Effect of Serial Intralesional Steroid Injections on Risk of Recurrence in Idiopathic Subglottic Stenosis

2.1 Patient Cohort Description

The study was approved by the research ethics board at both participating institutions (Western HSREB 115746, Unity Health REB 19–324). Written informed consent was obtained from all study participants. Patients 18 years of age or older with a diagnosis of iSGS were ascertained from Unity Health Toronto-St. Michael's Hospital (Toronto, Ontario, Canada) or London Health Sciences Centre (LHSC)—Victoria Hospital (London, Ontario, Canada). Only patients with at least 1 year of prospectively collected clinical data were included. Patients with a history of cricotracheal resection were excluded, given prior evidence indicating a substantially lower risk of recurrence compared to endoscopic dilations [4]. All participants consented to a biopsy of their affected subglottic tissue during a surgical dilation procedure for molecular analyses [2, 15]. The prospective follow-up period was defined as the time from the “index” surgical dilation during which the index biopsy was collected, to the most recent surgical or clinical appointment, until the study cut-off date of July 1, 2024. All data in this study were collected after approval by institutional board review.

All ILSIs were performed in the office under local anesthesia. The majority of the injections were performed trans-nasally via a channeled laryngoscope, and some patients received the steroids via the trans-cricothyroid approach due to a hyperactive gag reflex. Each patient received up to 3 mL of 40 mg/mL of triamcinolone at each injection.

2.2 Study Outcomes and Statistical Analyses

The primary outcome of the study was stenosis recurrence, defined as each surgical dilation following the index dilation that marked the start of the prospective follow-up period. Surgery-free interval was defined as the time between consecutive surgical dilations. To assess the effect of ILSIs administered during the surgery-free interval on the risk of recurrence, we used both a standard (time-to-first event) and the Andersen-Gill extension of the Cox proportional hazard model [21, 22]. First, a standard Cox proportional hazards model was applied to compare the time-to-first recurrence between patients who received one or more ILSI(s) before their first recurrence and those who did not. Second, to include all available longitudinal data points, including when patients experienced multiple recurrences (i.e., three or more surgical dilations), we utilized the Andersen-Gill extension of the Cox proportional hazards model. In this model, ILSI use status was annotated for each surgery-free interval, that is, treated as a time-varying covariate when patients received ILSI(s) during some surgery-free intervals but not others. Thus, specifically the surgery-free intervals where ILSIs were administered were compared with surgery-free intervals where ILSIs were not. Both univariate and multivariate analyses were performed, with the latter adjusting for age at diagnosis, years since diagnosis, body mass index (BMI), and antacid use. Categorical variables required a minimum count of 10 in each category to be included as covariates in the multivariate models. Additionally, years since diagnosis was coded as a time-varying covariate in the recurrent events model, indicating the number of years elapsed since diagnosis at the beginning of each surgery-free interval. For both models, patients were considered censored if their last available record corresponded to an ILSI or clinic appointment. The proportional hazards assumption in the standard Cox proportional hazards model was assessed using Schoenfeld residuals with the cox.zph function [23]. Plots of the scaled Schoenfeld residuals were generated using the ggcoxzph function and visually inspected for a downward or upward trend of the scaled Schoenfeld residuals when the test indicated a statistically significant (p < 0.05) violation of the proportional hazards assumption. Missing data were handled in the models using imputation, using the median value for one missing BMI and the most frequent value for one missing antacid use status.

The number needed to treat (NNT) with serial ILSIs to prevent one recurrence was calculated at multiple time points using Kaplan–Meier estimates of the time-to-first recurrence analysis. The absolute risk reduction (ARR) associated with ILSIs was determined at each time point, and the NNT was derived as the reciprocal of ARR (i.e., NNT = 1/ARR).

All analyses were conducted in R version 4.3.2. The survival package was used to implement the standard and recurrent events Cox proportional hazards models. survfit2 was used to create Kaplan–Meier plots. Statistical significance was defined as p < 0.05.

3.1 Cohort Description

In total, there were 75 patients with iSGS with at least 1 year of prospective follow-up data, and no history of cricotracheal resection (Table 1, Figure 1). All patients were female and White. The median age at diagnosis was 47 years (interquartile range (IQR) 34–55 years) and the median BMI was 28.3 kg/m² (IQR 24.5–34.3). Antacid use was reported by 31 (41.9%) patients. Smoking (current or past history), family history of iSGS (any affected first or second degree relative), and diabetes (type I or II) were infrequent in this cohort (Table 1).

Among these 75 patients, there were 687 recorded surgical or clinical appointments, including 151 surgical dilations (22.5%) and 200 ILSIs (29.8%) (Table S1, Figure 1). Of the 151 surgical dilations, 148 (98.0%) were balloon dilations while the remaining were rigid dilations. Nearly all dilations were performed with microsuspension laryngoscopy and were CO₂ laser assisted to divide the stenosis (Table S1). In most cases, intra-lesional steroids were administered at the end of the procedure (Table S1). The overall rate of surgical dilations was 0.82 dilations per year (IQR 0.68–1.21) (Figure S1). There were 30 patients (39.0%) who received only the initial surgical dilation and did not require a second dilation. Twenty-eight patients (37.3%) received two surgical dilations and the remaining received three to five dilations within the 1–4 year follow-up period (Table S1). Approximately half of the cohort (n = 39; 52.0%) had never received an ILSI at any point in the follow-up period while the remaining (n = 36; 48.0%) received between one to 14 ILSIs (Tables 1, S1, and S2).

