Standardized clinical test quality evaluation method

Abstract

The invention relates to a standardized clinical test quality evaluation method, and belongs to the technical field of clinical test quality control. According to the evaluation method, a clinical test quality control report is processed through a digital tool, and on the basis of initial quality control discovery classification and a keyword library, the quality control discovery classification and the keyword library are continuously improved through a self-training method; and forming a quality control discovery classification rule when the matching degree of the identifiable element information is improved to be greater than 95%. Subsequently, summarizing unmatched findings in a newly generated quality control report, and analyzing and perfecting through manual and mathematical statistical tools to form a new quality control discovery classification rule; and the quality control discovery classification rule is self-upgraded and updated through a self-training method until a new quality control discovery classification rule is obtained. According to the method, the quality and credibility of clinical test data are further improved, a common and measurable quality standard is provided for a clinical test institution, and the standardized development of the clinical research industry is promoted.

Description

Technical Field

[0001] The invention relates to a standardized clinical trial quality assessment method, belonging to the technical field of clinical trial quality control. Background Art

[0002] At present, there are 1,594 registered drug clinical trial institutions in my country. Due to differences in regional development, hospital size, practical experience and other objective factors, there are huge differences in trial quality management among clinical trial institutions. In the existing technology, there is a lack of standardized, feasible and replicable quality evaluation standards. Some existing quality management evaluation methods often take up a lot of manpower and material resources, and cannot detect risks well, and cannot prevent risks from turning into problems, which seriously restricts the progress of innovative drug research and development in my country.

[0003] Therefore, this technical field is in urgent need of a standardized clinical trial quality assessment method, through which a set of quality control discovery classification rules can be obtained, and this method can enable the existing quality control discovery classification rules and quality standards to be continuously iterated and upgraded, ultimately forming a set of clinical trial quality assessment standards that can be widely accepted by the industry and on which consensus can be reached, so as to promote the high-quality development of my country's biopharmaceutical industry. Summary of the invention

[0004] The purpose of the present invention is to solve the technical problem of how to obtain a standardized clinical trial quality assessment method, and to obtain standardized and self-updatable quality control discovery classification rules through this method to achieve consistency and uniformity of quality assessment.

[0005] In order to achieve the purpose of solving the above problems, the technical solution adopted by the present invention is to provide a standardized clinical trial quality assessment method, comprising the following steps:

[0006] Step 1: Obtain clinical trial quality management data, including clinical trial quality control reports;

[0007] Step 2: Pre-process the clinical trial quality control report through digital tools to obtain standardized governance data;

[0008] Extract element information by segmenting the text fields of clinical trial quality control reports;

[0009] Form an initial quality control discovery classification and keyword library through a small number of quality control reports; based on the initial quality control discovery classification and keyword library, continuously improve the quality control discovery classification and keyword library through self-training methods;

[0010] Step 3: Referring to the initial quality control discovery classification, continue to expand the quality control discovery and keyword library until the matching degree of identifiable element information is increased to greater than 95%, forming an initial quality control discovery classification rule 1 based on 4 levels and 54 types of discoveries; the initial quality control discovery classification rule 1 is specifically Table 1, Quality Control Discovery Classification Rule V1.0.

[0011] Table 1: Quality Control Finding Classification Rules V1.0

[0012]

[0013]

[0014]

[0015]

[0016]

[0017] Preferably, the above steps further include:

[0018] Step 4: Analyze the newly generated quality control report data by calculating the matching rate of identifiable element information; analyze the reasons why the matching rate is less than 95%;

[0019] Step 5: Summarize the unmatched findings in the newly generated quality control report, analyze and improve the unmatched findings through manual and mathematical statistical tools to form new quality control finding classification rules;

[0020] At the same time,

[0021] Step 6: Self-upgrade and update of quality control discovery classification rules: Use the self-training method of semi-supervised learning of digital tools to expand the sample set of training element information, repeatedly use the training results in each round of training, and continuously iterate and improve the quality control discovery classification rules;

[0022] Step 7: Repeat steps 4-6 above and calculate the matching rate of identifiable element information until the matching rate of element information reaches more than 95%, thus obtaining a new quality control discovery classification rule.

