“Artificial intelligence”: Which services, which applications, which results and which development today in clinical research? Which impact on the quality of care? Which recommendations?

doi:10.1016/j.therap.2018.12.003

Therapies

Volume 74, Issue 1, February 2019, Pages 155-164

https://doi.org/10.1016/j.therap.2018.12.003 Get rights and content

Summary

Artificial intelligence (AI), beyond the concrete applications that have already become part of our daily lives, makes it possible to process numerous and heterogeneous data and knowledge, and to understand potentially complex and abstract rules in a manner human intelligence can but without human intervention. AI combines two properties, self-learning by the successive and repetitive processing of data as well as the capacity to adapt, that is to say the possibility for a scripted program to deal with multiple situations likely to vary over time. Roundtable experts confirmed the potential contribution and theoretical benefit of AI in clinical research and in improving the efficiency of patient care. Experts also measured, as is the case for any new process that people need to get accustomed to, its impact on practices and mindset. To maximize the benefits of AI, four critical points have been identified. The careful consideration of these four points conditions the technical integration and the appropriation by all actors of the life science spectrum: researchers, regulators, drug developers, care establishments, medical practitioners and, above all, patients and the civil society. 1st critical point: produce tangible demonstrations of the contributions of AI in clinical research by quantifying its benefits. 2nd critical point: build trust to foster dissemination and acceptability of AI in healthcare thanks to an adapted regulatory framework. 3rd critical point: ensure the availability of technical skills, which implies an investment in training, the attractiveness of the health sector relative to tech-heavy sectors and the development of ergonomic data collection tools for all health operators. 4th critical point: organize a system of governance for a distributed and secure model at the national level to aggregate the information and services existing at the local level. Thirty-seven concrete recommendations have been formulated which should pave the way for a widespread adoption of AI in clinical research. In this context, the French “Health data hub” initiative constitutes an ideal opportunity.

Section snippets

Abbreviations

artificial intellligence

AMM

marketing authorization

ANR

Agence nationale de la recherche

ANSM

Agence nationale de sécurité du médicament et des produits de santé

CFDA

China food and drug administration

CNAMTS

Caisse nationale d’Assurance maladie des travailleurs salariés

CNIL

French commission nationale de l’informatique et des libertés

CPTS

territorial professional communities of health

DGOS

Direction générale de l’offre de soins

DMP

shared medical record

EHESP

École des hautes études en santé publique

EMA

1st critical point: artificial intelligence and clinical research, from promises to demonstration: what evidence?

The failure rate in clinical research is very high. Less than 10% (5,1% in cancerolgy) of drug candidates reach the market [1]. In phase III alone and a sample of 640 drug candidates, the failure rate is 54% according to a study published in 2016 [2]. While modeling and simulation is common in other sectors, e.g. automobile and aerospace industries, it is not yet prevalent in therapeutic research and development.

For medical devices, the failure rate is likely lower because “patient-device”

2nd critical point: create confidence to ensure the dissemination and acceptability of AI in health research, in a suitable regulatory context

Confidence in these new approaches will be built through information, the definition of a number legal and technical rules and safeguards, and due consideration for ethical issues.

Inform to enlighten and reassure public opinion: if “fear arises sooner than anything else” (Notebooks/Leonardo da Vinci), we know those ignited by artificial intelligence:

•
the AI is a black box where we do not know what is happening;
•
doubts about the reliability and robustness of the results (“rubbish in/rubbish out”);
•

3rd critical point: “Know-how and ability” – Availability of technical skills and ergonomic tools for data collection

Many preconceived ideas circulate about AI: today we are close to the saturation of available human resources and the shortage of skills and experts; there would be a delay in the adaptation of the educational system, in the establishment of quality vocational training for these new jobs. More specifically in healthcare, the sector lacks attractiveness, would face stark competition from the US tech giants (Google, Apple, Facebook, Amazon [GAFAM]), who already collaborate with “big-pharmas”, as

Organizational transformations to realize

The organizational impact of AI in clinical research has not been modeled. For healthcare companies, hospitals and research and development organizations, which will be the main users of artificial intelligence, this impact is far from being neutral. It has not yet been thought, theorized and anticipated.

The integration of AI in the R&D departments of pharmaceutical companies

“Big-pharmas” are multinational entities which have grown in size are complexity over the past 20 years, even if their plasticity has been put to the test by successive waves of mergers &

Conclusion

“Artificial intelligence will not replace doctors. But the doctors who will use AI will replace those who will not do it”. This is the catchy title of a recent article by Xavier Comtesse, mathematician and creator of several start-ups, and Daniel Walch, director of groupement hospitalier de l’Ouest lémanique (GHOL) [18]. Because the march of innovation is irresistible and does not wait, if France misses the train of AI, it will be relegated definitively, tomorrow, to supporting roles. In

Disclosure of interest

The authors declare that they have no competing interest.

