Predictive Parameters for Clinical Outcome in Patients with Critical Limb Ischemia Who Underwent Percutaneous Transluminal Angioplasty (PTA): A Systematic Review

Purpose To identify possible risk factors in predicting clinical outcome in critical limb ischemia (CLI) patients undergoing percutaneous transluminal angioplasty (PTA). Materials and Methods PubMed and EMBASE were searched for studies analyzing CLI and clinical outcome after PTA from January 2006 to April 2017. Outcome measures were ulcer healing, amputation free survival (AFS)/limb salvage and overall survival. Data on predictive factors for ulcer healing, AFS/limb salvage and survival were extracted. Results Ten articles with a total of 2448 patients were included, all cohorts and based on prospective-designed databases. For ulcers, it seems that complete healing can be achieved in most of the patients within 1 year. No significant predictive factors were found. AFS/limb salvage: AFS rates for 1, 2 and 3 years ranged from 49.5 to 75.2%, 37 to 58% and 22 to 59%, respectively. Limb salvage rates for 1, 2 and 3 years ranged from 71 to 95%, 54 to 93.3% and 32 to 92.7%, respectively. All studies had different univariate and multivariate outcomes for predictive factors; however, age and diabetes were significant predictors in at least three studies. Survival: Survival rates for 1, 2 and 3 years ranged from 65.4 to 91.5%, 45.7 to 76% and 37.3 to 83.1%, respectively. Different predictive factors were found; however, age was found in 2 out of 5 studies reporting on predictive factors. Conclusions In several studies two factors, age and diabetes, were found as predictive factors for AFS/limb salvage and survival in patients with CLI undergoing PTA. Therefore, we believe that these factors should be taken into account in future research. Level of Evidence Level 2a.


Introduction
Critical limb ischemia (CLI) due to peripheral arterial disease is a condition in which the lower extremity is threatened and is defined by ischemic rest pain, with or without ischemic tissue loss [1]. CLI has a great impact on healthcare and associated healthcare budget [2]. A number of risk factors are known to be associated with the development of CLI, which are diabetes mellitus, smoking, increased age, lipid abnormalities and low ankle-brachial pressure index [2].
Of the CLI patients, 10-40% will lose their leg within 6 months and the 1-year mortality rate is 25% in CLI patients who are not able to be revascularized [2][3][4].
Percutaneous transluminal angioplasty (PTA), with or without stenting, is an alternative approach to surgical bypass as a revascularization method in patients with CLI [5,6]. Compared to surgery, it involves advantages such as minimal access trauma and shorter hospital stay. Therefore, PTA is more suited and often suggested as first-line therapy for highrisk CLI patients with a lower life expectancy [7][8][9][10].
However, interpreting these clinical outcomes in this patient group is difficult, because of its heterogeneity in the risk factors such as comorbid diabetes, difference in age, renal failure or lifestyle factors such as smoking and obesity. We often see a discrepancy between a good revascularization result of the PTA, identified on digital subtraction angiography (DSA) and an unexpected poor clinical outcome with early amputation [9,18,19]. For future analysis of study results concerning endovascular treatment in CLI patients, it is important to identity which risk factors are associated with poor outcome.
Therefore, the aim of this systematic review was to identify risk factors in predicting poor clinical outcome in patients with CLI undergoing PTA with or without stenting. Drug eluting technologies were not included in the review to try to maintain homogeneity in the study population.

Materials and Methods
This review was conducted according to the preferred reporting items for systematic review and meta-analysis (PRISMA) guidelines [20]. The review protocol was not published or registered in advance.

Search Strategy
An electronic search was performed in the databases PubMed and EMBASE for studies analyzing CLI and clinical outcome after percutaneous revascularization. The search period was from January 2006 to April 2017. Search terms used for PubMed and EMBASE are listed below.
PubMed ''Critical limb ischemia OR critical limb ischemia AND (angioplasty OR endovascular revascularization OR percutaneous intentional extraluminal revascularization OR subintimal OR endovascular therapy) AND (major amputation OR amputation free survival OR death OR ulcer healing OR wound healing OR mortality OR survival) AND Humans''.
Embase: (Critical limb ischamia OR critical limb ischemia) AND (percutaneous transluminal angioplasty balloon OR percutaneous transluminal angioplasty OR angioplasty OR stent OR revascularization) AND mortality OR (amputation OR major amputation OR leg amputation) OR (ulcer healing OR wound healing) OR (survival).

Study Selection
Step 1 All retrieved articles were checked on title and abstract by one observer (X2). Duplicates, reviews, guidelines, comments, letters to the editor, conferences, case reports, study protocol and articles not containing CLI were excluded.
Step 2 All remaining articles were also checked on abstract by the same observer (X2). When studies contained less than fifty patients, patients did not receive PTA, the study was retrospective (we considered prospective database as prospective study) or the follow-up period was less than 1 year, these studies were excluded. To avoid exclusion of relevant articles, ambiguous articles were retrieved as full text and treated as potentially eligible articles. The observer double-checked step 2 and was not blinded to author and journal names.

