Determinants of survival and major amputation after peripheral endovascular intervention for critical limb ischemia

Abstract

Objective

Our objective was to analyze periprocedural and 1-year outcomes of peripheral endovascular intervention (PVI) for critical limb ischemia (CLI).

Methods

We reviewed 1244 patients undergoing 1414 PVIs for CLI (rest pain, 29%; tissue loss, 71%) within the Vascular Study Group of New England (VSGNE) from January 2010 to December 2011. Overall survival (OS), amputation-free survival (AFS), and freedom from major amputation at 1 year were analyzed using the Kaplan-Meier method. Cox proportional hazards models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs).

Results

The number of arteries treated during each procedure were 1 (49%), 2 (35%), 3 (12%), and ≥4 (5%). Target arterial segments and TransAtlantic Inter-Society Consensus classifications were aortoiliac, 27% (A, 48%; B, 28%; C, 12%; and D, 12%); femoral-popliteal, 48% (A, 29%; B, 34%; C, 20%; and D, 17%); and infrapopliteal, 25% (A, 17%; B, 14%; C, 25%; D, 44%). Technical success was 92%. Complications included access site hematoma (5.0%), occlusion (0.3%), and distal embolization (2.4%). Mortality and major amputation rates were 2.8% and 2.2% at 30 days, respectively. Overall percutaneous or open reintervention rate was 8.0% during the first year. At 1-year, OS, AFS, and freedom from major amputation were 87%, 87%, and 94% for patients with rest pain and 80%, 71%, and 81% for patients with tissue loss. Independent predictors of reduced 1-year OS (C index = .74) included dialysis (HR, 3.8; 95% CI, 2.8–5.1; P < .01), emergency procedure (HR, 2.5; 95% CI, 1.0–6.2; P = .05), age >80 years (HR, 2.2; 95% CI, 1.7–2.8; P < .01), not living at home preoperatively (HR, 2.0; 95% CI, 1.4–2.8; P < .01), creatinine >1.8 mg/dL (HR, 1.9; 95% CI, 1.3–2.8; P < .01), congestive heart failure (HR, 1.7; 95% CI, 1.3–2.2; P < .01), and chronic β-blocker use (HR, 1.4; 95% CI, 1.0–1.9; P = .03), whereas independent preoperative ambulation (HR, 0.7; 95% CI, 0.6–0.9; P = .014) was protective. Independent predictors of major amputation (C index = .69) at 1 year included dialysis (HR, 2.7; 95% CI, 1.6–4.5; P < .01), tissue loss (HR, 2.0; 95% CI, 1.1–3.7; P = .02), prior major contralateral amputation (HR, 2.0; 95% CI, 1.1–3.5; P = .02), non-Caucasian race (HR, 1.7; 95% CI, 1.0–2.9; P = .045), and male gender (HR, 1.6; 95% CI, 1.1–2.6; P = .03), whereas smoking (HR, .60; 95% CI, 0.4–1.0; P = .042) was protective.

Conclusions

Survival and major amputation after PVI for CLI are associated with different patient characteristics. Dialysis dependence is a common predictor that portends especially poor outcomes. These data may facilitate efforts to improve patient selection and, after further validation, enable risk-adjusted outcome reporting for CLI patients undergoing PVI. (J Vasc Surg 2015;62:655–64.)

The incidence of peripheral arterial disease (PAD) is increasing, affecting ~8 million Americans and 12% to 20% of Americans aged >65 years.¹ Endovascular peripheral vascular intervention (PVI) is increasingly used to treat lower extremity PAD, including critical limb ischemia (CLI), with a concomitant decline in the rate of lower extremity bypass (LEB).^2,3

The Bypass versus Angioplasty in Severe Ischaemia of the Leg (BASIL) trial demonstrated similar rates of overall survival (OS) and amputation-free survival (AFS) between the PVI and LEB groups at 2 years, but clinical outcomes >2 years were improved in patients initially randomized to surgery.^4,5 This finding suggests that infrainguinal bypass is a more durable treatment but that PVI is more suitable for older, less healthy patients who are poor surgical candidates with limited life expectancy.