References (18)

J.F. Dhainaut et al.
Utilisation des objets connectés en recherche clinique
Therapie
(2018)
D.W. Thomas et al.
Clinical development success rates 2006–2045
(2016)
T.J. Hwang et al.
Failure of investigational drugs in late-stage clinical development and publication of trial results
JAMA Intern Med
(2016)
Pharnext
Pharnext annonce les résultats positifs de l’essai pivot de Phase 3 de PXT3003 pour le traitement de la maladie de Charcot
(2016)
J. Simm et al.
Repurposing high-throughput image assays enables biological activity prediction for drug discovery
Cell Chem Biol
(2018)
E. Kogan et al.
Use of machine learning to determine stroke severity of patients diagnosed with stroke in integrated claims-medical records dataset
Circulation
(2018)
Avicenna Alliance
Asscation for predictive medicine
Conference report. International Avicenna Alliance conference report. European Parliament (Brussels)
(2018)
J.C. Rocha et al.
Using artificial intelligence to improve the evaluation of hyman blastocyst morphology. ESHRS congress, Genova, Switzerland
(2017)
C. Sonigo et al.
High-through put ovarian follicle counting by an innovative deep learning approach
Sci Rep
(2018)

There are more references available in the full text version of this article.

Cited by (9)

What place for intelligent automation and artificial intelligence to preserve and strengthen vigilance expertise in the face of increasing declarations?
2023, Therapies
Les Ateliers de Giens avaient consacré en 2018 une table ronde à l’intelligence artificielle (IA) et conduit ses experts à confirmer l’apport potentiel et le bénéfice théorique que représente l’IA en recherche clinique et en matière d’amélioration de l’efficience de soins. La table ronde de 2022 s’inscrit dans la continuité de cette réflexion sur l’IA et l’automatisation intelligente (AI) en la ciblant sur son apport en pharmacovigilance et les applications et tâches qu’elles pourraient optimiser afin de préserver et de renforcer l’expertise médicale et pharmacologique en pharmacovigilance. En effet, l’évolution du travail en pharmacovigilance est caractérisée par de nombreuses tâches à faible valeur ajoutée, une volumétrie croissante des déclarations de pharmacovigilance, la rareté des personnels médicaux ayant une expertise en pharmacologique clinique et en pharmacovigilance et des ressources humaines qui ne sont pas et ne seront pas en adéquation avec l’évolution constatée. Tous ces paramètres concourent à une embolisation du système de pharmacovigilance avec le risque de passer à côté de sa mission première qui est d’identifier et de caractériser un risque, voire une alerte sanitaire, sur un médicament. Les participants de la table ronde (représentants des centres régionaux de pharmacovigilance (CRPV), de l’Agence nationale de sécurité du médicament et des produits de santé (ANSM), des patients, de l’industrie du médicament ou encore de start-ups travaillant dans le développement de l’IA dans le domaine du médicament) ont partagé leurs expériences, leurs projets pilotes et leurs attentes sur les potentialités attendues, théoriques ou avérées, de l’IA et l’AI. Ce travail a permis de dégager les besoins et les enjeux réels que représentent, ou pas, l’IA ou l’AI dans les modes d’organisation actuelles ou futures de la pharmacovigilance. Cette démarche a permis l’élaboration d’une matrice SWOT (forces, faiblesses, opportunités, menaces), socle de réflexion pour identifier des points critiques et envisager quatre principales recommandations : (1) préserver et développer l’expertise métier en pharmacovigilance (incluant la partie recherche et développement en méthodes) avec l’intégration de nouvelles technologies ; (2) améliorer la qualité des déclarations en pharmacovigilance ; (3) adapter les moyens techniques et réglementaires ; (4) implémenter une stratégie de développements des outils AI et IA au service de l’expertise.
What place for intelligent automation and artificial intelligence to preserve and strengthen vigilance expertise in the face of increasing declarations?
2023, Therapies
Citation Excerpt :
In 2018, the Ateliers de Giens devoted a workshop to artificial intelligence (AI) and led its experts to confirm the potential contribution and theoretical benefit of AI in clinical research, pharmacovigilance, and in improving the efficiency of care [1].
In 2018, the “Ateliers de Giens” (Giens Workshops) devoted a workshop to artificial intelligence (AI) and led its experts to confirm the potential contribution and theoretical benefit of AI in clinical research, pharmacovigilance, and in improving the efficiency of care. The 2022 workshop is a continuation of this reflection on AI and intelligent automation (IA) by focusing on its contribution to pharmacovigilance and the applications and tasks could be optimized to preserve and strengthen medical and pharmacological expertise in pharmacovigilance. The evolution of pharmacovigilance work is characterized by many tasks with low added value, a growing volume of pharmacovigilance reporting of suspected side effects, and a scarcity of medical staff with expertise in clinical pharmacology and pharmacovigilance and human resources to support this growing need. Together, these parameters contribute to an embolization of the pharmacovigilance system at risk of missing its primary mission: to identify and characterize a risk or even a health alert on a drug. The participants of the workshop (representatives of the Regional Pharmacovigilance Centres (CRPV), the French National Agency for Safety of Medicinal Products (ANSM), patients, the pharmaceutical industry, or start-ups working in the development of AI in the field of medicine) shared their experiences, their pilot projects and their expectations on the expected potential, theoretical or proven, AI and IA. This work has made it possible to identify the needs and challenges that AI or IA represent, in the current or future modes of organization of pharmacovigilance activities. This approach led to the development of a SWOT matrix (strengths, weaknesses, opportunities, threats), a basis for reflection to identify critical points and consider four main recommendations: (1) preserve and develop business expertise in pharmacovigilance (including research and development in methods) with the integration of new technologies; (2) improve the quality of pharmacovigilance reports; (3) adapt technical and regulatory means; (4) implement a development strategy for AI and IA tools at the service of expertise.
Machine learning algorithms in the environmental corrosion evaluation of reinforced concrete structures - A review
2022, Cement and Concrete Composites
Citation Excerpt :
Due to the shortcomings and challenges of traditional methods in predicting corrosion process of RC structures, a new calculation method named machine learning (ML) algorithm has been gradually introduced in recent years. ML algorithms are a technology that studies the theory and method of computer simulating human relationship and activities, acquiring knowledge and skills, and enabling it the capability to take advantage of huge amounts of descriptors, i.e. influencing factors [30–34]. The earliest application of ML algorithms in the field of structure, thirty years ago, contained testing of diverse famous techniques on simple issues (e.g. structural analysis of steel components [35], design of control tools with reliable management structure [36]).
Accurate corrosion assessment of reinforced concrete (RC) structures is expected to improve the service life and durability of structures. However, traditional evaluation methods rely on simple regression and assumption models, which are easy to lead to unreliable evaluation results. The time-consuming and complex calculations in corrosion assessment are particularly suitable for machine learning (ML) and have already been deeply affected by the application of existing ML algorithms. The review analyzes recent ML methods for corrosion assessment of RC structures. These algorithms have recently had a significant impact on the estimation of the corrosion process, significant mechanical properties and durability of RC structures. In addition, some challenges that have emerged in corrosion evaluation and could be solved by ML algorithm are discussed critically. Through the detailed analysis of the challenges and future directions, researchers and engineers related industry will gain vital insight on the sustainable durability design of RC structures.
Artificial intelligence application in healthcare management: a review of challenges
2024, International Journal of Healthcare Technology and Management
Prediction of Sulfate Ion Penetration in Concrete Containing Nano-Silica and Micro-Silica Using Machine Learning
2023, SSRN
A review of trust in artificial intelligence: Challenges, vulnerabilities and future directions
2021, Proceedings of the Annual Hawaii International Conference on System Sciences

