Introduction

Patients who undergo major lower extremity amputation (mLEA) secondary to peripheral arterial occlusive disease (PAOD) have been reported to have a poor prognosis, likely due to the significant comorbidities and risk factors that exist in this population [1, 2]. Despite the decline in postoperative short-term mortality, no significant uniform improvement over time was observed at mid-to-long term follow-up [3, 4].

Prevention, early diagnosis, and aggressive medical and surgical treatment for patients with severe PAOD or infection has been studied, however, mortality rates remain high [3, 4]. Therefore, perioperative risk stratification may play a key role in patient counseling and improving postoperative outcomes [5,6,7].

The aims of this study were to analyze major clinical outcomes while identifying predictors of mortality to generate a risk index score in a contemporary cohort of patients after a first amputation for PAOD and/or infection.

Materials and methods

Study cohort

This is a single-center, retrospective, observational cohort study from a tertiary referral university hospital. We followed the checklist of items recommended by the STROBE statement [8]. For this study, all patients treated with above-the-knee amputation (AKA) or below-the-knee amputation (BKA) between January 1st, 2010 and June 30th, 2018 were identified. Post-hoc analysis identified those who underwent mLEA for PAOD, gangrene, infected non-healing wound. Medical records were reviewed by two senior surgeons (MF and GP). A consort diagram indicating all patients who underwent amputation during the period of study, including the study cohort from which this series was derived is reported in Fig. 1. People with a previous amputation distal to, and including, ankle disarticulation were included in the final analysis. Data collected included demographics, co-morbidities, severity of PAOD, surgical history, blood test (haemoglobin, leukocyte count, C-reactive protein, albumin) operative details (type of anaesthesia, duration of intervention, level of amputation), as well as postoperative events (amputation revision and mortality) during hospitalization and follow-up period. Owing to the retrospective nature of the present study, local Ethical Committee approval was not necessary according to the Italian National Policy in the matter of Privacy Act on retrospective analysis of anonymized data.

Fig. 1
figure 1

Consort diagram of lower extremity amputations during the period of study (January 1st 2010–June 30th 2018; n = number; AKA above-the-knee; BKA below-the-knee)

Full size image

Indication for interventions

Informed consent for prospective data collection and surgical intervention was signed by each patient. The interventions were performed according to the national guidelines of the Italian Society for Vascular and Endovascular Surgery (SICVE), which are consistent with the clinical practice guidelines on the diagnosis and treatment of PAOD of the European Society for Vascular Surgery (ESVS) [9, 10]. In general, primary amputation was performed in patients with extremely limited life expectancy, extensive necrosis or infectious gangrene, non-ambulatory status. Secondary amputation was performed when revascularization attempts failed and re-intervention was no longer possible because of the absence of a target vessel, or when the limb continued to deteriorate because of infection or necrosis despite adequate blood flow and optimal medical management. The level of amputation was determined based on the clinical judgment of the multidisciplinary team (vascular surgeon and anesthesiologist). Factors assessed included pre-existing limb-threatening ischemia and/or infection and decreased likelihood of salvageability. Computed tomography angiography magnetic resonance, or conventional angiography was not routinely performed to dictate the level of amputation. Indications for the choice of AKA rather than BKA included extensive gangrene or infection, flexion contracture of the knee ≥ 30°, or pre-existing prolonged non-ambulatory status. The type of anesthesia (general vs. spinal/epidural) was at anesthesiologist’s judgment. All patients received perioperative antibiotics. Patients who had no tissue loss or infection received short-term use of cefazolin (2gr b.i.d.; Cefamezin—Pfizer; Milan—IT). Those with tissue loss or infection received broad-spectrum antibiotics consisting of a glycopeptide (Vancotex®—Pharmatex; Milano—IT) and penicillin/beta-lactamase inhibitor (Textazo®—Pharmatex; Milano—IT), unless there was microbiological data already available with drug sensitivities. Postoperatively, electrocardiograms and cardiac enzyme analysis were performed. Generally, routine intensive care unit admission was not indicated per protocol, while rehabilitation transfer was offered to almost all patients to reach personalized outcomes.