Despite this evidence, patient selection remains challenging given the heterogeneity of patient presentation, variability in lesion severity, and evolving endovascular technology. Several models have been proposed to predict survival after LEB but few exist exclusively for PVI.^6–13 The BASIL survival prediction model is the only one derived from a randomized cohort treated by infrainguinal bypass or angioplasty.¹¹ Although determinants of survival and major amputation have been studied extensively in patients undergoing open surgical bypass, they have not been extensively studied in a multicenter cohort of patients undergoing a range of PVIs, including angioplasty, stenting, and atherectomy. The primary aim of this study was to examine factors predicting OS and major amputation after PVI for CLI using the Vascular Study Group of New England (VSGNE) Peripheral Vascular Intervention registry. Secondary aims included analysis of procedural technical details, morbidity, and 1-year reintervention rates.

METHODS

The use of deidentified data from the VSGNE, with waiver of informed consent for this analysis, was approved by the University of Vermont College of Medicine Institutional Review Board.

Patients and procedures

We reviewed 1414 PVI procedures performed for CLI within the VSGNE from January 2010 to December 2011. The VSGNE registry is a prospective, multi-institutional, regional quality improvement initiative that has been previously described.¹⁴ A total of 108 surgeons from 23 academic and community institutions contributed to the VSGNE PVI registry. The study included 1244 patients with 1301 limbs at risk. For the 155 patients (12%) undergoing more than one PVI during the study time period, only the first intervention was analyzed. The study analysis ended in December 2011 to allow for adequate long-term follow-up at 1 year.

Patients had rest pain (29%) or tissue loss (71%) as the indication for intervention. The study excluded patients with claudication (n = 3092), acute ischemia (n = 313), and aneurysmal (n = 17) disease and those who underwent concomitant PVI and infrainguinal (n = 47) or suprainguinal bypass (n = 5). Additional exclusions included 181 procedures performed in the absence of symptoms, including interventions on native vessels proximal or distal to failing bypass grafts and 17 procedures where the type of pathology requiring intervention was not recorded.

VSGNE PVI database

The VSGNE PVI registry includes procedures performed for lower extremity arterial occlusive disease of the aorta, iliac, femoral-popliteal, or infrapopliteal arteries or for peripheral aneurysms. Interventions on multiple arterial levels and concomitant femoral endarterectomy and PVI were included. Diagnostic angiograms and peripheral arterial thrombolysis-only procedures are not captured in the database.

Demographics, comorbidities, and indications were designed to mirror those collected for the VSGNE infrainguinal bypass database. The unit of analysis for demographics and comorbidities was the individual patient. The indication for PVI was classified as (0) asymptomatic, defined as documented peripheral disease with no symptoms, (1) claudication, (2) rest pain, (3) tissue loss, (4) acute ischemia, or (5) no documented peripheral disease. The pathology was classified as occlusive or aneurysmal disease according to the primary disease process.

An extensive list of 27 procedural variables was developed from a working group of surgeons representing centers across the VSGNE. Emergency procedures were defined as those requiring an operation ≤12 hours of admission to prevent limb loss. Procedural variables included access type and location, medications, fluoroscopy time, contrast volume, and method of arterial closure. Aortoiliac and femoral-popliteal lesions were classified based on the TransAtlantic Inter-Society Consensus (TASC I) for the Management of Peripheral Arterial Disease (TASC II) classification.¹⁵ Tibial-peroneal lesions were stratified by the TASC I classification because an updated infrapopliteal scoring system was not published with TASC II.¹⁶

Interventions were performed at the discretion of the interventionalist and included balloon angioplasty, including cryoplasty or cutting balloon, stent or stent graft placement, and atherectomy, including laser, excisional, or orbital techniques. The VSGNE PVI registry records interventions on up to six arteries. For this analysis, arteries were classified in four arterial segments: aortoiliac, common femoral-profunda, superficial femoral-popliteal, and infrapopliteal. Interventions performed on more the one arterial segment were classified as multilevel. For each case, a primary and secondary treatment was designated. If more than two treatment types were used, the two that most contributed to the final outcome in the opinion of the interventionalist were recorded. Adjunctive treatments were recorded, including mechanical or pharmacologic thrombolysis, use of re-entry or embolic protection devices, or simultaneous femoral endarterectomy.

Technical success was defined as residual stenosis of ≤30% or resting systolic pressure gradient of <10 mm Hg. If an intervention was attempted but the interventionalist could not cross the lesion or the procedure resulted in vessel occlusion, it was considered a technical failure.

