Content uploaded by Marjan Doostan
Author content
All content in this area was uploaded by Marjan Doostan on Aug 28, 2024
Content may be subject to copyright.
Content uploaded by Viswakanth Makutam
Author content
All content in this area was uploaded by Viswakanth Makutam on Aug 27, 2024
Content may be subject to copyright.
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
205
Journal Homepage: - www.journalijar.com
Article DOI: 10.21474/IJAR01/19245
DOI URL: http://dx.doi.org/10.21474/IJAR01/19245
RESEARCH ARTICLE
INTEGRATION OF ARTIFICIAL INTELLIGENCE IN ADAPTIVE TRIAL DESIGNS: ENHANCING
EFFICIENCY AND PATIENT-CENTRIC OUTCOMES
Viswakanth Makutam1, Sai Yashashwini Achanti1 and Marjan Doostan2
1. Pharm. D, MS Applied Clinical Research, ACRP-CP, Quality Assurance Associate.
2. Pharm. D, MS Applied Clinical Research, Clinical Research Coordinator.
……………………………………………………………………………………………………....
Manuscript Info Abstract
……………………. ………………………………………………………………
Manuscript History
Received: 05 June 2024
Final Accepted: 08 July 2024
Published: August 2024
Key words:-
Artificial Intelligence, Adaptive Trial
Designs, Clinical Trials, Data Analytics,
Predictive Modelling, Patient Outcomes
Background: Integrating artificial intelligence (AI) into adaptive trial
designs represents a transformative approach in clinical research,
promising enhanced efficiency and accuracy in trial outcomes. This
study aims to systematically review the current landscape of AI
applications in adaptive clinical trial designs.
Methods: A comprehensive search was conducted across multiple
databases, resulting in 6177 records initially identified. After removing
duplicates and ineligible records, 1476 studies were screened.
Following rigorous screening and eligibility assessment, 45 studies
were included in the final review. Inclusion criteria focused on peer-
reviewed articles, systematic reviews, and clinical trials discussing the
role of AI in adaptive trial designs. In contrast, exclusion criteria
eliminated non-relevant and low-quality studies.
Results: The selected studies demonstrate that AI significantly
improves adaptive trial designs through advanced data analytics,
predictive modelling, and real-time decision-making. AI's integration
facilitates dynamic randomization, optimised dosing strategies, and
efficient patient recruitment, thereby enhancing the overall
effectiveness of clinical trials.
Conclusion: AI integration in adaptive trial designs offers substantial
benefits regarding trial efficiency, precision, and patient outcomes.
Despite existing challenges such as data quality, ethical considerations,
and regulatory requirements, the findings underscore the potential for
AI to revolutionise clinical trials. Future research should address these
challenges to harness AI's capabilities in adaptive trial designs fully.
Copy Right, IJAR, 2024,. All rights reserved.
……………………………………………………………………………………………………....
Introduction:-
Clinical trials are the gold standard in clinical research, but when resources are few, conventional designs with set
sample numbers and restricted arms sometimes cannot adequately address contemporary, complicated research
issues. A solution is provided by adaptive clinical trial designs, which permit future adjustments in response to
gathering data. In 2019, the US Food and Drug Administration (FDA) released guidelines for adaptable designs that
included non-binding suggestions for their creation, use, and disclosure. Compared to traditional methods, adaptive
designs have several advantages. These include increased statistical efficiency (e.g., managing type I error rates and
power), ethical concerns addressed (e.g., increasing allocation to better-performing arms, stopping trials early for
Corresponding Author:- Viswakanth Makutam
Address:- Pharm. D, MS Applied Clinical Research, ACRP-CP, Quality Assurance
Associate.
