Methods for randomized, blinded, controlled evaluation of putative disease interventions in multi-laboratory, preclinical assessment networks

Methods for randomized, blinded, controlled evaluation of putative disease interventions in multi-laboratory, preclinical assessment networks | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Resource Methods for randomized, blinded, controlled evaluation of putative disease interventions in multi-laboratory, preclinical assessment networks Karisma Nagarkatti, Marcio Diniz, Ryan Cabeen, Monica Estrada, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3054771/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 18 Feb, 2026 Read the published version in Lab Animal → Version 1 posted You are reading this latest preprint version Abstract Science faces a reproducibility crisis, and public trust in science suffers when large clinical trials—qualified by promising preclinical studies—fail. While some clinical trial designs may have been inadequate, preclinical assessments of disease interventions have lacked key elements of rigor: treatment concealment, randomization, blinded outcomes, pre-specified and adequate sample sizes, and models including co-morbidities. To demonstrate feasibility and practicality of enhanced rigor in preclinical assessment, a six-laboratory network was designed that implemented rigorous study elements, using acute ischemic stroke for demonstration. This network enrolled 2615 animals in five different models and implemented a multi-stage, multi-arm statistical design that sequentially eliminated candidate interventions during interim analyses. The methods included centralized intervention packaging, randomization, data quality assessment, and data archiving. Blinded analysis of 9,274 video-recorded behavior tasks and 3,652 magnetic resonance images were evaluated. All tools and protocols are presented and could be adapted to preclinical assessment in other disease areas. Scientific community and society/Scientific community/Research management Health sciences/Medical research/Preclinical research Biological sciences/Drug discovery Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Science faces skepticism from the lay public, and scientists have described problems with rigor, transparency, and reproducibility. Many published findings—selected from high-impact journals—failed replication outside of the original laboratories 1-3 . Many factors contribute to reproducibility issues in science: inadequate sample size and proper power analysis prior to initiating experiments; lack of control for repeated significance testing (‘p-hacking’); inadequate blinding of the investigators; insufficient or inappropriate controls, among other deficiencies 1,4-7 . Many groups, including the National Academy of Science, have called on grant agencies and journals to enforce higher standards of rigor and experimental design to address these deficiencies, but appropriate methods to implement greater scientific rigor may be lacking or insufficiently developed 8 . Here we address one important type of scientific study, the use of preclinical animal disease models to assess the efficacy of proposed candidate interventions. Prior to launching pivotal clinical trials in patients, many funders, sponsors, and regulators require that therapeutic efficacy be documented in an accepted animal disease model. Typically, such animal disease models replicate some key aspects of human disease; it is assumed that results from the animal disease model anticipate the results of subsequent human clinical trials. Too often, however, promising interventions—despite positive results in preclinical studies using animal disease models—fail to translate into clinical trials, most recently a widely touted antibody targeting cerebral beta-amyloid in patients with Alzheimer’s disease, for example 9 . Similar failures have been noted in neuroscience, cardiology, and oncology, among other areas 2,3,10,11 . Although the failure in clinical translation may partly result from an inadequate design of the clinical trial, our concern here is to improve the quality and validity of the preclinical assessment of candidate interventions. Key elements of design-quality in preclinical assessments include treatment concealment during disease induction; subject randomization; blinded outcome assessment; and adequate, pre-specified sample size 5,12-14 . These key elements may be challenging to address in a single laboratory if too few personnel are available to isolate the randomization process from the treatments and assessments. Even if the steps can be separately tasked, the interventions under study often appear different, are dosed differently, or in some other way can be identified. Study outcome variables—behavior, histomorphometry, image analysis—should be assessed by a completely independent investigator unaware of treatment grouping. All these processes must be performed simultaneously yet independently, requiring yet another entity to combine all input into a common file so that data can be analyzed, also in a blinded manner. Historically, all these steps do not occur in most academic or contract laboratories due to a lack of funds to employ so many independent workers. We sought to create and organize the operational methods needed to conduct an effective, rigorous, and successful preclinical assessment of putative disease interventions 15,16 . We intended that our methods could be used in any field that requires rigorous preclinical demonstration of treatment efficacy in an animal disease model. As proof of concept, we conducted a study of six putative cerebroprotective interventions for acute ischemic stroke, an area in which failure to translate positive preclinical assessments into clinical trial success has been well documented 10,17-20 . The overall vision and implementation of our multi-laboratory stroke preclinical assessment network have been published 21 ; here, the specific operational methods are presented so that they could be adapted by other groups seeking to accomplish preclinical assessment with the utmost rigor. In addition, we provide extensive data involving five different animal models with acute ischemic stroke to demonstrate the feasibility, practicality, and generalizability of these methods. Further details and specific protocols are provided as supplementary materials. Results Initial Framework and Approach The Stroke Preclinical Assessment Network (SPAN) was funded by the National Institute of Health (NIH) and included a coordinating center (CC) and six research laboratories ( Fig. 1 a and Supplementary Table 6). The participating SPAN research laboratories and the six study interventions were selected via peer review 21 , but in other applications the collaborating laboratories might be selected through peer review or may self-select as part of a voluntary collaboration. To provide central coordination, a laboratory with prior experience in both preclinical modeling and managing multi-center clinical trials was selected as the CC through peer review. During SPAN, the CC was not directly involved in performing the animal disease model or generating any of the outcome data; instead, the CC served as the central data depot, managed intervention supply, randomized interventions, assigned digitally recorded behavioral assessments for blinded review, and performed all statistical analysis. The Principal Investigators (PI) of each research laboratory comprised the study Steering Committee (SC) which also included NIH representatives and was chaired by the CC PI. Over several months, the SC debated and finalized standard operating procedures (SOP) for every method or course of action, 22 including the selection of behavioral and imaging endpoints suitable for the chosen disease model (Fig. 1 b). The CC initiated the trial with an in-person kick-off meeting. The SC approved 57 laboratory specific SOPs and an additional seven CC-specific SOPs (available through www.spannetwork.org ), although not all SOPs were completed prior to initiating the trial. The adoption of SOPs was intended to standardize procedures across the network, but the investigators recognized the importance of embracing heterogeneity when attempting to model human disease in animal models 23 . Therefore, a stroke model was used that involved temporary occlusion of the middle cerebral artery (MCAo) in five different animal models: normal young adult mice, aging mice (15–17 months), mice with diet-induced obesity and hyperglycemia, normal young adult rats, and spontaneously hypertensive rats. Subject randomization was stratified by research laboratory, sex, and type of stroke model. After the randomized intervention assignment, subjects underwent stroke surgery, followed by behavioral and imaging assessments (Fig. 1 c). Digital video recordings of the behavioral assessments and the magnetic resonance images (MRI) were uploaded to the Imaging Digital Archive (IDA) of the Laboratory of Neuroimaging (LONI) 24 for blinded review and quantification (Fig. 1 d). Drug intervention vials were shipped as needed to the laboratories prior to randomization. The number of unique data-collection forms was relatively constant per stage (Fig. 1 e) allowing the four data stages to be combined for statistical analysis. Cohort Control and Subject Flow To sequentially evaluate six interventions in parallel, each tested against an appropriate control, SPAN used a multi-arm, multi-stage (MAMS) design, using four stages 25 – 27 . At the end of Stages 1–3, interim analyses were performed: study interventions that appeared futile or efficacious were to be dropped using pre-specified futility and efficacy criteria. At the end of Stage 1, none of the interventions were dropped; after Stage 2, there were three dropped; after Stage 3 there were two more dropped; leaving one intervention for final testing in Stage 4. At the end of Stage 4, a final analysis that included all data from all stages confirmed that the one remaining intervention exceeded the pre-specified boundary for declaring it efficacious. Attrition bias results when investigators control and censor individual subjects after randomization and intervention 28 . In clinical trials, attrition bias is managed with intention-to-treat (ITT) analysis: patients are grouped into cohorts as randomized, not as actually treated, and drop-out (i.e., lost-to-follow up) patients remain in the ITT analysis 29 . Cohort and subject control have not previously been addressed in preclinical trials, and we implemented a workflow to account for every subject (Fig. 2 a). To prevent investigator’s influence over subject cohorts, research laboratories assigned a unique subject identifier and attached an MRI-compatible, bar-coded ear tag (Supplementary Fig. 1), to each study subject upon arrival. Ear tags were to remain affixed throughout all study procedures. After randomized assignments were sent to the research laboratories, surgery to induce stroke was performed and the assigned intervention was administered. The ITT analysis population was defined as all subjects that were randomized (Fig. 2 b). The modified ITT analysis included subjects that completed the disease model surgery and began the intervention 30 . Subjects who completed all assigned intervention doses and survived five days comprised the per-protocol (PP) analysis (Fig. 2 b). Throughout the follow-up period, subject dropout occurred due to death. We used the Consolidated Standards of Reporting Trials (CONSORT) approach (Fig. 2 b) typically required in clinical trials to account for all subjects in this study 31 . Use of the correct intervention assignment was confirmed post-hoc and worked well for both intravenous (IV) (Fig. 2 c) and intraperitoneal (IP) (Fig. 2 d) interventions. The stratified randomization process worked as intended: equal numbers of subjects were enrolled at each research laboratory across all groups (Fig. 2 e). These data confirm that the methods devised for stratified randomization across multiple research laboratories and stages are feasible and worked well. These data may guide the planning of future networks in other disease areas. Concealment and Blinding To avoid bias during the performance of the animal disease model, it is essential that group assignment be concealed from the investigator generating the disease model, e.g., performing the stroke surgery. In each SPAN research laboratory, the surgeon performing the MCAo was responsible for the anesthetic level and other variables that could be manipulated subconsciously to bias in favor of one or another treatment group, which is one reason for the concealment. The same investigator was then responsible for administering interventions during or after stroke onset, representing another opportunity for manipulation, subconscious or otherwise. Since many putative treatments look different and could easily be identified, the challenge in this multi-arm, multi-laboratory trial was to conceal the identity of the interventions by packaging them identically. In SPAN, all drug interventions were prepared in identical appearing vials (Fig. 3 a and Supplemental Fig. 1), and packed in coded, labeled vial boxes (Fig. 3 b). The vial boxes were arranged identically across the research laboratories to simplify preparation at the CC (Supplementary Fig. 2). Vial box loading was confirmed independently by two investigators at the CC to assure correct loading. After packaging, vial boxes were shipped in thermo-protected containers with a temperature excursion monitor. Due to the multi-stage approach, in which some study interventions may be dropped, the number of shipped vials (Fig. 3 c) and the number of shipping containers (Fig. 3 d) decreased over time. Reduced sample size per stage and improved efficiencies allowed the total cost of shipping (Fig. 3 e) to decrease over time. These data may help planning future preclinical networks. Assessing Behavioral Outcomes The corner test was selected as the primary outcome assessment and used for the sample size/power calculation of the multi-stage statistical design 32 , 33 . The corner test is simple to perform without expensive or complex equipment (Fig. 4 a). The grid walk test (Fig. 4 b) and hanging wire test were selected as secondary outcomes but the hanging wire was eliminated after Stage 1 for being redundant 34 . To assure fully blinded, objective ratings of the behavioral assessments, anonymized digital video-recordings of each evaluation were assigned to blinded raters at research laboratories other than the one that generated the recording (Fig. 4 c). Neurological deficit scores 35 indicated mild-moderate stroke severity across the entire study (Fig. 4 d) and were balanced across research laboratories, intervention, and sex. Because interventions were dropped between stages, total uploads of recorded videos decreased over the 4 stages (Fig. 4 e) as did the number of ratings generated from those uploaded videos (Fig. 4 f). The corner test proved feasible and baseline values were remarkably similar across five different animal models. Across all animal models, MCAo provoked significant behavioral deficits at 7 days and 30 days. While there was some heterogeneity across research laboratories, overall, the corner test indicated similar insult severity 23 . Each recorded corner test was randomly assigned to three certified raters. Recordings contained no identifying labels as to intervention or originating research laboratory. After Stage 1, the intraclass correlation coefficient (ICC) was calculated to assess concordance among the three raters. Concordance among three human raters was good, but insufficient to reduce the number of raters, (ICC = 0.732; 95% CL, 0.71–0.76), adjusting for time and intervention (Extended Data Fig. 1 ). The grid walk recordings and assessments were rated similarly to the corner test. At the end of Stage 1, however, concordance among the three human raters was sufficient to reduce the number of raters to one in subsequent stages. Again, feasibility and completion rates were similar to the corner test results. Image Analysis Pipeline Magnetic resonance imaging of each subject was attempted two and 30 days after the stroke, and the anonymized data were analyzed via an imaging pipeline (Fig. 5 a). Imaging sequences included a T2 weighted scan (Fig. 5 b) and an ADC weighted image (Fig. 5 c). In a pilot study, the fully automated analysis was validated using 2,3,5 triphenyl tetrazolium chloride (TTC) 36 stained sections (Fig. 5 d). The resulting correlation was excellent. To assist the research laboratories with their protocol adherence over the course of the study, a control limit chart of the Day 2 lesion volume (Fig. 5 e) was sent on a regular basis 37 . Control limits were defined as two standard deviations around the study-wide mean lesion volume. Data Quality Assurance Data monitoring at the CC ran continuously throughout all four stages of the study. The CC investigators checked key data elements using a risk-based monitoring approach drawn from clinical trial design 38 , 39 . Data discrepancies were resolved by issuing a data query to the relevant research laboratory which then drew attention to certain fields to the data entry personnel. As a result, the total number of data queries dropped significantly over the four stages (Fig. 6 a). Since the stages differed in the number of interventions, and, therefore, the total number of subjects enrolled declined over the stages, the number of queries adjusted for the total enrollment in each stage was compared (Fig. 6 b), which again confirmed a significant improvement over time, suggesting the existence of a learning curve in this multisite pre-clinical experiment. One indicator of the commitment of the research laboratory is the time used to address and reply to the data queries. The research laboratories reduced their time to query resolution over the four stages (Fig. 6 c) but there was significant variation among them (Fig. 6 d). In future applications, network managers should be aware of the variable commitment to data quality across research laboratories and over time and plan QC activities accordingly. Discussion The successful design and execution of SPAN was founded on experience gained from two prior multi-site network-based studies 40 , 41 . We created and adopted network-wide SOPs, including one common stroke model (filament MCAo) in five animal models that all six research laboratories embraced. The MAMS statistical procedure allowed SPAN to rigorously identify interventions that fell below the futility boundary, sequentially eliminate them and to continue testing those that had not exceeded the efficacy boundary. After four stages, one intervention exceeded the efficacy boundary. The feasibility of the approach is further supported by the observation that SPAN began subject enrollment during a global pandemic with intermittent, mandatory lock-down of laboratory staff. SPAN took advantage of digital, Internet-based videoconferencing, video training, and centralized certification to maintain progress during COVID-19 pandemic lockdown periods. Traditionally, significant organizational barriers impede the implementation of a multi-laboratory collaboration. The SPAN investigators addressed several of these barriers. SPAN drafted a template application for the research laboratories to submit to their Institutional Animal Care and Use Committees (IACUC) after assuring compliance with NIH and other national guidelines 42 . Research laboratory contracting was handled via an NIH grant mechanism (RFA NS-18-34 and NS-18-033) but could have been challenging in a privately funded effort. Future planners should allow time for contract negotiations. A plethora of decisions was made involving every aspect of the study protocol. Decision-making required finding consensus after careful and thorough literature review of options and prior experience; a collaborative approach guided by the SC; and a commitment from the research laboratory investigators to reach agreements on key decisions expeditiously. Minor operational decisions (for the example, the color of the labels or brand of the vial boxes) were made by CC investigators for efficiency, and detailed decision logs were maintained by the CC. Enrollment proceeded toward a pre-specified sample size, based on the agreement that each research laboratory could enroll eight subjects per week. This enrollment rate proved feasible at all six research laboratories over the 2.5 years of enrollment. Furthermore, the careful coordination of behavioral assessments