Cell drug warning system for clinical research

Abstract

The invention relates to the technical field of drug early warning, in particular to a cell drug warning system for clinical research, which comprises an index anomaly marking module, an anomaly isolation cache module, a dynamic hysteresis monitoring module, a progressive risk evolution module and a warning state judgment module. According to the method, the inflammatory factor level index monitoring data of the subject are extracted in real time, anomaly distinguishing is carried out in combination with the preset monitoring interval, rapid identification of potential risk signals is achieved, dynamic tracking of the risk signals is achieved in combination with the intervention intensity grade and multi-time-point offset change sequence analysis in the hysteresis period, and the risk detection accuracy is improved. The real-time quantification of the risk evolution process is realized by calculating the offset characteristic quantity and the trend change, and the timely judgment capability of the high-risk state is improved through the dual comparative analysis of the maximum offset and the recovery time in combination with the linkage mapping of the risk level and the early warning threshold. And the identification, tracing, evaluation and early warning response levels of potential adverse reactions in clinical tests are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of drug early warning, and in particular to a cell drug alert system for clinical research. Background Art

[0002] The field of drug early warning technology involves the early identification and early warning of adverse reactions, toxic side effects and potential risks that may be caused by drugs during research and development, testing and clinical application. It covers drug safety monitoring, clinical data collection and analysis, drug interaction assessment, and the construction of risk assessment and early warning mechanisms based on historical and real-time data, including the dynamic collection and monitoring of clinical data, the establishment of risk assessment models for drug safety, the determination and reporting of early warning signals, and the construction and maintenance of drug vigilance databases. Among them, the traditional cell drug vigilance system used in clinical research refers to the process of clinical trials of cell drugs, which aims to identify adverse reactions and risk events caused by the complexity and individual differences of the biological activity of cell drugs. It usually identifies possible safety signals and risk trends by manually interpreting case report form records, summarizing and analyzing collected clinical safety data based on statistical software, referring to the changing trends of biochemical indicators in clinical observation records, and manually comparing the cell drug adverse event case library reported in previous literature.

[0003] In the current clinical cell drug research process, the main reliance is on manual interpretation of case report form information and post-summary analysis of safety data through statistical software. There is a lack of dynamic abnormality recognition capabilities for real-time clinical monitoring data. During the data collection stage, it is impossible to timely distinguish the transient abnormal changes of biomarkers such as inflammatory factors, which can easily lead to missed risk signals or delayed responses. In addition, due to the lack of immediate isolation and dynamic deviation trend analysis mechanisms for abnormal data, the evolution process of risk signals lacks a continuous quantitative basis, which in turn affects the timeliness of risk level changes. At the same time, in the face of cell drug clinical trials with large individual differences and frequent fluctuations in indicators, the accuracy and reliability of safety management can be easily reduced due to unclear historical data traceability or insufficient risk trend tracking capabilities. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a cellular drug vigilance system for clinical research.

[0005] To achieve the above-mentioned object, the present invention adopts the following technical solutions: a cellular pharmacovigilance system for clinical research, comprising an indicator abnormality marking module, an abnormality isolation and caching module, a dynamic hysteresis monitoring module, and a progressive risk evolution module; The indicator abnormality marking module obtains the clinical monitoring data of the subjects, extracts the monitoring data of the inflammatory factor level indicators, compares the monitoring values ​​with the inflammatory factor monitoring range, and adds an abnormal mark if it exceeds the range, and generates a list of abnormal mark data; The abnormal isolation cache module buffers abnormal data entries together with timestamps into the abnormal isolation cache area based on the abnormal marking data list, writes normal data into the normal data area and maintains continuous monitoring, and generates abnormal isolation cache objects; The dynamic hysteresis monitoring module extracts the intervention intensity level corresponding to the abnormal data based on the abnormal isolation cache object, obtains the inflammatory factor level indicator sequence continuously collected during the hysteresis period in combination with the hysteresis period baseline value, calculates the offset distance of each sampling point value in the sequence relative to the inflammatory factor monitoring range, and generates a hysteresis monitoring feature data list; Based on the hysteresis monitoring feature data list, the progressive risk evolution module extracts the maximum offset distance and the corresponding sampling time point, obtains the offset change sequence to determine the offset direction, calculates the stable time from the start of intervention to the first recovery, calculates the time difference with the standard hysteresis period, combines the offset pattern to form a risk evolution level value, and outputs a risk evolution analysis data set.

[0006] As a further solution of the present invention, the abnormal marking data list includes the abnormal data entry index, the abnormal indicator name, and the abnormal monitoring value; the abnormal isolation cache object includes the abnormal data timestamp, the abnormal data content, and the abnormal cache status; the hysteresis monitoring feature data list includes the monitoring sequence sampling time point, the offset distance of each sampling point, the maximum offset feature value, and the hysteresis cycle span; the risk evolution analysis data set includes the offset pattern category, the offset change trend, the intervention start time, the first stable recovery time, the time difference, and the risk evolution level value.

[0007] As a further solution of the present invention, the indicator abnormality marking module includes: The monitoring data extraction submodule obtains the real-time clinical monitoring data of the test subjects, detects the monitoring items related to the inflammatory factor levels, collects the monitoring data of the inflammatory factors, extracts the corresponding monitoring time, subject number and test value, and generates the inflammatory factor monitoring data set; The abnormal interval comparison submodule is based on the inflammatory factor monitoring data set and the corresponding inflammatory factor monitoring range. For each monitoring data, the detection value is compared with the corresponding inflammatory factor monitoring range interval, and it is determined whether the detection value exceeds the interval range. All data entries outside the interval are filtered out to obtain an interval deviation abnormality list; The abnormal mark generation submodule adds abnormal identification information to the monitoring data items that exceed the inflammatory factor monitoring range according to the interval deviation abnormal list, marks the abnormal category, abnormal degree and corresponding abnormal factor type, and generates an abnormal mark data list.

[0008] As a further solution of the present invention, the abnormality isolation cache module includes: The data tag screening submodule screens all data entries with abnormal tags based on the abnormal tag data list, calls the corresponding timestamp of each data entry, obtains the combination of the data entry and the timestamp, performs synchronous caching operations on all the combined contents, and obtains an abnormal combination cache set; The dual-area writing submodule determines the abnormal flag status of each data entry based on the abnormal combination cache set, extracts the data entries marked as normal, adjusts the sequence according to the timestamp, writes the adjusted data into the normal data area, performs continuous monitoring, and obtains the timing adjustment synchronization record; The isolation object generation submodule classifies the abnormal data entries based on the abnormal combination cache set and the timing adjustment synchronization record, obtains the continuous burst segments and concentrated offset types according to the timestamp and tag distribution, performs structure organization and label classification, and generates abnormal isolation cache objects.

