of adverse limb events can be replicated in a real-world cohort of patients eligible for both procedures, and when considering the competing risk of death, given high mortality rates.^6,7

Given prior comparative effectiveness research and updated randomized controlled trial data,^7,11,12 there is an urgent need for complementary real-world evidence. We therefore aimed to conduct a propensity-matched analysis using a real-world Vascular Quality Initiative (VQI) Medicare claims-linked cohort for patients undergoing LEB vs PVI for CLTI. Specifically, we assessed the association between LEB vs PVI for each of (1) 30-day all-cause mortality and 5-year all-cause mortality, as well as (2) 30-day and 5-year amputation outcomes, taking into account the competing risk of death.

METHODS

Study population.

Vascular procedures from the PVI and infrainguinal LEB modules from the VQI registry¹³ were included from over 900 sites in North America. Demographic, clinical, and procedural data elements were abstracted from patients’ medical records and were entered into the VQI by trained data collectors, with audits demonstrating errors in less than 2% of entered data.¹⁴ VQI registry data was linked with Medicare claims data through the Vascular Implant Surveillance and Interventional Outcomes Network (VISION) platform, providing accurate long-term outcome data.¹⁵ Direct and indirect linkage methods, with 93% matching results and >99% accuracy rate, were used to merge the VQI data with the Medicare data.^16,17

Inclusion criteria were: (1) index procedures for LEB or PVI for CLTI between January 1, 2014, and December 31, 2019; (2) age ≥18 years; (3) treatment to below the knee popliteal or tibial infrainguinal disease; (4) Rutherford classification IV to VI; and (5) Medicare eligibility. Exclusion criteria were: (1) patients with high rates of missing data (the upper limit for high rates of missing data was defined as the third quartile +1.5 × interquartile); (2) patients with concomitant LEB, PVI, or endarterectomy; (3) wheelchair-bound or bedridden preprocedural ambulatory status; (4) urgent or emergent indications; and (5) patients with one or more missing characteristics that were used for the propensity score calculation. This study was deemed exempt by the Yale University Institutional Review Board because of the deidentified nature of VQI data.

Exposure.

LEB included any procedure with surgical graft destinations including the popliteal artery below the knee, tibioperoneal trunk, anterior tibial artery, posterior tibial artery, dorsalis pedis artery, and plantar arteries. The graft destination site was used to define lesion location as it has been used similarly in prior studies within the VQI as angiographic data is not available.^18,19 These distal graft destinations were chosen to avoid the inclusion of isolated superficial femoral artery disease while maximizing cases of popliteal and below-the-knee disease, including multivessel disease. Conduits used for LEB included single-segment saphenous vein grafts, which is the gold standard in bypass surgery, as well as alternative autologous and prosthetic conduits.^1,10

Our primary analysis included patients who received all conduits. We additionally derived a cohort consisting of patients undergoing LEB with an optimal conduit and those undergoing PVI. An optimal conduit was defined as a single-segment greater saphenous vein graft.

PVI included any procedure on the popliteal, tibial, or inframalleolar arteries, and included procedures using plain balloon angioplasty, stents, drug-coated balloons, cutting balloons, and atherectomy. Isolated superficial femoral artery disease was not included, and we were unable to determine whether cases with multivessel disease involving the superficial femoral artery were included due to coding and variable design in the VQI registry. PVI location was based on data abstracted from the medical record as recorded in the VQI registry.

Outcomes.

Thirty-day and 5-year outcomes included all-cause mortality and amputation. All-cause mortality was derived from the Centers for Medicare and Medicaid Services vital status files. Amputation was also derived from the Centers for Medicare and Medicaid Services, and includes transmetatarsal amputations, amputation at the level of the femur, and amputation at the level of the tibia or fibula. Although prior randomized controlled trials have defined major amputation to include only above-ankle amputations,^7,12 we were unable to differentiate transmetatarsal amputations from above-ankle amputations based on available data. We therefore chose to include transmetatarsal amputations in our endpoint. Patients were followed up to 5 years after the index procedure or until December 31, 2019, or death, whichever occurred first. Detailed information on Current Procedural Terminology codes, subcategorization, and definitions of amputations are provided in Supplementary Table I (online only).