[0023] Preferably, the above step 2 also includes converting clinical trial quality control reports from various formats into structured data and importing them into a database to form structured data that can be processed; unifying the format of quality control report data written by different quality control personnel to ensure data consistency; and using natural language processing (NLP) tools to segment text fields and extract element information.

[0024] Preferably, the above step 6 also includes:

[0025] Step 6.1: Use the self-training method of semi-supervised learning with digital tools to expand the sample set of training element information, repeatedly use the training results in each round of training, and continuously iterate and improve the quality control discovery classification rules;

[0026] Step 6.2: Two data sets are formed through data collection, including labeled data, i.e., data with identifiable elements with keywords, and unlabeled data, i.e., continuously accumulated quality control report data;

[0027] Divide the labeled data into training and validation sets;

[0028] Use the labeled training set to train an initial supervised learning model and use the naive Bayes classifier model to perform the text classification task;

[0029] Use the initial model to predict the unlabeled data and obtain the predicted label and confidence;

[0030] Use high-confidence prediction results as new training data to expand the labeled training set;

[0031] Retrain the learning model using the expanded training set;

[0032] Cycle self-training, continue to cycle the above steps until the high confidence prediction results in the unlabeled data no longer increase or reach the predetermined number of iterations;

[0033] Step 6.3: Evaluate the final learned model using the validation set;

[0034] Step 6.4: The correspondence between keywords and elements and the automated verification results during the cyclic self-training process are compiled into a report for expert review.

[0035] The present invention provides quality control finding classification rules V1.0 and applications as shown in Table 1 above, obtained according to the above-mentioned standardized clinical trial quality assessment method. The quality control finding classification rules V1.0 include 5 categories, namely data, quality, regulatory compliance, safety and ethics, and finance. The 5 categories are further divided into 4 levels, namely levels 1-4 are progressively described in classification, and level 4 includes a total of 54 findings.

[0036] The present invention provides upgraded quality control finding classification rules V2.0 and applications as shown in Table 2 below, obtained according to the above-mentioned standardized clinical trial quality assessment method.

[0037] Table 2: Quality Control Finding Classification Rules V2.0

[0038]

[0039]

[0040]

[0041]

[0042]

[0043]

[0044]

[0045]

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention provides a standardized clinical trial quality assessment method, which helps to ensure the accuracy, reliability and consistency of clinical trial quality assessment, avoids result differences caused by adopting different evaluation standards and processes, further improves the quality and credibility of clinical trial data, and also provides common and measurable quality standards for clinical trial institutions, laying the foundation for mutual recognition of clinical trial quality data results. The present invention promotes the standardization and standardization of the clinical research industry. DETAILED DESCRIPTION

[0048] In order to make the present invention more clearly understood, the preferred embodiments are described in detail as follows:

[0049] As one of the first batch of clinical trial institutions in China, the Affiliated Cancer Hospital of Fudan University, the inventor of the present invention, has accumulated more than 2,000 clinical trial quality management experiences of various types, and through retrospective analysis of quality data over the years, has creatively formed a set of quality control discovery classification rules including data, quality, regulatory compliance, safety and ethics, finance and other five categories and four levels, for clinical trial institutions to carry out quality evaluation. However, in the actual application process, it was found that these used quality control discovery classification rules still have some problems, such as: ① There are some overlapping and unclear classification items; ② There are classification items that do not meet the classification requirements; ③ The definition and description of the classification items are not standardized and unified; ④ Some classification items are not compatible with the latest regulations or quality control requirements, etc. These problems make the used quality control discovery classification rules inconvenient and inaccurate in use. Through the used quality control discovery classification rules, the newly generated quality control discovery data of the Affiliated Cancer Hospital of Fudan University from 2022 to 2023 were retrospectively analyzed, and it was found that the matching rate between the newly generated quality control discovery data and the used quality control discovery classification rules can only reach 56.6%. In the root cause, in addition to the four internal reasons for the quality control discovery classification rules mentioned above, there are also external factors such as ① the update of quality management related regulations and the continuous improvement of quality requirements; ② the need to continuously improve the capabilities of quality controllers; ③ the continuous expansion of quality discovery. Therefore, this technical field is in urgent need of a method by which a set of quality control discovery classification rules can be obtained, and this method can continuously iterate and upgrade the existing quality control discovery classification rules and quality standards, and finally form a set of clinical trial quality assessment standards that can be widely accepted by the industry and can reach consensus.

