Risk factors for amputation are influenced by competing risk of death in patients with critical limb ischemia

Risk factors for amputation are influenced by competing risk of death in patients with critical limb ischemia Eva Torbjörnsson, RN, MSd,a,b Lena Blomgren, MD, PhD,c Ann-Mari Fagerdahl, RN, PhD,a,d Lennart Boström, MD, PhD,a,b Carin Ottosson, MD, PhD,a,d and Jonas Malmstedt, MD, PhD,a,b Stockholm and Örebro, Sweden

ABSTRACT Objective: Patients with critical limb ischemia (CLI) have a high risk of amputation and death. Death is a competing risk that affects the estimated amputation risk. Our aim was to find the specific risk factors for amputation for patients with CLI using competing risk analyses and compared these results with those from standard Cox regression analysis. Methods: Patients who had undergone revascularization for CLI (2009-2013, with follow-up data until 2017) in Stockholm were identified from the Swedish National Registry for Vascular Surgery. The main outcome was major amputation. The risk factors for amputation were assessed using competing risk analysis and compared with the risk factors for amputation-free survival identified using Cox proportional hazards regression analysis. Results: Of 855 patients with CLI, 178 had required a major amputation and 415 had died during the 8-year follow-up period. In the competing risk regression, age (subdistribution hazard ratio [sub-HR], 0.98; 95% confidence interval [CI], 0.97-1.00), ambulatory status (independent vs bedridden; sub-HR, 4.10; 95% CI, 2.14-7.86), and ischemic wound vs rest pain (sub-HR, 3.03; 95% CI, 1.72-5.36) were associated with amputation, considering death as a competing risk. In contrast, Cox regression analysis identified female vs male (hazard ratio [HR], 0.77; 95% CI, 0.64-0.94), age (HR, 1.02; 95% CI, 1.01-1.03), renal impairment (HR, 2.08; 95% CI, 1.61-2.67), ambulatory status (independent vs bedridden; HR, 3.45; 95% CI, 2.30-5.18), and ischemic wound vs rest pain (HR, 2.41; 95% CI, 1.78-3.25) as risk factors. Conclusions: The risk factors associated with amputation differed when analyzing the data using competing risk regression vs Cox regression. The differences between the analyses indicated that a risk exists for biased estimates using standard survival methods when a strong competing risk such as death is present. (J Vasc Surg 2019;-:1-10.) Keywords: Competing risk analysis; Critical limb ischemia; Major amputation

Peripheral arterial disease is common worldwide. The prevalence in Sweden was 18% in subjects aged $65 years.1 The incidence of peripheral arterial disease increases with older age, the presence of cardiovascular disease, and smoking.1 Treatment of critical limb ischemia (CLI), defined as ischemic rest pain, nonhealing ulceration or gangrene, aims to preserve the limb, relieve the pain, and maintain the patient’s ambulatory function.2 The possibility of achieving these goals must be weighed against the risk of complications and the worsening of symptoms if revascularization should fail.3 A patient with a poor prognosis after revascularization might benefit from primary amputation. However, predicting the need for amputation after intervention has been difficult because several technical, disease-related, and patient-related factors interplay.4

The risk factor evaluation for individual patients is important when considering vascular intervention. The mortality has been high in patients with CLI regardless of the use of revascularization; hence, the most important question for the patient and surgeon to consider is the risk of amputation after the intervention. CLI research has traditionally focused on limb salvage as the primary outcome.2 The use of limb salvage as the primary outcome is a problematic measure because it yields a positively biased estimate of amputation risk by disregarding mortality. Thus, the combined outcome of amputation-free survival (AFS) has been recommended for CLI studies because it includes mortality.2,5 The disadvantage of AFS is that its use as the outcome measure does not estimate the absolute risk of amputation.6 For example, a risk factor that is associated with an

From the Department of Clinical Science and Education, Karolinska Institutet,

Department of Surgery, Södersjukhuset (South Hospital), Stockholm 118 83,

Stockholma; the Division of Vascular Surgery, Department of Surgery, Söders-

Sweden (e-mail: [email protected]).

jukhuset AB, Stockholmb; the Department of Cardiothoracic and Vascular

The editors and reviewers of this article have no relevant financial relationships to

Surgery, Faculty of Medicine and Health, Örebro University Hospital, Örebroc;

disclose per the JVS policy that requires reviewers to decline review of any

and the Wound Centre, Södersjukhuset AB, Stockholm.d Author conflict of interest: none.

manuscript for which they may have a conflict of interest. 0741-5214

Additional material for this article may be found online at www.jvascsurg.org.

Copyright Ó 2019 by the Society for Vascular Surgery. Published by Elsevier Inc.

Correspondence: Jonas Malmstedt, MD, PhD, Department of Clinical Science

https://doi.org/10.1016/j.jvs.2019.07.074

and Education, Karolinska Institutet, and Division of Vascular Surgery,

1

2

Journal of Vascular Surgery

Torbjörnsson et al

---

increased risk of death could give a biased estimate of the risk of amputation when the research question is whether a certain group of patients will be more or less likely to undergo amputation despite the achievement of revascularization. A competing risk analysis is a method to study specific outcomes (eg, amputation) and to consider in the same analysis a competing event (eg, death). The method has been shown to provide less biased estimates of the absolute risk compared with methods using a composite end point such as AFS.7,8 A competing risk analysis studies the actual risk of an event and has been the recommended tool for the analysis of prognostic factors.7,9,10 Standard Cox regression analysis ignore the competing risk of death, resulting in a biased or incorrect interpretation of the results in populations with high mortality.11,12 We found one study that had used competing risk analysis to predict the rate of major amputation after revascularization for peripheral arterial disease.13 That study reported lower estimates for amputation compared with standard survival analysis using Cox regression.13 This highlights the importance of using appropriate methods to estimate the risk factors for amputation after revascularization and, thereby, provide further insight into which patients will have the greatest risk.14,15 The aim of the present study was to identify the specific risk factors for amputation in patients with CLI using competing risk analyses and to compare these results with those from standard Cox regression analysis.

