Background <p>Serum amyloid A (SAA) is an acute-phase responsive protein, and its kinetic changes may be related to the immune therapy response. Therefore, we assume that it can predict the outcomes of recurrent or metastatic nasopharyngeal carcinoma (RM-NPC) patients receiving immunotherapy.</p> Methods <p>We conducted a retrospective cohort study to analyze the clinical data of 186 patients with RM-NPC who received PD-1 inhibitor therapy. The patients were divided into three groups: the SAA flare-responder group (SAA increased by at least two times compared to baseline and then decreased below baseline), the SAA responder group (SAA decreased by ≥ 30% within three months without a prior flare), and the remaining patients were SAA non-responders. The primary endpoint was progression-free survival (PFS). Secondary endpoints included the objective response rate (ORR), disease control rate (DCR), and overall survival (OS).</p> Results <p>The median PFS (mPFS) of the entire cohort was 3.52 months, and the median OS (mOS) was 17.02 months. The ORR was 24.73% and DCR was 58.60%. Fourteen (7.53%) patients were classified as SAA flare-responders, 95 (51.08%) patients were defined as SAA responders, and 77 (41.40%) as SAA non-responders. The mPFS of the three SAA kinetic groups was 7.47, 3.97, and 1.87 months (<i>p</i> = 0.0008) and the mOS was 31.13, 26.93, and 15.87 months (<i>p</i> = 0.0283). The ORR of the three groups was 35.71%, 33.68%, and 11.69%, respectively (<i>p</i> = 0.0024). The DCR of the three groups was 85.71%, 71.58%, and 37.66%, respectively (<i>p</i> &lt; 0.0001). It is worth noting that SAA kinetics have no predictive value for RM-NPC patients with low EBV DNA levels.</p> Conclusion <p>SAA kinetics can predict the prognosis of RM-NPC patients treated with PD-1 inhibitors, and can reliably differentiate RM-NPC patients most likely to benefit from immunotherapy.</p>