The objectives of this work were to assess inter- and intraobserver variability of different staging systems for tympanic membrane (TM) retraction using otoendoscopy in children at risk of retraction from cleft palate, to compare hearing level with stage of retraction, and to propose optimum characteristics for monitoring TM retraction with endoscopy.
Endoscopic images of 245 TMs of children with cleft palate (mean age, 13.0 years) were assessed on two separate occasions by six observers using the Sade and Erasmus staging systems for pars tensa retraction and Tos system for pars flaccida retraction. Intra- and interobserver agreements were calculated. Extent of TM retraction was compared with hearing threshold. TMs with middle ear effusion, tympanostomy tubes, or perforation were excluded.
A total of 108 ear drums (44%) were rated as having pars tensa and/or flaccida retraction. Intraobserver agreement was fair to moderate (kappa = 0.3–0.37, P < .001) for the different staging systems and interobserver agreement slight to moderate (0.18–0.41 P < .001). Conductive hearing loss (four-tone average air-bone gap >25 dB HL) was present in 11 ears (15%). No correlation between hearing threshold and retraction stage was found. Isolated tensa retraction onto the promontory increased hearing threshold more than retraction involving the incus (P = .02; analysis of variance).
Endoscopic image capture may provide a clear objective record of TM retraction, but current staging systems have unsatisfactory reliability when applied to such images, and retraction stage correlates poorly with hearing threshold. Modification of retraction assessment to improve validity and clinical relevance is proposed.