OBJECTIVE: To analyze the effect of clinical subepileptic G-like discharges on sleep architecture and provide objective indications for clinical needs. METHODS: A full-night EEG polysomnography study was performed in 44 patients with benign childhood epilepsy G with central temporal spike waves (BECTS), and the data were processed by applying unconditional logistic regression analysis. The distribution of spike wave index in the sleep cycle of BECTS patients was II > I > III/IV > REM; the proportion of REM phase was shortened, latency was prolonged, the proportion of I sleep was increased, and the proportion of III/IV sleep was decreased in BECTS patients; the frequency of clinical sub-G discharges was positively correlated with the occurrence of sleep structure disorders; frequent clinical sub-G discharges (>10 discharges/min) were the main risk factor for sleep structure disorders. Frequent clinical subclinical G discharges (>10 discharges/min) were the main risk factor for sleep architecture disorders. Conclusion: Frequent clinical subclinical G-like discharges can cause sleep structure disorders and are an indicator for the need to administer antiepileptic G drugs. DISCUSSION: There are many factors that may influence the accuracy of monitoring results in PSG research work. The first-night effect and the effect of antiepileptic G drugs on sleep cannot be ignored. Although the regular dosage of sodium valproate (available for BECTS treatment, high doses have a stabilizing effect on sleep) in our study had no effect on the composition of sleep, we took care to minimize the effect of confounding factors on an equally comparable basis. The distribution of G discharges during the sleep cycle under clinical conditions: epileptic G activity is prone to be generalized and epileptic G-like discharges are more pronounced during the NREM period due to the dominant role of brain synchronization mechanisms, while brain desynchronization mechanisms are dominant during the REM and waking periods and therefore have an inhibitory effect on epileptic G activity. In childhood epilepsy, epileptic G activity during sleep is mainly regulated by the sleep spindle evoked mechanism. During stages I and II of NREM sleep, the reduction of afferents to the brainstem reticular formation causes the cell membrane surface potential to fluctuate in the frequency range of the sleep spindle; during stages III and IV, with the hyperpolarization of thalamocortical cells, the sleep spindle-like activity is gradually replaced by slow waves in the δ-wave range. Different previous studies have made different conclusions about the distribution of G discharges during sleep in each time phase. clemens et al. analyzed the frequency and morphology of spike waves in BECTS patients and found that the greatest spike wave density was found in phases III and IV, followed by phases I and II, REM and wakefulness, with the highest number of spikes in the first sleep cycle. nobili et al. did not find NREM in the first sleep cycle Nobili et al. did not find a statistical difference in spike indices in each time phase of the first sleep cycle, but the overall NREM phase epileptic G activity was stronger than that in the REM phase. We speculate that the different views of epileptic G activity across time phases in the NREM period may be related to differences in lead design of monitoring instruments and study methods. We found in sleep polysomnography monitoring that: patients with BECTS often have sleep fragmentation, especially when SED is frequent, and in severe cases it can cause absence in the REM phase, and even the temporal phases of the sleep phases cannot be accurately identified; the frequency of G discharges varies significantly in different sleep cycles. Considering the irregularity of the time and number of sleep cycles of patients, we chose the NREM and REM phases of the first and second sleep cycles during monitoring to compare the distribution of G discharges. The results showed that the distribution of SWI in the two sleep cycles was consistent, both being stage II > stage I > stage III/IV > REM stage, with P=0.000 by multiple correlation sample rank sum test.