A Predictive Model of Bilateral Sensorineural Hearing Loss in Meniere Disease Using Clinical Data

Abstract

Objectives: Meniere disease (MD) is defined by a clinical syndrome of recurrent attacks of spontaneous vertigo associated with tinnitus, aural fullness, and sensorineural hearing loss (SNHL). Most patients have unilateral SNHL, but some of them will develop contralateral SNHL during the course of the disease. Several studies have reported a frequency of 2 to 73% SNHL in the second ear, according to the duration of disease and the period of follow-up. We hypothesize that unilateral and bilateral MD are different conditions, the first would initially involve the apical turn of the cochlea, while bilateral MD would affect the entire length of the cochlea. The aim of the study is to search for clinical predictors of bilateral SNHL in MD to build a predictive model of bilateral involvement.

Design: A retrospective, longitudinal study including two cohorts with a total of 400 patients with definite MD was carried out. The inception cohort consisted of 150 patients with MD and the validation cohort included 250 cases. All of the cases were diagnosed of unilateral MD according to their hearing loss thresholds. The following variables were assessed as predictors of bilateral SNHL for the two cohorts: sex, age of onset, familiar history of MD, migraine and high-frequency hearing loss (HFHL, defined if hearing threshold >20 dB in two or more consecutive frequencies from 2 to 8 KHz). A descriptive analysis was carried out according to the presence of HFHL in the first audiogram for the main variables. By using multiple logistic regression, we built-up several predictive models for the inception cohort and validated it with the replication cohort and merged dataset.

Results: Twenty-three (19.3%) and 78 (41%) of patients with HFHL developed contralateral SNHL during the follow-up, in the inception and validation cohorts, respectively. In the inception cohort, the best predictive model included HFHL in the first audiogram (OR = 6.985, p = 0.063) and the absence of migraine (OR = 0.215, p = 0.144) as clinical predictors for bilateral SNHL [area under the curve (AUC) = 0.641, p = 0.002]. The model was validated in the second cohort (AUC = 0.621, p < 0.001). Finally, we merged both datasets to improve the precision of the model including HFHL in the first audiogram (OR = 3.168, p = 0.001), migraine (OR = 0.482, p = 0.036) and age of onset >35 years old (OR = 2.422, p = 0.006) as clinical predictors (AUC = 0.639, p < 0.001).

Conclusions: A predictive model including the age of onset, HFHL in the first audiogram and migraine can help to assess the risk of bilateral SNHL in MD. This model may have significant implications for clinical management of patients with MD.