Abstract
BACKGROUND: The optimal initial corticosteroid strategy for idiopathic sudden sensorineural hearing loss (SSNHL) remains uncertain, and randomized trials may not capture real-world variability. Observational comparisons are vulnerable to confounding, particularly when post-treatment variables are inappropriately adjusted. We evaluated the real-world comparative effectiveness of initial systemic versus intratympanic corticosteroids using a prespecified causal inference framework. METHODS: We conducted a retrospective cohort study (2012-2019) of adults with idiopathic SSNHL and complete baseline and discharge audiometry. Patients receiving combined systemic-intratympanic therapy or missing exposure data were excluded. The exposure was the initial corticosteroid route (systemic oral/IV vs. intratympanic). Propensity scores were estimated from baseline covariates including age, sex, symptom profile, comorbidities, baseline PTA, and time from symptom onset to treatment initiation. Primary estimation used augmented inverse probability of treatment weighting (AIPW) to retain all eligible patients; 1:1 propensity-score matching (caliper 0.2 SD of logit PS) with regression adjustment addressed residual imbalance as a sensitivity analysis. Outcomes were PTA gain (primary), complete recovery (≥30 dB gain), and effective improvement (≥15 dB gain). RESULTS: Among 284 adults (systemic = 240; intratympanic = 44), crude mean PTA gain was 13.8 vs. 3.8 dB (difference 10.0 dB). In AIPW analysis, systemic therapy was associated with greater PTA gain (adjusted mean difference 10.1 dB; 95% bootstrap CI 1.8-19.5) and a higher probability of complete recovery (adjusted risk difference 16.9%; 95% CI 9.4-23.4). The risk difference for effective improvement was 11.4% (95% CI - 15.3 to 30.9). Findings were directionally consistent in IPTW, overlap-weighted, and matched analyses. CONCLUSION: In this single-center real-world cohort, initial systemic corticosteroids were associated with greater short-term hearing improvement and a higher likelihood of complete recovery compared with intratympanic therapy. Weighting-based doubly robust estimators improved precision and generalizability, while matched analyses yielded consistent direction but wider uncertainty.