Objective To investigate the mechanism by which thioredoxin‑interacting protein (TXNIP) mediates oxidative stress in bupivacaine‑induced spinal neurotoxicity. Methods A total of 48 specific pathogen‑free (SPF) healthy male SD rats, weighing 110−130 g, were selected. According to the random number table method, they were divided into four groups (n=12): a control group (group C), a bupivacaine group (group B), a bupivacaine+adeno‑associated virus (AAV)‑TXNIP short hairpin RNA (shRNA) group (group BT), and a bupivacaine+shRNA group (group BS). Intrathecal catheterization was conducted at the L5‑L6 segments of rats in groups C, B, BT and BS. Rats in group C and group B were intrathecally injected with 5 μl normal saline, while those in group BT and group BS were intrathecally injected with 5 μl of AAV‑TXNIP shRNA (virus titer: 1.01×1012 VG/ml) and 5 μl of the nonsense control sequence shRNA, respectively. Subsequently, the sheath tube was removed. When all the rats were raised to 250−300 g, the L5‑L6 segment was intrathecally catheterized again. Groups B, BT, and BS were intrathecally injected with 5% bupivacaine at 0.12 μl/g for three times, with an interval of 90 min between each session. Group C was injected with the same amount of normal saline at the corresponding time points. After the corresponding treatments, the tail sensation and lower limb motor function of rats were measured by the percent of maximum possible effect (%MPE) converted from tail‑flick latency (TFL) in rats, and the Basso Beatty Bresnahan (BBB) scoring method on the day of intrathecal injection of bupivacaine or an equal amount of normal saline (T0), day 1 after intrathecal injection (T1), day 2 after intrathecal injection (T2), and day 3 after intrathecal injection (T3). The injury of spinal cord tissue was observed by hematoxylin eosin staining (H‑E staining), while Nissl staining was used to observe survival neurons. The contents of malondialdehyde (MDA), superoxide dismutase (SOD), catalase (CAT), and glutathione (GSH) were detected by biochemical methods. The levels of reactive oxygen species (ROS) were detected by immunofluorescence staining. The levels of TXNIP messenger RNA (mRNA) were examined by real‑time fluorescence quantitative polymerase chain reaction (FQ‑PCR). The levels of TXNIP and glutathione peroxidase 4 (GPX4) in spinal cord nerve tissue was detected by Western blot. Results Compared with group C, groups B, BT, and BS showed increases in %MPE and decreases in BBB scores at T0, T1, T2, and T3 (all P<0.05); group B and group BS had an increased number of oligodendrocytes and microglia, edema or atrophy in neuronal cells, and a decreased number of survival neurons (all P<0.05); groups B, BT, and BS showed increases in the contents of MDA (all P<0.05), and decreases in the contents of SOD, GSH, and CAT (all P<0.05); groups B and BS presented increases in the levels of ROS (all P<0.05), and TXNIP mRNA levels (P<0.05); group BT showed decreases in TXNIP mRNA levels (P<0.05); and groups B and BS presented increases in TXNIP protein expression (all P<0.05), and decreases in GPX4 protein expression (all P<0.05). Compared with group B, group BT showed decreases in %MPE, but increases in BBB score at T1, T2, and T3 (all P<0.05), as well as reduced edema and vacuoles in the gray and white matter, with a reduced number of damaged neurons, and an elevated number of survival neurons (P<0.05), decreases in MDA contents (P<0.05), increases in the contents of SOD, GSH and CAT (all P<0.05), decreases in the level of ROS (P<0.05), reductions in TXNIP mRNA and protein levels (all P<0.05), and increases in GPX4 protein expression (P<0.05). Conclusions Bupivacaine induces spinal neurotoxicity by activating TXNIP‑mediated oxidative stress. Knockdown of TXNIP can alleviate bupivacaine‑induced spinal neurotoxicity by inhibiting oxidative stress.