A Method for Screening Key Quality Attributes of Drugs Based on Serious Adverse Reaction Data

Abstract

This invention relates to a method for screening key quality attributes of pharmaceuticals based on serious adverse reaction (SAR) data. It utilizes SAR data to establish appropriate indicators for classifying the risk levels of SARs among different products of the same drug variety. Furthermore, it establishes a comprehensive method for screening key quality attributes of pharmaceuticals based on SAR data to explore critical control points in the pharmaceutical manufacturing process. The developed method can indicate potential correlations between drug quality attributes and the probability of SAR occurrence, providing direction for further toxicological research and pharmaceutical quality standard research. The method is not limited by the range of quality attributes selected, seeking only intrinsic correlations within the existing scope, resulting in high reliability. By using correlation to replace causation, logistic regression reveals possible objective phenomena, providing directions for further toxicity evaluation and offering guidance and application value for pharmaceutical process improvement and standard enhancement.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a method for screening key quality attributes of drugs based on serious adverse reaction data. Background Technology

[0002] Adverse reaction monitoring refers to the process of real-time, continuous, and systematic monitoring, reporting, and management of potential adverse reactions after drug use. An adverse reaction is a harmful reaction caused by a qualified drug at normal dosage and administration, unrelated to the expected therapeutic effect. Adverse reaction monitoring can be traced back to the early 20th century. At that time, doctors and pharmacists began to realize that certain drugs could cause adverse reactions in patients and began recording and reporting these reactions. Over time, adverse reaction monitoring has become increasingly important and has gradually become an important component of drug regulatory agencies worldwide. Internationally, adverse reaction monitoring is led and guided by the World Health Organization (WHO). The WHO collects, manages, and analyzes adverse drug reaction reports globally through its international drug surveillance agency (Adverse Drug Reaction Monitoring Center). National drug regulatory agencies are responsible for domestic adverse reaction monitoring, conducting monitoring and reporting of adverse drug reactions according to the guidance and requirements of the WHO. In my country, the National Adverse Drug Reaction Monitoring Center is responsible for the technical work of reporting and monitoring adverse drug reactions nationwide. Pharmaceutical manufacturers, pharmaceutical distributors, and medical institutions should establish adverse drug reaction reporting and monitoring management systems in accordance with relevant laws and regulations, and assign dedicated personnel to be responsible for adverse reaction reporting and monitoring. Local drug regulatory authorities and health administrative departments at all levels are responsible for the management and guidance of adverse drug reaction reporting and monitoring within their respective administrative regions, and must also establish and improve adverse drug reaction monitoring institutions. In addition, the National Center for Adverse Drug Reaction Monitoring is also responsible for the collection, evaluation, feedback, and reporting of adverse drug reaction reports and monitoring, as well as the construction and maintenance of the adverse drug reaction monitoring information network.

[0003] Currently, the adverse reaction monitoring data from the National Adverse Reaction Monitoring Center is statistically analyzed using a one-case-one-report method for each drug product. In addition to the total number of reported adverse reactions, the data is categorized into general and severe cases based on the severity of clinical manifestations. The adverse reaction reporting mechanism combines voluntary and mandatory reporting, with severe adverse reactions generally reported under a mandatory mechanism. Therefore, the data completeness and representativeness of severe adverse reactions are more comprehensive than those of general adverse reactions. Commonly used adverse reaction data analysis methods include the Proportional Reporting Ratio (PRP), the Holistic Medication Analysis (MHRA), the Reporting Odds Ratio (ROR), the Information Components (IC) method, and cluster analysis. These methods all study the univariate correlation between adverse reactions and drug products. However, for drug regulatory science, the study focuses more on the complex multivariate correlations of adverse reaction risks among different categories of drugs and different products of the same drug product, which requires big data analysis. Currently, relevant data management systems primarily function as data collection and management systems, while their big data analysis capabilities based on drug quality patterns are severely lacking.