View all citing articles on Scopus

^☆: Articles, analyses and proposals from Giens workshops are those of the authors and do not prejudice the proposition of their parent organization.

View full text

Giens Workshops 2018Clinical research“Artificial intelligence”: Which services, which applications, which results and which development today in clinical research? Which impact on the quality of care? Which recommendations?☆

Summary

Section snippets

Abbreviations

1st critical point: artificial intelligence and clinical research, from promises to demonstration: what evidence?

2nd critical point: create confidence to ensure the dissemination and acceptability of AI in health research, in a suitable regulatory context

3rd critical point: “Know-how and ability” – Availability of technical skills and ergonomic tools for data collection

Organizational transformations to realize

The integration of AI in the R&D departments of pharmaceutical companies

Conclusion

Disclosure of interest

Therapie

Clinical development success rates 2006–2045

Failure of investigational drugs in late-stage clinical development and publication of trial results

JAMA Intern Med

Pharnext annonce les résultats positifs de l’essai pivot de Phase 3 de PXT3003 pour le traitement de la maladie de Charcot

Repurposing high-throughput image assays enables biological activity prediction for drug discovery

Cell Chem Biol

Use of machine learning to determine stroke severity of patients diagnosed with stroke in integrated claims-medical records dataset

Circulation

Asscation for predictive medicine

Conference report. International Avicenna Alliance conference report. European Parliament (Brussels)

Using artificial intelligence to improve the evaluation of hyman blastocyst morphology. ESHRS congress, Genova, Switzerland

High-through put ovarian follicle counting by an innovative deep learning approach

Sci Rep

Giens Workshops 2018
Clinical research
“Artificial intelligence”: Which services, which applications, which results and which development today in clinical research? Which impact on the quality of care? Which recommendations?☆