[0009] As a further solution of the present invention, the dynamic hysteresis monitoring module includes: The intervention level extraction submodule extracts the intervention measure content corresponding to each abnormal inflammatory factor monitoring data recorded in the abnormal isolation cache object based on the above-mentioned object, matches the intervention behavior description with the preset intervention level table, maps the intervention text of the record entry to the standard intensity level value, classifies and organizes the extraction process according to the patient dimension, and establishes the corresponding relationship between the patient and the intervention intensity level based on the intervention behavior corresponding to each abnormal data, thereby obtaining the intervention intensity level record; The cycle sequence acquisition submodule sets the corresponding hysteresis cycle benchmark value based on the intervention intensity level record and the half-life t1 / 2 of the drug. The hysteresis cycle is set to 3 times t1 / 2 as the evaluation interval. The inflammatory factor monitoring value sequence within the collection period is extended backward based on the intervention start time point of each patient. The data pair structure is formed with the corresponding timestamp, and the monitoring data sequence is established and the corresponding monitoring time points are marked to obtain the hysteresis cycle indicator sequence. The feature quantity calculation submodule obtains the hysteresis offset value based on the hysteresis cycle indicator sequence, the inflammatory factor level indicator collected at each monitoring time point, and the upper limit value of the corresponding inflammatory factor monitoring range, thereby revealing the maximum relative offset degree within the cycle, binding it to the corresponding patient and intervention event label, and obtaining a list of hysteresis monitoring feature data.

[0010] As a further solution of the present invention, the progressive risk evolution module includes: The trend extraction submodule extracts the offset values ​​and corresponding times corresponding to each sampling time point after the start of the intervention based on the hysteresis monitoring feature data list, constructs an offset sequence array in chronological order, calculates the offset difference change sequence within all time periods, confirms the trend direction, and generates an offset trend pattern classification result; The stabilization time identification submodule retrieves the offset value of each time point after the start time of the trend stabilization section based on the offset trend pattern classification result, and determines whether the difference between the current offset value and the offset value of the previous time period is lower than the offset stabilization threshold. If the offset value difference is within the interval for three consecutive time periods, the corresponding time period is determined to be the first stable time point, and the time difference between the time point and the intervention start time is recorded to obtain the first stable time interval data; The risk level adjustment submodule sets the hysteresis period standard time according to the first stable time interval data, calculates the difference with the stable time interval value, forms a time offset index, and performs a level classification adjustment based on the progressive rule on the current object according to the offset pattern shown in the offset trend pattern classification result, calculates and obtains the risk evolution level value, and generates a risk evolution analysis data set.

[0011] As a further solution of the present invention, the system further includes an alert status determination module: The alert status determination module extracts the risk evolution level value based on the risk evolution analysis data set and compares it with the drug alert action limit. If the risk evolution level value is greater than or equal to the alert threshold level, the corresponding risk state is marked as an alert state, and a clinical cell drug alert status record is generated; The clinical cell drug alert status record includes a risk level identifier, an alert threshold level, an alert status label, an alert determination time point, and an associated risk analysis number.

[0012] As a further solution of the present invention, the alert status determination module includes: The alert value extraction submodule obtains each record in the risk evolution analysis data set, extracts the drug type information and adverse drug event level standard corresponding to the record object, constructs classification rules based on the AE level classification standard of the drug, combines the logical relationship between the pharmacological category and the clinical AE trigger record, and generates the drug alert threshold level standard; The risk comparison submodule compares the risk evolution level value extracted from the drug vigilance threshold level standard and the risk evolution level value in the risk evolution analysis data set with the corresponding alert threshold level value. If the risk evolution level value is greater than or equal to the alert threshold level value, the record is determined to be an object requiring alert status; otherwise, it is marked as a non-alert status, and an alert status identification label is obtained; The status record submodule retrieves the intervention time, risk level value, cell drug type, deviation trend classification, first stable time and risk deviation value content in turn according to the record object number identified as requiring alert in the alert status identification label, organizes and constructs the status list table structure, and numbers each record as an independent alert event unit to establish a clinical cell drug alert status record.

[0013] Compared with the existing technology, the advantages and positive effects of the present invention are: the present invention realizes rapid identification of potential risk signals by extracting the monitoring data of the inflammatory factor level indicators of the subjects in real time and distinguishing abnormalities in combination with the preset monitoring intervals, and utilizes the partitioning processing of abnormal data isolation and synchronous writing of continuous normal data to ensure the real-time integrity of the clinical monitoring data stream and the independent tracing of abnormal events. It combines the intervention intensity level with the offset change sequence analysis of multiple time points within the hysteresis period to realize dynamic tracking of risk signals in the time dimension, realizes real-time quantification of the risk evolution process by calculating the offset feature quantity and trend change, realizes accurate judgment of risk level changes by dual comparative analysis of maximum offset and recovery time, improves the ability to timely judge high-risk states by the linkage mapping of risk level and warning threshold, and effectively improves the identification, tracing, evaluation and warning response level of potential adverse reactions in clinical trials. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a system flow chart of the present invention; Figure 2 This is a flow chart of the indicator abnormality marking module of the present invention; Figure 3 This is a flow chart of the abnormal isolation cache module of the present invention; Figure 4 This is a flow chart of the dynamic hysteresis monitoring module of the present invention; Figure 5 This is a flowchart of the progressive risk evolution module of the present invention; Figure 6 This is a flow chart of the alert status determination module of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] See also Figure 1 , a cellular drug vigilance system for clinical research, including an indicator abnormality marking module, an abnormality isolation and caching module, a dynamic hysteresis monitoring module, a progressive risk evolution module and an alert status determination module; The indicator abnormality marking module obtains real-time clinical monitoring data of test subjects during the clinical research of cell-based drugs, extracts monitoring data of inflammatory factor levels, and compares the monitoring value with the corresponding inflammatory factor monitoring range (referring to the FDA biomarker qualification standard, such as IL-6 ≥ 50pg / mL is abnormal). If the monitoring value exceeds the range, an abnormal mark is added to the corresponding data entry to generate an abnormal mark data list; The exception isolation cache module filters data entries with exception tags based on the exception tag data list, buffers them together with timestamp information into the exception isolation cache area, writes normal data into the normal data area simultaneously and maintains continuous monitoring (using the real-time data flow monitoring mechanism of the HL7FHIR standard), and generates exception isolation cache objects; The dynamic hysteresis monitoring module extracts the intervention intensity level corresponding to the abnormal data in the cache based on the abnormal isolation cache object. Combined with the hysteresis period baseline value (based on 3 times the elimination time calculated based on the drug half-life: t1 / 2×3), it obtains the inflammatory factor level indicator sequence continuously collected within the hysteresis period. The offset distance of each sampling point value in the sequence relative to the inflammatory factor monitoring range is calculated, and the maximum offset distance is selected as the hysteresis monitoring feature value to generate a hysteresis monitoring feature data list. The progressive risk evolution module extracts the maximum offset distance and the corresponding sampling time point based on the hysteresis monitoring feature data list, obtains the offset change sequence, and determines the offset direction based on the trend of the offset change sequence. If the offset continues to increase, it is classified as an increasing offset mode, and if the offset continues to decrease, it is classified as a decreasing offset mode. The module calculates the stabilization time from the start of the intervention to the first recovery, calculates the difference between the stabilization time and the standard time of the hysteresis period, and adjusts the offset mode and time difference accordingly to form a risk evolution level value, outputting a risk evolution analysis data set. The alert status determination module extracts the risk evolution level value based on the risk evolution analysis data set and compares it with the drug alert action limit (referring to the AE level classification of the EudraVigilance system). If the risk evolution level value is greater than or equal to the alert threshold level, the corresponding risk status will be marked as an alert state, and a clinical cell drug alert status record will be generated.