Statistical analysis.

A comprehensive description of the statistical analysis is included in the Supplementary Appendix (Detailed Methods, online only). As our primary analysis for the LEB vs PVI comparisons, propensity score matching with complete cases was used to balance patient baseline characteristics between the treatment groups. Propensity matching provides the benefit of ease of interpretation, at the potential expense of needing to exclude unmatched cases.²⁰

Propensity scores, the probability of being assigned to LEB vs PVI using PVI as the reference, were calculated using logistic regression based on 15 variables, including region of the facility performing the procedure as well as patient characteristics previously shown to be predictors of morbidity, mortality, and functional status in CLTI (overview of patient-level factors are presented in the Table).^21–26 The preprocedural living at home variable was not included in the propensity score. We excluded bedbound or wheelchair-bound patients, and hypothesized that, as we had selected for a more functional population, there would not be significant variability between groups with this variable. Using the propensity scores, patients in each group were 1:1 matched to the nearest neighbor within a caliper width of 0.2 standard deviations of the logarithm of the propensity score and without replacement.²⁷ Distributions and kernel densities of the propensity score were inspected before and after matching (Supplementary Fig, online only). Patient and procedural characteristics before and after matching were described as means and standard deviations, as medians and interquartile ranges for continuous variables, and as frequencies and percentages for categorical variables. Standardized differences were calculated to evaluate the effect sizes of the differences between groups. An absolute standard difference lower than 0.1 and 0.2 was considered, as negligible and small, respectively.²⁸

Table.

Baseline characteristics of the overall complete-case cohort and by lower extremity bypass (LEB) vs peripheral vascular intervention (PVI) before and after propensity matching

	Before propensity matching			After propensity matching
	PVI (n = 8487)	LEB (n = 2105)	Standardized difference	PVI (n = 2075)	LEB (n = 2075)	Standardized difference
Demographics
Age,a years	73 (11)	71 (10)	0.19	71 (11)	71 (10)	0.01
Age, years	73 [66–82]	71 [65–78]		72 [65–79]	71 [65–78]
Male sexa	5299 (62)	1467 (70)	0.15	1420 (68)	1438 (69)	0.02
Racea			0.15		0.02
White	6169 (73)	1617 (77)		1581 (76)	1590 (76)
Black or African American	1720 (20)	368 (18)		372 (18)	365 (18)
Asian	107 (1.3)	22 (1.0)		21 (1.0)	22 (1.1)
Other or unknown	421 (5.0)	90 (4.3)		84 (4.0)	90 (4.3)
Hispanic ethnicitya	510 (6.0)	139 (6.6)	0.02	130 (6.3)	137 (6.6)	0.01
Preprocedural living location at home	7968 (94)	2032 (97)	0.12	1990 (96)	2002 (97)	0.02
Primary insurera			0.06		0.04
Medicare	8065 (95)	2018 (96)		1983 (96)	1988 (96)
Medicaid	51 (0.60)	16 (0.80)		11 (0.50)	16 (0.80)
Commercial	335 (3.9)	65 (3.1)		73 (3.5)	65 (3.1)
Medical history
Body mass index,a kg/m2	28 (6.4)	28 (6.2)	0.08	28 (6.5)	28 (6.2)	−0.01
Body mass index, kg/m2	27 [24–32]	27 [24–31]		27 [23–31]	27.0 [24–31]
Hypertensiona	7792 (92)	1901 (90)	0.05	1879 (91)	1872 (90)	0.01
Diabetesa	6009 (71)	1228 (58)	0.26	1208 (58)	1224 (59)	0.02
Smoking statusa			0.37		0.03
Never	3546 (42)	607 (29)		607 (29)	606 (29)
Prior	3614 (43)	885 (42)		861 (42)	884 (43)
Current	1327 (16)	613 (29)		607 (29)	585 (28)
Chronic kidney disease	4444 (55)	820 (41)	0.27	878 (45)	815 (42)	0.06
On dialysisa	1819 (21)	234 (11)	0.28	235 (11)	234 (11)	0.002
Coronary disease (CAD, CABG, and/or PCI)a	4174 (49)	927 (44)	0.10	912 (44)	923 (45)	0.01
Chronic lung diseasea	1709 (20)	501 (24)	0.09	496 (24)	488 (24)	0.01
Congestive heart failurea	2575 (30)	429 (20)	0.23	442 (21)	428 (21)	0.02
Prior amputation	2118 (25)	395 (19)	0.15	435 (21)	393 (19)	0.05
Rutherford classificationa			0.36		0.02
IV: ischemic rest pain	1480 (17)	693 (33)		683 (33)	663 (32)
V/VI: minor/major tissue loss	7007 (83)	1412 (67)		1392 (67)	1412 (68)

