Medical Device Link.

INSPECTION SYSTEMS

The PSIFEYE from The Value Engineering Alliance can overcome distortions that may result when proofreading large inserts.

Call it the multi-million-dollar decimal point. Or hyphen. Misplaced or missing, these typographical mainstays can cause the kinds of packaging mistakes that make pharmaceutical executives reach for their companies’ own heartburn medicine. Misprints of any kind can lead to very expensive FDA recalls and, in extreme cases, customer fatalities.

Most pharmaceutical companies use the eagle-eye talents of manual proofreaders. Their job is to ensure that inserts, outserts, and carton artwork remain accurate as they proceed through each departmental step on their way to hospital or pharmacy shelves. The process can take hour upon eye-watering hour, and fatigue and fallibility can result in errors. Even though GMPs require 200% human inspection, labeling mistakes occur. As a result, a small percentage of manufacturers have turned to computerized systems that can slash proofreading time, increase accuracy, and speed market launches.

Made by only a small universe of providers, automated proofreading systems look for deviations in content by comparing two or more documents that are converted to different formats, such as a Microsoft Word file to an Adobe PDF file. They also search differences among either electronic or hard-copy files with similar formats. The systems incorporate large-format scanners, flat-bed scanners, or CCTV cameras. Using computer images, the systems flag differences so that operators can compare versions in order to determine whether there’s an actual discrepancy. Suppliers claim that the automated systems can trim proofreading time from hours to literally minutes.

Manually read inserts, for instance, “take hours and hours” to review, notes Miriam Glassman, marketing communications manager for Global Vision Inc. (Montreal). Unfolding at lengths of 181¼ in. or larger and, carrying 5-point type, often in multiple languages, inserts require proofreaders to manually check wording “character by character.” Often the proofreaders don’t necessarily speak the languages they’re double-checking, she says.

“In the pharma world there’s no margin for error,” Glassman asserts. She raises the potentially life-threatening situation of, for instance, an en dash that goes missing in instructions for caplets, resulting in a dosage recommendation of 23 tablets instead of 2 to 3.

In September 2006, Global Vision and MeadWestvaco’s Paxonix division (Waltham, MA) announced a strategic alliance to combine automated proofreading and brand packaging management software in a single platform. The partnership brings together Global Vision’s Docu-Proof document and graphic comparison product and Paxonix’s PaxPro life-cycle management software.

Docu-Proof allows users to compare text and graphic files for spelling errors, missing text, added text, unwanted font changes, and similar mistakes. PaxPro is designed for product life-cycle management (PLM), an engineering-based program encompassing packaging, branding, and marketing. Both PaxPro and Docu-Proof meet 21 CFR Part 11 and other FDA requirements, according to the companies.

“The effect of a product recall could cost millions,” Glassman notes.

“No one actually has a line item in their budget for rework. ‘Okay, who’s got the screw-ups this quarter? I’ve got one over here for $300,000 or $400,000,’” jokes Kent St. Vrain, vice president of business development for Paxonix.

Speed to market, quality, and accuracy are three of the driving forces behind automated proofreading. Controlling the various versions of files ensures that “the wrong files aren’t moving around,” St. Vrain adds. Collaboration, annotation, electric notes, routing, and approval—all are offered in one streamlined package.

“Global Vision found a way to look at two different versions of the same file and essentially automatically proofread them, whether they’re in the same file or not,” St. Vrain notes. “Word has some nice proofreading [capabilities], and as soon as you take that off, and put it in a label…in most formats it’s no longer a Word document, it’s a graphic.”

The AutoProof system from Complete Inspection Systems compares inserts to electronic files in minutes.

Typically, a Word file arrives from the pharmaceutical house’s regulatory department with the approved wording, St. Vrain says, adding, “nobody screws things up intentionally. But when they take it from the Word document and put in Adobe Illustrator as a PDF file, all your proofreading capability disappears.” The proofreading system “looks at the Word document and looks at the PDF and tells you the difference.” The system checks everything automatically and creates two reports. If the proofreader is working on-screen, flashing lights indicate a discrepancy between the two versions. The proofing tool also creates an inspection report for the auditor, showing, for example, that “on line 6 there’s now the letter ‘a’ where there wasn’t in the approved version.”

The same process occurs between two versions of the same graphic, so that if the operator has two labels “that are now really graphics and no longer Word documents at all, other than stylized logos or a bar code or something like that, our system deals with the graphics.”

Gary Parish knows automated proofreading about as well as anyone, having started more than 28 years ago with his first system. His initial work began in the late 1970s and early 1980s at a lab in Irvine, CA, that had printed some silk-screened bottles with label mistakes. He was asked to devise a method to verify all the bottle copy. Now president of Complete Inspection Systems (Indiatlantic, FL), Parish worked through the years to perfect the automated proofreading system that is used today by companies such as Bayer HealthCare Inc. (Morristown, NJ) and Cardinal Health Inc. (Dublin, OH).