[0017] The abnormal marking data list includes the abnormal data entry index, abnormal indicator name, and abnormal monitoring value. The abnormal isolation cache object includes the abnormal data timestamp, abnormal data content, and abnormal cache status. The hysteresis monitoring feature data list includes the monitoring sequence sampling time point, the offset distance of each sampling point, the maximum offset feature quantity, and the hysteresis cycle span. The risk evolution analysis data set includes the offset pattern category, offset change trend, intervention start time, first stable recovery time, time difference, and risk evolution level value. The clinical cell drug alert status record includes the risk level identifier, alert threshold level, alert status label, alert judgment time point, and associated risk analysis number.

[0018] See also Figure 2 , the indicator anomaly marking module includes: The monitoring data extraction submodule obtains the real-time clinical monitoring data of the trial subjects, detects the monitoring items related to the levels of inflammatory factors, collects the monitoring data of inflammatory factors (including IL-6, TNF-α, and IL-1β), extracts the corresponding monitoring time, subject number, and test value, and generates an inflammatory factor monitoring data set; To obtain real-time clinical monitoring data of trial subjects, first confirm the subject numbers one by one at the clinical trial center according to the trial subject identity registration form, corresponding to numbers S001, S002, S003, etc., and collect inflammatory factor level monitoring samples from each subject at different time periods such as 08:00, 09:00 and 10:00 on June 1, 2024, and use enzyme-linked immunosorbent assay to quantitatively detect IL-6, TNF-α, and IL-1β. During the detection process, the instrument equipment is set to the corresponding standard curve, and the sample dilution ratio is 1:5. The concentration of each sample is converted according to the standard curve. The IL-6 detection values ​​are 45.6pg / mL, 52.3pg / mL, and 48.7pg / mL, respectively, and the TNF-α detection values ​​are 18.2pg / mL, respectively. , 20.5pg / mL, 19.7pg / mL, and the IL-1β detection values ​​were 10.1pg / mL, 11.3pg / mL, and 9.8pg / mL, respectively. All test data were entered into the LIMS system. The monitoring time, subject number, and test value were integrated to generate data rows. Then, all data were judged to be null values. If the test result was missing, it was marked as "null value to be filled". For example, if the IL-1β monitoring result of S003 was null, supplementary sampling was arranged immediately. After the supplementary test, the data was updated and re-imported. The overall data was checked for consistency through the data verification program to check for abnormal values ​​or entry errors. The test value range was 0-100pg / mL for preliminary rationality judgment, and the samples with errors outside the interval were excluded. After data cleaning, the inflammatory factor monitoring data set was obtained.

[0019] Table 1 Monitoring data of inflammatory factors of subjects:

[0020] The abnormal interval comparison submodule is based on the inflammatory factor monitoring data set and the corresponding inflammatory factor monitoring range. For each monitoring data, it compares the test value with the corresponding inflammatory factor monitoring range, determines whether the test value exceeds the range, filters all data entries outside the range, and obtains a list of abnormal deviations from the range. Based on the inflammatory factor monitoring data set, the abnormal interval range set by the FDA biomarker standard was called. First, the baseline interval was set according to the IL-6 abnormal threshold, with a lower limit of 0 pg / mL and an upper limit of 50 pg / mL. The thresholds for TNF-α and IL-1β were set by reference, 0-25 pg / mL for TNF-α and 0-15 pg / mL for IL-1β. The three monitoring data of S001, S002, and S003 were selected as sample data, and the IL-6 test value was compared with its abnormal interval for judgment. The S001 test value was 45.6 pg / mL. g / mL is less than the upper limit of 50pg / mL, marked as "no", S002 test value of 52.3pg / mL is greater than the upper limit, marked as "yes", S003 test value of 48.7pg / mL is less than the upper limit, marked as "no". The direct comparison method of interval boundary values ​​is adopted in the judgment process, and the test value greater than the upper limit or less than the lower limit is used as the basis for abnormal judgment. The subsequent processing method for TNF-α and IL-1β is the same. After the preliminary judgment, the abnormal data entries are summarized one by one according to the subject number, factor name, and test value, as shown in Table 2, to form a list of interval deviation abnormalities.

[0021] Table 2 List of interval deviation anomalies:

[0022] The abnormal mark generation submodule adds abnormal identification information to the monitoring data items that exceed the inflammatory factor monitoring range according to the interval deviation abnormal list, marks the abnormal category, abnormal degree and corresponding abnormal factor type, and generates an abnormal mark data list; According to the interval deviation abnormality list, first extract all monitoring data entries with a "yes" judgment. In this embodiment, the IL-6 item of subject S002 is compared with its actual detection value of 52.3pg / mL and the abnormal threshold interval of 50pg / mL, and it is determined to be an "exceeding the upper limit" category abnormality. Further, according to the abnormality degree classification standard, the mild abnormality interval is set to exceed the upper limit by 1pg / mL to 10pg / mL. S002 exceeds the limit by 2.3pg / mL, so it is determined to be "mildly abnormal". Subsequently, call the abnormal factor category, abnormality degree, detection value and other information, append the abnormal identification field, and finally establish the abnormal mark data entry in the abnormal list to form the abnormal mark data list.

