A multi-laboratory preclinical trial to assess treatment candidates for acute ischemic stroke
Data files
Aug 30, 2023 version files 1.26 MB
Abstract
Human diseases may be modeled in animals to allow preclinical assessment of putative new clinical interventions. Recent, highly publicized failures of large clinical trials called into question the rigor, design, and value of preclinical assessment. We established the Stroke Preclinical Assessment Network (SPAN) to design and implement a randomized, controlled, blinded, multi-laboratory network for the rigorous assessment of candidate stroke treatments combined with intravascular thrombectomy. Futility boundaries in a Multi-Arm Multi-Stage statistical design aimed to exclude less effective interventions, and efficacy boundaries allowed early declaration of success after each of four sequential stages. Six independent research laboratories performed a standard focal cerebral ischemic insult in five animal models that included equal numbers of males and females: young mice, young rats, aging mice, mice with diet-induced obesity, and spontaneously hypertensive rats. The laboratories adhered to a common protocol and efficiently enrolled 2,615 animals with full data completion and comprehensive animal tracking. SPAN successfully implemented treatment masking, randomization, pre-randomization inclusion and exclusion criteria, and blinded assessment of outcomes. The SPAN design and infrastructure provide an effective approach that could be used in similar preclinical, multi-laboratory studies in other disease areas, and should help improve reproducibility in translational science.
README
A multi-laboratory preclinical trial to assess treatment candidates for acute ischemic stroke
All data needed for the graphs published in Science Translational Medicine are provided.
Description of the data and file structure
Tables contain the results of the corner test, the MR imaging organized by subject ID. Treatment group data is available in the feasibility file. These are contained in the following files:
dt_feasibility_stm_paper_2023-08-21.xls
dt_outcome_corner_stm_paper_2023-08-21.xls
dt_outcome_mri_stm_paper 2023-08-21.xls
report_stm_paper_2023-07-29.html
The first 3 files are Excel files that contain data listings in tab 1 and data definition dictionay in tab 2. The treatment group variables can be found in the first file and applied to the remaining files by matching on the animal identity variable 'enro_animal_id'. The variable 'txas_tx_group' contains a code that gives the treatment group to which the subject was assigned. The variable 'txas_tx_group_actual' gives a code for the treatment group actually recieved. The corresponding drug treatment names can be found in tab 2 and in the main paper itself.
The final html file is a report used to define the various study populations, including the enrolled (pop_enro), the inelegible to be randomized, the intention to treat (pop_itt), the excluded animals, the procedural dropouts, the modified intention to treat (pop_mitt), the partial treatments, the per protocol treated animals, and the loss to followup (died) animals.
Sharing/Access information
Links to other publicly accessible locations of the data:
www.spannetwork.org is a repository of all SPAN SOPs
Code/Software
(Citations and figures can be found in the SPAN publication)
SPAN selected a recently developed MAMS design as it is more efficient than multiple individual trials. Validated, user-friendly MAMS R-code has been published (56) based on a generalized Dunnett’s procedure (33). SPAN implemented the MAMS design with four stages using predefined futility and efficacy boundaries( 33). Stages 1 – 3 ended with planned interim analyses performed after 25, 50 and 75% recruitment and one final analysis of the entire dataset after Stage 4. The Statistical Analysis Plans (see Supplemental Material) provide a detailed description of the planned analysis for each endpoint (30).
The primary outcome used in the MAMS was the 30-day corner test. To estimate needed sample size, simulations were performed using the MAMS R-package (92). From prior reports of the corner test in aged mice, we estimated mean of 0.55 and standard deviation (SD) of 0.262 turning index (left turns/total turns) for the control arm(47). We inflated the estimated standard deviation to 1.048 as a conservative assumption to account for probable variability among SPAN research laboratories. The power of 90% was calculated under the least favorable configuration, which was defined as the probability of rejecting only one null hypothesis in any stage(93). Total sample size was then adjusted assuming 10% animal death over 30 days.
At the end of each stage, an interim analysis of the primary outcome was performed that included all subjects up to that analysis. Each treatment was compared to the appropriate control group (pooled IV/IP or RIC sham), adjusting for laboratory, sex, baseline corner test index, animal model (stages 2-4) and whether the subjects remained under anesthesia during MCA occlusion (Stage 1 only). Stopping rules included an estimated treatment effect size ((intervention – control)/sd) greater than an upper limit of 50% (change of 0.5 x 0.55 = 0.275) and less than a lower limit of 6% (change 0.06 x 0.55 = 0.033) treatment effect size.” The projected boundaries at each interim analysis comprise a ‘triangle’ of upper and lower criteria for the five drug treatments compared against a pooled IV and IP control group (Fig. 3C). The RIC intervention was compared against the RIC sham group (Fig. 3D). The Steering Committee reviewed each interim analysis to decide which models to use and which interventions to continue in the next stage. At the end of Stage 3, the SC decided to study the one remaining intervention in young rats since all prior testing had been done in mice, other than SHRs. The Steering Committee making these decisions remained blinded to the treatment arm identity throughout all four stages.
Methods
SPAN consists of six research laboratories to conduct preclinical stroke modeling and a coordinating center to mask drugs, conduct centralized randomization, assess blinded outcomes, collect data, and perform statistics. The aim of SPAN was to determine whether such rigor is feasible and practical on a large scale in a multi-laboratory controlled, randomized preclinical study of six investigator-initiated putative interventions for acute ischemic stroke, selected through a rigorous NIH-peer review. Here we report the primary results of SPAN.
Usage notes
Tables contain the results of the corner test and the MR imaging organized by subject ID. Treatment group data is available in the feasibility file. These are contained in the following files:
dt_feasibility_stm_paper 2023-08-21
dt_outcome_corner_stm_paper 2023-08-21
dt_outcome_mri_stm_paper 2023-08-21
report_stm_paper 2023-07-29
The first 3 files are Excel files that contain data listings in tab 1 and data definition dictionary in tab 2. The treatment group variables can be found in the first file and applied to the remaining files by matching on the animal identity variable "enro_animal_id. The variable 'txas_tx_group' contains a code that gives the treatment group to which the subject was assigned. The variable 'txas_tx_group_actual' gives a code for the treatment group actually received. The corresponding drug treatment names can be found in tab 2 and in the main paper itself. The final html file is a report used to define the various study populations, including the enrolled (pop_enro), the inelegible to be randomized, the intention to treat (pop_itt), the excluded animals, the procedural dropouts, the modified intention to treat (pop_mitt), the partial treatments, the per protocol treated animals, and the loss to followup (died) animals.