Background Oxygen transport in the blood depends not only upon oxygen partial pressure (or saturation, SaO₂), but also the concentration of haemoglobin to which oxygen binds, and blood volume in any given period. In current clinical studies and guidelines, it is rare for these components to be discussed together.

Methods An observational cohort study was performed in 165 consecutive unselected patients with pulmonary arteriovenous malformations, including 98 before and after embolisation. 159 (96%) had hereditary haemorrhagic telangiectasia (HHT). Prior to statistical analyses, patients were stratified by self-reported exercise tolerance, and serum iron/ferritin.

Results Patients were aged 17–87 (median 49)ys, with median SaO₂ 95% (range 78·5–99%) Many had biochemical/haematinic evidence of iron deficiency, predominantly attributable to HHT-blood loss, but several had undergone venesection for secondary polycythaemia, as recommended in current UK haematological guidelines. Haemoglobin ranged from 7.7–20.9g/dl (median 14.1g/dl), and was higher in patients with lower SaO₂: For every 1% fall in SaO₂, haemoglobin rose on average, by 0·82g/dl (p < 0·0001). SaO₂ explained only 0·1% of the variance in oxygen carriage per unit blood volume. Embolisation treatment increased SaO₂ (median + 2·5%), but usual subsequent haemoglobin falls restored oxygen carriage/unit volume. Concurrent iron deficiency reduced erythrocyte haemoglobin content. Secondary erythrocytosis was achievable with low ferritin, but only if serum iron was high-normal. Oxygen carriage was lower if iron deficiency was present (median 16·1 [IQR 14·7, 18·1]mls/dL compared to 19·0 [IQR 17·4, 20·2]mls/dL, p < 0·0001), implying ~15% higher cardiac outputs were required to maintain oxygen delivery. At rest, this was not met by a higher pulse in the iron deficiency group, indicating stroke volume would need be higher. Self reported exercise tolerance was used as a crude indicator of whether compensating for reduced oxygen carriage was achievable in this patient group. There was a very clear inverse correlation between exercise grade and oxygen carriage such that exercise tolerance was worse in patients with lower oxygen carriage (coefficient -0.0012 (95% CI -0.0017, -0.00061) p < 0.0001). The trend was evident across all grades of exercise tolerance.

Conclusions Low oxygen carriage per unit volume, and not low PaO₂/SaO₂ per se, are predictors of impaired exercise tolerance in PAVM patients.