[0050] The present invention provides a standardized clinical trial quality assessment method, the method comprising:

[0051] Step 1: Obtain clinical trial quality management data, mainly clinical trial quality control reports; clinical trial quality control reports come from clinical research quality control inspections, which are usually conducted 2-3 times at different stages of each clinical trial project implementation;

[0052] Step 2: Preprocess clinical trial quality control reports through digital tools to obtain standardized governance data; form an initial classification and keyword library of quality control findings;

[0053] Step 2.1: Convert clinical trial quality control reports from various formats (PDF, WORD, Excel) into structured data and import them into the database to form structured data that can be processed;

[0054] Step 2.2: Perform data standardization and unify the format of the quality control report data written by different quality control personnel to ensure data consistency, such as unifying the date format;

[0055] Step 2.3: Use natural language processing (NLP) tools to segment the text field and extract element information; the element information includes but is not limited to keywords.

[0056] Step 2.4: Form an initial classification and keyword library of quality control findings;

[0057] A total of 456 quality control reports on phase I-IV new drug clinical trials from 13 tumor subspecialty groups (all tumor types), 48 research teams, and 123 international and domestic sponsors from the Affiliated Cancer Hospital of Fudan University from January 2016 to February 2021 were summarized; with reference to relevant literature and professional specifications, such as the "Good Clinical Practice for Drugs" and "Key Points for On-site Verification of Drug Clinical Trial Data", key quality control findings in quality control reports were classified according to the logic of personnel qualifications, process systems, and document management to form initial classification indicators for quality control findings; in the early stage, the number of quality control findings was used as the representative Based on the principle of scientificity, 21 quality control reports with the largest number of findings were selected as the starting samples, and the applicability of the initial classification indicators was tested and optimized. At the same time, the classification of human genetic resources, protection of subjects' rights and interests, ethical review, sample management, and financial-related issues was supplemented; the starting sample size was expanded to 41 quality control reports to verify the scientificity and feasibility of the classification indicators, and the classification indicators were further improved to form the initial classification of quality control findings; based on the classification results formed by the above 41 quality control reports, the keywords based on the quality control discovery classification were used as feature values ​​for manual annotation, and the keywords were summarized to form the initial keyword library of quality control findings;

[0058] Based on the initial quality control discovery classification and keyword library, a semi-supervised learning (SSL) self-training method is used. The manually labeled quality control report is used as the initial training set, and then the training element information sample set is expanded. A large number of unlabeled quality control reports are used to expand the training sample set. The training results are repeatedly used in each round of training, and the best classification results are selected and added to the training set. The generated results are used to train itself again, so as to continuously improve.

[0059] Step 3: Analyze 456 quality control reports with reference to the initial classification index of quality control findings, and the initial matching result is 67.0%; the reasons for unrecognized problems include: 1) the description of quality control findings is unclear or ambiguous; 2) the initial keyword library does not contain identification words; the first type of problems are eliminated or converted, and the second type of problems are supplemented with manual keyword annotation; continue to expand the quality control finding classification and keyword library, and automatically classify the quality control reports again until the matching degree is improved to 96.8%, so as to form a standardized quality control finding classification and form an initial quality control finding classification rule based on 4 levels and 54 types of findings;

[0060] As shown in Table 1; Table 1 includes 5 categories including data, quality, regulatory compliance, safety and ethics, and finance, which are horizontally divided into 4 levels, namely, levels 1-4 are progressive, and the classification description is more detailed. Level 4 includes 54 findings, which are 54 findings used for quality control in clinical trials;

[0061] Examples of the classification of problems used for quality control in clinical trials, such as the problem of handwritten medical records of outpatient medication records of subjects C1D1 and the failure to use an outpatient electronic medical record system, are classified into: Level 1: Data (medication records); Level 2: Source data collection and management (handwritten medical records); Level 3: Source data collection and / or records (C1D1 outpatient medication records); Level 4: Non-standard form of source data collection and / or records (non-standard form of records due to failure to use an outpatient electronic medical record system);

[0062] The initial classification rule 1 based on 54 discoveries at 4 levels is as follows in Table 1:

[0063] Table 1: Quality Control Finding Classification Rules V1.0

[0064]

[0065]

[0066]

