Considerations on the Application of Real-World Data for Post-Marketing Safety Studies in China

2021-08-26 | Press Releases

Post-marketing safety studies are studies conducted after a medicine has been authorized. The intent of these studies is to identify, characterize or quantify a safety hazard; confirm the safety profile of a medicine; or to measure the effectiveness of risk-management measures.

China’s Good Pharmacovigilance Practices was issued by the National Medical Products Administration (NMPA) on May 7, 2021. The new law goes into effect December 1, 2021 and marks the beginning of a new era in China’s administration of drug safety. According to the new guidelines, drug manufacturers should conduct post-marketing safety studies voluntarily or imposed by NMPA, depending on the risk profiles of their products.

In recent years, the fusion of big data and medical science supported by relevant policies leading to more use of real-world data (RWD) in medical studies. Real-world evidence (RWE) from real-world data has been used for post-marketing assessment of medical products to support decision-making on safety supervision.

In this article, we will discuss the application of RWD on post-marketing safety studies.


Common Data Sources for Post-Marketing Safety Studies

Post-marketing studies can be performed using raw data, i.e., data directly collected from medical staff or patients. Second-hand data, i.e., data collected for other studies. These data may originate from clinical trials or from real-world settings. Data from real-world settings can provide safety information of greater relevance to the wider patient population compared to data from clinical trials with stringent inclusion and exclusion criteria.

The provisional Guideline for the Application of Real-World Data for Generation of Real-World Evidence specified that RWD for post-marketing studies mainly included hospital information system data, medical insurance payment data, and registry data.

Hospital information data can include basic information from patient records, laboratory information, picture archiving and communication systems, pharmacy orders, and electronic medical records (EMR). These data are based on records of clinical diagnosis, treatment, and procedures. They can be used for studying rare adverse events. However, hospital information systems data are multi-source heterogenous data; they are often stored in a decentralized way in China and can be hard to access and process.

Medical insurance payment data are mainly from two sources: the public medical insurance system which can be accessed by the government and academic institutions; and private insurance databases. Medical insurance payment data are relatively uniform and highly standardized over a wider population compared to hospital information systems data, but with limited medical information. Furthermore, lack of accessibility to insurance data is often a big challenge to drug manufacturers in China.

Registry data, including medical product registry and disease registry data, are collected systematically in clinical trials or observational studies. They are collected in a relatively standardized way and often include outcomes sponsors are interested in. In China, there are limited large scale registry database available. They are generally more expensive than other forms of data and may not be accessible by drug manufactures.

Feasibility Assessment of Real-World Data for Post-Marketing Safety Studies

Real-world data reflect clinical experience across a broader and more diversified patient population. They often include larger sample sizes and longer follow-up periods than clinical trials, which can allow assessment of rare and long-term outcomes and provide additional safety information after drug approval. Studies using RWD are often cheaper and time efficient. However,  RWD have some limitations, such as incompleteness, missing  values, and inaccurate records due to the lack of rigorous quality control as there is often no uniform standard for data recording, collection, and storage in China. In addition, data quality varies a lot on different data sources. Therefore, it’s critical to select the appropriate data source suitable for the study objectives.

The original intention of collecting and recording RWD was not generally for clinical research. The data needs to be processed and legally authorized to use. Feasibility assessment of RWD for post marketing studies is necessary, which includes assessment of the raw data and a separate assessment of the processed data. When using RWD in prospective studies, the feasibility assessment of raw data is exempt.

The feasibility of raw data can be assessed in dimensions like accessibility, ethics, compliance, representativeness, integrity of key variables, sample size, and active status of data. In one study, Yang Yu et al. selected 40 key variables from the Mini-Sentinel and the Observational Medical Outcomes Partnership common data model and used them to assess three major electronic health information databases (a regional data set, China’s national medical insurance database, and an EMR database). The results showed that academic institutions could use all three databases, where each included more than 1 million patients covering 55-85% of key variables, preliminarily demonstrating the feasibility of these electronic databases for post-marketing studies. However, the integrity, accuracy, and consistency of the data was not assessed. Further evaluation is needed to identify the most appropriate database for post launch safety study .

Hospital systems data generally contain detailed safety information of drugs used in a hospital; however, information on the use of a drug outside the hospital is usually missing. These data are suitable for studies interested in immediate adverse events and adverse events that are identifiable based on information from testing, inspection,

or other medical procedures and disease course records.

Medical insurance payment data, on the other hand, contain a more complete record of patients’ visits to various medical institutions and their use of covered medications. Some clinical outcome related to the medications can be found in the database, but detailed information on lab values, imagines, and disease course are usually missing. Therefore, these data are generally suitable to study adverse events that requires treatment records.

Registry data generally contain numerous observational outcomes. These data are accurate and well-organized, making them suitable to study  the safety and effectiveness of a medication, as well as cost effectiveness and compliance.

Once the feasibility assessment is complete, the raw data must be processed before they can be used in post-marketing studies. Given the limited access to medical insurance data and registry data in China, hospital systems data are more frequently used in post-marketing studies. These data are generally multi-source heterogenous data, where data processing with authorization is necessary. Adverse events that can be directly identified from diagnosis, laboratory testing, radiographic imaging, and physician notes. However, clinical conditions and treatment information are often recorded as unstructured plain text in hospital systems and require manual review to extract relevant safety information. .

Happy Life Technology (HLT) has developed an intelligent pharmacovigilance monitoring tool 1using big data and artificial intelligence (AI) technology to detect adverse events more efficiently and reduce physician’s workload. This tool was established based on the Common Terminology Criteria for Adverse Events v4. 0. 3standard ( CTCAE 4. 0. 3). to automatically identify adverse events from hospital systems data.

This tool shows that useful information from RWD can be extracted and transferred. Special attention must be paid to the accuracy, transparency, and quality control of data during the governance process.

Other detailed requirements and procedures for data feasibility assessments are available in the provisional Guideline for the Application of Real-World Data for Generation of Real-World Evidence. Additional information is also available in the article, “Feasibility Assessment and Governance of RWD in China”.

Moreover, in post-marketing risk management, evidence from a single study might not be sufficient to support regulatory decision making. Multi-type, multi-dimensional, and complementary studies should be considered to assess the risk comprehensively and ensure safety requirements from regulatory agency are met.

1 JIANG MinZHU TongJI Jia FuWANG Xiao YunYU Wen BoApplication of an intelligent adverse event monitoring tool in detecting adverse event in clinical trial. Chin J Clin Pharmacol, 2019, 035(006):573-576.