Rheumatoid arthritis (RA) is an incurable chronic disorder that may induce autoinflammation and serious pain in the joints. Early diagnosis and treatment are important for RA prognosis. However, there is a lack of effective and objective diagnostic approaches. Levels of several immunity cytokines were found to change for patients with early RA, including IL-6, TNF-α, and IL-17 in serum. We assumed a combined change of these cytokines could predict early RA, and a total of 37 outpatients were found. After these patients with early symptoms had been followed for more than one year, 32 clinical cases of RA were diagnosed. The accuracy rate of the current method is >86%. We assumed the symptom relief could be achieved by regulating these cytokines and serum lipid-associated indicators. Thereafter, (R)-dihydrolipoic acid (R-DHLA)-stabilized gold nanoclusters (AuNCs) without (R-DHLA-AuNCs) and with cerium modification (R-DHLA-AuNCs-Ce) were employed for treatment of the RA rat model and . R-DHLA-AuNCs-Ce exhibited extraordinary reactive oxygen species-scavenging and anti-inflammation effects by regulating macrophage polarization, which was found to be more effective than methotrexate. The inflammation response of the joint microenvironment was also reduced with an exciting efficiency. By complex analysis of the pro-inflammatory cytokines and activity period indicators and , we concluded that macrophage-mediated inflammation exacerbated autoimmunity, which fully relieved the symptoms after administration of R-DHLA-AuNCs-Ce to RA rats.