Definition and primary outcomes

Medical comorbidity grading system and operative outcomes were defined according to the Society for Vascular Surgery (SVS) [11]. Chronic kidney disease was defined in agreement with the clinical practice guidelines of the Kidney Disease Improving Global Outcomes [12]. Chronic obstructive pulmonary disease was defined accordingly to the GOLD executive summary [13]. Rutherford classification was used to define critical limb ischemia (CLI) or acute limb ischemia (ALI) [9,10,11]. Patient’s frailty was assessed using the modified Frailty Index (mFI) [7]. The mFI consists of eleven parameters which generate a frailty score, by giving 1 point for each component and a maximum score of 11.Frailty patient was classified who had a cutoff of mFI > 2 [7]. Failure of the initial amputation was defined as the need for conversion to a higher level. Conversion of BKA to AKA was performed for failed BKA, defined as the presence of non-healing tissues with extensive deep infection or wound disruption, or extensive stump tissue ischemia. The Clavien-Dindo grading system was used to classify postoperative complications [14]. Follow-Up Index (FUI) describes follow-up completeness at a given study end date as a ratio between the investigated and the potential follow-up period [15]. Through December 2020, information on reintervention, vital status, and date of death of individual patients were validated by death certificate, electronic charts managed by the regional health care system, or certified data from Emergency Department admission. For this study, the primary outcome of interest was early (≤ 30 days) mortality. Secondary outcomes were postoperative complications and freedom from amputation stump revision/failure. Time to death was calculated from the date of the first amputation.

Statistical analysis [16]

Clinical data were recorded and tabulated in Microsoft Excel (Microsoft Corp—Redmond; Wash—USA) database. Statistical analysis was performed by means of SPSS 26.0 for Windows (IBM SPSS—Chicago; Ill—USA). Considering the reported median 9% rate mortality at 30-days, an ɑ cut-off of 0.05and a power of 90%, for a 15% expected mortality our cohort would have enrolled a total of 288 patients. Categorical variables were presented using frequencies and percentages. Continuous variables were presented with mean ± standard deviation (SD), or median with interquartile range (IQR) and ranges, based on data distribution. Categorical variables were analyzed with the χ2 test, and Fisher’s exact test when appropriate. Continuous variables were tested for normal distribution by the Shapiro–Wilk’s test and compared between groups with unpaired Student’s T-test for normally distributed values; otherwise, the Mann–Whitney U test was used. Tukey’s honest significance test was used as single-step multiple comparison to find a significant difference among means. Univariate analysis was used to identify potential predictors of mortality at 30-days. Associations that yielded a P value < 0.20 on univariate screen were then included in a binary logistic regression analysis using the Wald’s forward stepwise model. The strength of the association of variables with mortality was estimated by calculating the odd ratio (OR) and 95% confidence intervals [(95% CI): significance criteria 0.25 for entry, 0.05 for removal)]. Model discrimination was evaluated by using the area under the receiver operating characteristic (AUROC) curve, with ≥ 0.7 being considered significantly accurate. A risk score for mortality at 30-days was then constructed by dividing the β-coefficient of each significant predictor by 0.25 and then by rounding off to the nearest integer value. Cox’s regression analysis was used to assess the strength of the association of covariates with mortality. First, the univariate analysis to identify potential predictors of mortality using the Kaplan–Meier survival estimates and log-rank test for each covariate. Associations that yielded a P value < 0.20 on univariate screen were then included in a forward regression analysis, and the strength of association between covariates and mortality was estimated by calculating the hazard ratio (HR) and 95% CIs. All survival analyses were estimated with the Kaplan–Meier test and reported as percentage ± standard error (SE) with 95% CI. All reported P values were two-sided; P value < 0.05 was considered significant.

Results

Study cohort

During the study period, we identified 310 (77.7%) mLEAs performed on 286 patients. This group consisted of 188 (65.7%) men and 98 (34.3%) women. Considering the entire cohort, the median age was 79 years (IQR 69–83). Demographic data, comorbidities, and risk factors are reported in Table 1. The median mFI was 4 (IQR 3–6). Indications for major amputation were as follow: CLI in 235 (75.8%), ALI in 46 (14.8%), and infection in 29 (9.4%) unrelated to PAOD. There were 163 (52.6%) primary amputations and 147 (47.4%) secondary amputations. We performed 257 (82.9%) AKA and 53 (17.1%) BKA. In 70 (22.6%) cases, a prior ipsilateral minor amputation had been performed. The intervention was performed under general anesthesia in 212 (68.4%) cases and with spinal/epidural in 98 (31.6%).

Table 1 Demographic data, comorbidities, and risk factors of the entire cohort (n = 310)
Full size table

Early outcomes (< 30 days)