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
206
safety issues or limited benefit), and improved understanding of treatment effects (e.g., enriching trials with
participants expected to benefit). [1], [2], [3], [4], [5]
Adaptive designs (ADs) provide clinical trials with flexibility by allowing for pre-planned alterations depending on
interim results. This could improve trial quality and speed up or decrease the trial duration. Careful planning can
solve practical obstacles, including money, communication, and execution. Learning from prior experiences and
case studies is crucial, and more practitioners should write up their insights. Although inappropriate for all studies,
especially those with lengthy outcome measures, ADs should be an essential methodological tool for clinical
investigators. ADs protect against erroneous beginning assumptions but are not a solution for poorly thought-out
trials. Although bias is possible, it may be reduced by careful preparation, openness, skilled Independent Data
Monitoring Committees (IDMCs), and blinding triallists to modifications. Precise and thorough reporting of trial
outcomes and design details is essential for future education, and reporting procedures are codified into an AD
appendix to the CONSORT guidelines.[6]
Adaptive trial designs provide speed and flexibility through planned adjustments depending on interim data analysis,
a significant advancement in clinical research. These designs enhance statistical power and ethical standards by
allowing modifications like early halting for efficacy or futility, re-estimating sample numbers, and adjusting
treatment allocations. The group sequential design, which incorporates scheduled intermediate evaluations to
determine early termination based on predetermined criteria, is one essential component.[7], [8] As the trial continues,
adaptive randomisation using minimisation modifies the likelihood of treatment assignments in favour of more
successful therapies.[9]Reestimating the sample size enables adjustments to guarantee sufficient power, lowering the
possibility of underpowered research.[10]
Moreover, adaptive trials allow for adding or removing treatment arms in response to interim data, improving
resource allocation flexibility.[11] By concentrating on particular subgroups that are probably to benefit more from
the therapy, enrichment strategies increase the accuracy of trial results.[12] Phase II/III designs that are seamless
integrate these stages into a single, ongoing trial, using phase II results to guide phase III decisions without stopping
the investigation.[13],[14] Adaptive dose-finding maximises dose determination more effectively by modifying dosage
in response to participant reactions.[15]
Artificial intelligence (AI) is a field dedicated to replicating human-like abilities in perception, reasoning, learning,
and problem-solving, particularly within clinical contexts.[16]AI is increasingly seen as a critical driver for
sustainable and optimised drug development, with practical applications in clinical trials (CTs) now emerging. The
growth of randomised trials has led to vast amounts of complex clinical, molecular, and imaging data. This data
abundance is crucial for data-driven and personalised medicine trends. However, comprehensive AI models trained
on suitable datasets are essential to derive actionable insights. These models can significantly expedite and
streamline various activities within drug research, enhancing the efficiency and effectiveness of the process.[17]
Integrating AI into adaptive trial designs represents a revolutionary advancement in clinical research, offering
unprecedented efficiency and precision. Traditional clinical trials, often limited by fixed sample sizes and rigid
protocols, need help addressing contemporary research questions' dynamic and complex nature. However, Adaptive
trial designs allow for real-time modifications based on interim data, enhancing statistical power and ethical
standards. AI's role in this paradigm is transformative, facilitating advanced data analytics, predictive modelling, and
real-time decision-making, significantly improving trial quality and speed. AI enables dynamic randomisation,
optimised dosing strategies, and efficient patient recruitment, making clinical trials more responsive to emerging
data and patient needs. This study systematically reviews the current landscape of AI applications in adaptive trial
designs, highlighting the substantial benefits and addressing the challenges, such as data quality, ethical
considerations, and regulatory requirements. By providing a comprehensive overview of AI-driven innovations in
clinical trials, this research underscores the potential for AI to revolutionise the field, ultimately leading to faster,
more accurate, and patient-centric outcomes. As the healthcare industry embraces AI, understanding its integration
into adaptive trial designs becomes crucial for advancing medical research and improving patient care.
This review article explores the evolving role of adaptive clinical trial designs (ADs) and artificial intelligence (AI)
in modern clinical research. It will assess the advantages, challenges, and practical applications of ADs and AI,
emphasising how they can enhance clinical trials' efficiency, flexibility, and ethical standards.
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
207
Methodology:-
This review used preferred reporting items for systematic reviews and meta-analyses (PRISMA) recommendations.
The three authors, ViswakanthMakutam, Sai Yashahwini Achanti, and Marjan Doostan independently assessed all
eligible studies. If there were any cases of disagreement, a consensus was reached. We searched through the
electronic database PubMed (until June 2024) using the search terms “Artificial Intelligence in Clinical Trials,”
“Machine Learning in Clinical Trials,” and “Adaptive Clinical Trial designs.”
From the PubMed database, a comprehensive search yielded a total of 6177 articles, which included 2447 articles
for “Machine Learning in Clinical Trials”; 3593 articles were earmarked for “Artificial Intelligence in Clinical
Trials” and 137 articles for “Adaptive Clinical Trial designs”. However, after meticulous scrutiny, 6132 articles
were deemed irrelevant to the review's focus based on the inclusion and exclusion criteria and consequently
excluded. Ultimately, 45 articles remained for detailed analysis, each considered pertinent to the study's objectives.
