1、The AI outlook for clinical trials:discover new capabilities and efficiencies at every stageHow the latest developments are saving sponsors time and money while improving quality and safety Artificial intelligence(AI)is expected to be the most disruptive emerging technology in the pharma sector in 2

2、023,according to GlobalData.AI spending by pharma companies is predicted to grow at a combined annual growth rate(CAGR)of 18.5%to reach USD 0.9 billion in 2024,and medical device companies are expected to spend USD 0.5 billion by 2024,a CAGR of 20.6%.AI promises to transform clinical research at alm

3、ost every stage,offering new efficiency,accuracy and more powerful analytics than ever before and equipping sponsors to handle a clinical trials market that is growing in scope,scale and competition.2IntroductionAI is transforming the way we work in almost every sector.Advanced analytics,automation

4、and speed can be brought to bear on a diverse range of applications,from security and compliance to user experience and customer service.Not only can AI make processes more efficient by increasing speed and accuracy,it can also do things that humans cant,offering brand new,more powerful capabilities

5、.Combined with advanced technologies like big data,machine learning(ML)and knowledge graphs,AI is becoming ever more valuable and even essential for the future of business.Meanwhile,the clinical trials landscape is evolving.In recent years,the industry has been impacted by significant macro trends,i

6、ncluding the COVID-19 pandemic,climate pressures and geopolitical instability,which have affected supply chains,regulatory requirements and patient populations as well as transformed the way clinical research is conducted.As todays clinical trials grow in scale and scope,they are producing more data

7、 than ever before.At the same time,advances like adaptive design,personalisation and complex novel treatments mean sponsors must be able to handle ever more complex,agile trial design.The rate of clinical research is also increasing.Traditionally,it took up to 10 years to take a drug from concept to

8、 market,but pressures like increasing competition and the COVID-19 pandemic have forced sponsors to reduce their time frames.Yet many sponsors are still running paper-based clinical trials involving many highly manual processes.Traditional trial design is limited to the demands of todays environment

9、,and sponsors must deal with fragmented data and disconnected systems,excessive manual effort and repetitive tasks and trials that struggle to embrace innovations like digitisation.AI can be utilised in clinical trials with transformative results.It offers optimisation opportunities across the entir

10、e process of clinical research,from trial design and subject recruitment to documentation and post-market surveillance.AIs ability to process large amounts of structured,semi-structured and unstructured data input from a range of sources and to unify them into one digital source powers better predic

11、tive modelling,drives better decision making,and greater actionable insights.Sponsors save time and money while improving quality and safety at every stage.Ultimately,centralising data and using AI means researchers can get more out of it.Using AI to bolster clinical trials transforms the way we can

12、 bring drugs into market.Taimei AI accelerates the speed to market,from concept to approval”-Alonso Diaz,director of quality,Taimei“3Faster patient recruitment rate,better patient engagementThe patient recruitment and enrolment stage is considered one of the most challenging aspects in conducting cl

13、inical trials.This is where the most time-consuming phase happens,slowing down the start of the trial.Limited patient retention during conduct of trial affects conclusive results.Manual recruitment processes can be costly,repetitive and time-consuming.Sponsors must find and identify a set of subject

14、s before gathering subject information and using inclusion/exclusion criteria to filter them and select patients.With increasing competition,dynamic markets and regulatory pressure,todays clinical trials are working to tighter timelines than ever.Bottlenecks at the early stage can cause significant

15、delays and downstream challenges for sponsors.At the same time,high-quality patient recruitment which identifies the right participants is vital to ensure trial success.Once patients are recruited,they must be managed.Sponsors must oversee randomisation processes and procedures including emergency u

16、nblinding.As clinical trials grow in scale,some involving participants distributed over more than a hundred sites across multiple geographies,manual patient management poses an increasing challenge for sponsors.Taimei Technologys 360-Degree View ApproachBut when it comes to patient recruitment,AI of

17、fers a range of revolutionary functionalities.Since AI can handle large amounts of data,it can gather subject information and screen and filter potential candidates to select appropriate participants.For sponsors struggling to identify a set of subjects,AI can analyse medical records and even social

18、 media content to detect patient subgroups that might benefit from a particular trial,or identify geographical areas where a condition is especially prevalent.Advanced analysis can also be used to alert medical staff and patients about relevant trials,or to reuse subjects for future trials.“4In fact

19、,involving AI in patient recruitment leads to faster enrolment,accelerating the clinical trial and reducing costly delays.It can also enable sponsors to reach more diverse populations and more relevant participants,increasing the quality of clinical research and boosting the chance of the study bein

20、g valuable.Better patient recruitment can also help maximise patient retention,which is a crucial metric for study success.Not only this,but in cases where sponsors have difficulty finding enough subjects to participate in a trial,AI can analyse huge amounts of past data to compare results to the tr

21、eatment group,reducing the number of patients that are required.During the study,AI algorithms can also provide sponsors with real-time insights into study execution and patient adherence,enabling them to optimise study design and protect retention.AI enabled wearable devices and remote patient moni

22、toring will help recruit patients from geographically challenged areas.SHOWCASEeBalance is Taimeis randomisation and trial supply management tool,which covers randomisation,enrolment,emergency unblinding and other processes,as well as real-time monitoring and management of drugs and clinical trials.

