Prognostic Factors for Free Flap Failure in Head and Neck Reconstruction

ABSTRACT

Background

The failure rate of free flaps varies from 0.8% to 10.6% in the literature in head and neck reconstruction. The primary objective was to identify prognostic factors for free flap failure.

Methods

We prospectively included all consecutive free flaps performed between August 2021 and January 2024, and used a multivariate Cox proportional hazard model.

Results

We included 307 free flaps, performed on 274 patients. Age, cardiovascular risk, radiotherapy history, type of flap, type of arterial anastomosis, and ischemia duration were not statistically linked to the risk of flap failure. In multivariate analysis, a venous anastomosis to the anterior jugular vein or to the superior thyroid vein were associated with an increased risk of flap failure, such as per‐ or postoperative revision of the anastomoses.

Conclusions

The choice of venous anastomosis, appear to have a greater influence on the success or failure of a microvascularized transfer than patient characteristics.

1. Introduction

The use of microvascularized free tissue transfers or free flaps is now the standard in head and neck reconstruction for complex defects [1]. Depending on the structures to be reconstructed and the volume to be supplied, the flaps used may be osteocutaneous, such as the fibula free flap (FFF) or the iliac crest flap (deep circumflex iliac artery, DCIA), muscular or osteomuscular, such as the scapular system free flaps (SFFs), or fasciocutaneous, such as the radial forearm free flap (RFFF) or the lateral arm flap (LAF) [2].

The failure rate of free flaps, corresponding to tissue necrosis, most often due to arterial or venous thrombosis of the anastomoses, varies from 0.8% to 10.6% in the literature [3, 4].

Improvements in surgical techniques and perioperative management have led to a reduction in the failure rate and in patient morbidity and mortality [2]. However, failure leads to an increase in the cost and length of hospitalization and is always responsible for major functional, esthetic, and psychosocial consequences. It is, therefore, essential to understand the factors leading to success with free flaps, in order to further reduce this outcome.

The primary objective of this study was to identify prognostic factors for free flap failure through a prospective, standardized collection of pre‐, per‐, and postoperative patient data. The secondary objective was to identify factors influencing perioperative mortality in these patients.

2. Material and Methods

2.1. Data Collection

We prospectively included all consecutive free flaps performed in our maxillofacial surgery department between August 2018 and January 2024. The study was validated by our local ethics committee (CER‐2011‐022) in accordance with the Declaration of Helsinki. There were no exclusion criteria.

In a standardized way, we collected:

1. Preoperative and demographic data: Age, gender, cardiovascular SCORE risk evaluation [5], thrombotic risk factors (history of phlebitis or pulmonary embolism, thrombophilia, polycythemia, cirrhosis), active or past alcohol consumption, active smoking, body mass index (BMI), history of cervicofacial radiotherapy, indication, location of reconstruction, atherosclerosis/stenosis of lower limb arteries on CT‐scan in case of fibula flap;

2. Peroperative data: Type of free flap, presence of a skin paddle, cervical lymph node dissection, peroperative flap volume, choice of donor and recipient vessels, size of donor artery, type of venous anastomosis (sutures or coupler device) and, if applicable, size of coupler used, duration of arterial anastomosis, duration of venous anastomosis, flap ischemia duration, peroperative revision of anastomoses, operating time; volume was measured by dipping the flap into a graduated cup of saline marked every 25 mL, and measuring the surface level. The diameter of the donor artery was measured using a microsurgical meter.

3. Postoperative data: Return to the theater for revision of anastomoses, cervical hematoma, infection, partial or total necrosis, esthetic or functional remodeling of the flap, mortality during the hospital stay.

2.2. Data Analysis

A multivariate Cox proportional hazard model, including factors with a p‐value < 0.075 in univariate analysis was built to select clinical parameters associated with free flap failure (partial or total necrosis), and with perioperative mortality. An interaction parameter has been added between the selected variables for each model. The hypotheses of proportionality and log‐linearity have been verified. The model coefficients were compared to 0 using Student tests. Explanatory variables improving the log‐likelihood of the model and maintaining convergence were selected. Assumptions of normality and homoscedasticity of errors were tested. Survival distributions of the flaps according to factors were plotting using the Kaplan–Meier method.