[0004] The current state of pharmaceutical manufacturing processes in China reveals poor process control capabilities, fundamentally stemming from a lack of understanding of processes and a failure to clarify and focus on the relationships between critical quality attributes (CQAs), critical material attributes (CQAs), and critical process parameters. According to the QbD principle, CQAs refer to the physical, chemical, biological, and microbiological characteristics of the final product, which should be within appropriate limits or distributions to ensure drug quality.

[61] The quality attributes of a drug include identification, content determination, content uniformity, degradation products, residual solvents, drug release or solubility, moisture content, and microbial limits. The criticality of these attributes depends primarily on the severity of harm to patients if these attributes exceed their limits. Therefore, a causal relationship may exist between drug quality attributes and adverse reaction data, but this causal relationship is unknown. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for screening key quality attributes of pharmaceuticals based on serious adverse reaction data.

[0006] The technical solution adopted in this invention is: a risk level classification method based on serious adverse reaction data, which uses serious adverse reaction case data and clinical drug dosage data as a basis, and uses serious adverse reaction weighting coefficient and / or relative serious adverse reaction weighting coefficient to measure the risk of serious adverse reactions;

[0007] The formula is as follows:

[0008] Formula 1;

[0009] Formula 2;

[0010] W sADR RW is the weighting factor for serious adverse reactions.sADR This is the weighting coefficient for relatively severe adverse reactions.

[0011] The drug key quality attribute screening method based on serious adverse reaction data is based on the risk level classification method. For each level, the multinomial logistic regression method is used to analyze each quality attribute. The correlation between the quality attribute and the serious adverse reaction is determined by the probability distribution of each quality attribute.

[0012] Preferably, the specific steps are as follows:

[0013] Step 1: Based on the serious adverse reaction data, measure the risk of serious adverse reactions using the serious adverse reaction weighting coefficient and / or the relative serious adverse reaction weighting coefficient;

[0014] Formula 1;

[0015] Formula 2;

[0016] W sADR RW is the weighting factor for serious adverse reactions. sADR This is the weighting factor for relatively severe adverse reactions;

[0017] Step 2: Select quality attributes related to the occurrence of serious adverse reactions as key points of process control, use quality attributes as independent variables and adverse reaction incidence rate as dependent variable, and use multinomial logistic regression to identify key points of process control; according to the risk level, perform multinomial logistic regression analysis for the same risk level.

[0018] (Formula 4);

[0019] Among them, drug quality attributes QA1, QA2, ... QA m If the probability of a serious adverse reaction occurring is p, then the probability of no serious adverse reaction occurring is 1-p. β 0 It is a constant. β 1 ，β 2 ，…，β m For quality attributes QA 1 QA 2 QA m The logistic regression coefficient;

[0020] Step 3: Through β j (j=1,2,…,m)The output values ​​are analyzed to determine the contribution of quality attributes to different risk levels. β j for β 1 ，β 2 ，…，β m One of them is to determine the correlation between a quality attribute and the occurrence of serious adverse reactions by using the probability distribution of the corresponding quality attribute.

[0021] Preferably, in step one, based on the number of serious adverse reactions and clinical drug usage data, a logistic regression curve of the relative serious adverse reaction weight coefficient is established, and the risk level is classified by the serious adverse reaction weight coefficient. The n+1 levels are obtained by dividing the risk level by n serious adverse reaction weight coefficient division point values.

[0022] Preferably, in step two, the quality attributes include at least two of the following: pH value, polymer content (PIMP), individual impurity content, total impurity content (IMPT), moisture value, and API content value on anhydrous basis; wherein the individual impurity content includes one or more combinations of IMPA, IMPB, IMPD, IMPE, and IMPS.

[0023] Preferably, in step two, the probability p of the occurrence of serious adverse reactions at each dividing point value is calculated, and multiple serious adverse reaction-quality attribute logistic regression models are fitted.

[0024] Preferably, in step three, β j The larger the output value, the higher the corresponding quality attribute (QA). j It is more closely associated with the occurrence of serious adverse reactions.