and scheduling access to the MR scanners was necessary. To enhance feasibility and timely throughput, the SPAN CC created laboratory-specific ordering and surgery timelines which were distributed to all research laboratories prior to each of the four stages (Supplementary Table 3). Figures 1 to 3 provide descriptive data about many aspects of the trial. Collectively, these data paint a picture of the expected workload for future planners to set up their own preclinical assessment network. The SPAN investigators created operational methods to implement several desirable elements deemed essential to rigor in preclinical trial design 15 , 16 , 43 , 44 . All drug interventions were packaged in similar appearing, labeled vials (Fig. 3 a), to conceal the treatment assignment from the investigator who induced the disease model. Central randomization and 100% assignment of subjects avoided attrition bias and selection bias (Fig. 2 a). By distributing anonymized digital outcome data—behavior videos or MR images—group blinding was preserved through final data analysis (Fig. 1 d and Fig. 4 c). If rigor and scrupulous experimental design enhance the likelihood of future success in clinical trials, then the approach here provides future Investigators with a feasible and practical approach. From the outset, the methods created for SPAN and presented here were intended to serve more than one project. Principles of rigor and scientific excellence, e.g., concealment, blinding, randomization, and statistical power, transcend any specific intervention focus and generally apply to other disease areas. The data presented confirm that the same methods performed well in different stages of the investigation, across five different animal stroke models: young adult or aging C57BL/6J mice, obesity-induced hyperglycemia in C57BL/6J mice, young adult Sprague-Dawley rats, and spontaneously hypertensive rats. Although the same approach to inducing stroke—the filament MCAo—was used across the study, success in all five animal models suggests that the procedures and operational methods could work in other disease models. The SOPs were written to allow adaptation to any other disease areas. Although we believe our methods can be applied to other disease areas where investigators seek to assess putative disease-modifying intervention in animal models, our network has so far used only one method for inducing stroke, the filament occlusion model. Generalizability is suggested, but not a given. This study focused the assessment on only a few outcomes; other outcomes could perform differently, perhaps with varied concordance. Notice must be taken of results in the aging mouse model. A plethora of commentators suggested that putative stroke interventions must be tested in the aging model with co-morbid conditions. However, the mortality of MCAo in aging mice was over 50%. During this investigation, SPAN investigators attempted multiple maneuvers that have been described by others to promote survival in these mice: scrupulous temperature control, careful anesthesia, fluid resuscitation, limited handling, and careful genital hygiene in aging males. Unfortunately, over the course of Stages 2 and 3, which together included over 340 aging mice, we could not improve survival. Although stroke in aged patients has higher mortality than in younger patients, the mortality is nowhere near 50% 45 . Furthermore, from a cost and efficiency perspective, and even perhaps reproducibility, MCAo in aging mice may not be a viable future option for assessing putative interventions, although this question deserves further study. The methods presented here for establishing and maintaining a preclinical disease intervention network are feasible, practical, and generalizable. In addition, the specific operational methods should be straightforward to implement in other disease areas. Methods Network Capability Assessment Upon notification by NINDS of the six research laboratories selected for SPAN, the CC conducted in-person visits to establish the infrastructure available at each research laboratory (Fig. 1 a). Surgeon experience, resources available at the research laboratory, MRI capability, and several other elements were summarized. This compiled review of capabilities allowed the CC to create experimental protocols that would be feasible across all research laboratories. Communication systems were built, including a group email address at each research laboratory so that all team members could be addressed simultaneously. The CC established a hierarchy and organization for the flow of information to and from research laboratories and to the imaging repository (Fig. 1 d). Network Governance To expedite decision-making and to oversee protocol development, the governing body of the network was a Steering Committee (SC), convened by the Coordinating Center (CC) in conjunction with NINDS (Fig. 1 b). Steering Committee membership included the CC PI, the PI of each research laboratory, and NIH Program scientists. In addition, an independent External Advisory Board (EAB), appointed by and reporting to NINDS, was chartered to include basic, translational, and clinician scientists with expertise in cerebroprotection, representatives from the pharmaceutical and biotech industry, and experts in regulatory affairs, statistics, and clinical trial design. Trial Setup SPAN investigators began meeting in 2019 to select and design all aspects of the SPAN network, including structure, communication, animal models, outcomes, and protocols. The SPAN CC drafted clearly defined SOPs for all activities, which were edited and approved by the SC. An in-person kick-off meeting was held in Los Angeles, CA on September 9th and 10th, 2019 at which time protocol decisions were made. The CC also created and managed a separate Imaging Committee to design and approve the magnetic resonance imaging protocol used in SPAN and will be available from the corresponding author or by request through the website, www.spannetwork.org . Optimized Communication The SPAN CC team was in daily contact with staff at each research laboratory via email and phone. SC meetings were held monthly, and stage-specific meetings were held weekly. Initially, the SC meetings were limited to the research laboratory PIs but were expanded later to include all hands-on investigators to improve communication and understanding of the protocol and SOPs at the research laboratories. The SPAN CC sent a weekly Enrollment Report to the research laboratories, the NINDS and the EAB. The CC team visited each research laboratory twice in person and once virtually during the project. These visits allowed the CC to inspect the surgical and behavior-recording areas, audit data, meet staff, and disseminate best practices. The SPAN website, www.spannetwork.org , was used as a repository for distributing the SOPs and other needed information to all research laboratories. A public-facing webpage contains general information about the project and its members. A private access page allowed the CC to post template documents and SOPs. The website included a chat forum for investigators to share ideas and experiences. Conflict Resolution The SPAN CC sought to facilitate open, transparent communication and encouraged robust discussion on all topics. Consensus was achieved gradually and thoughtfully with multiple rounds of review. The NINDS Scientific Officers retained final authority to settle disputes if needed, but this never became necessary during the study. The CC regulated the decision-making deadlines to meet study timelines and coordinated a timely discussion through forums and emails with the SC. A pilot stage, consisting of 10 subjects per research laboratory, was helpful in troubleshooting the workflow of data collection, surgical procedures, behavior testing, and video upload. Interventions, Concealment, and Blinding Through rigorous peer review, six interventions were selected for study in this project 21 . One mechanical procedure, called remote ischemic conditioning or RIC, needed its own control group, making it impossible to conceal group assignment. The remaining five compounds (uric acid, tocilizumab, veliparib, fingolimod, and fasudil), were formulated as either IV or IP infusions. The CC met with biopharmaceutical companies and arranged for either purchased or donated drug interventions (Supplemental Note 1). Detailed information was collected from the investigators or manufacturers about preparation, valid excipients, aliquoting, and storage. The CC tried to obtain stability data on each compound. Where possible, drug interventions were resupplied with expiration dates based on this stability data. Where such stability information was not available, the CC conducted stability trials in-house using high-performance liquid chromatography (HPLC) (uric acid) or shipped test vials to research laboratories for them to perform bioassays (uric acid, fasudil, fingolimod). Additionally, as a further demonstration of stability, repeat bioactivity assays were performed on select drug interventions at the end of Stage 1 with unused vials at the research laboratory. These unused vials were tested near the end of their expiration date and all bioactivity was preserved. To facilitate blinding, the CC chose to administer two of the drug interventions intraperitoneally (IP; fasudil and fingolimod) and three intravenously (IV; veliparib, tocilizumab, uric acid) with matching 0.9% weight/volume (w/v) saline placebo controls. Dosing (sex and species) was determined by the research laboratory that proposed the intervention. The IV drug interventions were shipped in liquid form for a single 8 µL/g body weight infusion to be administered over 20 minutes starting five minutes before reperfusion. One of the IP drug interventions had a shorter shelf life once suspended so it was decided that all IP drug interventions would be lyophilized and resuspended immediately before use. Because of the larger volume of liquid needed for the six injections in 0.9% w/v saline, IP drug interventions were lyophilized in 5% w/v saline to minimize the volume and prevent a boil over during the lyophilization. For each randomized subject, an email prescription was generated that indicated the exact volume of sterile water to add to the coded vial to result in a final concentration of 0.9% w/v saline, and the correct concentration of resuspended drug interventions. The first IP injection was administered five minutes before reperfusion and then twice daily for five more doses. For each stage, the CC located and purchased glass vials (Supplementary Table 2) that were large enough to hold required quantities of drug interventions after estimating the expected dose ranges given the ages and weights of the mice and rats to be included (Supplementary Fig. 1a). At the CC, all vials were sealed with non-reactive rubber stoppers and secured with a crimped flip cap to maintain sterility of the septum until use (Supplementary Note 1). Placebo vials were prepared identically by loading 0.9% w/v saline or lyophilizing 5% w/v saline into matching vials. Multi-vial boxes (Fig. 3 b) for each research laboratory were loaded at the CC with the vials needed for one randomization block. Boxes were then labeled and shipped in qualified 2-Day summer shipping containers (ThermoSafe E3R2S, E6RR2S, E12R2S) with a temperature tag (ShockWatch WarmMark: WM 8/46). Research laboratories could read only the CC-applied labels and could not identify which intervention the vial contained. Surgical Procedures For this trial, the filament MCAo 46 was implemented at all research laboratories for several reasons. First, all research laboratories had prior experience performing a version of it, which reduced training and start up time. Second, the filament model can be easily accomplished in both mice and rats. Third, with the advent of thrombectomy in clinical practice 47 , in which an intravascular clot-retrieval catheter is used to recanalize an occluded artery, it is now possible to know precisely when reperfusion begins, and then immediately start intervention. We used the filament MCAo model to represent this clinical scenario of known recanalization with immediate initiation of the intervention. The SC drafted and agreed upon experimental protocols for each of the four stages, including standardization of ischemia and anesthesia. All SPAN experimental protocols are available through the website www.spannetwork.org . To ensure surgical reproducibility, all surgeons were certified prior to starting work. They performed the MCAo surgeries until they could produce 10 with measurable lesions, as demarcated with TTC. Once the CC received the TTC images and morphometry, these were reviewed, and if satisfactory, the surgeon was approved to begin. This process allowed new surgeons to join the network at any time. Process Control Throughout the trial, the CC desired to monitor quality across all research laboratories. To monitor stroke volume as a key indicator of research laboratories quality, we used a control limit chart 48 approach using the exponentially weighted moving average that weighted more recent data more heavily 37 : $${M}_{i}=\lambda {M}_{o}+\left(1- \lambda \right){M}_{i-1}$$ where \({M}_{o}\) is the average of volume lesion from the surgeon certification at Stage 1 or volume fraction from Stage 1 for the remaining stages; \({M}_{i}\) is the current moving average and \({M}_{i-1}\) is the previous moving average. The control limits were established based on the network-wide mean value \({M}_{o}\) ± L x \({V}_{i}\) with \({V}_{i}\) given by: $${V}_{i}^{2}= {\sigma }_{0}^{2}\left(1-{\left(1- \lambda \right)}^{2i}\right) \lambda /(2- \lambda )$$ where \({\sigma }_{0}^{2}\) is the variance of volume lesion from the surgeon certification at Stage 1 and volume fraction from Stage 1 for the remaining stages. Based on simulation studies, CC established \(\lambda =0.15\) and L = 3. Every 2 weeks the charts were updated, and research laboratories could observe their performance relative to the control limits. At a few points in the trial a research laboratory was briefly out of control. At these points, the CC informed the research laboratory and met with the surgical team to investigate potential explanations (was there a change in the equipment for example) for the out-of-control variations. In all cases, the research laboratories corrected procedures and returned the metric to within control limits (Fig. 5 e). Standardized Behavior Assessment To assure quality behavioral assessments, the CC arranged for on-line webinars during the global pandemic lockdown to illustrate proper technique. After each interactive webinar, research laboratories were given sample test videos to rate. After correctly scoring these test videos, an investigator was then considered certified to rate behavior in the study. Test videos included intentional violations of the recording SOP and the viewers were expected to detect and label > 90% of all these protocol violations. Raters who failed to correctly score the test videos, including failure to comment on protocol violations, were asked to re-watch the recorded training webinar and then rate a new set of test videos. This approach not only allowed for rigorous training and certification of raters initially, but it also allowed new raters to join the trial at any time later. Standardized Video Recording The corner test (Fig. 4 a) and grid walk test (Fig. 4 b) were both recorded at the research laboratories to be scored later by blinded, certified reviewers. Each research laboratory used a digital camera(GoPro Hero 7, 8 or 11), set to the lowest possible resolution (preferred 720 pixels) and frame rate (preferably 30 frames per second) to reduce file size and time needed to upload and download the videos for scoring. Standardized lighting was required to allow an optimum view of the behaving subject. The resulting, recorded digital video files were uploaded to the Imaging Digital Archive, a repository for the long-term preservation and sharing of biomedical research data 24 . Standardized Video Scoring/Rating Digital recordings of the Corner and Grid Walk tests were assigned at random by the CC to three trained and certified raters at research laboratories other than the one that recorded the video. Each week, raters received behavior scoresheets with lists of anonymized URL addresses and empty fields for data entry. Each rater scored the assigned video and entered the results into the scoresheet. Investigators were given 1 week to return completed scoresheets. After running quality checks on the returned data, the CC imported these results into a REDCap (Research Electronic Data Capture) database. Raters did not know the identity of the research laboratory, sex, behavior timepoint, or intervention given to the subject in the video they rated. Randomization Centralized randomization assures that subjects are allocated to intervention groups without bias or baseline differences. Stratification during randomization ensures balanced numbers in key variables that could influence outcome, such as sex and research laboratory; other stratification variables may apply in other disease-specific implementations. These needs required the design of a manual process that contained three essential tools: a custom REDCap database, a Master Intervention Spreadsheet (MISS) and a set of randomization tables generated for each research laboratory. These tools are described in detail to facilitate use in other preclinical assessment networks. Subjects began the centralized randomization process when the participating research investigator assigned the subject to a surgery date using an ITT form in REDCap. Stroke surgeries for males and females were to be performed concurrently and evenly distributed during each surgery day. Upon notification by the ITT form, the CC consulted the appropriate Randomization Table and assigned the subject to the next available row. The intervention assignment was entered into REDCap and the MISS (see Supplementary Note 3 for a more detailed description of the randomization process). After quality control steps to assure correct treatment assignment, a randomization email was sent to the research laboratory containing all information needed to treat: subject ID, assigned vial number, administration volume, route of administration, the volume of sterile water to add to resuspend if IP, and intervention schedule. After stroke surgery was performed, and intervention administered, correct administration was confirmed post-hoc. The REDCap Database Using REDCap 49 , 50 , a project was designed for data entry for both the end users and the CC. Several repeatable forms were designed to capture data over four stages of SPAN. Data Access Groups were established so that each research laboratory received access to their own records only. The REDCap Alerts & Notifications application was used to trigger email signals to notify the CC when subjects were ready for randomization and to notify the research laboratory when randomization was completed. Custom reports were created in REDCap to track intervention accuracy, scheduled vs. actual surgery dates, and incoming rater scores for invalid or missing fields, among many others. These reports facilitated quality control and management. Master Intervention Spreadsheet (MISS) Using Excel and the Visual Basic Macro Language, the MISS was designed to track inventory and assign vials during the manual randomization process (Supplementary Table 4). Using the MISS, CC staff tracked all inventory upon arrival at the CC, shipments to the research laboratories, and then tracked inventory as it was administered at the research laboratory. The MISS also calculated the dose to administer and provided an easy way to monitor for the shelf-life expiration of each vial. The MISS used relational logic to cross-check fields and had pre-programmed alerts (Supplementary Note 3 and Supplementary Table 5). Randomization Tables The randomization tables were created by the Statistical team using the pre-specified analysis plan, the stratification variables, and knowledge of the number of research laboratories and number of interventions. The tables and the MISS were devised in close collaboration. Rather than use a permuted block size, the randomization block sizes were unique to each stage and calculated from: multiples of the weekly subject enrollment at each research laboratory, the total sample size, the sex, the model, and the number of interventions being studied. Such variation in randomization block sizes allowed