Complete Inspection Systems’ AutoProof Pro Proofreading Suite 3.5 is offered as a complete inspection and proofreading tool that allows users to convert PDF files into images or actually convert the file to text in any language. Once the file is converted, the operator can compare a 20-page regulatory document to a one-page insert, finding any differences in any language. The AutoProof ATM version scans and compares up to 50 pages at one time, significantly reducing the time it takes an operator to compare those pages from hours to minutes.

AutoProof Pro is also designed to scan large press sheets up to 54 in. (137 cm) at speeds of 10 in./sec at 300 dpi resolution for accurate comparisons. Its pharma version complies with 21 CFR Part 11. Suite 3.5 has the ability to compare images as well as text copy. The latest version, Inspec, offers a new 10.3-megapixel camera and the ability to capture a 17 × 23-in. document in a single image.

Parish cautions that the types of images being used can affect proofreading accuracy. Manufacturers should ask: “What kind of picture am I using? By using different image formats there is a possibility you can lose data.” JPEG image formats allow pixel loss by compressing and reducing the file size. Editing the image by stretching or shrinking the picture could also cause information to be lost. It’s like going from a movie format to that of a television screen. “The key to quality comparisons is in the quality of the images and information provided by the camera, scanner, or graphics program,” says Parish.

Earlier-version automated proofreading systems lacked the accuracy of the latest tools, says Marcel Singleton, founder and business development manager of The Value Engineering Alliance (Cambridge, MA). Large- format documents posed particular problems because many of the earlier techniques used less-advanced alignment methods than current systems,” says Singleton, whose company sells the PSIFEYE proofreading tool. The systems either “were not finding the right defects or were finding that they weren’t defects at all.”

Distortions caused by stretching and warping of large paper documents caused the bulk of the problems, Singleton says, adding, “they distort in a nonlinear manner.” Comparing a stretched version with one that hasn’t stretched creates proofreading problems that the PSIFEYE system is designed to overcome. The tool uses algorithms developed by Michael Negin, PhD, founder and president of Value Engineering Affiliate, Mnemonics Inc. (Mt. Laurel, NJ), where it was developed and is sold under the trade name Avia Proofreader. “He developed something called ‘needle-in-a-haystack’ technology that’s designed to find small defects in large images,” Singleton says.

“The job of the computer is to do all this analysis and to funnel it down to a point where we do the thing that we do well,” he adds. The proprietary technology is the software that makes the analysis.

“That’s where the effort went into nonlinear comparison techniques to address these issues of stretching and warping and these subtle distortions that might occur.”

Eric Sarver has been using automated proofreading for more than 10 years as prepress manager for two Cardinal Health plants in New Jersey, one for folding cartons and another for inserts and labels. He has used systems from Complete Inspection Systems to proofread “text-intensive and content-intensive” pharmaceutical inserts for more than two years. Cardinal Health also uses systems from Global Vision.

Speed has been the biggest improvement in the tools. Earlier automated proofreading systems were “too slow to keep up,” Sarver says. That led to a lot of “close-enough guesswork,” he points out. “They used to rely on overhead vertical cameras to take their image. Now they can use electronic files that increase the speed enormously because you can rely on having one image on top of another.”

The drawback was the overhead camera, because “the source document had to be square” and any skewing would throw off the analysis. “Because of the focal view of the overhead camera, you could only grab a certain portion at a time of what you were looking at. It was a panel-by-panel kind of tiling,” Sarver says. “That was time-consuming.” Systems such as Complete Inspection’s tool see an entire image at once, “and the software automatically finds those tiles if it needs to. That’s been a major improvement.”

Sarver says that any company considering automated proofreading needs to “make a distinction between validation and performance qualification.” Even software engineers have a hard time validating software, he asserts. “They know they have bugs in it. To truly do a validation in software is an engineering feat that even the software manufacturers are unable to do. Too many variations go into it.” Therefore, put a system through its paces with some basic performance tests that determine the ability to catch mistakes, Sarver suggests.

Given the time-to-market pressures of the industry, Sarver praises a text-extraction feature of Complete Inspection’s system. It’s particularly handy because pharmaceutical companies are in a perpetual “launch mode,” he says. “The insert is the last piece that FDA will sign off on because it’s so content intensive, and FDA is truly looking at Word documents.” Sarver says it’s guaranteed that graphic layouts of the Word documents will face last-minute changes when a new document shows up.

“You’re starting over at the 24th or 25th hour,” explains the prepress manager. Complete Inspection’s tool is able to “extract the actual text out of the Word document, compare it with the text of your graphic document, and basically redline it at the very last second.”

Even though automated systems promise reductions in proofreading times of up to 95%, Parish estimates that only 5–10% of all pharmaceutical companies use automated proofreading. Administrative practices constitute the biggest obstacle to wider adoption, he believes. “Each time a company makes a change, it has to validate and document those new procedures. That takes time and effort and brings some risk to the ‘project champion’” who may question whether he or she took the right approach in terms of concerns such as costs and system lifetime.