Issued by the State Food and Drug Administration: statistical guidelines for bioequivalence research and technical guidelines for bioequivalence research of high variation drugs

In order to ensure the smooth development of the consistency evaluation of generic drugs, the State Drug Administration has formulated the statistical guidelines for bioequivalence research and the technical guidelines for bioequivalence research of high variation drugs, which are now published Notice is hereby given Annex: 1 Statistical guidelines for bioequivalence research 2 Technical guidelines for bioequivalence research of high variation drugs, SFDA, October 17, 2018 Annex 1: statistical guidelines for bioequivalence research 1 Overview of bioequivalence, Be) study is a study to compare whether the difference of absorption speed and absorption degree between the test preparation (T) and the reference preparation (R) is within the acceptable range It can be used for the listing application of the chemical drug generic, or for the application of the change of the listed drugs (such as the new specification, new dosage form, new administration way) At present, the method of average bioequivalence (ABE) is usually recommended in bioequivalence research The average bioequivalence method only compared the average level of pharmacokinetic parameters, without considering individual variation and variation caused by interaction between individual and preparation In some cases, other analytical methods may need to be considered For example, the population bioequivalence (PBE) method can be used in the in vitro be study of aerosols to evaluate the average level of pharmacokinetic parameters among preparations and whether the variation within individuals is equivalent This guideline aims to provide technical guidance for the research design, data analysis and result report of bioequivalence research with pharmacokinetic parameters as the end point evaluation index It is a general principle for statistical analysis of bioequivalence research data When carrying out bioequivalence research, in addition to the contents of this guiding principle, it is also necessary to comprehensively refer to the relevant guiding principles, such as the guiding principles of bioequivalence research technology for chemical drug generic human body with pharmacokinetic parameters as the end point evaluation index and the guiding principles of Biostatistics for drug clinical trials 2、 Research design (1) the overall design considering bioequivalence research can adopt cross design or parallel group design 1 The method of cross design is generally recommended for bioequivalence study of cross design The advantages of cross design include: it can effectively reduce the bias caused by the variation between individuals; in the case of equal sample size, cross design has higher test efficiency than parallel group design Cross design of two preparations, two cycles and two sequences is a common cross design, as shown in Table 1 Table 1 cross design of two preparations, two cycles and two sequences can be used if it is necessary to accurately estimate the intra individual variation of a preparation Repeat crossover design includes partial repeat (such as two preparations, three cycles, three sequences) or complete repeat (such as two preparations, four cycles, two sequences), as shown in Table 2 and table 3 Table 2 two preparation, three cycle, three sequence repeat crossover design table 3 two preparation, four cycle, two sequence repeat crossover design 2 Parallel group design can also be used in some specific cases (such as drugs with long half-life) Parallel group design has greater impact on the experiment due to individual variation than cross design, so there should be more strict conditions for the selection of subjects, such as age, gender, weight, disease history, etc., and reasonable randomization scheme should be used to ensure the baseline level balance between groups to get better group comparability 3 For other design methods such as adaptive design, please refer to the biostatistics guidelines for clinical trials of drugs, and communicate with regulatory authorities in advance （2） Before the sample size test, it is necessary to fully estimate the sample size required to ensure sufficient test efficiency, and the method and results of sample size estimation shall be detailed in the test plan In bioequivalence analysis using Abe method, the sample size should be reasonably estimated based on a clear formula Different designs have different sample size estimation formulas The factors to be considered in the sample size of cross design include: (1) test level α, usually 0.1 on both sides (0.05 on both sides); (2) test efficiency 1 - β, usually at least 80%; (3) within subject coefficient of variation (CVW%), which can be estimated based on literature reports or pre test results; (4) geometric mean ratio, GMR); (5) equivalence bound The sample size estimation of parallel group design can refer to the sample size calculation formula of general continuous variables If the analytical method used does not have a clear formula for calculating the sample size, the computer simulation method can also be used to estimate the sample size （3） In order to avoid the sample size shortage caused by the drop off of subjects in the research process, the applicant should consider increasing the sample size properly when estimating the sample size In general, subjects should not be added after the start of the trial Subjects assigned a random number are usually not replaceable （4） The cross-over design was used for be study to increase the accuracy of comparison through each subject's own control The basic assumption was that the comparative preparations did not have residual effect in the next cycle test, or the residual effect was similar If there are unequal residual effects in cross design, the estimation of GMR may be biased Residual effects should be avoided in the study design If residual effect is found, the applicant shall analyze the possible causes, provide corresponding judgment basis, and evaluate its impact on the final conclusion 3、 Data processing and analysis (1) data sets of data sets need to be clearly defined in the scheme in advance, including specific subject rejection criteria Generally, the data set of be study should at least include pharmacokinetic parameter set (PKPs) and bioequivalence set (BES) The number of subjects used for the analysis