NMPA published the “Medical Device Clinical Trial Data Submission Guideline” and “In Vitro Diagnostic Reagent Clinical Trial Data Submission Guideline” on November 25, finalizing the draft issued in mid-September. They are to direct manufacturers on how to collect, organize, analyze and submit clinical trial data, and improve the quality of clinical trial implementation and management.
Significantly referring to the Clinical Data Interchange Standards Consortium (CDISC) standards, the guidelines detail requirements on the original database, analysis database, explanatory documents and program codes.
Basic principles
(1) truth
The submitted clinical trial data are consistent with the original clinical trial records.
(2) traceability
According to the data, descriptive documents and program codes submitted by the registration applicant, the analysis database and the statistical analysis results in the clinical trial report can be reproduced from the original database, and the formed analysis database and statistical analysis results are consistent with the content submitted by the registration applicant Unanimous.
(3) readability
The database has a clear structure and detailed comments for easy review. Submitting clinical trial data in accordance with the relevant specifications of this guideline will help improve readability.
Highlights of the Device Clinical Trial Data Submission Guideline
(1) Original database
Each data set needs to include the subject’s unique identification variable to achieve the correlation of observations from different data sets of the same subject. If it involves data observed at different visit time points, the visit time variable should be used for identification. For example, for 3 months and 6 months after the operation of cardiac ultrasound-related observations, the visit time identification variables can be named Visit_3, Visit_6, etc. to distinguish them. If two or more clinical trials are involved, the data set must include clinical research identification variables. If randomization is used in clinical trials, the original database should contain variables such as random numbers.
(2) Analyze the database
The analysis data set can be named based on the corresponding statistical results it produces. For example, the data set that generates the comparison results of adverse events can be named “ADAE” (Adverse Event Analysis Data Set). It is recommended to add the prefix “AD” (analysis data) to the name of the analysis data set to identify the data set as an analysis data set.
(3) Program code
The codes that need to be submitted mainly include: the codes used for the original database to generate the analysis database, the codes for the analysis database to generate the statistical analysis results, etc. The relevant codes used to adjust the format or generate the form may not be submitted. The submitted code should conform to the usual programming format and programming specification, with a clear structure and easy to read. The program code should include sufficient comments to describe the purpose of different program codes and other content that needs to be explained to help reviewers better understand the code logic. If the submitted program code refers to a macro program, the corresponding macro program code must be provided, and the software version and system environment for running the program must be stated.
(4) Explanatory documents
1. Data description file
The description file of the analysis database needs to specify the generation rules of the derived variables, and the variables and calculation methods involved. For example, for the filling of missing values, the filling method should be clarified and the corresponding program code should be provided. It is recommended to list the program code files used to generate each analysis data set and the name of the original data set in the form of a table.
2. Program code use instruction file
The registration applicant should explain the encoding used in the original data set and analysis data set (such as UTF-8, EUC-CN, etc.) to avoid garbled codes in the submitted data set.
3. Annotate the case report form
There may be some redundant data collected in the CRF that is not related to the analysis of clinical trial results. These data may not be included in the submitted database, but it should be clearly marked as “not submitted” on the comment CRF, and the reason should be stated.