Having observed a large variation in the number and type of original preclinical publications for newly registered drugs, we have explored whether longitudinal trends and/or factors specific for certain drugs or their manufacturers may explain such variation. Our analysis is based on 1954 articles related to 170 newly approved drugs. The number of preclinical publications per compound declined from a median of 10.5 in 1991 to 3 in 2011. A similar trend was observed for the number of in vivo studies in general, but not in the subset of in vivo studies in animal models of disease. The percentage of compounds with studies using isolated human cells or cell lines almost doubled over time from 37 to 72%. Number of publications did not exhibit major differences between compounds intended for human versus veterinary use, therapeutic areas, small molecules versus biologicals, or innovator versus follow-up compounds; however, some companies may publish fewer studies per compound than others. However, there were qualitative differences in the types of models being used depending on the therapeutic area; specifically, compounds for use in oncology very often used isolated cells and cell lines, often from human origin. We conclude that the large variation in number and type of reported preclinical data is not easily explained. We propose that pharmaceutical companies should consistently provide a comprehensive documentation of the preclinical data they generate as part of their development programs in the public domain to enable a better understanding of the drugs they intend to market.

The present study aimed to explore early publication patterns (i.e. up to 1 year after obtaining regulatory approval) for newly registered drugs. From the website of the US Food and Drug Administration, all newly approved drugs for 6 calendar years between 1991 and 2011 were identified. Non-clinical original publications for these compounds were retrieved from PubMed and their abstracts analysed for type of study and reported data. This yielded 170 compounds for which 1954 original non-clinical publications were identified, i.e. a median/mean of 5/11.5 publications per compound; however, number of publications per compound varied widely (0-90) and this variation exhibited a non-Gaussian distribution. The earliest non-clinical publication typically was published less than 5 years before regulatory approval and more than 5 years after filing of the primary patent, but some compounds exhibited notable deviations from this pattern. Publications most often reported on efficacy related to the target indication and on potential future indications, with fewer studies addressing mechanisms of action, potency, selectivity, pharmacokinetics and toxic effects. For most compounds, data at the cellular and in vivo level were published, with fewer reports on effects on isolated tissues and even fewer at the molecular level. The preferred species for cellular studies was human, whereas for in vivo studies, it was rats and mice. In 75 % of cases, the lead author of the publication came from an academic institution, and most studies were published in classic pharmacology journals. We conclude that number, timing and scope of early non-clinical publications on newly approved drugs exhibit major variance. Factors potentially associated with such variance are explored in a companion paper.

No abstract