Qu Yi Wang Yuzhuo,Yuan Xiaoliang,Wang Chuang,Huang Zhe,Chen Yuwen

(1.School of Business Administration,Shenyang Pharmaceutical University,Shenyang 110016,China;2.Research Institute of Drug Regulatory Science,Shenyang Pharmaceutical University,Shenyang 110016,China)

Abstract Objective To introduce the design of observational trials under real-world study (RWS),and to provide guidance for clinical development of new research trials and reference for real-world researchers.Methods Relevant literature of RWS and observational experiments at home and abroad were reviewed and analyzed,and then the design of observational studies was summarized under RWS.Results and Conclusion The data and information provided in observational studies not only help to further verify the clinical study results obtained by randomized controlled trial (RCT) in clinical practice,but also objectively reflect the real situation in the process of clinical research and application.

Keywords: real-world study (RWS);randomized controlled trial (RCT);observational study

In recent years,real-world study (RWS) has gradually become a focus of medical circles.In 2007,the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) formulated a framework document for Real-World Data (RWD) to support decision-making,emphasizing the integration of RWD and randomized controlled trial (RCT) data for decision-making.In 2016,the United States implemented the “21st Century Cures Act”,clarifying that the FDA can use RWD as evidence for the approval of drug and medical device post-market research and the development of new indications under appropriate circumstances.In May 2019,the National Medical Products Administration (NMPA)issued the “Basic Principles for Drug Development Supported by Real-World Evidence (Draft for Comment)”.In January 2020,the “Guiding Principles for Drug Development and Evaluation Supported by Real-World Evidence (Trial)” was released.In September 2020,Center for Drug Evaluation issued the “Technical Guidelines for the Development and Evaluation of Children’s Drugs Supported by Real-World Study (Trial)”.RWS originates from practical clinical research for post-market evaluation of drugs,which is used for further evaluation of drugs in actual clinical practice.Many people think that RWS can only be an experimental study,but it is not.In fact,the basic design of RWS can be either experimental or observational.This article mainly discusses the design of observational studies in RWS[1].

1 The concept of observational study

There are essential differences between observational study and experimental study.Experimental study,also called interventional research,which means the researcher selects specific populations to give specific instructions as intervention methods and measures to avoid bias affecting the outcome of comparisons between groups.Then,specific follow-up and observation can be conducted.It is divided into RCT and non-randomized controlled trial.Experimental study under RWS usually includes pragmatic clinical trial (PCT) and a randomized controlled single-arm clinical trial that uses real-world evidence as external evidence.PCT is a test type between RCT and observational study[2].

Observational study,also called investigative research,refers to observing and recording the physiological characteristics of the research object objectively and the development of the disease in the absence of any intervention treatment or control measures in the early stage.Thus,researchers can have an accurate understanding of the process and law of the early occurrence and later development of a certain disease.The research method can be divided into descriptive study and analytical study[3].

2 The need for observational study

Although RCT is considered by clinical experts to be a gold standard for evaluating the safety and effectiveness of drugs,it also has many unavoidable problems and limitations.Firstly,RCT has strict entry and discharge control on the trial population.Therefore,the trial population cannot accurately represent the target population,causing clinical researchers to face great technical pressure and challenges when extrapolating their conclusions to their actual clinical applications.Secondly,due to the long-term lack of effective clinical treatment and prevention,it is difficult for RCT to get effective clinical treatment for rare diseases and some serious diseases that may be life-threatening.Thirdly,RCT requires high research time and cost[4].

Observational study is the non-interventional observation and evaluation of clinical research under natural conditions.Therefore,in a sense,the data and information provided by the results of observational study can not only further verify the results obtained by RCT in clinical practice,but also more accurately reflect the actual situation in the process of clinical research and application.With the development and improvement of research theoretical methods,experimental design,implementation process and statistical analysis,more and more observational studies have been published and recognized in some international journals.

