A total of 103 PTTS patients underwent SEMS insertion between January 2000 and December 2008 in our institution. According to the aforementioned inclusion criteria, 77 out of 103 (74.8%) PTTS patients were eventually included into this study (Figure 1). Table 1shows patients’ baseline characteristics. In 77 enrolled patients, the median age was 32.77±10.73 years and 75.3% were female. The symptoms-diagnosis time window was 5.62 ± 6.24 months. The mean duration of treat tuberculosis with anti-tubercular medications was 8.42 ± 5.30 month. All patients were diagnosed with airway stenosis related to transbronchial tuberculosis. 85 SEMS were implanted in 77 patients with PTTS during the study period. 51 (60.0%) uncovered SEMS (Nanjing Micro-Tech Co. Ltd, China) and 34 (40.0%) uncovered Ultraflex SEMS (Boston Scientific, USA) were deployed successfully for PTTS patients involving left main bronchus (n=65, 84.4%), right main bronchus (n=7, 9.1%) and right middle bronchus (n=5, 6.5%).
Short-term clinical outcomes and complications of SEMS
The cough, median mMRC scale and lumen diameters at the stenotic site for all patients 1 week after SEMS insertion (53.25%, 0.26 ± 0.05 and 8.23 ± 1.44, respectively) with no pending complications were significantly improved in comparison with those before SEMS insertion (32.47%, 1.60 ± 0.61 and 3.69 ± 1.35, respectively; Cough: P=0.009, mMRC scale: P<0.001, lumen diameters: P<0.001). Spirometry tests showed statistically significant increases in the mean values of forced expiratory volume in 1 second (FEV1), forced vital capacity (FVC), FEV1 predicted and FVC predicted from baseline. 6 patients with atelectasis of left lung resulting from complete occlusion of the left main bronchus experienced almost totally successful recruitment after 1-week SEMS placement. Compared with pre-stenting, there were statistically significant improvements in the pectoralgia, mean mMRC scale and lumen diameters of the stenotic segment after 6-month stenting. Furthermore, a statistical difference in the palliation of the cough was observed between 1 week and 6 months after stenting (P=0.04) (Table 2).
During the period of 6-month follow-up, a total of 23.4% (18 out of 77) patients developed the stent-related complications, including granulation proliferation (n=15, 19.5%), overgrowth of necrotic tissue (n=1, 1.3%), and migration (n=2, 2.6%). SEMS replacement occurred in 8 patients because of the inappropriate size of stents (n=6) and stent migration (n=2). All 8 patients underwent a successful removal of the SEMS without significant complications. In addition, 11 patients suffered restenosis due to overgrowth of granulation tissue in the 1st (n=2), 2nd (n=1), 3rd (n=2), 4th (n=3) and 5th (n=3) months.
Long-term clinical outcomes and complications of SEMS
At the average follow-up duration of 163.32 months, 48 patients (62.3%) did not develop restenosis after SEMS insertion, the other 29 patients (37.7%) developed. The results suggested that the type of stenosis of restenosis patients was of statistical difference with non-restenosis patients (P=0.042). For restenosis patients, mixed stenosis (n=13, 44.8%) and cicatricial stenosis (n=10, 34.5) were the major types of stenosis. Moreover, restenosis patients presented better epithelialization of SEMS by comparison of non-restenosis patients, with a statistical significance (P=0.003). However, there was no significant difference in age, sex, site of stenosis, the number of other interventional treatments before stenting, and the type of SEMS between restenosis and non-restenosis patients (Table 3).
During the stenting period of 152.19±54.31 months, almost all patients experienced stent-related late complications in which granulation tissue formation (63.6%) and scarring tissue proliferation (23.4%) commonly occurs. Furthermore, the rate of overgrowth of granulation tissue resulting in restenosis reached 33.8%. Figure 2A shows the Kaplan - Meier survival curve of restenosis after SEMS placement. The 1-month, 6-month, 1-year, 3-year and 5 year restenosis rates were 2.6%, 14.3%, 23.4%, 35.1% and 37.7%, respectively. But the median time of restenosis after stenting was not estimated by this curve.