for equal distribution of interventions amongst sexes and models, adequate rotation of models and efficiencies in timing re-supply shipments of interventions to research laboratories while reducing loss (expired vials). Laboratory investigators were unaware of this approach to randomization block sizes. Avoiding Sources of Bias: Attrition, Selection and Performance Every animal purchased for SPAN was tracked through the entire protocol (Fig. 2 b). Upon arrival at the research laboratory, a barcoded ear tag (RapID, San Francisco, CA) was affixed (Fig. 2 a and Supplementary Fig. 1c). Barcodes were read with a handheld scanner paired to a Bluetooth compatible device to ensure accurate ear tag entry. The research laboratory documented the ear tag color in the enrollment form. The ear tag color was verified during subsequent study procedures, for example during behavior video review and scoring. If a discrepant animal ID or tag color was noted, the animal record was flagged and investigated. Once enrolled, 100% of subjects were accounted for through the end of the study. To prevent surgeons from assigning interventions to individuals in a biased manner, we used group concealment 44 , described above. To assure unbiased assessment, the investigators performed all behavior ratings and image analysis blinded to any knowledge of intervention group assignment 12 , 43 , 51 . The methods are described above. Translational Relevance To increase the translational relevance of the study, we included disease models that incorporated some of the key risk factors for stroke, including age, hypertension, and hyperglycemia. After a successful first stage using young, disease-free subjects, comorbidities were included in Stages 2 and 3. The comorbid models used were: aging mice, diet-induced obesity in mice, and spontaneously hypertensive rats. Aging mice, 15–17 months at surgery were obtained from the National Institute on Aging (Aged Rodent Colonies, National Institute on Aging (nih.gov)). To induce obesity, C57BL/6J mice were fed a high-fat diet (EnvigoTeklad; TD.06414) for 12 weeks and were 15–18 weeks old at surgery 52 – 54 . Spontaneously hypertensive rats (Charles River: SHR/NCrl) were obtained from Charles River Laboratories 55 and were 15–17 weeks old at surgery. To ensure even distribution of experience across all research laboratories and to avoid either research laboratory practice effects or seasonal effects we instituted several measures: The CC randomized the models assigned to the research laboratories in each stage; the randomization process (drawing of lots) was recorded using a video camera and archived for later review if needed. During Stages 2 and 3, each research laboratory was assigned to perform two of the models; no more than two research laboratories could perform a particular model simultaneously; research laboratories had to alternate between their assigned two models at least twice per stage to prevent an experience effect. By the end of the study, all models were performed in at least four different research laboratories. All these scheduling requirements assured that the study was free of practice effects, research laboratory preference/experience with one or another model, and seasonal effects. MRI Methods A protocol for MRI included quantitative imaging sequences for estimating the apparent diffusivity coefficient (ADC) and T2 relaxation rate. Data were acquired from Bruker scanners with field strengths across research laboratories, including 7T, 9.4T, or 11T. A multi-echo T2 sequence was used to estimate the T2 rate using relaxometry, with echo times rating from 10 to 100 milliseconds. A diffusion-weighted imaging sequence was used to estimate ADC, with b-values 0, 500, and 1000 s/mm 2 . After data acquisition, image files were uploaded to the LONI IDA for long-term storage and subsequent analysis. Using the Quantitative Imaging Toolkit 56 (QIT), we created an automated image analysis pipeline (code posted at https://doi.org/10.48550/arXiv.2203.05714 ) for generating quantitative reports and visualizations of individual animals, summarized 56 in Fig. 5 a. We segmented the brain using a deep learning neural network approach with a U-net architecture 57 , and we segmented lesions, cerebrospinal fluid, and normal-appearing brain tissue using a combination of thresholding and mathematical morphology. We estimated midline shift by isolating cerebrospinal fluid closest to the midline, computing the centroid, and measuring the relative distances to the boundaries of the brain. To visualize our data in a common reference frame, we aligned our data to a group-averaged scan using rigid registration with Advanced Normalization Tools (ANTs) 58 . We measured the volume of segmentation brain areas and lesion, aggregated the results in tabular form, and generated mosaic plots for later qualitative assessment. Statistical Methods Data Management Data inconsistencies across forms in the database, or range-check errors, resulted in a data query from the CC to the research laboratory. Data queries (Fig. 6 a and b) were initiated by the CC using the REDCap database. Research laboratory technicians and PIs were instructed to check REDCap data queries frequently. If the laboratory did not resolve the query initiated in REDCap within 48 hours, the CC alerted the technician to the query by email. If queries went unresolved for longer, the CC would request that the research laboratory PI follow up with their technician to resolve the query (Fig. 6 c and d). Daily, the CC ran quality checks in REDCap to track study progress (surgery completion, dropouts, intervention administration, end of study) at each research laboratory and to reconcile video and MR image uploads into IDA. When inconsistencies in video and MRI uploads were identified, the CC contacted investigators at the research laboratory to upload the missing video or image (Supplementary Note 2). Statistical Analysis Interventions were assessed using a multi-arm, multi-stage (MAMS) design based on the Generalized Dunnett’s test 26 , 27 . The MAMS design provides large gains in efficiency over separate randomized trials of each treatment as it allows a shared control group. Futility (lower) and efficacy (upper) boundaries were calculated in advance. At each interim analysis, a Dunnett test statistic for the primary endpoint comparing each arm with the control arm as reference was calculated 66 . If the test statistic value fell below the futility boundary, the intervention was to be declared as non-promising, and enrollment into that arm was to stop. If the test statistic value fell above the efficacy boundary, the intervention was to be declared effective, and enrollment into that arm could be stopped. Otherwise, the intervention was declared promising and enrollment into that arm continued into the next stage. Power and Sample Size Considerations Two parallel multi-stage trials were performed with the corner test index as the primary endpoint after transformation. The power analysis used this variable to derive sample sizes: (A) the first trial included five drug intervention arms and one control arm, split equally between IP control and IV control under the assumption that there would be no difference between the two control treatments for the primary endpoint; (B) the second trial had 1 mechanical intervention arm and 1 control arm. We implemented the MAMS design with 4 stages using Triangular futility and efficacy boundaries 25 , 27 . Three interim analyses were planned after 25, 50 and 75% recruitment with an equal allocation among arms. Based on a published study of the corner test in aged subjects, we estimated a mean of 0.55 and a standard deviation (SD) of 0.262 for the control arm 59 . We then increased the assumed SD to 1.048 as a conservative scenario to account for predicted excessive variability among the research laboratories. Assuming that the transformed corner test index in different animal models follows independent Normal (µ, σ 2 ) distributions with known variance, we defined an interesting effect size (δ) as a decrease on the corner test index of at least 50% to declare an intervention as effective, and a non-interesting effect size (δ 0 ) as a decrease of at most 6% to declare an intervention as ineffective. We chose 50% for the efficacy boundary based on a meta-analysis of prior preclinical trials in which the mean reported effect size was 50% 44 . We chose 6% for the futility boundary as this value was the effect size used to power several large, Phase 3 trials of interventions for acute ischemic stroke 60 , 61 . We tested a family of null hypotheses (H 0 ) with each hypothesis stating that the intervention arm average is greater than or equal to the control arm average of the 30-day transformed corner test index. The family-wise error rate was defined as the probability of falsely rejecting one or more null hypotheses within this family at any trial stage with fixed value of 5%. The power of 90% was calculated under the least favorable configuration (LFC), which is defined as the probability of rejecting only one null hypothesis in any stage 62 . We computed the minimum, maximum, and expected sample sizes under the null and LFC hypothesis based on Triangular boundaries 63 – 65 . Total planned sample size was then adjusted assuming 10% animal death over 30 days. Abbreviations (ANTs) Advanced Normalization Tools (ADC) Apparent diffusivity coefficient (CONSORT) Consolidated Standards of Reporting Trials (CC) Coordinating Center (δ) Effect size (EAB) External Advisory Board (EOS) End of study (HPLC) High-performance liquid chromatography (IDA) Imaging Digital Archive (μ, σ 2 ) Independent Normal distributions (IACUC) Institutional Animal Care and Use Committees (ITT) Intention-to-treat analysis (δ) Interesting effect size (ICC) Intraclass correlation coefficient (IP) Intraperitoneal (IV) Intravenous (LONI) Laboratory of Neuroimaging (LFC) Least favorable configuration (L) Left turns (ml) Milliliter (MRI) Magnetic resonance images (MISS) Master Intervention Spreadsheet (µL/g) Microliters per gram (MCAo) Middle cerebral artery occlusion (MAMS) Multi-arm, multi-stage (NIA) National Institute of Aging (NIH) National Institute of Health (NINDS) National Institute of Neurological Disorders (NDS) Neurological deficit score (δ 0 ) Non-interesting effect size (μ, σ 2 ) Normal distribution (H 0 ) Null hypotheses (NIfTI) Neuroimaging Informatics Technology Initiative file (PP) Per-protocol (PI) Principal Investigators (QIT) Quantitative Imaging Toolkit (REDCap) Research Electronic Data Capture (QC) Quality control (R) Right turns (s/mm 2 ) Seconds per millimeter squared (SD) Standard deviation (SOP) Standard operating procedure (SC) Steering Committee (SPAN) Stroke Preclinical Assessment Network (TTC) 2,3,5 Triphenyl tetrazolium chloride (w/v) Weight/volume Declarations Data and materials availability SOPs, experimental protocols, and MRI pipeline raw data and code will be available from the corresponding author or by request through the website, www.spannetwork.org. Acknowledgments We thank Zachary Grodzinski, Liron Israel and Rameshwar Patel for their expertise and guidance in the pharmacology and packing of the drug interventions. We would like to thank the National Institute of Ageing for their generous donation of the aging mice. Finally, we would like to thank Novartis for their generous donation of Fingolimod. References Sayre, F. & Riegelman, A. The reproducibility crisis and academic libraries. College & Research Libraries 79, 2 (2018). Errington, T. M. et al. Investigating the replicability of preclinical cancer biology. eLife 10, e71601, doi: 10.7554/eLife.71601 (2021). Perrin, S. Preclinical research: Make mouse studies work. Nature 507, 423–425, doi: 10.1038/507423a (2014). Errington, T. M., Denis, A., Perfito, N., Iorns, E. & Nosek, B. A. Challenges for assessing replicability in preclinical cancer biology. eLife 10, e67995, doi: 10.7554/eLife.67995 (2021). Voelkl, B., Vogt, L., Sena, E. S. & Wurbel, H. Reproducibility of preclinical animal research improves with heterogeneity of study samples. PLoS Biol 16, e2003693, doi: 10.1371/journal.pbio.2003693 (2018). Benjamin, D. J. et al. Redefine statistical significance. Nat Hum Behav 2, 6–10, doi: 10.1038/s41562-017-0189-z (2018). Steckler, T. Editorial: preclinical data reproducibility for R&D - the challenge for neuroscience. SpringerPlus 4, 1, doi: 10.1186/2193-1801-4-1 (2015). National Academies of Sciences, E. & Medicine. Reproducibility and Replicability in Science . (The National Academies Press, 2019). Veening-Griffioen, D. H. et al. Are some animal models more equal than others? A case study on the translational value of animal models of efficacy for Alzheimer's disease. European Journal of Pharmacology 859, 172524, doi: https://doi.org/10.1016/j.ejphar.2019.172524 (2019). O'Collins, V. E. et al. Preclinical drug evaluation for combination therapy in acute stroke using systematic review, meta-analysis, and subsequent experimental testing. J Cereb Blood Flow Metab 31, 962–975, doi:jcbfm2010184 [pii] 10.1038/jcbfm.2010.184 (2011). Sughrue, M. E. et al. Data presentation in rodent stroke studies and the predictive value of confidence intervals. J Clin Neurosci 17, 11–15, doi:S0967-5868(09)00340-3 [pii] 10.1016/j.jocn.2009.05.008 (2010). Sena, E. S., Currie, G. L., McCann, S. K., Macleod, M. R. & Howells, D. W. Systematic reviews and meta-analysis of preclinical studies: why perform them and how to appraise them critically. J Cereb Blood Flow Metab 34, 737–742, doi: 10.1038/jcbfm.2014.28 (2014). van der Worp, H. B. et al. Can animal models of disease reliably inform human studies? PLoS medicine 7, e1000245, doi: 10.1371/journal.pmed.1000245 (2010). Macleod, M. R. et al. Reprint: Good laboratory practice: preventing introduction of bias at the bench. Int J Stroke 4, 3–5, doi:IJS241 [pii] 10.1111/j.1747-4949.2009.00241.x (2009). Bix, G. J. et al. Uncovering the Rosetta Stone: Report from the First Annual Conference on Key Elements in Translating Stroke Therapeutics from Pre-Clinical to Clinical. Transl Stroke Res 9, 258–266, doi: 10.1007/s12975-018-0628-9 (2018). Bosetti, F. et al. Translational Stroke Research: Vision and Opportunities. Stroke 48, 2632–2637, doi: 10.1161/strokeaha.117.017112 (2017). Lyden, P., Buchan, A., Boltze, J., Fisher, M. & Consortium*, S. X. Top Priorities for Cerebroprotective Studies-A Paradigm Shift: Report From STAIR XI. Stroke 52, 3063–3071, doi: 10.1161/STROKEAHA.121.034947 (2021). Feuerstein, G. Z. et al. Missing steps in the STAIR case: a Translational Medicine perspective on the development of NXY-059 for treatment of acute ischemic stroke. J Cereb Blood Flow Metab 28, 217–219, doi:9600516 [pii] 10.1038/sj.jcbfm.9600516 (2008). Savitz, S. I. A critical appraisal of the NXY-059 neuroprotection studies for acute stroke: a need for more rigorous testing of neuroprotective agents in animal models of stroke. Exp Neurol 205, 20–25 (2007). Lyden, P. & Lapchak, P. Sisyphus and translational stroke research. Sci Transl Med 4, 156ps120, doi:4/156/156ps20 [pii] 10.1126/scitranslmed.3005083 (2012). Lyden, P. D. et al. The Stroke Preclinical Assessment Network: Rationale, Design, Feasibility, and Stage 1 Results. Stroke 53, 1802–1812, doi: 10.1161/STROKEAHA.121.038047 (2022). Dirnagl, U. & Group, M. o. t. M.-S. O. P. Standard operating procedures (SOP) in experimental stroke research: SOP for middle cerebral artery occlusion in the mouse. Nature Precedings, doi: 10.1038/npre.2012.3492.3 (2012). Morais, A. et al. Embracing Heterogeneity in The Multicenter Stroke Preclinical Assessment Network (SPAN) Trial. Stroke 54, 620–631 (2023). Crawford, K. L., Neu, S. C. & Toga, A. W. The Image and Data Archive at the Laboratory of Neuro Imaging. Neuroimage 124, 1080–1083, doi: 10.1016/j.neuroimage.2015.04.067 (2016). Jaki, T. F., Pallmann, P. S. & Magirr, D. The R package MAMS for designing multi-arm multi-stage clinical trials. Journal of Statistical Software (2017). Wason, J. M. & Trippa, L. A comparison of Bayesian adaptive randomization and multi-stage designs for multi-arm clinical trials. Stat Med 33, 2206–2221, doi: 10.1002/sim.6086 (2014). Jaki, T., Magirr, D. & Pallmann, P. MAMS: designing multi-arm multi-stage studies, 2014. URL http://CRAN. R-project. org/package = MAMS. R package version 0.3 . Dumville, J. C., Torgerson, D. J. & Hewitt, C. E. Reporting attrition in randomised controlled trials. Bmj 332, 969–971, doi: 10.1136/bmj.332.7547.969 (2006). Group, R. i. S. W. Improving the Quality of NINDS-Supported Preclinical and Clinical Research through Rigorous Study Design and Transparent Reporting. (2011). Ranganathan, P., Pramesh, C. S. & Aggarwal, R. Common pitfalls in statistical analysis: Intention-to-treat versus per-protocol analysis. Perspect Clin Res 7, 144–146, doi: 10.4103/2229-3485.184823 (2016). Begg, C. et al. Improving the quality of reporting of randomized controlled trials. The CONSORT statement. JAMA 276, 637–639 (1996). Manwani, B. et al. Functional recovery in aging mice after experimental stroke. Brain Behav Immun 25, 1689–1700, doi: 10.1016/j.bbi.2011.06.015 (2011). Li, X. et al. Chronic behavioral testing after focal ischemia in the mouse: functional recovery and the effects of gender. Exp Neurol 187, 94–104, doi: 10.1016/j.expneurol.2004.01.004 (2004). Schaar, K. L., Brenneman, M. M. & Savitz, S. I. Functional assessments in the rodent stroke model. Exp Transl Stroke Med 2, 13, doi: 10.1186/2040-7378-2-13 (2010). Bederson, J. B. et al. Rat Middle Cerebral-Artery Occlusion - Evaluation of the Model and Development of A Neurologic Examination. Stroke 17, 472 (1986). Bederson, J. B. et al. Evaluation of 2,3,5-triphenyltetrazolium chloride as a stain for detection and quantification of experimental cerebral infarction in rats. Stroke 17, 1304 (1986). Hunter, J. S. The Exponentially Weighted Moving Average. Journal of Quality Technology 18, 203–210, doi: 10.1080/00224065.1986.11979014 (1986). Agrafiotis, D. K. et al. Risk-based Monitoring of Clinical Trials: An Integrative Approach. Clinical Therapeutics 40, 1204–1212, doi: https://doi.org/10.1016/j.clinthera.2018.04.020 (2018). Tantsyura, V. et al. Risk-Based Monitoring: A Closer Statistical Look at Source Document Verification, Queries, Study Size Effects, and Data Quality. Therapeutic Innovation & Regulatory Science 49, 903–910, doi: 10.1177/2168479015586001 (2015). Rewell, S. S. et al. Evolution of ischemic damage and behavioural deficit over 6 months after MCAo in the rat: Selecting the optimal outcomes and statistical power for multi-centre preclinical trials. PLoS One 12, e0171688, doi: 10.1371/journal.pone.0171688 (2017). Llovera, G. et al. Results of a preclinical randomized controlled multicenter trial (pRCT): Anti-CD49d treatment for acute brain ischemia. Sci Transl Med 7, 299ra121, doi: 10.1126/scitranslmed.aaa9853 (2015). Gettayacamin, M. & Retnam, L. AAALAC International Standards and Accreditation Process. Toxicological Research 33, 183–189, doi: 10.5487/TR.2017.33.3.183 (2017). Macleod, M. R. et al. Good laboratory practice: preventing introduction of bias at the bench. Stroke 40, e50-52, doi:STROKEAHA.108.525386 [pii] 10.1161/STROKEAHA.108.525386 (2009). Sena, E., van der Worp, H. B., Howells, D. & Macleod, M. How can we improve the pre-clinical development of drugs for stroke? Trends Neurosci 30, 433–439, doi: 10.1016/j.tins.2007.06.009 (2007). Lozano, R. et al. Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 380, 2095–2128, doi: 10.1016/S0140-6736(12)61728-0 (2012). Van Winkle, J. A. et al. Concurrent middle cerebral artery occlusion and intra-arterial drug infusion via ipsilateral common carotid artery catheter in the rat. J Neurosci Methods 213, 63–69, doi: 10.1016/j.jneumeth.2012.12.004 (2013). Goyal, M. et al. Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials. Lancet 387, 1723–1731, doi: 10.1016/S0140-6736(16)00163-X (2016). English, J. R., Krishnamurthi, M. & Sastri, T. Quality monitoring of continuous flow processes. Computers & industrial engineering 20, 251–260 (1991). Harris, P. A. et al. The REDCap consortium: Building an international community of software platform partners. Journal of biomedical informatics 95, 103208 (2019). Harris, P. A. et al. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. Journal of biomedical informatics 42, 377–381 (2009). Crossley, N. A. et al. Empirical evidence of bias in the design of experimental stroke studies: a metaepidemiologic approach. Stroke 39, 929–934, doi:STROKEAHA.107.498725 [pii] 10.1161/STROKEAHA.107.498725 (2008). Wang, C.