The object of this study was to compare the effect on the hearing of the insertion of a grommet, with or without adenoidectomy, against a non-surgically managed control ear in children with persistent hearing impairment due to otitis media with effusion. Seventy-eight children (44 boys, 34 girls, mean age 5.8 years) with documented bilateral otitis media with effusion associated with a bilateral hearing impairment (pure tone average air conduction thresholds over 0.5, 1 and 2 kHz of > or = 25 dB HL) over a three month period were admitted to a randomized, controlled trial. Each child was randomized to have or not to have an adenoidectomy. The ears in each child were then randomly allocated to have a grommet (tympanostomy tube) inserted. The children's hearing status was reviewed six and 12 months post-operatively. During follow-up, should a child redevelop a persistent bilateral hearing impairment (as defined above) for three months they were managed with a hearing aid. Thus no child had repeat insertion of a grommet. Surgery of each type had an effect on the hearing and the presence of otitis media with effusion at six months post-operatively but not at 12 months when it was no different from natural resolution. If resolution of the otitis media with effusion is the outcome measure, then adenoidectomy alone is significantly better than no surgery but only in boys rather than in girls. Even in boys it only resolves about 60 per cent of effusions. However, when combined with a grommet (one insertion) adenoidectomy gives no greater resolution (89 per cent compared with 86 per cent).(ABSTRACT TRUNCATED AT 250 WORDS)