CABG, Coronary artery bypass graft; CAD, coronary artery disease; PCI, percutaneous coronary intervention.

Data are presented as number (%), mean (standard deviation), or median [interquartile range].

^aCharacteristic used in the propensity score.

Kaplan-Meier curves by LEB vs PVI were constructed for all-cause mortality over 30 days and 5 years and were compared using the log-rank test.

After confirming that the assumption of proportionality of hazard was met using Schoenfeld residuals,²⁹ two hierarchical Cox regression models, with a random effect for center and operator (nested in center), were constructed to estimate the hazard ratio (HR) and 95% confidence interval (CI) associated with LEB vs PVI for 30-day and 5-year all-cause mortality outcomes.

As conventional Kaplan-Meier and Cox regression methods may lead to overestimation of the risk of amputation in the presence of a competing event like death, 30-day and 5-year cumulative incidence function curves by LEB vs PVI were constructed for amputation, taking into account all-cause mortality as a competing risk, and were compared using the Gray test for our primary analysis for amputation.³⁰ Two competing risk models were constructed, as proposed by Fine and Gray, to derive unbiased HRs, called sub-hazard ratio (sHR) with 95% CI for the 30-day and 5-year amputation outcomes.^31,32

To verify the robustness of our results, we performed three sensitivity analyses. First, we replicated the Kaplan-Meier, Cox regression, and Fine and Gray models using overlap weighting of the propensity score on the complete case cohort, which addresses the concern of excluding patients in the propensity-matched analyses and allows for wider generalizability and better precision, while accounting for extreme propensity scores without trimming.^33,34

Second, to address the issue of missing data in predictors for calculating the propensity score, missing data were imputed using multiple imputation method by chained equations generating 10 imputed datasets.³⁵ Propensity scores were calculated as previously described on each of the imputed data sets and were pooled using Rubin’s rule to replicate the primary analysis.³⁶

Lastly, we replicated the Kaplan-Meier, Cox regression, and Fine and Gray models on the propensity-matched cohort while excluding those patients undergoing LEB without an optimal conduit. This cohort allowed for direct comparison between LEB using an optimal conduit with PVI, as saphenous vein bypass grafts have been shown to be superior to prosthetic grafts and alternative autologous conduits.^1,10,37

E-value analysis was used to evaluate for unmeasured bias.³⁸ A detailed description of E-values is included in the Supplementary Appendix (online only). Detailed methods for justification of a sample size are also included in the Supplementary Appendix (online only).

Analyses were performed with STATA version 17 (Stata-Corp. 2021. Stata Statistical Software: Release 17), and the R package, coxme, was used for mixed Cox proportional hazard modeling (version 2.2–16, R Foundation for Statistical Computing).³⁹ Due to confidentiality reasons, cell sizes smaller than 11 were masked.

RESULTS

The complete-case cohort before matching included 10,592 patients (10,592 total procedures), of whom 8487 patients underwent PVI (80%) and 2105 patients underwent LEB (20%). The propensity-matched cohort included 2075 patients in each group (Fig 1).

Fig 1.

Flow chart for the propensity-matched cohort. CLTI, Chronic limb-threatening ischemia; LEB, lower extremity bypass; PVI, peripheral vascular intervention.

