Predicting risk factors that lead to free ﬂap failure and vascular compromise: A single unit experience with 565 free tissue transfers

Summary Background: Even though the beneﬁt of free tissue transfer is uncontested in complex reconstructive cases, vascular compromise and/or ﬂap failure remain a challenge for the surgeon and identiﬁcation of possible risk factors can aid in the preoperative planning. The aim of this study was to identify the individual risk factors leading to ﬂap failure and/or vascular compromise in free tissue transfers in a single institution over a period of 10 years and to create an index predicting these problems, as well as ﬁnding predictors of other postoperative complications. Methods: Data from all the patients undergoing free tissue transfers between 2009 and 2018 were retrospectively analyzed (demographics, comorbidities, ﬂap failure, vascular compromise, and other complications). The results from the univariate and multivariate analyses were used to create an index. Results: A predictability index with three classes (low, moderate, and high risk) was calculated for each patient, based on defect etiology and the presence of coronary heart disease, diabetes, smoking, peripheral arterial vascular disease, and arterial hypertension. A patient with moderate-risk index had 9.3 times higher chances of developing vascular compromise than those in the low-risk group, while a high-risk index had 18.6 higher odds ( p = 0.001). American Society of Anesthesiologists (ASA) classiﬁcation was found to be a predictor of complications in free tissue transfer ( p = 0.001). Conclusion: If patients at a high risk of vascular compromise could be identiﬁed preoperatively through this predictability index, patient counseling could be improved and the surgeon might adapt the reconstructive plan and choose an alternative reconstructive strategy.


Introduction
Since 1972, when McLean and Buncke reported the first free tissue transfer, a multitude of free flaps have been envisioned and described. 1 Free flaps are normally reserved for the most complex reconstructions and are often the only option left to the reconstructive surgeon. 2 Therefore, a better understanding of the risk factors associated with flap failure and return to theater with vascular compromise requiring anastomotic revision could potentially aid the surgeons to tailor their decision-making process and provide patients with improved counseling before consenting to such complex procedures. It could also potentially lead to a different surgical strategy taking into consideration individual patient ś surgical and medical history.
Even though great effort has been made in refining the microsurgical techniques and reducing the operatordependent technical errors, the postoperative complications are still high. 3 , 4 Nevertheless, patients' characteristics also seem to play an important role in the outcomes of free flaps. [5][6][7] Usually, reports are based on specific anatomical areas, for example, head and neck, 8 , 9 lower extremity 10 , 11 or breast reconstruction. 12 Therefore, it is not clear whether these previously identified risk factors for different regions have the same effect across the entire microsurgical reconstructive spectrum.
The primary aim of this study is to identify the individual risk factors leading to flap failure and re-exploration of the microsurgical anastomosis in free tissue transfers in a single institution over a period of 10 years and to create an index predicting these problems. Secondly, the study also examines predictors of other postoperative complications.

Patients and methods
All patients undergoing free tissue transfer between 2009 and 2018 in the Department of Plastic and Reconstructive surgery, University Hospital in Bern, Switzerland, were considered for this study. The following patient-related data were retrospectively collected for each patient: age at time of surgery, gender, American Society of Anesthesiologists (ASA) classification, alcohol and tobacco consumption, obesity, medical comorbidities (high blood pressure, diabetes mellitus, heart failure, coronary artery disease, and others), defect etiology, defect location, and flap used. Each free flap performed was considered as a single encounter; thus, patients who received multiple free flaps at once were considered as multiple encounters, as well as free flaps performed on the same patient but at different time points.
Flap failures, as well as vascular compromise necessitating return to theater, and all complications were recorded. In order to create an index predicting flap failure and/or vascular compromise, we analyzed surgical etiology, as well as comorbidities using the Mann-Whitney U test for the continuous variables, while Pearson's χ 2 test and Fischer's exact test were conducted for the categorical variables. Depending on the results, the presence of various conditions was scored in order to create an index with low, moderate and high risk for vascular compromise, return to theater, and flap failure.
The postoperative surgical complications (infection, partial flap necrosis, postoperative bleeding, hematoma, seroma, dehiscence, and others) were divided into minor and major complications as follows: minor complications were considered conditions that were treated conservatively, without returning to the operating theater, while major complications were the ones that required surgical reintervention. In order to analyze the complications, factors with a significance of p < 0.2 in the univariate analysis were included in the multivariate binominal logistic regression model to identify independent risk factors for postoperative complications. Statistical analysis was performed using SPSS 23.0 (SPSS Inc., Chicago, USA).
The study was conducted according to the Declaration of Helsinki principles and was approved by the local Research Ethics Committees (ID 2018-00312).

