Obstructive sleep apnoea (OSA) in children is commonly considered to occur predominantly in rapid eye movement (REM) sleep, but clinical experience suggests that this is not universally the case. We hypothesized that there would be a subgroup of children with OSA who have non-REM (NREM) predominance of obstructive events and that these children share certain clinical characteristics. Thus, we aimed to compare the obstructive apnoea-hypopnoea index (OAHI) in REM versus NREM sleep and to assess factors influencing the distribution of events by sleep state. Polysomnography (PSG) recordings of 102 children aged 0-18 years with moderate to severe OSA (OAHI >or=5 h(-1)) were reviewed. OAHI was calculated separately for REM and NREM sleep. A REM predominance index (RPI) was determined using log transformation [RPI = log (REM OAHI + 0.5) - log (NREM OAHI + 0.5)] and compared with possible influencing factors using multiple linear regression. Analysis showed that obstructive events were more common in REM sleep (median REM OAHI 21.4 h(-1), median NREM OAHI 8.3 h(-1), P < 0.001). Mean RPI was significantly greater than zero (P = 0.003). However, a substantial minority of children (30.4%) had a higher NREM than REM OAHI. The factors that were related significantly to NREM predominance were older age (P = 0.02), higher arousal index (P < 0.001) and higher SpO(2) nadir (P < 0.001). Our findings demonstrate that while OSA is a REM sleep-related problem in the majority of children, there is a significant subset of children with NREM predominance of obstructive events. This finding highlights the importance of considering sleep state distribution of events in studies of the pathophysiology and outcomes of OSA in childhood.