The ROSIE clinical trial was the first application of a Bayesian dose escalation design to balance efficacy-toxicity trade offs. The performance of this design was successful and particularly well suited for stroke thrombolytic trials because early efficacy (reperfusion) and toxicity (intracranial hemorrhage) outcomes are attainable. The model performed appropriately in that it escalated the dose when insufficient rates of reperfusion were achieved and de-escalated the dose when toxicity occurred. The model statistically eliminated abciximab with no reteplase as an acceptable dosing regimen because of insufficient efficacy after only 36 patients, whereas approximately 1200 patients were required to reach the same conclusion in Phase II and Phase III trials using conventional trial methodology. Our results demonstrate the value and efficiency of this approach to making the correct decisions about dose finding. From post-hoc exploratory analyses of the ROSIE trial data we have further refined the MRI inclusion criteria and reperfusion response that better approximates a dose-response relationship on reperfusion for a 24 hour window thrombolytic trial. This includes requiring an estimated volume of baseline perfusion defect of at least 5cc, and a definition of clinically meaningful reperfusion response as a 67% or greater reduction in ischemic volume by 24 hours (previously complete reperfusion was required). Our reperfusion half-life study described under Project 1 Z01 NS003043-01 modeled the time to reperfusion in tPA-treated and untreated patients and derived the half-life value that defines reperfusion rates in these populations. Future lytic trials will obtain sufficient serial MRI scans to model reperfusion rates against these two benchmarks. Lesion volume measurements by MRI are used as an objective outcome measure ischemic stroke studies, yet some fundamental questions about these measurements had not previously been addressed. We have completed studies quantifying the intra- and inter-reader lesion volume measurements using a standard semi-automated planimetric method on lesions acquired with diffusion weighted (DWI), mean transit time (MTT) perfusion and Fluid Attenuated Inversion Recovery (FLAIR) MRI at acute and chronic time points in 68 patients. Across in image types and time points the test-retest correlations for both intra- and interreader variability were excellent, ranging between .83 and .99 (intra-reader with higher), with test-retest volume differences of <5% or <2cc. We also studied whether there was a difference in final infarct volumes measured at 30 and 90 days after stroke onset. Although 90 day is standard for assessing final clinical outcome, if 90 day lesion volume can be estimated at day 30 then the follow-up period for imaging outcome stroke studies could be shorter. We performed a retrospective study of 18 patients that had lesions scanned with DWI acutely and with FLAIR at both 30 and 90 days. Mean (SD) volumes for 30, and 90 day images were 18.6 (14.0), and 15.9 (13.8) mL respectively linear regression analysis revealed a strong relationship, r=0.96 (p<0.001), between lesion volumes at 30 and 90 days. Although there is a small difference between the volumes measured and 30 and 90 days, the 30 day measured is a good estimate of 90 day lesion volume and may be acceptable as a final infarct volume outcome. The unrelenting failure of neuroprotective stroke trials to demonstrate clinical benefits brings to light shortcomings in the development of neuroprotective drugs for stroke. No compound that has reached late Phase II or Phase III trials, has ever been demonstrated to have (1) reached the target brain tissue, the ischemic penumbra, in adequate concentrations, (2) produced a pharmacological effect in the ischemic region know to occur in animal models or (3) caused an attenuation of infarct volume. It is our objective to develop novel methods to directly measure neuroprotective effects in the stroke patient. Studies of infarct volume have been previously described under this and other projects. We are conducting several other investigations toward that end. The optimal experimental setting for a test of neuroprotection in stroke would be to begin administration of the neuroprotective drug prior to the ischemic event. While that is not possible for spontaneous stroke presenting to the hospital emergency department, we propose that patients undergoing open heart surgery may be such a clinical model. Although disabling stroke is an infrequent adverse consequence of this surgery (<5% of cases), published work has indicated that the occurrence of new ischemic lesions, most not overtly symptomatic, occurs at a much higher rate. In this year we have nearly completed enrollment in our study to identify pre-operative factors associated with the highest rates peri-operative ischemic lesions. If we find at least a 30% risk of new lesions, then a proof of principle pretreatment neuroprotective trial to modify the rate and volume of ischemic lesions would be feasible. Preliminary analysis of the data suggest that new lesions may occur in greater than 50% of patients, if this result is confirmed, then this approach can be taken to the next stage. We have begun to use MR spectroscopy applications to investigate in stroke patients altered brain metabolism and drug pharmacokinetics and pharmacodynamics. Spectroscopy has had limited application to acute stroke patients because of excessive long acquisition times. Using the stroke dedicated 3 Tesla MRI scanner at our Washington Hospital research site, we have been modifying and optimizing pulse sequences for this purpose. Chemical Shift Imaging (CSI) can provide muti-voxel images of proton metabolites such as NAA, creatine, choline, and lactate. We found a 2D PRESS to be optimal for this purpose. Using this protocol, 2D CSI spectra with matrix size of 12x10, slice thickness 15 mm can be obtained within five minutes. In addition, fat contamination of the spectra is greatly reduced because of the spatial localization by PRESS. This is important for detecting the lactate peak which overlaps with the lipid peaks. A metabolite more specific to stroke and neuroprotective stroke therapy is glutathione. Glutathione (GSH) is one the major intracellular anti-oxidants that has neuroprotective effects in ischemic stroke models. Preserved or increased tissue concentrations of glutathione may be evidence of antioxidant neuroprotective response. Glutathione (GSH) is difficult to measure in vivo due to its low concentration (0.8 to 3 mM) and overlapping of its peak with more intense peaks from other metabolites, such as creatine. There has been one prior report of a GSH spectrum in human brain at 4 Tesla with an acquisition time of 39 minutes. We developed a MEGA-PRESS sequence for the 3T Philips scanners at the NIH NMR Center and Washington Hospital Center. After theoretical simulations and hundreds of phantom scans, as well as in vivo scans, we found that TE of 131 ms is optimal. Using a voxel size of 50 x 30 x 30 mm-cubed to further improve signal strength, we consistently get a good quality GSH spectrum in 9 minutes. A novel post-processing algorithm and computer program were also developed to process the data and generate spectra. Frequency and phase changes caused by system instability and small patient motions are corrected by registering each spectrum to a target spectrum. This target spectrum is generated from the original spectra by a fitting and selection process. The whole post-processing program is fully automatic and gives reliable results consistently without any manual adjustment. GSH measurements in stroke patients will permit us to assess antioxidant effects of putative neuroprotective therapies.