Immediate, in-hospital, or periprocedural complications were reported, including arterial dissection or perforation, distal embolization, access site occlusion or hematoma, and medical complications requiring admission. Hematoma was defined as (1) minor if visible or symptomatic but requiring no treatment beyond compression and might require admission for observation, (2) moderate if transfusion or thrombin injection was required, or (3) severe if surgery was required for repair. Postprocedure discharge medications were recorded. The unit of analysis for procedure indication, pathology type, and outcome measures was the individual limb.

Data were entered on-line by physicians, nurses, or trained data entry personnel into a Web-based form (Pathways; M2S, West Lebanon, NH) on a secure Web site. A long-term follow-up form was also constructed to capture key 1-year outcomes, including symptomatic status, patency, minor or major amputation, and interventional or surgical reinterventions. Additional information on the VSGNE PVI registry, including portable document forms containing data elements, is available at http://www.vascularqualityinitiative.org/about/procedures-collected.

Data validation

Cases submitted to the registry from each center were audited for comparison with hospital claims data to verify that all patients undergoing PVI had been entered by the participating physicians. Patient data were extracted using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedure codes for PVI. Patients within the VSGNE were matched to administrative data using unique identifiers such as Social Security numbers.

Outcomes

Outcome measures included 1-year OS, freedom from major amputation (FFA), and AFS as described in the Society for Vascular Surgery objective performance goals for catheter-based interventions for CLI.¹⁷ We also evaluated procedure technical details and morbidity. Surgical or percutaneous reintervention rates were recorded during the first year. Survival was also determined by matching data with the Social Security Death Index (SSDI).

Statistical analysis

Univariate analysis was performed using χ² and Fisher exact tests for known factors associated with death and amputation in patients with CLI as well as additional VSGNE variables that were identified in the literature. Risk factors univariably associated with the outcome of interest at the 0.10 significance level were then used in a multivariable Cox proportional hazards model. This model was used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for the 1-year OS and FFA after the PVI. Discrimination of the Cox models was tested using the Harrell C index. Kaplan-Meier curves were constructed for OS, AFS, and FFA during the first year. Analyses were performed using Excel (Microsoft Corp, Redmond, Wash) and STATA software (StataCorp LP, College Station, Tex).

RESULTS

Patient characteristics

From January 2010 to December 2011, 3878 PVI were performed. After exclusions for aneurysmal disease and indications other than CLI, our study cohort included 1253 patients undergoing 1414 PVIs. Mean follow-up was 250 days. All procedures were performed for CLI, with tissue loss (71%) as the most common indication (Table I). During the study time frame, 90% of patients underwent a single endovascular procedure, with two and three or more separate procedures performed in 9% and 1%, respectively. Patients were a mean age of 70 years, 58% were male, and 82% were Caucasian.

Table I.

Univariate analysis of patient and procedure characteristics of mortality at 1 year

Variable	Total (N = 1244), %	Percent died at 1 year		P valuea
		If variable absent	If variable present
Male gender	58	18	18	.88
Nonwhite race	11	18	18	.42
Age, years
<80	75	29	14
≥80	25	14	29	<.01
Preoperative status
Living in nursing home	8	16	42	<.01
Not independently ambulatory	32	14	28	<.01
Smoking (prior or current)	74	22	16	.07
Prior smokers	46	22	20
Current smokers	29	22	12	<.01b
BMI, kg/m2
<20	8	18	22
20–29	60	17	18
≥30	32	19	16	.64b
Diabetes
Nondiabetic	39	16	16
Diabetic
Oral medication	21	16	16
Insulin-dependent	40	16	21	.08b
Hypertension	90	15	18	.16
COPD	21	17	21	.08
Coronary artery disease	35	16	22	<.01
CHF	23	14	30	<.01
Renal function
Creatinine <1.8 mg/dL	80	38	13
Creatinine ≥1.8 mg/dL	8	17	30
Dialysis	12	14	44	<.01b
Tissue loss	71	13	20	<.01
Prior PVI	29	18	18	.86
Prior LEB	21	18	19	.11
Prior major amputation	8	17	29	.01
Preoperative medications
Antiplatelet agent	80	18	18	.84
Statin	67	20	17	.99
β-blocker
None	30	23	20
Preoperative	3	12	18
Chronic	64	21	14	<.01b
Procedure variables
Emergency procedure	1	18	40	.06
Concomitant CFA endarterectomy	8	18	14	.36
Arteries treated, No.
1	49	16		16
2	35	17		20
≥3	16	18	18	.12b
Arterial segment treated
Aortoiliac	24	18	12
SFA-popliteal	33	18	15
Tibial-peroneal	15	17	18
CFA/PFA	3	17	21
Multilevelc	25	15	22	.01b
TASC classification
Aortoiliac
A (referent)	48	14	15
B	28	14		15
C	12	15		14
D	11	15	11	.57b
Femoral-popliteal
A (referent)	29	19	19
B	34	20		19
C	20	19		18
D	16	19	18	.91b
Tibial-peroneal
A (referent)	18	21	26
B	17	22		2
C	25	20		26
D	40	25	17	.79b

