Marcio Augusto Diniz

BACKGROUND: Negative clinical trials foster therapeutic nihilism and public skepticism, especially if they follow optimistic preclinical results. One way to increase the likelihood of successful clinical trials is to enhance the preclinical testing of new treatments. Recently, multicenter preclinical testing networks have gained attention as one tool to enhance rigor and reduce the likelihood of erroneously recommending a new candidate treatment for advancement to clinical trial testing. METHODS: The SPAN (Stroke Preclinical Assessment Network) consists of 6 research laboratories performing the experiments and 1 coordinating center at which treatments are prepared in a blinded manner, subjects are randomized, and statistical analysis is performed. We conducted a rigorous trial that included both sexes, multiple comorbidities, randomization, concealment of treatment status, blinded outcome assessments, prespecified sample sizes, and comprehensive data tracking. Five promising treatments were evaluated: BPN-27332, a lipoxygenase inhibitor; uric acid, a potent natural scavenger of free radicals; tatCN19o, a peptide inhibitor of activated CaMKII (Ca ²⁺ /CaM–dependent protein kinase II); GSK2256294, an inhibitor of sEH (soluble epoxide hydrolase); and NVX-058, an oxygen-carrying compound. A behavioral battery was selected as the end point for the primary analysis. The primary analysis used the modified intention-to-treat population. RESULTS: In total, 1768 rodents (882 females and 886 males) were enrolled: 441 young healthy mice, 444 aging mice, 450 mice with obesity-induced hyperglycemia and metabolic syndrome, and 433 spontaneously hypertensive rats. Data missing due to death before 30 days of follow-up were imputed from the same treatment group, animal model, sex, and stage. Although none of the candidates met the prespecified efficacy boundary, GSK2256294 showed an odds ratio for benefit of 1.28 (95% CI, 0.99–1.65), which was significant at P <0.05. CONCLUSIONS: The second SPAN study reaffirms the importance and feasibility of rigorous, multilaboratory preclinical testing networks to potentially reduce false positives in the development pipeline.

Background: Depression affects nearly half of patients with heart failure (HF), which is associated with increased morbidity and reduced health-related quality of life (HRQoL). This study evaluated the comparative effectiveness of behavioral activation (BA) versus antidepressant medication (MEDS) among patients with depression and HF who adhered to study interventions.Methods: This analysis is based on a pragmatic randomized comparative effectiveness trial conducted from 2018 to 2022, including 1-year follow-up. 416 participants diagnosed with HF and a DSM-5 depressive disorder were randomized to BA or MEDS, and data were analyzed as intention-to-treat.Outcomes: The primary outcome was depressive symptom severity (PHQ-9) at 6 months, and the secondary outcomes were physical and mental HRQoL (SF-12v2-PC and SF-12v2-MC, respectively) and HF-specific HRQoL (Kansas City Cardiomyopathy Questionnaire; KCCQ) overall and clinical scores (KCCQ-OS and KCCQ-CS), at 3, 6, and 12 months. High adherence was defined as completion of greater than 75 to 100% of intervention components. Overall adherence rates were similar between BA and MEDS. At 6 months, among patients with >75-100% adherence who followed treatment as directed in both arms, BA demonstrated statistically significantly higher Mental HRQoL (SF-12v2-MC; p=0.023) and higher HF-specific HRQoL (KCCQ-OS; p<0.001 and KCCQ-CS; p<0.001). There were no significant differences in depressive symptoms or physical HRQoL at 6 months.Interpretation: Both BA and MEDS were equally effective treatments for depression in HF. However, compared to MEDS, BA provided the added benefit of significantly greater improvement in both mental and HF-specific HRQoL in highly adherent patients.

