Bumetanide, a drug being studied in autism spectrum disorder (ASD) may act to restore gamma-aminobutyric acid (GABA) function, which may be modulated by the immune system. However, the interaction between bumetanide and the immune system remains unclear. Seventy-nine children with ASD were analysed from a longitudinal sample for a 3-month treatment of bumetanide. The covariation between symptom improvements and cytokine changes was calculated and validated by sparse canonical correlation analysis. Response patterns to bumetanide were revealed by clustering analysis. Five classifiers were used to test whether including the baseline information of cytokines could improve the prediction of the response patterns using an independent test sample. An immuno-behavioural covariation was identified between symptom improvements in the Childhood Autism Rating Scale (CARS) and the cytokine changes among interferon (IFN)-γ, monokine induced by gamma interferon and IFN-α2. Using this covariation, three groups with distinct response patterns to bumetanide were detected, including the best (21.5%, n = 17; Hedge's g of improvement in CARS = 2.16), the least (22.8%, n = 18; g = 1.02) and the medium (55.7%, n = 44; g = 1.42) responding groups. Including the cytokine levels significantly improved the prediction of the best responding group before treatment (the best area under the curve, AUC = 0.832) compared with the model without the cytokine levels (95% confidence interval of the improvement in AUC was [0.287, 0.319]). Cytokine measurements can help in identifying possible responders to bumetanide in ASD children, suggesting that immune responses may interact with the mechanism of action of bumetanide to enhance the GABA function in ASD.