The statistical analyses were performed on R 4.2.2 using the nlme [6], survival [7], and ggplot [8] packages. All statistical tests were two‐sided, and p‐values ≤ 0.05 were considered to be statistically significant.

3. Results

3.1. Data Description

We included 307 free flaps, performed on 274 patients, ranging in age from 19 to 92 years (mean: 61.8 ± 15.4 years). Forty‐two percent were female. The mean BMI was 24.1 ± 4.9 kg/m². Cardiovascular SCORE risk was low for 37% of patients, moderate for 32%, high for 14% and very high for 7%; venous thrombotic risk was identified for 7% of patients. With regard to toxic consumption, 69% were active smokers and 31% chronic drinkers. Among patients who underwent angio‐CT of the lower limbs for reconstruction with a FFF, 28% had normal arteries, 47% had proximal atherosclerosis or stenosis (aorta, iliac vessels, femoral or popliteal vessels), 26% had distal atherosclerosis or stenosis (anterior tibial, posterior tibial or fibular vessels). 33% of patients had a history of cervicofacial radiotherapy.

The most frequent surgical indications were malignant tumors with primary reconstruction (n = 181, 59%), resection of osteoradionecrosis (n = 50, 16%), sequelae of tumor excision or radiotherapy (n = 39, 13%), benign tumors (n = 21, 7%), and trauma (n = 8, 3%). Other causes are detailed in Table 1. The most common anatomical structures reconstructed were the mandible (38%), the maxilla (21%), the tongue (12%), and the floor of the mouth (9%).

TABLE 1.

Data description.

N	307
Patients	274
Age (mean ± SD)	61.8 ± 15.4
Females	129/307 (42%)
BMI (mean ± SD)	24.1 ± 4.9
Cardiovascular SCORE risk
Low risk	113/302 (37%)
Moderate risk	96/302 (32%)
High risk	42/302 (14%)
Very high risk	51/302 (17%)
Venous thrombotic risk	23/307 (7%)
Angio‐CT of the lower limbs
Normal	16/58 (28%)
Proximal atherosclerosis/stenosis	27/58 (47%)
Distal atherosclerosis/stenosis	15/58 (26%)
Toxics
Active smoker	164/237 (69%)
Chronic drinker	73/237 (31%)
Indication
Malignant tumors	181/307 (59%)
Osteoradionecrosis	50/307 (16%)
Malignant tumor sequellas	39/307 (13%)
Benign tumors	21/307 (7%)
Other indications a	16/307 (5%)
Reconstructed structure
Mandible	116/307 (38%)
Maxilla	64/307 (21%)
Tongue	37/307 (12%)
Floor of mouth	29/307 (9%)
Maxillo mandibular commissure	23/307 (7%)
Cheek mucosa	12/307 (4%)
Other structures a	24/307 (8%)
Cervicofacial radiotherapy history	102/307 (33%)
Flap
RFFF	104/307 (34%)
FFF	99/307 (32%)
SFF a	70/307 (23%)
LAF	18/307 (6%)
Periosteal lateral brachial flap	8/307 (3%)
DCIA	3/307 (1%)
Other flaps a	4/307 (1%)
Skin paddle	271/307 (88%)
Lymph nodes dissection	151/307 (49%)
Flap perop. volume (mean ± SD, cc)	102 ± 79
Donnor artery
Facial	150/268 (56%)
Superior thyroid	44/268 (16%)
Lingual	36/268 (13%)
Transverse cervical	14/268 (5%)
External carotid	12/268 (4%)
Superficial temporal	7/268 (3%)
Other donnor arteries a	4/268 (2%)
Donnor artery diameter (mean ± SD, mm)	2.2 ± 0.9
Recipient vein
External jugular	68/250 (27%)
LFT	62/250 (25%)
Common facial	53/250 (21%)
Superior thyroid	15/250 (6%)
Internal jugular	13/250 (5%)
Lingual	12/250 (5%)
Transverse cervical	9/250 (4%)
Anterior jugular	8/250 (3%)
Other recipient veins a
Vein anastomosis
Coupler	223/232 (96%)
Sutures	9/232 (4%)
Type of vein anastomosis
Terminoterminal	240/250 (96%)
Terminolateral	10/250 (4%)
Coupler size (mean ± SD)	3.2 ± 0.6
Anastomosis duration (min)
Artery	29.5 ± 22.1
Vein	14.0 ± 10.8
Ischemia duration (mean ± SD, min)	80.1 ± 42.1
Operating time (mean ± SD, min)	361 ± 103
Perop. anastomosis revision	17/307 (6%)
Artery	15/307 (5%)
Vein
Postop. anastomosis revision	36/307 (12%)
Artery	10/25 (40%)
Vein	15/25 (60%)
Flap salvage after anastomosis revision	25/36 (69%)
Postoperative infection drainage	28/307 (9%)
Cervical hematoma drainage	13/307 (4%)
Flap remodeling	57/307 (19%)
Partial flap necrosis	14/307 (5%)
Total flap necrosis	24/307 (8%)
Other secondary surgeries a	16/307 (5%)
Death	42/307 (14%)
Perioperative death	8/307 (3%)
Surgery specific survival	97.3% [0.955–0.992]

