Pharmacokinetic data management and statistical calculation methods that meet regulatory requirements

In some clinical trials, the results of bioequivalence (BE) and pharmacokinetics (PK) are the main endpoints of the study. Therefore, the data management, statistical calculation process and the compliance, consistency and completeness of the results It will be the focus of review and on-site verification. Recently, the State Food and Drug Administration (CFDA) successively promulgated the “Guiding Principles of Biostatistics for Clinical Trials of Drugs”, “Technical Guidelines for the Management of Clinical Trial Data”, and “Requirements for the Application of New Registration Classification of Chemical Drugs (Trial)”. In the guiding principles, there are content about the management of PK data and statistical calculations. In the current regulatory environment, it is necessary to consider how to conduct PK data management and statistical calculations to improve the quality of clinical research. In addition, these data also involve electronic records, electronic signatures, etc. In the case that my country’s corresponding laws and regulations have not yet strictly regulated, how to conduct standardized electronic data management requires serious consideration by researchers. In accordance with the requirements of domestic laws and regulations, the author has carried out a relatively complete data management and statistical calculation of PK, and has been successfully applied to the application and research of new drugs. The purpose of this article is to preliminarily introduce how to manage clinical PK data and statistical calculations so that the calculation process and results meet the requirements of regulation, consistency and completeness.

1 Research data and software

The research data used for verification in this article includes BE research data and PK research data, all of which come from specific clinical trials. The statistical software used in this article is PharsightWinNonlinV6.4 (California, USA) and MicrosoftEXCEL20lo (Microsoft, USA).

2 method

2.1 Process PK data management

Preparation before analysis. Develop data management and statistical plans related to PK in accordance with the test plan, and sign it into effect; conduct personnel authorization and training; whether the calculation software has been verified.

Determine the data to be analyzed and the data related to the PK and BE statistical calculations can be provided by different departments, including: ①test data, including the lowest limit of quantification; ②the subject’s drug random table; ③clinical trial data, including the subject’s Drug status, sample code, compliance status, time difference between actual sampling point and planned sampling point, whether there are subjects to be eliminated, etc.

Data format and transmission In order to facilitate statistical calculations, data statistical units and generating units need to specify the data format in advance. The naming of data fields can refer to the recommended requirements of CDISC (Clinical Data interchange standard consortium) to be unified. There are many ways to transfer data. The most commonly used files are EXCEL files, and the CSV format is a plain text format, which can be directly read by various statistical software. However, the EXCEL file lacks trajectory and can cause unconscious entry errors. It has been suggested to use less at present. For data transmission, e-mail is an economical and feasible method, and at the same time it can save a backup for easy checking. The important data should be encrypted and transmitted, and the password and data should be transmitted separately.

Data acceptance, review and questioning After the statistics department accepts the data, it needs to be reviewed as soon as possible, including the quantity, format, and effective figures of the data, which should be consistent with the requirements of the plan. If necessary, logical inspections can also be carried out. If you have any questions, fill in the data question form and submit it to the data provider for review and correction. If you receive data that needs to be manually entered, the statistical unit needs to verify it after double entry to ensure that it is correct. In order to reduce errors, manual data entry should be reduced in statistics.

Pre-analysis and locking of data For the received data, even if a very complete review is carried out, there may be some problems. Pre-analysis can reduce the occurrence of problems in this area. Pre-analysis should be detailed in the statistical plan. After there is no problem in the pre-analysis, all the data received will be formally analyzed after being locked in accordance with the locking procedure. If you need to unlock it for some reason, you need to follow the requirements of the SOP and re-lock it. The reason for unlocking generally needs to be explained in the summary report.

Formal PK calculation and statistical analysis The locked data is subjected to complete PK calculation and statistical analysis. The procedures used need to be reviewed in advance to ensure that they are correct. Since most of the experiments have the same calculation principles, the previously compiled and reviewed procedures can be reused, but it should be noted that there may be differences between different experiments and cannot be simply applied.

Locking, reporting and archiving of results The calculated results can be locked after being reviewed by the quality assurance department, and the relevant data can be exported for writing summary reports and archiving.

3 PK statistical method

3.1 Calculation of non-compartmental model

For the obtained individual blood drug concentration-time curve data, the PK calculation method recommended by the regulations is the non-compartmental model method. The calculation principle is the linear regression after logarithmic transformation of 3~5 points of the elimination phase in the best fitting manner. The negative value of the regression coefficient is the elimination rate constant (λz), and the elimination half-life can be calculated by this parameter. (t½), the calculation formula is t½=In2/λz:. The calculation methods of other PK parameters are: AUC0-t is calculated by the trapezoidal method; Cmax and tmax measured values; CL/F=Dose/AUC, Vd/F=CL/F/λz; if it is an intravenous preparation, F=1; AUCo -inf=AUCo-t+Ct/λz.

3.2 Judgment of linear PK

The logarithmic transformation of PK parameters conforms to the normal distribution, and the linear mixed effects model can be used to calculate the regression coefficient and its 90% confidence interval between the logarithmic dose and the transformed PK parameter. If the confidence interval of the regression coefficient is 80% Within ~125%, it can be judged that the linear relationship is established (also known as the dose proportional relationship).

3.3 Analysis of gender differences

According to the assumption of the PK parameter, the logarithm conversion of the main PK parameter conforms to the normal distribution, which can be carried out with the t test and can provide an interval estimate of the overall mean.