-Y. & Liao, J. K. in mTOR: Methods in Molecular Biology Vol. 821 (ed T. Weichhart) 421–433 (Springer, 2012). Surwit, R. S., Kuhn, C. M., Cochrane, C., McCubbin, J. A. & Feinglos, M. N. Diet-induced type II diabetes in C57BL/6J mice. Diabetes 37, 1163–1167 (1988). Yang, Y., Smith Jr, D. L., Keating, K. D., Allison, D. B. & Nagy, T. R. Variations in body weight, food intake and body composition after long-term high‐fat diet feeding in C57BL/6J mice. Obesity 22, 2147–2155 (2014). Okamoto, K. & Aoki, K. Development of a strain of spontaneously hypertensive rats. Jpn Circ J 27, 282–293, doi: 10.1253/jcj.27.282 (1963). Cabeen, R., Laidlaw, D. & Toga, A. Quantitative Imaging Toolkit: Software for Interactive 3D Visualization, Data Exploration, and Computational Analysis of Neuroimaging Datasets , 2018). Ronneberger, O., Fischer, P. & Brox, T. in International Conference on Medical Image Computing and Computer-Assisted Intervention. 234–241 (Springer International Publishing). Avants, B., Tustison, N. & Song, G. Advanced Normalization Tools: V1.0. doi: 10.54294/uvnhin (2009). Manwani, B. et al. Differential effects of aging and sex on stroke induced inflammation across the lifespan. Exp Neurol 249, 120–131, doi: 10.1016/j.expneurol.2013.08.011 (2013). Shuaib, A. et al. NXY-059 for the treatment of acute ischemic stroke. N Engl J Med 357, 562–571 (2007). Zivin, J. A. et al. Effectiveness and safety of transcranial laser therapy for acute ischemic stroke. Stroke 40, 1359–1364, doi:STROKEAHA.109.547547 [pii] 10.1161/STROKEAHA.109.547547 (2009). Dunnett, C. W. in Design of Experiments: Ranking and Selection (ed Tamhane AC Santer T.J.) 47–66 (Marcel Dekker, 1984). Whitehead, J. The design and analysis of sequential clinical trials . (John Wiley & Sons, 1997). O'Brien, P. C. & Fleming, T. R. A multiple testing procedure for clinical trials. Biometrics 35, 549–556 (1979). Magirr, D., Stallard, N. & Jaki, T. Flexible sequential designs for multi-arm clinical trials. Stat Med 33, 3269–3279, doi: 10.1002/sim.6183 (2014). Magirr, D., Jaki, T. & Whitehead, J. A generalized Dunnett test for multi-arm multi-stage clinical studies with treatment selection. Biometrika 99, 494–501 (2012). Diehl, K., Hull, R. & Morton, D. Pfi ster R, Rabemampianina Y, Smith D, Vidal JM, van de Vorstenbosch C (2001) A good practice guide to the administration of substances and removal of blood, including routes and volumes. Journal of Applied Toxicology 21, 15–23. Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryTables1.xlsx Supplementary tables SupplementaryMaterials.docx Cite Share Download PDF Status: Published Journal Publication published 18 Feb, 2026 Read the published version in Lab Animal → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3054771","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Resource","associatedPublications":[],"authors":[{"id":211665472,"identity":"e9a2ada0-94a7-4c6c-9ec0-0f699852e382","order_by":0,"name":"Karisma Nagarkatti","email":"","orcid":"","institution":"Keck School of Medicine of USC","correspondingAuthor":false,"prefix":"","firstName":"Karisma","middleName":"","lastName":"Nagarkatti","suffix":""},{"id":211665473,"identity":"34f4ca49-b1bd-4934-946c-77a83c430e9c","order_by":1,"name":"Marcio Diniz","email":"","orcid":"","institution":"Cedars-Sinai Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Marcio","middleName":"","lastName":"Diniz","suffix":""},{"id":211665474,"identity":"07964477-183c-46dc-b1fe-5d4c3df03910","order_by":2,"name":"Ryan Cabeen","email":"","orcid":"","institution":"Keck School of Medicine of USC","correspondingAuthor":false,"prefix":"","firstName":"Ryan","middleName":"","lastName":"Cabeen","suffix":""},{"id":211665475,"identity":"114db456-a0a0-40bf-90e8-eec9eaee3710","order_by":3,"name":"Monica Estrada","email":"","orcid":"","institution":"Keck School of Medicine of USC","correspondingAuthor":false,"prefix":"","firstName":"Monica","middleName":"","lastName":"Estrada","suffix":""},{"id":211665476,"identity":"d316eb4d-8a3b-4898-a51f-aab5c5d2aa9e","order_by":4,"name":"Karen Crawford","email":"","orcid":"","institution":"Keck School of Medicine of USC","correspondingAuthor":false,"prefix":"","firstName":"Karen","middleName":"","lastName":"Crawford","suffix":""},{"id":211665477,"identity":"9c187360-e2df-48f2-8bd3-7940991ed757","order_by":5,"name":"Andre Rogatko","email":"","orcid":"","institution":"Cedars-Sinai Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Andre","middleName":"","lastName":"Rogatko","suffix":""},{"id":211665478,"identity":"703e66d7-90eb-48c6-b946-cd7d7375560d","order_by":6,"name":"Sungjin Kim","email":"","orcid":"","institution":"Cedars-Sinai Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Sungjin","middleName":"","lastName":"Kim","suffix":""},{"id":211665479,"identity":"8edad6f0-64a2-4d4e-8d67-86ae1e4cc8d5","order_by":7,"name":"Cenk Ayata","email":"","orcid":"https://orcid.org/0000-0002-3644-6042","institution":"Neurovascular Research Unit, Department of Radiology, Massachusetts General Hospital, Harvard Medical School","correspondingAuthor":false,"prefix":"","firstName":"Cenk","middleName":"","lastName":"Ayata","suffix":""},{"id":211665480,"identity":"60c9cc72-8289-45a8-97b6-f505f07c1008","order_by":8,"name":"Anil Chauhan","email":"","orcid":"","institution":"Carver College of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Anil","middleName":"","lastName":"Chauhan","suffix":""},{"id":211665481,"identity":"4a326e85-af80-4da4-a544-d76858be6f29","order_by":9,"name":"David Hess","email":"","orcid":"","institution":"Medical College of Georgia","correspondingAuthor":false,"prefix":"","firstName":"David","middleName":"","lastName":"Hess","suffix":""},{"id":211665482,"identity":"3d9406a9-0131-4c7c-9c4b-62cd56cf1957","order_by":10,"name":"Mohammad Khan","email":"","orcid":"","institution":"Medical College of Georgia","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"","lastName":"Khan","suffix":""},{"id":211665483,"identity":"1badf245-5714-44b2-845a-32de76a6483f","order_by":11,"name":"Mariia Kumskova","email":"","orcid":"","institution":"Carver College of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Mariia","middleName":"","lastName":"Kumskova","suffix":""},{"id":211665484,"identity":"bf8a459f-8b5b-49f4-911c-03657d14e8a9","order_by":12,"name":"Enrique Leira","email":"","orcid":"","institution":"Carver College of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Enrique","middleName":"","lastName":"Leira","suffix":""},{"id":211665485,"identity":"19bad815-ee4c-49f5-8bec-5428604c47f2","order_by":13,"name":"Rakeshkumar Patel","email":"","orcid":"","institution":"Carver College of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Rakeshkumar","middleName":"","lastName":"Patel","suffix":""},{"id":211665486,"identity":"99bce9d5-2189-4bbc-a484-f3e75fddfeae","order_by":14,"name":"Patrick Lyden","email":"","orcid":"","institution":"Keck School of Medicine of USC","correspondingAuthor":false,"prefix":"","firstName":"Patrick","middleName":"","lastName":"Lyden","suffix":""},{"id":211665487,"identity":"24804663-1ad5-40c5-b81c-b9cc8de7b7c3","order_by":15,"name":"Jessica Lamb","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAqklEQVRIiWNgGAWjYBACPgYGNgaJCghHgigtbGAtZ0jWwthGkhb202kPLOcdtudvYD54m4coLTy52w0ktx1OnHGALdmaOC0MudskgFoSDBh4zKSJ08L/FqhlzmF7Awb+b0RqkQDZ0nCYcQMDDxuxWt5uN5A4lp444zCbseUcYrTw8+dueyxRY23P39788MYbYrSAADM4PpiJVQ4CjB9IUT0KRsEoGAUjDwAAXDMoUF3kpwwAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-0342-9983","institution":"Keck School of Medicine of USC","correspondingAuthor":true,"prefix":"","firstName":"Jessica","middleName":"","lastName":"Lamb","suffix":""}],"badges":[],"createdAt":"2023-06-12 19:00:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3054771/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3054771/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41684-026-01683-z","type":"published","date":"2026-02-18T05:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":73518974,"identity":"afa771f9-503d-4dfd-a535-1b849d9ce14e","added_by":"auto","created_at":"2025-01-10 18:05:33","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":211831,"visible":true,"origin":"","legend":"\u003cp\u003eDescription of the network. a, Geographical representation of key elements of the network. b, Organizational components of the network. c, General experimental timeline through end of study (EOS). d, Pathways of data flow from the research laboratories to central storage and analysis. e, Total number of REDCap forms for data capture for each stage. MCAo=middle cerebral artery occlusion; NDS=neurological deficit score; MRI=magnetic resonance imaging\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/b5ef2a06d71f184adac49bdc.jpg"},{"id":73518976,"identity":"b8adeb27-1236-4fa9-914e-5cb1b421f0ef","added_by":"auto","created_at":"2025-01-10 18:05:33","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":280596,"visible":true,"origin":"","legend":"\u003cp\u003eSubject flow from enrollment to MCAo, administration errors, CONSORT diagram and total randomized subjects per research laboratory. a, Workflow for the randomization process for every animal enrolled. b, Population groups following CONSORT guidelines. c, Total animals for each stage that received IV interventions and total medication errors for all 4 stages combined. There were no errors in the RIC treatment group. d, Total animals for each stage that received IP interventions and total medication errors for all 4 stages combined. e, Total animals randomized during each stage at each of the six research laboratories.\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/6d865c86dbabe47ca821d27d.jpg"},{"id":73520876,"identity":"f34dbd73-e790-4f94-a045-ad3375297965","added_by":"auto","created_at":"2025-01-10 18:29:33","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":190967,"visible":true,"origin":"","legend":"\u003cp\u003eIntervention concealment and supply to research laboratories. a, Example labeled vial. b, Example set of IV vials ready to be shipped to research laboratory. c, Total number of vials shipped to research laboratories for each stage. Each subject received treatment from one vial. \u0026nbsp;D, Total Thermos-controlled shipping boxes/containers shipped to research laboratories for each stage. This number indicated the total number of shipments from the CC to the laboratories. e, Total cost of shipping for each stage, which declined due to a combination of efficiency and diminishing sample size as some treatments dropped.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/d427dbf5646750959a2ef810.jpg"},{"id":73520433,"identity":"5be661d0-0e48-408f-9e38-0344fe55deb4","added_by":"auto","created_at":"2025-01-10 18:21:33","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":249500,"visible":true,"origin":"","legend":"\u003cp\u003eBehavioral testing workflow and feasibility. a, Sketch of the corner test set-up with camera mounted above the apparatus. b, Sketch of the grid walk test set-up with the camera and light source below the grid. c, Workflow of the behavior test videos recorded, then uploaded, then sent out for blinded scoring. Multiple quality control (QC) steps throughout the workflow assured data integrity. d, NDS of animals at Day 1 and Day 2 post-MCAo for each research laboratory and combined. e, Numbers of the corner and grid behavior test videos uploaded to IDA for each stage. f, Numbers of corner and grid behavior test assignments completed by research laboratories for each stage. None of the research laboratories rated their own videos.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/9077e008512d782c01d92cb2.jpg"},{"id":73518980,"identity":"086eb730-fc40-4b99-b038-60ccdc189329","added_by":"auto","created_at":"2025-01-10 18:05:33","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":409324,"visible":true,"origin":"","legend":"\u003cp\u003eLesion volume estimation using an automated image analysis pipeline. a, MR images of each subject were analyzed via an imaging pipeline. b-c, Imaging sequences included a T2 weighted scan (b) and an ADC weighted image (c). d, In a pilot study, the fully automated analysis was validated using TTC stained sections. e, To assist the research laboratories with their protocol adherence over the course of the study, a control limit chart of day two lesion volume was sent on a regular basis. Control limits were defined as 2 standard deviations around the study-wide mean lesion volume. Fig. 5a reprinted from [ https://doi.org/10.48550/arXiv.2203.05714].\u003c/p\u003e","description":"","filename":"5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/34721142966c7e05eeab030b.jpg"},{"id":73520253,"identity":"3972839d-9c29-4489-95f0-cfcc6f42042c","added_by":"auto","created_at":"2025-01-10 18:13:33","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":88212,"visible":true,"origin":"","legend":"\u003cp\u003eQueries. a, Total queries created in REDCap across the 4 stages showing improved data entry. b, To correct for the decreased number of subjects and interventions included in each stage, the number of queries per subject is calculated the improvement remains but with a slight uptick in Stage 4. c, Research laboratories took a varied number of days each stage to resolve queries, however the response time is still low. d, Chart shows the varied time that each research laboratory took to resolve open queries over the 4 stages.\u003c/p\u003e","description":"","filename":"6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/d82eb8f0c1b74d1db4c9d8c3.jpg"},{"id":102977093,"identity":"d8f558ef-3e98-44fc-8ea0-71e1d280caba","added_by":"auto","created_at":"2026-02-19 08:07:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2437776,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/61ef93dd-f203-451c-8408-c60ae06623e3.pdf"},{"id":73520251,"identity":"38c5f8c0-2d96-42fb-8c7a-e4b202c57b52","added_by":"auto","created_at":"2025-01-10 18:13:33","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":59669,"visible":true,"origin":"","legend":"Supplementary tables","description":"","filename":"SupplementaryTables1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/cc516699b08276dccf1146bb.xlsx"},{"id":73518981,"identity":"08dea494-40c2-4639-80de-d6fe571f3ec9","added_by":"auto","created_at":"2025-01-10 18:05:34","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":5507961,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-3054771/v1/625669f37d539355bae0ef1f.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Methods for randomized, blinded, controlled evaluation of putative disease interventions in multi-laboratory, preclinical assessment networks","fulltext":[{"header":"Introduction","content":"\u003cp\u003eScience faces skepticism from the lay public, and scientists have described problems with rigor, transparency, and reproducibility. Many published findings\u0026mdash;selected from high-impact journals\u0026mdash;failed replication outside of the original laboratories\u003csup\u003e1-3\u003c/sup\u003e. Many factors contribute to reproducibility issues in science: inadequate sample size and proper power analysis prior to initiating experiments; lack of control for repeated significance testing (\u0026lsquo;p-hacking\u0026rsquo;); inadequate blinding of the investigators; insufficient or inappropriate controls, among other deficiencies\u003csup\u003e1,4-7\u003c/sup\u003e. Many groups, including the National Academy of Science, have called on grant agencies and journals to enforce higher standards of rigor and experimental design to address these deficiencies, but appropriate methods to implement greater scientific rigor may be lacking or insufficiently developed\u003csup\u003e8\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eHere we address one important type of scientific study, the use of preclinical animal disease models to assess the efficacy of proposed candidate interventions. Prior to launching pivotal clinical trials in patients, many funders, sponsors, and regulators require that therapeutic efficacy be documented in an accepted animal disease model. Typically, such animal disease models replicate some key aspects of human disease; it is assumed that results from the animal disease model anticipate the results of subsequent human clinical trials. Too often, however, promising interventions\u0026mdash;despite positive results in preclinical studies using animal disease models\u0026mdash;fail to translate into clinical trials, most recently a widely touted antibody targeting cerebral beta-amyloid in patients with Alzheimer\u0026rsquo;s disease, for example\u003csup\u003e9\u003c/sup\u003e. Similar failures have been noted in neuroscience, cardiology, and oncology, among other areas\u003csup\u003e2,3,10,11\u003c/sup\u003e. Although the failure in clinical translation may partly result from an inadequate design of the clinical trial, our concern here is to improve the quality and validity of the preclinical assessment of candidate interventions.\u003c/p\u003e\n\u003cp\u003eKey elements of design-quality in preclinical assessments include treatment concealment during disease induction; subject randomization; blinded outcome assessment; and adequate, pre-specified sample size\u003csup\u003e5,12-14\u003c/sup\u003e. These key elements may be challenging to address in a single laboratory if too few personnel are available to isolate the randomization process from the treatments and assessments. Even if the steps can be separately tasked, the interventions under study often appear different, are dosed differently, or in some other way can be identified. Study outcome variables\u0026mdash;behavior, histomorphometry, image analysis\u0026mdash;should be assessed by a completely independent investigator unaware of treatment grouping. All these processes must be performed simultaneously yet independently, requiring yet another entity to combine all input into a common file so that data can be analyzed, also in a blinded manner. Historically, all these steps do not occur in most academic or contract laboratories due to a lack of funds to employ so many independent workers.\u003c/p\u003e\n\u003cp\u003eWe sought to create and organize the operational methods needed to conduct an effective, rigorous, and successful preclinical assessment of putative disease interventions\u003csup\u003e15,16\u003c/sup\u003e. We intended that our methods could be used in any field that requires rigorous preclinical demonstration of treatment efficacy in an animal disease model. As proof of concept, we conducted a study of six putative cerebroprotective interventions for acute ischemic stroke, an area in which failure to translate positive preclinical assessments into clinical trial success has been well documented\u003csup\u003e10,17-20\u003c/sup\u003e. The overall vision and implementation of our multi-laboratory stroke preclinical assessment network have been published\u003csup\u003e21\u003c/sup\u003e; here, the specific operational methods are presented so that they could be adapted by other groups seeking to accomplish preclinical assessment with the utmost rigor. In addition, we provide extensive data involving five different animal models with acute ischemic stroke to demonstrate the feasibility, practicality, and generalizability of these methods. Further details and specific protocols are provided as supplementary materials.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec2\" class=\"Section2\"\u003e \u003ch2\u003eInitial Framework and Approach\u003c/h2\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eThe Stroke Preclinical Assessment Network (SPAN) was funded by the National Institute of Health (NIH) and included a coordinating center (CC) and six research laboratories (\u003c/span\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eand Supplementary Table\u0026nbsp;6). The participating SPAN research laboratories and the six study interventions were selected via peer review\u003c/span\u003e\u003csup\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/span\u003e\u003c/sup\u003e, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ebut in other applications the collaborating laboratories might be selected through peer review or may self-select as part of a voluntary collaboration. To provide central coordination, a laboratory with prior experience in both preclinical modeling and managing multi-center clinical trials was selected as the CC through peer review. During SPAN, the CC was not directly involved in performing the animal disease model or generating any of the outcome data; instead, the CC served as the central data depot, managed intervention supply, randomized interventions, assigned digitally recorded behavioral assessments for blinded review, and performed all statistical analysis. The Principal Investigators (PI) of each research laboratory comprised the study Steering Committee (SC) which also included NIH representatives and was chaired by the CC PI.