Abbreviations: BMI = body mass index; CT = computed tomography; DCIA = deep circumflex iliac artery; FFF = fibula free flap; LAF = lateral arm flap; LFT = linguo‐facial trunk; perop. = peroperative; postop. = postoperative; RFFF = radial forearm free flap; SCORE = systematic coronary risk evaluation; SD = standard deviation; SFF = scapular system free flap.

^a Other indications: traumatology 8/307 (3%), vascular malformations 3/307 (1%), dysmorphic syndromes 3/307 (1%), infection 1/307 (0%), cocaine lesions 1/307 (0%). Other structures: cervical 6/307 (2%), lip 5/307 (2%), orbital region 5/307 (2%), scalp 4/307 (1%), soft palate 2/307 (1%), nose 2/307 (1%). SFF: scapular tip 38/307 (12%), Latissimus dorsi flap 21/307 (7%), scapular crest 7/307 (2%), thoracodorsal artery perforator 3/307 (1%), serratus 1/307 (0%). Other flaps: helix 2/307 (1%), gracilis 1/307 (0%), soleus 1/307 (0%). Other donor arteries: previous pedicle 3/268 (1%), middle thyroid 1/268 (1%), internal mammary 1/268 (1%). Other recipient veins: superficial temporal 3/250 (1%), retromandibular vein 3/250 (1%), middle thyroid 3/250 (1%), previous pedicle 1/250 (0%). Other secondary surgeries: margins revision 10/307 (3%), false aneurism 3/307 (1%), cervical coverage by a pectoralis major flap 2/307 (1%), tracheal resection 1/307 (0%).

The distribution of flaps performed was as follows: 104 (34%) RFFF, 99 (32%) FFF, 70 (23%) SFF (including 38 scapula tip flaps, 21 LD, 7 scapular ridges, 3 TDAP, and 1 serratus flap), 26 (9%) LAF (including 8 with brachial periosteum harvesting), 3 (3%) DCIA, 2 (1%) helix, 1 (0%) gracilis, and 1 soleus flap (0%). A skin paddle was harvested in 88% of cases. Peroperative flap volume averaged was 102 ± 79 cc.

The most frequent donor arteries were the facial artery (56%), the superior thyroid artery (16%), the lingual artery (13%), and the transverse cervical artery (5%), with a mean diameter of 2.2 ± 0.9 mm (min/max: 1/8 mm). Mean arterial anastomosis time was 29.5 ± 22.1 min. The most frequent recipient veins were the external jugular vein (27%), the lingual‐facial trunk (25%), the common facial vein (21%), the superior thyroid vein (6%), and the internal jugular vein (5%); 96% of anastomoses were performed with a coupler. The mean size of the coupler used was 3.2 ± 0.6 mm (min/max: 1/4 mm). Mean venous anastomosis time was 14.0 ± 10.8 min. Ischemia duration ranged from 17 to 389 min, with a mean of 80.1 ± 42.1 min. For 12% (n = 17) of patients, the anastomoses had to be revised before the end of the procedure. Mean operating time was 361 ± 103 min, ranging from 119 to 770 min.

Regarding postoperative follow‐up, 8% (n = 24) of flaps had to be removed for complete necrosis, and 5% (n = 14) were partially necrotic. Cervical hematoma and infection had to be drained in 4% and 9% of cases, respectively.

