Masterarbeit

Drug Name (Un-)Likeness - Methods and Strategies to Find an Approvable Drug Name ***

Dr. Gilbert Müller (2024)

SUMMARY 

Language: English

The EMA Name Review Group has rejected between 40% and 50% of the submitted drug name candidates in the last 10 years. FDA also strictly reviews name candidates to prevent medication errors attributable to confusable drug product names.

To improve the approval likelihood of a drug name candidate, name similarity searching via a new web application is proposed to find similar approved drug and/or registered trade names that would prevent the name candidate from being submitted to Health Authorities and/or Trademark Offices. 

The web application provides an user interface to search a drug name candidate in a database compiled from approved drug names in EU and US as well as registered trade names from the ‘Deutsche Patent- und Markenamt‘. 282725 names from centrally approved drugs in EU, from the EU Article 57 product database, from the US FDA POCA website, INNs as well as German, European and International trademarks were collected to create the GlobalProducts database. 

A name candidate may be tested via 3 different approaches: ‘Full Name Similarity‘ and ‘Split Name Similarity‘ searches, which conducts a similarity search across the full (i.e., unchanged) names or across names having been split into single words. In addition, a complementary string detection method may be used to search for word parts independent of their similarity. Since approved drug/trade names often consist of more than one word, the ‘Full Name Similarity‘ search might not retrieve relevant multi word names which may be found by the ‘Split Name Similarity‘ search. 

Following a ‘One Global Brand‘ approach, a process to develop a first name candidate targeting approvals at least in EU and US as well as registration as a trademark is proposed, including different strategies for name development. 

The web application allows to search a name candidate against the names in the GlobalProducts database and to improved the name in a ‘search and change‘ cycle. Search results are presented as lists of similar approved drug/trade names, provided with similarity values and often information about owner, identifies, status, etc., and are sorted by relevance. Search results can be exported to an Excel file.

The ‘search and change‘ cycle ends when the name candidate is confirmed in searches on the POCA website as well as on TMO trademark websites and no similar drug/trade names are found or, alternatively, when drug/trade names are considered to be moderately similar but not confusable to the name candidate. 

If moderately similar search hits remain after the ‘search and change‘ cycle, aspects like prescription, marketing, formulation, strength, dose, route of administration, or setting of administration might be identified that help separating the name candidate from approved drug names. Confusability scrutiny may start in the web application but is mainly conducted outside of it.

Once the name candidate is considered to not be confusable, cleared for having promotional, offensive or inappropriate connotations, and free of misrepresentations, it can be submitted to HAs/TMOs and the proposed process ends in best case with name approvals.

The similarity measure to conduct the similarity searches within the web application was selected from 10 different similarity families, of which 6 were re-programmed from literature and 4 were invented new within this thesis. NED_via_strdist_v4_nal_p1_s0_min is determined to be the optimal measure regarding precision/recall performance on a test set containing confusable drug names and a test set of ‘drug name - altered drug name‘ pairs, as well as computation speed.

By comparing similarities between approved drugs in the web application database and acknowledging US GLs, a cut-off value 0.7 was determined as maximum similarity a name candidate should have to be sufficiently different from approved drug/trade names. To be comparable with the FDA similarity assessment tool (POCA), the POCA Orthographic Score was modelled as a second similarity measure and included into the web application.

Names in the GlobalProducts database are single word names by 59.9%, have a name length of mostly 7 or 8 characters, and consist mostly of 3 syllables. Letters ‘A’ and ‘E’ are most often used vowels while ‘U’ is underrepresented. Letters ‘R’, ‘N’, and ‘T’ are the most often used consonants. Together with statistics on 2- and 3-grams, the understanding of such properties might be helpful when a name candidate is altered in the ‘search and change‘ cycle.

It should be noted that the name development process is set-up to disqualify a name candidate for which (highly) similar approved drug/trade names are found while understanding that the similarity measures used in the web application focus on orthographic similarity. Additional criteria need to be fulfilled to receive HA approval, including but not limited to non-confusability, avoiding promotional, offensive or inappropriate connotations, and avoiding misrepresentations. 

When these aspects are considered, the web application should be a useful tool for developing a name candidate approvable by EMA and FDA as well as acceptable for trademark registration.

Pages: 134

Annexes: 4; Pages: 89