Risk of Aminoglycoside-Induced Hearing Loss among Patients with Drug-Resistant Tuberculosis in South Africa
Embargo until
2019-12-01
Date
2018-11-27
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
Problem Statement
Individuals treated for multidrug-resistant tuberculosis (MDR-TB) with aminoglycosides (AGs) in resource-limited settings often experience permanent hearing loss, but there is no practical, cost-effective means to identify those at higher risk. This dissertation aimed to estimate the risk of AG-induced hearing loss for MDR-TB-infected individuals in South Africa.
Methods
We nested this analysis within a cluster randomized trial of nurse-led case management in 10 South African TB hospitals. All participants ≥13 years old received kanamycin or amikacin. Hearing loss was defined as a poorer hearing threshold compared to baseline clinical and audiometric evaluation. We developed the prediction model using data from 265 patients at hearing frequencies from 250 to 8,000Hz and validated the model using data from 114 separate patients at both normal (250-8,000Hz) and ultrahigh frequencies (9,000-16,000Hz). We estimated standardized weekly AG exposure as:
{prescribed daily AG dose (mg) x frequency of dosing per week} ÷ weight (kg)
Cox proportional hazard and logistic regression were used for multivariable adjustment.
Results
Of 936 participants, 54% were male; mean age was 36 years; 75% were HIV coinfected at baseline. Comparing patients with high (≥75mg/kg/week) versus low (<75mg/kg/week) AG exposure, the adjusted hazard (aHR) of regimen cessation due to ototoxicity was 1.33 (p=0.006); aHR for audiometric hearing loss was 1.34 (p=.038). Pre-existing hearing loss (aHR=1.71, p<.001) and age (aHR=1.02, p=.031) were also associated with increased hazard of hearing loss. Predictors of ototoxicity in the final prediction model included: standardized weekly AG exposure, HIV status, CD4 count, age, serum albumin, BMI, and pre-existing hearing loss. This model demonstrated moderate discrimination (AUC=0.72) and good calibration (χ2[8]=6.10, p=.64) at normal frequencies and better discrimination (AUC=0.81) at ultrahigh frequencies that might represent early manifestations of AG ototoxicity. Discrimination for AG regimen cessation due to ototoxicity (among 671 patients without baseline audiometric data) was weaker (AUC=0.60). Using a cutoff of 85% predicted probability of hearing loss, the positive predictive value was 100% and the negative predictive value was 41%.
Conclusions
This model identifies patients at high risk for AG-induced hearing loss and may inform clinical guidelines regarding which patients to prioritize for injectable-free regimens.
Description
Keywords
drug-resistant tuberculosis, aminoglycoside, hearing loss