[0067]

[0068]

[0069] Step 4: Verification of the quality control discovery classification rules: that is, calculating the matching rate of identifiable element information;

[0070] The matching rate is calculated by counting the ratio of identifiable quality control findings to all quality control findings;

[0071] The match rate refers to the proportion of quality control findings that are correctly classified; the match rate for each category can be calculated using the following formula:

[0072] Matching rate = (number of correctly classified QC findings / total number of QC findings) x 100%

[0073] Among them, "Number of correctly classified QC findings" is the number of QC findings that are accurately classified into the corresponding categories in the rules according to the used rules, and "Total number of QC findings" is the total number of all QC findings;

[0074] The above calculation can identify the element information matching rate analysis and newly generated quality control report data;

[0075] Calculate the matching rate of identifiable element information of newly generated quality control report data;

[0076] The reasons why the analysis matching rate is less than 95% include but are not limited to the update of relevant quality management regulations and the improvement of quality requirements; the improvement of quality control capabilities of quality controllers; the continuous expansion of quality discoveries, etc.;

[0077] Step 5: Summarize the unmatched findings in the quality control report (quality control findings cannot be correctly classified), analyze and improve the unmatched findings through manual and mathematical statistical tools to form a new quality control finding classification rule;

[0078] For example: Failure to match includes, but is not limited to, repetitive questions, accuracy questions, inappropriate questions, and non-inclusion questions;

[0079] in:

[0080] 5.1. Repetitive problems were found using cluster analysis, which is an unsupervised learning method using SPSS statistical software to divide objects in a data set into groups or categories with similar characteristics to identify differences between similar problems and regular classifications;

[0081] For example, 8,246 pieces of quality management data collected between 2022 and 2023 according to the used version rules were collected and collated to ensure that the data format was unified and contained necessary description fields;

[0082] The input for cluster analysis is usually a keyword or phrase describing the discovery, including the discovery category, description, and key terms, such as EDC+not entered.

[0083] By using hierarchical clustering, keywords are captured and grouped into the same cluster, and the quality control reports within each cluster are analyzed to mainly identify quality control findings of the same category but classified into different categories;

[0084] Cluster analysis revealed controversial situations where the same problem existed in multiple categories but was classified into different rule categories. For example, in the four-level classification, the problem of "plan violation" appeared in different categories, which caused trouble for quality control work;

[0085] For repetitive problems: By combining cluster analysis results with different authoritative quality control standards and expert experience, the overlapping rules are refined and re-summarized to eliminate ambiguity and improve the clarity of the rules;

[0086] 5.2. Accuracy issues are discovered through manual identification;

[0087] The accuracy problem is that the feature information is too broad or not specific, lacking sufficient details, making it difficult to match the exact classification;

[0088] The problem of "inconsistent records of the informed process" is not clearly defined. In response to these problems, the rules are further refined and improved.

[0089] Regarding the accuracy issue: refine the definition by combining different authoritative quality control standards and expert experience;

[0090] 5.3. Inappropriate questions are identified manually;

[0091] Inappropriate questions are defined as examples that are not clear or relevant enough to accurately reflect the content of the rule definition;

[0092] For inappropriate questions: Improve the description by combining different authoritative quality control standards and expert experience;

[0093] 5.4. Problems not included are discovered manually and responsible persons are added to each rule;

[0094] The non-inclusion problem is defined as missing key information, which makes the rule unable to fully meet the needs of users;

[0095] For example, some examples in some rules were not clear or relevant enough to accurately reflect the content of the rule definition. To address this issue, industry experts were invited to review and supplement the examples in the rules and confirm the direct person in charge of each link in the rules.

