Does surgical treatment improve behaviour in OME?

Background. For older children with OME, the acknowledged presentation includes behaviour problems. However, it has never been specified which aspects of behaviour are most affected, nor whether treatments improving physical health and hearing (efficacy) also improve such indirect outcomes (effectiveness). Methods. We developed a parent‐reported, age‐appropriate, 18‐item measure (BEH‐18) of developmental behaviour problems and showed it to be affected by otitis media with effusion (OME) in a large‐sample case‐control comparison. (SD effect size 0.72, both ns >500) Correlations with hearing level (HL) and ear infection scores in cases showed a maximum with behaviour lagging in time by at least 3 months. The TARGET database offers follow‐up times 6 and 12 months after randomisation so provides a fair test of benefits from ventilation tubes (VTs – grommets). We pooled over grommets‐alone and grommets plus adenoidectomy (+ad), as +ad showed no effects on BEH‐18. We adjusted for baselines and for expectancy and halo biases by analysis of covariance. On top of the usual attendance bias with sample attrition, a social desirability reporting bias in offspring's behaviour and the lagged outcome period led to much missing data, making imputation essential. We partially imputed the two‐occasion outcome for missing data by simple part‐whole regression, raising the used sample from 263 with complete data on all variables to 328 (70 up to 87% of the 376 randomised). Results. An unimputed analysis with the most favourable bias adjustment, admitting some earlier beneficial effects on HL and ear infections into the estimated treatment effect suggested a marginally significant reduction in behaviour problems ( P  = 0.14, 0.15 for 6 and 12 months). However with the same favourable adjustment, the combined and imputed analysis produced a Bias‐adjusted Treatment Effect Size (BaTES) of only 0.10 SD ( P  = 0.423). It is known, and increasingly heeded, that misleading results (in either direction) can be obtained in trials that fail to impute for missing data (when over about 10% of cases). Conclusions. Wider benefits from grommets (beyond those to hearing and ear infections) exist but those are limited and do not include benefits to behaviour, so should not be oversold.