METHODS We performed a population-based cohort study of 855 patients with CLI who had undergone first-time arterial revascularization in the index leg in a catchment area covering one half of the population of Stockholm County (2.2 million). The cohort included patients who had undergone open or endovascular surgery from 2009 to 2013, with follow-up data collected until 2017. We defined CLI in accordance with the Inter-Society Consensus for the Management of Peripheral Arterial Disease as chronic ischemic rest pain, ulcer, or gangrene attributable to objectively proven arterial occlusive disease.3 Patients with CLI secondary to acute ischemia or femoral/popliteal aneurysms were excluded. The regional ethics committee of Stockholm, Sweden, approved the present study (approval no. 2014-801-31/1). Patients and databases. The Swedish National Registry for Vascular Surgery (Swedvasc) identified all patients who had undergone infrainguinal revascularization for CLI. Swedvasc has an external validity of 93% for infrainguinal procedures.16 For further validation, we used a random sample of 103 patients (12%) from the study cohort to assess the accuracy. The operative and comorbidity data from Swedvasc were compared with the data from the corresponding medical records, and the

2019

ARTICLE HIGHLIGHTS d

d

d

Type of Research: Retrospective analysis of prospectively collected data from the Swedish National Registry for Vascular Surgery database Key Findings: The risk of major amputation was estimated using a competing risk framework for 855 patients treated for critical limb ischemia. The risk factors for amputation differed from those for death. A competing risk analysis resulted in different risk estimates compared with those from Cox regression analysis. Take Home Message: Biased estimates can arise if the competing risk from death is ignored for patients with critical limb ischemia.

sensitivity, specificity, and accuracy were calculated. All variables had an accuracy of $90%, except for cardiac disease (Supplementary Table I, online only). The main outcome measure, major amputation, was retrieved by cross-linking the study cohort with the Swedish National In-Patient Registry using the unique personal identification number assigned to all Swedish residents.17 The Swedish National In-Patient Registry contains individual information regarding the presence and level of amputation according to the Swedish version of the Nordic Medico-Statistical Committee surgical procedure codes (available at: www.nordclass.uu. se/index_e.htm). We retrieved all the medical records for the patients who had undergone amputation to ascertain the level, laterality, and date of the amputation. The coding was correct for 96% of the amputations. The incorrect data were corrected, resulting in complete follow-up data for our main outcome measure, amputation. The patients were followed from their first (ie, primary) infrainguinal revascularization procedure (index procedure) until the first major ipsilateral amputation, death, or the end of the follow-up period (December 31, 2017). Swedvasc provided data regarding demographic details, comorbidity, and revascularization procedures (Tables I and II). The comorbidities included cardiac disease (eg, current heart failure, angina pectoris, myocardial infarction, coronary bypass, percutaneous coronary intervention), pulmonary disease (eg, chronic obstructive pulmonary disease, emphysema), diabetes (type 1 and 2 treated with oral antidiabetic agents or insulin), hypertension (ie, systolic blood pressure >150 mm Hg, the use of antihypertensive medication), stroke, and renal impairment (serum creatinine, >150 mmol/L). Smoking was categorized as current smoker (smoking at the time of, or within 4 weeks of, the previous index operation) vs previous or nonsmoker. The ischemic status of the ipsilateral leg was categorized using the Society for Vascular Surgery Wound, Ischemia, and foot Infection

Journal of Vascular Surgery Volume

-,

Number

Torbjörnsson et al

3

-

Table I. Baseline characteristics Characteristic

Total population

No amputation

Major amputation

Patients

855 (100)

677 (79)

178 (21)

Male sex

408 (48)

310 (46)

Age, years (range) Current smoker

P value NA

98 (55)

.028

77 (77-78)

76 (74-77)

NA

152 (18)

116 (17)

36 (20)

.334

316 (37)

249 (37)

67 (38)

.832 .064

77 (76-78)

Comorbidity Cardiac disease Pulmonary disease

129 (15)

110 (16)

19 (11)

Diabetes

354 (41)

272 (40)

82 (46)

.156

Hypertension

655 (77)

523 (77)

132 (74)

.385

Stroke

132 (15)

98 (15)

34 (19)

.129

Renal impairment

99 (12)

66 (10)

33 (19)

.001 .713

Ischemic grade, WIfI systema 0

12 (2)

10 (2)

2 (1)

1

77 (10)

65 (10)

12 (8)

2

192 (24)

159 (25)

33 (22)

3

515 (65)

412 (64)

103 (69)

Independent

369 (43)

317 (47)

52 (29)

Walking with aid

<.001

Ambulatory status 403 (47)

307 (45)

96 (54)

Wheelchair-bound

50 (6)

36 (5)

16 (9)

Bedridden

33 (4)

19 (3)

14 (8)

173 (20)

160 (24)

13 (7)

682 (80)

517 (76)

165 (93)

<.001

Indication Rest pain Ischemic wound

ABI, Ankle-brachial index; NA, not applicable; WIfI, Wound, Ischemia, and foot Infection. Data are presented as number (%), unless noted otherwise. a Grade 0, ABI $0.80, toe pressure $60 mm Hg; grade 1, ABI $0.6-0.79, toe pressure 40-59 mm Hg; grade 2, ABI $0.4-0.59, toe pressure 30-39 mm Hg; grade 3, ABI #0.39, toe pressure <30 mm Hg.

(WIfI) system18: grade 0 (no ischemia), ankle-brachial index (ABI) $0.80 and toe pressure $60 mm Hg; grade 1 (mild ischemia), ABI $0.6 to 0.79 and toe pressure 40 to 59 mm Hg; grade 2 (moderate ischemia), ABI $0.4 to 0.59 and toe pressure 30 to 39 mm Hg; and grade 3 (severe ischemia), ABI #0.39 and toe pressure <30 mm Hg. The patients’ ambulatory status at 2 weeks preoperatively was classified as independent, walking with aid, wheelchair-bound, or bedridden. Revascularization was categorized as an endovascular procedure alone or open surgery (including hybrid procedures). The anatomic extent of revascularization procedures was classified as femoropopliteal, popliteal-crural, or multilevel (both femoropopliteal and crural). This classification was used as a proxy for the extent of the atherosclerotic burden. Major amputation was defined as amputation above the ankle. Swedvasc contained incomplete data for smoking, renal impairment, ABI, toe pressure, and ambulatory status. A review of the medical records resulted in complete data for these variables, except for four missing for smoking and 59 for ABI and/or toe pressure.