\u003c/span\u003e\u003c/p\u003e \u003cp\u003eOver several months, the SC debated and finalized standard operating procedures (SOP) for every method or course of action,\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e including the selection of behavioral and imaging endpoints suitable for the chosen disease model (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb). The CC initiated the trial with an in-person kick-off meeting. The SC approved 57 laboratory specific SOPs and an additional seven CC-specific SOPs (available through \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003ca href=\"http://www.spannetwork.org\" target=\"_blank\"\u003ewww.spannetwork.org\u003c/a\u003e\u003c/span\u003e\u003cspan address=\"http://www.spannetwork.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e ), although not all SOPs were completed prior to initiating the trial. The adoption of SOPs was intended to standardize procedures across the network, but the investigators recognized the importance of embracing heterogeneity when attempting to model human disease in animal models\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Therefore, a stroke model was used that involved temporary occlusion of the middle cerebral artery (MCAo) in five different animal models: normal young adult mice, aging mice (15\u0026ndash;17 months), mice with diet-induced obesity and hyperglycemia, normal young adult rats, and spontaneously hypertensive rats.\u003c/p\u003e \u003cp\u003eSubject randomization was stratified by research laboratory, sex, and type of stroke model. After the randomized intervention assignment, subjects underwent stroke surgery, followed by behavioral and imaging assessments (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec). Digital video recordings of the behavioral assessments and the magnetic resonance images (MRI) were uploaded to the Imaging Digital Archive (IDA) of the Laboratory of Neuroimaging (LONI)\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e for blinded review and quantification (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ed). Drug intervention vials were shipped as needed to the laboratories prior to randomization. The number of unique data-collection forms was relatively constant per stage (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ee) allowing the four data stages to be combined for statistical analysis.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eCohort Control and Subject Flow\u003c/h3\u003e\n\u003cp\u003eTo sequentially evaluate six interventions in parallel, each tested against an appropriate control, SPAN used a multi-arm, multi-stage (MAMS) design, using four stages\u003csup\u003e\u003cspan additionalcitationids=\"CR26\" citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. At the end of Stages 1\u0026ndash;3, interim analyses were performed: study interventions that appeared futile or efficacious were to be dropped using pre-specified futility and efficacy criteria. At the end of Stage 1, none of the interventions were dropped; after Stage 2, there were three dropped; after Stage 3 there were two more dropped; leaving one intervention for final testing in Stage 4. At the end of Stage 4, a final analysis that included all data from all stages confirmed that the one remaining intervention exceeded the pre-specified boundary for declaring it efficacious.\u003c/p\u003e \u003cp\u003eAttrition bias results when investigators control and censor individual subjects after randomization and intervention\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. In clinical trials, attrition bias is managed with intention-to-treat (ITT) analysis: patients are grouped into cohorts as randomized, not as actually treated, and drop-out (i.e., lost-to-follow up) patients remain in the ITT analysis\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. Cohort and subject control have not previously been addressed in preclinical trials, and we implemented a workflow to account for every subject (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). To prevent investigator\u0026rsquo;s influence over subject cohorts, research laboratories assigned a unique subject identifier and attached an MRI-compatible, bar-coded ear tag (Supplementary Fig.\u0026nbsp;1), to each study subject upon arrival. Ear tags were to remain affixed throughout all study procedures. After randomized assignments were sent to the research laboratories, surgery to induce stroke was performed and the assigned intervention was administered. The ITT analysis population was defined as all subjects that were randomized (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). The modified ITT analysis included subjects that completed the disease model surgery and began the intervention\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. Subjects who completed all assigned intervention doses and survived five days comprised the per-protocol (PP) analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). Throughout the follow-up period, subject dropout occurred due to death. We used the Consolidated Standards of Reporting Trials (CONSORT) approach (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb) typically required in clinical trials to account for all subjects in this study\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. Use of the correct intervention assignment was confirmed post-hoc and worked well for both intravenous (IV) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec) and intraperitoneal (IP) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed) interventions. The stratified randomization process worked as intended: equal numbers of subjects were enrolled at each research laboratory across all groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee). These data confirm that the methods devised for stratified randomization across multiple research laboratories and stages are feasible and worked well. These data may guide the planning of future networks in other disease areas.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eConcealment and Blinding\u003c/h2\u003e \u003cp\u003eTo avoid bias during the performance of the animal disease model, it is essential that group assignment be concealed from the investigator generating the disease model, e.g., performing the stroke surgery. In each SPAN research laboratory, the surgeon performing the MCAo was responsible for the anesthetic level and other variables that could be manipulated subconsciously to bias in favor of one or another treatment group, which is one reason for the concealment. The same investigator was then responsible for administering interventions during or after stroke onset, representing another opportunity for manipulation, subconscious or otherwise. Since many putative treatments look different and could easily be identified, the challenge in this multi-arm, multi-laboratory trial was to conceal the identity of the interventions by packaging them identically. In SPAN, all drug interventions were prepared in identical appearing vials (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea and Supplemental Fig.\u0026nbsp;1), and packed in coded, labeled vial boxes (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). The vial boxes were arranged identically across the research laboratories to simplify preparation at the CC (Supplementary Fig.\u0026nbsp;2). Vial box loading was confirmed independently by two investigators at the CC to assure correct loading. After packaging, vial boxes were shipped in thermo-protected containers with a temperature excursion monitor. Due to the multi-stage approach, in which some study interventions may be dropped, the number of shipped vials (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec) and the number of shipping containers (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed) decreased over time. Reduced sample size per stage and improved efficiencies allowed the total cost of shipping (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee) to decrease over time. These data may help planning future preclinical networks.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eAssessing Behavioral Outcomes\u003c/h3\u003e\n\u003cp\u003eThe corner test was selected as the primary outcome assessment and used for the sample size/power calculation of the multi-stage statistical design\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. The corner test is simple to perform without expensive or complex equipment (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). The grid walk test (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb) and hanging wire test were selected as secondary outcomes but the hanging wire was eliminated after Stage 1 for being redundant\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. To assure fully blinded, objective ratings of the behavioral assessments, anonymized digital video-recordings of each evaluation were assigned to blinded raters at research laboratories other than the one that generated the recording (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec). Neurological deficit scores\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e indicated mild-moderate stroke severity across the entire study (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed) and were balanced across research laboratories, intervention, and sex. Because interventions were dropped between stages, total uploads of recorded videos decreased over the 4 stages (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee) as did the number of ratings generated from those uploaded videos (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ef).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe corner test proved feasible and baseline values were remarkably similar across five different animal models. Across all animal models, MCAo provoked significant behavioral deficits at 7 days and 30 days. While there was some heterogeneity across research laboratories, overall, the corner test indicated similar insult severity\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eEach recorded corner test was randomly assigned to three certified raters. Recordings contained no identifying labels as to intervention or originating research laboratory. After Stage 1, the intraclass correlation coefficient (ICC) was calculated to assess concordance among the three raters. Concordance among three human raters was good, but insufficient to reduce the number of raters, (ICC\u0026thinsp;=\u0026thinsp;0.732; 95% CL, 0.71\u0026ndash;0.76), adjusting for time and intervention (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe grid walk recordings and assessments were rated similarly to the corner test. At the end of Stage 1, however, concordance among the three human raters was sufficient to reduce the number of raters to one in subsequent stages. Again, feasibility and completion rates were similar to the corner test results.\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eImage Analysis Pipeline\u003c/h2\u003e \u003cp\u003eMagnetic resonance imaging of each subject was attempted two and 30 days after the stroke, and the anonymized data were analyzed via an imaging pipeline (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea). Imaging sequences included a T2 weighted scan (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb) and an ADC weighted image (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ec). In a pilot study, the fully automated analysis was validated using 2,3,5 triphenyl tetrazolium chloride (TTC) \u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e stained sections (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ed). The resulting correlation was excellent. To assist the research laboratories with their protocol adherence over the course of the study, a control limit chart of the Day 2 lesion volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ee) was sent on a regular basis\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. Control limits were defined as two standard deviations around the study-wide mean lesion volume.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eData Quality Assurance\u003c/h2\u003e \u003cp\u003eData monitoring at the CC ran continuously throughout all four stages of the study. The CC investigators checked key data elements using a risk-based monitoring approach drawn from clinical trial design\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e,\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. Data discrepancies were resolved by issuing a data query to the relevant research laboratory which then drew attention to certain fields to the data entry personnel. As a result, the total number of data queries dropped significantly over the four stages (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). Since the stages differed in the number of interventions, and, therefore, the total number of subjects enrolled declined over the stages, the number of queries adjusted for the total enrollment in each stage was compared (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb), which again confirmed a significant improvement over time, suggesting the existence of a learning curve in this multisite pre-clinical experiment. One indicator of the commitment of the research laboratory is the time used to address and reply to the data queries. The research laboratories reduced their time to query resolution over the four stages (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec) but there was significant variation among them (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed). In future applications, network managers should be aware of the variable commitment to data quality across research laboratories and over time and plan QC activities accordingly.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe successful design and execution of SPAN was founded on experience gained from two prior multi-site network-based studies\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. We created and adopted network-wide SOPs, including one common stroke model (filament MCAo) in five animal models that all six research laboratories embraced. The MAMS statistical procedure allowed SPAN to rigorously identify interventions that fell below the futility boundary, sequentially eliminate them and to continue testing those that had not exceeded the efficacy boundary. After four stages, one intervention exceeded the efficacy boundary. The feasibility of the approach is further supported by the observation that SPAN began subject enrollment during a global pandemic with intermittent, mandatory lock-down of laboratory staff. SPAN took advantage of digital, Internet-based videoconferencing, video training, and centralized certification to maintain progress during COVID-19 pandemic lockdown periods.\u003c/p\u003e \u003cp\u003eTraditionally, significant organizational barriers impede the implementation of a multi-laboratory collaboration. The SPAN investigators addressed several of these barriers. SPAN drafted a template application for the research laboratories to submit to their Institutional Animal Care and Use Committees (IACUC) after assuring compliance with NIH and other national guidelines\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. Research laboratory contracting was handled via an NIH grant mechanism (RFA NS-18-34 and NS-18-033) but could have been challenging in a privately funded effort. Future planners should allow time for contract negotiations. A plethora of decisions was made involving every aspect of the study protocol. Decision-making required finding consensus after careful and thorough literature review of options and prior experience; a collaborative approach guided by the SC; and a commitment from the research laboratory investigators to reach agreements on key decisions expeditiously. Minor operational decisions (for the example, the color of the labels or brand of the vial boxes) were made by CC investigators for efficiency, and detailed decision logs were maintained by the CC.\u003c/p\u003e \u003cp\u003e Enrollment proceeded toward a pre-specified sample size, based on the agreement that each research laboratory could enroll eight subjects per week. This enrollment rate proved feasible at all six research laboratories over the 2.5 years of enrollment. Furthermore, the careful coordination of behavioral assessments and scheduling access to the MR scanners was necessary. To enhance feasibility and timely throughput, the SPAN CC created laboratory-specific ordering and surgery timelines which were distributed to all research laboratories prior to each of the four stages (Supplementary Table\u0026nbsp;3). Figures\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e to \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e provide descriptive data about many aspects of the trial. Collectively, these data paint a picture of the expected workload for future planners to set up their own preclinical assessment network.\u003c/p\u003e \u003cp\u003eThe SPAN investigators created operational methods to implement several desirable elements deemed essential to rigor in preclinical trial design\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. All drug interventions were packaged in similar appearing, labeled vials (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea), to conceal the treatment assignment from the investigator who induced the disease model. Central randomization and 100% assignment of subjects avoided attrition bias and selection bias (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). By distributing anonymized digital outcome data\u0026mdash;behavior videos or MR images\u0026mdash;group blinding was preserved through final data analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ed and Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec). If rigor and scrupulous experimental design enhance the likelihood of future success in clinical trials, then the approach here provides future Investigators with a feasible and practical approach.\u003c/p\u003e \u003cp\u003eFrom the outset, the methods created for SPAN and presented here were intended to serve more than one project. Principles of rigor and scientific excellence, e.g., concealment, blinding, randomization, and statistical power, transcend any specific intervention focus and generally apply to other disease areas. The data presented confirm that the same methods performed well in different stages of the investigation, across five different animal stroke models: young adult or aging C57BL/6J mice, obesity-induced hyperglycemia in C57BL/6J mice, young adult Sprague-Dawley rats, and spontaneously hypertensive rats. Although the same approach to inducing stroke\u0026mdash;the filament MCAo\u0026mdash;was used across the study, success in all five animal models suggests that the procedures and operational methods could work in other disease models. The SOPs were written to allow adaptation to any other disease areas. Although we believe our methods can be applied to other disease areas where investigators seek to assess putative disease-modifying intervention in animal models, our network has so far used only one method for inducing stroke, the filament occlusion model. Generalizability is suggested, but not a given. This study focused the assessment on only a few outcomes; other outcomes could perform differently, perhaps with varied concordance.\u003c/p\u003e \u003cp\u003eNotice must be taken of results in the aging mouse model. A plethora of commentators suggested that putative stroke interventions must be tested in the aging model with co-morbid conditions. However, the mortality of MCAo in aging mice was over 50%. During this investigation, SPAN investigators attempted multiple maneuvers that have been described by others to promote survival in these mice: scrupulous temperature control, careful anesthesia, fluid resuscitation, limited handling, and careful genital hygiene in aging males. Unfortunately, over the course of Stages 2 and 3, which together included over 340 aging mice, we could not improve survival. Although stroke in aged patients has higher mortality than in younger patients, the mortality is nowhere near 50%\u003csup\u003e45\u003c/sup\u003e. Furthermore, from a cost and efficiency perspective, and even perhaps reproducibility, MCAo in aging mice may not be a viable future option for assessing putative interventions, although this question deserves further study.