A model for assessing drug manufacturing risk index, its construction method, and its application.
Patent Information
- Application Number
- CN202411012431.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2025-10-28
- Estimated Expiration
- 2044-07-26
AI Technical Summary
Existing technologies fail to quantify and manage drug risk factors, resulting in insufficient drug regulatory effectiveness and making it difficult to achieve scientific and rigorous risk management and early warning.
A drug production risk index model is constructed. The weight values and risk coefficients of each indicator are determined by the pecking order graph method. The model comprehensively considers the inherent risks, production process risks, quality control and assurance risks, compliance risks and product complaint and recall risks in the drug production process, thus forming a TRI model.
It has achieved quantitative management of drug risk factors, and the assessment results are in good agreement with the expert evaluation results, which has improved the pertinence and timeliness of drug supervision. It can dynamically assess the risks of drug production lines and promptly initiate risk disposal procedures.
Smart Images
Figure QLYQS_1 
Figure QLYQS_2 
Figure QLYQS_3
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence applications in drug regulation, specifically to a model for assessing drug production risk index, its construction method, and its application. Background Technology
[0002] The newly revised Drug Administration Law explicitly states that drug administration should be centered on people's health, adhere to the principles of risk management, whole-process control, and social co-governance, establish a scientific and rigorous supervision and management system, comprehensively improve drug quality, and ensure the safety, efficacy, and accessibility of drugs. Thus, the concept of risk management has been explicitly introduced into the field of drug regulation in the law. To innovate drug regulation methods, explore ways to improve regulatory efficiency, save administrative resources, and control drug production quality risks, this paper attempts to utilize information technology and risk management technology to establish a drug production risk index model, scientifically classify drug production risk levels, establish a risk early warning mechanism, and implement a risk-based drug inspection plan, thereby improving the ability of drug regulatory departments to control drug production quality risks.
[0003] The US FDA's drug regulatory model is widely recognized and admired by the global pharmaceutical industry, government departments, and international organizations for its transparency and authority. From the perspective of drug inspection, some of its practices are worth learning from. For example, faced with the reality that drug inspection demands far exceed the regulatory body's capacity, in 2005, the US FDA began implementing a risk-based site selection model in its routine cGMP surveillance inspections. This model prioritizes and plans inspections of production sites, creating a site surveillance inspection checklist. In 2012, the US FDA amended Section 510(h) of the Federal Food, Drug, and Cosmetic Act (FD&C Act), changing the frequency of site surveillance inspections from a fixed minimum of once every two years to a risk-based inspection planning strategy. For example, 25-30% of the total cost of clinical trials and approximately 70% of the time are spent on source data verification and error correction during on-site monitoring, while its significance in improving the quality of clinical trials is debatable. In 2013, in response to the problems existing in on-site clinical trial verification and to reduce on-site monitoring costs and improve monitoring efficiency, the US FDA issued new monitoring guidance principles for clinical trials – "Clinical Research Surveillance – A Risk-Based Regulatory Approach," introducing risk management into the on-site monitoring of drug clinical trials. This change in inspection approach has played an important role in improving inspection efficiency and saving resources and costs.
[0004] Since the implementation of the newly revised Drug Administration Law in 2019, the concept of drug risk management has gradually penetrated into all aspects of drug research, production, operation, and use in my country. As a management tool, risk management plays an increasingly important role in drug production practices. Domestic scholars and regulatory departments have conducted a series of studies. For example, Zhu Jialiang et al. studied the construction ideas of the national drug sampling data sharing platform from the perspective of big data; Lin Weiqiang et al. [Research on Drug Quality Risk Early Warning Model Based on Inspection Data [J]. International IT Media Brand, 2018, 39(12):127-130.] conducted research on drug quality risk early warning model based on inspection data; Sun Lingling et al. [Current Status and Problems of Post-Marketing Sampling Inspection Model of Drugs in China [J]. Chinese Journal of Modern Applied Pharmacy, 2012, 29(8):762-765.] studied the current status and problems of post-marketing sampling inspection model of drugs in China. However, the application of this tool in drug regulatory activities has lagged behind. The Jiangsu Evaluation Center explored a risk management-based drug GMP inspection initiation mechanism [Risk Management-Based Drug GMP Inspection Initiation Mechanism [J]. Pharmaceutical Administration, 2020, 28(3), 229-232.], and established a drug production risk assessment model. Wang Guangping et al. [Design and Methodology of Drug Safety Risk Early Warning Big Data Decision Model [J]. China Food and Drug Administration, 2022(9):138-147.] conducted research on the design and methodology of drug safety risk early warning big data decision model, using the analytic hierarchy process to analyze the model design criteria of the drug safety risk early warning big data decision system. However, there is currently no practice of quantitatively managing drug risk factors. Summary of the Invention
[0005] To address the shortcomings of existing technologies in quantifying and managing drug risk factors, this invention provides a model for assessing drug production risk index, its construction method, and its application.