[0025] The advantages and positive effects of this invention are: defining a "relatively severe adverse reaction weighting coefficient" and establishing an adverse reaction analysis method suitable for group analysis, which is used for classifying the risk levels of different processes under the same product; furthermore, establishing a screening method for key quality attributes related to adverse reactions based on quality data and adverse reaction data;

[0026] The established method can indicate the possible association between the quality attributes of a drug and the probability of serious adverse reactions, providing direction for further toxicological research and drug quality standard research. The method is not affected by the selection range of quality attributes, but only seeks the intrinsic correlation within the existing range. The results are reliable, and the correlation is used to replace the causal relationship. The logistic regression results reveal possible objective phenomena and give directions for further toxicity evaluation. It has guiding significance and application value for drug process improvement and drug standard enhancement. Attached Figure Description

[0027] Figure 1 Methods and steps for screening key quality attributes of drugs based on serious adverse reaction data;

[0028] Figure 2 Adverse reaction data for cefotiam for injection from 2015 to 2019;

[0029] Figure 3 Frequency of serious adverse reactions to cefotiam for injection (N) sADR Weighting factor RW for relative severity of adverse reactions sADR Relationship diagram; (a): N sADR With RW sADR Relationship diagram; (b): ln(N) sADR ) and RW sADR Relationship diagram;

[0030] Figure 4 RW sADR Results of normal fit and logistic fit;

[0031] Figure 5 RW sADR The cumulative probability distribution and tangent point;

[0032] Figure 6 Normal probability plot of the Pearson fit residuals of the logistic regression model for cefotiam for injection. Detailed Implementation

[0033] The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0034] Current statistical methods measure drug quality by the proportion of serious adverse reactions (SARs) in total batches, with SAR data based on the number of serious cases reported by manufacturers. While this frequency of SARs (cases / 10,000 units) can reflect the relationship between adverse reactions and manufacturers to some extent, it is an absolute value calculated from data of each manufacturer (process), exhibiting independence. The correlation and comparability between results from different manufacturers are weak, making it unsuitable for direct analysis of the correlation between adverse reactions and drug manufacturing processes. To further mine key control point information related to process correlations, it is necessary to quantitatively or semi-quantitatively characterize the risk level of serious adverse reactions.

[0035] This invention relates to a risk level classification method based on severe adverse reaction data and a method for screening key quality attributes of pharmaceuticals. First, appropriate indicators are established using severe adverse reaction data to classify the risk levels of severe adverse reactions in different products of the same drug category. Then, a method for screening key quality attributes of pharmaceuticals based on adverse reaction data is established from a holistic perspective to explore the critical control points in the pharmaceutical manufacturing process. Through adverse reaction data analysis, correlation information is mined with pharmaceutical quality attributes. A big data approach is used to study the correlation between adverse reactions and key quality control attributes, and information on key process control points is screened and obtained. Identifying the factors influencing the occurrence of severe adverse reactions in pharmaceuticals through data analysis is beneficial for improving pharmaceutical manufacturing processes, thereby promoting the continuous improvement of pharmaceutical production quality.

[0036] First, appropriate indicators are established using serious adverse reaction (sADR) data to classify the risk levels of serious ADRs among different products of the same drug category. The relative values ​​of the incidence of serious ADRs are represented by the proportion of serious ADRs and the proportion of clinical drug use, respectively, and are defined as the weight coefficient of serious ADR (W). sADR ) and the relative weight coefficient of sADR (RW) sADR The formula is as follows:

[0037] (Formula 1);

[0038] (Formula 2);

[0039] Secondly, from a holistic perspective, a method for screening critical quality attributes of drugs based on adverse reaction data is established to explore the key control points of the processes involved in drug production. Specifically, critical quality attributes are identified by seeking the relationship between the probability of serious adverse reactions and drug quality attributes. Those quality attributes related to the occurrence of serious adverse reactions (such as certain impurities, pH value, API purity, etc.) are considered key points for process control. Here, quality attributes may be continuous or discrete, but the risk level of serious adverse reactions is discrete. Using quality attributes as independent variables and the incidence rate of adverse reactions as the dependent variable, multinomial logistic regression is used to identify key points for process control.