Outcome and end points. The main outcome of interest was the interval to the first ipsilateral major amputation. Patients were followed from the date of the revascularization procedure (index date) to the first ipsilateral amputation or the end of the follow-up period (December 31, 2017). Amputation after trauma was excluded (n ¼ 2). We also estimated the AFS, defined as the first ipsilateral amputation above the ankle or death from any cause, whichever occurred first. Statistical analysis. Continuous variables are presented as the mean 6 standard deviation or the median and interquartile range and categorical variables as frequencies and percentages. The baseline differences in the categorical variables were evaluated using the Pearson c2 test. We estimated the rate of amputation using the cumulative incidence function. The effect of the risk factors on amputation was estimated using the subdistribution hazard ratio (sub-HR) computed using the Fine-Gray competing risks regression analysis, and significance was assessed using the Gray test.15 One issue with

4

Journal of Vascular Surgery

Torbjörnsson et al

---

Table II. Perioperative characteristics No amputation

Major amputation

P value

677 (79)

178 (21)

NA

171 (25)

42 (24)

Bypass with vein

80 (47)

14 (34)

Bypass with prosthetic

15 (9)

4 (10)

Othera

76 (44)

24 (56)

495 (73)

135 (76)

PBA

210 (42)

61 (45)

BMS þ SAP

Characteristic Patients Surgical procedure Open surgery

Endovascular treatment

.488

143 (29)

25 (19)

SAP

79 (16)

32 (24)

Otherb

63 (13)

17 (13)

Hybrid procedure

11 (2)

1 (1)

Nonelective vs planned surgeryc

59 (9)

33 (19)

Level of revascularization Femoropopliteal Crural/pedal d

Multilevel

<.001 .219

470 (69)

112 (63)

38 (6)

14 (8)

169 (25)

52 (29)

BMS, Bare metal stent; NA, not applicable; PBA, plain balloon angioplasty; SAP, subintimal angioplasty. Data are presented as number (%). P values were calculated using the Pearson c2 test. a Included thromboendarterectomy, open patch reconstruction, or bypass. b Included drug-eluting stent or combinations of BMS, PBA, and SAP. c Acute or unplanned surgery vs planned surgery. d Both femoropopliteal and crural.

2019

from the crude analysis. We arrived at the final simplified model by removing the factors that were neither associated (P > .05) with risk of amputation nor substantially altered the estimate of the covariates associated with risk of amputation. We compared the different models using the Bayes information criterion. Patient age and sex were retained in all the models. To compare the result from the analysis using competing risk regression with previously reported data, an analysis using Cox regression for AFS and all-cause mortality was performed (Supplementary Tables II and III, online only). The proportional hazards assumption was confirmed using Schoenfeld residuals in the Cox regression analysis (testing for interaction between the Schoenfeld residuals and time) and including an interaction term with time in the competing risk model.20 We performed a prespecified sensitivity analysis to confirm that the conclusion was valid when patients with any event within 30 days postoperatively had been excluded (Supplementary Table IV, online only). The difference in survival between patients who had undergone amputation and those without amputation was evaluated using a Cox model that included amputation as a time-dependent covariate. Two-tailed P values < .05 were considered statistically significant. The analyses were performed using SPSS Statistics, version 25 (IBM Corp, Armonk, NY). The competing risk analysis was performed using R statistical analysis, version 3.4.3 with the package cmprsk, version 2.2-7 (Foundation for Statistical Computing, Vienna, Austria; available at: http://www.R-project.org).

RESULTS standard survival methods such as the Kaplan-Meier method and Cox proportional hazards regression is that they depend on the assumption that the censoring is independent (ie, that those who have been censored had had an equal risk of the event as those who have remained in the study).6 Censoring because of mortality is not independent for patients with CLI because the risk of amputation will be influenced by the patients’ morbidity.6 The cumulative incidence function overcomes this issue by producing separate estimates for the outcome of interest (eg, amputation) and the competing event (eg, death).7,15 The results from competing risk regression were compared with those from standard Cox proportional hazards regression,19 using AFS as the outcome measure. Further details regarding competing risk regression are provided in the Appendix (online only). We calculated a crude unadjusted sub-HR using competing risk regression and a cause-specific hazard ratio (HR) using Cox regression for each risk factor. We then built multivariable competing risk and Cox regression models that included clinically significant factors

Perioperative characteristics The cohort included 855 patients (408 men), with a mean age of 77 years (range, 41-97). Diabetes, previous stroke, and renal impairment were more common in patients who had required an amputation during the follow-up period compared with those without. In contrast, current smoking, hypertension, and cardiac disease did not differ significantly between the two groups (Table I). The revascularization procedures were similar in patients with and without amputation, and most (n ¼ 630) had undergone an endovascular procedure. The most common procedure for open surgery was bypass with a vein vs a plain balloon for the group treated with an endovascular method. The type of revascularization procedure was not significantly associated with amputation (P ¼ .49). However, those who had received a bare metal stent and had undergone subintimal angioplasty had an increased rate of amputation compared with those who had undergone other endovascular procedures. The anatomic location of the treated segment was not associated with the incidence of amputation (Table II).

Journal of Vascular Surgery Volume

-,

Number

Torbjörnsson et al

5

-

Fig 1. Cumulative incidence and 95% confidence interval for amputation and death during the follow-up period for the cohort.

Amputation and mortality during follow-up The median follow-up period for survival was 6.1 years (interquartile range, 5.2-7.4), equivalent to 3613 personyears of observation. The shortest and longest event-free follow-up period was 1472 and 3278 days, respectively. During the follow-up period, 178 patients had required an ipsilateral major amputation, of which 130 were below the knee. The mean interval to amputation was 390 days (95% confidence interval [CI], 298-481). The cumulative incidence of ipsilateral amputation at 1 year was 15.0% and was 20.0% at 5 years. During the follow-up period, 415 patients had died. The cumulative incidence for all-cause mortality at 1 and 5 years was 12.5% and 32.3%, respectively. Most of the amputations had been performed during the first year after revascularization. In contrast, the death rate was constant during the follow-up period (Fig 1). Patients who had required an amputation during the follow-up period had lower 5-year survival (38% vs 59%; P < .001, log-rank test; Cox model with amputation as the time-dependent covariate). Risk factors associated with amputation Univariable analysis using competing risk vs Cox regression. Overall, fewer risk factors were associated with major amputation on univariable analysis using the competing risk method compared with the Cox

regression method. Additionally, the competing risk analysis produced different estimates of the risk compared with the Cox regression analysis. The competing risk analysis identified five factors associated with amputation: male sex, older age, renal impairment, worse ambulatory status, and ischemic wound as the indication for surgery (Table III). In contrast, the Cox regression analysis yielded 8 factors associated with shorter AFS: older age, cardiac disease, diabetes, stroke, renal impairment, worse ambulatory status, ischemic wound as the indication for surgery, and multilevel revascularization. Additional information regarding the details in the results from the analyses using Cox regression are provided in Supplementary Table II (online only). In the univariable competing risk analysis, patients with ischemic wounds had a nearly 4 times greater risk of amputation compared with those with rest pain (sub-HR, 3.6; 95% CI, 2.1-6.3). Renal impairment was also associated with an increased relative risk of amputation (sub-HR, 2.0; 95% CI, 1.4-2.9). Patients with renal impairment had a more than tripled incidence of amputation during the follow-up period compared with those without (13.4 [95% CI, 9.2-18.8] and 4.3 [95% CI 3.6-5.1] per 100 person-years, respectively).