In the propensity-matched cohort, the mean age was 71 ± 11 years, 68.9% were male, and 76% were white. Overall, 90% of patients were hypertensive, 57% had diabetes, 71% were active or former smokers, 44% had coronary artery disease, and 43% had chronic kidney disease, with 11% who were dialysis-dependent. Most patients in the cohort presented with tissue loss (68%). Demographics, medical comorbidities, functional living status, and severity of peripheral vascular disease were balanced between groups, with standardized differences less than 0.1 (Table).

Procedural characteristics are presented for LEB and PVI in Supplementary Table II (online only) and Supplementary Table III (online only). Most patients undergoing LEB received a saphenous vein graft conduit (90%), whereas prosthetic grafts were used as conduits in 4.1% of patients. The remainder of conduits were alternative autologous conduits or composite grafts. For patients undergoing PVI, plain balloon angioplasty was used most frequently (88%), either alone or in combination with another method.

Missing covariate information used in the propensity score ranged from 0% to 0.4% before the complete-case cohort was derived. Missing demographic variables ranged from 0% to 5.5% (for chronic kidney disease) in the propensity score-matched cohort. The median follow-up was 19.8 months (interquartile range, 7.9–37.9 months), and 411 patients (9.9%) were lost to follow up at 5 years.

For the primary results, there was no association between all-cause mortality over 30 days and LEB vs PVI (cumulative incidence, 2.3% vs 2.3% by Kaplan Meier; log-rank P-value = .906; HR, 0.95; 95% CI, 0.62–1.44; P-value = .80; E-value for HR 1.29 and lower limit of CI 1.00) (Fig 2, A). All-cause mortality over 5 years was lower for LEB vs PVI (cumulative incidence, 55.9% vs 60.1% by Kaplan Meier; log-rank P-value < .001; HR, 0.77; 95% CI, 0.70–0.86; P-value < .001; E-value for HR 1.69 and lower limit of CI 1.60) (Fig 2, B). At 30 days, 93 patients (2.2%) had died without an amputation. At 5 years, 1162 patients (28.0%) had died without an amputation. Amputation over 30 days was also lower in LEB vs PVI (cumulative incidence function, 1.9% vs 3.0%; Fine and Gray P-value = .025; subhazard ratio [sHR], 0.63; 95% CI, 0.42–0.95; P-value = .025; E-value for HR 2.55 and lower limit of CI 1.29) (Fig 3, A). There was no association between amputation over 5 years and LEB vs PVI (cumulative incidence function, 22.6% vs 23.4%; Fine and Gray P-value = .184; sHR, 0.91; 95% CI, 0.79–1.05; P-value = .184; E-value for HR 1.34 and lower limit of CI 1.00) (Fig 3, B).

Fig 2.

A, Kaplan-Meier curves by lower extremity bypass (LEB) vs peripheral vascular intervention (PVI) for 30-day all-cause mortality (A) and 5-year all-cause mortality (B) in the propensity matched cohort (n = 4150).

Fig 3.

Cumulative incidence function by lower extremity bypass (LEB) vs peripheral vascular intervention (PVI) for 30-day amputation (A) and 5-year amputation (B) in the propensity matched cohort (n = 4150).

The results from the three sensitivity analyses were concordant with those of the primary analysis. For the sensitivity analysis on the complete-case cohort excluding patients undergoing LEB without a single-segment greater saphenous vein graft conduit, there was no association between all-cause mortality over 30 days and LEB vs PVI (cumulative incidence, 2.3% vs 2.3% by Kaplan Meier; log-rank P-value = .904; HR, 0.90; 95% CI, 0.58–1.39; P-value = .630) and; all-cause mortality over 5 years was also lower for LEB vs PVI (cumulative incidence, 55.2% vs 60.1% by Kaplan Meier; log-rank P-value < .001; HR, 0.75; 95% CI, 0.67–0.84; P-value < .001). Amputation over 30 days was similarly lower in LEB vs PVI (cumulative incidence function, 1.9% vs 3.8%; Fine and Gray P-value = .001; sHR, 0.49; 95% CI, 0.33–0.74; P-value = .001), and there was no association between amputation over 5 years and LEB vs PVI (cumulative incidence function, 21.6% vs 23.0%; Fine and Gray P-value = .0074; HR, 0.86; 95% CI, 0.74–1.01; P-value = .0074).