The flowchart depicts the systematic process of identifying, screening, and selecting studies for a review on the role
of artificial intelligence (AI) in clinical trials and adaptive trial designs. Initially, 6177 records were identified from
databases, but no records were found in registers. Before screening, 3251 records were marked ineligible by
automation tools, and 1450 were removed for other reasons, leaving 1476 records to be screened. Following the
screening, no records were excluded at this stage. All 1476 records were assessed for eligibility, excluding 1035
records that were not related to AI in clinical trials or adaptive trial designs and 396 records with unclear
methodology or insufficient data. Ultimately, 45 studies were included in the review. This flowchart ensures a
rigorous selection, enhancing the review's credibility and relevance.
Figure 1:- Flow Diagram of Papers Researched.
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
208
Results:-
Clinical Trial Design with AI:
Designing clinical trials is complex, requiring accuracy, reliability, power, and generalizability. Adaptive designs,
which modify trials using interim data, show promise but need robust, unbiased statistical analysis. AI can
significantly enhance this process by simulating multiple rounds of data analysis, such as tenfold validation or the
parametric G formula.[18]In statistical planning, AI's predictive analytics can improve interim and final analyses,
operationalise study design through stratified cohort identification, and optimize designs involving multiple
biomarkers and treatments from various modalities. The primary limitation of current trials is their generalizability
and external validity. Advanced methodologies like Markov chain Monte Carlo, decision rules, Bayesian analysis,
and novel AI techniques can address these issues. AI can select targeted populations by analyzing multiple variables,
such as ethnicity, lifestyle, diet, income, and geographical location, improving the external validity of findings.
Additionally, AI can identify drug efficacy and adverse reactions, reducing time and costs compared to traditional
trials. For instance, AI applied to wearable technology can integrate real-world variables into trial protocols. AI also
facilitates genome-driven basket/umbrella trials, enhancing statistical methods and power by selecting genes or
single nucleotide variants in small populations with rare heterogeneous syndromes.[19]
Prediction of Clinical Trial Outcomes:
AI and machine learning (ML) significantly enhance clinical trials (CTs) by supporting early disease detection,
predicting molecular features, target sensitivity, bioavailability, and toxicity. These technologies reduce late-stage
trial failures, increasing the likelihood of Phase II/III trials advancing to regulatory approval. The benefits extend to
conserving human and financial resources and ensuring participant safety, positively impacting public perception of
CTs. ML models, informed by CT design and patient data, can predict regulatory approval and estimate success
probabilities in phase transitions, considering factors such as protocol complexity, clinical endpoints, interventional
arms, and eligibility criteria. Additionally, risk score-based models can help design safer and more effective trials. In
oncology, AI builds in-silico trials using clinical data to simulate cohorts and model treatment effectiveness,
identifying better responders and reducing development failures. However, lacking high-quality, curated datasets
remains a significant challenge, limiting AI's full potential. Overall, AI and ML offer transformative improvements
in trial design, execution, and success rates.[20], [21], [22]
Forecasting Clinical Outcomes in Precision Medicine:
AI is pivotal in predicting clinical outcomes, crucial for precision medicine, and optimizes trial design by
minimizing statistical variability. AI can simulate data to identify efficient statistical outcome measures and predict
participant outcomes, potentially shortening trial durations. By analyzing electronic medical records (EMRs), AI can
also predict clinical trial dropouts, allowing targeted interventions to retain participants and reduce the overall
sample sizes needed. In cardiovascular trials, targeting potential dropouts with additional education has shown
promise in extending participation. Machine learning (ML) prediction models have demonstrated a significant
impact, reducing cancer mortality by 15-25% across various trials. These models utilize large biological databases to
correlate drug-related predictive biomarkers with survival data stratified by genetic and environmental factors. For
instance, ML tumor growth models have been validated in non-small cell lung cancer trials to predict tumor
response and survival rates based on biomarker status and other complex drivers. AI tools enhance drug selection
and adapt investigational drugs to specific cancer histologies, potentially increasing survival rates. As ML models
incorporate comprehensive multi-omic data, they are poised to transform treatment paradigms and redefine precision
trial design and recruitment, making them more efficient and targeted.[22], [23], [24]
Randomization:
Randomization is crucial in RCTs to ensure that treatment arms have similar patient characteristics, making the
treatment assignment the only differing factor. Traditional coin flip randomization is simple but lacks control over
allocations and covariate adjustment. Stratified block randomization improves on this by using pre-specified
sequences within patient subgroups, though it struggles with categorizing continuous variables, managing many
covariates, and potential selection bias. Covariate-adaptive designs, such as the minimization method, address these
limitations by balancing multiple covariates simultaneously. This method reduces imbalance across covariates but is