Journal of Vascular Surgery

Torbjörnsson et al

6

---

Table III. Univariable analysis for predicting amputation using competing risk regression Variable

Amputation (death as competing risk)

Sex, female vs male

0.71 (0.53-0.96)

Age, continuous

0.99 (0.97-1.00)

Current smoker

1.18 (0.82-1.69)

Comorbidity Cardiac risk

1.05 (0.80-1.43)

Pulmonary disease

0.63 (0.40-1.01)

Diabetes Hypertension Stroke Renal impairment

1.27 (0.95-1.70) 0.84 (0.60-1.18) 1.32 (0.92-1.91) 2.00 (1.36-2.95)

Ischemic grade, WIfI systema 0

Reference

1

0.93 (0.21-4.22)

2

1.01 (0.24-4.26)

3

1.24 (0.30-5.08)

Ambulatory status Independent Walking with aid

Reference 1.82 (1.30-2.54)

Wheelchair-bound

2.70 (1.52-4.80)

Bedridden

4.06 (2.17-7.61)

Ischemic wound vs rest pain Method (open/hybrid vs endovascular)

3.59 (2.05-6.29) 0.90 (0.64-1.27)

Level of revascularization Femoropopliteal

Reference

Crural/pedal

1.48 (0.85-2.56)

Multilevelb

1.27 (0.92-1.77)

ABI, Ankle-brachial index; WIfI, Wound, Ischemia, and foot Infection. Data are presented as subdistribution hazard ratio from univariate Fine-Gray model with 95% confidence intervals in parentheses. a Grade 0, ABI $0.80, toe pressure $60 mm Hg; grade 1, ABI $0.6-0.79, toe pressure 40-59 mm Hg; grade 2, ABI $0.4-0.59, toe pressure 3039 mm Hg; grade 3, ABI #0.39, toe pressure <30 mm Hg. b Both femoropopliteal and crural.

Between each of the ambulatory status levels, the risk of amputation increased. Thus, patients using a walking aid had a nearly 2 times greater risk (sub-HR, 1.8; 95% CI, 1.3-2.5) of amputation compared with those walking independently preoperatively (data on the other levels are provided in Table III). The greatest risk of amputation was in bedridden patients (sub-HR, 4.1; 95% CI, 2.2-7.6; Fig 2). The cumulative incidence of amputation among the patients who had used a walking aid preoperatively was 6% at 30 days compared with 2% at 30 days for those who had been walking independently preoperatively (P < .001, Gray’s test). Current smoking and hypertension were not associated with amputation, nor was the method of intervention.

2019

Multivariable analysis using competing risk vs Cox regression analysis. We built a multivariable model that included the risk factors that were statistically significant (P # .05) on univariable analysis. The multivariable competing risk model yielded 4 factors (ie, male gender, older age, worse ambulatory status, and ischemic wound as surgical indication) that were associated with an increased risk of amputation. In contrast, older age, current smoking, cardiac disease, diabetes, renal impairment, worse ambulatory status, and ischemic wound as an indication for surgery were associated with decreased AFS in the multivariable Cox regression model. Thus, their relative importance differed compared with the results from the competing risk model (Supplementary Table III, online only). Our main analysis included all amputations, both early (#30 days) and long term. Because perioperative amputations can result from different risk factors than late amputations, we performed a subanalysis in which we excluded all the patients who had died or had required an amputation within 30 days postoperatively. The reason was to determine whether the risk factors for long-term amputation was influenced by those for an early event. However, no differences were found in the risk factors between those requiring an early postoperative amputation compared with a late amputation, except for women vs men (Supplementary Table IV, online only). Final reduced risk model for amputation. For the final reduced model, we selected risk factors associated with the risk of amputation from the initial multivariable model (P # .05). The reduced competing risk model identified older age, worse ambulatory status, and ischemic wound as the surgical indication as risk factors associated with amputation (Table IV). In further analyses, with the aim of comparing the different competing risk models, we found that the inclusion of variables such as smoking status, pulmonary disease, and WIfI grade did not change the model or estimates, nor did the Bayes information criterion improve. In summary, our findings have shown that an impaired preoperative ambulatory status has a high effect on amputation even when the impairment is small. Older age and the presence of an ischemic wound were also important variables in the risk assessment for patients undergoing revascularization. Sex-specific analysis We performed a prespecified sex-specific analysis. The women in the study cohort were older (mean age, 80 vs 74 years for men). The amputation incidence was slightly increased among the men (5.8 amputations per 100 person-years [95% CI, 4.7-7.0] vs 4.2 amputations per 100 person-years [95% CI, 3.2-5.2] for women; P ¼ .032, mid-P exact). The cumulative incidence of an amputation at 1 year was 13% for women compared

Journal of Vascular Surgery Volume

-,

Number

Torbjörnsson et al

7

-

Fig 2. Cumulative incidence and 95% confidence interval for amputation vs walking ability during the follow-up period for the cohort.

with 18% for men (P ¼ .02, Gray’s test). The mortality rate did not differ between the men and women; however, the mean interval until death was slightly shorter for the men (645 days vs 700 days for women).

DISCUSSION The present cohort study of patients with CLI who had undergone lower extremity revascularization showed that the risk factors for amputation differed when analyzing the data using a competing risk analysis15 vs Cox regression analysis19 when using AFS as the outcome measure. Older age, worse ambulatory status, and ischemic wound as an indication for intervention were the only risk factors that remained associated with amputation in the reduced competing risk model. The differences in the risk factors between the 2 models can be explained by the high mortality in the cohort (mortality of 49% in our population during the followup period). The findings from previous studies have supported the existence of a risk for biased estimates using standard survival methods when a strong competing risk such as death is present.6,13 The results from a Cox regression analysis will only be valid if the censoring is independent (ie, those censored have the same future

prognosis to experience the event as those who remained in the study).7 Because the risk of amputation was increased for those who had died during the followup period, we concluded that a risk of biased estimates exists using Cox regression analysis. By focusing on the factors specifically associated with amputation, a competing risk model could be a valuable tool to support the choice of treatment strategy. Vascular and endovascular surgery can cause severe complications, highlighting the importance of identifying those with a high risk of amputation and thereby requiring extended postoperative and follow-up care after revascularization. A competing risk model could also increase the possibility to improve patient selection for vascular intervention, because, for some patients, primary amputation might be preferable. Moreover, qualitative studies have reported that some patients regretted not undergoing amputation earlier, because they had then experienced relief of the pain and wounds.21 We believe that some of the commonly established risk factors for a poor outcome after revascularization for CLI, such as renal impairment, diabetes, and cardiac disease, are the factors affecting survival, and not the specific