Inclusion of Relevant Articles
Three observers (X1, X2 and X3) independently checked all remaining articles for inclusion and exclusion criteria. Two observers (X1 and X2) each checked half of the relevant articles, and the findings were discussed with observer 3 (X3) who has experience on data extraction of 25 meta-analyses.
The inclusion criteria were as follows: (1) prospective study or prospective database (we considered prospective database as prospective study, hospital billing and other registries as retrospective); (2) patients with CLI as defined by Fontaine class III-IV or Rutherford class IV-VI (rest pain, non-healing ulcer or gangrene); (3) patients underwent (regular) PTA (no drug eluting stents); (4) [50 patients with CLI undergoing PTA; (5) data on outcome were available for at least 1 year of follow-up (outcomes were healing, AFS (major of minor) and overall survival); (6) separate data on CLI and PTA were available (in studies that included a variety of patients or treatments, for example data on CLI patient who underwent PTA or bypass surgery); and (7) finally, data on predictive factors were reported. Exclusion criterion was duplicate data.

Data Extraction
Two reviewers (X1, radiologist with experience in extracting data of two reviews and X2, medical student) used a standardized form to extract data independently on study design characteristics, patient selection, baseline patient characteristics, procedure description, angiographic outcomes and complications, follow-up and dropout patients, clinical outcomes and predictive factors. Again, each observer extracted data of half of the articles and were double-checked by the third reviewer with experience on data extraction of 25 meta-analyses.
Study design characteristics The following data on study design characteristics were extracted: (1) study type (cohort, part of RCT or other); (2) study design (single center or multicenter and prospective study or prospective database retrospectively analyzed); (3) setting initiation institute (academic, tertiary or other); (4) department initiation by first author (radiology, surgery or other); (5) period of recruitment; (6) institutional review board approval (approved and informed consent obtained/waived, not approved or unclear); and (7) funding or a potential role of funders in the study (conflict of interest).
Patient selection The following data on patient selection were retrieved: (1) consecutive sample of patients enrolled (yes or no); (2) inclusion and exclusion criteria defined; and (3) spectrum of patients representative for CLI patients normally receiving PTA.
Angiographic outcomes and complications data were extracted on how articles defined (1) technical success; (2) partial success/failure; (3) complete technical failure; (4) major complications; and (5) minor complications and how many successes, failures and complications occurred.
Follow-up and dropout patients The following data were extracted regarding follow-up: (1) a summary of follow-up time and scheme; (2) if all patients underwent the same follow-up (yes or no) and (3) were dropout patients adequately reported (yes or no, with or without reasons for dropout or unclear).
Clinical outcomes and predictive factors Data were extracted on the three previously defined outcome variables: (1) ulcer healing; (2) AFS (major of minor) or limb salvage and (3) overall survival at baseline and at least 1-year follow-up with a maximum of 5-year follow-up. Data on predictive factors either in terms of regression analysis (univariate or multivariate) were extracted.

Data Analysis
All data at baseline were presented as number plus percentage, with the exception of age, which is presented as a mean. Because standard deviation was not available in all datasets, result on baseline could not be pooled.
Data on ulcer healing, AFS and overall survival at baseline and at least 1-year follow-up were recorded. Data on predictive factors for ulcer healing, AFS (also limb salvage) and survival were extracted as reported in papers. As anticipated, the number of studies was limited. The data were heterogeneously presented so even meta-analysis with random effect approach would not be suitable for pooling predictive values. All data are therefore presented per study.

Search, Selection and Inclusion of Relevant Articles
The search yielded 1635 studies: 734 from Pubmed and 901 from EMBASE (see Appendix 1).

Study Design Characteristics
Of the ten articles included, all were cohort studies; most studies were performed based on prospective-designed databases and were single center. In all studies, there was no role of funders (see Table 1).

Patient Selection
The patient selection was consecutive in most of the studies. In all studies, patients were included with CLI; however, the spectrum of patients was equivocal, as in one study only patients [ 80 years were included [21], only diabetic patients [22], only hemodialysis patients [27] or patients with Rutherford V and VI [28] (see Table 2).

Baseline Patient Characteristics
In total, 2448 patients were included who were CLI patients and underwent PTA with or without bare metal stent placement. Mean ages ranged from 50 to 85.9 years. Male-to-female ratio was 816:534 in the seven studies mentioning this ratio [21][22][23][27][28][29][30]. In addition, a broad range of risk factors was present: smoking rate from 6.9 to 58.3%, diabetes from 49.1 to 100%, hypertension from 51.6 to 98%, dyslipidaemia from 21.1 to 65% and renal disease up to 100%. Other risk factors such as coronary artery disease, cerebrovascular disease and stroke were also present in the majority of patients (see Table 3).
ABI was mentioned in only small number of studies, other measurements such as toe pressure and ankle pressure were only mentioned in the study of Strom et al. (toe pressure mean 30 mmHg [range 0-60 mmHg] and ankle pressure mean 50 mmHg [range 0-60 mmHg]) [30]. The TcPO2 was not mentioned in any of the studies. The disease severity in terms of Fontaine classification or Rutherford category was described heterogeneously (see Table 4).