[0040] The steps for screening key quality attributes of drugs based on serious adverse reaction data are as follows:

[0041] Step 1: Establish appropriate indicators using serious adverse reaction data to classify the risk levels of serious adverse reactions for different products under the same product category;

[0042] Step 2: Select quality attributes related to the occurrence of serious adverse reactions as key points of process control, use quality attributes as independent variables and adverse reaction incidence rate as dependent variable, and use multinomial logistic regression to identify key points of process control; specifically, multinomial logistic regression analysis can be performed for the same risk level according to the risk level classification.

[0043] (Formula 4);

[0044] in, β 0 It is a constant. β 1 ，β 2 ，…，β m For quality attributes QA 1 QA 2 QA m The logistic regression coefficient;

[0045] Step 3: Determine the key quality attributes that cause each serious adverse reaction risk level through the corresponding probability distribution; specifically, through... β j (j=1,2,…,m) The output values ​​are analyzed to determine the contribution of quality attributes to different risk levels. β j for β 1 ，β 2 ，…，β m one of the.

[0046] The specific method for identifying key points of process control using multinomial logistic regression in step two is as follows.

[0047] Assuming that in the drug quality attributes QA1, QA2, ..., QA m If the probability of a serious adverse reaction occurring under certain conditions is p, then the probability of not experiencing a serious adverse reaction is 1-p. The ratio of the probability of a serious adverse reaction occurring to the probability of not experiencing a serious adverse reaction (p / 1-p) is defined as the odds of serious adverse reaction (SADR). sADR Then the logit transformation of p is:

[0048] (Formula 3);

[0049] We selected quality attributes that may be related to serious adverse reactions and established a logistic regression model of serious adverse reactions and quality attributes (Formula 4) to screen the quality attributes for key factors.

[0050] (Formula 4);

[0051] Where β0 is a constant, β 1 ，β 2 ，…，β m For quality attributes QA 1 QA 2 QA m The logistic regression coefficient. In some embodiments of the present invention, the analysis objects are classified according to the relative severity adverse reaction weight coefficients. The classification is divided into n+1 levels by n severity adverse reaction weight coefficient division points. Different groups in multiple levels form multiple models, and the number of models is n.

[0052] Define the relative risk (RR) of a serious adverse reaction. sADR (relative sADR risk), which is the ratio of the incidence of serious adverse reactions to different risk levels, describes the magnitude of the effect of quality attributes on the probability of adverse reactions. Since the incidence of serious adverse reactions is relatively low, RR... sADR It can be approximated as:

[0053] (Formula 5);

[0054] Where p0 and p1 represent the probability of serious adverse reactions occurring at different risk levels.

[0055] According to Formula 5, we can obtain:

[0056] (Formula 6);

[0057] Then RR sADR With logistic regression coefficient β j The relationship is:

[0058] (Formula 7);

[0059] Among them, QA j0 QA j1 QA j The average level at the probabilities of serious adverse reactions p0 and p1.

[0060] In step three, it can be done through β j The output values ​​are analyzed to determine the contribution of quality attributes between different risk levels, i.e., with all other conditions remaining constant, as QA... j The change in p1 corresponds to a higher risk level of serious adverse reaction than p0, indicating an increased likelihood of a serious adverse reaction occurring (exp(β)). j () times. The key quality attributes causing each serious adverse reaction risk level are determined by the corresponding probability distribution.

[0061] RW sADR For discrete continuous variables, first according to RW sADR The values ​​are converted into ordinal variables, and then multinomial logistic regression is used to screen the quality attributes and obtain a reasonable interpretation of the analysis results.

[0062] To verify the reasonableness of the obtained data, the probability distribution of Pearson residuals was used to check the quality of the model fit. The formula for calculating Pearson residuals is as follows:

[0063] (Formula 8);

[0064] Among them, o i It is the observation probability, e i It is a prediction probability. It is the standard deviation estimate of the predicted probability, where N is the number of independent samples multiplied by the number of models n.