BMI, Body mass index; CFA, common femoral artery; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; LEB, infrainguinal lower extremity bypass; PFA, profunda femoris artery; PVI, peripheral vascular intervention; SFA, superficial femoral artery; TASC, TransAtlantic Inter-Society Consensus.

^aP value by log-rank test.

^bAcross groups.

^cMultilevel includes >1 of any segment in any combination.

Comorbidities were typical for patients with advanced PAD, including a history of hypertension in 90%, prior or current smoking in 75%, and diabetes in 61%. Renal insufficiency was present in 8%, with an additional 12% requiring dialysis. Prior LEB had been performed in 21%, and 8% had a prior major contralateral amputation.

Characteristics of interventions and procedure complications

The number of arteries treated during the intervention ranged from 1 (49%), 2 (35%), to ≥3 (16%; Table I). The arterial segments undergoing intervention were aortoiliac in 25%, common femoral-profunda femoral in 4%, superficial femoral-popliteal in 32%, and infrapopliteal in 14% arteries, whereas 25% underwent intervention on more than one segment. TASC classifications varied by arterial segment. A variety of treatment modalities were used based on the interventionalist’s discretion (Table II). Technical success was 92%, with failure to cross the lesion occurring in 4%. Complication rates included access site hematoma (5.0%) or occlusion (0.3%), and distal embolization (2.3%; Table III). The in-hospital mortality rate was 1%, the 30-day mortality rate was 2.8%, and the major amputation rate was 2.0%, reflective of the patient population with CLI.

Table II.

Type of peripheral vascular interventions (PVI) according to arterial segment and TransAtlantic Inter-Society Consensus (TASC) classification

	Arterial segment
	Aortoiliac, No. (%)		Femoral-popliteal, No. (%)		Tibial-peroneal, No. (%)
	TASC class
Type of PVI	A/B	C/D	A/B	C/D	A/B	C/D
PTA alone	46 (14.2)	12 (10.0)	230 (49.6)	102 (32.7)	102 (77.9)	225 (73.5)
Stent	242 (74.7)	72 (60.0)	180 (38.8)	134 (42.9)	9 (6.9)	18 (5.9)
Stent graft	33 (10.2)	34 (28.3)	9 (1.9)	13 (4.2)	0 (0)	0 (0)
Atherectomya	3 (0.9)	2 (1.7)	45 (9.7)	63 (20.2)	20 (15.3)	63 (20.6)

PTA, Percutaneous transluminal angioplasty.

^aIncludes laser, excisional, or orbital atherectomy combined.

Table III.

Procedural outcomes and complications

Variable	No. (%)
Technical result
Successfula	1975 (92)
Residual stenosis >30%	76 (4)
Failure to cross lesion	85 (4)
Hematoma	64 (5)
Minor	48 (3.9)
Moderate	11 (0.9)
Major	5 (0.4)
Occlusion of access site	4 (0.3)
Distal embolization	25 (2.3)
Arterial perforation
Iliac	6 (0.5)
Femoral-popliteal	7 (0.6)
Tibial-peroneal	2 (0.2)
Complication requiring admission	53 (4)
30-day major amputation	15 (2)
30-day mortality	35 (2.8)

^aNumber exceeds the total procedures because it includes interventions on ≥1 artery per procedure.

OS

The OS estimate during follow-up was 82%. The OS estimate was lower for patients treated for tissue loss than for patients treated for rest pain (log-rank, P = .0005), with 1-year survival estimates of 80% and 87%, respectively (Fig 1). In univariate analysis, we found several preoperative variables that were associated with decreased survival at 1 year, including comorbidities, presence of tissue loss, and anatomic factors (Table I). Lower survival was observed for common femoral-profunda interventions and for patients undergoing an intervention on more than one arterial segment. Anatomic factors, such as TASC class within each arterial segment and the treatment of more than one vessel within the same arterial segment, were not associated with mortality.

