Pulse Oximetry Versus Arterial Co-Oximetry During Exercise: A Systematic Review of Agreement and Sensor Technology Reporting Gaps

Pulse oximetry is an optical biosensor for peripheral oxygen-saturation monitoring, but its agreement with arterial co-oximetry during exercise remains uncertain. We searched three databases through 30 August 2026 and conducted supplementary Embase searching in September. Sixteen studies yielded 46 device- or condition-level outcomes. Across four peak/maximal cohorts, mean bias ranged from −0.88 to +2.63 percentage points. Three reported paired SDs ranged from 2.75 to 4.60 points; a fourth was conditionally reconstructed. Primary synthesis was structured and descriptive. Post hoc random-effects mean-bias analyses of three reported-SD cohorts and a conditional fourth were imprecise and clinically heterogeneous, and did not characterize individual agreement. A descriptive technology audit found complete device identification in 14 studies and sensor-site reporting in 15, but no study fully reported numerical emitter wavelengths. Averaging/update behavior was fully reported in two studies, algorithm/post-processing detail in three, motion handling in six, and failure/missingness in seven. The author-agreed appraisal was interpreted descriptively without validated aggregate risk or certainty grades. These data do not establish a universal correction factor or a transportable substitution rule. No prospective registration, incomplete supplementary-search export reconciliation and reference-construct ambiguity limit interpretation. More complete sensor, reference-method and agreement reporting is needed.