[0006] The first aspect of this invention provides a model for assessing the risk index of pharmaceutical production:
[0007] The drug production risk index model:
[0008] TRI = α max +β1+β2+γ+δ;
[0009] In the formula, TRI represents the risk index value of a certain pharmaceutical production line; α max β1 is the maximum inherent risk index among all online products on the production line; β2 is the production process risk index; γ is the quality control and quality assurance risk index; and δ is the product complaint and recall risk index.
[0010] The risk index value = 100 × W i ×Kj ;
[0011] In the formula, the risk index value, i.e., RI, is the risk index value of a single indicator; W i K represents the weight value of a single indicator determined using the pecking order graph method. j Risk coefficient for a single indicator determined by experts.
[0012] Calculate W using the pecking order graph method. i value:
[0013] Experts evaluated the importance of each indicator, with higher scores indicating greater importance. The average expert scores for each indicator (X1, X2, ..., X...) were calculated. n );
[0014] Compare the average values of the calculated expert scores pairwise. If index X... i Ratio index X j If it is important, then X i 1 point; if equally important, then X i 0.5 points are awarded; if indicator X j Ratio index X i If it is important, then X i 0 points; score a ij (where i, j = 1, 2, ..., n);
[0015] After comparing each pair of data to obtain the score, the scores for each row are summed to obtain the score A for each indicator. i (where i = 1, 2, ..., n);
[0016] The formula for calculating the total score for all indicators is as follows:
[0017]
[0018] Calculate the weight value W of each indicator i (where i = 1, 2, ..., n), the formula is as follows:
[0019]
[0020] K j Determination:
[0021] Risk coefficient K j This refers to the level of risk corresponding to the next lower level indicator (i.e., the fourth level indicator), which is determined by experts and set as five levels: A, B, C, D, and E, with corresponding risk coefficients of 1.0, 0.8, 0.6, 0.4, and 0.2, respectively.
[0022] The second aspect of this invention provides a method for constructing the drug production risk index model:
[0023] S1. Establish risk indicators: Based on pharmaceutical production supervision practices, establish various levels of indicators for pharmaceutical production risks from five aspects: inherent product attributes, production process elements, quality control and assurance, GMP compliance, and complaints and recalls.
[0024] S2. Questionnaire Survey: A questionnaire survey was designed to obtain expert evaluations of the importance of each indicator. Higher scores indicate greater importance. The average expert scores for each indicator (X1, X2, ..., X...) were calculated. n );
[0025] S3. Weight Calculation: The weights are calculated using the pecking order graph method.
[0026] S4. Drug Production Risk Index Model:
[0027] TRI = α max +β1+β2+γ+δ;
[0028] In the formula, TRI represents the risk index value of a certain pharmaceutical production line; α max β1 is the maximum inherent risk index among all online products on the production line; β2 is the production process risk index; γ is the quality control and quality assurance risk index; δ is the compliance risk index; and δ is the product complaint and recall risk index.
[0029] The risk index value = 100 × W i ×K j ;
[0030] In the formula, the risk index value, i.e., RI, is the risk index value of a single indicator; W i K represents the weight value of a single indicator determined using the pecking order graph method. j Risk coefficient for a single indicator determined by experts.
