Therapeutic hypothermia (HT) is standard care for term infants with hypoxic-ischaemic (HI) encephalopathy. However, the efficacy of HT in preclinical models, such as the Vannucci model of unilateral HI in the newborn rat, is often greater than that reported from clinical trials. Here, we report a meta-analysis of data from every experiment in a single laboratory, including pilot data, examining the effect of HT in the Vannucci model. Across 21 experiments using 106 litters, median (95% CI) hemispheric area loss was 50.1% (46.0-51.9%; n = 305) in the normothermia group, and 41.3% (35.1-44.9%; n = 317) in the HT group, with a bimodal injury distribution. Median neuroprotection by HT was 17.6% (6.8-28.3%), including in severe injury, but was highly-variable across experiments. Neuroprotection was significant in females (p < 0.001), with a non-significant benefit in males (p = 0.07). Animals representing the median injury in each group within each litter (n = 277, 44.5%) were also analysed using formal neuropathology, which showed neuroprotection by HT throughout the brain, particularly in females. Our results suggest an inherent variability and sex-dependence of the neuroprotective response to HT, with the majority of studies in the Vannucci model vastly underpowered to detect true treatment effects due to the distribution of injury. The use of animal research has dramatically advanced human knowledge of physiology, pathology, and sciencebased medicine 1. However, despite centuries of improvement in the methodological and ethical approaches to experiments involving animals 1 , preclinical research across all fields of medicine has also significantly underperformed with regards to the translation of robust treatments for complex diseases in humans 2. For perinatal asphyxia and subsequent hypoxic-ischaemic encephalopathy (HIE), just one treatment so far, therapeutic hypothermia (HT), has emerged from the preclinical research to provide a robust treatment effect in term newborns with moderate-to-severe neurological injury 3. A recent meta-analysis of those clinical trials found that the numbers needed to treat (NNT) were 11 to reduce mortality, and 8 to reduce major neurodevelopmental disability 3. However, HT is not universally neuroprotective. Meta-analyses of the original large clinical trials of HT found that 40-50% of treated infants still experienced a poor outcome (death or severe disability) 4 , though more recent trials suggest it is now around 30% 5. In the most widely-researched model of neonatal hypoxic-ischaemic (HI) brain injury, the Vannucci model of unilateral HI, our group and others have traditionally found that HT provides around 40% reduction in neuropathology score and tissue loss 6-9. This neuroprotective effect of HT was corroborated in larger animal models 10-12 ,