Objective: To provide a qualitative view and quantitative measure of sleep disturbances across and between early stages – clinical ultra high-risk and first episode – of psychotic and bipolar disorders. Methods: Electronic databases (PubMed, Cochrane, Embase, PsychINFO) were searched up to March 2021 for studies comparing sleep measures between individuals with an early stage and controls. Standard mean deviations (Cohen’s d effect sizes) were calculated for all comparisons and pooled with random-effects models. Chi-square tests were used for direct between-subgroups (ultra high-risk vs first episode) comparisons of standard mean deviations. The effects of age, sex ratio, symptoms and treatment were examined in meta-regression analyses. Results: A database search identified 13 studies that contrasted sleep measures between individuals with an early stage ( N = 537) and controls ( N = 360). We observed poorer subjective sleep quality (standard mean deviation = 1.32; 95% confidence interval, [1.01, 1.62]), shorter total sleep time (standard mean deviation =−0.44; 95% confidence interval, [−0.67, −0.21]), lower sleep efficiency (standard mean deviation = −0.72; 95% confidence interval, [−1.08, −0.36]), longer sleep onset latency (standard mean deviation = 0.75; 95% confidence interval, [0.45, 1.06]) and longer duration of wake after sleep onset (standard mean deviation = 0.49; 95% confidence interval, [0.21, 0.77]) were observed in early stages compared to controls. No significant differences were observed for any of the reported electroencephalographic parameters of sleep architecture. No significant between-subgroups differences were observed. Meta-regressions revealed a significant effect of the age and the antipsychotic status on subjective measures of sleep. Conclusion: The early stage population presents with significant impairments of subjective sleep quality continuity, duration and initiation. Systematic assessments of sleep in early intervention settings may allow early identification and treatment of sleep disturbances in this population.