There were no intraoperative deaths. Duration of the intervention was < 60 min in 158 (50.9%) patients and > 60 min in 152 (49.1%). The median length of hospitalization was 8 days (IQR 5–15 days). Complications were observed in 42 (13.5%) cases, which are described in Table 2. An intervention performed for Rutherford stage 5–6 (OR 10.3, 95% CI 2.20–47.76; P = 0.003) and BKA (OR 3.88, 95% CI 1.58–9.54; P = 0.012) was independently associated with the development of a postoperative complication. Early death occurred in 49 (15.8%) patients with the causes of death listed in Table 3. Early mortality did not differ among the different quartiles of age (15.4% vs. 14.3% vs. 15.4% vs. 19.5%, P = 0.826). Binary logistic regression analysis identified three predictive variables associated with 30-day mortality: age > 80 years (OR 2.24, 95% CI 1.17–4.31; P = 0.015), chronic obstructive pulmonary disease (OR 2.12, 95% CI 1.11–4.06; P = 0.023), and hemodialysis (OR 2.52, 95% CI 1.15–5.52; P = 0.021), listed in Table 4. The integer score assigned to each covariate was used to calculate an individual risk score for mortality at 30 days. The score ranged from 0 to 10 (median 3; IQR 0–4) owing to the sum of the three predictors (Table 5). On the basis of the assigned score, we identified two subgroups with varying mortality rates at 30 days: a lower-risk subgroup (score < 4, 10.8%) and a higher-risk subgroup (score ≥ 4: 28.7%; OR 3.2, 95% CI 1.63–6.32; P < 0.001 compared to the lower-risk group). The ROC analysis (AUROC 0.66, 95% CI 0.58–0.75) had reasonably good discrimination for the obtained multivariable model (Fig. 2). None of the blood tests nor operative variables were significantly associated with the development of a complication.

Table 2 Postoperative complication classified with the Clavien-Dindo severity grade system
Full size table
Table 3 Causes of early death
Full size table
Table 4 Univariate screen and multivariable analysis for early mortality and postoperative complications
Full size table
Table 5 Proposed preliminary score for early mortality estimate
Full size table
Fig. 2
figure 2

Receiver operating characteristic curve for the multivariate model evaluating the risk score for mortality at 30 days (AUROC area under the receiver operating characteristic curve)

Full size image

Late outcomes

During the follow-up period, 175of the 261 (67%) patients died. The median FUI was 0.3 (IQR 0–1). Median survival was 19 months (IQR 7–43): estimated overall survival was 55.7% (SE 0.28; 95% CI 50.2–61.1) at 1-year, 36.6% (SE 0.29; 95% CI 31.1–42.4) at 3-year, and 25.4% (SE 0.28; 95% CI 20.3–31.2) at 5-year (Fig. 3). Long-term survival was different between the two categories of risk for early mortality. The risk of mortality in the higher-risk group was 1.8x (60.8% vs. 33.3%; log-rank χ2 = 12.9, P < 0.001) that of the lower-risk group (Fig. 3.). Long-term analysis through the Cox’s regression analysis identified four variables associated with mortality: need for AKA (HR 1.61, 95% CI 1.04–2.50; P = 0.032), age > 80 years (HR 1.69, 95% CI 1.28–2.24; P < 0.001), end-stage renal disease (HR 1.37, 95% CI 1.03–1.82; P = 0.028), and congestive heart failure (HR 1.60, 95% CI 1.22–2.11; P = 0.001). During the study time period, 24 (7.7%) patients underwent bilateral major amputation (AKA, n = 13; BKA, n = 11). Surgical revision of the amputation stump was required in 25 (8.1%) patients. Failure of a BKA to heal occurred in 3 (1.1%) cases at 2, 10, and 13 months after the initial amputation requiring conversion to an AKA. Freedom from amputation stump revision/failure was similar between BKA and AKA (Log-rank χ2 = 1.77, P = 0.183) as reported in Fig. 4. No preoperative and intraoperative variables were associated with the need for stump revision.

Fig. 3
figure 3

Risk analysis stratified by risk score categories

Full size image
Fig. 4
figure 4

Kaplan–Meier estimates of survival stratified by level of amputation (AKA above-the-knee; BKA below-the-knee)

Full size image

Discussion

There are three findings of significance in this study: mortality after mLEA remains high and unchanged through the years, risk stratification is not adequately sensitive, and older age (e.g., ≥ 80) is the most concrete predictor of a major adverse outcome following a mLEA.

Year after year, many studies have reported consistently high mortality rates after mELAs, notwithstanding a more aggressive policy toward peripheral revascularization, better medical management, and preoperative optimization, in addition to anesthetic improvements [2, 17,18,19]. In our experience, early mortality remained unchanged for the past decade and is consistent with the 7.6–22.5% reported in several real-world experiences, not falling below 14% in the four quartiles of the period of study (Table 6) [3, 5,6,7, 17,18,19,20,21,22,23,24,25,26,27]. Our results are similar to Jones et al. [3], who analyzed 186,338 older patients with identified PAOD who underwent mLEA, namely the largest cohort published up to date. Though, there appears to have been a decline in the short-term, mortality rates were similar at the beginning and end of his study never falling below 12.7%. One must also consider the fact that nearly 63% of our patients were on best medical therapy at the time of mLEA, underscoring how truly frail are these patients [19].