Results
From 2009 to 2018, 580 free flap procedures were identified, with 15 of them having missing data. Therefore, only 565 free flaps were included in this study. Patients' demographic and clinical characteristics are outlined in Table 1 . The distribution of defect etiology, defect location and type of flap used for reconstruction are illustrated in Figures 1-3 , respectively.

Flap failures and return to theater
There were 35 patients with vascular compromise, 14 arterial and 21 venous. In the arterial group, 10 flaps were taken back on the first postoperative day, but only 1 was saved. Both of the flaps with arterial occlusion on the second postoperative day, one flap on the third postoperative day and one flap on the forth postoperative day were lost despite revision. Therefore, we recorded a 7.1 % salvage rate among the flaps with arterial occlusion (1 out of 14). Vein occlusion occurred in 21 flaps and had a salvage rate of 85.71 % (18 out of 21). Seventeen flaps were revised on the first postoperative day, with success in 15 cases, while the other four were taken back to theater on the second postoperative day, with three of them being saved. When looking exactly at the etiology of vascular compromise that preceded flap failure, the 13 cases due to arterial occlusion proved to be statistically significant ( < 0.001) when compared to the venous occlusion. In the patients with arterial and venous occlusion, there was no statistically significant difference in terms of comorbidities, gender, or age.

Figure 1
Defect etiology: 1. Elective: free tissue transfer for various cancers that were reconstructed in a delayed fashion more than three months after the initial resection, chronic osteomyelitis with plate infection or other posttraumatic complications ensuing more than 3 months after the initial surgical intervention and osteoradionecrosis; 2. Cancer: patients with various malignancies undergoing free tissue transfer immediately or at a maximum of 3 months after tumor resection; 3. Trauma: patients with soft-tissue trauma and fractures, including infections, undergoing reconstructive surgery immediately or at a maximum of 3 months after the incident. Table 2 summarizes the characteristics of the patients with and without vascular compromise. In order to create an index predicting flap failure and/or vascular compromise, each of the following comorbidities were assigned a score of 1: smoking, the presence of arterial hypertension, diabetes mellitus, coronary heart disease, and peripheral arterial vascular disease. We analyzed the different etiological groups and found a trend in flap failure and/or vascular compromise between the various etiologies ( p < 0.2): in the elective group, there were 3.8% cases with vascular compromise, while the cancer group recorded 6.9% and the trauma group had a vascular compromise rate of 8.9%. A statistical significance was reached when the elective group was compared to the trauma group ( p = 0.038). Based on this analysis, we assigned the following predictability score to each etiological group: elective -1, cancer -2 and trauma -3. This allowed us to create a predictability index ranging from 1 to 8 points calculated for each patient. In order to predict the risk of vascular compromise, we subdivided the predictability index in three groups: low risk (index 1 and 2), moderate risk (index 3, 4, and 5), and high risk (index 6, 7, and 8). Table 3 shows the vascular compromise rate in each group. In the binominal logistic regression analysis, increasing index group was associated with an increased likelihood of exhibiting vascular compromise ( p = 0.001): a patient with a moderate-risk index had 9.3 times higher chances of developing vascular compromise than those in the low-risk group, while a high-risk index had 18.6 higher odds. Therefore, our predictability index appeared to be an accurate predictor of flap failure and/or vascular compromise. However, it is not able to predict flap salvage. These findings were supported to a certain extent by the ASA score, but only when comparing ASA 4 to ASA 1, since the other scores

