Imprimir Republish

Good practices

Reproducibility put to the test

Léo Ramos Chaves Tumor cells targeted by drugs such as those tested under the Reproducibility InitiativeLéo Ramos Chaves

The results of the Reproducibility Initiative, a U.S. project launched in 2013 to assess the reproducibility of 50 cancer drug candidates, are now being made public after the biopharmaceutical company Amgen reported having failed to replicate (with consistent results) 47 of 53 experiments described in such journals as Nature, Science and Cell. Five studies were published in the January issues of Nature and eLife. A test measuring the effects of an anti-tumor peptide failed, those of two anti-cancer compounds indicated data that were almost identical to the original study and the results of another two tests were unclear, making comparisons impossible to conduct.

Erkki Ruoslahtia, a researcher at the Sanford Burnham Prebys Medical Discovery Institute in California who oversaw the unsuccessful anti-tumor peptide test, reported to Nature that at least 10 laboratories in the United States, Europe, China, and Japan had obtained the same result. He said that the fact that a given experiment does not replicate the initial study faithfully does not mean that it cannot obtain the same results; and he also says that he fears irreproducibility of results will prejudice the approval process for the funding of future research.

The failure to replicate results need not imply that conclusions are erroneous, warns Tim Errington, manager of the Reproducibility Initiative and a researcher at University of Virginia’s Center for Open Science. According to Errington, researchers should view negative results as information, not a negative judgement, unless others also fail to obtain the expected results. In Nature, Errington remarks that perhaps the project’s most obvious conclusion is the fact that many articles include little in the way of detail concerning the methodology employed in the experiments.

In its editorial, Nature points out that reproducibility is one of the central assumptions when doing science and highlighted, without justifying the point, the difference between the results of initial experiments and their replications. In one of the cases, carcinogen-induced lab mice treated with an anti-tumor drug survived nine weeks in the initial study but only one week in experiments that attempted to replicate the results of the initial trial.