23、The use of AI-smart software will boost clinical trial participation through subject-centric interactions and by tapping into hard to reach demographic and underrepresented areas,diversifying clinical trials.”-Alonso Diaz,director of quality,Taimei“5Accelerating study buildThe process of case report

24、 form(CRF)creation and database building can also take weeks before a trial can begin and risks becoming another bottleneck for sponsors.Since CRFs and databases determine what information is collected about patients,it has a direct impact on the quality and accuracy of clinical research data.Theref

25、ore,precision is essential at the study build stage.Traditionally,data managers must read the study protocol,which details information including the instructions for the study,what data must be collected,what methods must be used,what measurements will be taken and what the visit matrix looks like,a

26、s well as generate as many as 50 to 60 CRFs for data collection.For site visits,forms must contain fields including demography,basic information and medical history.They can also include treatment group,lab reports and readings from medical devices.Each trial type has different form requirements,and

27、 different therapeutic areas focus on different information.For example,when working on breast cancer research,genotype and other gene data may be relevant.Once the CRFs are generated,data managers must create the data verification plan.Typically,data entry is a highly manual processes which is open

28、 to human error.For example,data is subject to typos and duplication,or it might be uploaded to the wrong location.In order to ensure accuracy in a clinical trial,data must be cleaned after entry into the system,requiring manual edit checks by the data manager.The entire study-building process can t

29、ake eight to 12 weeks.If the protocol is subsequently updated or modified,the whole procedure must be repeated.6From 10 weeks to 1Taimeis advanced AI has sophisticated text-reading capabilities,meaning it can parse,categorise and stratify entire corpora of words and automatically correlate data elem

30、ents across electronic data capture in electronic case report forms(eCRF).AI reads the clinical trial protocol and automatically generates eCRFs and a matrix directly from the documentation provided into the system,providing faster database build times.Optical character recognition(OCR)technology ca

31、n address structured and unstructured native documents to extract information,including content form text,and tables and graphs,using general processing logic and special processing logic.This step uses completions and post-processing to enable CRFs to be generated in the right format.The programme

32、utilises a built-in protocol electronic data capture verification and edit checks,which are automatically tested on simulated data entry.For example,it can cover basic checks such as data format,range,or logical criteria such as ensuring that the second visit date is later than the first visit date.

33、But the systems can also prioritise high-risk checks to ensure that data points related to safety and efficacy are recorded correctly.Overall,leveraging AI for study build can reduce the timeframe from around 10 weeks to just one.Data managers time is freed up and the study can get underway more pro

34、mptly,helping to accelerate the overall research timeline.The responsibility of the personnel is to improve the knowledge base and give feedback on the systems.In an average study there will be around 600 to 1,000 checks with many parameters each,which can take four to eight weeks to test and build.

35、AI can do this automatically in a fraction of the time.AI can improve the efficiency of the whole process.We have tools to help faster study building to save time before the system goes live.Taimeis solution is also integrated with KODA to automate medical coding.This is a big part of AI application

36、.”-Shou Yuan,director,Global Solutions,TaimeiOur concept is dont do this yourself,leave it to the system.”-Shou Yuan,director,Global Solutions,Taimei“7Flexible,robust electronic data managementOnce the database is built and eCRFs are created,study data can be collected.But data management is a prima

37、ry problem for clinical researchers.Modern clinical trials produce large amounts of data and the input is often diverse and unstructured,since medical records from different institutions and patients can take varying formats and qualities.Meanwhile,innovations in data collection are advancing trial

38、decentralisation and reducing reliance on in-person sites.Technologies such as wearables,telemedicine and at-home self-reporting are excellent tools to increase patient adherence and reduce errors but they also present more diverse data sources for sponsors to manage.As studies increase in scale and

39、 scope and trial decentralisation and virtualisation technologies continue to be adopted,data management systems must be able to cope with increasing trial complexity.AI-enabled data management is flexible,agile and robust.Using electronic data capture(EDC),sponsors cut out the need to manage paper-

40、based documentation.It also means data is centralised in a secure,easy-to-access location.Using knowledge graphs,data management AI can work on diverse data inputs and the vast amounts of clinical trial data input can be used to train algorithms to make them more accurate.Automated EDC also enables

