Measuring reproducibility of regional brain metabolic responses to lorazepam using statistical parametric maps

Statistical parametric mapping (SPM) is a method for localizing differences in brain activation patterns without the need for anatomic predefined constraints. The purpose of this study was to assess the reproducibility of the patterns of activation obtained with SPM for baseline measures and for metabolic changes in response to lorazepam on a test-retest design. The results were compared with those we previously published using region-of-interest (ROI) methods.

Sixteen healthy right-handed men were scanned twice with PET and [18F]fluorodeoxyglucose (FDG): before placebo and before lorazepam (30 microg/kg). The same double FDG procedure was repeated 6-8 wk later to assess test-retest reproducibility. Image datasets were analyzed by using SPM95 software. Difference images between baseline and lorazepam were compared for the first and second evaluations, both for relative decreases as well as increases in metabolism. Significance level was systematically varied to P < 0.001, P < 0.01 and P < 0.05.

There were no differences in the baseline SPM maps obtained for the first and second evaluations. SPM showed similar, although not identical, differences in response to lorazepam between the two evaluations. Both evaluations showed significant decreases in occipital cortex (9.7% and 10%) and significant relative increases in left temporal pole (6.8% and 10.4%). However, the second evaluation showed a decrease in the left frontal cortex (areas 6 and 8), which was not present in the first evaluation. The results were very similar to those we had obtained with ROI methods, except for the activation in the left temporal pole, which we had not observed with ROI analyses.

Although the overall pattern of lorazepam-induced activation depicted by SPM was reproducible in pattern and magnitude, there were some differences that included a left frontal area of deactivation during the second but not the first evaluation. Results with SPM are similar to those with the ROI method, and, because it systematically analyses the whole brain, SPM can uncover patterns not seen with the ROI method.