[0023] Table 3 Abnormal marking data list:

[0024] See also Figure 3 , the exception isolation cache module includes: The data tag filtering submodule filters all data entries with abnormal tags based on the abnormal tag data list, calls the corresponding timestamp of each data, obtains the combination of data entry and timestamp, and performs synchronous caching operations on all combined contents to obtain the abnormal combination cache set; Based on the abnormal marker data list, in the cell drug vigilance scenario used in clinical research, it is necessary to identify and screen the inflammatory factor level data collected by the test subjects during real-time monitoring. The abnormal marker data list comes from the operation process executed by the previous indicator abnormal marker module. This module sets the inflammatory factor monitoring range according to the FDA biomarker qualification standard and makes abnormal judgments on the collected data. For example, the normal upper limit of IL-6 is set at 50pg / mL. When the IL-6 monitoring value exceeds this value, it is marked as abnormal. In actual operation, the subject monitoring record is first extracted from the monitoring database, including the patient number, sampling timestamp and various inflammatory factor values. Here, taking the IL-6 indicator as an example, the collection records are as follows: Subject A On the first, second, and third days of the experiment, IL-6 was detected at 42 pg / mL, 58 pg / mL, and 67 pg / mL, respectively. Compared with the IL-6 monitoring threshold of 50 pg / mL, the first item was normal, and the latter two items were abnormal. Therefore, an abnormal flag value of 1 was added to the corresponding record entries on the second and third days, and a flag value of 0 was added on the first day. In the process of extracting abnormal data entries, the system performed a screening operation based on the flag field, and only buffered the records with a flag value of 1 as abnormal data. At the same time, its sampling timestamp value was extracted to construct a data combination structure. The buffer queue was configured using a first-in-first-out mechanism during caching, and each abnormal data was written to the cache with a standard structure of [patient ID, timestamp, IL-6 value, abnormal flag], as shown in Table 4. Table 4 Abnormal inflammatory factor data cache table:

[0025] As shown in Table 4, the IL-6 values ​​on the second and third days exceeded the threshold, so they were marked and cached. During the storage process, the system allocated 12 bytes of cache space for each record and set the cache overflow warning threshold to 96 bytes. The total cache space of the current record is 24 bytes, and no warning is triggered. Finally, an abnormal combination cache set is generated.

[0026] The dual-area writing submodule determines the abnormal flag status of each data entry based on the abnormal combination cache set, extracts the data entries marked as normal, adjusts the sequence according to the timestamp, writes the adjusted data into the normal data area, performs continuous monitoring, and obtains the timing adjustment synchronization record; Based on the structured records extracted from the abnormal combination cache, it is necessary to further identify and extract normal data entries marked as 0. Normal data is used to maintain the continuous clinical observation data stream and is written to the normal data area to ensure the integrity of the data record timeline. In this example, Patient A's IL-6 value on Day 1 was 42 pg / mL, which did not exceed the abnormal threshold of 50 pg / mL and was therefore determined to be a normal entry. The extracted timestamp is Day 1, or t = 1, and compared with the latest normal timestamp recorded in the current system. If the data continues the last recorded time without a jump, it is marked as time-series consistent. In the current process, the last write timestamp is recorded as 0, indicating a one-day interval and no abnormal interval. Therefore, this data can be directly written to the normal data area. During the write process, the system calls the data monitoring protocol based on the HL7FHIR standard to perform synchronization checks on the write event. The check content is to determine the continuity and sequence. If the continuous time interval exceeds 2 days, it is marked as a sequence offset. The current data does not meet this condition, and the record write is successfully completed. At the same time, the current write sequence offset is reported as 0 days, and a timing adjustment synchronization record is generated.

[0027] The isolation object generation submodule classifies abnormal data entries based on the abnormal combination cache set and timing adjustment synchronization records, obtains continuous burst segments and concentrated offset types according to timestamp and tag distribution, performs structure organization and tag classification, and generates abnormal isolation cache objects; Abnormal inflammatory factor data is categorized based on the anomaly combination cache set and time-adjusted synchronization records. In this example, the two abnormal records in the cache set are monitoring data from days 2 and 3 of the trial. The system performs time-series aggregation based on timestamps, identifying temporally consecutive abnormal data entries with a 1-day aggregation window. The difference between timestamps 2 and 3 is determined to be 1 day, which is less than the maximum window value of 2 days. Therefore, they are classified as the same sudden abnormal segment. Subsequently, the two IL-6 values ​​of 58 pg / mL and 67 pg / mL are offset and the average offset is calculated as [(58-50) + (67-50)] / 2 = 12.5 pg / mL. Segments with offsets exceeding 10 pg / mL are labeled "moderate sudden inflammatory abnormality." An abnormal structure group is also constructed, recording information such as patient number, abnormality label, maximum offset, and duration. These are ultimately integrated into a structured object for subsequent clinical data early warning aggregation and the generation of an abnormality isolation cache object.

[0028] See also Figure 4 , the dynamic hysteresis monitoring module includes: The intervention level extraction submodule extracts the intervention measures corresponding to each abnormal inflammatory factor monitoring data record based on the abnormal isolation cache object, matches the intervention behavior description with the preset intervention level table, and maps the intervention text of the record entry to the standard intensity level value. The extraction process is classified and organized according to the patient dimension. Combined with the intervention behavior corresponding to each abnormal data, the corresponding relationship between the patient and the intervention intensity level is established to obtain the intervention intensity level record; Based on the abnormal isolation cache object, the intervention information text field attached to each record is first identified, and the text content is disassembled to extract intervention behavior keywords, such as "suspend medication", "increase the dose of oral steroids", "increase the dose of injectable antibodies", etc., which are converted into structured intervention items, and an intervention intensity level comparison table is established. The intervention level is set from 1 to 5 in ascending order. Level 1 represents no drug adjustment, only observation, level 3 can represent "suspend the drug for 24 hours" or "reduce the dose by 50%", and level 5 means immediately stopping the drug and performing anti-inflammatory injection treatment, etc. Combined with the record form filled out by the clinical research subjects during the actual treatment process, if an abnormal record text shows "the patient suspended the medication on the 4th day and supplemented intravenous hormones If "treatment" is used, the two intervention elements "suspension of medication" and "intravenous hormones" are extracted and mapped to levels 3 and 5 respectively. According to the preset rules, the main intervention level with the highest weight is selected and assigned a value of 5. This assignment mechanism is implemented through keyword indexing and conditional comparison. In actual operation, the field matching structure is called and the judgment is completed according to the set rules. If there are multiple intervention behaviors in the patient record, they are screened according to the behavior priority order defined in the intervention intensity level table. For example, "only oral hormones" has a priority of 3, and "intravenous antibodies" has a priority of 4. The system finally assigns a value of 4. In addition, when recording the level results, the implementation time of the corresponding intervention needs to be marked for subsequent cycle calculation steps to call, and finally the intervention intensity level record is obtained.

