Learning Sensory Nerve Coaptation in Free Flap Breast Reconstruction

Abstract Background  The aims of this study were to assess whether sensory nerve coaptation in free flap breast reconstruction is subject to learning, and to elucidate challenges of this technique. Methods  In this single-center retrospective cohort study, we reviewed consecutive free flap breast reconstructions performed between March 2015 and August 2018. Data were extracted from medical records, and missing values were imputed. We assessed learning by exploring associations between case number and probability of successful nerve coaptation using a multivariable mixed-effects model. Sensitivity analysis was performed in a subgroup of cases with evidence of attempted coaptation. Recorded reasons for failed coaptation attempts were grouped into thematic categories. Multivariable mixed-effects models were used to examine associations between case number and postoperative mechanical detection threshold. Results  Nerve coaptation was completed in 250 of 564 (44%) included breast reconstructions. Success rates varied considerably between surgeons (range 21–78%). In the total sample, the adjusted odds of successful nerve coaptation increased 1.03-fold for every unit increase in case number (95% confidence interval 1.01–1.05, p  < 0.05), but sensitivity analysis refuted this apparent learning effect (adjusted odds ratio 1.00, 95% confidence interval 1.00–1.01, p  = 0.34). The most frequently recorded reasons for failed nerve coaptation attempts were inability to locate a donor or recipient nerve. Postoperative mechanical detection thresholds showed a negligible, positive association with case number (estimate 0.00, 95% confidence interval 0.00–0.01, p  < 0.05). Conclusion  This study does not provide evidence in support of a learning process for nerve coaptation in free flap breast reconstruction. Nevertheless, the identified technical challenges suggest that surgeons may benefit from training visual search skills, familiarizing with relevant anatomy, and practicing techniques for achieving tensionless coaptation. This study complements prior studies exploring therapeutic benefit of nerve coaptation by addressing technical feasibility.


INTRODUCTION
Losing a breast is a common reality for women around the world: over 2 million women were diagnosed with breast cancer in 2020 1 , and mastectomy is performed in 20 to 97% of cases. 2 2 Breast reconstruction mitigates the toll mastectomy may take on a patient's wellbeing. 3,4 Using a free flap from elsewhere on the body, a natural looking breast can be recreated, with high levels of patient satisfaction. 5 But functionally, a reconstructed breast remains a pale imitation of the one it replaces: the transplanted tissue cannot lactate and rarely regains normal sensation. A reconstructed breast devoid of sensation is prone to burns and other injuries [6][7][8] , may feel unnatural to the patient 9 , and provides limited sexual benefit. 10 Nerve coaptation is one of several innovations surgeons have proposed for restoring sensory feedback. 11,12 The technique involves connecting a donor nerve in the territory of the reconstruction flap to a transected recipient nerve in the chest, and is thought to improve sensory recovery by facilitating axonal regeneration. The concept of sensate free flaps dates to the 1970s, when their benefit was first explored in extremity [13][14][15] and head and neck reconstruction. 16,17 In 1992, Slezak and colleagues introduced the sensate flap in breast reconstruction. 18 Nowadays, various techniques are used in a range of flaps. [19][20][21] Although evidence from observational studies is mounting in support of nerve coaptation 19 , its benefit remains to be confirmed by controlled clinical trials. Exploring learning effects can help determine how much practice surgeons require before trial initiation. 22 Therefore, the main aim of this study was to assess whether sensory nerve coaptation in free flap breast reconstruction is subject to learning, i.e., whether success rates improve with increasing experience. We evaluated this both on a surgeon and department level. To understand which aspects require practice, our second aim was to elucidate technical challenges of this technique. Lastly, we assessed whether increasing experience translates into better postoperative sensory return. 3 Following institutional ethical approval, we conducted a retrospective cohort study at a single teaching hospital in Maastricht, the Netherlands. We included consecutive deep inferior epigastric perforator (DIEP) and lateral thigh perforator (LTP) flap breast reconstructions performed by one of four attending reconstructive microsurgeons between March 2015 and August 2018. To limit missing data, we excluded cases where operative reports were unavailable.

