OBJECTIVES Rheumatoid arthritis is an autoimmune disorder of unknown aetiology. Morning stiffness, a characteristic feature of rheumatoid arthritis, shows a 24-h rhythm. Noticing this rhythm, we hypothesized the presence of a similar rhythm for a rheumatoid arthritis indicator, in addition to dosing-time dependency of the anti-rheumatic effect of methotrexate in arthritis induced by collagen in rats and mice, which reflect the symptomatology of rheumatoid arthritis patients. METHODS To measure tumour necrosis factor (TNF)-alpha concentration, blood was taken at different times (2, 6, 10, 14, 18 or 22 h after the light was turned on (HALO)) in collagen-induced arthritic mice. Methotrexate was administered at two different dosing times based on these findings to estimate arthritis. KEY FINDINGS The arthritis score was significantly lower in the 22 HALO-treated group than in the control and 10 HALO-treated groups in collagen-induced arthritic rats and mice. Plasma TNF-alpha concentrations showed obvious 24-h rhythms, with higher levels at light phase and lower levels at dark phase after rheumatoid arthritis crisis. Arthritis was relieved after administration of methotrexate during the dark phase in synchronization with the 24-h rhythm. CONCLUSIONS Our findings suggest that choosing an optimal dosing time associated with the 24-h cycling of TNF-alpha could lead to effective treatment of rheumatoid arthritis by methotrexate.