Fig 1.

Kaplan-Meier curve of overall survival (OS) after peripheral vascular intervention (PVI) for critical limb ischemia (CLI). EST, Estimated survival; SE, standard error.

In the primary analysis, the multivariable model identified eight factors associated with reduced survival at 1 year (Table IV), including dialysis (HR, 3.8; 95% CI, 2.8–5.1; P < .01), an emergency procedure (HR, 2.5; 95% CI, 1.0–6.2; P = .046), age >80 years (HR, 2.2; 95% CI, 1.7–2.8; P < .01), dependent preoperative living status (HR, 2.0; 95% CI, 1.4–2.8; P < .01), creatinine >1.8 mg/dL (HR, 1.9; 95% CI, 1.3–2.8; P < .01), congestive heart failure (CHF; HR, 1.7; 95% CI, 1.3–2.2; P < .01), and chronic β-blocker use (HR, 1.4; 95% CI, 1.0–1.9; P = .03). Independent preoperative ambulation was associated with improved survival (HR, 0.7; 95% CI, 0.5–0.9; P = .01). The C index was 0.74 for the survival model.

Table IV.

Multivariable Cox proportional hazards model of factors associated with 1-year survival after peripheral vascular intervention (PVI) for critical limb ischemia (CLI)^a

Preoperative characteristic	HR	95% CI	P value
Dialysis dependence	3.8	2.8–5.1	<.01
Emergency procedure	2.5	1.0–6.2	.046
Age >80 years	2.2	1.7–2.8	<.01
Not living at home preoperatively	2.0	1.4–2.8	<.01
Creatinine >1.8 mg/dL	1.9	1.3–2.8	<.01
CHF	1.7	1.3–2.2	<.01
Chronic β-blocker use	1.4	1.0–1.9	.03
Independent ambulation preoperatively	0.7	0.6–0.9	.014

CHF, Congestive heart failure; CI, confidence interval; HR, hazard ratio.

^aC index = 0.74.

AFS

Univariate analysis of AFS is shown in Supplementary Table I (online only). The AFS estimate was lower for patients treated for tissue loss compared with patients treated for rest pain (log-rank, P < .0001), with 1-year estimates of 71% and 87%, respectively (Fig 2). Multivariable analysis showed six variables were predictive of worse AFS during the first year (Table V). These included dialysis dependence (HR, 2.4; 95% CI, 1.6–3.5; P < .01), dependent preoperative living status (HR, 1.8; 95% CI, 1.1–3.0; P = .03), CHF (HR, 1.7; 95% CI, 1.2–2.3; P < .01), tissue loss (HR, 1.6; 95% CI, 1.1–2.5; P = .02), male gender (HR, 1.5; 95% CI, 1.1–2.0; P = .02), and the arterial segment treated. Interventions on more than one arterial segment were associated with decreased AFS compared with aortoiliac (HR, 1.6; 95% CI, 1.0–2.5; P = .04) or common femoral-profunda interventions (HR, 3.4; 95% CI, 1.2–9.5; P = .02). Tibial-peroneal interventions were also associated with decreased AFS compared with aortoiliac (HR, 2.1; 95% CI, 1.3–1.5; P < .01) or superficial femoral-popliteal interventions (HR, 1.6; 95% CI, 1.0–2.5; P = .03).

Fig 2.

Kaplan-Meier curve of amputation-free survival (AFS) after peripheral vascular intervention (PVI) for critical limb ischemia (CLI). EST, Estimated survival; SE, standard error.

Table V.

Multivariable Cox proportional hazards model of factors associated with amputation-free survival (AFS) after peripheral vascular intervention (PVI) for critical limb ischemia (CLI)^a

Variable	HR	95% CI	P value
Dialysis dependence	2.4	1.6–3.5	<.01
Not living at home preoperatively	1.8	1.1–3.0	.03
CHF	1.7	1.2–2.3	<.01
Tissue loss	1.6	1.1–2.5	.02
Male gender	1.5	1.1–2.0	.02
Arterial segment
Multilevel vs CFA-PFAb	3.4	1.2–9.5	.02
Tibial-peroneal vs aortoiliac	2.1	1.3–3.5	<.01
Multilevel vs aortoiliacb	1.6	1.0–2.5	.04
Tibial-peroneal vs SFA-popliteal	1.6	1.0–2.5	.03
CFA-PFA vs tibial-peroneal	0.2	0.1–.06	<.01

CFA, Common femoral artery; CHF, congestive heart failure; CI, confidence interval; HR, hazard ratio; PFA, profunda femoris artery; SFA, superficial femoral artery.