Journal of Vascular Surgery

Torbjörnsson et al

8

---

Table IV. Reduced multivariable risk model for amputation using competing risk regression Variable

Amputation (death as competing risk)

Sex Male Female Age, continuous Renal impairment

Reference 0.77 (0.56-1.05) 0.98 (0.97-1.00) 1.48 (0.97-2.25)

Ambulatory status Independent

Referencea

Walking with aid

1.79 (1.27-2.52)

Wheelchair-bound

2.34 (1.30-4.19)

Bedridden

4.10 (2.14-7.86)

Indication Rest pain Ischemic wound

Reference 3.03 (1.72-5.36)

Data are presented as subdistribution hazard ratio (multivariate FineGray model) with 95% confidence intervals in parentheses. a P (overall) < .0001, Wald test (c2 ¼ 24.0, df ¼ 3).

amputation risk in the combined AFS end point. Heikkila et al13 reported a 5-year incidence of major amputation after revascularization with lower estimates using a competing risk approach than with using Cox regression. The investigators concluded that a risk exists for the incorrect interpretation of results when ignoring the competing risk of death using standard survival methods.13 In contrast to our findings, Heikkila et al13 found a greater risk of amputation among those who had undergone open revascularization. This could be explained by the different inclusion criteria; in addition to patients with CLI, they had included patients with claudication. Furthermore, they had not separated contralateral amputations from ipsilateral amputations. However, we included ipsilateral amputation exclusively. In our study, the preoperative ambulatory status was strongly associated with the risk of amputation. Even a small impairment had an effect on the outcome, such that a patient using a walking aid had a worse outcome than a patient walking independently. The reason for this was not clear. However, a selection bias could have been present such that a less ambulatory patient would not have been selected for repeat surgery if the vascular reconstruction had failed. The ambulatory status has previously been shown to predict the likelihood of death after revascularization22,23 and to have a significant effect on graft occlusion and amputation after surgery.24 However, these studies only compared independent ambulatory status and nonindependent ambulatory status. Current smoking was, surprisingly, not associated with amputation in our competing risk analysis. However, smoking was associated with worse AFS in our standard Cox regression analysis, consistent with the findings from previous studies.25,26 The latter finding indicates that

2019

most of the decreased AFS among smokers will be driven by their greater mortality.26 Women had a lower incidence of amputation compared with men. Previous studies regarding sex differences and CLI did not support these findings.27 Our baseline data showed that men had increased comorbidities compared with women. It is possible that the preoperative selection of patients could have contributed to this finding. The use of competing risk regression has been criticized for presenting a hypothetical situation (ie, patients who have died would have no interest in whether they might require an amputation). However, Cox regression and competing risk regression analyses have different applications. If the aim is to compare the failure rate of different procedures or to study which treatment will result in the best patency, Cox regression should be used. However, if the aim is to identify high-risk patients and communicate the amputation risk to individual patients, the use of competing risk regression analysis to estimate the absolute risk would be preferable.7,28 Strengths and limitations. The use of the competing risk approach was a strength of the present study, because it allowed for an estimation of the absolute risk of amputation and death separate from each other.8 Another strength was the inclusion of patients who had undergone open and endovascular surgery, providing a cohort representative of a contemporary vascular service.29 Furthermore, our cohort was population-based and prospective, with long-term follow-up data available. One limitation was that our study was a single-center study. However, the nearly complete data available for the variables and that all amputations had been validated during a review of the medical records provided high internal validity. We believe the external validity was also high, at least in Sweden, because the differences between centers in Sweden are very small.16 Another limitation of the present study was that we only had data available for the ankle and toe pressures and lacked information regarding the other two components in the WIfI classification (wound size and infection). Of the 855 patients, 12 had had an ischemic status of grade 0 using the WIfI system. Those patients were included because they had undergone revascularization to treat diabetic foot ulcers that had not healed with conservative treatment despite the presence of only minor ischemia according to the European Society for Vascular Surgery guidelines.30

CONCLUSIONS The use of competing risk analysis could improve the prediction of the risk of amputation and, thereby, could be a valuable tool to support the choice of the treatment strategy for individual patients. In our study, only 3 factors (ie, older age, worse ambulatory status, ischemic wound

Journal of Vascular Surgery Volume

-,

Number

Torbjörnsson et al

9

-

as the surgical indication) were associated with amputation after revascularization using the competing risk approach. The differences between the results using Cox regression analysis and those using competing risk regression showed that a risk exists for biased estimates using standard survival methods when a strong competing risk such as death is present. The present study would not have been possible without the continuous registration efforts in the Swedvasc registry by the vascular surgeons, interventional radiologists, and nurses at Södersjukhuset, Stockholm, Sweden.

AUTHOR CONTRIBUTIONS Conception and design: ET, LB, AMF, LBL, CO, JM Analysis and interpretation: ET, JM Data collection: ET, JM Writing the article: ET, LBL, JM Critical revision of the article: ET, LB, AMF, LBL, CO, JM Final approval of the article: ET, LB, AMF, LBL, CO, JM Statistical analysis: ET, JM Obtained funding: Not applicable Overall responsibility: JM

10.

11.

12.

13.

14.

15.

16.

17.

18.

REFERENCES 1. Sigvant B, Lundin F, Wahlberg E. The risk of disease progression in peripheral arterial disease is higher than expected: a meta-analysis of mortality and disease progression in peripheral arterial disease. Eur J Vasc Endovasc Surg 2016;51:395-403. 2. Conte MS. Understanding objective performance goals for critical limb ischemia trials. Semin Vasc Surg 2010;23:129-37. 3. Norgren L, Hiatt WR, Dormandy JA, Nehler MR, Harris KA, Fowkes FG, et al. Inter-Society Consensus for the Management of Peripheral Arterial Disease (TASC II). Eur J Vasc Endovasc Surg 2007;33(Suppl 1):S1-75. 4. Schanzer A, Hevelone N, Owens CD, Belkin M, Bandyk DF, Clowes AW, et al. Technical factors affecting autogenous vein graft failure: observations from a large multicenter trial. J Vasc Surg 2007;46:1180-90. 5. Bradbury AW, Adam DJ, Bell J, Forbes JF, Fowkes FG, Gillespie I, et al. Bypass versus Angioplasty in Severe Ischaemia of the Leg (BASIL) trial: an intention-to-treat analysis of amputation-free and overall survival in patients randomized to a bypass surgery-first or a balloon angioplasty-first revascularization strategy. J Vasc Surg 2010;51(Suppl):5S-17S. 6. Andersen PK, Geskus RB, de Witte T, Putter H. Competing risks in epidemiology: possibilities and pitfalls. Int J Epidemiol 2012;41:861-70. 7. Wolbers M, Koller MT, Stel VS, Schaer B, Jager KJ, Leffondré K, et al. Competing risks analyses: objectives and approaches. Eur Heart J 2014;35:2936-41. 8. Wongworawat MD, Dobbs MB, Gebhardt MC, Gioe TJ, Leopold SS, Manner PA, et al. Editorial: estimating survivorship in the face of competing risks. Clin Orthop Relat Res 2015;473:1173-6. 9. Van Der Pas S, Nelissen R, Fiocco M. Different competing risks models for different questions may give similar results

19. 20. 21.