Science faces a reproducibility crisis, and public trust in science declines when large clinical trials, which had been qualified by promising preclinical studies, fail. While some clinical trial designs may have been inadequate, preclinical assessments of disease interventions might have lacked key elements of rigor such as treatment concealment, randomization, blinded outcomes, prespecified and adequate sample sizes, and models including comorbidities. Here, to demonstrate feasibility and practicality of enhanced rigor in preclinical assessment, we designed a six-laboratory network that implemented rigorous study elements, using acute ischemic stroke for demonstration. This network enrolled 2,615 rodents in 5 different models and implemented a multistage, multiarm 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 behavioral 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.

Background: Stroke is a sexually dimorphic disease, with different risk factors, incidence, outcomes, and treatment responses in men and women. While sex differences have been documented in preclinical studies, these findings often come from single‐site studies with small sample sizes and require validation across diverse research settings. Methods: We used data from the SPAN (Stroke Preclinical Assessment Network), a randomized, placebo‐controlled, blinded, multilaboratory trial, to determine if sex differences in neurological outcomes are present in preclinical stroke models. We analyzed data from 665 stroke animals treated with saline, including young mice, diet‐induced obese mice, aging mice, young rats, and spontaneously hypertensive rats. We compared the corner test index and brain morphology between the sexes using linear random effect models and assessed the mortality rate using Cox proportional hazard regression models. Results: No significant sex differences were found in neurological outcome measured with the corner test on either day 7 or day 30 after stroke, regardless of the mouse or rat stroke model used. Additionally, female and male mice exhibited similar infarct sizes on day 2 magnetic resonance imaging and on brain atrophy measures on day 30 after stroke, indicating a lack of sex differences in brain injury. Similarly, no sex differences were observed in acute or chronic sensorimotor or tissue outcomes in young rats. In 1 subanalysis, sex differences were seen in the spontaneously hypertensive rats cohort. Female rats exhibited a higher corner test index on day 30 than males, indicating more severe sensorimotor injury. Conclusions: In this multicenter preclinical study, we did not detect sex differences in stroke outcomes in mice, although sex differences in behavioral outcomes were observed in spontaneously hypertensive rats. These findings highlight that sex differences may be model‐specific and subtle, emphasizing the need for methodological consistency and thoughtful inclusion of diverse animal models in translational stroke research to better understand if sex‐specific responses contribute to stroke outcomes.

Preclinical stroke research faces a critical translational gap, with animal studies failing to reliably predict clinical efficacy. To address this, the field is moving toward rigorous, multicenter preclinical randomized controlled trials (mpRCTs) that mimic phase 3 clinical trials in several key components. This collective statement, derived from experts involved in mpRCTs, outlines considerations for designing and executing such trials. mpRCTs offer advantages such as increased sample sizes, robust statistical design, incorporation of heterogeneity, and standardized protocols, but they face challenges in finding the right balance between standardization and heterogeneity, appropriate stroke model selection, and outcome measures, as well as the implementation of complex network infrastructure. We discuss the importance of rigorous study design, including appropriate stroke models, representation of biological variables and comorbidities, functional outcome readouts, and handling of attrition and mortality. Statistical considerations such as adaptive sequential designs, covariate adjustments, and appropriate handling of missing data are also addressed. The integration of machine learning, the implementation of common data elements, and the selection of appropriate therapeutic candidates are crucial for maximizing the efficiency and utility of mpRCTs. Furthermore, the transition toward mpRCT platforms, akin to clinical trial platforms, holds promise for facilitating continuous evaluation of therapies. Finally, we discuss data-sharing practices and the collateral benefits of mpRCTs, emphasizing their potential to improve preclinical stroke research and bridge the translational gap. Altogether, we hope that this article will serve as a starting point for a lasting debate on the future of stroke mpRCTs and their evolution toward a universally accepted set of principles.

Supplementary Table S2 provides information about the number of patients with missing adverse events at each timepoint for anastrozole treatment.

Supplementary Table S1 outlines all predictors employed to predict treatment discontinuation, including the predictor variable name, variable description, variable type, and whether the data was collected over time or only at baseline.