Professor Chen Yao: Real-world research from the eyes of a statistician

Foreword: The 2022 China International Drug Information Conference/DIA Annual Conference and Exhibition will be held in Suzhou International Expo Center in Ju.

Although the relevant understandings and expressions are different, the data characteristics of RWS are the same: it is no longer accurate small sample research data, but mixed medical big da.

To meet the above challenges, it has become a consensus that real-world research requires the collaboration of multidisciplinary exper.

Yao Chen: Currently Director of the Medical Statistics Office of Peking University First Hospital, Deputy Director of Peking University Clinical Research Institute and Deputy Director of Hainan Real World Data Research Institu.

ever-changing

The so-called real-world study (RWS) refers to a study that obtains a variety of data in real clinical, community or family settings to evaluate the real impact of a treatment on a patient's heal.

As we all know, the concept of RWS was proposed for traditional clinical trials (RCT.

Therefore, RWS may use some statistical methods (such as propensity scores, e.

In Yao Chen's view, RWS is not an independent and new method, but a formulation of "old wine in a new bottl.

It is true that China has a vast territory and the economic level of patients varies great.

On the one hand, if the traditional RCT method is used, the clinical research data still relies on the manual input and review of the clinical research coordinator (CRC), which will not only lead to high cost and low efficiency, but also cannot fully guarantee the authenticity and accuracy-this is worrying However, it is the true portrayal of most of the RWS in China at prese.

On the other hand, how to convert real-world data (RWD) into real-world evidence (RWE) depends on the extent to which it is us.

Scientific innovation is not achieved overnight, it takes time to accumula.

Changes in the making

In recent years, medical big data processing and data security technologies have gradually matured, and intelligent data collection models have emerged, especially the processing of hospital diagnosis and treatment records, giving birth to another possibility that is different from traditional manual transcripti.

Benefiting from the agglomeration effect of Hainan Boao Lecheng International Medical Tourism Pilot Zone in terms of imported pharmaceutical and equipment policies, medical institutions, product application scenarios, e.

At the end of 2021, the special issue of "Real-World Data Research and Drug Device Supervision Innovation" jointly established by "China Food and Drug Administration" magazine and Hainan Real-World Data Research Institute will be published, which systematically introduces the real-world data research results and cases in Boao Lecheng, Hain.

In order to evaluate the application effect of ESR in clinical practice, Professor Yao Chen's team carried out the deployment of ESR tools and the trial of real-world data research projects in Boao Yiling Life Care Cent.

The test results show that compared with the traditional clinical research process, the ESR-based eSource method can improve the collection efficiency of source data and reduce the workload required to complete data transcripti.

The results of this study demonstrate the feasibility and application value of E.

A detailed study of holistic practice has been published in the UK journal BMC MedicalInformatics and Decision Maki.

According to reports, the pilot work of ESR tools in mainland hospitals is being gradually promoted, including multidisciplinary consultation records for rare diseases and clinical research data management initiated by researchers in medical institutio.

"Everyone may ask: RWS cannot control variation through randomization and blinding, how to ensure the credibility of research results?" Professor Yao Chen is not shy about the shortcomings of real-world research: "We emphasize more on mixing various mixed The factors are recorded, and then statistical methods are used to correct the bi.

For example, the correction of known confounding factors is more commonly used now called propensity score, which is to record the patient's basic situation, accompanying disease treatment and living environment factors, e.

Relative balance between analysis groups is ensured by calculated propensity score matchi.

However, there may also be a loss of sample si.

”

"So the sample size required by RWS is relatively large, and it is more dimensional data," Professor Yao Chen added: "Let's not mention big data, it means large sample size, in fact, more dimensional data is big Da.

Why? You have not recorded other important confounding factors, and no statistician can do anythi.

Although there are some methodological advances, such as the application of instrumental variables, these applications are still limited by some conditio.

So I personally I feel that the easiest and best way is to record these data as completely as possible according to the requirements of the research program, and then conduct multi-factor analysis to explore potential patter.

”

"It is unrealistic to rely entirely on manual metho.

Modern information technology is a very promising application direction to solve current data quality problems," Pr.

Yao Chen further explained: "Automatically convert text through speech and natural language recognition technology (NLP), which can quickly and accurately record data and conduct structured processing without affecting the doctor's diagnosis and treatment behavi.

Similar to the case report form (CRF) filled out by RCT in the past according to the requirements of the research program; RWS also needs to be based on the requirements of the research progr.

To obtain the most original and authentic source data, and then use NLP technology to extract and review and verify, thereby improving efficien.

”

At the same time, the demands and challenges of data standardization cannot be ignor.

"Real world data and clinical research data standards are not the same, we can't ask for unification, but there should be a process of mapping and transformation between the two data standar.

"

In response to this difficult problem, Professor Yao Chen's team explored and designed a set of methods to standardize data, which can be used to automatically populate electronic case report forms that meet the standards of the Clinical Data Interchange Standards Association (CDISC) from EMR source data ( eCRF), and meet the data submission requirements of the regulatory authoriti.

It is reported that the current research results "standard translational research from real-world data to clinical research data" have been published in the November 2021 special issue of "China Food and Drug Administratio.

In Yao Chen's view, the unification of data submission standards is an international tre.

Once implemented, regulators can perform data merging and model comparison when reviewing various statistical reports, which will shorten the approval time; it will also facilitate the application of various models by statisticians; there will also be great opportunities for data sharing in the futu.

Meaning, effectively avoid fragmentati.

At the end of the interview, Professor Yao Chen said that due to the limitations of the methodology of real-world research, it is impossible to obtain an assessment of direct effects like RCT, but more of an assessment of indirect effects, so as to put forward reasonable assumptions for further resear.

We need to keep a clear head, objectively evaluate the current staged results, and face up to the bottleneck of data quality and standardizati.

In short, a clear and clear data collection and quality control process is the guarantee to achieve the rationality and reliability of research conclusio.

It is believed that more and more exchanges and cooperation between statisticians and clinical medical experts will surely promote clinical Research innovation and development!