22.

23.

24.

25.

26.

in arthroplasty registers in the presence of few events. Acta Orthop 2018;89:145-51. Wolkewitz M, Cooper BS, Bonten MJ, Barnett AG, Schumacher M. Interpreting and comparing risks in the presence of competing events. BMJ 2014;349:g5060. Wolbers M, Koller MT, Witteman JC, Steyerberg EW. Prognostic models with competing risks: methods and application to coronary risk prediction. Epidemiology 2009;20: 555-61. Marzona I, Baviera M, Vannini T, Tettamanti M, Cortesi L, Riva E, et al. Risk of dementia and death in patients with atrial fibrillation: a competing risk analysis of a populationbased cohort. Int J Cardiol 2016;220:440-4. Heikkila K, Loftus IM, Mitchell DC, Johal AS, Waton S, Cromwell DA. Population-based study of mortality and major amputation following lower limb revascularization. Br J Surg 2018;105:1145-54. Berry SD, Ngo L, Samelson EJ, Kiel DP. Competing risk of death: an important consideration in studies of older adults. J Am Geriatr Soc 2010;58:783-7. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999;94: 496-509. Troeng T, Malmstedt J, Bjorck M. External validation of the Swedvasc registry: a first-time individual cross-matching with the unique personal identity number. Eur J Vasc Endovasc Surg 2008;36:705-12. Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim JL, Reuterwall C, et al. External review and validation of the Swedish national inpatient register. BMC Public Health 2011;11:450. Mills JL Sr, Conte MS, Armstrong DG, Pomposelli FB, Schanzer A, Sidawy AN, et al. The Society for Vascular Surgery lower extremity threatened limb classification system: risk stratification based on wound, ischemia, and foot infection (WIfI). J Vasc Surg 2014;59:220-34.e1-2. Cox DR. Regression models and life-tables. J R Statist Soc 1972;34:187-220. Schoenfeld D. Partial residuals for the proportional hazards regression model. Biometrika 1982;69:239-41. Torbjörnsson E, Ottosson C, Blomgren L, Boström L, Fagerdahl AM. The patient’s experience of amputation due to peripheral arterial disease. J Vasc Nurs 2017;35:57-63. Goodney PP, Likosky DS, Cronenwett JL; Vascular Study Group of Northern New England. Predicting ambulation status one year after lower extremity bypass. J Vasc Surg 2009;49:1431-9.e1. McDermott MM, Tian L, Liu K, Guralnik JM, Ferrucci L, Tan J, et al. Prognostic value of functional performance for mortality in patients with peripheral artery disease. J Am Coll Cardiol 2008;51:1482-9. Goodney PP, Nolan BW, Schanzer A, Eldrup-Jorgensen J, Bertges DJ, Stanley AC, et al. Factors associated with amputation or graft occlusion one year after lower extremity bypass in northern New England. Ann Vasc Surg 2010;24: 57-68. Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, AdairRohani H, et al. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 2012;380:2224-60. Armstrong EJ, Wu J, Singh GD, Dawson DL, Pevec WC, Amsterdam EA, et al. Smoking cessation is associated with decreased mortality and improved amputation-free survival among patients with symptomatic peripheral artery disease. J Vasc Surg 2014;60:1565-71.

10

Torbjörnsson et al

Journal of Vascular Surgery ---

27. Ferranti KM, Osler TM, Duffy RP, Stanley AC, Bertges DJ; Vascular Study Group of New England. Association between gender and outcomes of lower extremity peripheral vascular interventions. J Vasc Surg 2015;62:990-7. 28. Sayers A, Evans JT, Whitehouse MR, Blom AW. Are competing risks models appropriate to describe implant failure? Acta Orthop 2018;89:256-8. 29. Chung J, Modrall JG, Valentine RJ. The need for improved risk stratification in chronic critical limb ischemia. J Vasc Surg 2014;60:1677-85. 30. Aboyans V, Ricco JB, Bartelink MEL, Bjorck M, Brodmann M, Cohnert T, et al. 2017 ESC Guidelines on the Diagnosis and Treatment of Peripheral Arterial Diseases, in collaboration with the European Society for Vascular

2019

Surgery (ESVS): document covering atherosclerotic disease of extracranial carotid and vertebral, mesenteric, renal, upper and lower extremity arteries. Endorsed by: the European Stroke Organization (ESO), the Task Force for the Diagnosis and Treatment of Peripheral Arterial Diseases of the European Society of Cardiology (ESC) and of the European Society for Vascular Surgery (ESVS). Eur Heart J 2018;39:763-816. Submitted Apr 18, 2019; accepted Jul 15, 2019.

Additional material for this article may be found online at www.jvascsurg.org.

Journal of Vascular Surgery Volume

-,

Number

Torbjörnsson et al

10.e1

-

APPENDIX (online only). Comments regarding competing risk regression. To understand the role of amputation after revascularization it is important to consider which patients will be at risk. The composite outcome of amputation-free survival and interval to death or major amputation (whichever has occurred first) has been recommended as the outcome measure instead of limb salvage, which, by disregarding mortality, could yield biased estimates of the risk of amputation. The issue with limb salvage as the end point is the high mortality among patients with critical limb ischemia, which will lead to a high number of censored patients. Both Kaplan-Meier and Cox proportional hazards regression analyses depend on the assumption that the censored data were noninformative or independent (ie, that those who were censored because of death would have had an equal risk of the event of interest [amputation] as those who had remained in the study). Censoring because of mortality is not independent for patients with critical limb ischemia,

because the risk of amputation is influenced by the patient’s morbidity. The issue with amputation-free survival is that the composite construction prohibits estimation of the amputation risk alone. Thus, the cumulative incidence function would be a better method for use in populations with competing risks because it gives separate estimates of the primary outcome (ie, amputation) and the competing event (ie, death) and avoids the bias due to censoring. For further details regarding competing risk regression, we recommend the following sources: Andersen PK, Geskus RB, de Witte T, Putter H. Competing risks in epidemiology: possibilities and pitfalls. Int J Epidemiol 2012;41:861-870. Wolbers M, Koller MT, Stel VS, Schaer B, Jager KJ, Leffondré K, Heinze G. Competing risks analyses: objectives and approaches. Eur Heart J 2014;35:2936-2941. Wongworawat MD, Dobbs MB, Gebhardt MC, Gioe TJ, Leopold SS, Manner PA, Rimnac CM, Porcher R. Editorial: estimating survivorship in the face of competing risks. Clin Orthop Relat Res 2015;473:1173-1176.

