Effect of lack of blinding on outcome assessment in rodent object recognition tasks

Background: Blinding of the outcome assessor in experimental research is an important measure to reduce bias (Bespalov et al., 2020). In a blinded condition, the observer is unaware of the experimental groups to which an individual belongs and is therefore less vulnerable to observer bias (Mahtani et al., 2018). Blinding of outcome assessment is believed to influence outcomes, with lack of blinding often exaggerating the estimate of the effect of an experimental condition on measured outcome (Hrobjartsson et al., 2012; Bello et al., 2014; Macleod et al., 2008). Despite this, blinding is not always reported in preclinical settings. To date, there is little quantitative experimental evidence measuring the impact of blinding on outcome assessment, particularly in the context of preclinical behavioural research. Study description: We have designed a randomised controlled study testing the impact of blinding on outcome assessment in rodent object recognition tasks (Novel Object Recognition (NOR) and Object Place (OP), see Figure s1), which will be scored by 3rd year undergraduate students (presumed to be inexperienced observers) in the context of one of their course modules in animal behaviour. Students will score novel object exploration in previously recorded videos of an experiment testing cognition in mice exposed to different dietary regimes. The students will be introduced to the hypothesis that cognition is impaired in mice on high fat diets. Students will be randomly allocated to being blinded or unblinded for all video evaluations. They will each be asked to score 24 behaviour videos selected from a pool of 88 videos. Each animal is represented in 2 of these 88 videos, once for the OP task and once for the NOR task. Each of these videos will be presented either with (“unblinded”) or without (“blinded”) embedded text identifying the treatment group to which the animal belongs as well as the nature of the test (OP or NOR) and the identity (left/right) of the novel object. Appropriate versions will be presented according to the student status as blinded or unblinded. Each student will conduct 2 sessions of scoring with 12 videos in each session; and each session will contain 3 videos for each of 4 conditions (OP followed by NOR, with LFD or HFD presented in random order). We created sets of videos to ensure balanced representation between the first and the second scoring sessions and across participants. The intervention (addition of video labels to remove blinding) is intended to create an expectation in the mind of the scorer that control mice will explore a novel object more than they will a familiar one; and that high-fat diet-fed mice will not be able to discriminate between a novel and familiar object. Each of these tests provides a discrimination score, which is a measure of the difference in time spent investigating a familiar object in a novel place (OP) or a novel object (NOR). For each of OP and NOR we will report the effects of the dietary intervention in data aggregated from blinded and from unblinded observations. Our primary outcome measure is the proportion of students finding statistically significant differences (p