^aC index = 0.70.

^bMultilevel includes any combination of >1 segment.

FFA

By univariate analysis, we found several preoperative variables that were associated with major amputation at 1 year (Supplementary Table II, online only). Treatment type did not affect the outcome, with morality and amputation rates equal across groups receiving angioplasty alone, stent/stent graft, or atherectomy The FFA estimate was lower during follow-up for patients treated for tissue loss compared with patients treated for rest pain (log-rank, P = .0002), with 1-year estimates of 81% and 94%, respectively (Fig 3). Multivariable analysis found six variables were predictive of major amputation at 1 year (Table VI): dialysis dependence (HR, 2.7; 95% CI, 1.6–4.5; P < .01), tissue loss (HR, 2.0; 95% CI, 1.1–3.7; P = .02), prior major contralateral amputation (HR, 2.0; 95% CI, 1.1–3.5; P = .02), non-Caucasian race (HR, 1.7; 95% CI, 1.0–2.9; P = .045), and male gender (HR, 1.6; 95% CI, 1.1–2.6; P = .03). Paradoxically, current or former smoking (HR, 0.6; 95% CI, 0.4–1.0; P = .042) was associated with improved FFA.

Fig 3.

Kaplan-Meier curve of major amputation after peripheral vascular intervention (PVI) for critical limb ischemia (CLI). EST, Estimated survival; SE, standard error.

Table VI.

Multivariable Cox proportional hazards model of factors associated with freedom from major amputation (FFA) after peripheral vascular intervention (PVI) for critical limb ischemia (CLI)^a

Variable	HR	95% CI	P value
Dialysis dependence	2.7	1.6–4.5	<.01
Tissue loss	2.0	1.1–3.7	.02
Prior major contralateral amputation	2.0	1.1–3.5	.02
Nonwhite race	1.7	1.0–2.9	.04
Male gender	1.6	1.1–2.6	.03
Current or former smoker	0.6	0.4–1.0	.04

CI, Confidence interval; HR, hazard ratio.

^aC index = 0.69.

DISCUSSION

Patients with CLI present a challenge to the vascular specialist because of their comorbidities and advanced atherosclerotic disease. Physicians must balance the risks and benefits of conservative management compared with revascularization by endovascular, open, or hybrid techniques. Treatment decisions must often be made in the setting of declining functional status, comorbidities that limit life expectancy, and the realistic prospects for meaningful recovery. The proliferation of endovascular therapies has expanded the options for treatment, while at the same time making decision making more challenging. Although endovascular therapies have the potential to reduce procedural morbidity and mortality, the issues of reintervention rates and long-term durability remain important considerations. Decision making is further complicated by the wide range of potential endovascular therapies and the relative paucity of quality outcome data. Recognizing this, we sought to identify predictors of mortality and major amputation in patients undergoing PVI in an effort to improve patient selection and provide a mechanism for risk-adjusted outcome reporting.

Similar to the published outcomes for LEB in the VSGNE, outcomes after PVI for CLI within our region were comparable to the SVS suggested objective performance goals (OPG) for catheter-based interventions of CLI.^17,18 Mortality and major amputation rates at 30 days were 2.8% and 2.0% compared with 2.7% and 1.9% reported from the randomized, controlled trials that served as basis for SVS safety and efficacy OPGs. Similarly, our reported limb salvage and OS at 1 year of 85% and 83% compare favorably with SVS OPGs of 84% and 80%.¹⁷ The 23% mortality at 1 year after PVI is higher than the rate of 13% previously reported in the VSGNE after infrainguinal bypass for CLI.¹³ This difference likely reflects physicians offering higher-risk patients PVI instead of LEB given the perceived decreased short-term morbidity and mortality.