more complex and requires additional resources. Response-adaptive designs modify allocation ratios based on
ongoing trial success, aiming to allocate more participants to effective arms. While potentially increasing no
responders and introducing biases from temporal trends, this method can be advantageous in multiarmed trials
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
209
where the control group maintains a fixed allocation. However, caution is needed due to risks of unblinding and
estimation variability. [9], [25], [26], [27], [28], [29], [30], [31], [32], [33]
Randomization in clinical trials often involves assigning participants to a control group, which can be resource-
intensive. AI models, trained on extensive clinical, genetic, and imaging data, offer the potential to replace
traditional control arms with virtual controls, significantly reducing recruitment needs. For instance, deep learning
models have shown promise in predicting long-term disease progression from baseline data. Validating these AI
models in independent populations is essential to ensure accuracy. While consuming clinical metadata, DL models
have demonstrated a proof of concept by being able to forecast the appearance of a Humphrey visual field (HVF)
from a single baseline HVF up to 5.5 years in advance.[34] Incorporating synthetic control arms alongside traditional
placebo groups in trials can help evaluate AI predictions without affecting results. Large-scale collaborations and
standardisedprotocols are crucial for developing robust AI algorithms. Despite challenges like dependency on well-
labelled data and AI's "black-box" nature, methods to elucidate AI decision-making can integrate AI effectively into
RCTs, enhancing efficiency and participation rates.[35]
Treatment arm:
Adaptive treatment arm selection improves efficiency and relevance by adjusting research designs by including or
excluding treatment arms. Adaptable dose-finding, seamless drop-the-losers, and adaptable platform designs are
essential techniques. Like the ongoing reassessment approach, adaptive dose-finding establishes the ideal dose early
in research. Adaptive seamless designs eliminate phase transition time and reduce total sample size by combining
the objectives of two trial phases. Drop-the-losers strategies concentrate on the most promising therapies, discarding
less effective treatments following interim assessments. Adaptive platform trials focus on the condition rather than
individual therapies by examining several interventions for a given condition and allowing treatments to be added or
eliminated based on real-time data. These techniques enhance the relevance, efficiency, and adaptability of trials.[36],
[37], [38]
AI can create external control arms, making trials more patient-centric and efficient. Unlearn'sTwinRCTsTM
combines AI, digital twins, and advanced statistical methods to improve trial success rates with fewer
patients.[39]Digital twins predict disease progression using historical control data, enhancing confidence in treatment
effects. This approach, suggested by the European Medicines Agency for phase II and III trials, is appealing because
it reduces reliance on traditional control groups, increasing patient likelihood of receiving treatment and speeding up
enrollment. However, rigorous guidelines are needed to standardize synthetic patient generation and ensure
comparability with traditional placebo arms.[40]
Patient Selection:
Patient selection in clinical trials faces multiple challenges, such as eligibility, suitability, and patient motivation.
Traditional methods often lead to delays and high failure rates, especially in Phase III trials where recruitment
problems are common. AI and machine learning (ML) systems can significantly enhance patient cohort composition
and recruitment efficiency by analyzing comprehensive clinical, genetic, and imaging data. This allows AI to
accurately identify eligible and suitable patients, predict disease progression, and facilitate better trial matches
without manual reviews. AI can assist in clinical trial enrichment by selecting patients who are more likely to
demonstrate the drug's effect. Sophisticated analytics methods combine omic data with electronic medical records
(EMR) and other patient data to identify suitable biomarkers and patient subpopulations. Techniques like natural
language processing (NLP) and optical character recognition (OCR) automate data extraction and harmonization,
enhancing patient profiling and selection. Electronic phenotyping within health informatics reduces population
heterogeneity by identifying patients with specific characteristics, further refined by advanced ML methods for
improved prognostic and predictive enrichment.[41], [42]
Currently, AI is most frequently applied to patient recruitment in clinical trials. Tools like Deep Six use NLP to
analyses physicians' notes, pathology reports, and lifestyle data to match patients to suitable trials, while Mendel.ai
interprets unstructured data from clinicaltrials.gov to find qualified patients. AI can also leverage EHRs to identify
eligible patients or define criteria based on past trials. Despite some AI systems facing criticism for biases and lack
of standard evaluation techniques, advancements in explainable AI and regulatory frameworks pave the way for
more effective patient selection. By improving recruitment, stratification, and optimization processes, AI enhances
both the efficiency and accuracy of clinical trials, ultimately leading to better trial outcomes and more reliable
design.[19], [43], [44]
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
210
Discussion:-
Adaptive clinical trial designs (ADs) and artificial intelligence (AI) are revolutionizing clinical research by
enhancing efficiency, flexibility, and ethical standards. ADs offer advantages over traditional trials by allowing pre-
planned modifications based on interim results. This flexibility improves trial quality and can shorten duration.