3.2 Effect of Serial Intralesional Steroid Injections on Risk of Recurrence

We first compared the time-to-first recurrence, defined as the first subsequent surgical dilation, between patients who received (n = 22) versus those who did not receive (n = 53) one or more ILSI(s) prior to their first recurrence or date of censoring (Figure 2, Table S3). ILSI use was associated with significantly lower risk of recurrence in a univariate model (hazard ratio (HR) = 0.23, 95% confidence interval (CI) 0.10–0.54) and in a multivariate model (HR = 0.20, 95% CI 0.08–0.49) (Table 2). The median time to recurrence among those who received ISLIs was 2.5 years compared to 1.4 years among those who did not (Figure 2). In the multivariate model, older age at diagnosis (HR = 0.96, 95% CI 0.94–0.98) and longer time since diagnosis (HR = 0.88, 95% CI 0.80–0.98) were also associated with significantly reduced risk of recurrence (Table 2). We evaluated the proportional hazards assumption using Schoenfeld residuals and identified a deviation from model assumptions for ILSI use status near the end of the follow-up period, beginning at approximately 2.2 years (Figure S2). To address this, we repeated the analysis censoring all data beyond 2.2 years. This adjustment resulted in minimal differences in the outcomes (Table S4).

We also utilized a recurrent event-adapted Cox proportional hazards model to account for all recurrences in our dataset (Table 2). This analysis included 43 surgery-free intervals (from n = 36 patients) during which one or more ILSI(s) were administered and 102 surgery-free intervals without ILSI use (from n = 55 patients; these include intervals where no ILSI was administered in patients who had received ILSIs during other intervals). Consistent with the findings from the time-to-first event analysis, ILSI use was associated with a significantly lower risk of recurrence in both the univariate model (HR = 0.44, 95% CI 0.26–0.73) and the multivariate model (HR = 0.44, 95% CI 0.26–0.75). Similarly, older age at diagnosis and a longer duration since diagnosis were significantly associated with a reduced risk of recurrence in the multivariate model (Table 2).

3.3 Number Needed to Treat With Intralesional Steroid Injections

We calculated the NNT with ISLIs to prevent one recurrence at 1.0, 1.5, and 2.0 years follow-up, based on the ARR at each time point derived from the Kaplan–Meier estimates in the time-to-first recurrence analysis. The percent of recurrence-free individuals in the ILSI group versus the no-ILSI group at 1.0, 1.5, and 2.0 years was 100.0% vs. 67.9%, 95.5% vs. 44.4%, and 77.6% vs. 32.6%, respectively (Figure 2B). These correspond to an NNT of three, two, and two at 1.0, 1.5, and 2.0 years, respectively (Table S5).

3.4 Frequency of Intralesional Steroid Injections

Given that ILSIs are typically administered in a serial manner, we quantified their usage frequency among all patients who had ever received an ILSI (n = 36 patients). Among 43 surgery-free intervals where one or more ILSI(s) were received, either between consecutive surgical dilations or from dilation to censoring, the median number of ILSIs received was four (range 1–8) (Figure 3A). Among the 200 recorded ILSIs, the median time from either the previous surgical dilation or the previous ILSI was 53 days (IQR 42–91) (Figure 3B).

4.1 Advantages and Limitations

To our knowledge, this is one of two prospective studies on the effect of serial ILSIs as an adjunctive treatment for patients with iSGS [20]. This study included 36 patients who received ILSIs and 39 who did not, representing a substantially larger sample size than previous retrospective studies, where fewer than 10 patients with documented surgical dilations prior to ILSI initiation were available for comparison [6, 9, 12]. Additionally, this study exclusively focused on patients with iSGS, in contrast to many previous studies that included more heterogeneous causes of subglottic stenosis [6, 9, 12]. In contrast to a previous prospective study on serial ILSI [20], our data capture the time point that each ILSI was received and the total number of ILSIs received between each recurrence. We applied both time-to-first event and recurrent event models to account for all available surgical events, providing a comprehensive analysis. Furthermore, the availability of detailed clinical data allowed for multivariate models that included other covariates associated with recurrence risk.

This study has several limitations. Although a minimum of 1 year of follow-up was required for inclusion, a longer follow-up period would likely provide a more accurate assessment of the recurrent nature of iSGS. When testing the proportional hazards assumption using Schoenfeld residuals, we observed a trend towards a higher risk of recurrence in the ILSI group after approximately 2.2 years (Figure S2). While this may suggest an upper limit to the efficacy of ILSIs in maintaining surgery-free intervals, it could also represent a spurious trend due to the small number of patients at risk beyond that time point (Figure 1B). Additionally, the relatively higher hazard ratio from the recurrent events model (HR = 0.44, 95% CI 0.26–0.75) compared to the time-to-first event model (HR = 0.20, 95% CI 0.08–0.49) may indicate that the relative effectiveness of serial ILSIs diminishes with successive surgery-free intervals. Longer follow-up with more patients extending beyond the first surgery-free interval will be essential to evaluate the long-term durability of serial ILSIs. While we regard requiring a surgical dilation to be the most relevant clinical outcome of restenosis, our study lacks measurement of functional outcomes, such as peak expiratory flow, to assess direct physiological changes that may be conferred by ILSI use. Future studies will also be necessary to investigate potential side effects of serial ILSIs [6]. Finally, despite adjusting for covariates, the results from this non-randomized study may still reflect bias arising from factors such as variations in surgeon preference, patient access to healthcare, and other demographic features.