Predictors of recurrent strictures after oesophageal atresia repair

Oesophageal stricture, Delayed repair, Prematurity, Oesophageal dilatation Background: Anastomotic strictures continue to complicate the outcome after oesophageal atresia (OA) repair. Multiple variables contribute to the development of strictures, and oesophageal dilatations are the mainstay of treatment. We aim to analyse the factors that impact the timing for initiation of oesophageal dilatations, the duration, frequency, and success of the dilatation regimen for OA.


INTRODUCTION
The surgical management of oesophageal atresia (OA) without a tracheo-oesophageal fistula (TOF) (Gross Type A), with a proximal fistula (Type B), distal fistula (Type C), or proximal and distal fistulas (Type D) is frequently complicated by anastomotic strictures in 18% to 59% of cases. [1][2][3][4][5][6][7] Oesophageal dilatation is the mainstay of treatment with a reported success rate of 70-100% after three median dilatation sessions. [3,6,8,9] Several factors play a role in the aetiology of strictures. These include prematurity and low birth weight; long oesophageal gap length resulting in anastomotic tension, and due to its segmental blood supply vascular compromise to the lower oesophagus from excessive mobilisation; suture type; anastomotic leak; and gastro-oesophageal reflux (GOR). [3][4][5][6]10,11] There is limited published data on whether these factors can cause strictures to become recurrent or refractory and therefore difficult to manage. [9] However, it is accepted that the different anatomic variants have different outcomes, with Type A being associated with long oesophageal gap [2] and refractory strictures. [9] We aim to analyse these factors and determine how they can impact the timing for initiation of oesophageal dilatations, the duration, frequency, and overall success of the dilatation regimen for Type C OA.

METHODS
After obtaining institutional review board approval, a retrospective review of data of all children who underwent repair of Gross Type C oesophageal atresia between October 2004 and December 2017 was performed. Data on demographics, comorbidities, strictures, leaks, and dilatations were obtained for the cohort.
Significant anastomotic leaks were defined as those that presented early with either radiological opacification of the right hemithorax or with saliva draining out of the thoracostomy drain. Small, contained leaks were defined as leaks identified at contrast swallow. Oesophageal strictures were defined as anastomotic narrowing on contrast oesophagram in children that were symptomatic with regurgitation or dysphagia and required management with oesophageal dilatations.
A long gap between the oesophageal ends, when encountered, was documented but its definition is not standardized. However, when gap length was significant, delayed oesophageal anastomosis at a second procedure was performed after assessment for improved gap length. Therefore, children that underwent delayed repair were analysed as somewhat representative of the long gap group.
Children that had died, had an oesophageal traction technique or a primary oesophageal replacement were excluded.
Acid suppression was prescribed to all children, but the duration of use and compliance was not consistently documented and therefore not included in the analysis. Likewise, gastro-oesophageal reflux was not consistently assessed with pH monitoring for the entire cohort and therefore not analysed.
Fisher's Exact test was used to test the association between categorical variables. Student's T-test for comparing means of parametric continuous data and Mann-Whitney U test was used to compare nonparametric data sets. Spearman's Rho test was used for non-parametric correlation analysis. Stepwise linear regression analysis was used to analyse the relationship between dependent variables and independent categorical and continuous variables. Significance was set at P < 0.05.

RESULTS
There were 91 children with Gross Type C anomaly. Nineteen children were excluded and data of the remaining 72 children were analysed. The follow-up duration had a range (median) of 2.4 to 148 (80.5) months (Mean 75.3 ±45 months). Their mean (range) birth weight (BW) and gestational age (GA) were 2.67kg (0.98 to 3.86kg) and 37.4weeks (29.6 to 42 weeks), respectively. Table 1 shows the number of children with associated comorbidities, complications, and further related procedures including anti-reflux surgery. Not shown are the other severe comorbidities encountered in the children that had died during the study period.  Total study population =72 There were 10 (13.9%) children who due to long gap length underwent ligation of their fistula and formation of a gastrostomy only. This was then followed by delayed oesophageal anastomosis. The delayed repair was performed post-birth at a median (interquartile range) of 132 days (87 to 178 days). Figure 1 shows that 50% (n=5) of children who had a delayed repair developed strictures compared to 35.5% (n=22) who had a primary repair (P=0.48).
Children that had a delayed repair developed significantly more anastomotic leaks compared to children that had a primary repair (40% vs 11.3%; P=0.04; Figure 1) There was no statistical difference in the mean BW of children who developed strictures compared to those with no strictures (2.74kg vs 2.63Kg; P =0.548).
Similarly, there was no difference in their mean GA respectively (37.4 weeks vs 37.3 weeks; P=0.9). Several treatment modalities were employed for strictures and frequently different treatment options were combined in the same child (Table 2). Six episodes (5%) of oesophageal perforation occurred after 120 dilatation sessions. Balloon + filiform + Maloney + stent 1 aThis child had a delayed repair. Stricture resection was performed after 5 dilatation sessions. One further dilatation was required after stricture resection. bThis child had a primary repair, developed a leak and later a recurrent tracheo-oesophageal fistula identified at the first dilatation session. Stricture resection and recurrent fistula ligation were performed.
One further dilatation session was required.
Overall a median of 3 dilatation sessions were required per child. Those who suffered leaks compared to those without leaks did not require significantly more dilatation sessions respectively (6.5 vs 3 median sessions; P= 0.2). However, those that underwent a delayed repair required significantly more dilatation sessions (12 vs 2 median sessions; P =0.001) ( Table  3). Children with anastomotic leaks compared to those without leaks respectively did not have a significantly earlier onset of initiation of dilatations (62.5 vs 179 median days; P = 0.58) or a significantly more prolonged duration of treatment (580 vs 121 median days; P = 0.3) and their symptom-free period postdilatation was not significantly shorter (37 vs 33 median months; P=0.73). Likewise, children that had a delayed repair compared to those with a primary repair respectively did not have a significantly earlier onset of initiation of dilatations (59 vs 229 median days; P=0.11) and their symptom-free period was not significantly shorter (22 vs 38 median months; P= 0.6). However, they had a significantly prolonged duration of treatment (610 vs 63 median days; P = 0.013) corresponding to the significantly more dilatation sessions required (Table 4).  The linear regression model revealed that delayed repair increased the number of dilatations by an average of 7 sessions compared to primary repair. For every 1 week increase in GA, the number of dilatations reduced by 0.4 sessions. Finally, for every 333 days from primary surgery to first dilatation, the number of dilatation sessions reduced by 1. The delayed repair was the most important factor related to the number of dilatation sessions required (p <0.001); this was followed by lower GA or BW (p = 0.0265) and early onset of dilatations (p=0.0471).s