Examples include adaptive dose-finding designs that optimize dose determination early and adaptive seamless
designs that merge trial phases to reduce overall sample sizes and transition times. Drop-the-losers and adaptive
platform trials focus resources on the most promising treatments, increasing relevance and efficiency. Ethical
considerations are addressed by increasing allocation to better-performing arms and enabling early trial termination
for safety issues or limited benefits. These designs ensure participants receive the most effective treatments,
improving patient outcomes and trust in clinical research. Additionally, by allowing modifications like re-estimating
sample sizes and adjusting treatment allocations, ADs enhance statistical power and the precision of treatment effect
estimates. Enrichment strategies focusing on subgroups likely benefit from the therapy further refine trial results.
Figure 2:- AI and ML enhancing adaptive Clinical trial designs.
This diagram illustrates the integration of Artificial Intelligence (AI) and Machine Learning (ML) in the adaptive
design of clinical trials (CTs). The adaptive design methods include Group Sequential Design, Adaptive
Randomization, Sample Size Re-estimation, and Adaptive Treatment Arms. AI and ML enhance these methods by
enabling more efficient CT designs, predicting outcomes, forecasting precision medicine, optimizingrandomization
and treatment arms, and improving patient selection. The flow chart visually represents how AI and ML support
each adaptive method and ultimately lead to enhanced clinical trial processes and outcomes, with the benefits
summarized in the yellow box.
Table 2:- Description of various clinical trial adaptive designs and AI usage.
Adaptive designs
Description
Enhancing Clinical
Trial Design with AI
AI enhances clinical trial design by simulating multiple rounds of data analysis and
improving statistical planning. It helps operationalize study design through
stratified cohort identification and optimizes designs involving multiple biomarkers
and treatments. AI can address generalizability and external validity issues using
methodologies like Markov chain Monte Carlo, decision rules, Bayesian analysis,
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
211
and novel AI techniques. It also identifies drug efficacy and adverse reactions,
integrating real-world variables through wearable technology and genome-driven
trials.
Prediction of Clinical
Trial Outcomes
AI and ML support early disease detection, predict molecular features, target
sensitivity, bioavailability, and toxicity, reducing late-stage trial failures. ML
models predict regulatory approval and success probabilities, aiding safer and more
effective trials. AI builds in-silico trials to simulate cohorts, identifying better
responders and reducing development failures. However, high-quality, curated
datasets are needed to realize AI's potential fully. AI and ML offer transformative
improvements in trial design, execution, and success rates.
Predicting Clinical
Outcomes in
Precision Medicine
AI predicts clinical outcomes, which is crucial for precision medicine. It optimizes
trial design by minimizing statistical variability. AI simulates data to identify
efficient statistical outcome measures and predict participant outcomes. It analyses
EMRs to predict clinical trial dropouts, allowing targeted interventions to retain
participants. ML models have significantly impacted cancer trials, reduced
mortality and enhancing drug selection. ML models incorporate multi-omic data,
transforming treatment paradigms and redefining precision trial design and
recruitment.
Randomization
Randomization in RCTs ensures treatment arms have similar patient characteristics.
Traditional coin flip randomization lacks control over allocations and covariate
adjustment. Stratified block randomization and covariate-adaptive designs address
these limitations, balancing multiple covariates simultaneously. Response-adaptive
designs modify allocation ratios based on ongoing trial success. AI models offer
the potential to replace traditional control arms with virtual controls, significantly
reducing recruitment needs. Validating AI models in independent populations
ensures accuracy.
Treatment Arm
Adaptive treatment arm selection improves efficiency by adjusting research
designs. Techniques include adaptive dose-finding, seamless, drop-the-losers, and
adaptive platform designs. AI can create external control arms, making trials more
patient-centric and efficient. Digital twins predict disease progression using
historical control data, enhancing confidence in treatment effects. This approach
reduces reliance on traditional control groups, increasing patient likelihood of
receiving treatment and speeding up enrollment. Rigorous guidelines are needed to
standardize synthetic patient generation.