We had to revise anastomoses in 36 flaps (12%), with arterial thrombosis in 40% and venous thrombosis in 60%. In 69% (n = 25) of cases, the flap was saved.

Eight patients (2.7%) died during the hospital stay (survival rate 97.3% [95.5–99.2]).

3.2. Partial or Total Flap Failure

In univariate analyses, age, gender, BMI, SCORE risk, venous thrombotic risk, lower limbs angio‐CT results, toxics consumption, indication, radiotherapy history, type of flap, flap volume, the choice of donor artery and its diameter, type of vein anastomosis, anastomosis, ischemia and surgery durations, postop. infection/hematoma drainages were not statistically linked to the risk of flap failure. The variables included in the multivariate Cox model were the choice of recipient vein, and intra‐ or postoperative revision of anastomoses.

In multivariate analysis, a venous anastomosis to the anterior jugular vein or to the superior thyroid vein were associated with an increased risk of partial or total flap failure (adjusted HR = 22.19 [1.702–289.3], p = 0.018 and adjusted HR = 37.99 [2.649–545.0], p = 0.007 for the anterior jugular and superior thyroid veins, respectively). Furthermore, per‐ or postoperative revision of the anastomoses resulted in an excess risk of failure of 5.655 [1.073–29.82] (p = 0.041). These results are summarized in Tables 2 and 3 and in Figure 1.

TABLE 2.

Cox model results for partial or total flap failure using clinical factors.

	Univariate analysis
	HR	95% CI	p
Age
< 50	0.399	[0.138–1.157]	0.091
> 74	1.301	[0.525–3.224]	0.570
Gender
Female	1.231	[0.608–2.495]	0.564
BMI	1.038	[0.970–1.112]	0.278
Cardiovascular SCORE risk
Moderate	1.266	[0.578–2.771]	0.555
High	1.759	[0.677–4.571]	0.247
Very high	2.334	[0.637–8.546]	0.201
Venous thrombotic risk	0.696	[0.242–1.997]	0.500
Angio‐CT of the lower limbs
Proximal aterosclerosis/stenosis	0.152	[0.008–2.812]	0.206
Distal aterosclerosis/stenosis	0.185	[0.017–1.970]	0.162
Toxics
Chronic drinker	0.766	[0.357–1.648]	0.496
Active smoker	0.495	[0.223–1.098]	0.084
Indication (reference: others)
Malignant	0.916	[0.121–6.958]	0.933
ORN/sequella	0.809	[0.105–6.230]	0.839
Cervicofacial radiotherapy history	1.047	[0.546–2.009]	0.890

Note: Bolded p‐values indicate statistical significance.

Abbreviations: BMI = body mass index; CI = confidence interval; CT = computed tomography; HR = hazard ratio; ORN = osteoradionecrosis; SCORE = systematic coronary risk evaluation.

TABLE 3.

Cox model results for partial or total flap failure using surgical factors.

	Univariate analysis			Multivariate analysis
	HR	95% CI	p	Adjusted HR	95% CI	p
Flap (reference: RFFF)
LAF	0.362	[0.086–1.528]	0.167
FFF	1.227	[0.401–3.753]	0.720
SFF	1.189	[0.386–3.666]	0.763
Skin paddle	1.174	[0.356–3.871]	0.792
Lymph nodes dissection	1.074	[0.558–2.066]	0.831
Flap perop. volume	1.005	[0.999–1.011]	0.074
Donnor artery (reference: facial)
External carotid	2.330	[0.648–8.378]	0.195
Transverse cervical	0.313	[0.063–1.567]	0.158
Lingual	1.125	[0.313–4.043]	0.856
Superior thyroid	0.449	[0.164–1.234]	0.121
Superficial temporal	0.860	[0.112–6.617]	0.885
Old pedicle	0.406	[0.052–3.169]	0.390
Donnor artery diameter	1.773	[0.480–6.550]	0.390
Recipient vein (reference: common facial)
Internal jugular	3.153	[0.813–12.23]	0.097	1.225	[0.292–5.139]	0.782
External jugular	0.663	[0.222–1.978]	0.461	0.863	[0.137–5.417]	0.875
Anterior jugular	3.239	[0.750–13.99]	0.115	22.19	[1.702–289.3]	0.018
Transverse cervical	2.042	[0.232–17.95]	0.520	0.775	[0.083–7.220]	0.823
Superior thyroid	14.65	[1.324–162.2]	0.029	37.99	[2.649–545.0]	0.007
LFT	1.397	[0.454–4.301]	0.560	2.623	[0.541–12.71]	0.231
Vein anastomosis (reference: coupler)
Suture	2.507	[0.317–19.80]	0.383
Coupler size	1.045	[0.593–1.841]	0.880
Anastomosis duration
Artery	0.997	[0.989–1.005]	0.420
Vein	0.989	[0.964–1.015]	0.397
Ischemia	0.999	[0.995–1.004]	0.770
Surgery duration	0.999	[0.996–1.001]	0.319
Anastomosis revision
Per‐/postop.	1.975	[0.992–3.933]	0.053	5.655	[1.073–29.82]	0.041
Postop. infection drainage	0.538	[0.222–1.307]	0.171
Cervical hematoma drainage	0.927	[0.318–2.700]	0.889

