Nurses' Knowledge, Attitudes, Practices and Perceived Barriers to Early Mobilization of Patients in Intensive Care Units in Southeastern Iran.

In many fields of medical research, drugs that show promising results in preclinical studies in animals frequently fail to do the same in human clinical trials (). This ‘translation failure’ is one of the biggest challenges in biomedical research today: it is estimated that around two-thirds to 95% of therapeutics found to be safe and effective during animal testing fail when tested in humans (; ; ; ). Translatability has been defined as ‘the ability to apply research discoveries from experimental models to applications that directly benefit humans’ (). Reasons for low translatability are multifaceted. However, the pervasiveness of suboptimal study design, analysis, and reporting, potentially resulting in a lack of reproducibility (i.e. the ‘extent to which the results of a study agree with those of replication studies’ ), has been flagged as a key concern (; ). Animal studies often demonstrate deficiencies such as inaccurate or inconsistent data collection procedures, poor reporting of key variables, including the age and sex of animals used, and a lack of measures to reduce risks of bias, including the absence of randomization or blinding (; ; ). In terms of statistical methodology, frequent issues include small sample sizes, leading to low-powered studies, inadequate control for confounding variables, and insufficient description of statistical methods or reporting of uncertainty measures (; ; ; ). Publication bias, the phenomenon in which the decision to publish a study is based on the direction or strength of its findings, is also rampant. Animal studies reporting positive and statistically significant results are more likely to be published than those with negative or statistically non-significant findings, meaning that subsequent human studies may be based upon biased conclusions (; ). Such issues are a detriment to both animals, whose lives are wasted when incorrect conclusions are drawn from research performed on them, and humans, who are put at unnecessary risk during clinical trials when an intervention’s reported safety or efficacy is overstated or outright false (; ). Recently, there has been a growing interest in replications of previously published studies. A replication is defined as a ‘study that repeats all or part of another study and allows researchers to compare their findings’ (). To perform a replication study, researchers could, for example, use the same methodology and/or analysis as presented in an original study on newly collected data. They then attempt to determine if the results from the replication study are consistent with those in the original study (). A multitude of metrics have been used or proposed to quantify the consistency of results, and ultimately to decide if a replication was ‘successful’ or not (). We will refer to these metrics as ‘replication success metrics’. They might compare, for example, the p-values or the magnitude, direction, or uncertainty of estimated treatment effects obtained from the original and replication studies. Other metrics, such as the one based on a meta-analysis, combine results from an original study and its replication attempt(s) to estimate an overall effect size (; ; ). So far, studies attempting to estimate how often translation failure occurs have largely utilized the simple statistical significance criterion, that is assessing if the animal and human studies both report a statistically significant treatment effect in the same direction, often referred to as the two-trials rule (). To our knowledge, the usage of alternative replication success metrics in a translation setting has not yet been investigated. Translation contrasts with replication in that animal and human studies examine different populations and often have different experimental designs, and thus inherently produce different results. As such, a human study is a ‘conceptual’ rather than a ‘direct’ replication of the animal study (). As a result, metrics that are useful for measuring replication may not be as applicable for translation. This distinction motivates the current simulation study. In this paper, we define ‘translation success’ statistically: a metric flags translation success when a certain condition, depending on the metric, is fulfilled. This condition reflects the translation goal, which could be, for example, confirming that a beneficial effect exists in both animals and humans, or assessing the similarity of effect sizes across animals and humans. This statistical definition is narrower than biological translation, which concerns the underlying mechanisms linking animal and human physiology. Several frameworks have recently been proposed to structure the use of preclinical evidence in decisions about progression to human trials (; ). Quantitative tools that assess the consistency between animal and human efficacy findings, including the metrics evaluated here, can provide an empirical basis for one component of such frameworks.