[0029] The cycle sequence acquisition submodule records the intervention intensity level and sets the corresponding hysteresis cycle benchmark value in combination with the drug's half-life t1 / 2. The hysteresis cycle is set to 3 times t1 / 2 as the evaluation interval. The inflammatory factor monitoring value sequence within the collection period is extended backward from the intervention start time point of each patient, forming a data pair structure with the corresponding timestamp. The monitoring data sequence is established and the corresponding monitoring time points are marked to obtain the hysteresis cycle indicator sequence. The intervention start time point was extracted according to the intervention intensity level record, and the half-life t1 / 2 parameter value defined in the drug instructions was called. The system defaulted to using the t1 / 2 value of drug A as 8 hours. Based on this, the hysteresis period benchmark value was calculated to be 3 times of t1 / 2, that is, 24 hours. The intervention time was T0, and the monitoring value of the inflammatory factor IL-6 was collected every 4 hours from T0 until the 24-hour period was covered. During the implementation process, several pre-set times were 12:00 on the second day, and the collection point times were 16:00, 20:00, 0:00 the next day, and 4:00 0, 8:00, and 12:00, a total of 6 time points, the data were recorded in a structure array with the IL-6 value sequence and its corresponding time point, which was organized in a table as shown in Table 3. A structural check was used to determine whether the time points were evenly spaced. If any time point was missing data, it was marked as a missing item and eliminated in subsequent analysis. After data integration, a complete cycle indicator sequence for each patient was constructed, and the sequence length, timestamp span, and whether it covered the complete cycle were checked to eliminate data records that did not meet the cycle conditions, and finally a delayed cycle indicator sequence was obtained.

[0030] Table 5 Inflammation index deviation calculation data table:

[0031] As shown in Table 5, the sampling data sequence meets the requirements of continuous time interval and complete coverage period.

[0032] The characteristic quantity calculation submodule adopts the formula based on the hysteresis cycle indicator sequence, the inflammatory factor level indicators collected at each monitoring time point, and the upper limit of the corresponding inflammatory factor monitoring range: ; The hysteresis offset value is obtained by calculation, revealing the maximum relative offset degree within the cycle, which is bound to the corresponding patient and intervention event label to obtain the hysteresis monitoring feature data list; among them, Indicates the hysteresis offset value, which is used to reflect the peak value of the offset amplitude of the inflammation index within the cycle. Indicates the The inflammatory factor monitoring value at each sampling time point is expressed in pg / mL. Indicates the upper limit threshold of the corresponding inflammatory factor monitoring, the unit is pg / mL, Indicates the The time difference between the sampling point and the intervention start time, in hours, Indicates the inflammatory response delay factor, which is defined as the delay between the patient's last response peak and the intervention moment, in hours. represents the hysteresis time scale conversion factor, the unit is pg²·h⁻¹·mL⁻²; According to the hysteresis cycle indicator sequence, the deviation degree of the inflammatory factor (calculated as IL-6) index value at each sampling point is calculated. Combined with the monitoring upper limit set to 50pg / mL, the formula is substituted. Taking the two data points of patient P001 as an example, the sampling points are 4 hours and 8 hours respectively, and the IL-6 values ​​are 53 and 67pg / mL respectively. The parameters are set as follows: Monitoring upper limit , delay factor , conversion factor (The setting is based on the inverse of the proportional function between the average decline rate of the inflammatory factor IL-6 at different time points and the upper limit of the standard monitoring. That is, when the average decline rate of IL-6 in the past 96 hours is 4.5 pg / mL per hour, the conversion factor is set to 0.4. The conversion factor value is adjusted with the type of inflammatory factor, the subject's drug metabolism rate (through t1 / 2 back-calculation), and the fluctuation range of IL-6. Under the same intervention plan, its variation range remains between 0.3 and 0.6). Substituting it into the first time point, the calculation is: ; The second point is: ; Taking the maximum value of all results, we get ,The result shows that the patient experienced a maximum deviation of 0.319 in this cycle, and finally generated a list of hysteresis monitoring feature data.

[0033] The hysteresis offset value is a quantitative indicator used to measure the maximum deviation of the inflammatory factor level from the upper limit of its normal monitoring range within a specific hysteresis period after intervention. It reflects whether the hysteresis and intensity abnormalities of the inflammatory response in the time dimension are obvious in the subject during the specified observation period after the start of drug intervention. This value is constructed by integrating the IL-6 test value at the sampling time point, the monitoring upper limit, the observation time after the intervention, and the individualized inflammatory response delay factor. The larger the value, the more significant the reaction deviation of the inflammatory factor and the slower the recovery within the period. Therefore, this value not only has the function of reflecting the current intensity of inflammatory activation, but also integrates the reaction hysteresis characteristics. It is a composite biological indicator with both longitudinal time sensitivity and lateral deviation intensity. It can serve as an important quantitative basis for determining delayed drug response or insufficient intervention.

[0034] The formula is designed to quantify the maximum abnormal response intensity within the hysteresis period. First, the numerator Indicates the The offset between the collected value at each time point and the monitoring threshold reflects the absolute difference between the monitoring value above (or below) the upper limit of the standard. The larger the value, the stronger the level of inflammatory activation. The denominator is composed of three factors, among which As a standard reference value, it provides normalization so that the offset value will not be amplified due to the excessive monitoring level itself. The time interval is mapped to a dimension that is consistent with the concentration offset to represent the time hysteresis effect of the response. The multiplication structure is used to achieve linear expansion and adjustment. The sum of the two represents the patient's individual inherent reaction lag, and the square root operation is performed on the sum, that is, , is to express the influence of hysteresis factor on total offset as nonlinear amplification, that is, the longer the hysteresis, the weaker its inhibitory effect. The square root operation can slow down the growth of time and delay factor, thereby avoiding excessive amplification of the calculation result by the extreme value of time. Finally, the absolute value symbol is added to the outer layer of the whole expression. The directional differences of the upper and lower offsets are handled uniformly, and the maximum value function of the outermost layer It means that the one with the largest deviation among all monitoring time points is selected as the representative value of the cycle, which is used as the characteristic indicator of the degree of hysteresis anomaly, and is used to trigger the monitoring process and generate the data list in the future.

