Preparing CDISC Submissions for China’s NMPA [Updated]
October 30, 2020
(Originally published on October 6, 2020. Last Updated on October 30, 2020)
Effective October 1st, 2020, China’s NMPA will accept CDISC submissions. To support this initiative, Pinnacle 21 has released a new Chinese-language validation engine, available now in both Enterprise and Community. This engine supports datasets with Chinese-encoded characters and displays rule messages and descriptions in Chinese translation.
NMPA & CDISC Publications
First and foremost, talk to the NMPA reviewers for your submission to learn their requirements for your study data. In this stage of their CDISC adoption, the content of your study may matter more than formal standard compliance. Some possible talking points:
Study Parameters: Ensure you meet the NMPA’s expectations about your trial design, population race, special reporting for Human Genetic Resources (HGR) collected from patients in China, etc.
IG Versions: Propose to the NMPA the IG versions and validation ruleset you plan to use and invite their feedback. Over 80% of the companies that P21 supports use SDTM-IG 3.2 and ADaM-IG 1.1, which are likely good starting points.
Dataset Translation: The minimum requirements may or may not apply to your submission. Check with your NMPA reviewers about the scope of translation needed.
The NMPA hasn’t yet published their standards catalog. It’s still unknown:
If/when CDISC conformance will be mandatory, or how to handle legacy studies.
If there will be NMPA-specific business rules and severity levels (Error, Warning, Reject).
CDISC is in the process of translating their standards (dataset and variable labels).
In progress: SDTM-IG 3.3, ADaM-IG 1.1, and CDISC Controlled Terminology (CT)
Published: only SDTM-IG 3.2, available when logged in to CDISC’s website. It is expected that CDISC will publish the remaining documents here when they are completed; we will also inform our users when they do. Please subscribe to our mailing list for industry news.
With over 30,000 terms to manage, CDISC will likely prioritize a certain subset, and leave some terms in English. E.g., in the YES/NO codelist, "Y" and "N" remain in English; but the terms "unknown" and "not applicable" are translated into Chinese.
Datasets: Translation & Encoding
The NMPA requires Chinese-language datasets in SAS XPORT V5 format, with dataset labels, variable labels, adverse events, medical history, concomitant medications, and clinical endpoints in Chinese. They also require Chinese-language Define.XML/PDFs, aCRFs, and xDRGs.
If your datasets were originally in English, you’ll need to translate them into Chinese by working with a language services team in-house or at an external vendor. Effort will be needed to keep translations consistent among different datasets and documents (e.g., clinical summary reports).
NMPA accepts no other versions besides SAS XPORT V5. If an agency begins to accept these V8 or V9, we will add support as needed.
Note: Japan’s PMDA accepts submissions in English and requires translation of only certain datasets into Japanese. Submit datasets in their original language, and place the translations in a separate folder, as instructed.
Note: 1 Chinese character ≠ 1 byte. Depending on your encoding, 1 Chinese character might be up to 4 bytes. Thus, the number of characters in the English-language original data might not fit the NMPA’s limits when translated into Chinese:
Dataset and variable names ≤ 8 bytes
Variable labels ≤ 40 bytes
Variable length ≤ 200 bytes
P21 NMPA Engine
As of October 30, 2020, our NMPA Engine now runs 340 rules for SDTM-IGs 3.1.3 and 3.2, 135 rules for Define.xml, and 262 rules for ADaM-IG 1.1 validation—the rulesets with highest confidence.
We’ll update our NMPA Engine with more rules as CDISC Controlled Terminology, IGs, and NMPA-specific rules are released. Until the NMPA clarifies their requirements, please view our NMPA validation rules as a beta version, which could undergo frequent updates.
Our NMPA Engine scans datasets that contain Chinese, and its validation report outputs rule messages and descriptions in Chinese.
If needed, use the rule IDs, which are consistent across engines and languages, to look up the rule’s original English.
Please use this Engine only for data you’re preparing for the NMPA, since key FDA/PMDA checks are disabled.
The Engine allows a mix of English and Chinese.
The rules that check for consistency between Define and the datasets (SD0037, SD1228) work perfectly with a mixed translation, e.g., some test names in VLM and values in the codelist are in Chinese, and the rest are in English in the Define.
The rules that check for consistency between the dictionaries and the datasets are working only for Chinese-language MedDRA. We are adding support next for WHODD Global Chinese.
For now, we disabled checks again CDISC CT (CT2001-6) until CDISC publishes their Chinese translations.
You can access the NMPA Engine after upgrading to Community 3.1 (NMPA 1907.5 Engine) and Enterprise 4.2 (NMPA 1907.2 Engine).
Community upgrades automatically when you login with your Pinnacle ID.
Enterprise upgrades per the schedule from your Customer Success Manager.
Note: We’re considering how to best communicate the differences among the P21/FDA/PMDA Engines and the NMPA Engine.
Artifacts: Define.xml & xDRGs
The P21 Enterprise Define.xml Designer has long been able to handle Chinese character metadata, and its functionality is preserved in this release.
Your Define.xml may mix English and Chinese. In these cases, our Define.xml Designer will reflect your data truly: your data’s mix of English and Chinese will appear as such in the Define.xml output.
Our Define.xml Designer can autofill page numbers from Chinese aCRFs.
We’ve made Define.xml stylesheets available in English, Chinese, and Japanese.
Enterprise users: Download or use them from our Define.XML Designer.