Among 29 restenosis patients, the median duration of restenosis was 10 months (range from 1 to 60 months). Granulation tissue ingrowth (90.0%), scarring tissue proliferation (51.7%) and secretion retention (10.3%) were three common complications of SEMS (Table 3). A variety of repeat interventional treatments had been performed to deal with the aforementioned complications. During at least 10 years of follow-up, 4 (13.8%) patients displayed a stable clinical condition, 12 (41.4%) required persistent interventional therapies, 11 (37.9%) had bronchial atresia and 9 of them were referred for surgery, and 2 (6.9%) lost follow-up. The mean duration of SEMS implantation in restenosis group was 132.38 ± 64.05 months (Figure 1). With respect to prognosis, the cough (37.0%) and median mMRC scale (1.78 ± 1.01) were not significantly improved compared with pre-stenting (24.1% and 1.66 ± 0.61, respectively; Cough: P=0.224, mMRC scale: P=0.43). The worse shortness of breath occurred in 2 patients with poor response to stenting, severely affecting quality of life. In all restenosis patients, significant improvements in the pectoralgia and lung function were observed at 10 years after stenting (Table 4).
At the stenting duration of 164.17 ± 44.00 months, the overall incidence of stent-related complications was 66.7% (32/48), including granulation tissue formation (47.9%), scarring tissue proliferation (6.3%), necrotic tissue overgrowth (2.1%), mucus plugging (4.2%), stent fracture (4.2%), and infections (2.1%) (Table 3). These were of minor severity, and were self-correcting or appropriately managed with observation or endoscopic intervention. During the period of follow-up, 93.8% (45 out of 48) patients exhibited a stable clinical condition in which no severe complication developed due to the well-tolerated SEMS. The other 3 patients were lost to follow-up (Figure 1). As regard to prognosis, the median luminal diameter of the stenotic segment increased from 3.51 ± 1.42 mm to 6.29 ± 1.10 mm (P<0.001). The functional effects of which were embodied as significant improvements on spirometry, the pectoralgia and the mMRC Scale. However, no significant improvement in the cough was examined between pre-stenting and 10 years after stenting (P=0.084) (Table 4).
Restenosis Score prediction model
Univariate and multivariate cox regression analysis
A total of 13 variables which were selected in this cross-sectional study were bound up with restenosis in patients after stenting and readily available in the clinical setting. These variables consisted of sex, age, symptoms-diagnosis time window, pre-stenting anti-tubercular therapy, initial treatment, type of stenosis, the number of pre-stenting interventional treatments (i.e. thermal ablation, cryotherapy and balloon dilation), type of SEMS, the difference value of the luminal diameter pre- and post-stenting, the difference value between the external diameter of SEMS and the luminal diameter before stenting, and the difference value between the length of SEMS and the length of the stenosis segment. Univariate cox regression analysis indicated that the statistical effects of the difference value between the length of SEMS and the length of the stenosis segment, the number of pre-stenting thermal ablation and cryotherapy were significant. Type of stenosis did not reached a statistical significance (P=0.09), however, it was included in the followed multivariate cox regression analysis on the basis of clinical evidences and published articles [13, 14]. On the contrary, considering small sample size in the training cohort, the number of pre-stenting cryotherapy was excluded.
The construction of Restenosis Score prediction model
After final multivariate cox regression analysis, the Restenosis Score retained type of stenosis, the difference value between the length of SEMS and the length of the stenosis segment, and the number of pre-stenting thermal ablation (Table 5). A detailed description of the Restenosis Score is also displayed in Table 5. The Restenosis Score of all patients is from -8 to 8, and higher score is interrelated to greater predicted incidence of restenosis. The area under the receiver-operating characteristic curve (AUROC) in the development group was 0.83 (95% CI: 0.74-0.92, P＜0.001), implying that there was a significant discrimination with the Restenosis Score prediction model (Figure 3). Hosmer-Lemeshow χ2 of 5.8 (P=0.33) in the development group signified good model calibration. The development group and validation group were separated into two risk stratification depending on the cut-off point: ≤0 (low risk), ＞0 (high risk). The restenosis rates of the development group in low-risk and high-risk patients were 20.8% and 65.5%, with a statistical significance (P＜0.001). There were significant differences in the Kaplan - Meier restenosis survival curve between low-risk and high-risk patients after stenting (Log Rank or Breslow Test, P<0.001). The median time of restenosis with high-risk patients was 13.00 ± 3.44 months, however, low-risk patients failed to be estimated (Figure 2B).
Verification of model performance
We then employed the external validation group to verify the above findings from the development group. The external validation group comprising 10 patients demonstrated a restenosis rate of 80%. ROC analysis utilizing the Restenosis Score displayed an excellent discrimination with an AUC of 0.94 (95% CI: 0.77-1.00) and the Hosmer–Lemeshow analysis for the Restenosis Score showed good calibration (χ2 = 3.29, P = 0.771). Applying the aforesaid risk classification to the external validation cohort yielded a restenosis rate of 33.3% for low-risk (n=3), and 100.0% for high-risk (n=7) patients, respectively. The difference of restenosis rate between low-risk and high-risk patients reached a marginally statistical significance (Fisher’s Exact Test, P=0.067).