This study examines the impact of health education on prevalence of and expenditure on treatment of self-reported arthritis-related illness among elderly persons in rural Bangladesh. An intervention study was conducted, including 1135 elderly persons (>or=60 years) from eight randomly selected villages, four each of an intervention and a control area. The analyses include 839 elderly persons who participated in both pre- and post-intervention surveys (intervention area: n = 425, control area: n = 414). Participants of the intervention area were further categorized as compliant (n = 315) and non-compliant (n = 110) based on adherence to the intervention instructions. The intervention that lasted for 15 months comprised home-based physical exercise, dietary instructions and other aspects of management. Results show that although there was no significant difference in self-reported arthritis-related illness between the compliant and non-compliant groups at baseline, it was significantly lower in the compliant group (71%) at post-intervention compared with the non-compliant (81%). Related monthly expenditure on treatment was significantly reduced in the compliant group (from Taka 104 to Taka 52) but not in the other two groups. Logistic regressions further showed that the control group had a higher probability of increased treatment-related expenditure compared with the compliant group (OR 2.0, 95% CI 1.4-2.8).