(T1230-02-07) Risk-Based Decision Matrix for Identification of High-Risk Pharmaceutical Products that Need a Full-Scale Toxicological Evaluation

Purpose: As the 47th Participating Authority of the Pharmaceutical Inspection Co-operation Scheme (PIC/S), the Hong Kong pharmaceutical manufacturing industry must transition from PIC/S GMP version 11 to version 17. This includes new requirements for toxicological evaluations to establish residue limits for cleaning validation, in accordance with PIC/S Annex 15 and the EMA Guideline on Permitted Daily Exposure. A recent survey of Hong Kong manufacturers identified 4084 products from 21 manufacturers needing evaluation. This study aims to create a decision-tree matrix to streamline the process by identifying low-risk products, reducing the workload for toxicological assessment, and ensuring efficient compliance.

Methods: To develop a risk-based matrix, we collaborated with GMP (Good Manufacturing Practice) experts and toxicologists to identify key parameters from a dataset of 4084 pharmaceutical products. We selected six parameters (as listed in Table 1) that are readily accessible from either the products themselves or the APIs (Active Pharmaceutical Ingredients) contained within. The parameters derived directly from our survey of the 4084 products include: (1) Potency of API: This parameter quantifies the amount of API in the minimum dosage form of the product. For example, in solid dosage forms, it refers to the amount in one tablet, capsule, or patch. For solutions, it is the amount in 5mL, and for semi-solids like creams or ointments, it is per gram. A lower dose indicates a higher potency of the API. (2) Cumulative Exposure: This measures the duration over which the products are to be used. Longer exposure periods are associated with increased risk. (3) Route of Administration: This describes how the product is administered into the body. Routes that directly access the bloodstream, such as injections, are considered higher risk compared to indirect routes like oral or topical administrations. (4) Medical Oversight: This assesses whether the administration of the product is supervised medically. Products requiring clinical oversight are deemed higher risk. The remaining two parameters focus on the intrinsic characteristics of the APIs: (5) Carcinogenicity: Based on the Lhasa Carcinogenicity Database, APIs are categorized into three levels: Level 1: APIs identified in the database with one or more tumour findings marked as "Positive". Level 2: APIs not fitting into Levels 1 or 3. Level 3: APIs with results marked as "Negative". (6) Highly Sensitizing Active: This classification concerns the allergenic potential of the API: Level 1: Assigned to Beta-lactam APIs known for higher sensitization risk; Level 2: Used when the sensitization potential is unknown; Level 3: Applied to non-Beta-lactam APIs. Each parameter is assigned a basic score from 1 to 3 (Levels 3 to 1, respectively), with "Cumulative Exposure" and "Route of Administration" weighted at 5, and the other four parameters at 10. To compute the risk score for each API in a product, we use the equation: S_API=SUM(w₁*S₁,...,w_i*S_i,...,w₆*S₆), where S_i is the score for the i-th parameter and w_i is the weight assigned to that parameter. The overall score for a product containing multiple APIs is defined as: S_product=max(S_API,1,...,S_API,j,...,S_API,n), where S_API,j is the score for the j-th API, ensuring the product risk score reflects the highest risk API present. This method allows for a systematic and quantifiable approach to assessing the relative risk of pharmaceutical products based on their API characteristics and usage parameters which could be used to assess the need for further toxicological consideration.

Results: Using Perl scripts, we calculated risk scores for 4,084 pharmaceutical products and their respective APIs. The score distributions, shown in Figure 1, indicate that most APIs studied pose a relatively low risk, exhibiting a right-skewed distribution. Conversely, product scores display a slight shift towards a more normal, yet still right-skewed, distribution. This shift arises because a product's score is determined by the highest API score within it. We set a threshold score of 90 to identify high-risk products, with these products highlighted in red in Figure 2's Manufacturer-Product-API network. Out of the total, 171 products were classified as high-risk and require detailed toxicological evaluations for setting safe residue limits. Notably, about 37% of these high-risk products (64 products) contain triamcinolone acetonide, a corticosteroid known for its carcinogenic potential.

Conclusion: This study developed a six-parameter risk-based matrix utilizing readily accessible API and product usage data to identify high-risk pharmaceutical products for further toxicological evaluation. The methodology described provides a practical approach for relative risk assessment where complete toxicological information is unavailable, prioritizing products for detailed evaluation and enhancing safety and compliance in pharmaceutical manufacturing.

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)


Table 1. The identified six risk parameters and their weight, levels, and related rationales.


Figure 1. Histograms showing the distribution of API scores (Upper) and products scores (Bottom). The x-axis represents the calculated risk-based scores and the y-axis represents the count of Product-API combination or products.


Figure 2. Potential high risk products (red balls) in the Manufacturer-Product-API network, in which Manufacturers are light yellow balls, Products are green/red balls and APIs are grey balls.