Non-publication and publication bias in animal research is a core topic in current debates on the "reproducibility crisis" and "failure rates in clinical research". To date, however, we lack reliable evidence on the extent of non-publication in animal research. We collected a random and stratified sample (n = 210) from all archived animal study protocols of two major German UMCs (university medical centres) and tracked their results publication. The overall publication rate was 67%. Excluding doctoral theses as results publications, the publication rate decreased to 58%. We did not find substantial differences in publication rates with regard to i) the year of animal study approval, ii) the two UMCs, iii) the animal type (rodents vs. non-rodents), iv) the scope of research (basic vs. preclinical), or v) the discipline of the applicant. Via the most reliable assessment strategy currently available, our study confirms that the non-publication of results from animal studies conducted at UMCs is relatively common. The non-publication of 33% of all animal studies is problematic for the following reasons: A) the primary legitimation of animal research, which is the intended knowledge gain for the wider scientific community, B) the waste of public resources, C) the unnecessary repetition of animal studies, and D) incomplete and potentially biased preclinical evidence for decision making on launching early human trials. Results dissemination should become a professional standard for animal research. Academic institutions and research funders should develop effective policies in this regard.