[0096] At the same time,

[0097] Step 6: Quality control finds that the classification rules are self-updated and updated:

[0098] Step 6.1: Use the self-training method of semi-supervised learning with digital tools to expand the sample set of training element information, repeatedly use the training results in each round of training, and continuously iterate and improve the quality control discovery classification rules;

[0099] Step 6.2: Two data sets are formed through data collection, including labeled data (data with identifiable elements of keywords) and unlabeled data (continuously accumulated quality control report data);

[0100] Divide the labeled data into training and validation sets;

[0101] Use the labeled training set to train an initial supervised learning model and use the Naive Bayes classifier model to perform text classification tasks;

[0102] Use the initial model to predict the unlabeled data and obtain the predicted label and confidence;

[0103] Use high-confidence prediction results as new training data to expand the labeled training set;

[0104] Retrain the learning model using the expanded training set;

[0105] Cycle self-training, continue to cycle the above steps until the high confidence prediction results in the unlabeled data no longer increase or reach the predetermined number of iterations;

[0106] Step 6.3. Evaluate the final learning model using the validation set;

[0107] The correspondence between keywords and elements in the cyclic self-training process and the automatic verification results are generated into a report for experts to proofread;

[0108] Step 7: Repeat steps 4-6 above to calculate the matching rate of identifiable element information to more than 95%.

[0109] Through the above process, combined with different authoritative quality control standards, expert experience and communication experience, a new version of the quality control findings classification rules V2.0 (Appendix 2) was formed. The quality control findings classification rules V2.0 improve the clarity, accuracy and practicality of the rules, ensuring the consistency and accuracy of quality control work.

[0110] See Table 2 for details: Quality Control Finding Classification Rules V2.0

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121] Application of QC Finding Classification Rules:

[0122] 1. Application by quality control inspectors:

[0123] Beneficial effects of the application: ① Follow the unified quality control discovery classification rules to reduce the deviation caused by professional knowledge and ability of different quality control personnel, and ensure the objectivity of quality assessment. ② Quality control personnel carry out project quality control according to the unified quality control discovery classification rules, and form a standardized quantitative and analyzable quality control report with traceable classification; ③ Based on the quantitative and analyzable project quality control, the root cause analysis and attribution of quality control findings of different types and frequencies are carried out, and the project team is urged to formulate corresponding rectification measures and tracking mechanisms to ensure that various quality control findings are accurately improved or rectified.

[0124] 2. Application in clinical trial institutions:

[0125] Beneficial effects of the application: ① Based on the quality control data of quality control discovery classification rules, it breaks the limitations of traditional single points and single projects, and conducts quantitative multi-dimensional data analysis. In the researcher dimension, it can analyze the quality control data of any researcher's previous clinical trial projects, analyze the distribution of quality control findings, and make horizontal comparisons between different researchers. It can also conduct real-time vertical and horizontal analysis and summary of multiple dimensions such as sponsors and key findings. ② Based on the actual situation, set up a risk warning mechanism, quantify and predict risks, accurately guide the research team to pay close attention, and further improve the data management of the clinical trial process, and carry out targeted training.

[0126] 3. Apply the quality control findings classification rules to evaluate the quality of clinical trials:

[0127] ① The quality control of traditional clinical trial institutions is based on internal requirements to issue quality control reports. The deviation of the experience of quality control personnel can easily cause differences in quality control quality. Moreover, since the quality control results are independent and incoherent, it is impossible to summarize the rules from multiple quality controls, let alone find the quality trend. The clear quality control discovery classification rules provided by the present invention make it easier for quality control personnel to quickly find the appropriate discovery items, shorten the quality control time and improve the standardization of quality control reports, making it easier to summarize the rules and find the quality trend.

[0128] ②Quality control personnel can combine the quality control discovery classification rules to clearly indicate the key focus and scope of quality control inspections, so as to promote different quality control personnel to more comprehensively discover risks and then prevent risks from turning into problems. Targeted inspections of clinical research data integrity, protocol compliance, safety monitoring, data accuracy, and source data file management can improve the accuracy of quality control discovery identification, make quality control reports more detailed, and unify evaluations.

[0129] ③ Ensure the consistency of quality control result evaluation. According to the quality control discovery classification rules, the distribution of project problems, the frequency of problem occurrence and the key issues can be viewed, and warnings can be issued to research participants. Research participants can rectify the project based on the report, reach an objective and consistent quality evaluation, promptly learn about quality defects, and take effective preventive measures.

[0130] Through the method of the present invention, as the number of quality control reports increases, a set of constantly updated and upgraded quality control discovery classification rules can be obtained, so that the existing quality control discovery classification rules and quality standards can be continuously iterated and upgraded, and finally a set of clinical trial quality assessment standards that are adapted to a certain time and space and can be widely accepted by the industry in this technical field and can reach a consensus can be formed, which lays the foundation for establishing unified quality assessment standards with sponsors, regulators, etc., so as to promote the high-quality development of my country's biopharmaceutical industry.