[0031] S5. Model Validation: Validation is carried out using the production line of a pharmaceutical manufacturing enterprise. The consistency between the risk level obtained by the model and the expert evaluation results is calculated. The kappa value is calculated. A kappa coefficient value between 0.61 and 0.80 indicates that the model evaluation and the expert evaluation results are consistent.
[0032] Furthermore,
[0033] The S1 establishes various levels of indicators for drug production risk, including 5 secondary indicators and 22 tertiary indicators. Under each tertiary indicator, experts establish 2 to 5 quaternary indicators according to the principle of factor classification, and assign corresponding risk levels based on their risk magnitude, such as A, B, C, D, and E, with corresponding risk coefficients of 1, 0.8, 0.6, 0.4, and 0.2, respectively.
[0034] Furthermore, the S3 weight calculation is as follows:
[0035] The average scores of experts for drug production risk indicators obtained through the S2 questionnaire were compared pairwise; if indicator X i Ratio index X j If it is important, then X i 1 point; if equally important, then X i 0.5 points are awarded; if indicator X j Ratio index X i If it is important, then X i 0 points; score a ij (where i, j = 1, 2, ..., n);
[0036] After comparing each pair of data to obtain the score, the scores for each row are summed to obtain the score A for each indicator. i (where i = 1, 2, ..., n); Calculate the total score for all indicators using the following formula:
[0037]
[0038] Calculate the weight value W of each indicator i (where i = 1, 2, ..., n), the formula is as follows:
[0039]
[0040] The third aspect of this invention is the application of the above-described drug production risk index model or the model constructed by the above-described drug production risk index model construction method in the assessment of drug production risk index.
[0041] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0042] 1. The model constructed in this invention realizes the practice of quantitative management of drug risk factors and is used for the assessment of drug production risk index. The assessment results have good consistency with the expert evaluation results. The kappa calculated by SPSS 26.0 is 0.697, p<0.001, and the consistency rate is 80%.
[0043] 2. The model validation results show good consistency between the drug production risk index and expert evaluation results. This model can effectively integrate resources such as enterprise basic information, drug record information, drug regulatory information, inspection and monitoring information, and human resource information to collect objective and quantifiable indicators and dynamically assess the risks of drug production lines. Based on drug regulatory resources and legal regulations, risk warning conditions can be reasonably set, and risk disposal procedures can be promptly initiated when these conditions are met, increasing the targeting and timeliness of drug regulation. Attached Figure Description
[0044] none Detailed Implementation
[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. It should be noted that unless specific conditions are specified in the examples, conventional conditions are followed. The following descriptions are merely some embodiments of this application and do not limit the present invention in any way. Those skilled in the art can make various adjustments and improvements without departing from the concept of the present invention. All technologies implemented based on the above content of the present invention fall within the protection scope of the present invention.
[0046] 1. A model for assessing the risk index of pharmaceutical production.
[0047] The drug production risk index model:
[0048] TRI = α max +β1+β2+γ+δ
[0049] In the formula, TRI represents the risk index value of a certain pharmaceutical production line; α max β1 is the maximum inherent risk index among all online products on the production line; β2 is the production process risk index; γ is the quality control and quality assurance risk index; and δ is the product complaint and recall risk index.
[0050] Risk Index (RI) = 100 × W i ×K j
[0051] In the formula, RI is the risk index value of a single indicator; W i K represents the weight value of a single indicator determined using the pecking order graph method. j Risk coefficient for a single indicator determined by experts.
[0052] W i Calculation:
[0053] Calculate W using the pecking order graph method. i Values. Experts evaluate the importance of each indicator; higher scores indicate greater importance. The average expert scores for each indicator are calculated (X1, X2, ..., X...). n The calculated average expert scores are compared pairwise. If index X... i Ratio index X j If it is important, then X i 1 point; if equally important, then X i 0.5 points are awarded; if indicator X j Ratio index X i If it is important, then X i0 points. Score a ij (where i, j = 1, 2, ..., n).