Observational study can be used to replace phase I or phase II clinical trials conducted in conventional clinical trials to initially explore clinical effectiveness and safety.Based on the existing data of observational studies,RCT or PCT is used to further confirm the effectiveness and safety of some traditional Chinese medicines,and provide supporting evidence for the registration,marketing and application of them.After evaluating by relevant departments,the high-quality RWD can be taken as real-world evidence through well-designed observational studies.Then company can communicate with the drug quality supervision and management department and use the evidence to support its drug registration and marketing[5].

3 Characteristics of observational study

3.1 The principle of non-random,non-intervention and openness

RCT mainly adopts the principle of randomization,blinding and control.Observational study mainly adopts non-random,non-intervention and open research principles.The main purpose of the nonrandom principle is not to randomize the grouping of subjects.Researchers generally choose the most suitable grouping according to the actual situation of the subjects.At the same time,the researcher will not impose any intervention on the subjects.Only the safety and actual situation of the subjects will be recorded,and the data will be analyzed.Therefore,it is the principle of non-intervention.The main purpose of the openness principle is to not add any blind design to the research,and both the researchers and the subjects have a complete understanding of the entire research design.Thus,the subjects can obtain a better causal effect during the research[6].

3.2 Large sample size without strict entry and discharge standard

Due to various factors such as funds,resources,and time,the sample size of RCT research is usually small.Generally,researchers will choose a scientific and rigorous statistical method to calculate the sample size for a minimum sample quantity.For observational study,to better obtain a more realistic situation and improve the clinical treatment effect,it is necessary to cover a large number of people.Therefore,the sample size of observational study is large.Besides,there is no need to set strict entry and discharge standards for the population participating in the study.Under normal circumstances,as long as the population participating in the study can meet their clinical requirements and drug quality standards,they can be included[7].

3.3 High external validity

Although RCT has high internal validity,its external validity is relatively poor,which makes it a major disadvantage.The environment of observational study is close to the actual clinical research environment of diagnosis or treatment,so it has higher external validity.In the process of analysis,it is often impossible to avoid and control all confounding factors in a timely manner.Therefore,in clinical research,there are confounding factors that are not correctly observed or controlled,which directly leads to the bias of observational study results[8].

4 Type of study

4.1 Descriptive study

By investigating or collecting data to record various disease-related problems such as the main causes and disease distribution in various groups of people,we can establish scientific hypotheses and provide theoretical support for clinical etiology research and clinical work.

Descriptive studies,also known as prevalence studies,are mainly cross-sectional studies based on the requirements of pre-research design that collect information about certain disease or health condition in a specific population within a specific time range.Sampling surveys or general surveys are used to describe the specific distribution of a certain disease or health status of population and its influencing factors.From the perspective of research time,the prevalence study mainly refers to research completed in a certain time or within a short period of time,so it is also called cross-sectional study[9].

Cross-sectional study has its obvious advantages.Firstly,sampling surveys are often used,and the analysis results have strong promotion and reference significance.Secondly,it can provide evidence for etiological research,which is an important theoretical basis for carrying out analytical etiological research.Thirdly,the cross-sectional study has a control group composed of the same population,which is highly referable.But it also has the following limitations.For instance,it is difficult to determine the temporal relationship of the first cause and effect.Besides,the data of incidence rate cannot be accurately obtained.Lastly,prevalence disease of the study population may be underestimated[10].

The basic steps of the cross-sectional study include selecting the topic and determining the main purpose of the research,research object,sample size and sampling method,designing and drafting the quality questionnaire,determining the sample measurement and inspection method,formulating and establishing necessary quality and safety control measures,and analyzing the data[10].

4.2 Analytical study

By carrying out clinical research and analyzing the relationship between various common diseases and their influencing factors,we can explore and verify the etiology and the main influencing factors on the disease.Then we can effectively prevent the occurrence of various common diseases and propose disease control methods.

4.2.1 Cohort study

The cohort study divides the objects of study into exposure group and control group,and the grouping standard is based on whether they need to be exposed to factors.Then the objects of study should be followed up for a certain period of time.After that,the differences in disease incidence and mortality between the two groups should be compared.Finally,the degree of association between these exposure factors and the disease will be analyzed[11].