Postoperative complications
An overview of the postoperative complications is presented in Table 4 . There were 183 flaps (32.4 %) with other postoperative complications, with some patients experiencing only minor or major complications, while others were treated both conservatively and surgically for the postoperative complications. In total, there were 42 minor complications (7.4 %) and 152 major complications (26.9 %). The results of the univariate and multivariate analysis are summarized in Tables 5 and 6 , respectively. The ASA score, along with the presence of diabetes and cerebrovascular accident, showed statistical significance ( p < 0.05). Since the ASA score was overlapping with the various comorbidities, we decided to use the ASA score as the sole comorbidity representative in the logistic regression, as well as the other variables with p < 0.2. Of all the predictors, only two were statistically significant: the ASA score and flap type. Patients with ASA 2 had 2.2 times higher odds to develop postoperative complications, while ASA 3 increased the chances by 2.6. Patients with ASA 4 had 11.8 times increased chances of postoperative complications in comparison to patients with ASA 1. Moreover, the flap type also had an influence on postoperative complications: patients undergoing free osteo-septo-cutaneous fibula flap recorded a 2.33 times higher chance of developing postoperative complications.

Flap failures and return to theater
The advances achieved in microsurgery have given surgeons the armamentarium to deal with some of the most complex defects. Free tissue transfer has achieved widespread popularity and can be considered a safe and reliable method, yielding success rates between 92.3% and 98%. 9 , 10 , 13-15 Even though the benefit of free flap reconstruction is uncontested in successful cases, vascular compromise and/or flap failure still remain a challenge for the surgeon, and identification of possible risk factors can aid in the preoperative planning and counseling of the patient. Our index allows the reconstructive surgeon to stratify the risk for vascular compromise and in certain situations, it may aid in adjusting the reconstructive plan. Even though our flap success rate was well within the range of acceptability of reported series (97.2%), we aimed to identify the factors playing a significant role in flap failure and/or vascular compromise in our hands and if possible, produce an index predicting these problems in practice and hopefully then be able to adjust or modify our reconstructive plan in the future.
Most series in the literature looking at flap failure and risk factors is restricted to a specific practice, for example, head and neck or breast reconstruction, while our series deals with the entire spectrum of reconstructive surgery. On the one hand, studies 16 , 17 looking at head and neck cases found that smoking, diabetes, heart disease and the presence of comorbidities were not risk factors for flap thrombosis in their univariate analysis. The lack of these variables as risk factors was also reported in a breast reconstruction study, 12 although smoking was near the statistical significance limit. On the other hand, in our series covering the entire reconstructive field, the presence of arterial hypertension, diabetes, coronary heart disease, peripheral arterial occlusive disease and smoking, in combination with the etiology of the surgery constructed as an index, was highly associated with flap failure and/or vascular compromise. The vessel architecture and tissue perfusion changes induced by diabetes and therefore the increased risk for flap failure were reported by various authors. 18 , 19 Moreover, the peripheral arterial occlusive disease affecting mainly    the donor vessels also seems to play a significant role in flap failure. 7 , 19 The endothelial dysfunction caused by smoking might explain its relation with increased vascular compromise as a contributing factor to our predictability index, but we could not reach the same statistical significance when differentiating between arterial and venous occlusion, as O'Neill et al. 12 did in their study.
Even though the ASA score was used to assess the perioperative risk of patients undergoing surgery, the extreme ASA 4 also seemed to be related with the odds of developing vascular compromise, therefore acting as a different marker for validating our index. Mücke et al. 17 also stated the presence of a significant inverse correlation between the ASA score and free flap survival, but its simplicity was considered a major drawback and does not include etiological factors like the index.
The low flap salvage rate in the patients with arterial thrombosis was also observed in other studies. Nakatsuka et al. 20 recorded 85% failure rate in the arterial thrombosis group and only 40% failure rate in the venous thrombosis group, and Chiu et al. 21 obtained statistically significant val-ues when looking at flap salvage among arterial and venous thrombosis, with a venous thrombosis salvage rate of 68.8%. This can be explained by the easier detection of venous congestion and also by the compensatory oozing through the flap margins and therefore delay of the irreversible damage on the flap. In their head and neck series, they report a thrombosis rate of 4%, while their lower extremity rate was around 19.7%. In our series, 6.2% of the patients recorded vascular compromise, but we could not find any statistically significant difference when looking at defect location. Even though the increased rate of atherosclerosis and vascular damage found among patients with advanced age might suggest an increased thrombosis rate, our study, as well as other authors, does not support this belief, 17 , 22 making free tissue transfer in elderly a safe and reliable procedure. Arterial thrombosis usually occurs on the first postoperative day. 23 Our results support this finding and also reinforce the fact that arterial thrombosis has a lower salvage rate, with only one flap saved out of the 14 with arterial thrombosis. This difference between arterial and venous occlusion has also been postulated by others, 16 , 24 with the conclusion that a thrombus will damage the arterial intima, while the vein seems to be more resilient. Overall, our flap salvage rate   30 -9 . 9 -Wink et al. 29 18 13.1 29.5 Handschel et al. 27 -2 3 -Cornejo et al. 26 --22.35 Pohlenz et al. 28 35.7 22 -Chang et al. 25 --33.3 Wettstein et al. 4 28 -4 0 among the flaps with vascular compromise (54.3%) is close to the one of 59.7% reported by Chiu et al.

