We propose to continue a novel, adaptive, secured system for parallel testing of promising interventions
designed to improve outcome after ischemic stroke when compared to reperfusion alone, the Stroke Preclinical
Assessment Network. SPAN will screen and select highly promising candidate interventions for possible study
in StrokeNet. The applicant PI and research team—the SPAN Coordinating Center (CC)—created and
implemented novel solutions to several key barriers impeding successful pre-clinical network implementation.
The result is a highly successful, multi-site network for testing putative cerebroprotectants in animal stroke
models that include equal numbers of females and males and co-morbidities such as aging, hypertension and
hyperglycemia. We invented novel methods of case enrollment and tracking; centralized randomization;
centralized drug preparation/masking/bottling; group concealment from the surgeons; centralized, blinded
behavioral assessment; and centralized, blinded evaluation of MRI scans. Each aspect of SPAN went through
a thorough, organized process: literature review, rounds of debate and re-review, and finally decisions were
made and documented. Using this process, SPAN developed SOPs concerning the choice of animal models,
surgical methods, behavior assessments, assessor training and certification. SPAN uses a Multi-Arm Multi-
Stage design and state-of-the-art experimental rigor to successfully reduce or eliminate common sources of
bias. Stroke is administered and interventions are provided in blind fashion. Outcome data (behavioral tests
and MRI scans) are uploaded by the SPAN Testing Laboratories to a central repository, and then randomly
assigned by the CC to independent raters at other labs. We, the SPAN CC, coordinate, network
communication via daily contact with the sites; weekly enrollment update reports; monthly Steering Committee
meetings; annual investigator meetings; and semi-annual site visits. To improve reproducibility across all labs,
we devised training sessions and certification tests for all surgeons and behavioral raters targeted to specific
tasks, e.g., surgery, recording behavior testing, corner test rating, etc. These training and certification tools will
facilitate rapid, rigorous addition of new Testing Labs in SPAN 2.0. Throughout SPAN 1.0 in fact, all SOPs,
protocols, and infrastructure were designed to facilitate rapid, simple addition of new sites and easy transition
to SPAN 2.0 with minimal down-time. Presently, we are devising innovative enhancements for the next funding
cycle: we are training a machine learning algorithm to score behavior videos to allow rapid, reproducible
scoring using a digital pipeline; we are developing a blood clot/thrombolysis model to allow testing in the
presence of thrombolytics. This present application, if funded, will allow the SPAN CC to continue to improve
and advance preclinical development by implementing critical technical innovations, including novel
assessment tools using machine learning. SPAN 2.0 will continue our track record of successful enrollment
and technical innovation, providing a model for pre-clinical networks in other disease areas.