cGMPs

From the Editor: Brush Up on Statistics (Or Face the Consequences)

Are you using statistics to support the status quo, or to get to true root cause?

By Agnes Shanley, Editor in Chief

For several years now, we’ve heard about all the new reviewers and inspectors that FDA has hired to help improve the level of scientific understanding at the Agency. A far cry from traditional hires in the past, some recruits have had experience in very different industries, some have engineering degrees.

I interviewed one of this new generation of FDA staffers recently: Karthik Iyer, statistician and senior policy advisor at CDER’s Office of Manufacturing and Product Quality (OMPQ). Karthik, who has been with FDA for two years, has a B.S. in chemical engineering and an MBA, is a Certified Quality Engineer, and also certified by ASQ as a Six Sigma Black Belt, and has over 11 years in the chemicals, refining and consumer products industries. 

His main role is to analyze the statistics most relevant to cGMP compliance and drug manufacturing, to support OMPQ. When that office sees a case, the facility has already failed to comply with cGMPs. 

We’ll be publishing that interview, with Mr. Iyer and colleague Rick Friedman, in our next issue. But, at AOAC International (the Association of Analytical Communities) conference last quarter, speaking as a statistician rather than on behalf of FDA, Mr. Iyer summarized some trends that he has seen recur in FDA 483’s and warning letters that show some basic disconnects. (Click here for a copy of the presentation.)

The good news is that more facilities are turning to statistical process control and process capability analyses in their day to day operations and quality control. The bad news? These tools aren’t always being used correctly, or for the right problem.

In the worst cases, statistics are being used to hide some incorrect assumptions or to cover some shortcut. We’ve seen the fruits of this approach in 483’s, which cut to the quick of lame or fuzzy excuses citing “operator or analyst error.”
In other cases, statistics are being used, to paraphrase Andrew Lang, as a drunken person uses a lamppost, more for support, rather than a source of light for better process understanding.

Mr. Iyer pointed to some lingering misunderstanding of sampling and other basic concepts.  In one case, a company pooled x number of vials but then used only one reportable value (even though n=x had been used in its sampling plan). So the wording of the resulting 483 shouldn’t have been too much of a surprise. “You based batch acceptance criteria on a single reportable value averaged from a pooled sample.”

The company’s response to the 483, Iyer said, showed that it didn’t understand how to use or interpret sampling plans, or concepts such as Acceptable or Limiting Quality Level, and the operating characteristic curve of a sampling plan.

Iyer provided another example, from a manufacturer that makes four tablets of various strengths. The company’s initial process qualification used a single-sided tablet press. However, the manufacturer was actually using a double-sided press during day to day manufacturing, and had not qualified compression using this press.

The company tried to use process capability analysis (Cpk) to show that the single and double sided presses were statistically equivalent, when, instead, Iyer said, it should have compared means and variances, and used distribution analysis or parametric approaches to show equivalence. Process capability alone wasn’t sufficient, Iyer said, because it requires a normal underlying distribution and demonstrated state of statistical process control.

He mentioned another problem involving a large-volume parenteral packaged in a dual-chambered bag. Over a period of 2.5 years, there were reports of leaks, bursts and premature activation, traceable to bag and closure manufacturing problems, in particular, variability in film thickness and critical sterility and stability problems.  The bag used two seals, and the supplier of container closures had a spotty product quality history.  

Pharma uses statistics in so many operations, and on a sophisticated level. But is there more to be learned about applying statistics to manufacturing? FDA is using sampling methods developed in other industries, and standardized by ASTM, in its day to day operations. It’s also employing people like Mr. Iyer, experts with broad-based experience, to look at how you’re using statistics.

Are you also employing people like Mr. Iyer? 

At the very least, brushing up on statistics and how they’re being applied in the broader manufacturing world can not only help your line, facility and company stay out of trouble with regulators, but ensure the kind of process understanding that can only lead to consistently safe products for patients, continuous improvement and lower costs.

Free Subscriptions

Pharma Manufacturing Digital Edition

Access the entire print issue on-line and be notified each month via e-mail when your new issue is ready for you. Subscribe Today.

pharmamanufacturing.com E-Newsletters

A mix of feature articles and current new stories that are critical to staying up-to-date on the industry, delivered to your inbox. Choose from an assortment of different topics and frequencies. Subscribe Today.