DISCUSSION

In this contemporary, “real-world,” propensity score-matched cohort of patients with CLTI, using national quality registry-linked Medicare claims data, there was no association between 30-day all-cause mortality outcomes and LEB vs PVI. Long-term mortality risk for LEB vs PVI was lower, with an absolute difference of 4.2% and relative risk reduction over 20%. Amputation risk, taking into account the competing risk of death, for LEB vs PVI was lower at 30 days with a small absolute difference of 1.1% and a relative risk reduction of nearly 40%. No difference for 5-year amputation risk was found between the two procedures, taking into account the long-term mortality risk.

This study builds on prior work by providing the largest contemporary comparative effectiveness analysis for LEB vs PVI in a real-world cohort to date, using a well-balanced, propensity score-matched cohort with high-quality long-term outcomes data using a Medicare claims-linked database. In addition to high-quality outcomes data, robust methodology was used in our analysis, including competing risk analysis and multiple sensitivity analyses. Although we chose to use propensity matching for ease of interpretation, we complemented this with the use of overlap weighting in sensitivity analysis, which accounts for extreme propensity scores without trimming.^33,40 Leveraging real-world data to complement comparative effectiveness evidence for the management of CLTI in more homogenous populations is critical to help contextualize prior and current randomized controlled trial efforts, as inclusion criteria are more strict, and results did not account for the competing risk of death in this high-risk population.⁷ Much clinical equipoise for patients with infrainguinal CLTI exists as they are weighing their options as candidates for both procedures, and guidelines do not advocate for one option over the other.¹ Yet, despite advances in technology, overall mortality for CLTI remains high, with the current study showing a mortality rate of nearly 60% at 5 years. Generalizability and validation of prior trial results in real-world cohorts are needed to help support preference-sensitive and tailored medical decision-making approaches for revascularization options in CLTI.

Although 30-day mortality risk was comparable for LEB vs PVI, the risk of mortality was lower for the 5-year outcomes. Potential explanations as to why patients in the surgical group had better survival outcomes are better surveillance and associated quality of care and risk management (including graft surveillance and use of goal-directed medical therapy), and the potential for unmeasured selection bias, which allowed for more “fit” patients to be triaged for surgical revascularization. This hypothesis, specifically the potential difference in more aggressive long-term surveillance in the surgical group vs the endovascular group in terms of overall risk management, warrants further mechanistic outcomes studies.

Thirty-day amputation risk was lower for LEB compared with PVI, but there was no difference in amputation over 5 years between groups. There was no difference in long-term amputation between groups in prior randomized controlled trials,¹² but BEST-CLI demonstrated lower long-term risk of above-the-ankle amputation associated with LEB.⁷ It is notable, however, that the competing risk of death has not been taken into account in prior studies, which may bias results such as reintervention and amputation.^41–43 Patients with CLTI are older and chronically ill, with high rates of mortality regardless of employed treatment for revascularization. The use of traditional survival analysis in the case of CLTI would result in upward biasing of amputation risk, as has been seen in other cardiovascular illnesses subject to competing events.^41–43 Given the high long-term mortality rate in patients with CLTI, replicated in the current study, it is critical to take this competing risk into consideration when examining the risk of amputation in CLTI. Higher 30-day amputation associated with PVI may be due to higher rates of periprocedural technical failure and reintervention, which was not examined in the current study. As differences in reintervention tend to be most pronounced in the first 6 months after revascularization,⁷ this may explain the lower risk of short-term amputation associated with LEB with no difference in long-term outcomes. Selection bias, as noted previously, may also have influenced our results.