Note: Bolded p‐values indicate statistical significance.

Abbreviations: CI = confidence interval; DCIA = deep circumflex iliac artery; FFF = fibula free flap; HR = hazard ratio; LAF = lateral arm flap; LFT = linguo‐facial trunk; perop. = peroperative; postop. = postoperative; RFFF = radial forearm free flap; SFF = scapular system free flap.

FIGURE 1.

Kaplan–Meier curves, to compare flap failures, for (A) the different types of flaps, (B) whether or not the anastomoses were revised surgically, and (C) the recipient vein chosen. DCIA = deep circumflex iliac artery; FFF = fibula free flap; LAF = lateral arm flap; LFT = linguo‐facial trunk; RFFF = radial forearm free flap; SFF = scapular system free flap. [Color figure can be viewed at wileyonlinelibrary.com]

3.3. Perioperative Death

Perioperative mortality was 2.7% (survival of 97.3% [95.5%–99.2%]) at 30 days. Age, gender, the flap type and a cervicofacial radiotherapy history were not statistically associated with a higher rate of perioperative death. The variables included in the multivariate Cox model were the BMI, the cardiovascular SCORE risk, venous thrombotic risk factors, and operative time. In multivariate analysis, an increase in BMI (adjusted HR = 0.764 [0.596–0.980] per point of BMI, p = 0.034) or a moderate SCORE risk (adjusted HR = 0.010 [0.002–0.053], p < 0.001) appeared to be protective factors for the perioperative mortality. Survival at 30 days was 99.1% [97.4–100] for patients at low cardiovascular risk, compared with 92.2% [85.1%–99.8%] for patients at high risk. Venous thrombotic risk factors lead to an excess risk of perioperative mortality of 101.9 [18.29–567.8] (p < 0.001). In patients with venous thromboembolic risk factors, survival fell from 98.6% [97.2%–99.9%] to 82.6% [68.5%–99.6%] at 30 days. A very high cardiovascular SCORE risk was associated with an excess risk of 6.782 [1.311–35.08] (p = 0.022). Finally, operative duration was significantly associated with increased perioperative mortality, with an adjusted HR = 1.015 [1.010–1.019] (p < 0.001) after taking into account the various confounding factors (Figure 2, Table 4). Of the eight perioperative deaths, four were caused by pneumopathy, two by pulmonary embolism, and two by massive bleeding.

FIGURE 2.

Kaplan–Meier curves, to compare perioperative mortality, by (A) different cardiovascular SCORE risks, (B) venous thrombotic risk factors, (C) BMI scores, and (D) operative time. A cut‐off time of 6 h was used for the operating time, as this was the median in our series. BMI = body mass index. [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 4.

Multivariate analysis results for perioperative mortality.