10.e2

Journal of Vascular Surgery

Torbjörnsson et al

---

2019

Supplementary Table I (online only). Validation of Swedvasc registry dataa Variable Sex

Sensitivity, %

Specificity, %

Accuracy, %

NA

NA

100.0

Indication

98.1

99.1

98.8

Main revascularization method

98.1

99.1

98.7

Operated side (right vs left)

NA

NA

100.0

Ambulatory status (independent/walking with aid vs wheelchair-bound/bedridden)

100.0

100.0

100.0

Diabetes

100.0

92.8

95.3

Stroke

71.4

97.7

92.5

Cardiac disease

65.3

96.6

82.2

Hypertension

89.1

93.3

89.7

Renal impairment

92.9

100.0

99.1

Pulmonary disease

84.2

97.7

95.3

Smoking

95.9

72.7

88.8

All values calculated from medical record data as standard, with missing values assumed to be negative. Sensitivity ¼ true positive  100/(true positive þ false negative); specificity ¼ true negative  100/(true negative þ false positive); accuracy (percentage agreement) ¼ (true positive þ true negative)  100/(true positive þ true negative þ false positive þ false negative). NA, Not applicable. a Data show the validation of Swedvasc using a random sample of 103 patients (12%) from the study cohort. Data on operative details and comorbidity from Swedvasc were compared with corresponding medical records and sensitivity, specificity, and accuracy calculated. All variables had an accuracy of $90%, except for cardiac disease.

Journal of Vascular Surgery Volume

-,

Number

Torbjörnsson et al

10.e3

-

Supplementary Table II (online only). Univariable competing risk regression vs Cox regression models for predicting amputation, amputation-free survival, and mortality after revascularizationa Amputation (death as competing risk; sub-HR (95% CI)

Amputation-free survival (HR; 95% CI)

Sex, female vs male

0.71 (0.53-0.96)

0.86 (0.71-1.02)

Age, continuous

0.99 (0.97-1.00)

1.02 (1.00-1.03)

1.03 (1.02-1.04)

Current smoker

1.18 (0.82-1.69)

1.08 (0.85-1.36)

1.04 (0.81-1.34)

Variable

All-cause mortality (HR; 95% CI) 0.90 (0.74-1.09)

Comorbidity Cardiac risk

1.05 (0.80-1.43)

1.52 (1.27-1.83)

1.74 (1.43-2.11)

Pulmonary disease

0.63 (0.40-1.01)

1.25 (0.98-1.59)

1.44 (1.12-1.85)

Diabetes Hypertension Stroke

1.27 (0.95-1.70)

1.46 (1.22-1.75)

1.40 (1.16-1.70)

0.84 (0.60-1.18)

0.89 (0.72-1.10)

0.91 (0.73-1.14)

1.32 (0.92-1.91)

1.33 (1.04-1.68)

1.33 (1.04-1.72)

2.00 (1.36-2.95)

2.36 (1.84-3.01)

2.68 (2.09-3.45)

0

Reference

Reference

Reference

1

0.93 (0.21-4.22)

1.56 (0.62-3.94)

2

1.01 (0.24-4.26)

1.05 (0.43-2.59)

0.85 (0.35-2.11)

3

1.24 (0.30-5.08)

1.49 (0.61-3.60)

1.30 (0.54-3.16)

Renal impairment Ischemic grade, WIfI systemb

1.46 (0.58-3.70)

Ambulatory status Independent Walking with aid

Reference

Reference

Reference

1.82 (1.30-2.54)

1.81 (1.48-2.21)

1.92 (1.55-2.38)

Wheelchair-bound

2.70 (1.52-4.80)

2.88 (2.01-4.13)

2.85 (1.95-4.18)

Bedridden

4.06 (2.17-7.61)

4.00 (2.67-5.98)

3.03 (1.96-4.69)

Indication (ischemic wound vs rest pain)

3.59 (2.05-6.29)

2.89 (2.15-3.89)

2.76 (2.01-3.79)

Method (open/hybrid vs endovascular)

0.90 (0.64-1.27)

0.91 (0.73-1.12)

0.95 (0.76-1.18)

Level of revascularization Femoropopliteal

Reference

Reference

Reference

Crural/pedal

1.48 (0.85-2.56)

0.98 (0.66-1.46)

0.90 (0.59-1.38)

Multilevelc

1.27 (0.92-1.77)

1.25 (1.02-1.54)

1.20 (0.96-1.49)

ABI, Ankle-brachial index; CI, confidence interval; HR, hazard ratio (univariate Cox regression); sub-HR, subdistribution hazard ratio (univariate FineGray model); WIfI, Wound, Ischemia, and foot Infection. a Comments regarding the analyses: the risk associated with an ischemic wound was lower but still significant on Cox regression analysis (HR, 2.9; 95% CI, 2.2-3-9) compared with the analysis using competing risk regression. Worse preoperative ambulation status was associated with amputationd patients using a walking aid had a nearly 2 times greater risk (HR, 1.8; 95% CI, 1.5-2.2) for amputation compared with those walking independently preoperatively on Cox regression analysis. Renal impairment was associated with an almost similar increased relative risk using Cox regression or competing risk regression analysis (sub-HR, 2.0; 95% CI, 1.4-2.9 vs HR, 2.4; 95% CI, 1.8-3.0). In contrast to the analysis using competing risk regression, severe ischemia was not associated with decreased amputation-free survival on Cox regression analysis. b Grade 0, ABI $0.80, toe pressure $60 mm Hg; grade 1, ABI $0.6-0.79, toe pressure 40-59 mm Hg; grade 2, ABI $0.4-0.59, toe pressure 30-39 mm Hg; grade 3, ABI #0.39, toe pressure <30 mm Hg. c Both femoropopliteal and crural.