[0065] Since Pearson residuals are the original residuals normalized by the estimated standard deviation and have the same variance, they approximate a normal distribution when the model fits the data reasonably. This invention uses a normal probability plot of Pearson residuals to compare the distribution of Pearson residuals with a normal distribution to check whether it follows a normal distribution. The established method is validated using adverse reaction data from various manufacturers. Based on the average quality attribute levels of each manufacturer, a multinomial logistic regression model is used to weight severe adverse reactions (RW). sADR Predict the distribution range.

[0066] This method analyzes adverse reaction data and mines its correlation with drug quality attributes, establishing a method for obtaining key process control point information based on correlation research using big data thinking. It employs a product-manufacturer model for group analysis, ensuring sufficient representativeness of the manufacturer group and minimizing the impact of individual sample size, thus enabling the application of big data thinking to the analysis of group data of any size. Adverse reactions are direct evaluation indicators of drug quality and safety. A "relatively severe adverse reaction weighting coefficient" is defined, and an adverse reaction analysis method suitable for group analysis is established for classifying different process risk levels within the same product. Furthermore, a screening method for key quality attributes related to adverse reactions based on quality data and adverse reaction data is established. The developed method can reveal potential correlations between drug quality attributes and the probability of severe adverse reactions, providing direction for further toxicological research and drug quality standard research, and offering guidance and application value for drug process improvement and standard enhancement.

[0067] The present invention will now be described in conjunction with specific embodiments.

[0068] Example:

[0069] Taking the statistical data of various adverse reactions of cefotiam for injection from 2015 to 2019 as an example (e.g.) Figure 2 According to the classification of affected systems, skin-like reactions such as rash and itching accounted for the majority, while gastrointestinal reactions such as nausea, vomiting, abdominal pain, and diarrhea were also observed, as well as cardiovascular reactions such as chest tightness, dizziness, and palpitations. The adverse reaction reporting mechanism combines voluntary and mandatory reporting, with serious adverse reactions generally reported under the mandatory mechanism. Therefore, the data on serious adverse reactions is more complete and representative than that of general adverse reactions. Thus, this invention uses serious adverse reaction data for correlation analysis. The number of serious cases in the serious adverse reaction data is based on the corresponding number reported by the manufacturer.

[0070] 1. Using adverse reaction data of cefotiam for injection from 2015 to 2019, an index was established to classify the risk level of serious adverse reactions.

[0071] Table 1 shows the number of serious adverse reactions and the amount of clinical medication used by different manufacturers. W is calculated according to Formula 1 and Formula 2. sADR and RW sADR Calculate the frequency N of serious adverse reactions. sADR See Table 1 for details.

[0072] Table 1. Number and frequency of serious adverse reactions to cefotaxime for injection, and weight of serious adverse reactions.

[0073]

[0074] Weighting factor W for serious adverse reactions of cefotiam for injectionsADR The results are shown in Table 1. The severe adverse reaction coefficient (SACC) for cefotaxime for injection produced by M3 was the lowest, at -0.19; while the SACC for cefotaxime for injection produced by M1 was the highest, at 1.97. These results are consistent with the characterization method for the frequency of severe adverse reactions. The relationship between the frequency of severe adverse reactions and the weighting coefficient of severe adverse reactions is as follows: Figure 3 As shown in (a), W sADR The values ​​of W are relatively evenly distributed within the range of [-0.5, 2.0], indicating that W... sADR A higher frequency of severe adverse reactions is more conducive to defining the risk level of adverse reactions and further analysis; by Figure 3 (b) It can be seen that ln(N) sADR ) and W sADR The relationship is linear, with a correlation coefficient of 0.9999. Data fitting yields N. sADR W sADR and RW sADR Relationship:

[0075] (Formula 9);

[0076] (Formula 10);

[0077] (Formula 11);

[0078] 2. Using national evaluation sampling data from 2019, combined with adverse reaction risk analysis of cefotaxime for injection over the past five years, multinomial logistic regression was used to identify key process control points.

