May 29, 2015
Reproducibility of Research and Preclinical Validation: Problems and Solutions
Many scientific reports and literature present a considerable lack of reproducibility which causes high rates of failure when validating highly promising preclinical observations in clinical research. In these types of studies and especially in cancer research, there is a list of technical reasons that prevent from reproducing the experiments. The combination of all these reasons result in the publication of underdeveloped results, so it’s essential to identify all of them in order to make alternative plans and improve reproducibility.
Advances in biology and medicine have made possible the increase of life expectancy as well as an improvement in the quality of life. In this way, drug industries have been developing without stopping. Recently, the cost of drugs has increased considerably due to failed experiments researching new drugs. There are estimations that point out that 80% of phase II and 50% of phase III studies are failed in most cases because of a lack of efficacy. What is worse is that many published results limit reproducibility. Some reports reflect this problem. There is a research based on the published papers of a year in one high-impact journal which states that 50% of the results of some published clinical studies couldn’t be reproduced and only 33% of them could be reproduced to some extent. These facts, at the same time, weaken public trust in science and philanthropic support for research and without a doubt, they are the reason why the scientific progress is slow and there are some useless and failed experiments. This entire situation is caused in general by poor documentation trials, a selective publication of positive results, misinterpretation of technical noise and fabrication data. Organizations such as the NIH in the USA are promoting a responsible and reproducible research campaign as well as a list of recommendations. In addition, Nature published a new editorial policy in May 2013 in order to improve the quality of the experiments.
The current scientific field situation rewards positive results in regards to grants and funding, creating atension among true experimentation, integrity and the career interest of a study to transform the negative results into positive results. There is a study that found that 33% of researchers answering a survey admitted data manipulation and 70% of them reported questionable data manipulation by others.
Recently, there is an increase of hypothesis-driven studies, which is one of the causes of cognitive biases. When it’s impossible to prove the scientist inquiries it’s necessary to reformulate the hypothesis. However, the scientists are focused on proving one’s own hypothesis and this is called by psychologists “confirmation bias.” Consequently, there is a tendency to think the experiment was wrong; instead of thinking the hypothesis was wrong.
Distributing information on good experimental design, appropriate statistical techniques and documentation could improve reproducibility of scientific experiments results. Nevertheless, these factors are not sufficient. In the following box there are some other recommendations looking for the same purpose.
Offering funding and grants to replicate published experiments and validate the results could be a good step towards the solution of this current problem. Investing in reproducibility research is the key to erase poor practices. A three-pronged grant system has been proposed to encourage laboratories to pay attention to document methodology and ensure reproducibility from the initial publication. In the following box there is a summary of this proposal.
According to these methods, changing the objectives from the hypothesis to the research question in the grant preparation process could be a very good way to reduce confirmation bias. All experiments start with an undefined question. However, many grant applications fail to state that question explicitly or as a brief part of the introductions whereas the hypothesis occupies an independent section of the application. This is the reason of this problem and they have to reestablish the priorities and strategy.
Apart from that, negative results are as important as the positive ones despite a few publications that include fail experiments. Nowadays, there are more and more open-access journals like PLoS ONE and Scientific Reports that provide high visibility of result-neutral publications.
The transparency of the clinical trial process has increased substantially in the past 10 years since the International Committee of Medical Journal Editors (ICMJE) introduced a policy in 2005 that they would consider trials for publication only if the trial had been registered in the NIH clinical trial registry. This has made access to information about therapeutic clinical trials easier by making the synopses of trials available to the public.
Unfortunately, similar data repositories do not exist for basic and preclinical research data, possibly due to the difficulty to normalize these sorts of experiments.
Written by Dr. David Alcantara and Paula Ruíz for The All Results Journals.