10.e4

Journal of Vascular Surgery

Torbjörnsson et al

---

2019

Supplementary Table III (online only). Multivariable full risk model for amputation and death using competing risk regression vs Cox regressiona Amputation (death as competing risk; sub-HR; 95% CI)

Amputation-free survival (HR; 95% CI)

Sex, female vs male

0.71 (0.51-0.98)

0.83 (0.68-1.01)

0.87 (0.71-1.08)

Age, continuous

0.98 (0.97-1.00)

1.02 (1.01-1.03)

1.03 (1.02-1.04)

Current smoker

1.08 (0.73-1.61)

1.37 (1.07-1.77)

1.48 (1.13-1.94)

Variable

All-cause mortality (HR; 95% CI)

Comorbidity Cardiac disease

0.98 (0.71-1.36)

1.27 (1.05-1.54)

1.41 (1.15-1.73)

Pulmonary disease

0.65 (0.40-1.06)

1.26 (0.98-1.62)

1.44 (1.11-1.87)

Diabetes

0.89 (0.63-1.26)

1.33 (1.09-1.62)

1.30 (1.05-1.60)

Stroke

1.26 (0.85-1.86)

1.11 (0.87-1.42)

1.10 (0.85-1.42)

Renal impairment

1.37 (0.88-2.13)

1.88 (1.45-2.44)

2.34 (1.79-3.05)

0

Referencec

Referenced

Reference

1

0.82 (0.18-3.78)

1.45 (0.57-3.69)

1.42 (0.56-3.64)

2

1.03 (0.24-4.44)

1.14 (0.46-2.83)

0.94 (0.38-2.34)

3

1.09 (0.26-4.60)

1.31 (0.54-3.20)

1.13 (0.46-2.76)

Referencee

Referencef

Ischemic grade, WIfi systemb

Ambulatory status Independent

Reference

Walking with aid

1.93 (1.37-2.74)

1.46 (1.18-1.80)

1.43 (1.14-1.79)

Wheelchair-bound

2.37 (1.26-4.46)

2.06 (1.40-3.02)

1.87 (1.24-2.82)

Bedridden

3.84 (1.98-7.45)

3.45 (2.28-5.23)

2.67 (1.70-4.18)

Indication Rest pain Ischemic wound

Reference

Reference

Reference

2.87 (1.63-5.03)

2.21 (1.63-3.00)

2.08 (1.50-2.88)

ABI, Ankle-brachial index; CI, confidence interval; HR, hazard ratio (multivariate Cox regression analysis); sub-HR, subdistribution hazard ratio (multivariate Fine-Gray model). a Comments regarding the analyses: risk factors for amputation differed between the analyses using competing risk regression and Cox regression. In the competing risk model, male sex, older age, worse ambulatory status, and ischemic wound were associated with an increased risk of amputation. In contrast, in the multivariable Cox regression model, older age, current smoking, cardiac disease, diabetes, renal impairment, worse ambulatory status, and ischemic wound as an indication for surgery were associated with decreased amputation-free survival; thus, their relative importance differed between the competing risk and Cox regression analyses. b Grade 0, ABI $0.80, toe pressure $60 mm Hg; grade 1, ABI $0.6-0.79, toe pressure 40-59 mm Hg; grade 2, ABI $0.4-0.59, toe pressure 30-39 mm Hg; grade 3, ABI #0.39, toe pressure <30 mm Hg. c P (overall) ¼ .003, Wald test (c2, 16.2; df ¼ 4). d P (overall) ¼ .005, Wald test (c2 ¼ 14.8; df ¼ 4). e P (overall) < .0001, Wald test (c2 ¼ 24.2; df ¼ 3). f P (overall) < .0001, Wald test (c2 ¼ 42.6; df ¼ 3).

Journal of Vascular Surgery Volume

-,

Number

Torbjörnsson et al

10.e5

-

Supplementary Table IV (online only). Multivariable full risk model for amputation for those with no event within 30 days of index revascularization using competing risk analysis vs Cox regression Amputation (death as competing risk; sub-HR; 95% CI)

Amputation-free survival (HR; 95% CI)

Sex, female vs male

0.74 (0.51-1.07)

0.82 (0.65-1.02)

Age, continuous

0.98 (0.96-1.00)

1.03 (1.02-1.04)

Current smoker

1.23 (0.81-1.88)

1.50 (1.13-1.98)

Cardiac disease

0.97 (0.68-1.40)

1.25 (1.01-1.55)

1.36 (1.08-1.71)

Pulmonary disease

0.85 (0.50-1.44)

1.41 (1.08-1.84)

1.47 (1.11-1.94)

Diabetes

0.89 (0.60-1.31)

1.39 (1.11-1.74)

1.34 (1.06-1.69)

Variable

All-cause mortality (HR; 95% CI) 0.83 (0.65-1.05) 1.4 (1.03-1.06) 1.53 (1.13-2.07)

Comorbidity

Stroke

1.39 (0.90-2.16)

1.16 (0.88-1.52)

1.11 (0.83-1.48)

Renal impairment

1.29 (0.77-2.17)

1.66 (1.20-2.29)

2.10 (1.52-2.92)

Ischemic grade, WIfi systema 0

Reference

Reference

Reference

1

0.76 (0.15-3.74)

1.92 (0.68-5.41)

1.89 (0.67-5.35)

2

0.90 (0.19-4.17)

1.38 (0.50-3.78)

1.20 (0.44-3.32)

3

0.78 (0.17-3.55)

1.41 (0.52-3.81)

1.30 (0.48-3.54)

Reference

Reference

Reference 1.40 (1.09-1.79)

Ambulatory status Independent Walking with aid

1.72 (1.17-2.53)

1.38 (1.09-1.75)

Wheelchair-bound

1.98 (0.97-4.05)

1.99 (1.30-3.05)

1.75 (1.10-2.77)

Bedridden

3.27 (1.49-7.18)

3.08 (1.82-5.20)

2.53 (1.44-4.43)

2.85 (1.55-5.24)

2.25 (1.62-3.12)

2.13 (1.50-3.02)

Indication (ischemic wound vs rest pain)

ABI, Ankle-brachial index; CI, confidence interval; HR, hazard ratio (univariate Cox regression); sub-HR, subdistribution hazard ratio (univariate FineGray model). a Grade 0, ABI $0.80, toe pressure $60 mm Hg; grade 1, ABI $0.6-0.79, toe pressure 40-59 mm Hg; grade 2, ABI $0.4-0.59, toe pressure 30-39 mm Hg; grade 3, ABI #0.39, toe pressure <30 mm Hg.