[0079] RW of cefotiam for injection sADR As dependent variables, pH value, polymer content (PIMP), individual impurity content (IMPA, IMPB, IMPD, IMPE, IMPS (other maximum)), total impurity content (IMPT), moisture content, and API content (calculated as anhydrous) were selected as independent variables. These quality attributes were screened to identify influencing factors related to serious adverse reactions, and then the degree of influence of these factors on serious adverse reactions was analyzed.

[0080] Continuous variable RW sADR The results of normal fitting and logistic fitting are as follows Figure 4 As shown, the normal fit has a kurtosis of 0.6460, a skewness of 0.4198, and a variance of 0.1762; the logistic fit has a kurtosis of 0.6888, a skewness of 0.2292, and a variance of 0.1728. It can be seen that the logistic fit and the normal fit are highly consistent, and their cumulative probability distributions are basically the same, as shown below. Figure 5As shown, three "cutting points" C1, C2, and C3 are obtained at cumulative probabilities of 30%, 60%, and 90%, respectively, corresponding to RW. sADR The values ​​were 0.426, 0.752, and 1.184, respectively. The distribution was divided into four risk regions for logistic regression classification: RW sADR ≤0.462 is Level 1, 0.462 < RW sADR ≤0.752 is level 2, 0.752 < RW sADR ≤1.184 is level 3, RW sADR >1.184 is level 4.

[0081] Based on the national evaluative sampling data in 2019, the risk areas of quality attributes (QA) were statistically determined for pH value, polymer content (PIMP), single impurity content (IMPA, IMPB, IMPD, IMPE, IMPS (other maximum)), total impurity content (IMPT), moisture value, and API content value, as shown in Tables 2 and 3.

[0082] Table 2

[0083]

[0084]

[0085]

[0086]

[0087] Table 3 Cefotaxime for Injection (W) ADR Risk level and quality attributes

[0088]

[0089] A multinomial logistic regression model was fitted using quality attributes and adverse reaction data from 148 batches of cefotaxime for injection:

[0090]

[0091]

[0092]

[0093] like Figure 6 As shown, the fitting residuals of the adverse reaction-quality attribute model basically conform to a normal distribution, and the distribution trends of each risk level show significant differences. Therefore, the established model has a good fit and can be used for the screening and analysis of quality attributes. The model has 406 degrees of freedom, and the overall fit bias is 262.40, relative to the chi-square distribution of the 95% confidence interval. The model is significant.

[0094] The model parameters are shown in Table 4. The coefficient estimate for IMPE is the largest at 16.21, followed by pH (7.87) and PIMP (2.43). The p-value matrix results of the significance test show that p... IMPE p pH p PIMP All values ​​were less than 0.05, indicating that IMPE, pH, and PIMP all have a significant impact on the risk of serious adverse reactions. Therefore, IMPE, PIMP, and moisture are significantly correlated with the incidence of serious adverse reactions.

[0095] Table 4 Parameters of the Adverse Reactions-Quality Attribute Logistic Regression Model

[0096]

[0097] The established method was validated using adverse reaction data from various manufacturers. Based on the average quality attribute levels of each manufacturer in Table 3, the adverse reaction-quality attribute model was used to assign weights (RW) to serious adverse reactions. sADR The distribution range was predicted, and the prediction results and their 95% confidence intervals are shown in Table 5. The results show that, except for M6 which deviated slightly, the prediction results for the other manufacturers were consistent with the actual RW. sADR The results are consistent, therefore the method is quite accurate. The prediction results for M6 show that there is a 10% probability that it is consistent with the risk level in Table 3, and a 26.7% and 58.5% probability that it is risk level 2 and 3, respectively. This reveals that the manufacturer may potentially develop towards higher risks, and its serious adverse reaction data should be continuously monitored.

[0098] Table 5. Prediction results of the adverse reaction-quality attribute model and their 95% confidence intervals.