[0054] After comparing each pair of data to obtain the score, the scores for each row are summed to obtain the score A for each indicator. i (Where i = 1, 2, ..., n). The total score for all indicators is calculated using the following formula:
[0055]
[0056] Calculate the weight value W of each indicator i (Where i = 1, 2, ..., n). The formula is as follows:
[0057]
[0058] K j Determination:
[0059] Risk coefficient K j This refers to the level of risk corresponding to the next lower level indicator, which is determined by experts and set as five levels: A, B, C, D, and E, with corresponding risk coefficients of 1.0, 0.8, 0.6, 0.4, and 0.2, respectively.
[0060] 2. Research Methods
[0061] 2.1 Establish risk indicators
[0062] This study attempts to construct a risk assessment index system for pharmaceutical production. Based on my country's pharmaceutical production supervision practices, 22 sub-items of 5 major pharmaceutical production risk indicators (i.e., the tertiary indicators in Table 1) were established through recommendations from 10 pharmaceutical production experts and senior drug inspectors, covering five aspects: inherent product attributes, production process elements, quality control and assurance, GMP compliance, and complaints and recalls.
[0063] Under each tertiary indicator, experts establish 2 to 5 quaternary indicators according to the factor classification principle, and assign corresponding risk levels based on their risk magnitude, such as levels A, B, C, D, and E, with corresponding risk coefficients of 1, 0.8, 0.6, 0.4, and 0.2, respectively. See Table 1 for details.
[0064] Table 1. Level 3 and Level 4 Indicators and Their Risk Levels
[0065]
[0066]
[0067]
[0068]
[0069] 2.2 Determine the weight calculation method
[0070] Currently, commonly used weight calculation methods include: pecking order graph method, analytic hierarchy process (AHP), and entropy method. The pecking order graph method, first proposed by Moody in the United States, is a subjective judgment-based approach. It compares multiple indicators or objectives pairwise to determine their importance or priority. If indicator A is more important than indicator B, A receives 1 point; if they are equally important, A receives 0.5 points; if indicator B is more important than indicator A, A receives 0 points. This method is suitable for both qualitative and quantitative problems. This study uses the pecking order graph method for weight calculation.
[0071] 2.3 Model Construction
[0072] 2.3.1 Questionnaire Survey
[0073] Based on the content of 5 major categories and 22 minor categories of drug production risk indicators, an expert questionnaire was designed to collect expert evaluations of the importance of each indicator; the higher the score, the more important it is. The average expert score for each indicator (X1, ...) was calculated.
[0074] X2, ..., X n ).
[0075] 2.3.2 Weight Calculation
[0076] Compare the average values of the calculated expert scores pairwise. If index X... i Ratio index X j If it is important, then X i 1 point; if equally important, then X i 0.5 points are awarded; if indicator X j Ratio index X i If it is important, then X i 0 points. Score a ij (where i, j = 1, 2, ..., n).
[0077] After comparing each pair of data to obtain the score, the scores for each row are summed to obtain the score A for each indicator. i (Where i = 1, 2, ..., n). The total score for all indicators is calculated using the following formula:
[0078]
[0079] Calculate the weight value W of each indicator i (Where i = 1, 2, ..., n). The formula is as follows:
[0080]
[0081] 2.3.3 Model Establishment
[0082] 2.3.3.1 Risk indicators at each level. These indicators were scored by experts and weighted according to their respective values W. i The formula calculates the weight values of the 22 sub-indicators (i.e., tertiary indicators), as shown in Table 2. The risk coefficient refers to the level of risk corresponding to each tertiary indicator. Each tertiary indicator is further divided into 2 to 5 tertiary indicators, with corresponding risk levels set according to their risk levels: A, B, C, D, and E, with corresponding risk coefficients of 1.0, 0.8, 0.6, 0.4, and 0.2, as shown in Table 1.
[0083] Table 2. Distribution of Secondary and Tertiary Indicators and Weights
[0084]
[0085] 2.3.3.2 The drug production risk index in this study is used to assess the risk level of a drug production line. It is closely related to the attributes of the corresponding product and its production quality management system. In this paper, it refers to the drug production line risk index. The calculation method for the production line risk index is as follows: ① Risk Index (RI) = 100 × Weight × Risk Coefficient; ② Production Line Risk Index (TRI) = {Inherent Risk Index of Co-line or Dedicated Line Products} max +Production process risk index +Quality control and quality assurance risk index +Compliance risk index +Product complaint and recall risk index.