	Univariate analysis			Multivariate analysis
	HR	95% CI	p	Adjusted HR	95% CI	p
Age	1.041	[0.986–1.099]	0.145
Gender
Female	0.841	[0.201–3.519]	0.813
BMI	0.817	[0.666–1.003]	0.053	0.764	[0.596–0.980]	0.034
Cardiovascular SCORE risk
Moderate	2.295	[0.208–25.31]	0.498	0.010	[0.002–0.053]	< 0.001
High	—	—	—	—	—	—
Very high	9.096	[1.017–81.39]	0.048	6.782	[1.311–35.08]	0.022
Venous thrombotic risk	12.77	[3.194–51.07]	< 0.001	101.9	[18.29–567.8]	< 0.001
Cervicofacial radiotherapy history	0.664	[0.134–3.290]	0.616
Flap (reference: RFFF)
LAF	—	—	—
FFF	5.523	[0.645–47.28]	0.119
SFF	3.031	[0.275–33.43]	0.365
Operative time	1.009	[1.004–1.014]	< 0.001	1.015	[1.010–1.019]	< 0.001

Note: Bolded p‐values indicate statistical significance.

Abbreviations: BMI = body mass index; CI = confidence interval; FFF = fibula free flap; HR = hazard ratio; LAF = lateral arm flap; RFFF = radial forearm free flap; SCORE = systematic coronary risk evaluation; SFF = scapular system free flap.

4. Discussion

To our knowledge, this is the largest prospective cohort with standardized peroperative measurements studying factors associated with the risk of free flap failure in head and neck reconstruction. Most studies in the literature are retrospective, with many missing data.

Free flap failure rates vary from 0.8% to 10.6% in the literature [3]. However, flap failure is not consensual. Indeed, some authors consider as a failure of a free flap the need for revision of the anastomoses [1, 2, 4, 9], the partial necrosis of the flap [10, 11, 12], while others do not [13, 14, 15, 16, 17, 18, 19, 20]. Then free flap failure rates are not standardized and difficult to compare.

Rosenberg et al. [21] explained that the success of a free flap was more related to the microsurgical technique itself than to patient characteristics. Our findings are consistent with this, since factors such as age, sex, cardiovascular risk including diabetes, venous thrombotic risk or previous radiotherapy did not appear to be significantly associated with free flap failure. However, some studies are not in accordance with these findings showing such as Ishimaru et al., who found a higher failure rate in patients with diabetes, peripheral vascular disease or renal insufficiency [2, 17, 22, 23, 24, 25].

Las et al. [26] or Bollig et al. [27] showed a better reliability in RFFF while Kwok and Agarwal [28], in accordance with our study, found no significant differences between the different types of flap, in term of flap failure. In particular, there was no excess risk of failure in bone free flaps.

Furthermore, we found no significant relationship between operative, ischemia, and anastomosis times with the risk of flap failure, as some authors have [1, 2, 29]. Most studies had no access to the duration of ischemia and used operating time to represent it [4]. We demonstrated here that these different durations had no significant impact. It should be noted, however, that Serletti et al. [29] found a higher risk after 600 min of surgery, which corresponded to very few patients in our cohort. Unfortunately, we did not have any information on the tourniquet time during flap removal.

The choice of donor artery was identified as a significant risk factor of free flap failure. Indeed, Las et al. [26] which advised against anastomoses with the superficial temporal artery, which would promote vasospasm due to its small diameter. Zhang et al. [9] recommends to select, in the following order, the facial artery, the superior thyroid artery, the transverse cervical artery, and finally the external carotid artery. We found no significant relationship between the choice of donor artery and the risk of flap failure. Of particular note is the absence of any additional risk associated with the use of the superficial temporal artery, contrary to what has been suggested by Las et al. [26]. However, it is difficult to draw conclusions, as only seven anastomoses with superficial temporal vessels were performed in our series. Similarly, we found no relationship with the diameter of the donor artery.