Several factors, including age >80 years, dialysis, emergency procedures, and CHF were, similar to predictors of survival after infrainguinal bypass previously reported from the VSGNE.¹³ In contrast to patients undergoing LEB, diabetes, male gender, and prior major amputation were not associated with reduced survival after PVI for CLI. It is important to realize that the two VSGNE studies are not directly comparable because the former study included claudicant patients, whereas the present study was confined to patients with CLI. Although there are several common predictive factors between LEB and PVI, there are also important differences in risk factors. Recognizing this, we see value in reporting a PVI-specific survival model to improve decision making and provide risk adjustment to outcomes.

In this study, the independent factors associated with OS, AFS, and FFA after PVI differed. The only common risk factor for OS, AFS, and FFA was dialysis, emphasizing the importance of renal function in patient prognosis (Table VII). In contrast, CHF was associated with decreased OS and AFS. The differences in risk factors for survival vs amputation highlight the difficulty in predicting composite end points such as AFS. Survival and amputation are different clinical end points are thus are likely to be driven by different factors. Causes of death may range from cardiovascular causes, such as myocardial infarction, to stroke and to cancer. Conversely, prior major amputation was associated with a later major amputation but not with survival. Statistical models are often strengthened when they are used to predict the end point of a single pathophysiologic process.¹⁹

Table VII.

Comparison of variables associated with overall survival (OS), amputation-free survival (AFS), and freedom from major amputation (FFA)

Variable	OS	AFS	FFA
Dialysis dependence	↑	↑	↑
Congestive heart failure	↑	↑
Not living at home preoperatively	↑	↑
Tissue loss		↑	↑
Male gender		↑	↑
Age >80 years	↑
Creatinine >1.8 mg/dL	↑
Emergency procedure	↑
Independent ambulation preoperatively	↓
Chronic β-blocker use	↑
Arterial segmenta		↑ or ↓
Nonwhite race			↑
Smoking (current or former)a			↑
Prior major contralateral amputation			↑

^aRefer to the multivariate models for specific arterial segments.

A surprising finding was that smoking was correlated with an improvement in FFA, a finding that persisted for past and current smokers. We first postulated that the presence of CLI and subsequent PVI were the impetus for smoking cessation, which was in turn responsible for the improvement in amputation outcomes. However, we found no differences in FFA among patients who quit compared with those who continued to smoke. Further research will be needed to determine if this is a reproducible finding.

Relatively few studies have analyzed survival after endovascular treatment of CLI. Using a large Canadian administrative data set, Al-Omran et al²⁰ reported risk factors for death after arterial bypass or angioplasty without stratifying by indication. Although this database lacked the level of patient and procedural detail available in the VSGNE, the authors also identified age and male gender as risk factors for late mortality and amputation after angioplasty.

The BASIL trial prediction model for survival after angioplasty or bypass for CLI remains the best evidence reported to date.¹¹ We found several risk factors in common with the BASIL model, including advanced age and renal insufficiency (Table VIII). These factors also appear in the Edifoligide for the Prevention of Infrainguinal Vein Graft Failure (PREVENT) III model for open revascularization.⁹ Unlike the BASIL trial, presence of tissue loss, coronary artery disease, and body mass index were not associated with mortality in this study.

Table VIII.

Summary of four risk-prediction models for survival or amputation-free survival (AFS) after revascularization^a

Study	VSGNE PVI	BASIL	PREVENT III	FINNVASC
Revascularization	PVI (n = 1253)	PVI (n = 228) or LEB (n =224)	LEB (n = 1404)	LEB (n = 3925)
Outcome	1-year AFS	2-year AFS	1-year AFS	30-day AFS
Variable
Renal insufficiencyb	1.9–3.8	2.0	2.8
Agec	2.2	1.7	1.6
Emergency procedure	2.5			1.8
Not living at home	2.0
CHF	1.7
Chronic β-blockers	1.4
Independent ambulation	0.7
Tissue lossd		2.6	2.2	1.5
CAD		↓	1.4	1.4
Diabetes, type 1 or 2		1.2		1.5
Bollinger score below knee 5–8		1.9
Smoking		1.7
BMI <20 kg/m2		1.5
Stroke/TIA		↓
Ankle pressurese		↓
Hematocrit <30			1.6
Model discrimination: reported C index validation8–12	0.74	–	–	0.60
	Not validated	0.65–0.70	0.58–0.63	0.50–0.58

BASIL, Bypass versus Angioplasty in Severe Ischaemia of the Leg; BMI, body mass index; CAD, coronary artery disease; CHF, congestive heart failure; CLI, critical limb ischemia; FINNVASC, Finland National Vascular prediction model; LEB, lower extremity bypass; PREVENT III, Edifoligide for the Prevention of Infrainguinal Vein Graft Failure; PVI, peripheral vascular intervention; TIA, transient ischemic attack; VSGNE, Vascular Study Group of New England.

