Skip navigation

Randomised versus nonrandomised OME children


Clinical bottom line

The results obtained using grommets (ventilation tubes) in children with otitis media with effusion were the same whether they were randomised into the trial, or not. In this instance randomised trial results appear to be generalisable to the population.

One of those big questions that keeps coming up is whether randomised trial results can be used to treat patients as a whole. The argument is that trials exclude so many patients that their results are not generalisable to all the patients with that complain, because, for instance, most will have some comorbid condition.

While that can sometimes be the case, it is increasingly common to see randomised trial which are more inclusive, and with exclusions that make pretty good sense because of drug interactions, or because of frank dangers of treating some patients with a particular condition with the intervention under test.

That still leaves a nagging doubt about whether the patient we are treating are like those in the trial. Perhaps people not consenting to a trial are just, well, different! An investigation into the use of grommets for glue ear looked at just that point.


MM Rovers et al. Generalizability of trial results based on randomized versus nonrandomized allocation of OME infants to ventilation tubes or watchful waiting. J Clin Epidemiol 2001 54: 789-794.


This was a randomised trial comparing grommet insertion with watchful waiting in Holland. A routine screening test at nine months of age was followed by rescreening one month later if unsatisfactory. Failure of three successive tests resulted in referral to ENT for assessment. This occurred in 1081 of 30,099 cases.

Persistent (4-6 months) bilateral otitis media with effusion (glue ear) confirmed by tympanometry and otoscopy were invited to participate in a randomised trial. There were 386 at this stage. randomisation was to be between grommet insertion, or watchful waiting.

Parents of 199 children did not agree to randomisation and 66 disagreed with any form of follow up. The remaining 133 were treated as agreed with the ENT surgeon, and consented to a full year of follow up. In addition 187 children were randomised.


The main results are shown in Table 1. Follow up at 12 months was less exhaustive in the nonrandomised children, but results were identical for the children randomised and those not randomised. Children who were not randomised differed from those who were randomised by being more likely to have an older sibling, to be less likely to be attending day care, and to be less likely to have mothers with a low educational level.

Table 1: Results for randomised and nonrandomised children with grommets of watchful waiting over one year





Watchful waiting


Watchful waiting






12 month follow up (%)





Grommets inserted (%)



Mean time with effusion (%, 95% CI)





Mean improvement at 12 months (dB)






This is an interesting paper that investigates one of those dark corners of evidence, that of generalising results of trials. It is easy to carp. It's much more difficult to do useful work. This is one such, and it is another example of where the results of randomised trials are generalisable.