Objectives: To determine whether GPs should advise their older patients with chronic knee pain to use topical or oral non-steroidal anti-inflammatory drugs (NSAIDs). Design: An equivalence study was designed to compare the effect of advice to use preferentially oral or topical ibuprofen (an NSAID) on knee pain and disability, NSAID-related adverse effects and NHS/societal costs, using a randomised controlled trial (RCT) and a patient preference study (PPS). Reasons for patient preferences for topical or oral preparations, and attitudes to adverse effects, were explored in a qualitative study. Setting: Twenty-six general practices in the UK. Participants: Participants comprised 585 people with knee pain, aged 50 years or over; 44% were male, mean age 64 years. The RCT had 282 participants: 144 in the oral group and 138 in the topical group. The PPS had 303 participants: 79 in the oral group and 224 in the topical group. Interventions: Advice to use preferentially oral or topical NSAIDs for knee pain. Outcome measures: The primary outcome measure was the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). Secondary outcome measures were the Short Form with 36 Items (SF-36), perceived troublesomeness of knee pain, satisfaction with health status, major adverse effects (unplanned hospital admissions and deaths) and minor adverse events over 12 months. The health economic analysis measured the comparative cost per quality-adjusted life-year (QALY) from both an NHS and a societal perspective over 1 and 2 years. Results: Changes in the global WOMAC score at 12-months were equivalent in both studies: topical - oral, RCT difference = 2 [95% confidence interval (CI) -2 to 6], PPS difference = 1 (95% CI -4 to 6). There were no differences in the secondary outcomes, except for a suggestion, in the RCT, that those in the topical group were more likely to have more severe overall pain and disability as measured by the chronic pain grade, and more likely to report changing treatment because of inadequate pain relief. There were no differences in the rate of major adverse effects but some differences in the number of minor ones. In the RCT, 17% and 10% in the oral and the topical group, respectively, had a defined respiratory adverse effect (95% CI of difference -17% to -2.0%); after 12 months, the change in serum creatinine was 3.7 mmol/l (95% CI 0.9 to 6.5) less favourable in the oral than in the topical group, and 11% of those in the oral group reported changing treatment because of adverse effects compared with 1% in the topical group (p = 0.02). None of these differences were seen in the PPS. Oral NSAIDs cost the NHS £191 and £72 more per participant over 1 year in the RCT and PPS respectively. In the RCT the cost per QALY in the oral group, from an NHS perspective, was in the range £9000-12,000. In the PPS it was £2564 over 1 year, but over 2 years the oral route was more cost-effective. Patient preference for medication type was affected by previous experience of medication (including adverse reactions), other illness, pain elsewhere, anecdotes, convenience, severity of pain and perceived degree of degeneration. Lack of understanding about knee pain and the action of medication led to increased tolerance of symptoms. Potentially important symptoms may inadvertently have been disregarded, increasing participants' risk of suffering a major adverse effect. Conclusions: Advice to use either oral or topical preparations has an equivalent effect on knee pain, but oral NSAIDs appear to produce more minor adverse effects than topical NSAIDs. Generally, these results support advising older people with knee pain to use topical rather than oral NSAIDS. However, for patients who prefer oral NSAID preparations rather than a topical NSAID, particularly those with more widespread or severe pain, the oral route is a reasonable treatment option, provided that patients are aware of the risks of potentially serious adverse effects from oral medication. Further research is needed into strategies to change prescribing behaviour and ensure that older patients are aware of the potential risks and benefits of using NSAIDs. Observational studies are needed to estimate rates of different predefined minor adverse effects associated with the use of oral NSAIDs in older people as are long-term studies of topical NSAIDs in those for whom oral NSAIDs are not appropriate.