Back in late 2016, the US Food and Drug Administration (FDA) released a new version of the Phonetic and Orthographic Computer Analysis (POCA) program called POCA 4.0. For those of us in the pharmaceutical trademark clearance and/or branding industries, this sounded like a good thing.
Program version updates come with the expectation of improvement. The original POCA was released years ago, and while not perfect, it remains the pharma industry standard in terms of ranking pharmaceutical marks on name safety similarity to a proposed candidate. POCA isn’t a definitive tool, and it’s just one of many methods and tools in the FDA’s arsenal for determining name safety issues in the pharmaceutical industry.
The POCA 4.0 algorithm has indeed been dramatically changed from the original pre-4.0 version. According to the FDA, the orthographic component of the POCA algorithm was revised to better capture errors that are being reported due to the increased use of electronic prescribing. The algorithm has been revised to put more emphasis on the similarity that occurs at the beginning of a name — especially the first three letters and on exact letter matches — and also to emphasize where names share letters that are not consecutive.
Sounds good, right? The FDA is improving POCA to make it work better with script errors that happen when using drop-down menus and other user interfaces for creating an electronic prescription. While perhaps an improvement over handwritten scripts, new technologies come with new problems.
Unfortunately, these changes were not without an “adverse event.” At Corsearch, we examined POCA 4.0 and in our comparison to prior versions of POCA, we noticed a significant increase in the result set being returned from POCA 4.0 queries. Where we used to get a few hundred results, we were now getting thousands of results. Results in the threshold ranges of 60 and below appear to show the most significant increases in results, but all threshold ranges have been affected by increases.
In short, the new POCA 4.0 algorithm is essentially a shotgun approach with respect to orthographic similarity. The focus on the three-letter prefix, exact letter matches, and shared letters that are not consecutive have resulted in POCA users now getting reports up to 10 times larger than the prior POCA algorithm. This is due to the new algorithm raising the scores of many marks based on the criteria mentioned above. So get ready for weeding through quite a bit more name safety data than you did with the old POCA. In our testing, we saw data reports that had over 15,000 returned marks, even with thresholds set at 60. Additional good results do come up, but this is due to the extremely broad nature of the algorithm changes.
On the bright side, the FDA has raised the thresholds to help reduce the burden of the new higher-yield algorithm. “Highly Similar” is still 70% or greater. “Moderately Similar” is now between 55% (formerly 50%) and 69%. And finally, “Low Similarity” is now 54% or lower. Despite these changes, POCA users can still expect to see vastly larger result sets with POCA 4.0, and with that, increased review and analysis time.