Patient Selection
AI and ML systems enhance patient cohort composition and recruitment efficiency
by analyzing comprehensive datasets, identifying eligible and suitable patients,
predicting disease progression, and facilitating better trial matches without manual
reviews. Techniques like NLP and OCR automate data extraction, enhancing
patient profiling and selection. AI tools like Deep Six and Mendel.ai use NLP to
match patients to suitable trials. Despite challenges like biases and lack of standard
evaluation techniques, advancements in explainable AI and regulatory frameworks
are improving patient selection.
AI significantly enhances the trial design process by simulating multiple rounds of data analysis and improving
statistical planning. AI operationalises study designs through stratified cohort identification and optimises designs
involving multiple biomarkers and treatments, addressing the primary limitation of traditional trials: their
generalizability and external validity. AI and machine learning (ML) models improve early disease detection and
predict molecular features, bioavailability, and toxicity. These technologies reduce late-stage trial failures,
increasing the likelihood of Phase II/III trials advancing to regulatory approval. AI also predicts clinical trial
dropouts, enabling targeted interventions to retain participants and reduce sample sizes.
Furthermore, AI improves patient selection and recruitment by analysing clinical, genetic, and imaging data. Tools
like Deep Six and Mendel.ai use natural language processing (NLP) to match patients to suitable trials, enhancing
recruitment efficiency and outcomes. Electronic health records (EHRs) identify eligible patients and define criteria
based on past trials, facilitating better trial matches without manual reviews. While AI offers numerous benefits,
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
212
challenges such as dependency on well-labelled data, biases, and the "black-box" nature of AI models must be
addressed to maximise their potential.
Figure 3:- AI and ML application in adaptive CTs.
Figure 3 outlines the role of AI and ML in enhancing clinical trials (CTs) through adaptive designs. Starting from
designing clinical trials, the flowchart branches into adaptive designs and highlights limitations of current trials and
advanced methodologies, including AI for external validity. Critical areas of AI application include statistical
planning, drug efficacy, cost reduction, wearable tech integration, and genome-driven trials. These enhancements
contribute to early disease detection, predicting clinical outcomes, and reducing trial failures. The diagram
emphasises the comprehensive impact of AI and ML on improving the effectiveness and efficiency of clinical trials.
Unsuccessful randomized controlled trials (RCTs) often need help with poor patient selection, inadequate
randomization leading to confounding factors, insufficient sample sizes, and poor endpoint selection.[45]By
leveraging well-curated large datasets that include clinical and multimodal imaging data, AI models can be trained
to efficiently identify suitable study participants, predict their natural history using advanced statistical methods, and
assess study endpoints in a data-driven manner. These capabilities of AI offer the potential for more efficient
execution and greater statistical power compared to traditional RCTs, reducing the reliance on costly manual
reviews.[35]
Orphan drug approval and design pose unique challenges that can benefit significantly from adaptive clinical trial
designs (ADs) and artificial intelligence (AI). Orphan drugs that treat rare diseases often need help with limited
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
213
patient populations, complicating traditional trial methodologies. ADs and AI can provide tailored solutions to these
challenges, enhancing orphan drug trials' efficiency and success rates. The rejection of Sohonos by the European
Medicines Agency (EMA) is a pertinent example of the challenges faced in orphan drug approval.[46] Sohonos is
intended to treat fibrodysplasia ossificans progressive (FOP).[47] FOP is a rare and highly debilitating autosomal
dominant condition marked by frequent episodes of soft tissue pain, edema, and the growth of tumors in muscle and
connective tissue.[48] This drug faced significant hurdles due to limited clinical trial data and concerns about the
robustness of the evidence presented. The EMA's decision underscores the critical need for innovative trial designs
and analytical methods to overcome the inherent difficulties in developing treatments for rare diseases for which
incorporation of AI and adaptive trial designs may foster the drug approval process.
Conclusion:-
This review systematically examined the integration of artificial intelligence (AI) in adaptive trial designs within
clinical research. By rigorously identifying and analysing 45 relevant studies, it has been demonstrated that AI
significantly enhances the efficiency and effectiveness of clinical trials through advanced data analysis, adaptive
randomisation, and real-time decision-making. The findings highlight the potential of AI to streamline trial
processes, improve patient outcomes, and accelerate the development of new treatments. However, challenges such
as data quality, ethical considerations, and the need for regulatory frameworks remain. Future research should
address these challenges and explore the broader implementation of AI-driven adaptive trial designs across diverse
clinical settings. Overall, integrating AI in adaptive trial designs promises revolutionising clinical research, making
it more responsive, precise, and patient-centric.