3.4 Analysis of variance and calculation of 1-2α confidence interval

Here mainly refers to the analysis of variance of the 2×2 crossover experiment design and the calculation of the 1-2α confidence interval. The calculation principle can refer to the literature.

4 Implementation of PK data management and statistical calculation process

Using the above methods, the actual data transfer, review, lock-in, and formal statistical calculations were carried out based on the actual PK and BE research. The specific data management situation will not be described in particular, you can refer to the literature, here we mainly explain the whole process of using WinNonlin software to realize PK calculation and BE statistics.

4.1 PK calculation

In order to reduce errors in the calculation process, the best option is to no longer perform human operations during the calculation process, so that the entire calculation process is automatically carried out in accordance with the program designed in advance. This approach is also conducive to the audit by the quality department. If it is related to data correction, replacement, deletion and other operations, and the processing of the data provider is involved, it needs to be completed by the data provider; when it is completed by the statistical calculation unit, the best way is to complete the processing directly by the program through programming. This processing method can ensure the reliability and consistency of the results, and facilitate subsequent repeated calculations. Based on the above ideas, the following points should be paid attention to in actual operation: ①The data sent by the data generating unit can be directly read through the PK software. If it involves data integration from different sources, it can also be done through the data integration function of the software. If it involves the inconsistency between the planned sampling time and the actual sampling time, you can use the data to automatically replace the operation. ②The lowest limit of quantification (LLOQ) data can be entered into the BQL Rules, so that the data can be processed in accordance with the BQL regulations. ③A new table can be generated for the original data of the guide, and it is easy to check with the original data after output. ④The generation of various charts should be drawn in accordance with the requirements of the CTD format in the BE study, and the format of these charts should be adjusted so that they can be directly added to the final summary report after being exported. In addition, you can visually check whether there is abnormal data through the drawn graphs and tables. ⑤ The calculation of the non-compartmental model and the judgment of linear PK can be performed in accordance with the method of PK statistical calculation.

4.2 BE Statistics

In the BE calculation, the basic requirements of the involved PK calculation part are the same as the aforementioned PK calculation. The difference between PK and BE statistical calculations is: in the BE calculation, the wireless PK judgment requires that there is no need to analyze the gender difference, but the analysis of variance and the calculation of the 1-2α confidence interval are added. In the chart drawn, it is necessary to add a comparison chart of the main PK parameters drawn according to the cycle and the order of administration, which can be displayed with a box chart (BOX chart).

5 Results

From the actual completed data management and PK statistical calculations, using the above data management process and PK calculation method, it has clear data trajectories and nodes, strong traceability, consistent data, high efficiency, convenient recalculation, and no errors. And other characteristics, can meet the requirements of current laws and regulations. According to this method, the unit has carried out data management and PK statistical calculations for multiple projects, and has successfully passed the on-site inspection by the drug regulatory department.

6 Discussion

Maintaining the traceability, trajectory, and consistency of data is an important part of improving the quality of research. But in clinical research, it will inevitably involve the correction of data, the elimination of subjects, and the modification of the research report after review, repeated demonstrations and other operational requirements. This article introduces a simple data management and PK statistical calculation method, which can be conveniently used to deal with many situations in actual research. It should be noted that some specific operations involved in this article need to be specified in standard operating procedures and strictly implemented in actual operations. This aspect will not be further introduced in this article.

With the PK calculation flow chart, the generated data can be processed in a variety of ways, and finally the data and charts that meet the requirements can be generated. The current software can connect these different tasks in series and complete various tasks in one running command, avoiding human manipulation of the program and reducing the occurrence of errors.

When performing statistical calculations, legal requirements require verification of the software. At present, many domestic users have insufficient knowledge in this aspect. Any software may have some shortcomings, and unpredictable problems may occur during use, which may bring risks. It is necessary to confirm the installation, operation and performance of the software through some means to ensure that the software used can meet the needs of statistical calculation purposes. In addition, the software used should have an audit trail to facilitate the audit and find the cause when a problem occurs.

When using WinNoIllin software for statistical calculation, it is currently impossible to analyze gender differences, and it is also difficult to calculate the cumulative rate of urinary excretion. This part of the statistics needs to be completed with other software. When encountering the above calculation situation, it is impossible to complete all the PK statistical calculation tasks with one running command. One solution is to use the non-compartmental model to calculate all the PK parameters, export the data, and use other professional statistical software, such as SAS, to complete the rest of the statistical analysis, which includes the judgment of linear PK and gender differences. Analysis, analysis of variance, calculation of 1-2α confidence interval, etc. The above calculation procedure can be found in some literatures. In addition, some charts can be drawn with SAS software. For some special forms with fixed formats, you need to fill in the calculated summary data, print out the statistical results, and then manually fill in the form. In addition to time-consuming and error-prone, this method requires additional personnel for verification. When it comes to repeated calculations of data, and when there are multiple report versions, the probability of error is greater. The required summary chart is directly generated through the SAS program. If there is no problem with the calculation program, the charts generated by repeated calculations can meet the traceability requirements. These softwares also save the trajectory during statistical calculation, which is convenient for later review. In addition, SAS programming can also connect multiple tasks in series, run once to complete all statistical calculation tasks, and even complete PK calculations for non-compartmental models.