The Reproducibility Project: Cancer Biology, led by Tim Errington, just reported the results of an eight-year project where they tried to replicate experiments from 193 preclinical studies published in 53 cancer journals. Preclinical studies refer to studies conducted in animals before being carried out in humans. Those studies were published between 2010 and 2012.
The team could only reproduce 50 experiments out of 193 because of a lack of data, reporting issues, or access to materials.  
To then assess the replication of the 50 experiments, they used five criteria and “found that just 18% succeeded on all five, while 20% failed on all five. Overall, 46% of effects succeeded on most criteria.”
The results are reported in two articles published in eLife, Challenges for assessing replicability in preclinical cancer biology, and Investigating the replicability of preclinical cancer biology.

Nature just published a comment titled, “Five keys to writing a reproducible lab protocol” for better transparency and to avoid similar issues in the future.