References:-
1. Center for Drug Evaluation and Research. Adaptive Design Clinical Trials for Drugs and Biologics Guidance for
Industry. Available from: FDA. Accessed: Jun. 30, 2024.
2. Li Q, Lin J, Lin Y. Adaptive design implementation in confirmatory trials: methods, practical considerations and
case studies. Contemp Clin Trials 2020;98:106096.
3. Center for Drug Evaluation and Research. Enrichment Strategies for Clinical Trials to Support Approval of
Human Drugs and Biological Products. Available from: FDA. Accessed: Jun. 30, 2024.
4. Center for Biologics Evaluation and Research. Interacting with the FDA on Complex Innovative Trial Designs for
Drugs and Biological Products. Available from: FDA. Accessed: Jun. 30, 2024.
5. Chow SC, Chang M, Pong A. Statistical consideration of adaptive methods in clinical development. J Biopharm
Stat 2005;15(4):575-91.
6. Pallmann P, et al. Adaptive designs in clinical trials: why use them, and how to run and report them. BMC Med
2018;16(1):29.
7. Pre-hospital Assessment of the Role of Adrenaline: Measuring the Effectiveness of Drug Administration in
Cardiac arrest (PARAMEDIC-2): Trial protocol. Available from: PMC. Accessed: Jul. 01, 2024.
8. Burnett T, Jennison C. Adaptive enrichment trials: What are the benefits? Stat Med 2021;40(3):690-711.
9. Taves DR. The use of minimisation in clinical trials. Contemp Clin Trials 2010;31(2):180-4.
10. Power Priors and Their Use in Clinical Trials. Available from: The American Statistician. Accessed: Jul. 01,
2024.
11. Adaptive platform trials: definition, design, conduct and reporting considerations. Available from: PubMed.
Accessed: Jul. 01, 2024.
12. Adaptive Enrichment Designs in Clinical Trials. Available from: PMC. Accessed: Jul. 01, 2024.
13. Cuffe RL, Lawrence D, Stone A, Vandemeulebroecke M. When is a seamless study desirable? Case studies from
different pharmaceutical sponsors. Pharm Stat 2014;13(4):229-37.
14. Richter J, Friede T, Rahnenführer J. Improving adaptive seamless designs through Bayesian optimisation. Biom
J 2022;64(5):948-63.
15. Adaptive Dose-Finding Studies: A Review of Model-Guided Phase I Clinical Trials. Available from: PMC.
Accessed: Jul. 01, 2024.
16. Reddy S, Fox J, Purohit MP. Artificial intelligence-enabled healthcare delivery. J R Soc Med 2019;112(1):22-8.
17. Delso G, Cirillo D, Kaggie JD, Valencia A, Metser U, Veit-Haibach P. How to Design AI-Driven Clinical Trials
in Nuclear Medicine. Semin Nucl Med 2021;51(2):112-9.
18. Johnson KW, Glicksberg BS, Hodos RA, Shameer K, Dudley JT. Causal inference on electronic health records
to assess blood pressure treatment targets: an application of the parametric g formula. Pac Symp
Biocomput2018;23:180-91.
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
214
19. Krittanawong C, Johnson KW, Tang WW. How artificial intelligence could redefine clinical trials in
cardiovascular medicine: lessons learned from oncology. Pers Med 2019;16(2):83-8.
20. Zhavoronkov A, Vanhaelen Q, Oprea TI. Will Artificial Intelligence for Drug Discovery Impact Clinical
Pharmacology? Clin Pharmacol Ther 2020;107(4):780-5.
21. The Case for AI-Driven Cancer Clinical Trials – The Efficacy Arm In Silico. Available from: PMC. Accessed:
Jul. 01, 2024.
22. Siah KW, Khozin S, Wong CH, Lo AW. Machine-Learning and Stochastic Tumor Growth Models for
Predicting Outcomes in Patients With Advanced Non-Small-Cell Lung Cancer. JCO Clin Cancer Inform 2019;3:1-
11.
23. A Comparative Study on Machine Learning Algorithms for Predicting Breast Cancer Prognosis in Improving
Clinical Trials. Available from: IEEE Xplore. Accessed: Jul. 01, 2024.
24. Schperberg AV, Bouchard A, Tsigelny IF, Richard SB, Kurzrock R. Machine learning model to predict
oncologic outcomes for drugs in randomised clinical trials. Int J Cancer 2020;147(9):2537-49.
