Research has shown efficacy of melatonin treatment to advance sleep-wake rhythms in insomnia. In healthy adults, direction and magnitude of the phase shift depends on the timing of administration relative to the phase position of the circadian system. Therefore, in the present study we investigated whether in children with chronic sleep onset insomnia (SOI) efficacy of melatonin treatment in the early evening could be predicted from dim light melatonin onset (DLMO), a phase marker of the circadian system. We combined data of two previously published double blind, randomized, placebo-controlled trials in 110 participants, aged 6-12 years. Sleep was actigraphically estimated, and saliva collected, at baseline and in the third week of a 4-week treatment period with 5 mg melatonin or placebo at 18:00 or 19:00 hours. Primary outcome measures were pre- to post-treatment changes in dim light melatonin onset (DeltaDLMO), sleep onset (DeltaSO), sleep latency (DeltaSL), and total sleep duration (DeltaTSD). Melatonin advanced DLMO with +1:12 h (P < 0.001), SO with +0:42 h (P = 0.004), SL decreased with 25 min (P = 0.019), and TSD did not change significantly, as compared with placebo. In the melatonin-treated group, but not in the placebo-treated group, pretreatment DLMO was significantly related to DeltaDLMO [F(1, 29) = 7.28, P = 0.012] and DeltaSO [F(1, 25) = 7.72, P = 0.010]. The time interval between treatment administration and pretreatment DLMO (INT) was only significantly related to DeltaSO [F(1,26) = 5.40, P = 0.028]. The results suggest that in children with SOI, the efficacy of early evening melatonin to advance sleep onset and endogenous melatonin onset increases the later the pretreatment DLMO is.