Table 6 Summary of the literature including the largest experiences reporting on major lower extremity amputations and mortality analyses
Full size table

Despite the large number of mLEAs performed every year, risk stratification in this clinical context is still meager [5, 7]. Since no single clinical or physiologic parameter has been able to reliably predict a poor outcome after mLEA, the use of a risk-prediction score may be a more accurate method to optimize the risk stratification [27]. Taking advantage of the large number of patients contained in the American College of Surgeons National Surgical Quality Improvement Program, Easterlin et al. [5] aimed to create a risk index to predict 30-day mortality after mLEAs for PAOD. Their scoring system included eleven covariates and showed to have similar discriminatory power to several renowned risk scores used to predict surgical outcomes. The risk score developed from our cohort relies on fewer covariates, thus simplifying the process, but decreasing the accuracy shown by our model. However, our findings are worthy of several observations and conclusions. First, it constitutes our institutional audit, which is an important method of professional quality improvement based on examination of outcomes and correction of substandard practice [28]. Secondly, the covariates identified by our model have been already confirmed to be associated with mortality after mLEAs in several experiences, thus known and reliable predictors [1, 3, 5, 18, 29,30,31]. Third, the model allowed us to generate two markedly different categories of risk for early mortality, a distinction that was also associated with a significant difference in long-term survival in favor of the lower-risk group [5]. This could mean that lower-risk patients, those who are least likely to die, are the ones most likely to survive longer. In light of the fact that overall early mortality has been unchanged over the years, and the fact that patients within the lower-risk category are those who benefit from longer survival, we may have to reverse how to interpret the significance of our score model. Although our results must find future confirmation, patients who are more likely to survive the past 30-days might benefit from additional improvement of the intensity of perioperative care, which could ultimately further improve survival rates. Better perioperative care could be centered on specific risk factors, optimization of blood tests, medical therapy enhancement, and also taking advantage of delaying a non-urgent intervention [3, 6, 32].

The literature is abundant with studies confirming that older age is an important predictor for adverse outcomes [1, 3, 5, 18]. Therefore, it is not surprising that older age, specifically ≥ 80 years, has been shown to be the most important predictor of mortality following mLEAs [33]. In particular, it is interesting to note that in the index score built by Easterlin et al. [5],this same age distinction was their most powerful covariate. Despite all commendable attempts to refine risk stratification, we are undoubtedly in need of a more accurate risk prediction system. Nevertheless, non-operative management in high-risk patients should still be avoided if possible [6]. We hope our findings may lead to further initiatives on this aspect.

Amputation level remains critical to outcomes. In this study, an AKA was more frequently associated with mortality compared to a BKA, a result similar to several other investigators [3, 17, 19, 24]. Indeed, the difference between the two groups seems to be mainly determined by the risk profile of our patients. However, while prior investigators have reported that an increased perioperative mortality rate in AKA patients was associated with the presence of advanced ischemia, the most determining factor for AKA in our experience was older age [5, 19]. On one hand, it further underlines the impact of age on mLEA outcomes as BKA patients had a higher incidence of diabetes, chronic kidney disease, or hemodialysis [17, 26, 30, 31]. The association of these comorbidities with BKA is the main rationale as to why a BKA was significantly associated with postoperative complications, and more frequently required proximal revision surgery because of stump failure [17].

Considering their overall frailty status, patients needing mLEA for PAOD have been shown to be at high-risk for major adverse events. In our experience, most of the interventions have been performed under general anesthesia which can be considered a potential risk factor in these patients [34]. In our experience the type of anesthesia did not impact negatively on both major outcomes. While older studies reported a potential benefit of regional anesthesia in comparison with general anesthesia, our result data finds support in several recent experiences which reported the mode of anesthesia, did not have a significant effect on perioperative outcomes after mLEA [5, 19, 22, 29, 34]. Although it is difficult to give an unquestionable explanation, the combination of advancement in perioperative care and anesthesia management along with multidisciplinary evaluation might have been beneficial on tempering worse outcome with a specific anesthesia regimen [34].

Limitations

The present study has several limitations. It is retrospective in nature, the sample size is small and prospective evaluation of the risk score is needed to clarify the potential utility in the decision process. Nonetheless, although large databases have significant value through increased power and sample size, a single institutional analysis may offer granular detail that may not be available in the larger study. All operations were managed by members of our service only and not by different divisions or departments. Although all these features may not allow for the generalizability of our findings, our data compare well with the available literature owing to the consistency of follow-up data validated by official health documents and will allow us to further refine our processes and perform continuous quality improvement.

Conclusions

In our experience, mLEAs continue to be associated with a disturbingly high mortality rate that remains greater 14% over the years. Our predictive score discriminated two categories of patients with significantly different risks of early mortality and long-term survival. In particular, the group with a score > 4 is characterized by significantly higher early as well as long-term mortality: in these patients, major amputation likely represents a marker of advanced illness that significantly limit survival independently of the possibility to perform this intervention. Therefore, prospective validation will help to refine our risk stratification and treatment policy for these patients who would also potentially represent a logical population to engage in a proactive discussion of end-of-life.