[0086] 2.3.4 Model Validation
[0087] 2.3.4.1 Expert Evaluation. Ten pharmaceutical manufacturing experts and senior drug inspectors familiar with the pharmaceutical manufacturing enterprises in the province were selected to conduct an overall evaluation of the GMP compliance of 12 pharmaceutical manufacturing enterprises. The evaluation grades were rated as Good, Good, Average, and Poor, with scores of 4, 3, 2, and 1 respectively. Each production line was classified into high, relatively high, average, and low risk levels according to its process complexity and risk, with scores of 4, 3, 2, and 1 respectively. The expert evaluation score for each production line was calculated using the following formula: Production Line Expert Evaluation Score = (4 - Average GMP Compliance Score) × Production Line Risk Level Score. A higher score indicates a higher risk level. The 50 expert evaluation scores were divided into risk levels in a 3:3:4 ratio: "++" indicates high risk (30%), "+" indicates medium risk (30%), and "-" indicates low risk (40%). The evaluation results are shown in Table 3.
[0088] 2.3.4.2 Model Evaluation. Data was entered online by 12 pharmaceutical manufacturers. Using risk assessment model software, the risk index values of 50 production lines were calculated, and the risk levels were divided into 3:3:4: "++" indicates high risk (30%), "+" indicates medium risk (30%), and "-" indicates low risk (40%). The model evaluation results are shown in Table 3.
[0089] Table 3 Risk Level Table for Expert Evaluation and Model Assessment
[0090]
[0091]
[0092] 2.3.4.3 Validation. Validation was conducted on 50 production lines from 12 pharmaceutical manufacturers. The consistency between the risk level calculated by the model and the expert evaluation results is shown in Table 4.
[0093] Table 4. Validation Results of Risk Level Model
[0094]
[0095] The kappa value calculated using SPSS 26.0 was 0.697, p < 0.001, indicating a consistency rate of 80%. Statistically, the kappa coefficient ranged from 0.61 to 0.80, demonstrating strong consistency between the model assessment and expert evaluation results, and this consistency is statistically significant. In terms of compliance, the model assessment and expert evaluation results showed a compliance rate of 80%. In summary, the validation results show that the production line risk index model's risk level assessment results for 50 production lines of 12 pharmaceutical manufacturing companies are highly consistent with the expert evaluation results. This model can be used for assessing the risk index of pharmaceutical production.
[0096] 3. Application of Results
[0097] The model validation results show good consistency between the drug production risk index and expert evaluation results. This model can effectively integrate resources such as enterprise basic information, drug record information, drug regulatory information, inspection and monitoring information, and human resource information to collect objective and quantifiable indicators and dynamically assess the risks of drug production lines. Based on drug regulatory resources and legal regulations, risk warning conditions can be reasonably set, and risk disposal procedures can be promptly initiated when these conditions are met, increasing the targeting and timeliness of drug regulation.
Claims
1. A model for assessing the risk index of pharmaceutical production, characterized in that: The drug production risk index model: TRI=a max +β1+β2+γ+δ; In the formula, TRI represents the risk index value of a certain pharmaceutical production line; α max β1 is the maximum inherent risk index among all online products on the production line; β2 is the production process risk index; γ is the quality control and quality assurance risk index; δ is the compliance risk index; and δ is the product complaint and recall risk index. The inherent risk tertiary indicators for all online products on the production line include: ingredients, category, therapeutic index, target population, route of administration, and adverse drug reactions (ADRs); production process risk tertiary indicators include: changes in key facilities and equipment, changes in raw materials and packaging materials, co-production, process complexity, degree of process control, and deviation management; quality control and quality assurance risk tertiary indicators include: personnel turnover, traceability of inspection data, and annual reporting; compliance risk tertiary indicators include: inspection categories, inspection frequency, defect categories, administrative penalties, and product sampling; and product complaint and recall risk tertiary indicators include: complaints and recalls. The risk index value = 100 × W i ×K j ; In the formula, the risk index value, i.e., RI, is the risk index value of a single indicator; W i K represents the weight value of a single indicator determined using the pecking order graph method. j Risk coefficient for a single indicator determined by experts.