According to the start time of the observation,cohort studies can generally be subdivided into prospective cohort study,retrospective cohort study and ambi-spective study.The first refers to the observation time of the study starts from a certain point of the current time to the future.While the second means the observation time of the study is from a certain point in the past to the present.The last indicates the observation time of the study starts from a certain point in the past to the present and then to the future[11].

4.2.2 Case-control study

A case-control study takes patients with a specific disease as the case,and other individuals without the disease but with a certain reference as a control.The exposure history of various risk factors that may occur before should be collected.The exposure ratio of risk factors in the case group and other individual control group is analyzed and compared.After testing,if the difference between the two groups has a certain significance,it can be confirmed that there is a certain statistical association between the factors and the disease[11].

Case-control studies are used to explore the causal relationship between risk factors and diseases.It can be widely used to study the causes of various diseases with unknown or multiple factors.In addition,it can also be used to study the risk factors of accidental injuries and various health-related events.Lastly,case-control studies are also applied to specifically evaluate the quality and effect of immunization[11].

Case-control studies are not as good as cohort studies in terms of the scientific diagnosis of the etiology,the quality,and reliability of the results.However,due to the characteristics of time saving,convenience,low cost,easy discovery and implementation,they are widely used in practice.They can also reduce the occurrence of bias and improve the quality and reliability of the results through rigorous design and strict quality risk management control[11].

5 Test design

5.1 Clarifying the purpose of the study

When conducting observational study based on RWD,we should fully consider the questions to be answered by research evidence in future medical decisions,and clarify the purpose of the study.The use of RWD to evaluate clinical outcomes is generally for post-marketing drugs,medical devices,or existing diagnosis and treatment strategies[12].

The purpose of the cross-sectional study is to describe the main distribution of disease or health status in the population,analyze the scientific hypothesis of the cause by exploring the relationship between certain influencing factors and disease or health status,study how to effectively formulate medical and health policies and measures,and provide evidence for scientific detection of diseases[10].

The purpose of the cohort study is to test the clinical etiology and scientific hypothesis of a new drug,evaluate its effect after clinical treatment,study the early disease history of patients who take the new drug,and carry out long-term monitoring after the new drug is marketed[13].

The purpose of the case-control study is to study the possible occurrence of various diseases in the clinical process and their influencing factors,research on the influencing factors of health status and other events,and analyze the effect of clinical prognosis on various factors[14].

5.2 Constructing research questions and research hypotheses

After clarifying the purpose of the study,we should construct four key elements of the research question,that is,whether the P (population),I(intervention),C (comparable control),and O(outcome) of the study can be compared with the existing RWD.For example,whether population data that meets the research requirements can be extracted from RWD,whether the existing interventions can constitute exposures that meet the research definition,whether there are sufficient comparable controls,and whether it includes the outcome evaluation indicators and standardized measurement results[12].

5.3 Assessing the degree of matching between the database and the research question

Whether it is a database based on existing health information and medical data,or a registered database based on prospective collection,it is necessary to conduct a preliminary matching between RWD and research-related issues before formulating research methods.This evaluation can help experts and researchers choose the appropriate research design to ensure the accuracy and feasibility of their research[12].

5.4 Selecting research design

When conducting observational studies based on RWD,researchers often use cross-sectional studies,cohort studies (including prospective,retrospective,and ambi-spective cohorts),and case-control studies to evaluate treatment outcomes.Researchers can choose different research designs according to the degree of matching between the database and the research question.If necessary,they can also choose multiple research designs at the same time[12].

5.5 Selecting the object of the study

When conducting observational study based on RWD,the relationship between the population covered by the database,the target population and the research population should be clearly given.The specific determination and selection of the study objects need to be considered but not limited to the following key factors,such as screening patients with the target disease,selecting the first medication case,defining the relevant time point of exposure,minimizing the population who may be lost to followup,and clarifying the number of cases included in the study[14].

