It is current practice to use a single diagnostic sleep study in the diagnostic workup of obstructive sleep apnoea (OSA). However, a relevant night-to-night variability (NtNV) of respiratory events has been reported.We evaluated the NtNV of respiratory events in adults with suspected or already diagnosed OSA who underwent more than one diagnostic sleep study. Data sources were PubMed, Cochrane and Embase up to 23 January 2019. Random-effects models were used for evidence synthesis. For moderator analysis, mixed-effects regression analysis was performed. The study was registered with PROSPERO (CRD42019135277).Of 2143 identified papers, 24 studies, comprising 3250 participants, were included. The mean Apnoea-Hypopnoea Index (AHI) difference between the first and second night was -1.70/hour (95% CI -3.61 to 0.02). REM time differences (first to second night) were significantly positive associated with differences in mean AHI (β coefficient 0.262 (95% CI 0.096 to 0.428). On average, 41% (95% CI 27% to 57%) of all participants showed changes of respiratory events >10/hour from night to night. Furthermore, 49% (95% CI 32% to 65%) of participants changed OSA severity class (severity thresholds at 5/hour, 15/hour and 30/hour) at least once in sequential sleep studies. Depending on the diagnostic threshold (5/hour, 10/hour or 15/hour), on average 12% (95% CI 9% to 15%), 12% (95% CI 8% to 19%) and 10% (95% CI 8% to 13%) of patients would have been missed during the first night due to single night testing.While there was no significant difference between mean AHI in two sequential study nights on a group level, there was a remarkable intraindividual NtNV of respiratory events, leading to misdiagnosis and misclassification of patients with suspected OSA.