[0131] The above is only a preferred embodiment of the present invention, and is not any formal or substantial limitation of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and supplements without departing from the present invention, and these improvements and supplements should also be regarded as the protection scope of the present invention. Any technician familiar with this profession, without departing from the spirit and scope of the present invention, can make some changes, modifications and equivalent changes made by using the technical content disclosed above, which are equivalent embodiments of the present invention; at the same time, any changes, modifications and evolutions of any equivalent changes made to the above embodiments based on the essential technology of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A standardized method for assessing the quality of clinical trials, It is characterized in that The following steps are involved: Step 1: Obtain clinical trial quality management data, including clinical trial quality control reports; Step 2: Pre-process the clinical trial quality control report through digital tools to obtain standardized governance data; Extract element information by segmenting the text fields of clinical trial quality control reports; Form an initial quality control discovery classification and keyword library through a small number of quality control reports; based on the initial quality control discovery classification and keyword library, continuously improve the quality control discovery classification and keyword library through self-training methods; Step 3: Referring to the initial quality control discovery classification, continue to expand the quality control discovery and keyword library until the matching degree of identifiable element information is increased to greater than 95%, forming an initial classification rule 1 based on 4 levels and 54 types of discoveries; the initial classification rule 1 is the quality control discovery classification rule V1.0 described in Table 1 in the manual.

2. A standardized clinical trial quality assessment method according to claim 1, It is characterized in that The steps also include: Step 4: Analyze the newly generated quality control report data by calculating the matching rate of identifiable element information; analyze the reasons why the matching rate is less than 95%; Step 5: Summarize the unmatched problem findings in the newly generated quality control report, analyze and improve the unmatched problem findings through manual and mathematical statistical tools to form a new quality control finding classification rule; At the same time, Step 6: Self-upgrade and update of quality control discovery classification rules: Use the self-training method of semi-supervised learning of digital tools to expand the sample set of training element information, repeatedly use the training results in each round of training, and continuously iterate and improve the quality control discovery classification rules; Step 7: Repeat steps 4-6 above and calculate the identifiable element information matching rate until the element information matching rate reaches more than 95% to obtain a new quality control discovery classification rule.

3. A standardized clinical trial quality assessment method according to claim 1, It is characterized in that The step 2 also includes converting the clinical trial quality control report from various formats into structured data and importing it into a database to form structured data that can be processed; unifying the quality control report data written by different quality control personnel into a unified format to ensure data consistency; Use natural language processing (NLP) tools to segment text fields and extract feature information.

4. A standardized clinical trial quality assessment method according to claim 2, It is characterized in that The step 6 also includes: Step 6.1: Use the self-training method of semi-supervised learning with digital tools to expand the sample set of training element information, repeatedly use the training results in each round of training, and continuously iterate and improve the quality control discovery classification rules; Step 6.2: Two data sets are formed through data collection, including labeled data, i.e., data with identifiable elements with keywords, and unlabeled data, i.e., continuously accumulated quality control report data; Divide the labeled data into training and validation sets; Use the labeled training set to train an initial supervised learning model and use the naive Bayes classifier model to perform the text classification task; Use the initial model to predict the unlabeled data and obtain the predicted label and confidence; Use high-confidence prediction results as new training data to expand the labeled training set; Retrain the learning model using the expanded training set; Cycle self-training, continue to cycle the above steps until the high confidence prediction results in the unlabeled data no longer increase or reach the predetermined number of iterations; Step 6.3: Evaluate the final learned model using the validation set; Step 6.4: The correspondence between keywords and elements and the automated verification results during the cyclic self-training process are compiled into a report for expert review.

5. According to a standardized clinical trial quality assessment method described in claim 1, the quality control finding classification rules V1.0 and their applications in Table 1 of the instructions are obtained, wherein the quality control finding classification rules V1.0 include 5 categories: data, quality, regulatory compliance, safety and ethics, and finance. The 5 categories are divided into 4 levels, i.e., levels 1-4 are progressive in sequence, and the classification description includes 54 questions in total in level 4.

6. A standardized clinical trial quality assessment method according to claim 2 obtains the quality control finding classification rules V2.0 and application in Table 2 of the instructions.

Images