Acute otitis media (AOM) has become increasingly difficult to treat in the 1990s, the decade of drug-resistant pneumococcus. Throughout the world, drug-resistant strains of this pathogen are being recovered from 20 to 50% of cases of initial untreated AOM, and from 45 to 90% of refractory AOM. Almost as alarming is that beta-lactamase-producing strains of Haemophilus influenzae are currently being isolated in 40 to 50% of cases of AOM in the US. Clinicians can no longer expect 'Pollyanna-like' high rates of clinical resolution for this disease. It is now imperative that they become aware of the regional prevalence of these drug-resistant bacteria and, just as importantly, their patterns of antibacterial resistance. Although some authors would hold that any antibacterial, or even placebo, should be adequate for most cases of AOM, clinical practice appears to suggest otherwise. Amoxicillin, still the first-line therapeutic choice for initial nonrefractory AOM, will often fail. The real dilemma begins when clinicians search for clinical data to select an antibacterial for therapeutic failures--few data are available. Thus, to give optimal treatment to a child who has failed antibacterial therapy--the true actual indication for all second-line antibacterials--they must instead become familiar with the following in vivo and in vitro data: 1. 'In vivo sensitivity data': otherwise known as bacteriological efficacy, in which repeat tympanocentesis is performed in mid-therapy. This reveals the bacterial 'Achilles heel' or weakness for the individual antibacterial agents. 2. Clinical efficacy data: analysis of rates of clinical resolution after therapy in comparative trials which use a single tympanocentesis initially and a 'gold standard' comparator antibacterial. 3. 'Bug to drug' data: comparison of reported middle ear concentrations for each individual antibacterial agent relative to the respective minimum inhibitory concentrations of isolates, particularly drug-resistant pneumococcus and H. influenzae (if possible, obtained from the paediatric respiratory tract). The selection of an antibacterial agent for AOM in any particular case should not be merely a random process. It involves awareness of the pathogens most likely to be observed: with co-infections; after failure with a particular antibacterial (the bacterial 'Achilles heel' of the drug); and at different points in time, whether initially or after therapeutic failures (e.g. first-line versus fourth-line failure).