What Ails Data AnalysisBy Kim S. Nash | Posted 2002-06-12 Email Print
Pharmaceuticals need to get smarter about using information technology to manage data. The alternative? Multimillion-dollar government penalties.
What Ails Data Analysis
One obvious defect begging for a fix: weak data analysis.
In 1999, the FDA noted that SmithKline Beecham of London, then worth $8.3 billion, had enough data available to indicate recurring problems with keeping needle assemblies sterile, but the company's quality control department failed to interpret the statistics correctly.
Quality control is an area ripe for better technology, says Terryn Barill, CEO of Terryn Barill Inc., a management consulting firm that helps drug companies comply with federal rules.
Technicians doing quality assurance must send one copy of their results to regulators, keep another in the lab and, usually, ship a third to the product manager or clinical trial manager, Barill explains. "Software would make it so much easier," she says. "You'd get e-mail to everyone and there would be a file copy on the network server."
Schering-Plough declined interview requests, but a company insider close to the issue says the information technology group is being reorganized to help deal with that firm's $500 million mess. Part of the $60 million the company has said it will spend to rectify problems at its plants in Puerto Rico and Kenilworth, N.J., will go to upgrading the company's supply chain systems, the executive says.
Though the number of warning letters issued by the FDA has declined steeply since 1998, the scope of the criticisms in them has grown much wider. In fact, the number of warnings issued has dropped by more than half in only three yearsfrom 179 in 1998 to just 86 in 2001. But the numbers are deceiving.
In the past, federal inspectors would cite many individual infractions, such as drug dust on the floor or a failure to document changes to written specifications. But now the FDA sees such incidents as symptoms of bigger problems in how a drugmaker operatesmuch more serious issues that are not smoothed out with mea culpas and the payment of a small fine.
Schering-Plough two years ago failed to keep all laboratory tests as required and had no backup files for some tests, the FDA says. The tests were of a sensitive machine used to measure the effects of light on some finished pharmaceuticals. It wasn't life or death stuff. Still, the government wants companies to keep the data to prove that they properly monitor production and to let them trace later problems. But most of the data, if it existed, was on paper. It got losta basic reason to computerize.
Merck was written up in November 2000 for submitting false test results to the FDA related to a weigh station at a Puerto Rico factory, then got dinged again for a management review process that "failed to prevent or even detect the problem."
With auditing software that enforces a chain of custody similar to what lawyers use when handling court evidence, Merck managers could possibly have spotted the falsification and acted.
Of course, if workers want to ignore rules, they will. Software won't stop them. But with the FDA more alert than ever, it's time for drug companies to use the technology available to them, Barill says. "A lot of the consent orders coming out are due to pharmaceutical firms not wanting to admit that the FDA really has changed."