[0035] See also Figure 5 , the progressive risk evolution module includes: The trend extraction submodule extracts the offset values ​​and corresponding times corresponding to each sampling time point after the start of the intervention based on the hysteresis monitoring feature data list, constructs an offset sequence array in chronological order, calculates the offset difference change sequence in all time periods, confirms the trend direction, and generates the offset trend pattern classification results; Based on the hysteresis monitoring feature data list, the time series data extracted from each subject record were parsed. First, the offset values ​​of each subject at each sampling time point after the intervention were obtained and recorded in array form. For example, the IL-6 values ​​of subject A001 collected at the 4th, 8th, 12th, 16th, 20th, and 24th hours after the intervention were [0.12, 0.18, 0.26, 0.34, 0.41, 0.49], respectively. Then, the time axis was marked [4, 8, 12, 16, 20, 24] to construct an offset sequence pair. Then, the difference calculation was performed on every two adjacent offset values, such as 0.18-0.12=0.06, 0.26-0.18=0.08, In this way, the offset difference sequence [0.06, 0.08, 0.08, 0.07, 0.08] is constructed, and each difference is judged to be greater than 0. If all are positive, the trend is judged to be increasing. If the difference is negative, such as [-0.07, -0.06, -0.05, -0.04], it is judged to be a decreasing trend. During the judgment process, the judgment window is set to 4 consecutive segments. The trend direction is locked if the difference direction in the window is consistent. If there is inconsistency in the direction in the first 4 segments, it is recorded as an invalid trend. If an object is in an increasing direction in four consecutive segments, its trend mode is finally marked as increasing. If it is continuously decreasing, it is marked as decreasing. If there is no obvious direction, it is not marked. Finally, the trend result of each object in this stage is recorded as the offset trend mode classification result.

[0036] The stable time identification submodule retrieves the offset value of each time point after the start time of the trend stable segment based on the offset trend pattern classification results, and determines whether the difference between the current offset value and the offset value of the previous period is lower than the offset stability threshold. If the offset value difference in three consecutive time periods is within the interval, the corresponding time period is determined to be the first stable time point, and the time difference between this point and the intervention start time is recorded to obtain the first stable time interval data; According to the classification results of the offset trend pattern, the monitoring objects in the increasing or decreasing mode are screened, and then the offset value sequence of the corresponding sampling point is obtained. According to the starting point of the trend stable segment, the difference between the current offset value and the previous offset value is calculated in sequence. For example, the offset value sequence is [0.45, 0.48, 0.49, 0.50], and its difference sequence is [0.03, 0.01, 0.01]. The difference is judged segment by segment to see whether it is within the range of ±0.05. If all the differences are met, the segment is confirmed to be an offset stable segment, and the starting time point of the stable segment is recorded. For example, the offset value 0.45 corresponds to the 26th hour, and the first stable time is recorded as 26, combined with the intervention time of 0 hours, the stable time interval value is 26 hours. The setting of the threshold of ±0.05 is derived from the natural fluctuation experimental results of the inflammatory factor IL-6. In the 72-hour continuous monitoring of 300 patients, the natural fluctuation of IL-6 did not exceed ±0.045. Therefore, 0.05 is selected as the stable threshold boundary here. For example, the stable segment of object A001 is [26, 30, 34, 38], and the corresponding offset difference is [0.03, 0.01, 0.01], all of which are lower than 0.05. 26 hours is determined to be the first stable time, and the difference between it and the intervention start time is finally obtained as the first stable time interval data.

[0037] The risk level adjustment submodule sets the hysteresis period standard time of 24 hours based on the first stable time interval data, calculates the difference between the hysteresis period and the stable time interval value, and forms a time offset index. Based on the offset pattern shown in the offset trend pattern classification result, the risk level adjustment submodule performs progressive rule-based level classification adjustment on the current object using the formula: ; The risk evolution level value is obtained by calculation and a risk evolution analysis data set is generated, where: Indicates the risk evolution level value, Indicates the absolute time value of the first stable time point, in hours; Indicates the starting time of intervention, in hours; Indicates the system-set hysteresis cycle standard time in hours, with a recommended value of 24; Indicates the maximum deviation value difference of the deviation trend segment, in pg / mL; It represents the average value of all sampling offset values ​​within the trend segment, in pg / mL; The difference between the first stable time interval data and the fixed hysteresis period standard time of 24 hours is calculated to obtain the stable time offset amplitude. At the same time, the maximum offset difference ΔP of the trend segment identified in the offset trend pattern classification result and the mean offset μP of the segment are used and substituted into the formula for calculation. For example, for object A001, the first stable time is 26 hours, the intervention time is 0 hours, ΔP is 18.6, μP is 9.3, and Tstd is set to 24. The calculation is performed as follows: ; ; The sum of the two is , using this as the risk evolution level value, and comparing it with the set interval to divide the risk level, and finally recording it as the risk evolution analysis data set. Table 6 shows the calculation results of the example object.

[0038] Table 6 Risk evolution index calculation table:

[0039] The risk evolution level value is a dimensionless numerical indicator used to reflect the comprehensive risk performance of the target object in the two dimensions of recovery speed and offset stability after the intervention process. This value forms a quantitative expression of the risk progression trend by integrating the time delay ratio experienced by the object from the start of the intervention to the first offset stabilization, as well as the severity of the offset value during this period. The larger the value, the longer the time required for the object to recover or the more severe the offset fluctuation, reflecting a higher degree of instability during the monitoring period. This value can not only be used to track the trend of individual risk changes, but also for horizontal comparison of risk status among multiple objects. It is a key core indicator used for level classification and trend warning in the entire risk assessment system.

