Bad data: new methods needed to catch errors in clinical trials
“We do not yet know how often scientists misbehave,” says Chris Hartgerink, a researcher from the Tilburg University, The Netherlands. However, inaccurate data is not always the result of misbehavior – often, it is the product of human error or poor data handling.
Thus, Hartgerink is leading a new Dutch Fulbright proposal that seeks to provide additional quality control layers for clinical trial results reported in the ClinicalTrials.gov database.
“Many people complain about pharma,” he says, then asking, “Do the numbers with respect to data fabrication support or disconfirm their complaints?”
In the pharmaceutical industry especially, bad data is not only costly, but life threatening
For example, beta-blockers were originally prescribed to cardiac patients to decrease perioperative mortality, but after a meta-analysis detected erroneous data in the related clinical trials, it was determined that the medication actually increased mortality risk.
When asked if he thinks the 2% response rate (of researchers who admit to manipulating data) is accurate, Hartgerink tells us, “The 2% rate is an underestimation, so no, I do not think it is accurate.”
Actually, he thinks that the number is likely double, maybe even more; however, he admits that it is pure speculation without the research, further highlighting its importance.
The next steps
While Hartgerink isn’t currently able to suggest methods for tacking reporting issues (his research proposal is about finding these new methods) he does explain they all apply the same principal: “People (including scientists) cannot properly process probabilities, and this will reflect itself in the data. If groups are randomly assigned, they should not show differences prior to an intervention, for example.”
The systematic reporting in the ClinicalTrials.gov database will facilitate the application of these methods – the next step is to carry out the research proposed in the project to develop the new approaches.
According to Hartgerink, The proposal has been submitted, but the team does not know yet if they will receive funding, without which they will be unable to complete their research.