Procedure Description, Outcomes and Complications
In most studies, it was not clear who performed the procedure. Moreover, the experience of the operator was not defined in any of the studies. In none of the studies, the procedure was described in sufficient detail to replicate. The angiographic outcome in terms of technical success was defined well, and complications were reported in detail.
All data on procedure description and outcomes are given in detail in Table 5.

Follow-Up and Dropout Patients
The follow-up was not homogeneous, but in general 1 month, 3-, 6-and 12-month follow-up was done. Patients did not undergo the same follow-up in seven studies, while in three studies patients did undergo the same follow-up. Dropout rates are poorly reported. Only one study [24] accurately reported dropouts, with missing baseline information as most frequent reason for dropout. Follow-up ranged from less than 1 month up to 109 months. All details are given in Table 6.

Ulcer Healing
In three studies [22,25,28], data on ulcer healing were given. It seems that complete healing can be achieved in most of the patients within 1 year [25,28]. Details are given in Table 7.

Ankle-brachial index (ABI)
Mean ± SD; median ? range Fontaine classification, Rutherford category or other classification Anticoagulation/antiplatelet medication at baseline (N and percentage) [

Predictive Factors
When data were available on predictive values, these data were also extracted (see Table 10). However, these data were presented heterogeneously. We extracted all data as given in the studies. In general for univariate analysis, data were given either (1) at a time point (e.g., AFS at 2 years) by Fisher exact test or Chi-square test (2 9 2 tables) or Student's t test (continuous normally distributed data) or Mann-Whitney tests (continuous not normally distributed data) or by association tests (continuous data) or (2) as time dependent by Kaplan-Meier analysis (with log rank test, for binary data) or Cox regression analysis (for multinomial or continuous data). Finally, multivariate analysis in either stepwise multiple regression analysis was used (at one time point) or Cox proportional regression analysis (for time dependent data) was performed.

Predictive Factors in Ulcer Healing
Number of studies reporting predictive factors is limited [22,28] with different predictive factors (see Table 10).

Predictive Factors in Survival Analysis
Also for the survival analysis, different predictive factors were found; however, age was found in 2 [22,25] out of 5 studies reporting on predictive factors (see Table 10). Based on these findings, age and diabetes should be at least taken into account when searching for predictive factors. One patient (2%) presented with acute abdomen and respiratory distress suspected of acute mesenteric ischemia Two patients died within 30 days (perioperative mortality; 3%) due to toxicity awaiting amputation (n = 1) and cerebral hemorrhage occurring after a minor amputation (n = 1)

Discussion Summary
In this review, we summarized the findings on predictive factors for wound healing, AFS and survival in CLI patients who underwent a PTA. As stated, the data were heterogeneously reported and presented. In addition, none of the studies found the same predictive factors. However, in several studies age and diabetes were found as predictive factors for AFS or limb salvage and survival. Several univariate studies showed age and diabetes as predictors [12,[31][32][33].

Compared with Other Studies
To our knowledge, no such systematic review has been published. There is a review [34] in which the authors    [23] Univariate analysis in Kaplan-Meier and log rank or associations (p \ 0.05) Cox proportional regression analysis summarized risk stratification models for CLI with a summary of the respective strengths and limitations of each. These models were developed from prospective cohorts to identify and quantify variables that can subsequently predict outcome in individual patients. In the prospective cohort, treatment options generally were compared (e.g., open and endovascular therapies) and new therapeutics were evaluated. The outcomes were not specific for defining risk models in patients with CLI patients undergoing PTA.

Strength of this Review
The major strength of our study is that we focussed on patients with CLI who underwent PTA to identify possible predictive factors for clinically relevant outcomes. We have done this to create a homogeneous and clinically relevant population, in order to draw conclusions. We included studies which aimed to study predictive values of all types of risk factors. In addition, we only selected prospective studies or studies that used a prospective database, to have a predefined design without missing a lot of data. It is known that missing data are much more common in retrospective studies, in which routinely collected data are subsequently used for a different purpose [35].

Limitations of this Review
Although all studies were performed prospectively or a prospective database was present with a spectrum of patients which are represented, the data were presented too heterogeneously. Even the AFS or survival analysis was not reported homogeneously. The presented data on the predictive values varied even more, making general conclusions difficult.

Conclusion and Recommendations
It is not clear which risk factors should be taken into account. However, in several studies two factors, age and diabetes, were found as predictive factors for AFS or limb None of the factors tested was significant a In this study, although data (ulcer healing, AFS or survival) were reported separately for PTA, data of regression analysis was presented combined both groups: PTA and bypass surgery b In this study, although data (ulcer healing, AFS or survival) were reported separately for PTA with Bare Metal Stent (BMS), data of regression analysis was presented for both PTA with BMS and PTA with drug eluting stent. The cox regression showed no difference between both groups salvage and survival in patients with CLI undergoing PTA. Therefore, we believe that these factors should be taken into account in the future when searching for predictive factors and when analyzing study data on endovascular treatments for CLI. More research on this topic is needed. A trial with registry of all risk factors and the outcomes up to 12 months would be very important. Future research is needed to simplify and improve the accuracy and generalizability of risk stratification in CLI.