Objective: To determine ventilation tube (VT) patency, or presence of an eardrum perforation, where otoscopy is not available or corroboration is required, tympanometry may be used, with a firm acoustical rationale. However, published literature shows little evidence that tympanometric criteria for patency have been optimized or are in routine clinical use. A randomized trial of VTs in otitis media with effusion offered large case numbers, assessed by uniform protocol, to define and evaluate a tympanometric patency criterion. Design: Children had been randomized to one of three interventions (VT insertion with or without adjuvant adenoidectomy and nonsurgical observation). The study examines 165 left and 171 right ears with functioning VTs and 103 left and 102 right noninserted ears, in children aged 3 3/4 to 7 1/4 yrs at first postintervention visit. Experienced otolaryngologists judged VT patency otoscopically. Measured equivalent ear-canal volume (Veq) was compared across the VT-inserted and the not-inserted groups, and also within the VT-inserted group between the pre- and postintervention visits. With otoscopy as reference, patency coding errors in both VT-inserted and not-inserted groups were analyzed as a function of Veq. Three methods of determining optimum cutoff were considered: equal error (cross-over point of the errors in each group), equated "cost" of error (maximum summed sensitivity and specificity), and minimum combined error (determined from the minimum of a polynomial fitted to the mean of the errors in each group). Cutoffs were evaluated in terms of classification accuracy against otoscopy after intervention. Results: The between-group comparison gave cutoffs by the three methods at Veq ≥0.95 mL, Veq ≥1.10 mL, and Veq ≥1.33 mL, respectively, pooled to 1.13 mL. The same 1.13 mL was also found for the within-group comparison. The corresponding cutoffs for the pre- and postintervention difference in Veq occurred at 0.23, 0.44, and 0.39 mL, respectively, with a mean of 0.35 mL. Within the range studied, age did not influence Veq, nor the optimum Veq cutoff, but boys had significantly larger Veqs (by 0.09 mL) than girls. Raw Veq, pre- and postinterven postintervention Veq difference or a combination of these definitions differed little in accuracy of predicting otoscopic status. Conclusion: In 3 3/4 to 7 1/4 years olds, the Veq criteria for VT patency, based on rigorous and transparent derivations, offer a supplementary information source for clinical practice, as well as a sole objective marker in research. We recommend for general use the pooled mean cutoff at Veq ≥1.13 mL, slightly higher than the 1.00 mL given by Shanks et al. The pooled pre- and postintervention difference Veq criterion was slightly lower than the 0.4 mL of Shanks et al. However, the extra effort in its use, alone or in combination with raw Veq, was not justified. Given the wide (flat-bottomed) error functions, users have the option of declaring slightly lower or higher cutoffs, reflecting differing "costs" on the two types of errors, e.g. penalizing false "patent" decisions more heavily.
Bibliographical notePublisher Copyright:
Copyright © 2008 by Lippincott Williams & Wilkins.