[0099]

[0100] According to the adverse reaction-value attribute model, IMPE, pH value, and PIMP correspond to the relative risk (RR) of serious adverse reactions. sADR The values ​​were 1.098E+07, 220.8, and 86.91, respectively. This means that, all other things being equal, the probability of serious adverse reactions increases by 1.098E+07 times, 220.8 times, and 86.91 times, respectively, with increasing IMPE, pH, and PIMP. Therefore, attention should be paid to the three key process control points: impurity E, pH, and polymer. Impurity E and polymer limits should be controlled, and pH should be kept within a reasonable range.

[0101] The embodiments of the present invention have been described in detail above, but the content described is only a preferred embodiment of the present invention and should not be considered as limiting the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the patent coverage of the present invention.

Claims

1. A method for screening key quality attributes of drugs based on serious adverse reaction data, characterized in that: This method is used to classify the risk levels of serious adverse reactions of different products under the same product category, and then establish a screening method for key quality attributes of drugs based on adverse reaction data from an overall perspective. Based on the number of serious adverse reactions and clinical drug usage data, the risk of serious adverse reactions is measured by the weighting coefficient of serious adverse reactions and / or the relative weighting coefficient of serious adverse reactions. Drugs produced by different processes under the same product are classified into risk levels. For each level, the multinomial logistic regression method is used to analyze each quality attribute. The correlation between quality attributes and serious adverse reactions is determined by the probability distribution corresponding to each quality attribute. The specific steps are as follows: Step 1: Based on the serious adverse reaction data, measure the risk of serious adverse reactions using the serious adverse reaction weighting coefficient and / or the relative serious adverse reaction weighting coefficient; Formula 1: Formula 2: W sADR RW is the weighting factor for serious adverse reactions. sADR This is the weighting factor for relatively severe adverse reactions; Step 2: Select quality attributes related to the occurrence of serious adverse reactions as key points of process control, use quality attributes as independent variables and adverse reaction incidence rate as dependent variable, and use multinomial logistic regression to identify key points of process control; according to the risk level, perform multinomial logistic regression analysis for the same risk level. Formula 4: Among them, drug quality attributes QA1, QA2, ... QA m If the probability of a serious adverse reaction occurring is p, then the probability of no serious adverse reaction occurring is 1-p. β 0 It is a constant. β 1 ，β 2 ，…，β m For quality attributes QA 1 QA 2 QA m The logistic regression coefficient; Quality attributes include at least two of the following: pH value, polymer content (PIMP), individual impurity content, total impurity content (IMPT), moisture value, and API content as anhydrous matter. Step 3: Through β j The output values ​​are analyzed to determine the contribution of quality attributes to different risk levels. β j for β 1 ，β 2 ，…，β m One of them is to determine the correlation between a quality attribute and the occurrence of serious adverse reactions by using the probability distribution of the corresponding quality attribute.

2. The method for screening key quality attributes of drugs based on serious adverse reaction data according to claim 1, characterized in that: In step one, based on the number of serious adverse reactions and clinical drug usage data, a logistic regression curve of the relative serious adverse reaction weight coefficient is established. The risk level is classified by the serious adverse reaction weight coefficient, and n+1 levels are obtained by dividing the risk level by n serious adverse reaction weight coefficient division point values.

3. The method for screening key quality attributes of drugs based on serious adverse reaction data according to claim 1, characterized in that: In step two, the content of a single impurity includes one or more combinations of IMPA, IMPB, IMPD, IMPE, and IMPS.

4. The method for screening key quality attributes of drugs based on serious adverse reaction data according to claim 1, characterized in that: In step two, the probability p of serious adverse reactions at each dividing point is calculated, and multiple logistic regression models of serious adverse reactions and quality attributes are fitted.

5. The method for screening key quality attributes of drugs based on serious adverse reaction data according to claim 1, characterized in that: In step three, β j The larger the output value, the higher the corresponding quality attribute (QA). j It is more closely associated with the occurrence of serious adverse reactions.

Images