DISCUSSION
After a successful OA repair, understanding the factors that predispose to recurrent and refractory strictures facilitates counselling information provided to families and helps anticipate this complication to prevent ongoing morbidity from aspiration, feeding difficulties, failure to thrive, dysphagia, and food bolus obstruction. Several reports have described factors that predispose to the development of strictures and leaks after OA repair. [3][4][5][6][7][10][11][12] Few reports have correlated these factors with the onset or intensity of dilatations required to manage strictures. [3,9] Vergouwe et al [9] looked at factors that predisposed to refractory strictures and defined these as an anastomotic stricture requiring ≥5 dilatations at a maximum of 4-week intervals; strictures occurring after 4-week intervals were considered recurrent. The authors included Gross Type C and A variants, end-to-end, oesophagojejunal, and oesophagogastric anastomosis and concluded that isolated OA, anastomotic leak, and the need for oesophageal dilatation ≤28 days after anastomosis were risk factors for refractory strictures.
In this report, we focused on Type C variant with endto-end anastomosis only and found that children that underwent a delayed repair were more likely to develop recurrent or refractory strictures based on their need for significantly more dilatation sessions over a longer period. Delayed repair in our cohort represents those whose gaps were too long to be closed at the primary operation, and so initially underwent fistula ligation only. Their oesophageal anastomoses were deferred and reattempted after gap studies showed improved gap length (no oesophageal elongation techniques used). It has been reported that if tension exists between the oesophageal ends leaks and strictures are more likely. [6,11] However, Thakkar et al [12] showed that as long as the primary repair is possible, measured gap length at the time of surgery was not a risk factor for anastomotic leak or stricture. We theorize that the scarring of the distal oesophagus that occurs after the fistula has been previously ligated and the subsequent mobilisation required results in compromised tissues that are not ideal for suturing. [3] These technical difficulties faced at delayed repair, rather than the actual gap length, maybe the more significant cause for difficult to manage strictures. [13] Furthermore, children who had delayed repair developed significantly more anastomotic leaks. Although anastomotic leaks alone were not associated with difficult strictures, when coupled with a long gap and a delayed repair, it may have resulted in more aggressive scarring that was difficult to dilate. Children with long gap atresia and anastomotic leaks have been reported to suffer more severe strictures. [3,5,6] Premature children also required more dilatation sessions. These smaller children with a smaller calibre oesophagus may have been more technically challenging and more prone to symptomatic luminal narrowing. The initial dilatation sessions may also have been milder and thus requiring more sessions over time.
Interestingly, like another report, leaks alone were not predictive of difficult strictures for this cohort. [5] This differed from the finding by Vergouwe et al., although they included jejunal and gastric anastomosis in their cohort. [9] The development of symptomatic strictures early was also indicative for difficult strictures that required more dilatation sessions. This suggests that early dilatation had a negative impact on the anastomosis or that these strictures had a more aggressive aetiology; this finding was also noted in the report by Vergouwe et al. and Chittmittrapap et al. [9,14] In the latter study they observed that children who developed strictures within 6 months of surgery required five or more dilatations or surgical intervention. This was an independent finding because children who had a delayed repair or developed leaks or were premature respectively were not at more risk for early strictures.
The findings of this report are limited by our inability to adjust for the impact of GOR [3,5] anti-reflux medications [6,10,11] and anti-reflux surgery [11,12] as well as jejunal feeding. Other studies have highlighted the difficulties analysing the GOR variable. [9] Furthermore, several surgeons performed these operations with variable techniques and suture preference. [4] Several different dilatation devices [6,11] were used with some sessions complicated by perforation, while a few subsequent dilatation sessions were scheduled routinely as part of a stricture management regimen. [11] CONCLUSION The development of strictures after OA repair is multifactorial. When strictures occur in the early postoperative period, in premature babies, or after a delayed repair for long gap atresia, recurrent/refractory strictures that require several more dilatation sessions should be anticipated. Where possible, every reasonable attempt at a primary repair should be made and a more frequent clinical review of these subsets of children is recommended, so prompt diagnosis can be made to prevent ongoing morbidity from strictures. The occurrence of an anastomotic leak does not predict a difficult to manage stricture.