Surgical procedure
Free flap breast reconstructions were performed in standard fashion, as previously described. [23][24][25] From March 2015 onwards, four surgeons started regularly attempting sensory nerve coaptation at their own discretion, according to the following procedure. The lead surgeon harvests a donor nerve along with the reconstruction flap. In DIEP flaps, a cutaneous branch of the 10 th , 11 th , or 12 th intercostal nerve is isolated from accompanying motor branches and dissected as a donor nerve. In LTP flaps, the posterior or occasionally anterior branch of the lateral femoral cutaneous nerve is used. A senior reconstructive surgery trainee dissects the internal mammary vessels in the second or third intercostal space, resecting the rib cartilage only if necessary for adequate exposure. In the same surgical field, the anterior cutaneous branch of the corresponding intercostal nerve is dissected as a recipient nerve. 26 After completing vascular anastomoses and flap shaping, direct end-to-end nerve coaptation is performed with 9-0 nylon and a drop of fibrin glue. No nerve conduits or allografts are used.
In reconstructions involving two stacked flaps, the reinnervated flap is positioned on top when technically feasible.

Variables
Patient characteristics and operative details were extracted from medical records. Each breast reconstruction was considered a separate case. The primary outcome studied to assess learning was sensory nerve coaptation as documented in the operative report, a dichotomous variable (i.e., successful nerve coaptation versus no nerve coaptation). The secondary outcome was postoperative mechanical detection threshold, for which data from prior studies were reused. [27][28] In these prior studies, nine anatomical points per breast were measured using a 20-piece Touch-test™ Sensory Evaluator (North Coast Medical, Inc., Gilroy, CA, USA). 23 Scores represent the logarithm of 10 times the force in milligrams required to bow the monofilament; lower values indicate better sensation. For each breast, the mean of nine scores was entered for analysis. Measurements made less than one year after reconstruction were excluded. Case number, taken to represent surgeon experience, was the primary independent variable of interest. This variable was quantified by arranging included breast reconstructions in chronological order of operation date and sequentially numbering them per surgeon.
Simultaneous bilateral reconstructions were both given the same number, and procedures performed collaboratively by two surgeons were included only in the lead surgeon's series.
Patient characteristics collected included age, body mass index (BMI), smoking status, comorbidities (i.e., hypertension and diabetes mellitus), prior radiotherapy to the chest wall, prior chemotherapy, and preexistent abdominal scars. Operative details collected included indication for mastectomy, prior lumpectomy, prior (infected) implant or tissue expander, timing and laterality of the reconstruction, flap type, flap weight, operative time from incision to closure, ischemia time, rib resection, intraoperative vascular compromise, intraoperative vascular revision (defined as takedown of an anastomosis with revision, or supercharging), and abdominal pedicle(s) used. Recorded reasons for failed nerve coaptation attempts were extracted from operative reports and grouped into thematic categories. Donor and recipient site complications were also registered. The four surgeons studied are referred to using Roman numerals I, II, III, and IV. 5 Normalcy of data distributions was assessed using Shapiro-Wilk tests ( = .01). We compared baseline characteristics of complete cases by outcome group, and, to explore variation in case mix, by surgeon. Missing values were assumed to be missing at random and were accounted for using multiple imputation, generating one hundred imputed datasets. 29 Missing mechanical detection thresholds were not imputed. To examine associations between increasing surgeon experience and probability of successful nerve coaptation, regression analyses were performed on each imputed dataset and results pooled into final estimates according to Rubin's rules. Associations between individual predictors and the outcome were modeled using univariable logistic mixed-effects regression analyses, with breast reconstructions (i.e., the unit of analysis) nested within patients, and patients within surgeons. Based on clinical experience and univariable associations, covariates for a multivariable mixed-effects model were agreed upon by four authors. We checked for multicollinearity (defined as variance inflation factor > 2), high correlation (defined as Pearson's r > .70), and tested for a nonlinear relationship between case number and the outcome by fitting quadratic and cubic models to the data. As nerve coaptation was not routine practice during the study period, the sample likely included breast reconstructions where nerve coaptation was not attempted. To test robustness of results, we conducted sensitivity analysis of procedures performed with intent to coapt a nerve. We first created an intention subgroup of cases that met at least one of the following criteria: any recorded notation of sensate reconstruction discussed with the patient, any mention of donor or acceptor nerve selection or dissection in the operative report, or, in case of simultaneous bilateral reconstruction, evidence of attempted or completed nerve coaptation in the opposite breast. We then recalculated the association between surgeon experience and probability of successful nerve coaptation by restricting multivariable analysis to this subgroup. Associations between increasing surgeon experience and mechanical detection threshold were assessed using multivariable linear mixed-effects regression analysis 6 using the hierarchical structure described above. Covariates were selected based upon prior analyses. 29 All statistical analyses were performed using R version 4.0.5 (R Foundation for Statistical Computing, Vienna, Austria), with  = .05.