Cross-sectional study is mainly selecting a specific population in a certain period to conduct a general survey or partial survey (such as a sample survey) on a certain disease.After obtaining specific data of some factors,the disease rate can be finally analyzed (generally speaking,the current prevalence rate rather than the morbidity rate)[10].

Cohort study is selecting a specific population,grouping them into cohorts according to the exposure situation,observing them for a period of time,and analyzing the relationship between risk factors and research outcomes[13].

Case-control study is to select cases (usually new cases) and appropriate control cases respectively for comparison.The ratio of these exposures in the case group and the control group is observed to help researchers judge the possible effects of these exposures on the disease[14].

5.6 Definition of exposure

When conducting observational studies based on RWD,it is essential to accurately define drug exposures to avoid misclassification bias.Therefore,researchers need to re-establish the algorithm of exposure by analyzing and referencing the variable information of some existing algorithms in the realworld database to verify the accuracy and reliability of these exposures.First,patients decide whether to take the study drug.Then,the drug exposure time should be clarified.Next,we must accurately assess whether there are behaviors such as withdrawal,exchange,and contamination of the drug.If necessary,we also calculate the cumulative exposure dose based on the information in the real-world database,such as the drug formulation,dosage,and use mode.Observational study based on RWD cannot avoid the lack of exposure information.Whether it is the lack of medication information before the intervention or after the intervention,it will affect the definition of drug exposure,leading to misclassification bias.Therefore,different algorithms can be designed to accurately define exposure.Besides,sensitivity data analysis can be conducted for different exposure categories[12].

5.7 Deciding follow-up time

When conducting observational studies based on RWD,it is important to decide the follow-up time reasonably,because the basic condition of the study objects should meet the selection criteria at the beginning of the follow-up rather than after it.In RWS,there can be two situations.The first is the follow-up starts when the enrollment criteria are met,such as the patient receives percutaneous coronary intervention (PCI) or coronary artery bypass graft(CABG) treatment for the first time.The second situation is that in many cases,there may be multiple time points for the study objects who can meet the inclusion criteria.For example,when evaluating the effect of estrogen replacement therapy,the inclusion criteria of the study objects is women who have no history of chronic disease after menopause and have not been treated with hormones in the past two years.Since some women may be 51-65 years old,they have many time points that can meet the inclusion criteria.In order to control the bias introduced by the different start-up time of the two groups,the following processing method can be considered.If the original data is collected according to a determined followup time interval,such as in some ongoing cohort studies,the routine follow-up is performed every two years,and then the time baseline that meets the inclusion and exclusion criteria can be randomly selected and defined as the start time of the follow-up.If the original data does not follow the pre-determined follow-up time interval for collection,such as patients with chronic diseases,the start time of follow-up can be determined as the time point with the largest interval from the last treatment to ensure that they experience a long drug washout period[12].

Compared with other study designs,cohort studies usually require a larger sample size.Therefore,they should be observed in different research sites,or research centers,which directly demands that staffs who conduct the observation should have a unified guidance and training.During the cohort study,the patient’s diagnosis and admission criteria should be accurate and consistent to avoid bias.Special attention should be paid to immediacy and accuracy in collecting and processing the information.Due to the long follow-up duration of cohort studies in different research centers,some treatment methods and management measures that can effectively improve the compliance of long-term follow-up patients should be formulated in time based on the actual situation of the study.Besides,the loss of follow-up rate should be controlled within 5%-10%.

5.8 Evaluating study outcomes

Under normal circumstances,the assessment of the outcome in RWS is not done through blind evaluation and other methods,so the accuracy of all outcomes cannot be guaranteed.When conducting observational studies,we should try to choose objective endpoint indicators[12].

6 Conclusions

Observational study is an important type of trial design for RWS.Its main characteristics include the principles of non-randomization,non-intervention and openness.Besides,the sample size is usually large,and there is no strict trial entry and discharge standard,which has good external effects.Since observational study is a non-interventional evaluation and observation conducted in a normal state,the data and information provided in observational study can be further used in clinical practice to analyze and verify the study results obtained by RCT.Besides,it can objectively reflect the real situation in the clinical research and application process.