[0040] The formula is based on the superposition of two risk sources: time delay and offset amplitude. First, the first It represents the normalized ratio of the time taken for stable recovery to the standard hysteresis period. Its logic is to quantify the degree to which the individual deviates from the expected recovery process. If the value is greater than 1, it means that the recovery process is slower than the standard period, otherwise it is faster than the standard period. This item constitutes a quantitative expression of the risk of recovery delay. It represents the ratio between the maximum change amplitude of the offset value and its average level in the same offset trend segment. This item is used to measure the relative intensity of drastic changes in the offset process. That is, if the maximum offset value is much larger than the mean, it means that the offset is unstable or rebounds violently. This item constitutes a quantitative expression of the offset volatility risk. The two items are connected by addition to represent the common impact of the two types of risk sources on the overall risk level. Because they are essentially dimensionless ratios, they are structurally reasonable and can be directly added without multiplication or square root processing, making the overall structure concise and clear and having the ability to distinguish.

[0041] See also Figure 6 , the alert status determination module includes: The alert value extraction submodule obtains each record in the risk evolution analysis dataset, extracts the drug type information and adverse drug event level standards corresponding to the record object, constructs classification rules based on the AE level classification standards of the drug, and combines the logical relationship between pharmacological categories and clinical AE trigger records to generate drug alert threshold level standards; Obtain each record in the risk evolution analysis data set, identify and extract the drug name of the corresponding object in each record, read the drug field and parse its classification name, then call the EudraVigilance adverse reaction rating library to perform drug category mapping and judgment operations, and assign the identified drugs to the corresponding drug group. For example, if the drug corresponding to object P001 is "T cell injection", the system maps it to Cytokine drugs, and sets the threshold level to 3. If the drug used by object P002 is "gefitinib", which is a small molecule EGFR inhibitor, it is matched and classified as an oral small molecule. Category, the corresponding threshold level value is 2. If the drug field of object P003 is marked as "injectable anti-IL-6 antibody", it is mapped to the intravenous antibody category, and the threshold level value is set to 3. The threshold level values ​​are derived from the AE (adverse event) statistical classification table set by the EudraVigilance system. After statistically analyzing the AE number distribution curve, the 95% trigger alert level is selected as the category threshold. In actual operation, if the number of objects is multiple batches and involves different types of drugs, the object drug fields are batch extracted and a drug-threshold level mapping table is constructed, as shown in Table 6, to finally obtain the drug vigilance threshold level standard.

[0042] The risk comparison submodule compares the risk evolution level value extracted from the drug vigilance threshold level standard and the risk evolution analysis data set with the corresponding alert threshold level value. If the risk evolution level value is greater than or equal to the alert threshold level value, the record is determined to be an object in need of alert status; otherwise, it is marked as a non-alert status and an alert status identification label is obtained; A comparison operation is performed based on the pharmacovigilance threshold level standard and the risk evolution level values ​​extracted from the risk evolution analysis dataset. Each object number is read and its risk level value and threshold level value are extracted in sequence. The risk level value of P001 is 3.4, corresponding to a threshold level of 3. When performing the comparison judgment, 3.4 and 3 are directly compared for magnitude. Since 3.4 is greater than 3, it is determined to be an out-of-limit record and assigned a value of "yes". The risk level value of object P002 is 1.8, and the threshold level value is 2, which is less than the threshold, so it is assigned a value of "no". Then, the risk value of P003 is 2.6, which is less than the threshold of 3, and is assigned a value of "no". During the judgment process, it is important to note that all values ​​must be expressed in unified units and normalized. The use of unnormalized risk indicators is prohibited. The judgment basis is whether the R value is greater than or equal to the L value (threshold). If it is equal, it should be marked as "yes" and an alert status identification label should be generated.

[0043] The status record submodule retrieves the intervention time, risk level, cell drug type, deviation trend classification, first stable time, and risk deviation value of the record object number marked as requiring alert in the alert status identification label, organizes and constructs the status list table structure, and numbers each record as an independent alert event unit to establish the clinical cell drug alert status record; Based on the alert status identification label, object records marked "yes" are screened to identify individuals with risk level values ​​exceeding the drug threshold level. These objects are numbered and their corresponding intervention time field, drug category field, deviation trend type field, and first stable time field are further obtained. These contents are organized into a structure record in a unified field order. The field combination is constructed as follows: object number, drug category, intervention time, risk level value, threshold level value, and whether it exceeds the limit. These fields are integrated into a structure list and output to form a status record. For example, for object P001, the drug category is T cell injection, the intervention time is 0 hours, the risk level value is 3.4, the threshold level value is 3, and the judgment result is "yes". After the records are summarized, they are output as a status record. Other objects such as P002 and P003, although included in the analysis but did not meet the limit-exceeding standard, are not included in the alert record list. Finally, they are summarized into the clinical cell drug alert status record.

[0044] Table 7 Warning limit comparison and judgment table:

[0045] As shown in Table 7, only the P001 object triggering the warning threshold is judged as an over-limit record, and its corresponding status will be included in the formal record list.

[0046] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A cellular pharmacovigilance system for clinical research, characterized in that: It includes indicator anomaly marking module, anomaly isolation and caching module, dynamic hysteresis monitoring module and progressive risk evolution module; The abnormal indicator marking module obtains the clinical monitoring data of the subject, extracts the monitoring data of the inflammatory factor level indicator, compares the monitoring value with the inflammatory factor monitoring range, and adds an abnormal mark if it exceeds the range, and generates an abnormal mark data list; The abnormal isolation cache module buffers abnormal data entries together with timestamps into the abnormal isolation cache area based on the abnormal marking data list, writes normal data into the normal data area and maintains continuous monitoring to generate abnormal isolation cache objects; The dynamic hysteresis monitoring module extracts the intervention measure intensity level corresponding to the abnormal data based on the abnormal isolation cache object, obtains the inflammatory factor level indicator sequence continuously collected within the hysteresis period in combination with the hysteresis period baseline value, calculates the offset distance of each sampling point value in the sequence relative to the inflammatory factor monitoring range, and generates a hysteresis monitoring feature data list; Based on the hysteresis monitoring feature data list, the progressive risk evolution module extracts the maximum offset distance and the corresponding sampling time point, obtains the offset change sequence to determine the offset direction, calculates the stable time from the start of intervention to the first recovery, calculates the time difference with the standard hysteresis period, combines the offset pattern to form a risk evolution level value, and outputs a risk evolution analysis data set.

2. The cellular pharmacovigilance system for clinical research according to claim 1, characterized in that: The abnormal marking data list includes the abnormal data entry index, abnormal indicator name, and abnormal monitoring value; the abnormal isolation cache object includes the abnormal data timestamp, abnormal data content, and abnormal cache status; the hysteresis monitoring feature data list includes the monitoring sequence sampling time point, the offset distance of each sampling point, the maximum offset feature value, and the hysteresis cycle span; the risk evolution analysis data set includes the offset pattern category, offset change trend, intervention start time, first stable recovery time, time difference, and risk evolution level value.