Reporting
Study methods and results are reported as recommended in the Strengthening the Reporting of OBservational studies in Epidemiology (STROBE) statement. 30

Characteristics of the study sample
During the study period, 632 free flap breast reconstructions were performed, 564 of which met the eligibility criteria (figure 1). Nerve coaptations were performed in 250 cases (44%), encompassing various flap types and medial as well as lateral row-based abdominal perforator flaps (tables 1 and 2). Rib resection, intraoperative vascular compromise, and intraoperative vascular revision were significantly more common in breast reconstructions without nerve coaptation. Nerve coaptation rates did not significantly differ between unilateral and bilateral cases. The intention subgroup comprised 354 breast reconstructions (i.e., 63% of the total sample; figure 1). Mechanical detection thresholds measured at least one year after reconstruction were available for 258 cases (46%).

Surgeon characteristics and case mix
We studied two female and two male reconstructive microsurgeons, who at the start of the cohort had two to 14 years' experience after qualification. Surgeons I, II, III, and IV performed 175 (31%), 175 (31%), 141 (25%), and 73 (13%) of included breast reconstructions respectively. Fourteen procedures were performed collaboratively by two surgeons yet analyzed only for the lead surgeon. The ratio of immediate versus delayed reconstructions ranged from 3:5 (surgeon I) to 5:2 (surgeon IV; supplemental material 1). 7 Surgeons I, III, and IV performed mainly standard DIEP breast reconstructions; surgeon II performed the majority of included stacked and LTP breast reconstructions.

Learning effects and technical aspects
Upon univariable analysis, we found a positive association between case number and probability of successful nerve coaptation. Patient BMI, delayed breast reconstruction, rib resection, intraoperative vascular compromise, and intraoperative vascular revision were inversely associated with the outcome (supplemental material 2). Intraoperative vascular revision was omitted from the multivariable model because of its strong correlation with intraoperative vascular compromise (Pearson's r = 0.95, 95% confidence interval 0.95 -0.96, p < .001) and weaker predictive value. In our experience, large-volume flaps complicate tensionless coaptation; by including patient BMI as a covariable in the model, we indirectly accounted for flap weight (Pearson's r = 0.68, 95% confidence interval 0.63 -0.72, p < .001).
Multivariable regression analysis in the total sample showed that for every unit increase in consecutive case number, the adjusted odds of successful nerve coaptation increased 1.03-fold (95% confidence interval 1.01 -1.05, p < .05; table 3). This apparent learning effect shrank when repeating analysis in the intention subgroup (adjusted odds ratio 1.00, 95% confidence interval 1.00 -1.01, p = .34). Including consecutive case number as a quadratic or cubic term did not improve the model in either sample.
With an overall nerve coaptation rate of 78%, surgeon II outperformed surgeons I, III, and IV, whose rates equaled 21%, 27%, and 53% respectively (figure 2). Upon visual inspection, ratios between failed and successful nerve coaptation attempts (medium versus dark grey levels, figure 2) vary heavily between surgeons, but show no obvious experience-related patterns.
The most frequently recorded reasons for failed nerve coaptation attempts were inability to locate a recipient nerve or an abdominal donor nerve (figure 3). Abdominal scarring was 8 present in 14 of 22 recorded cases where no (suitable) donor nerve could be identified. In one instance, surgeon II used a 15 mm autograft from an abdominal intercostal nerve to bridge the nerve gap. In 39% of cases, operative reports did not say why nerve coaptation failed.

Complications
Fewer donor and recipient site complications occurred in breast reconstructions with nerve coaptation than in breast reconstructions without nerve coaptation (figure 5). Only two cases of abdominal bulge were recorded in the total cohort, and zero cases of abdominal wall herniation. Complication rates did not vary significantly between surgeons (supplemental material 3).