41、matrix design and eCRF version management,with a clear layout to assign forms to visits.Advanced EDCs have the flexibility to cope with protocol amendments which would otherwise mean weeks of re-working by data managers.As well as this,they are able to display data in real time for easy visualisatio

42、n by management and other stakeholders,and AI can be used to auto-populate reports and analyses.It means usual trends can be quickly identified,allowing sponsors to implement proactive and effective actions before a trend becomes a crisis.Behind these technologies,we use a very large knowledge graph

43、 to ensure the quality and robustness of our system.Our knowledge graph is based on expertise and experience.”-Tao Yang,PhD,director of Fleming Research Centre,Taimei“SHOWCASEeCollect is Taimeis robust EDC system specifically designed for agile and complex clinical studies.It is AI automated,robust

44、and data driven,offering data management capabilities that can adapt to complexity and constant change.Intelligent AI applies ML to generate eCRFs directly from the protocol,with auto-testing edit checks for all data types.It can seamlessly adapt to protocol amendments and has proved its scalability

45、 from single site to global multiphase and multi-arm studies.It provides real-time data visualisation using iMensa technology.8Implementing AI:How its doneThroughout the lifetime of a clinical trial,sponsors can adopt or customise new process flows.When it comes to implementing AI,the process is as

46、follows:1.Establish a proof of concept build a small model,gather data and run a prediction to see whether the result meets the requirements2.Build a customised knowledge base3.Train the model to solve the problem on a large scaleHowever,before AI can be employed efficiently,algorithms must be train

47、ed to remove bias and ensure accuracy.Because its algorithms learn from data input,AI enables advanced flexibility and constant improvements.When AI can work on structured and unstructured sources,huge amounts of data become available,and leveraging knowledge graphs can create more sophisticated,rob

48、ust and unbiased algorithms.AI from other established sources can also be brought to bear on clinical trials,thanks to open APIs.For example,existing natural language processing APIs can be integrated into the system to optimise document parsing and creation.This flexibility enables best-in-class ad

49、vances to be integrated into clinical trial systems,fostering inter-industry collaboration,or co-opitition,to yield the most powerful results.The further applications of AI in clinical trials are diverse.For example,in study design,AI can be leveraged to analyse data from past and present trials to

50、inform future research.ML and natural language processing technologies can work on vast data sets from a large number of trials to suggest factors such as appropriate endpoints,site strategies and enrolment models.Better study design leads to fewer protocol amendments and a higher chance of study su

51、ccess.Similarly,in the area of pharmacovigilance,AI,OCR and ML could be used to analyse large amounts of structured and unstructured data to provide insights,which can be used to promptly identify adverse events and to assess risks and performance more efficiently.Natural language processing can als

52、o be used to parse regulatory documentation to make it more useable.One of the most promising AI applications across the medical industry is image and data analysis for diagnosis.An AI algorithm can interpret and analyse far larger volumes of data than a human diagnostician.This means data,including

53、 past medical history and related genes,could be utilised on a diagnosis,yielding more accurate results.AI enables the reuse of big data,saving organisations time and effort in data collection at the start of a trial.Taimeis technology leverages smart automation combined with a large repository of p

54、rotocol examples,making clinical development efficient.Thats the power:AI combined with big data.”-Alonso Diaz,director of quality,Taimei“9The outlookIn the short term,AI promises to take repetitive work away from human personnel.Tasks such as building CRFs,performing edit checks and cleaning data c

55、an be automated,reducing the risk of human error and freeing up entire teams.Thanks to natural language processing and OCR dealing with unstructured data,study building can be entirely automated.The responsibility of personnel will shift to maintaining the knowledge base and providing feedback to th

56、e system.The long-term implications for clinical research of AI are significant.As trial management becomes automated,the focus can shift from trial implementation to drug discovery,with the vision to find better treatments for patients.This,too,can be empowered by AI.For example,during the COVID-19

57、 pandemic,AI was leveraged to find vaccine candidates by going through vast amounts of early-stage clinical data to prioritise candidates for novel treatments.AI can analyse far more trial data than human researchers can.AI promises to transform clinical research,and this is good news for sponsors a

58、nd patients alike.Faster,more accurate and cost-effective clinical trials will ultimately produce more effective treatments faster.The role of the data manager will be shifted.They can leave all the repetitive work to the system.”-Shou Yuan,director,Global Solutions,Taimei“10At Taimei,AI-empowered d

59、igital technology solutions and innovative services are empowering new drug research and development,accelerating new medicine launches for customers and improving business performance.To find out more,visit or reach out to us at .Taimei has a very targeted product and service portfolio offering digital solutions for its partners across the entire clinical process.”-Alonso Diaz,director of quality,Taimei“