Las et al. [26] recommends against using the lingual vein for venous drainage, due to its cranial position under the mandible. Zhang et al. [9] recommends venous anastomosis on branches of the internal jugular vein, rather than on the external jugular vein, to avoid torsion and compression of the pedicle. According to our data, anastomosis with the external jugular vein does not entail any additional risk. However, we observed an increased risk of flap failure using the anterior jugular vein, after controlling for various confounding factors (HR = 22.19 [1.702–289.3], p = 0.018), probably due to the risk of pedicle compression, in particular, by performing a transient tracheotomy. Of the eight patients for whom a venous anastomosis with the anterior jugular vein was performed, six (75%) had a tracheotomy. Of these six patients, two had a flap failure, and one required revision of the venous anastomosis. In the two patients who underwent anastomosis with the anterior jugular vein without tracheotomy, no flap necrosis was observed. With these results, we strongly advise against anastomosis with the anterior jugular vein in cases of tracheotomy. The risk with this recipient vein is probably not increased in the absence of tracheotomy, but this result needs to be confirmed on a larger series of patients. We also found an excess risk of flap failure when using the superior thyroid vein (HR = 37.99 [2.649–545.0], p = 0.007). We could explain this result by several hypotheses: the anastomosis with the superior thyroid vein could lead to changes in pedicle orientation, greater dissection, and skeletonisation of the vessels, or greater tension applied to the anastomoses. We are, therefore, in agreement with Lin et al. [4] who explained why the choice of recipient vein seems to have a greater influence on flap success than the choice of donor artery, since venous anastomosis is more sensitive to extrinsic compression, and venous failure is more frequently found than arterial failure (60% vs. 40% in our data). It should also be noted that our team mainly uses a single venous anastomosis; for example, for a RFFF, we perform an anastomosis downstream of the communicating vein, in order to preserve the superficial and deep venous network in the harvest. The number of patients for whom a double venous anastomosis was performed was too low in our series to allow a comparison on this factor. This parameter could also be interesting to study for the risk of failure of a free flap.

Flap salvage rates vary from 43% to 61% [19, 21, 30, 31] in the literature. In our cohort, 36 flaps (12%) showed signs of vascular distress, and required rapid surgical revision. Of these, 25 were saved (69%). We, therefore, recommend rapid surgical recovery in the event of signs of vascular suffering.

However, some data could not be collected, such as tourniquet duration, use of papaverine or anticoagulants and antiaggregants. Lin et al. [4] showed an increased success rate after the protocolized use of peroperative papaverine by local application.

Perioperative death ranged from 0.3% to 3.2% in the literature [10, 11, 14, 16, 20, 21], which was consistent with our results, as perioperative mortality was 2.7% at 30 days. Multivariate analysis enabled us to identify strong factors increasing perioperative mortality, such as cardiovascular risk factors (HR = 6.782 [1.311–35.08], p = 0.022) and venous thromboembolic risk factors (HR = 101.9 [18.29–567.8], p < 0.001). We, therefore, reiterate the importance of a good preoperative and postoperative assessment of patients, in order to optimize personal treatments with the cardiologist and carry out the necessary investigations to anticipate any cardiac failure during the period of hospitalization. In our series, the risk of venous thromboembolism appeared to be high, and we, therefore, emphasize the importance of preventive anticoagulation, early mobilization of patients and vigilance for signs of phlebitis or pulmonary embolism. An increase in BMI (and conversely a decrease in undernutrition) appears to be a protective factor after taking confounding factors into account (HR = 0.764 [0.596–0.980], p = 0.034). It, therefore, seems essential to us to insist on nutritional optimization for patients in the pre‐ and postoperative period. A dietetic assessment and prescription of oral nutritional supplements, or sometimes enteral nutrition, should be carried out, as previously suggested by Herzog et al. [32]. Finally, although it does not affect the success of the free flap, operating time is a factor that significantly increases perioperative mortality (HR = 1.015 [1.010–1.019] per minute of surgery, p < 0.001). These factors should, therefore, be taken into account before deciding on free flap reconstruction.

Finally, we feel it is important to emphasize that the patient's age in itself has no effect on the failure of a free flap, as several teams have suggested [13, 29, 33, 34], nor on perioperative mortality according to our results. In practice, in our center, patients are referred to an oncogeriatric consultation if they have a significant ONCODAGE G8 score [35]. A score of 14 or less on 17 indicates a patient's vulnerability, which may contraindicate surgery. This score includes items on age, BMI, weight loss, number of medications and cognitive impairment. According to our results, it seems more interesting in a preoperative context to consider cardiovascular comorbidities and risk factors for thromboembolic complications, in addition to undernutrition status, rather than the patient's age itself.

5. Conclusion

We have identified prognostic factors for free flap failure, such as the choice of venous anastomoses with the anterior jugular vein, in case of tracheotomy, and superior thyroid vein, or the need for revision of the anastomoses, after taking into account various confounding factors, on a large prospective cohort of free flaps performed in our department.

Microsurgical factors, and, in particular, the choice of venous anastomosis, appear to have a greater influence on the success or failure of a microvascularized transfer than patient characteristics, particularly, the age.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.