(M1130-11-62) Network Analysis of Pharmaceutical Products from Hong Kong PIC/S GMP Manufacturers

Purpose: The Pharmaceutical Inspection Co-operation Scheme (PIC/S) Annex 15 states that “Limits for the carryover of product residues should be based on a toxicological evaluation” [1]. At the time of this study, there are 21 PIC/S GMP (Good Manufacturing Practice) manufacturers with products in production, and all of them are voting members of the Hong Kong Pharmaceutical Manufacturers Association (HKPMA). There are at least a total of 4084 products manufactured locally and registered. To assess the feasibility of significantly reduction on the workload and cost of the toxicological evaluation for all active ingredients, we conducted two surveys through HKPMA in February 2023 and March 2024. In this work, we will analyze the unique characteristics of Hong Kong pharmaceutical products using network visualization techniques.

Methods: The basic product information such as Product Name, Permit No., Active Ingredients and Company Name were downloaded from the website of Hong Kong Drug Office [2] for the 21 manufacturers. An Excel template for each manufacturer was prepared to allow the manufacturer to check its own product information and to add required supplement information not shown on Drug Office’s website. The manufacturer could add their new products or delete those expired products following the format of the Excel template. Such Excel templates were sent to the 21 manufacturers by HKPMA, and the collected results were combined and cleaned. At last, a total of 4084 products were collected from the 21 manufacturers. It was found that the same ingredient was written in different formats in different products (i.e., cloxacillin (as sodium) and cloxacillin sodium), and the same API (Active Pharmaceutical Ingredient) may have different salts (i.e., diclofenac potassium and diclofenac sodium). To avoid ambiguity, a unique code has been assigned to each active ingredient. For example, the unique codes for both diclofenac potassium and diclofenac sodium are “A00120”. As a result, a total of 438 active ingredients were identified. In addition, we used the Permit No. of each product as their unique code and codes of M01 to M21 were assigned to the 21 manufacturers according to their names’ alphabetical order. Using Neo4j (Version 5.12.0), a graph database was developed with the three types of nodes (Manufacturer, Product and API) and two kinds of edges (HAS_PRODUCT to connect one Manufacturer and one Product, HAS_API to connect one Product and one API). All nodes have at least two properties (code and name). Additional properties for a Product node include Dosage Form, Sales Requirement, Legal Classification, Administration Route, and Count of Ingredients. Exploration analysis of the graph database was performed in Neo4j Browser Desktop and visualization of the whole network was implemented in Neo4j Bloom (Version 2.11.0).

Results: The two surveys identified 4084 pharmaceutical products that required toxicological evaluation. As shown in Figure 1, most products (yellow nodes) are concentrated in the central location of the developed Manufacturer-Product-API network. The centered products are connected by a few APIs and involve majority manufacturers, which show that Hong Kong manufacturers’ products are highly homogeneous. About 5% APIs (22) are found in 49% products (2017 of 4084). This could be further elucidated by a sub-network of the top 4 APIs (chlorphenamine maleate, phenylephrine hydrochloride, caffeine citrate and paracetamol) in Figure 2. About 25% products (1040 among 4084) of 19 manufacturers contains at least one ingredient from the top 4 APIs. On the other hand, properties of products demonstrated the following unique characteristics as demonstrated in Figure 3: (A) the top 4 dosage forms are Tablets, Syrup, Cream and Capsules which account for 82% of the total 4084 products; (B) most of products are administered orally (74%) or topical/external used (23%) and only 3% of products were given through other administration routes, such as Parenteral, Nasal, Rectal, etc. (C) for sales requirement, 41% of products need no prescription and 40% of products do need a prescription and the other 19% of products could be sold in Pharmacy; (5) for legal classification, 71% of products are Part 1 or 2 poison and the others are not a poison (23%) or Antibiotic (6%).

Conclusion: A graph database was developed for all PIC/S GMP manufacturers’ products in Hong Kong. Visualization analysis indicated that most of the 4084 products are oral tablets/capsules/syrup or topical/external creams whose ingredients are mostly shared by multiple products and manufactured by multiple manufacturers. To improve the efficiency of toxicological evaluation and avoid the waste of social and regulatory resources, based on the characteristics of Hong Kong drugs, we recommend that the toxicological evaluation of all drugs should be conducted in a centralized manner.

References: [1] https://picscheme.org/users_uploads/news_news_documents/ps_inf_11_2015_pics_gmp_revised_annex_15.pdf (Visited at April, 2024)
[2] https://www.drugoffice.gov.hk/eps/do/en/consumer/news_informations/reListRPP_index2.html (Visited at April, 2024)

Acknowledgements: This work is financially supported by Innovation Technology and Commission and Hong Kong Pharmaceutical Manufacturers Association of the Hong Kong Special Administrative Region, China (Reference number: ITP/005/23LP)


Figure 1. A network constructed by the 4084 products (yellow nodes) from 21 manufacturers (blue nodes) and their 438 APIs (red nodes) in these products. All these products, manufacturers and APIs were connected by their relationships (lines).


Figure 2. A sub-network of the top 4 APIs (chlorphenamine maleate, phenylephrine hydrochloride, caffeine citrate and paracetamol, red nodes from left to right) and their related products and manufacturers.


Figure 3. Packed bubble charts for Dosage Form (A), Administration Route (B), Sales Requirement (C) and Legal Classification (D) of the 4084 products registered in Hong Kong. Different categories use different colors, and the size of the balls is adjusted according to their count of products in these categories.