Results from this study provide clinicians and patients with contemporary outcomes data and help to validate recently published and pending randomized controlled trial data in a broad, “real-world” cohort. Further work, however, is necessary to fully examine additional “real-world” outcomes, including periprocedural complications. These results also highlight the need for further research in developing comprehensive risk stratification and shared decision-making models with preference-sensitive indications for LEB vs PVI. Investment in technology for evaluation and for treatment of infrainguinal CLTI is needed. Despite persistently high morbidity and mortality associated with CLTI, PVI has continued to rely on plain balloon angioplasty, with nearly 90% of this cohort receiving plain balloon angioplasty. There remains uncertainty in risk assessment for patients with CLTI, as there may be significant variability in outcomes based on patient and anatomic characteristics, arguing for patient-level risk assessment as opposed to a universal approach. In the era of precision medicine, further tailored analyses in generalizable cohorts are required to inform decision-making.

Limitations.

Our study has several limitations. First, we used propensity score matching in our analysis to minimize the risk of heterogeneity between comparators and selection bias. Although we were successful in creating balanced cohorts based on the propensity score variables, we acknowledge the impact of unmeasured confounding variables that ideally would be included in the propensity score, such as anatomical complexity, lesion-specific data not available in the VQI, and frailty. E-value analysis does not suggest that unmeasured confounding is likely to change these results; however, these values are dependent on the strength of association and number of measured confounding variables in the model.⁴⁴ Additionally, the lack of lesion-specific data in the VQI limits our ability to fully define the location and extent of disease in the LEB group. In cases with more proximal graft destination, this may result in inclusion of some cases with predominantly superficial femoral artery disease. Additionally, although multivessel disease including the superficial femoral artery was included in the lower extremity bypass arm, we were unable to capture multivessel disease including the superficial femoral artery in the PVI cohort. The addition of available angiographic data, such as a variable for culprit lesion, to the VQI in the future, will allow for further investigation. We chose to include these to capture as many cases of multivessel disease and below-the-knee disease as possible. Second, we are unable to differentiate between above-ankle amputations and transmetatarsal amputations in the available data, and acknowledge that transmetatarsal amputations, traditionally considered a minor amputation, may account for differences in amputation outcomes between the two groups. A lack of benefit associated with LEB vs PVI for long-term amputation outcomes in our cohort, in contrast to recent randomized controlled trial data, may be attributable to the inability to exclude transmetatarsal amputations. However, this is not currently verifiable in the VQI, and further evaluation using other registry data is warranted.⁷ Third, we did not have access to mortality causes, which prevented us from examining whether the mortality differences were due to cardiovascular vs other causes and would have allowed us to understand potential mechanisms that explain the differential risk. Fourth, given the low proportion of patients with prosthetic and alternative autologous conduits in this study, our results may not be generalizable to patients without an optimal conduit. Fifth, laterality of reintervention was not available in the VISION data, limiting evaluation of reintervention. Lastly, our cohort was limited to Medicare beneficiaries and may not be generalizable to patients with private insurance or who lack insurance.

CONCLUSIONS

LEB was associated with lower 30-day amputation risk and 5-year all-cause mortality when compared with PVI in the VQI registry with Medicare-linked outcomes data. These results provide contemporary comparative effectiveness data for CLTI and may serve as a critical component in the validation of randomized controlled trial data.

ARTICLE HIGHLIGHTS.

• Type of Research: Retrospective review of prospectively collected Vascular Quality Initiative data with Medicare claims-linked outcomes data

• Key Findings: In a propensity-matched cohort of 4150 patients with infrainguinal chronic limb-threatening ischemia undergoing revascularization, lower extremity bypass was associated with a lower risk of all-cause mortality over 5 years (hazard ratio, 0.77; 95% confidence interval, 0.70–0.86; P-value < .001) and amputation over 30 days (hazard ratio, 0.63; 95% confidence interval, 0.42–0.95; P-value = .025) compared with peripheral vascular intervention.

• Take Home Message: Lower extremity bypass was associated with a lower risk of 5-year all-cause mortality and 30-day amputation compared with peripheral vascular intervention in a real-world cohort with administrative claims-linked outcomes data. These results may serve to validate recently published randomized controlled trial data, and to broaden the comparative effectiveness evidence base for chronic limb-threatening ischemia.