of different pharmacokinetic parameters may vary Pharmacokinetic parameter set (PKPs): includes the data set of pharmacokinetic parameters obtained from subjects who have received at least one study drug The purpose of this data set is to make descriptive statistics of pharmacokinetic parameters of subjects Bioequivalence set (BES): a statistical analysis set that usually includes at least one cycle and has at least one evaluable pharmacokinetic parameter This data set is the main data set for inferring whether the test preparation and the reference preparation are bioequivalent （2） Data conversion it is recommended to use natural logarithm for data conversion of pharmacokinetic parameters (such as AUC and Cmax) The logarithm conversion mode selected shall be consistent during the test and shall be specified in the scheme In bioequivalence research, it is difficult to determine the distribution of data due to the small sample size Therefore, it is not recommended to use the original data for statistical analysis on the basis that the data after logarithmic transformation does not obey the normal distribution or the original data obey the normal distribution （3） Statistical hypothesis and inferential average bioequivalence require that the difference between the test preparation and the reference preparation is within a certain acceptable range, and statistical inference is conducted through the following hypothesis test Original hypothesis H0: or alternative hypothesis H1: where μ t is the total mean of the pharmacokinetic parameters after the logarithmic transformation of the test preparation, μ R is the total mean of the pharmacokinetic parameters after the logarithmic transformation of the reference preparation, and θ is the bioequivalence boundary value Under the set test level, if the original hypothesis H0 is rejected, it indicates bioequivalence Generally, θ = ln (1.25), - θ = ln (0.8), that is to say, bioequivalence requires GMR of test preparation and reference preparation to fall within the range of 80.00% - 125.00% The bioequivalence criteria shall be applied to all major pharmacokinetic parameters, including Cmax, auc0-t and auc0 - ∞ In general, if the study drug contains more than one component, each component should meet the bioequivalence standard When Tmax is closely related to the clinical efficacy of drugs, the paired nonparametric method is usually used to test the difference of Tmax （4） Data analysis 1 Overview for the bioequivalence criteria mentioned above, it is usually the constructed bilateral 90% confidence interval If this confidence interval falls within the interval, it can be inferred that the test preparation and reference preparation meet the bioequivalence This method is equivalent to the double one-sided hypothesis test at the test level of 0.05 The proper calculation method of confidence interval should be selected according to different test designs After the calculated double-sided 90% confidence interval, the double-sided 90% confidence interval of GMR of the original data of the test preparation and the reference preparation can be obtained through the inverse logarithmic transformation (exponential transformation) 2 For cross design, it is recommended to use linear mixed effect model for analysis and calculation 3 In parallel group design, it is suggested to use the confidence interval construction method based on the difference of mean value of normal distribution （5） Outlier removal is generally not recommended in outlier data processing If necessary, sensitivity analysis should be carried out for outliers, that is, to evaluate the impact of outliers on bioequivalence results If the conclusions are inconsistent, explain and analyze the reasons （6） For other problems, if a cross design is conducted at two or more centers, the center effect should be considered in the statistical model The model used should be able to estimate the effect of different centers, reflect the actual situation of different centers, and explain whether the test data from different centers can be combined for analysis If there are multiple test preparations and / or multiple reference preparations, there are usually multiple bioequivalence hypothesis tests If multiple hypothesis tests need to be satisfied at the same time, there is no need to adjust the type I error; if not, there is a need to adjust the type I error by Bonferroni method, Hochberg method, etc 4、 The following contents shall be detailed in the result report （1） Randomization shall specify the randomization system and scheme used for the test, including the factors of randomization control, block group, seed number, etc., and attach the randomization digital table The results of randomization should be described in a table, including the number of subjects, the medication situation in each cycle, and the factors of randomization control The randomization results can be shown in the appendix （2） The statistical methods should include the calculation method of pharmacokinetic parameters, analytical model, equivalence test method, logarithm transformation, etc Also indicate the name and version number of the software used （3） The statistical analysis results shall provide the test results of the concentration of the tested components of each subject after administration At the same time, the arithmetic coordinates and the logarithmic coordinates of the drug time curve of each subject after administration and the average drug time curve of different drug preparations shall be given in the appendix Results of pharmacokinetic parameters of each subject shall be provided, including arithmetic mean, geometric mean, standard deviation and coefficient of variation of the test preparation and reference preparation Results of a mixed effect model including sequence nested subjects, sequence, cycle, and formulation factors shall be provided If there are other factors to consider, they should also be included in the model The geometric mean ratio of pharmacokinetic parameters and its confidence interval estimation results shall be provided 5、 Data management the data management of bioequivalence research for the purpose of registration and listing can refer to the relevant technical requirements of clinical trial data management In bioequivalence study, data such as biological sample analysis are external data, which should be kept in a blind state during sample analysis and relevant data transmission, and should be prepared in advance