In a 6-month pilot study of 7 patients with rheumatoid arthritis, we tested a treatment involving the extracorporeal photoactivation of biologically inert methoxsalen (8-methoxypsoralen) by ultraviolet A energy to a form that covalently cross-links lymphocyte DNA; the injured cells are reinfused into the patient. Prior experimental studies had indicated that this regimen produces an immune reaction against antigens on treated T cells, and a clinical trial in patients with cutaneous T cell lymphoma had been successful. The current study patients were treated monthly, on 2 successive days (or biweekly, later on). Between 12 and 16 weeks of therapy, there appeared to be a breakpoint, after which the joint counts and joint scores of 4 of the patients began to improve. In 3 of the 4 patients, these measures eventually diminished by a mean of 71% and 80% of baseline values, respectively, and there was variable improvement in less direct indicators of clinical response. The joint counts and scores of the fourth patient improved modestly (by 33% and 59% of baselines, respectively) but he required alternative therapy, and those of the remaining 3 study patients did not improve. Mean slopes for the joint counts were significantly different from zero after the apparent breakpoint (but not before), whether one examined the 4 apparent responders (P = 0.01) or the entire group of 7 patients (P = 0.01). After completion of therapy, there was also a delay, of 2-3 months, in the clinical deterioration of those patients who had improved. There was only 1 mechanical adverse effect--a single episode of transient hypotension in 102 treatment sessions--and no toxic effects.(ABSTRACT TRUNCATED AT 250 WORDS)