An apparent glitch in statistical software used in clinical trials has cast doubt on the reliability of data used to approve some drugs for use.
An official at the London-based European Medicines Agency (EMA) told Reuters that the issue, involving Thermo Fisher Scientific's Kinetica package, would be discussed by European regulators at a meeting next week.
It is believe that the issue affects generic copies of original drugs in particular. The US-based maker of laboratory equipment and life science research tools said it was looking into the matter, which was first raised by independent experts in a scientific paper.
The problem could mean some medicines have been approved on incorrect data. Others may have been rejected, or never submitted, even though they might have been good enough for use. The scale of the potential problem is unclear, but may extend to medicines submitted for approval in Europe, the US and beyond.
Kinetica is one of the most widely used statistical software packages for analysing so-called bioequivalence in clinical trials. It is integrated into data analysis processes at both pharmaceutical companies and contract research organisations, which conduct clinical tests on behalf of drug manufacturers.
Proving bioequivalence, which means two drugs with the same active ingredient are absorbed equally when taken into the body, is a prerequisite for approving cheap, generic versions of medicines after patents on the originals expire.
Generics account for around 80 per cent of such bioequivalence studies, although innovative pharmaceutical companies also use them for testing drug-drug interactions, food effects and formulation changes.
"It's an unprecedented situation with unclear implications," said Anders Fuglsang, a former medicines regulator who now runs a pharmaceutical consultancy in Denmark.
He and two colleagues reported the questionable results with the Kinetica software in a highly technical paper published last month in the journal of the American Association of Pharmaceutical Scientists (AAPS).
In a statement, Thermo Fisher said: "We are currently conducting a careful review of the issue raised in the AAPS journal article regarding Kinetica software. Because the review is ongoing, we decline to comment further."
In the light of the findings the authors of the paper wrote to the EMA, Europe's version of the US Food and Drug Administration, to express their concerns, prompting the agency to look into the matter.
"The paper and its findings will be further discussed by the CHMP's (Committee for Medicinal Products for Human Use) Pharmacokinetic Working Party at its next meeting 21 to 22 October 2014 to determine next steps," an EMA spokeswoman said.
Fuglsang and his associates found that Kinetica gave results that were inconsistent with those from other software packages when there were more subjects in one sequence of treatment than another. Such imbalance is very common in clinical trials because patients can drop out for various reasons.
In the case of these imbalanced trials, the authors found the confidence intervals used to assess bioequivalence could be unreliable with the Kinetica software package. Confidence intervals are used as an indication of how a clinical trial result would be reflected in the general patient population outside of a study.
A confidence interval falling within the limits set by regulatory agencies means a drug can be approved, since it implies the product is as safe and efficacious as the one to which it has been compared.
Christian Schneider, medical head of division for medicines licensing and availability at the Danish Health and Medicines Authority, who was not involved in the research, said any impact on drug approvals needed to be assessed by scientists, such as those on the EMA's expert committee.
"A bioequivalence study, albeit being of central importance to any generic medicine dossier, is only part of a huge set of data that is filed for approval, involving for example a whole battery of analytical tests," he said.
"Therefore, any regulatory decision for approval is not based on a bioequivalence study alone. Nevertheless, one will have to see what impact the results of this article might have.”