ARTICLE IN PRESS
In the progression of the index groups, the stepwise increase in the proportion of vascular compromise could be useful in the daily decision-making process when dealing with patients undergoing such complex procedures. The 18.6 times increased incidence of a high-risk patient when compared to a low-risk one to develop vascular compromise and/or flap failure should prompt the surgeon to reassess the indication and may be adopt an alternative surgical strategy.

Postoperative complications
Our minor (7.4%), major (26.9%) and overall surgical complications (32.4 %) situated themselves among the values reported in the literature, as shown in Table 7 . 3 , 4 , 25-30 It is of course difficult to adequately assess the scope of postoperative complications ranging from donor site complications with a minimal impact on reconstructive outcome to recipient site complication leading in the worst case scenario to flap or reconstructive failure. Nonetheless, the presence of a postoperative complication leading to surgical reintervention definitely has a decisive impact on extremely relevant issues like treatment-associated morbidity and cost.
Even though many studies have tried to identify risk factors involved in postoperative complications, [31][32][33] there are many contradictory results, and therefore, relevant risk factors are yet to be definitively identified. The only statistically significant variables in the univariate analysis were the ASA score, the presence of diabetes mellitus and history of cerebrovascular accident. Some studies have also reported increased complications in patients with diabetes, 33 , 34 while others reported no correlation between the disease and the outcome. 14 , 32 The presence of at least one comorbidity, 8 , 34 as well as the preoperative medical condition of the patient assessed through the Kaplan-Feinstein index, 35 ASA status, 8 or the Charlson comorbidity index, 36 was found to influence postoperative complications.
In contrast to other studies, where age, 8 smoking, 5 obesity 5 and gender 37 were reported to increase postoperative complications, our study could not identify any statistically significant differences regarding these variables between patients with and without complications. Our results coincide with some of the latest studies, where complications were not influenced by gender, 32 defect location, 15 obesity, 38 smoking 11 , 14 or age. 11 While isolated variables are unlikely to be reliable predictors for all forms of morbidity and since most of the results in the literature arise from statistical tests that do not take into account the confounding factors, 27 , 33 we conducted the multivariate analysis. The increased risk of suffering postoperative complications in patients undergoing free fibula transfer in comparison to the ALT flap (Odds Ratio = 2.334, [95% CI] 1.3-4.2) was also reported in other studies, where reintervention, fistula, hematoma and partial necrosis rates were statistically significantly higher in bone-containing free flaps than in pure soft-tissue transfers. 17 , 30 Moreover, O'Brien et al. 39 and Suh et al. 40 found this type of flap to be associated with higher rates of flap failure.

Limitations
The retrospective design that can account for observer bias is a limitation of our study. Since we envisioned this predictability index, we started collecting prospectively data on free tissue transfer, but this study is still ongoing. We expect the results to further validate our index. Moreover, operative time was not considered because it included the overall operation time, including other procedures such as tumor ablation, osteosynthesis, or intraoperative radiotherapy.

Conclusion
Microvascular free tissue transfer for complex reconstructions has proven to be reliable and safe with a low flap failure rate, although the associated complication risk is high. The patients with postoperative vascular compromise could benefit from an index that estimates this risk. If patients at a high risk of arterial or venous occlusion could be identified preoperatively, additional safety measures could be taken or the surgeon may select an alternative reconstructive strategy. This predictability index could be used in the preoperative setting to improve patient counseling, as well as treatment algorithms.

Declaration of Competing Interest
None declared.

Funding
None.

Ethical approval
The study was conducted according to the Declaration of Helsinki principles and approved by the local Research Ethics Committees.