The prescription of long-term oxygen (LTOT) is underpinned by the measurement of arterial PO2, generally obtained by radial artery puncture. This test is commonly associated with patient discomfort and a test that is reliable, well-tolerated and non-invasive would be advantageous. Cutaneous oximetry has not proved sufficiently accurate. Arterialized earlobe capillary sampling has been proposed, with some authors stating that it is under-utilized. However, to date studies have yielded conflicting results and the clinical utility remains uncertain. Our regional oxygen service based at a specialist respiratory hospital undertook a prospective study of consecutive patients with chronic respiratory disease undergoing assessment for LTOT. Simultaneous radial artery and arterialized earlobe sampling was performed. Rigorous steps were taken to ensure optimal arterialization of the earlobe samples. Agreement between arterial and arterialized PO2 and PCO2 was compared using the Bland-Altman method. One hundred patients were studied. Procedural difficulties (insufficient sample or air in sample) were similar for both procedures, however clotting occurred more frequently in arterialized earlobe samples. Sixty-four sample pairs were available for comparison. The bias and limits of agreement between arterialized and arterial PO2 were wide, mean (+/- 2 SD), -048 (-2.05-1.09) kPa. The bias and limits of agreement for PCO2 were smaller. Using the absolute criterion (arterial PO2 < 7.3 kPa), 9/55 (16%) patients would receive oxygen inappropriately based on the arterialized earlobe sample. Conversely, no patients would have been denied LTOT. Radial artery puncture gave rise to significantly greater discomfort (P < 0.0001) and level of concern (P < 0.0001). Patient preference strongly favoured arterialized earlobe sampling. However, despite rigorous attention to arterialization earlobe sampling was insufficiently accurate to replace radial artery puncture in the prescription of LTOT.