3. The cellular pharmacovigilance system for clinical research according to claim 1, characterized in that: The indicator abnormality marking module includes: The monitoring data extraction submodule obtains the real-time clinical monitoring data of the test subjects, detects the monitoring items related to the inflammatory factor levels, collects the monitoring data of the inflammatory factors, extracts the corresponding monitoring time, subject number and test value, and generates the inflammatory factor monitoring data set; The abnormal interval comparison submodule is based on the inflammatory factor monitoring data set and the corresponding inflammatory factor monitoring range. For each monitoring data, the detection value is compared with the corresponding inflammatory factor monitoring range interval, and it is determined whether the detection value exceeds the interval range. All data entries outside the interval are filtered out to obtain an interval deviation abnormality list; The abnormal mark generation submodule adds abnormal identification information to the monitoring data items that exceed the inflammatory factor monitoring range according to the interval deviation abnormal list, marks the abnormal category, abnormal degree and corresponding abnormal factor type, and generates an abnormal mark data list.

4. The cellular pharmacovigilance system for clinical research according to claim 1, characterized in that: The abnormal isolation cache module includes: The data tag screening submodule screens all data entries with abnormal tags based on the abnormal tag data list, calls the corresponding timestamp of each data entry, obtains the combination of the data entry and the timestamp, performs synchronous caching operations on all the combined contents, and obtains an abnormal combination cache set; The dual-area writing submodule determines the abnormal flag status of each data entry based on the abnormal combination cache set, extracts the data entries marked as normal, adjusts the sequence according to the timestamp, writes the adjusted data into the normal data area, performs continuous monitoring, and obtains the timing adjustment synchronization record; The isolation object generation submodule classifies the abnormal data entries based on the abnormal combination cache set and the timing adjustment synchronization record, obtains the continuous burst segments and concentrated offset types according to the timestamp and tag distribution, performs structure organization and label classification, and generates abnormal isolation cache objects.

5. The cellular pharmacovigilance system for clinical research according to claim 1, characterized in that: The dynamic hysteresis monitoring module includes: The intervention level extraction submodule extracts the intervention measure content corresponding to each abnormal inflammatory factor monitoring data recorded in the abnormal isolation cache object based on the above-mentioned object, matches the intervention behavior description with the preset intervention level table, maps the intervention text of the record entry to the standard intensity level value, classifies and organizes the extraction process according to the patient dimension, and establishes the corresponding relationship between the patient and the intervention intensity level based on the intervention behavior corresponding to each abnormal data, thereby obtaining the intervention intensity level record; The cycle sequence acquisition submodule sets the corresponding hysteresis cycle benchmark value based on the intervention intensity level record and the half-life t1 / 2 of the drug. It extends the inflammatory factor monitoring value sequence within the acquisition period backward based on the intervention start time point of each patient, forms a data pair structure with the corresponding timestamp, establishes a monitoring data sequence and marks the corresponding monitoring time point, and obtains the hysteresis cycle indicator sequence; The feature quantity calculation submodule obtains the hysteresis offset value based on the hysteresis cycle indicator sequence, the inflammatory factor level indicator collected at each monitoring time point, and the upper limit value of the corresponding inflammatory factor monitoring range, thereby revealing the maximum relative offset degree within the cycle, binding it to the corresponding patient and intervention event label, and obtaining a list of hysteresis monitoring feature data.

6. The cellular pharmacovigilance system for clinical research according to claim 5, characterized in that: The hysteresis period is set to 3 times of t1 / 2 as the evaluation interval.

7. The cellular pharmacovigilance system for clinical research according to claim 1, characterized in that: The progressive risk evolution module includes: The trend extraction submodule extracts the offset values ​​and corresponding times corresponding to each sampling time point after the start of the intervention based on the hysteresis monitoring feature data list, constructs an offset sequence array in chronological order, calculates the offset difference change sequence within all time periods, confirms the trend direction, and generates an offset trend pattern classification result; The stabilization time identification submodule retrieves the offset value of each time point after the start time of the trend stabilization section based on the offset trend pattern classification result, and determines whether the difference between the current offset value and the offset value of the previous time period is lower than the offset stabilization threshold. If the offset value difference is within the interval for three consecutive time periods, the corresponding time period is determined to be the first stable time point, and the time difference between the time point and the intervention start time is recorded to obtain the first stable time interval data; The risk level adjustment submodule sets the hysteresis period standard time according to the first stable time interval data, calculates the difference with the stable time interval value, forms a time offset index, and performs a level classification adjustment based on the progressive rule on the current object according to the offset pattern shown in the offset trend pattern classification result, calculates and obtains the risk evolution level value, and generates a risk evolution analysis data set.

8. The cellular pharmacovigilance system for clinical research according to claim 1, characterized in that: The system further includes an alert status determination module, which extracts a risk evolution level value based on the risk evolution analysis data set and compares it with the drug alert action limit. If the risk evolution level value is greater than or equal to the alert threshold level, the corresponding risk state is marked as an alert state, and a clinical cell drug alert status record is generated; The clinical cell drug alert status record includes a risk level identifier, an alert threshold level, an alert status label, an alert determination time point, and an associated risk analysis number.

9. The cellular pharmacovigilance system for clinical research according to claim 8, characterized in that: The alert status determination module includes: The alert value extraction submodule obtains each record in the risk evolution analysis data set, extracts the drug type information and adverse drug event level standards corresponding to the record object, constructs classification rules, and generates drug alert threshold level standards; The risk comparison submodule compares the risk evolution level value extracted from the drug vigilance threshold level standard and the risk evolution level value in the risk evolution analysis data set with the corresponding alert threshold level value. If the risk evolution level value is greater than or equal to the alert threshold level value, the record is determined to be an object requiring alert status; otherwise, it is marked as a non-alert status, and an alert status identification label is obtained; The status record submodule retrieves the intervention time, risk level value, cell drug type, deviation trend classification, first stable time and risk deviation value content in turn according to the record object number identified as requiring alert in the alert status identification label, organizes and constructs the status list table structure, and numbers each record as an independent alert event unit to establish a clinical cell drug alert status record.

10. The cellular pharmacovigilance system for clinical research according to claim 9, characterized in that: The classification rules are constructed based on the AE grade classification standards of the drugs to which they belong, combined with the logical relationship between pharmacological categories and clinical AE trigger records.

Images