The potential of Oxygen-15 positron emission tomography (PET) to investigate acute brain injury and the effect of therapeutic interventions is limited by inter- and intra-subject variability. We have tried to quantify these limitations when steady state PET is used to measure cerebral blood (CBF), cerebral blood volume (CBV), oxygen metabolism (CMRO2) and oxygen extraction fraction (OEF). We obtained PET images from 10 healthy volunteers and 24 patients within 10 days of head injury with a median (range) Glasgow Coma Score of 7 (3 – 13). Using a region of interest map comprising 15 regions covering the whole brain we calculated inter-subject variability in both groups, while the intra-subject variability was calculated from two consecutive PET studies in a cohort of six patients (test-retest reproducibility). The acquisition of PET data in two frames provided the opportunity of using individual frames to calculate two independent sets of metabolic images, which could be used to assess reproducibility within the context of a single pre-intervention baseline PET study (within-session reproducibility). In this way we obtained repeated measurements of CBF, CMRO2 and OEF, but not CBV, within the same study session in controls and patients under unchanged physiological conditions. We compared assessments of reproducibility obtained using these two approaches: the test-retest reproducibility (two independent studies) and the within-session reproducibility (two independent frames within a single study). Inter-subject coefficients of variation (CoV) for CBF, CBV, CMRO2 and OEF in patients (32.9 +/− 2.2, 15.2 +/− 2.1, 23.2 +/− 2.0 and 22.5 +/− 3.4 %) were generally larger than in controls (13.5 +/− 1.4, 22.5 +/− 2.8, 12.8 +/− 1.1 and 7.3 +/− 1.2 %). The CoV for the test-retest reproducibility measurements in patients were 2.1 +/− 1.5, 3.8 +/− 3.0, 3.7 +/− 3.0 and 4.6 +/− 3.5 %, respectively. The within-session reproducibility CoV for CBF, CMRO2 and OEF in patients (4.2 +/− 4.6, 2.2 +/− 2.0 and 4.8 +/− 4.7 %) were smaller than in controls (8.4 +/− 7.6, 5.3 +/− 3.9, 5.7 +/− 4.4 %). Although large inter-subject variability creates difficulties with studies of small patient groups, estimated sample sizes are moderated by the fact that the changes in cerebrovascular physiology in disease are often dramatic. These figures are irrelevant to the design of interventional studies, where the subject is his or her own control, and the relevant parameter is intra-subject variability or reproducibility. The CoV figures for reproducibility are substantially smaller and provide important reference data. In addition, the within-session and test-retest reproducibility measurements are comparable. This implies that within-session reproducibility measurements can be used to assess the significance of any change in physiology following a therapeutic intervention. These data provide guidance for designing interventional studies, and suggest it should be possible to detect differences of approximately 10 – 15% in controls, and 5 – 10% in head injured patients where systemic physiology is tightly controlled. While each centre should develop its own bank of such data, the figures provided will allow initial generic approximations of sample size for such studies.
