Obstructive sleep apnea (OSA) is a major risk factor for stroke and cognitive decline, although its effects on cerebrovascular reactivity (CVR) remain unclear. However, it remains unknown whether blood-oxygen-level–dependent (BOLD) MRI captures consistent CVR impairment or compensatory changes in OSA.

Here, we systematically synthesize BOLD MRI studies of CVR in adults with OSA to resolve these discrepancies.

We searched five databases through October 2025 for human studies reporting hypercapnia- or breath-hold–evoked BOLD CVR in polysomnography-confirmed OSA, with duplicate screening, JBI risk-of-bias assessment, and random-effects pooling of PETCO₂-referenced %BOLD·mmHg⁻¹ where possible.

Six studies met criteria; four case-control comparisons (n = 84 OSA; n = 55 controls) and two CPAP trials. Contrary to the prevailing view that OSA uniformly blunts CVR, pooled PETCO₂-referenced data revealed higher global CVR in OSA than controls (mean difference +0.05 %BOLD·mmHg⁻¹, 95 % CI 0.03–0.06; P < 0.001), with similarly increased grey and white matter reactivity. Studies using alternative CVR metrics uncovered regionally reduced vasodilatory volume or perfusion-based flow reactivity in OSA, aligning previously unresolved dissociations between BOLD and perfusion-derived CVR. Within OSA, one therapeutic CPAP trial demonstrated regional increases in CVR, whereas short-term CPAP withdrawal did not alter BOLD CVR.

These findings reveal that BOLD CVR in OSA is heterogeneous rather than uniformly reduced and establish PETCO₂-referenced BOLD MRI as a marker of cerebrovascular dysregulation, providing a new framework for CVR standardisation and opening the door to biomarker-based cerebrovascular risk stratification and CPAP response monitoring.