^aHazard ratios (HRs) or odds ratios (ORs) from original publication. ↓ indicates decreased overall survival (OS) or amputation free survival (AFS) when HR not reported.

^bCreatinine ≥1.78 mg/dL or dialysis for VSGNE, creatinine >4.7 mmol/L for BASIL, on dialysis for PREVENT III.

^cAge >80 years for VSGNE PVI and BASIL, >75 years for PREVENT III.

^dDefined as any tissue loss for PREVENT III and BASIL (with ankle pressure <50 mm Hg) and gangrene for FINNVASC.

^eIncludes number of serial ankle pressures obtained and maximum pressure.

We report additional risks factors of reduced OS including CHF, chronic β-blocker use, emergency procedure, and dependent living status, and found independent preprocedure ambulation to be protective. Independent ambulation and living in a nursing home likely reflect patient functional status. Chronic β-blockers are markers for chronic heart disease. The only similarity we found with the Finland National Vascular (FINNVASC) prediction model of 30-day death or major amputation after open revascularization was urgency of the procedure.⁶ In our study the poor outcomes in patients on dialysis was particularly sobering, with mortality rates of 44% at 1 year. Renal insufficiency and dialysis dependence have been also been reported as negative predictors in single-center series and in the PREVENT III prediction model for AFS after infrainguinal bypass.^8,9,21 The discrimination of the VSGNE PVI survival model was higher (C index = 0.74) than the three other models, likely due to the inclusion of several additional variables. Validation of the VSGNE PVI model is required to verify its discriminative ability.

Our analysis cannot be compared directly with the BASIL trial because we did not record stroke history, Bollinger scores, or serial ankle pressure measurements. We analyzed lesion severity by the number of vessels undergoing intervention, the arterial segment treated, and TASC classification. The type of treatment did not influence outcomes, with no differences between angioplasty alone, stenting, or atherectomy groups in major amputation or mortality. This suggests appropriate patient and lesion selection for the type of intervention.

Soga et al²² recently reported a multicenter retrospective Japanese registry on PVI or LEB for CLI involving 1053 patients with infrainguinal disease. This propensity-matched study found no difference in OS, limb salvage, or AFS between the endovascular and bypass groups. Similar to our findings, age, heart failure, hemodialysis, and nonambulatory status were independent predictors of reduced OS. Unlike our study, chronic obstructive pulmonary disease and body mass index were also independently associated with reduced survival. Similar to our findings, tissue loss, nonambulatory status, heart failure, and hemodialysis were independent predictors of reduced AFS.

Several limitations should be considered when these results are evaluated. First, the SSDI, which was used to validate Vascular Quality Initiative (VQI) data on mortality, has known limitations that may lead to underestimation of mortality rates.

Second, we only analyzed hard end points of death and amputation, without an in-depth study of symptomatic, ambulatory, or functional status. These end points are relevant because they ultimately influence quality of life after revascularization.

Third, our study population only included those with CLI and cannot be extrapolated to claudicant patients.

Fourth, the results only apply to PVI and not infrainguinal bypass.

Fifth, although we were able to verify survival using the SSDI, amputation was self-reported, with long-term follow-up available in 67% of patients. When we examined the mortality rates of the group with (186 [24%]) and without (98 [20%]) long-term follow-up, we found no difference (P = .10), suggesting that bad outcomes are not systemically under-reported.

Finally, we attempted to account for factors associated with the outcomes of interest; however, as with any retrospective study, we cannot completely account for all possible confounders. Additional work is under way to see if these results are generalizable beyond the New England region to the United States as a whole using data from entire VQI. A matched comparison of the PVI and infrainguinal VQI registries may further elucidate the survival difference between patients undergoing open and endovascular revascularization.

CONCLUSIONS

OS, AFS, and major amputation after PVI for CLI are associated with several preoperative patient-specific characteristics. Many of these characteristics differ between survival and amputation. Dialysis is a common risk factor and the strongest predictor of poor outcomes. These data may facilitate efforts to improve patient selection and, after further validation, enable risk-adjusted outcome reporting for CLI patients undergoing PVI.