April Fool’s, Your Data’s Aberrant!

It was two months prior to the scheduled filing date for an improved formulation of our oral contraceptive. We’d collected 12 months of stability data and were just sending it off for formal statistical analysis when an analyst noted that two data points—one at 3 months and one at 6 months—were potentially aberrant with drops of over 4% from trend. Unfortunately, the aberrant data widens the 95% confidence interval so we can no longer expect a 24-month shelf-life label based on these results. By eliminating them we would support well over 36 months.

With no laboratory investigation on file for the data and no comment documented, someone appears to have blindly entered results without checking historical trends. An initial laboratory inspection of the raw data and calculations found no obvious errors, though tablet count or a weighing error during sample or standard prep are obvious probable causes. So, we either justify eliminating the data or we file with an 18-month shelf-life.

Either way, we’re in for some major changes in the way stability data is reviewed. What are some strategies your company uses that would have prevented this?

Something more automated would be nice, but meanwhile, the stability group could provide regularly updated data trends for each product and require documentation that stability test results entered by the laboratory are checked against the historical trends. Finally, require that verifiers document their attested agreement with the trend assessment prior to signing off on the result.