Description:The practice of selectively reporting results from a ’representative’ experiment is very common in preclinical research (Frommlet and Heinze, 2020). The most frequent approach is to report the best out of k experiments, where the ’best’ experiment is characterized by the one with the most significant result. This thesis will investigate an approach to adjust the corresponding effect estimate, p-values and confidence interval for this kind of selective reporting. The thesis will then investigate the effect of selective reporting on the probability to obtain a significant result in a replication study (without selective reporting). Data from several replication projects (Errington et al, 2021, The Brazilian Reproducibility Initiative, 2025) will be used to infer the amount of selective reporting in preclinical research.

Contact: leonhard.held@uzh.ch

References:
Frommlet and Heinze (2020). Experimental replications in animal trials. Laboratory Animals, 55(1):65–75. doi: 10.1177/0023677220907617.
Errington et al (2021). Investigating the replicability of preclinical cancer biology. eLife. doi: 10:e71601.
The Brazilian Reproducibility Initiative (2025). Estimating the replicability of Brazilian biomedical science. bioRxiv. doi: 10.1101/2025.04.02.645026.

Description:The standard to analyse a non-inferiority (NI) trial comparing a new treatment to an establised one is the fixed margin approach, where a non-inferiority margin is selected based on historical evidence that the comparator in the historical trial was effective. The margin is often taken to be the limit of the 95%-CI of the historical trial effect which is closest to the null effect and the NI trial then has to show that the new treatment is significantly better than the fixed margin. An alternative approach is to synthesize the evidence from both trials in order to investigate whether the new treatment would have been superior to placebo in a hypothetical direct comparison (Althunian, 2017). The relevance of both approaches is reflected in recent guidelines by the FDA and EMA. This Master thesis will compare the two methods and investigate the applicability of alternative methods in this setting (e.g. square-and-add for binary outcomes).

Contact: leonhard.held@uzh.ch

References:
Althunian et al (2017). Defining the noninferiority margin and analysing noninferiority: An overview. British Journal of Clinical Pharmacology. doi: 10.1111/bcp.13280.
EMA (2025). Guideline on non-inferiority and equivalence comparisons in clinical trials (Draft). PDF
FDA (2016). Non-Inferiority Clinical Trials to Establish Effectiveness. PDF