The prevalence of antimicrobial drug resistance is now so high that all patients infected with Helicobacter pylori should be considered as having resistant infections. Ideally, therapy should be based on pretreatment antibiotic-susceptibility testing but this strategy is not currently practical. At present, clarithromycin-containing triple therapies do not reliably produce a > or =80% cure rate on an intention-to-treat basis and are, therefore, no longer acceptable as empiric therapy. In this Review, we discuss concepts of resistance that have become part of mainstream thinking for other infectious diseases but have not yet become so with regard to H. pylori. We also put data on the pharmacokinetics and pharmacodynamics of the drugs used in H. pylori therapy and the effect of host cytochrome P450 genotypes in context with treatment outcomes. Our primary focus is to address the problem of H. pylori resistance from a novel perspective, which also attempts to anticipate the direction that research will need to take to provide clinicians with reliable approaches to this serious infection. We also discuss current therapies that provide acceptable cure rates when used empirically (i.e. sequential therapy; four-drug, three-antibiotic, non-bismuth-containing 'concomitant' therapy; and bismuth-containing quadruple therapy) and how they might be further improved.