25. An Extended Minimization Method to Assure Similar Means of Continuous Prognostic Variables between
Treatment Groups. Available from: JSTAGE. Accessed: Jul. 01, 2024.
26. Endo A, Nagatani F, Hamada C, Yoshimura I. Minimization method for balancing continuous prognostic
variables between treatment and control groups using Kullback-Leibler divergence. Contemp Clin Trials
2006;27(5):420-31.
27. Pocock SJ, Simon R. Sequential treatment assignment with balancing for prognostic factors in the controlled
clinical trial. Biometrics 1975;31(1):103-15.
28. Taves DR. Minimization: a new method of assigning patients to treatment and control groups. Clin Pharmacol
Ther 1974;15(5):443-53.
29. Ciolino JD, Palac HL, Yang A, Vaca M, Belli HM. Ideal vs. real: a systematic review on handling covariates in
randomised controlled trials. BMC Med Res Methodol2019;19:136.
30. The Clinical Trial. Available from: NEJM. Accessed: Jul. 01, 2024.
31. Design of Experiments with R. Available from: SpringerLink. Accessed: Jul. 01, 2024.
32. Korn EL, Freidlin B. Time trends with response-adaptive randomisation: The inevitability of inefficiency. Clin
Trials Lond Engl 2022;19(2):158-61.
33. Rosenberger WF, Sverdlov O, Hu F. Adaptive randomisation for clinical trials. J Biopharm Stat 2012;22(4):719-
36.
34. Forecasting future Humphrey Visual Fields using deep learning. Available from: PMC. Accessed: Jul. 01, 2024.
35. Lee CS, Lee AY. How Artificial Intelligence Can Transform Randomized Controlled Trials. Transl Vis Sci
Technol 2020;9(2):9.
36. Neal D, Casella G, Yang MCK, Wu SS. Interval estimation in two-stage, drop-the-losers clinical trials with
flexible treatment selection. Stat Med 2011;30(23):2804-14.
37. Lin J, Bunn V. Comparison of multi-arm multi-stage design and adaptive randomisation in platform clinical
trials. Contemp Clin Trials 2017;54:48-59.
38. Adaptive Seamless Phase II/III Designs—Background, Operational Aspects, and Examples. Available from:
SAGE Journals. Accessed: Jul. 01, 2024.
39. Unlearn Closes $50 Million Series B Funding to Advance the Use of Its Machine Learning-Powered
TwinRCTs™ in Clinical Trials. Available from: Business Wire.
40. Zhang B, Zhang L, Chen Q, Jin Z, Liu S, Zhang S. Harnessing artificial intelligence to improve clinical trial
design. Commun Med 2023;3(1):1-3.
41. Miotto R, Weng C. Case-based reasoning using electronic health records efficiently identifies eligible patients
for clinical trials. J Am Med Inform Assoc 2015;22(e1):e141-50.
42. Harrer S, Shah P, Antony B, Hu J. Artificial Intelligence for Clinical Trial Design. Trends Pharmacol Sci
2019;40(8):577-91.
43. Glicksberg BS, et al. Automated disease cohort selection using word embeddings from Electronic Health
Records. Pac Symp Biocomput2018;23:145-56.
44. Mendel.ai nabs $2 million to match cancer patients with the latest clinical trials | TechCrunch. Available from:
https://techcrunch.com/2017/07/01/mendel-ai-nabs-2-million-to-match-cancer-patients-with-the-latest-clinical-
trials/. Accessed: Jul. 01, 2024.
45. Challenging issues in randomised controlled trials - PubMed. Available from:
https://pubmed.ncbi.nlm.nih.gov/20413119/. Accessed: Jul. 01, 2024.
46. Sohonos | European Medicines Agency. Available from:
https://www.ema.europa.eu/en/medicines/human/EPAR/sohonos. Accessed: Jul. 01, 2024.
ISSN: 2320-5407 Int. J. Adv. Res. 12(08), 205-215
215
47. Pignolo RJ, et al. Reduction of New Heterotopic Ossification (HO) in the Open-Label, Phase 3 MOVE Trial of
Palovarotene for Fibrodysplasia Ossificans Progressiva (FOP). J Bone Miner Res 2023;38(3):381-94.
48. Viswakanth M. REVIEW ARTICLE ON FIBRODYSPLASIA OSSIFICANS PROGRESSIVA. World J Pharm
Pharm Sci 2018;7(6):123-34.