\u003c/p\u003e \u003cp\u003eThe methods presented here for establishing and maintaining a preclinical disease intervention network are feasible, practical, and generalizable. In addition, the specific operational methods should be straightforward to implement in other disease areas.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003eNetwork Capability Assessment\u003c/h2\u003e\n \u003cp\u003eUpon notification by NINDS of the six research laboratories selected for SPAN, the CC conducted in-person visits to establish the infrastructure available at each research laboratory (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ea). Surgeon experience, resources available at the research laboratory, MRI capability, and several other elements were summarized. This compiled review of capabilities allowed the CC to create experimental protocols that would be feasible across all research laboratories. Communication systems were built, including a group email address at each research laboratory so that all team members could be addressed simultaneously. The CC established a hierarchy and organization for the flow of information to and from research laboratories and to the imaging repository (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ed).\u003c/p\u003e\n \u003cp\u003e\u003cspan type=\"BoldUnderline\" class=\"BoldUnderline\" name=\"Emphasis\"\u003eNetwork Governance\u003c/span\u003e\u003c/p\u003e\n \u003cp\u003eTo expedite decision-making and to oversee protocol development, the governing body of the network was a Steering Committee (SC), convened by the Coordinating Center (CC) in conjunction with NINDS (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eb). Steering Committee membership included the CC PI, the PI of each research laboratory, and NIH Program scientists. In addition, an independent External Advisory Board (EAB), appointed by and reporting to NINDS, was chartered to include basic, translational, and clinician scientists with expertise in cerebroprotection, representatives from the pharmaceutical and biotech industry, and experts in regulatory affairs, statistics, and clinical trial design.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eTrial Setup\u003c/h2\u003e\n \u003cp\u003eSPAN investigators began meeting in 2019 to select and design all aspects of the SPAN network, including structure, communication, animal models, outcomes, and protocols. The SPAN CC drafted clearly defined SOPs for all activities, which were edited and approved by the SC. An in-person kick-off meeting was held in Los Angeles, CA on September 9th and 10th, 2019 at which time protocol decisions were made. The CC also created and managed a separate Imaging Committee to design and approve the magnetic resonance imaging protocol used in SPAN and will be available from the corresponding author or by request through the website, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.spannetwork.org\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eOptimized Communication\u003c/h2\u003e\n \u003cp\u003eThe SPAN CC team was in daily contact with staff at each research laboratory via email and phone. SC meetings were held monthly, and stage-specific meetings were held weekly. Initially, the SC meetings were limited to the research laboratory PIs but were expanded later to include all hands-on investigators to improve communication and understanding of the protocol and SOPs at the research laboratories. The SPAN CC sent a weekly Enrollment Report to the research laboratories, the NINDS and the EAB. The CC team visited each research laboratory twice in person and once virtually during the project. These visits allowed the CC to inspect the surgical and behavior-recording areas, audit data, meet staff, and disseminate best practices. The SPAN website, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.spannetwork.org\u003c/span\u003e\u003c/span\u003e, was used as a repository for distributing the SOPs and other needed information to all research laboratories. A public-facing webpage contains general information about the project and its members. A private access page allowed the CC to post template documents and SOPs. The website included a chat forum for investigators to share ideas and experiences.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eConflict Resolution\u003c/h2\u003e\n \u003cp\u003eThe SPAN CC sought to facilitate open, transparent communication and encouraged robust discussion on all topics. Consensus was achieved gradually and thoughtfully with multiple rounds of review. The NINDS Scientific Officers retained final authority to settle disputes if needed, but this never became necessary during the study. The CC regulated the decision-making deadlines to meet study timelines and coordinated a timely discussion through forums and emails with the SC. A pilot stage, consisting of 10 subjects per research laboratory, was helpful in troubleshooting the workflow of data collection, surgical procedures, behavior testing, and video upload.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003eInterventions, Concealment, and Blinding\u003c/h2\u003e\n \u003cp\u003eThrough rigorous peer review, six interventions were selected for study in this project\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. One mechanical procedure, called remote ischemic conditioning or RIC, needed its own control group, making it impossible to conceal group assignment. The remaining five compounds (uric acid, tocilizumab, veliparib, fingolimod, and fasudil), were formulated as either IV or IP infusions. The CC met with biopharmaceutical companies and arranged for either purchased or donated drug interventions (Supplemental Note 1). Detailed information was collected from the investigators or manufacturers about preparation, valid excipients, aliquoting, and storage. The CC tried to obtain stability data on each compound. Where possible, drug interventions were resupplied with expiration dates based on this stability data. Where such stability information was not available, the CC conducted stability trials in-house using high-performance liquid chromatography (HPLC) (uric acid) or shipped test vials to research laboratories for them to perform bioassays (uric acid, fasudil, fingolimod). Additionally, as a further demonstration of stability, repeat bioactivity assays were performed on select drug interventions at the end of Stage 1 with unused vials at the research laboratory. These unused vials were tested near the end of their expiration date and all bioactivity was preserved.\u003c/p\u003e\n \u003cp\u003eTo facilitate blinding, the CC chose to administer two of the drug interventions intraperitoneally (IP; fasudil and fingolimod) and three intravenously (IV; veliparib, tocilizumab, uric acid) with matching 0.9% weight/volume (w/v) saline placebo controls. Dosing (sex and species) was determined by the research laboratory that proposed the intervention. The IV drug interventions were shipped in liquid form for a single 8 \u0026micro;L/g body weight infusion to be administered over 20 minutes starting five minutes before reperfusion. One of the IP drug interventions had a shorter shelf life once suspended so it was decided that all IP drug interventions would be lyophilized and resuspended immediately before use. Because of the larger volume of liquid needed for the six injections in 0.9% w/v saline, IP drug interventions were lyophilized in 5% w/v saline to minimize the volume and prevent a boil over during the lyophilization. For each randomized subject, an email prescription was generated that indicated the exact volume of sterile water to add to the coded vial to result in a final concentration of 0.9% w/v saline, and the correct concentration of resuspended drug interventions. The first IP injection was administered five minutes before reperfusion and then twice daily for five more doses. For each stage, the CC located and purchased glass vials (Supplementary Table\u0026nbsp;2) that were large enough to hold required quantities of drug interventions after estimating the expected dose ranges given the ages and weights of the mice and rats to be included (Supplementary Fig.\u0026nbsp;1a). At the CC, all vials were sealed with non-reactive rubber stoppers and secured with a crimped flip cap to maintain sterility of the septum until use (Supplementary Note 1). Placebo vials were prepared identically by loading 0.9% w/v saline or lyophilizing 5% w/v saline into matching vials. Multi-vial boxes (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eb) for each research laboratory were loaded at the CC with the vials needed for one randomization block. Boxes were then labeled and shipped in qualified 2-Day summer shipping containers (ThermoSafe E3R2S, E6RR2S, E12R2S) with a temperature tag (ShockWatch WarmMark: WM 8/46). Research laboratories could read only the CC-applied labels and could not identify which intervention the vial contained.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003eSurgical Procedures\u003c/h2\u003e\n \u003cp\u003eFor this trial, the filament MCAo\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e was implemented at all research laboratories for several reasons. First, all research laboratories had prior experience performing a version of it, which reduced training and start up time. Second, the filament model can be easily accomplished in both mice and rats. Third, with the advent of thrombectomy in clinical practice\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e, in which an intravascular clot-retrieval catheter is used to recanalize an occluded artery, it is now possible to know precisely when reperfusion begins, and then immediately start intervention. We used the filament MCAo model to represent this clinical scenario of known recanalization with immediate initiation of the intervention. The SC drafted and agreed upon experimental protocols for each of the four stages, including standardization of ischemia and anesthesia. All SPAN experimental protocols are available through the website \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.spannetwork.org\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eTo ensure surgical reproducibility, all surgeons were certified prior to starting work. They performed the MCAo surgeries until they could produce 10 with measurable lesions, as demarcated with TTC. Once the CC received the TTC images and morphometry, these were reviewed, and if satisfactory, the surgeon was approved to begin. This process allowed new surgeons to join the network at any time.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003eProcess Control\u003c/h2\u003e\n \u003cp\u003eThroughout the trial, the CC desired to monitor quality across all research laboratories. To monitor stroke volume as a key indicator of research laboratories quality, we used a control limit chart\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e approach using the exponentially weighted moving average that weighted more recent data more heavily\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e:\u003c/p\u003e\n \u003cdiv id=\"Equa\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e$${M}_{i}=\\lambda {M}_{o}+\\left(1- \\lambda \\right){M}_{i-1}$$\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({M}_{o}\\)\u003c/span\u003e\u003c/span\u003e is the average of volume lesion from the surgeon certification at Stage 1 or volume fraction from Stage 1 for the remaining stages; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({M}_{i}\\)\u003c/span\u003e\u003c/span\u003eis the current moving average and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({M}_{i-1}\\)\u003c/span\u003e\u003c/span\u003e is the previous moving average. The control limits were established based on the network-wide mean value\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({M}_{o}\\)\u003c/span\u003e\u003c/span\u003e \u0026plusmn; L x \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({V}_{i}\\)\u003c/span\u003e\u003c/span\u003e with \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({V}_{i}\\)\u003c/span\u003e\u003c/span\u003e given by:\u003c/p\u003e\n \u003cdiv id=\"Equb\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e$${V}_{i}^{2}= {\\sigma }_{0}^{2}\\left(1-{\\left(1- \\lambda \\right)}^{2i}\\right) \\lambda /(2- \\lambda )$$\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\sigma }_{0}^{2}\\)\u003c/span\u003e\u003c/span\u003eis the variance of volume lesion from the surgeon certification at Stage 1 and volume fraction from Stage 1 for the remaining stages. Based on simulation studies, CC established \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\lambda =0.15\\)\u003c/span\u003e\u003c/span\u003e and L\u0026thinsp;=\u0026thinsp;3. Every 2 weeks the charts were updated, and research laboratories could observe their performance relative to the control limits. At a few points in the trial a research laboratory was briefly out of control. At these points, the CC informed the research laboratory and met with the surgical team to investigate potential explanations (was there a change in the equipment for example) for the out-of-control variations. In all cases, the research laboratories corrected procedures and returned the metric to within control limits (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ee).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eStandardized Behavior Assessment\u003c/h2\u003e\n \u003cp\u003eTo assure quality behavioral assessments, the CC arranged for on-line webinars during the global pandemic lockdown to illustrate proper technique. After each interactive webinar, research laboratories were given sample test videos to rate. After correctly scoring these test videos, an investigator was then considered certified to rate behavior in the study. Test videos included intentional violations of the recording SOP and the viewers were expected to detect and label\u0026thinsp;\u0026gt;\u0026thinsp;90% of all these protocol violations. Raters who failed to correctly score the test videos, including failure to comment on protocol violations, were asked to re-watch the recorded training webinar and then rate a new set of test videos. This approach not only allowed for rigorous training and certification of raters initially, but it also allowed new raters to join the trial at any time later.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003ch2\u003eStandardized Video Recording\u003c/h2\u003e\n \u003cp\u003eThe corner test (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ea) and grid walk test (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eb) were both recorded at the research laboratories to be scored later by blinded, certified reviewers. Each research laboratory used a digital camera(GoPro Hero 7, 8 or 11), set to the lowest possible resolution (preferred 720 pixels) and frame rate (preferably 30 frames per second) to reduce file size and time needed to upload and download the videos for scoring. Standardized lighting was required to allow an optimum view of the behaving subject. The resulting, recorded digital video files were uploaded to the Imaging Digital Archive, a repository for the long-term preservation and sharing of biomedical research data\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\n \u003ch2\u003eStandardized Video Scoring/Rating\u003c/h2\u003e\n \u003cp\u003eDigital recordings of the Corner and Grid Walk tests were assigned at random by the CC to three trained and certified raters at research laboratories other than the one that recorded the video. Each week, raters received behavior scoresheets with lists of anonymized URL addresses and empty fields for data entry. Each rater scored the assigned video and entered the results into the scoresheet. Investigators were given 1 week to return completed scoresheets. After running quality checks on the returned data, the CC imported these results into a REDCap (Research Electronic Data Capture) database. Raters did not know the identity of the research laboratory, sex, behavior timepoint, or intervention given to the subject in the video they rated.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\n \u003ch2\u003eRandomization\u003c/h2\u003e\n \u003cp\u003eCentralized randomization assures that subjects are allocated to intervention groups without bias or baseline differences. Stratification during randomization ensures balanced numbers in key variables that could influence outcome, such as sex and research laboratory; other stratification variables may apply in other disease-specific implementations. These needs required the design of a manual process that contained three essential tools: a custom REDCap database, a Master Intervention Spreadsheet (MISS) and a set of randomization tables generated for each research laboratory. These tools are described in detail to facilitate use in other preclinical assessment networks.\u003c/p\u003e\n \u003cp\u003eSubjects began the centralized randomization process when the participating research investigator assigned the subject to a surgery date using an ITT form in REDCap. Stroke surgeries for males and females were to be performed concurrently and evenly distributed during each surgery day. Upon notification by the ITT form, the CC consulted the appropriate Randomization Table and assigned the subject to the next available row. The intervention assignment was entered into REDCap and the MISS (see Supplementary Note 3 for a more detailed description of the randomization process). After quality control steps to assure correct treatment assignment, a randomization email was sent to the research laboratory containing all information needed to treat: subject ID, assigned vial number, administration volume, route of administration, the volume of sterile water to add to resuspend if IP, and intervention schedule. After stroke surgery was performed, and intervention administered, correct administration was confirmed post-hoc.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\n \u003ch2\u003eThe REDCap Database\u003c/h2\u003e\n \u003cp\u003eUsing REDCap\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e49\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e, a project was designed for data entry for both the end users and the CC. Several repeatable forms were designed to capture data over four stages of SPAN. Data Access Groups were established so that each research laboratory received access to their own records only. The REDCap Alerts \u0026amp; Notifications application was used to trigger email signals to notify the CC when subjects were ready for randomization and to notify the research laboratory when randomization was completed. Custom reports were created in REDCap to track intervention accuracy, scheduled vs. actual surgery dates, and incoming rater scores for invalid or missing fields, among many others. These reports facilitated quality control and management.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\n \u003ch2\u003eMaster Intervention Spreadsheet (MISS)\u003c/h2\u003e\n \u003cp\u003eUsing Excel and the Visual Basic Macro Language, the MISS was designed to track inventory and assign vials during the manual randomization process (Supplementary Table\u0026nbsp;4). Using the MISS, CC staff tracked all inventory upon arrival at the CC, shipments to the research laboratories, and then tracked inventory as it was administered at the research laboratory. The MISS also calculated the dose to administer and provided an easy way to monitor for the shelf-life expiration of each vial. The MISS used relational logic to cross-check fields and had pre-programmed alerts (Supplementary Note 3 and Supplementary Table\u0026nbsp;5).\u003c/p\u003e\n \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e\n \u003ch2\u003eRandomization Tables\u003c/h2\u003e\n \u003cp\u003eThe randomization tables were created by the Statistical team using the pre-specified analysis plan, the stratification variables, and knowledge of the number of research laboratories and number of interventions. The tables and the MISS were devised in close collaboration. Rather than use a permuted block size, the randomization block sizes were unique to each stage and calculated from: multiples of the weekly subject enrollment at each research laboratory, the total sample size, the sex, the model, and the number of interventions being studied. Such variation in randomization block sizes allowed for equal distribution of interventions amongst sexes and models, adequate rotation of models and efficiencies in timing re-supply shipments of interventions to research laboratories while reducing loss (expired vials). Laboratory investigators were unaware of this approach to randomization block sizes.