Diagnostic value of tympanometry using subject-specific normative values

Abstract

Tympanometry has proven valuable in the diagnosis of otitis media in infants. The classification used in the Finnish otitis media studies has resulted in high specificity yet only moderate sensitivity. We aimed to explore the benefit of using each subject's individual previous test results (obtained at previous visits of the children when diagnosed healthy) as normative values. Specific aim was to assess this approach in increasing sensitivity in the diagnosis of otitis media. We used tympanometric (GrasonStadler GSI 38) results of children in the Finnish otitis media vaccine trial obtained at pre-scheduled 7-month visits (if ears found healthy) and at the first subsequent sick visit with respiratory infections and potential otitis media within a 90-day period after the 7-month visit. Children with recent otitis media within the previous month and tympanometric tests with documented interfering factors were excluded. Pneumatic otoscopy was used as the reference method for the diagnosis of otitis media. Altogether 2476 children at the 7-month visits were examined; after exclusions data on 1759 children were available for analysis. Of these children, 630 were brought to at least one sick visit defined above. Finally, data on 906 ears were available for analysis. Entirely flat type B tympanograms predicted otitis media accurately; also high-peaked tympanograms were accurate in predicting normal middle ear status. For the moderate-peaked curves in-between, absolute and relative changes in the tympanometric variables from the 7-month visit to the sick visit were explored for accurate prediction of otitis media; decrease in curve height by at least 2/3 was found most valuable. Sensitivity of the type B tympanogram was 61% and specificity 99% at the sick visit. The sensitivity increased to 67% with a specificity of 98%, if results with the ratio of static acoustic admittance (height of the curve) at the sick visit compared to the previous healthy visit <0.35 were also considered abnormal together with type B curve in indicating otitis media. The subject-specific normative values were not found especially useful, although a slight increase in sensitivity was achieved without compromizing specificity.