2. The model for assessing a drug production risk index according to claim 1, characterized in that: Experts evaluated the importance of each indicator; the higher the score, the more important it was. The average expert scores for each indicator were calculated as X1, X2, ..., X... n ; Compare the average values of the calculated expert scores pairwise. If index X... i Ratio index X j If it is important, then X i 1 point; if equally important, then X i 0.5 points are awarded; if indicator X j Ratio index X i If it is important, then X i 0 points; score a ij Where i, j = 1, 2, ..., n; After comparing each pair of data to obtain the score, the scores for each row are summed to obtain the score A for each indicator. i , where i = 1, 2, ..., n; The formula for calculating the total score for all indicators is as follows: Calculate the weight value W of each indicator i Where i = 1, 2, ..., n, the formula is as follows: 。 3. The model for assessing a drug production risk index according to claim 1, characterized in that: Risk coefficient K j This refers to the level of risk corresponding to the next lower level indicator, namely the fourth level indicator. It is determined by experts and set as five levels: A, B, C, D, and E, with corresponding risk coefficients of 1.0, 0.8, 0.6, 0.4, and 0.2, respectively.
4. A method for constructing a pharmaceutical production risk index model as described in any one of claims 1-3, characterized in that: S1. Establish risk indicators: Based on pharmaceutical production supervision practices, establish indicators at all levels for pharmaceutical production risks from five aspects: inherent product attributes, production process elements, quality control and assurance, GMP compliance, complaints and recalls; S2. Questionnaire Survey: A questionnaire survey will be designed to obtain expert evaluations of the importance of each indicator. Higher scores indicate greater importance. The average expert scores for each indicator will be calculated as X1, X2, ..., X... n ; S3. Weight Calculation: The weights are calculated using the pecking order graph method; S4. Drug Production Risk Index Model: TRI=a max +β1+β2+γ+δ; In the formula, TRI represents the risk index value of a certain pharmaceutical production line; α max β1 is the maximum inherent risk index among all online products on the production line; β2 is the production process risk index; γ is the quality control and quality assurance risk index; δ is the compliance risk index; and δ is the product complaint and recall risk index. The risk index value = 100 × W i ×K j ; In the formula, the risk index value, i.e., RI, is the risk index value of a single indicator; W i is the weight value of a single indicator determined according to the priority diagram method; K j Risk coefficient for a single indicator determined by experts; S5. Model Validation: Validation was conducted using a pharmaceutical manufacturing company's production line. The consistency between the risk level derived from the model and the expert evaluation results was calculated, and the kappa value was calculated. A kappa coefficient value between 0.61 and 0.80 indicates that the model evaluation and the expert evaluation results have good consistency.
5. The method for constructing the drug production risk index model according to claim 4, characterized in that: The S1 establishes various levels of indicators for drug production risk, including 5 secondary indicators and 22 tertiary indicators. Under each tertiary indicator, according to the principle of factor classification, experts establish 2 to 5 quaternary indicators and assign corresponding risk levels A, B, C, D, and E according to their risk magnitude, with corresponding risk coefficients of 1, 0.8, 0.6, 0.4, and 0.2, respectively.
6. The method for constructing the drug production risk index model according to claim 5, characterized in that: The S3 weight calculation: The average scores of experts for drug production risk indicators obtained through the S2 questionnaire were compared pairwise; if indicator X i Ratio index X j If it is important, then X i 1 point; if equally important, then X i 0.5 points are awarded; if indicator X j Ratio index X i If it is important, then X i 0 points; score a ij Where i, j = 1, 2, ..., n; After comparing each pair of data to obtain the score, the scores for each row are summed to obtain the score A for each indicator. i Where i = 1, 2, ..., n; calculate the total score for all indicators using the following formula: ； Calculate the weight value W of each indicator i Where i = 1, 2, ..., n, the formula is as follows: 。 7. The application of the drug production risk index model as described in any one of claims 1-3 or the model constructed by the method for constructing the drug production risk index model as described in any one of claims 4-6 in the assessment of the drug production risk index.