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec24\" class=\"Section2\"\u003e\n \u003ch2\u003eAvoiding Sources of Bias: Attrition, Selection and Performance\u003c/h2\u003e\n \u003cp\u003eEvery animal purchased for SPAN was tracked through the entire protocol (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eb). Upon arrival at the research laboratory, a barcoded ear tag (RapID, San Francisco, CA) was affixed (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ea and Supplementary Fig.\u0026nbsp;1c). Barcodes were read with a handheld scanner paired to a Bluetooth compatible device to ensure accurate ear tag entry. The research laboratory documented the ear tag color in the enrollment form. The ear tag color was verified during subsequent study procedures, for example during behavior video review and scoring. If a discrepant animal ID or tag color was noted, the animal record was flagged and investigated. Once enrolled, 100% of subjects were accounted for through the end of the study. To prevent surgeons from assigning interventions to individuals in a biased manner, we used group concealment\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e, described above. To assure unbiased assessment, the investigators performed all behavior ratings and image analysis blinded to any knowledge of intervention group assignment\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e43\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e. The methods are described above.\u003c/p\u003e\n \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e\n \u003ch2\u003eTranslational Relevance\u003c/h2\u003e\n \u003cp\u003eTo increase the translational relevance of the study, we included disease models that incorporated some of the key risk factors for stroke, including age, hypertension, and hyperglycemia. After a successful first stage using young, disease-free subjects, comorbidities were included in Stages 2 and 3. The comorbid models used were: aging mice, diet-induced obesity in mice, and spontaneously hypertensive rats. Aging mice, 15\u0026ndash;17 months at surgery were obtained from the National Institute on Aging (Aged Rodent Colonies, National Institute on Aging (nih.gov)). To induce obesity, C57BL/6J mice were fed a high-fat diet (EnvigoTeklad; TD.06414) for 12 weeks and were 15\u0026ndash;18 weeks old at surgery \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e52\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. Spontaneously hypertensive rats (Charles River: SHR/NCrl) were obtained from Charles River Laboratories\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e and were 15\u0026ndash;17 weeks old at surgery. To ensure even distribution of experience across all research laboratories and to avoid either research laboratory practice effects or seasonal effects we instituted several measures: The CC randomized the models assigned to the research laboratories in each stage; the randomization process (drawing of lots) was recorded using a video camera and archived for later review if needed. During Stages 2 and 3, each research laboratory was assigned to perform two of the models; no more than two research laboratories could perform a particular model simultaneously; research laboratories had to alternate between their assigned two models at least twice per stage to prevent an experience effect. By the end of the study, all models were performed in at least four different research laboratories. All these scheduling requirements assured that the study was free of practice effects, research laboratory preference/experience with one or another model, and seasonal effects.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e\n \u003ch2\u003eMRI Methods\u003c/h2\u003e\n \u003cp\u003eA protocol for MRI included quantitative imaging sequences for estimating the apparent diffusivity coefficient (ADC) and T2 relaxation rate. Data were acquired from Bruker scanners with field strengths across research laboratories, including 7T, 9.4T, or 11T. A multi-echo T2 sequence was used to estimate the T2 rate using relaxometry, with echo times rating from 10 to 100 milliseconds. A diffusion-weighted imaging sequence was used to estimate ADC, with b-values 0, 500, and 1000 s/mm\u003csup\u003e2\u003c/sup\u003e. After data acquisition, image files were uploaded to the LONI IDA for long-term storage and subsequent analysis. Using the Quantitative Imaging Toolkit\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e (QIT), we created an automated image analysis pipeline (code posted at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2203.05714\u003c/span\u003e\u003c/span\u003e) for generating quantitative reports and visualizations of individual animals, summarized\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ea. We segmented the brain using a deep learning neural network approach with a U-net architecture\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e, and we segmented lesions, cerebrospinal fluid, and normal-appearing brain tissue using a combination of thresholding and mathematical morphology. We estimated midline shift by isolating cerebrospinal fluid closest to the midline, computing the centroid, and measuring the relative distances to the boundaries of the brain. To visualize our data in a common reference frame, we aligned our data to a group-averaged scan using rigid registration with Advanced Normalization Tools (ANTs)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. We measured the volume of segmentation brain areas and lesion, aggregated the results in tabular form, and generated mosaic plots for later qualitative assessment.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e\n \u003ch2\u003eStatistical Methods\u003c/h2\u003e\n \u003cdiv id=\"Sec28\" class=\"Section4\"\u003e\n \u003ch2\u003eData Management\u003c/h2\u003e\n \u003cp\u003eData inconsistencies across forms in the database, or range-check errors, resulted in a data query from the CC to the research laboratory. Data queries (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ea and b) were initiated by the CC using the REDCap database. Research laboratory technicians and PIs were instructed to check REDCap data queries frequently. If the laboratory did not resolve the query initiated in REDCap within 48 hours, the CC alerted the technician to the query by email. If queries went unresolved for longer, the CC would request that the research laboratory PI follow up with their technician to resolve the query (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ec and d). Daily, the CC ran quality checks in REDCap to track study progress (surgery completion, dropouts, intervention administration, end of study) at each research laboratory and to reconcile video and MR image uploads into IDA. When inconsistencies in video and MRI uploads were identified, the CC contacted investigators at the research laboratory to upload the missing video or image (Supplementary Note 2).\u003c/p\u003e\n \u003c/div\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec29\" class=\"Section2\"\u003e\n \u003ch2\u003eStatistical Analysis\u003c/h2\u003e\n \u003cp\u003eInterventions were assessed using a multi-arm, multi-stage (MAMS) design based on the Generalized Dunnett\u0026rsquo;s test\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. The MAMS design provides large gains in efficiency over separate randomized trials of each treatment as it allows a shared control group. Futility (lower) and efficacy (upper) boundaries were calculated in advance. At each interim analysis, a Dunnett test statistic for the primary endpoint comparing each arm with the control arm as reference was calculated\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e. If the test statistic value fell below the futility boundary, the intervention was to be declared as non-promising, and enrollment into that arm was to stop. If the test statistic value fell above the efficacy boundary, the intervention was to be declared effective, and enrollment into that arm could be stopped. Otherwise, the intervention was declared promising and enrollment into that arm continued into the next stage.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003ePower and Sample Size Considerations\u003c/h3\u003e\n\u003cp\u003eTwo parallel multi-stage trials were performed with the corner test index as the primary endpoint after transformation. The power analysis used this variable to derive sample sizes: (A) the first trial included five drug intervention arms and one control arm, split equally between IP control and IV control under the assumption that there would be no difference between the two control treatments for the primary endpoint; (B) the second trial had 1 mechanical intervention arm and 1 control arm. We implemented the MAMS design with 4 stages using Triangular futility and efficacy boundaries\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. Three interim analyses were planned after 25, 50 and 75% recruitment with an equal allocation among arms. Based on a published study of the corner test in aged subjects, we estimated a mean of 0.55 and a standard deviation (SD) of 0.262 for the control arm\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. We then increased the assumed SD to 1.048 as a conservative scenario to account for predicted excessive variability among the research laboratories. Assuming that the transformed corner test index in different animal models follows independent Normal (\u0026micro;, \u0026sigma;\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e) distributions with known variance, we defined an interesting effect size (\u0026delta;) as a decrease on the corner test index of at least 50% to declare an intervention as effective, and a non-interesting effect size (\u0026delta;\u003csub\u003e0\u003c/sub\u003e) as a decrease of at most 6% to declare an intervention as ineffective. We chose 50% for the efficacy boundary based on a meta-analysis of prior preclinical trials in which the mean reported effect size was 50%\u003csup\u003e44\u003c/sup\u003e. We chose 6% for the futility boundary as this value was the effect size used to power several large, Phase 3 trials of interventions for acute ischemic stroke\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e60\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e. We tested a family of null hypotheses (H\u003csub\u003e0\u003c/sub\u003e) with each hypothesis stating that the intervention arm average is greater than or equal to the control arm average of the 30-day transformed corner test index. The family-wise error rate was defined as the probability of falsely rejecting one or more null hypotheses within this family at any trial stage with fixed value of 5%. The power of 90% was calculated under the least favorable configuration (LFC), which is defined as the probability of rejecting only one null hypothesis in any stage \u003csup\u003e\u003cspan class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e. We computed the minimum, maximum, and expected sample sizes under the null and LFC hypothesis based on Triangular boundaries\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e63\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. Total planned sample size was then adjusted assuming 10% animal death over 30 days.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e(ANTs)\u0026nbsp;\u0026nbsp;Advanced Normalization Tools\u003c/p\u003e\n\u003cp\u003e(ADC)\u0026nbsp; \u0026nbsp;\u0026nbsp;Apparent diffusivity coefficient\u003c/p\u003e\n\u003cp\u003e(CONSORT)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Consolidated Standards of Reporting Trials\u003c/p\u003e\n\u003cp\u003e(CC)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Coordinating Center\u003c/p\u003e\n\u003cp\u003e(\u0026delta;)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Effect size\u003c/p\u003e\n\u003cp\u003e(EAB)\u0026nbsp; \u0026nbsp;\u0026nbsp;External Advisory Board\u003c/p\u003e\n\u003cp\u003e(EOS)\u0026nbsp; \u0026nbsp;\u0026nbsp;End of study\u003c/p\u003e\n\u003cp\u003e(HPLC)\u0026nbsp;\u0026nbsp;High-performance liquid chromatography\u003c/p\u003e\n\u003cp\u003e(IDA)\u0026nbsp; \u0026nbsp; \u0026nbsp;Imaging Digital Archive\u003c/p\u003e\n\u003cp\u003e(\u0026mu;, \u0026sigma;\u003csup\u003e2\u003c/sup\u003e)\u0026nbsp; \u0026nbsp;Independent Normal distributions\u003c/p\u003e\n\u003cp\u003e(IACUC)\u0026nbsp;Institutional Animal Care and Use Committees\u003c/p\u003e\n\u003cp\u003e(ITT)\u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Intention-to-treat analysis\u003c/p\u003e\n\u003cp\u003e(\u0026delta;)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Interesting effect size\u003c/p\u003e\n\u003cp\u003e(ICC)\u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Intraclass correlation coefficient\u003c/p\u003e\n\u003cp\u003e(IP)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Intraperitoneal\u003c/p\u003e\n\u003cp\u003e(IV)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Intravenous\u003c/p\u003e\n\u003cp\u003e(LONI)\u0026nbsp; \u0026nbsp;Laboratory of Neuroimaging\u003c/p\u003e\n\u003cp\u003e(LFC)\u0026nbsp; \u0026nbsp; \u0026nbsp;Least favorable configuration\u003c/p\u003e\n\u003cp\u003e(L)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Left turns\u003c/p\u003e\n\u003cp\u003e(ml)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Milliliter\u003c/p\u003e\n\u003cp\u003e(MRI)\u0026nbsp; \u0026nbsp;\u0026nbsp;Magnetic resonance images\u003c/p\u003e\n\u003cp\u003e(MISS)\u0026nbsp; \u0026nbsp;Master Intervention Spreadsheet\u003c/p\u003e\n\u003cp\u003e(\u0026micro;L/g)\u0026nbsp; \u0026nbsp;\u0026nbsp;Microliters per gram\u003c/p\u003e\n\u003cp\u003e(MCAo)\u0026nbsp;Middle cerebral artery occlusion\u003c/p\u003e\n\u003cp\u003e(MAMS)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Multi-arm, multi-stage\u003c/p\u003e\n\u003cp\u003e(NIA)\u0026nbsp; \u0026nbsp; \u0026nbsp;National Institute of Aging\u003c/p\u003e\n\u003cp\u003e(NIH)\u0026nbsp; \u0026nbsp; \u0026nbsp;National Institute of Health\u003c/p\u003e\n\u003cp\u003e(NINDS)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;National Institute of Neurological Disorders\u003c/p\u003e\n\u003cp\u003e(NDS)\u0026nbsp; \u0026nbsp;\u0026nbsp;Neurological deficit score\u003c/p\u003e\n\u003cp\u003e(\u0026delta;\u003csub\u003e0\u003c/sub\u003e)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Non-interesting effect size\u003c/p\u003e\n\u003cp\u003e(\u0026mu;, \u0026sigma;\u003csup\u003e2\u003c/sup\u003e)\u0026nbsp; \u0026nbsp;Normal distribution\u003c/p\u003e\n\u003cp\u003e(H\u003csub\u003e0\u003c/sub\u003e)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Null hypotheses\u003c/p\u003e\n\u003cp\u003e(NIfTI)\u0026nbsp; \u0026nbsp;Neuroimaging Informatics Technology Initiative file\u003c/p\u003e\n\u003cp\u003e(PP)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Per-protocol\u003c/p\u003e\n\u003cp\u003e(PI)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Principal Investigators\u003c/p\u003e\n\u003cp\u003e(QIT)\u0026nbsp; \u0026nbsp; \u0026nbsp;Quantitative Imaging Toolkit\u003c/p\u003e\n\u003cp\u003e(REDCap)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Research Electronic Data Capture\u003c/p\u003e\n\u003cp\u003e(QC)\u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Quality control\u003c/p\u003e\n\u003cp\u003e(R)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Right turns\u003c/p\u003e\n\u003cp\u003e(s/mm\u003csup\u003e2\u003c/sup\u003e)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Seconds per millimeter squared\u003c/p\u003e\n\u003cp\u003e(SD)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Standard deviation\u003c/p\u003e\n\u003cp\u003e(SOP)\u0026nbsp; \u0026nbsp;\u0026nbsp;Standard operating procedure\u003c/p\u003e\n\u003cp\u003e(SC)\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Steering Committee\u003c/p\u003e\n\u003cp\u003e(SPAN)\u0026nbsp;\u0026nbsp;Stroke Preclinical Assessment Network\u003c/p\u003e\n\u003cp\u003e(TTC)\u0026nbsp; \u0026nbsp; \u0026nbsp;2,3,5 Triphenyl tetrazolium chloride\u003c/p\u003e\n\u003cp\u003e(w/v)\u0026nbsp; \u0026nbsp; \u0026nbsp;Weight/volume\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData and materials availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSOPs, experimental protocols, and MRI pipeline raw data and code will be available from the corresponding author or by request through the website, www.spannetwork.org.\u003c/p\u003e\n\u003cp\u003eAcknowledgments\u003c/p\u003e\n\u003cp\u003eWe thank Zachary Grodzinski, Liron Israel and Rameshwar Patel for their expertise and guidance in the pharmacology and packing of the drug interventions. We would like to thank the National Institute of Ageing for their generous donation of the aging mice. Finally, we would like to thank Novartis for their generous donation of Fingolimod.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSayre, F. \u0026amp; Riegelman, A. The reproducibility crisis and academic libraries. College \u0026amp; Research Libraries 79, 2 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErrington, T. M. \u003cem\u003eet al.\u003c/em\u003e Investigating the replicability of preclinical cancer biology. eLife 10, e71601, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7554/eLife.71601\u003c/span\u003e\u003cspan address=\"10.7554/eLife.71601\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerrin, S. Preclinical research: Make mouse studies work. Nature 507, 423\u0026ndash;425, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/507423a\u003c/span\u003e\u003cspan address=\"10.1038/507423a\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErrington, T. M., Denis, A., Perfito, N., Iorns, E. \u0026amp; Nosek, B. A. Challenges for assessing replicability in preclinical cancer biology. eLife 10, e67995, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7554/eLife.67995\u003c/span\u003e\u003cspan address=\"10.7554/eLife.67995\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVoelkl, B., Vogt, L., Sena, E. S. \u0026amp; Wurbel, H. Reproducibility of preclinical animal research improves with heterogeneity of study samples. PLoS Biol 16, e2003693, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pbio.2003693\u003c/span\u003e\u003cspan address=\"10.1371/journal.pbio.2003693\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBenjamin, D. J. \u003cem\u003eet al.\u003c/em\u003e Redefine statistical significance. Nat Hum Behav 2, 6\u0026ndash;10, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41562-017-0189-z\u003c/span\u003e\u003cspan address=\"10.1038/s41562-017-0189-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteckler, T. Editorial: preclinical data reproducibility for R\u0026amp;D - the challenge for neuroscience. SpringerPlus 4, 1, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/2193-1801-4-1\u003c/span\u003e\u003cspan address=\"10.1186/2193-1801-4-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNational Academies of Sciences, E. \u0026amp; Medicine. \u003cem\u003eReproducibility and Replicability in Science\u003c/em\u003e. (The National Academies Press, 2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVeening-Griffioen, D. H. \u003cem\u003eet al.\u003c/em\u003e Are some animal models more equal than others? A case study on the translational value of animal models of efficacy for Alzheimer's disease. European Journal of Pharmacology 859, 172524, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ejphar.2019.172524\u003c/span\u003e\u003cspan address=\"10.1016/j.ejphar.2019.172524\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO'Collins, V. E. \u003cem\u003eet al.\u003c/em\u003e Preclinical drug evaluation for combination therapy in acute stroke using systematic review, meta-analysis, and subsequent experimental testing. \u003cem\u003eJ Cereb Blood Flow Metab\u003c/em\u003e 31, 962\u0026ndash;975, doi:jcbfm2010184 [pii] \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/jcbfm.2010.184\u003c/span\u003e\u003cspan address=\"10.1038/jcbfm.2010.184\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSughrue, M. E. \u003cem\u003eet al.\u003c/em\u003e Data presentation in rodent stroke studies and the predictive value of confidence intervals. J Clin Neurosci 17, 11\u0026ndash;15, doi:S0967-5868(09)00340-3 [pii] \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jocn.2009.05.008\u003c/span\u003e\u003cspan address=\"10.1016/j.jocn.2009.05.