Published in the American Association of Pharmaceutical Scientists (AAPS) Journal last month, the paper calls for proper software validation after attempting to validate software used to assess bioequivalence in generic drugs.
“Very few datasets with known results are in the public domain, and the few that are published are of moderate size and balanced,” the paper states, and therefore the study introduced reference datasets of varying complexity calculated using four commercial packages - EquivTest, Kinetica, SAS and WinNonlin.
“We set out to put reference datasets for 2-treatment, 2-sequence, 2-period bioequivalence studies into the public domain along with known results,” one of the authors - Anders Fuglsang, a consultant and ex-regulator – told Outsourcing-Pharma.com. “To make sure we had a reasonable confidence in about the ‘known results’ we ran the datasets through different statistical software packages.”
However, he continued, “not all software packages produced the same results when the datasets had imbalance, which is a quite common phenomenon as volunteers in these studies can withdraw their consent to participate any time during the studies.”
Fuglsang added the purpose of the study was not to judge any particular brand of software or its fitness for bioequivalence evaluation, but rather the fact that imbalance between sequences provoked questionable results with one package illustrates the need for proper software validation.
An error with such software could mean some drugs may have been inaccurately approved.
According to a spokesperson from the European Medicine Agency (EMA), “the paper by Fluglsang et al. and its findings will be further discussed by the CHMP’s Pharmacokinetic Working Party at its next meeting 21-22 October 2014 to determine next steps.”
Outsourcing-Pharma.com also approached Thermo Scientific which said it is “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.”
According to the company, the software lets CROs and drug developers reuse and share standardized settings, streamline decision making and ensure more consistent results when analyzing PK/PD data, and can help decrease costs when determining toxicity and dose optimization.
We also spoke with Steve Toon, Managing Director of Certara Simcyp which owns the WinNonlin system also analysed in the paper.
While he told us WinNonlin has been in use for more than 15 years with no reported issues of this nature, he added that any user of such software systems have the responsibility to prove validity of “output as outlined in GLP and GCP standards and that user should be performing routine checks of performance with standard data sets.”
If data findings are found to be inaccurate with the Kinetica platform, as the study suggests, “the next action for the sponsor would be to reanalyse the data using an alternate tool,” Toon said, which is fairly quick and straight-forward.
“The only repercussion would be if the re-analysis shows that the prior statement of bioequivalence is unfounded. Then, the trials would have to be re-run. Ultimately, the responsibility falls on the sponsor.”