DISCUSSION
The aims of this study were to explore learning effects of nerve coaptation in free flap breast reconstruction, and to elucidate challenges of this technique. Our results do not provide evidence in support of a learning process: we found a statistically significant increase in collective success rates with surgeon experience, but sensitivity analysis refuted this apparent effect. We observed major disparities in nerve coaptation rates between surgeons, ranging from 21 to 78%. We found a minor, positive association between case number and mechanical detection threshold, meaning that increasing surgeon experience did not improve sensory return. At the coaptation stage, tensionless repair is crucial, for even minimal tension between donor and recipient nerve can impair functional recovery. 39 Our cohort included 10 cases where a gap prohibited direct coaptation, one of which was managed with an autograft from the abdomen. Alternatively, nerve conduits or allografts can be used 20,40 , with the disadvantage of entailing additional financial costs. If dissected donor and recipient nerves are of sufficient length, or the gap can be resolved, the mere act of suturing nerve ends should be a straightforward task to the experienced microsurgeon.
In addition to technical issues linked to specific nerve coaptation stages, overall procedural difficulty deserves consideration. We found lower probability of successful nerve coaptation in case of rib resection, intraoperative vascular compromise, and intraoperative vascular revision. Also, more major complications were recorded in the outcome group without nerve coaptation ( figure 4). These data suggest that providing a sensate reconstruction dropped on the surgeon's priority list during procedures that were particularly challenging, a theory that resonates with surgeons reporting 31% more physical demand and 78% more fatigue when DIEP procedures last longer than 9 hours. 41 Free flap breast reconstruction is a team effort, and variability in team composition might influence outcomes. In our hospital, a dedicated team of scrub nurses assists all free flap procedures, and all procedures routinely involve a senior reconstructive surgery trainee. Lack of variability in these factors prevented us from studying their potential association with the probability of successful nerve coaptation. In a retrospective study by Jubbal and colleagues, trainee involvement in free flap breast reconstruction was not associated with increased likelihood of major complications. 42 Based on this finding, we assume that trainee 11 involvement does not indirectly lower the probability of successful nerve coaptation by increasing overall procedural difficulty.
This study has three important limitations. First, successful nerve coaptation, our primary outcome, is a measure of surgical process that only indirectly relates to patient outcome. 43 We therefore also explored associations between increasing experience and sensory recovery, the patient-relevant outcome in this context. However, the complex sum of factors that determines sensory recovery makes its use as a measure of surgical skill questionable at best.
Second, operative reports yielded partly missing and ambiguous data, such that that the intention subgroup might in fact be larger and nerve coaptation rates higher than reported.
Third, our single-center design limits generalizability of findings to other surgeons, centers, and nerve coaptation techniques.
Benefits minus costs determine a treatment's utility. Observational studies previously explored therapeutic benefit of nerve coaptation 19 ; our study complements these studies by discussing technical feasibility. Donor site complications and added operative time constitute potential costs to be considered in the nerve coaptation utility equation. Previous studies report an anecdotal extra 8 to 35 minutes required for nerve coaptation, but none report data supporting these estimates. 40,[45][46][47] For our surgical approach, we estimate that nerve coaptation increases net operative time by maximum 20 minutes in unilateral breast reconstruction (i.e., 5-10 minutes for donor nerve dissection and 10 minutes for coaptation; recipient nerve dissection does not add time because flap dissection is always a lengthier process). Further study is required to allow for data-driven conclusions regarding added operative time.

CONCLUSION
We found no evidence of learning effects for nerve coaptation in free flap breast reconstruction. The main technical challenges identified by this study are donor nerve identification, recipient nerve identification, and tensionless coaptation. We suggest that prior

Funding statement
No funding was received for this work.

Conflict of interest
None declared.   This article is protected by copyright. All rights reserved.

Accepted Manuscript
Values are n (%) for 104 failed nerve coaptation attempts. This article is protected by copyright. All rights reserved.

Accepted Manuscript
Supplemental Material 1  This article is protected by copyright. All rights reserved.

Accepted Manuscript
Supplemental Material 2 Results are derived from univariable logistic mixed-effects regression analyses. Odds ratios represent the effect that a given predictor has on the probability of successful nerve coaptation. a p < .05. b Model failed to converge due to few events, lack of variance, or both.