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSena, E. S., Currie, G. L., McCann, S. K., Macleod, M. R. \u0026amp; Howells, D. W. Systematic reviews and meta-analysis of preclinical studies: why perform them and how to appraise them critically. J Cereb Blood Flow Metab 34, 737\u0026ndash;742, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/jcbfm.2014.28\u003c/span\u003e\u003cspan address=\"10.1038/jcbfm.2014.28\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan der Worp, H. B. \u003cem\u003eet al.\u003c/em\u003e Can animal models of disease reliably inform human studies? PLoS medicine 7, e1000245, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pmed.1000245\u003c/span\u003e\u003cspan address=\"10.1371/journal.pmed.1000245\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacleod, M. R. \u003cem\u003eet al.\u003c/em\u003e Reprint: Good laboratory practice: preventing introduction of bias at the bench. Int J Stroke 4, 3\u0026ndash;5, doi:IJS241 [pii] \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/j.1747-4949.2009.00241.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1747-4949.2009.00241.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBix, G. J. \u003cem\u003eet al.\u003c/em\u003e Uncovering the Rosetta Stone: Report from the First Annual Conference on Key Elements in Translating Stroke Therapeutics from Pre-Clinical to Clinical. \u003cem\u003eTransl Stroke Res\u003c/em\u003e 9, 258\u0026ndash;266, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s12975-018-0628-9\u003c/span\u003e\u003cspan address=\"10.1007/s12975-018-0628-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBosetti, F. \u003cem\u003eet al.\u003c/em\u003e Translational Stroke Research: Vision and Opportunities. Stroke 48, 2632\u0026ndash;2637, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/strokeaha.117.017112\u003c/span\u003e\u003cspan address=\"10.1161/strokeaha.117.017112\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLyden, P., Buchan, A., Boltze, J., Fisher, M. \u0026amp; Consortium*, S. X. Top Priorities for Cerebroprotective Studies-A Paradigm Shift: Report From STAIR XI. Stroke 52, 3063\u0026ndash;3071, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/STROKEAHA.121.034947\u003c/span\u003e\u003cspan address=\"10.1161/STROKEAHA.121.034947\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeuerstein, G. Z. \u003cem\u003eet al.\u003c/em\u003e Missing steps in the STAIR case: a Translational Medicine perspective on the development of NXY-059 for treatment of acute ischemic stroke. J Cereb Blood Flow Metab 28, 217\u0026ndash;219, doi:9600516 [pii] \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/sj.jcbfm.9600516\u003c/span\u003e\u003cspan address=\"10.1038/sj.jcbfm.9600516\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSavitz, S. I. A critical appraisal of the NXY-059 neuroprotection studies for acute stroke: a need for more rigorous testing of neuroprotective agents in animal models of stroke. Exp Neurol 205, 20\u0026ndash;25 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLyden, P. \u0026amp; Lapchak, P. Sisyphus and translational stroke research. \u003cem\u003eSci Transl Med\u003c/em\u003e 4, 156ps120, \u003cdiv class=\"ExternalRefDOI\"\u003edoi:4/156/156ps20\u003c/div\u003e [pii] 10.1126/scitranslmed.3005083 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLyden, P. D. \u003cem\u003eet al.\u003c/em\u003e The Stroke Preclinical Assessment Network: Rationale, Design, Feasibility, and Stage 1 Results. Stroke 53, 1802\u0026ndash;1812, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/STROKEAHA.121.038047\u003c/span\u003e\u003cspan address=\"10.1161/STROKEAHA.121.038047\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDirnagl, U. \u0026amp; Group, M. o. t. M.-S. O. P. Standard operating procedures (SOP) in experimental stroke research: SOP for middle cerebral artery occlusion in the mouse. Nature Precedings, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/npre.2012.3492.3\u003c/span\u003e\u003cspan address=\"10.1038/npre.2012.3492.3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorais, A. \u003cem\u003eet al.\u003c/em\u003e Embracing Heterogeneity in The Multicenter Stroke Preclinical Assessment Network (SPAN) Trial. Stroke 54, 620\u0026ndash;631 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrawford, K. L., Neu, S. C. \u0026amp; Toga, A. W. The Image and Data Archive at the Laboratory of Neuro Imaging. Neuroimage 124, 1080\u0026ndash;1083, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2015.04.067\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2015.04.067\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJaki, T. F., Pallmann, P. S. \u0026amp; Magirr, D. The R package MAMS for designing multi-arm multi-stage clinical trials. Journal of Statistical Software (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWason, J. M. \u0026amp; Trippa, L. A comparison of Bayesian adaptive randomization and multi-stage designs for multi-arm clinical trials. Stat Med 33, 2206\u0026ndash;2221, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/sim.6086\u003c/span\u003e\u003cspan address=\"10.1002/sim.6086\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJaki, T., Magirr, D. \u0026amp; Pallmann, P. MAMS: designing multi-arm multi-stage studies, 2014. \u003cem\u003eURL http://CRAN. R-project. org/package = MAMS. R package version 0.3\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumville, J. C., Torgerson, D. J. \u0026amp; Hewitt, C. E. Reporting attrition in randomised controlled trials. Bmj 332, 969\u0026ndash;971, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmj.332.7547.969\u003c/span\u003e\u003cspan address=\"10.1136/bmj.332.7547.969\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGroup, R. i. S. W. Improving the Quality of NINDS-Supported Preclinical and Clinical Research through Rigorous Study Design and Transparent Reporting. (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRanganathan, P., Pramesh, C. S. \u0026amp; Aggarwal, R. Common pitfalls in statistical analysis: Intention-to-treat versus per-protocol analysis. Perspect Clin Res 7, 144\u0026ndash;146, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4103/2229-3485.184823\u003c/span\u003e\u003cspan address=\"10.4103/2229-3485.184823\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBegg, C. \u003cem\u003eet al.\u003c/em\u003e Improving the quality of reporting of randomized controlled trials. The CONSORT statement. JAMA 276, 637\u0026ndash;639 (1996).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManwani, B. \u003cem\u003eet al.\u003c/em\u003e Functional recovery in aging mice after experimental stroke. Brain Behav Immun 25, 1689\u0026ndash;1700, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.bbi.2011.06.015\u003c/span\u003e\u003cspan address=\"10.1016/j.bbi.2011.06.015\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, X. \u003cem\u003eet al.\u003c/em\u003e Chronic behavioral testing after focal ischemia in the mouse: functional recovery and the effects of gender. Exp Neurol 187, 94\u0026ndash;104, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.expneurol.2004.01.004\u003c/span\u003e\u003cspan address=\"10.1016/j.expneurol.2004.01.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2004).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchaar, K. L., Brenneman, M. M. \u0026amp; Savitz, S. I. Functional assessments in the rodent stroke model. Exp Transl Stroke Med 2, 13, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/2040-7378-2-13\u003c/span\u003e\u003cspan address=\"10.1186/2040-7378-2-13\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBederson, J. B. \u003cem\u003eet al.\u003c/em\u003e Rat Middle Cerebral-Artery Occlusion - Evaluation of the Model and Development of A Neurologic Examination. Stroke 17, 472 (1986).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBederson, J. B. \u003cem\u003eet al.\u003c/em\u003e Evaluation of 2,3,5-triphenyltetrazolium chloride as a stain for detection and quantification of experimental cerebral infarction in rats. Stroke 17, 1304 (1986).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHunter, J. S. The Exponentially Weighted Moving Average. Journal of Quality Technology 18, 203\u0026ndash;210, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/00224065.1986.11979014\u003c/span\u003e\u003cspan address=\"10.1080/00224065.1986.11979014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (1986).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAgrafiotis, D. K. \u003cem\u003eet al.\u003c/em\u003e Risk-based Monitoring of Clinical Trials: An Integrative Approach. Clinical Therapeutics 40, 1204\u0026ndash;1212, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.clinthera.2018.04.020\u003c/span\u003e\u003cspan address=\"10.1016/j.clinthera.2018.04.020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTantsyura, V. \u003cem\u003eet al.\u003c/em\u003e Risk-Based Monitoring: A Closer Statistical Look at Source Document Verification, Queries, Study Size Effects, and Data Quality. Therapeutic Innovation \u0026amp; Regulatory Science 49, 903\u0026ndash;910, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/2168479015586001\u003c/span\u003e\u003cspan address=\"10.1177/2168479015586001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRewell, S. S. \u003cem\u003eet al.\u003c/em\u003e Evolution of ischemic damage and behavioural deficit over 6 months after MCAo in the rat: Selecting the optimal outcomes and statistical power for multi-centre preclinical trials. PLoS One 12, e0171688, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pone.0171688\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0171688\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLlovera, G. \u003cem\u003eet al.\u003c/em\u003e Results of a preclinical randomized controlled multicenter trial (pRCT): Anti-CD49d treatment for acute brain ischemia. Sci Transl Med 7, 299ra121, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/scitranslmed.aaa9853\u003c/span\u003e\u003cspan address=\"10.1126/scitranslmed.aaa9853\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGettayacamin, M. \u0026amp; Retnam, L. AAALAC International Standards and Accreditation Process. Toxicological Research 33, 183\u0026ndash;189, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.5487/TR.2017.33.3.183\u003c/span\u003e\u003cspan address=\"10.5487/TR.2017.33.3.183\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacleod, M. R. \u003cem\u003eet al.\u003c/em\u003e Good laboratory practice: preventing introduction of bias at the bench. Stroke 40, e50-52, doi:STROKEAHA.108.525386 [pii] \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/STROKEAHA.108.525386\u003c/span\u003e\u003cspan address=\"10.1161/STROKEAHA.108.525386\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSena, E., van der Worp, H. B., Howells, D. \u0026amp; Macleod, M. How can we improve the pre-clinical development of drugs for stroke? Trends Neurosci 30, 433\u0026ndash;439, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.tins.2007.06.009\u003c/span\u003e\u003cspan address=\"10.1016/j.tins.2007.06.009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLozano, R. \u003cem\u003eet al.\u003c/em\u003e Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 380, 2095\u0026ndash;2128, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(12)61728-0\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(12)61728-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Winkle, J. A. \u003cem\u003eet al.\u003c/em\u003e Concurrent middle cerebral artery occlusion and intra-arterial drug infusion via ipsilateral common carotid artery catheter in the rat. J Neurosci Methods 213, 63\u0026ndash;69, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jneumeth.2012.12.004\u003c/span\u003e\u003cspan address=\"10.1016/j.jneumeth.2012.12.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoyal, M. \u003cem\u003eet al.\u003c/em\u003e Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials. Lancet 387, 1723\u0026ndash;1731, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(16)00163-X\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(16)00163-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEnglish, J. R., Krishnamurthi, M. \u0026amp; Sastri, T. Quality monitoring of continuous flow processes. Computers \u0026amp; industrial engineering 20, 251\u0026ndash;260 (1991).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarris, P. A. \u003cem\u003eet al.\u003c/em\u003e The REDCap consortium: Building an international community of software platform partners. Journal of biomedical informatics 95, 103208 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarris, P. A. \u003cem\u003eet al.\u003c/em\u003e Research electronic data capture (REDCap)\u0026mdash;a metadata-driven methodology and workflow process for providing translational research informatics support. Journal of biomedical informatics 42, 377\u0026ndash;381 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrossley, N. A. \u003cem\u003eet al.\u003c/em\u003e Empirical evidence of bias in the design of experimental stroke studies: a metaepidemiologic approach. Stroke 39, 929\u0026ndash;934, doi:STROKEAHA.107.498725 [pii] \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/STROKEAHA.107.498725\u003c/span\u003e\u003cspan address=\"10.1161/STROKEAHA.107.498725\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, C.-Y. \u0026amp; Liao, J. K. in \u003cem\u003emTOR: Methods in Molecular Biology\u003c/em\u003e Vol.\u0026nbsp;821 (ed T. Weichhart) 421\u0026ndash;433 (Springer, 2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSurwit, R. S., Kuhn, C. M., Cochrane, C., McCubbin, J. A. \u0026amp; Feinglos, M. N. Diet-induced type II diabetes in C57BL/6J mice. Diabetes 37, 1163\u0026ndash;1167 (1988).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang, Y., Smith Jr, D. L., Keating, K. D., Allison, D. B. \u0026amp; Nagy, T. R. Variations in body weight, food intake and body composition after long-term high‐fat diet feeding in C57BL/6J mice. Obesity 22, 2147\u0026ndash;2155 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOkamoto, K. \u0026amp; Aoki, K. Development of a strain of spontaneously hypertensive rats. Jpn Circ J 27, 282\u0026ndash;293, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1253/jcj.27.282\u003c/span\u003e\u003cspan address=\"10.1253/jcj.27.282\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (1963).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCabeen, R., Laidlaw, D. \u0026amp; Toga, A. \u003cem\u003eQuantitative Imaging Toolkit: Software for Interactive 3D Visualization, Data Exploration, and Computational Analysis of Neuroimaging Datasets\u003c/em\u003e, 2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRonneberger, O., Fischer, P. \u0026amp; Brox, T. in \u003cem\u003eInternational Conference on Medical Image Computing and Computer-Assisted Intervention.\u003c/em\u003e 234\u0026ndash;241 (Springer International Publishing).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAvants, B., Tustison, N. \u0026amp; Song, G. Advanced Normalization Tools: V1.0. doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.54294/uvnhin\u003c/span\u003e\u003cspan address=\"10.54294/uvnhin\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManwani, B. \u003cem\u003eet al.\u003c/em\u003e Differential effects of aging and sex on stroke induced inflammation across the lifespan. Exp Neurol 249, 120\u0026ndash;131, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.expneurol.2013.08.011\u003c/span\u003e\u003cspan address=\"10.1016/j.expneurol.2013.08.011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShuaib, A. \u003cem\u003eet al.\u003c/em\u003e NXY-059 for the treatment of acute ischemic stroke. N Engl J Med 357, 562\u0026ndash;571 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZivin, J. A. \u003cem\u003eet al.\u003c/em\u003e Effectiveness and safety of transcranial laser therapy for acute ischemic stroke. Stroke 40, 1359\u0026ndash;1364, doi:STROKEAHA.109.547547 [pii] \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/STROKEAHA.109.547547\u003c/span\u003e\u003cspan address=\"10.1161/STROKEAHA.109.547547\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDunnett, C. W. in \u003cem\u003eDesign of Experiments: Ranking and Selection\u003c/em\u003e (ed Tamhane AC Santer T.J.) 47\u0026ndash;66 (Marcel Dekker, 1984).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhitehead, J. \u003cem\u003eThe design and analysis of sequential clinical trials\u003c/em\u003e. (John Wiley \u0026amp; Sons, 1997).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO'Brien, P. C. \u0026amp; Fleming, T. R. A multiple testing procedure for clinical trials. Biometrics 35, 549\u0026ndash;556 (1979).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMagirr, D., Stallard, N. \u0026amp; Jaki, T. Flexible sequential designs for multi-arm clinical trials. Stat Med 33, 3269\u0026ndash;3279, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/sim.6183\u003c/span\u003e\u003cspan address=\"10.1002/sim.6183\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMagirr, D., Jaki, T. \u0026amp; Whitehead, J. A generalized Dunnett test for multi-arm multi-stage clinical studies with treatment selection. Biometrika 99, 494\u0026ndash;501 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiehl, K., Hull, R. \u0026amp; Morton, D. Pfi ster R, Rabemampianina Y, Smith D, Vidal JM, van de Vorstenbosch C (2001) A good practice guide to the administration of substances and removal of blood, including routes and volumes. Journal of Applied Toxicology 21, 15\u0026ndash;23.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-3054771/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3054771/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eScience faces a reproducibility crisis, and public trust in science suffers when large clinical trials—qualified by promising preclinical studies—fail. While some clinical trial designs may have been inadequate, preclinical assessments of disease interventions have lacked key elements of rigor: treatment concealment, randomization, blinded outcomes, pre-specified and adequate sample sizes, and models including co-morbidities. To demonstrate feasibility and practicality of enhanced rigor in preclinical assessment, a six-laboratory network was designed that implemented rigorous study elements, using acute ischemic stroke for demonstration. This network enrolled 2615 animals in five different models and implemented a multi-stage, multi-arm statistical design that sequentially eliminated candidate interventions during interim analyses. The methods included centralized intervention packaging, randomization, data quality assessment, and data archiving. Blinded analysis of 9,274 video-recorded behavior tasks and 3,652 magnetic resonance images were evaluated. All tools and protocols are presented and could be adapted to preclinical assessment in other disease areas.\u003c/p\u003e","manuscriptTitle":"Methods for randomized, blinded, controlled evaluation of putative disease interventions in multi-laboratory, preclinical assessment networks","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-10 18:05:29","doi":"10.21203/rs.3.rs-3054771/v1","editorialEvents":[],"status":"published","journal":{"display":false,"email":"[email protected]","identity":"lab-animal","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"laban","sideBox":"Learn more about [Lab Animal](http://www.nature.com/laban/)","snPcode":"","submissionUrl":"","title":"Lab Animal","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"f9a4d67d-9e25-4bd0-9f99-ac6fbd5f866b","owner":[],"postedDate":"January 10th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":22575890,"name":"Scientific community and society/Scientific community/Research management"},{"id":22575891,"name":"Health sciences/Medical research/Preclinical research"},{"id":22575892,"name":"Biological sciences/Drug discovery"}],"tags":[],"updatedAt":"2026-02-19T08:07:20+00:00","versionOfRecord":{"articleIdentity":"rs-3054771","link":"https://doi.org/10.1038/s41684-026-01683-z","journal":{"identity":"lab-animal","isVorOnly":false,"title":"Lab Animal"},"publishedOn":"2026-02-18 05:00:00","publishedOnDateReadable":"February 18th, 2026"},"versionCreatedAt":"2025-01-10 18:05:29","video":"","vorDoi":"10.1038/s41684-026-01683-z","vorDoiUrl":"https://doi.org/10.1038/s41684-026-01683-z","workflowStages":[]},"version":"v1","identity":"rs-3054771","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3054771","identity":"rs-3054771","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}