By Poonam Balani
Healthcare data and technology are intertwined and the regulatory framework governing the medicines regulation is looking to update the regulations to bring them in tune with the use of new technologies and access to new platforms for handling healthcare data in real time. The enormous amount of data captured across multiple channels and devices offers an opportunity to use this data for individual characterisation of diseases, develop personalised treatments and improve the performance of medicinal products in individual healthcare systems. Artificial intelligence (AI) adoption in the pharmaceutical industry holds the promise of accelerating the drug development and, at the same time, decrease the cost of those drugs and devices to consumers. AI also holds the potential to navigate the ever-changing regulatory landscape. Technologies such as AI and Machine Learning (ML) enable capturing of data in different data types and forms, outside of a traditional health care context, known as real-world data (RWD), and provide valuable evidence that can be transformed using AI algorithms into meaningful data that can be used as real-world evidence (RWE) in the healthcare industry.
Most healthcare organisations do not have the capacity to collect the data necessary for optimal training of algorithms to (a) “fit” the specific population and/or the local practice patterns, and (b) to be able to question the bias to guarantee that the algorithms perform consistently, thereby making them reliant on big IT firms for the processing, handling and storage of sensitive data.
The progress and acceptance of AI for use in healthcare industry raises fundamental questions regarding the ownership of the healthcare data, privacy, confidentiality, data security, and informed consent, which need to be addressed, which can be achieved only through public discourse and policy intervention. Recent developments in the past years such as the implementation of data privacy regulations like the EU General Data Protection Regulation (GDPR)1 or California’s Consumer Privacy Act2, while having good intentions, may be restricting access to patient health data, which is the fundamental requirement to building AI algorithms and Machine Learning (ML) platforms. Other initiatives like the one to give the patients access to their healthcare data, including new proposals from the Center for Medicare and Medicaid Services3 can ease the requirement for having consent to their data for developing AI solutions. While these considerations apply to resource-poor settings as well, their relevance largely depends on differences in culture, literacy, patient–provider relationships, available IT infrastructure and regulatory issues. Although there have been some initial efforts to address these issues, including the Data Sharing Principles in Developing Countries put forward in Nairobi in 2014, widespread adoption lags behind."The key principle for the data sharing principle is that data generated with public funds should be viewed as a public good and should be made freely accessible."
Furthermore, Many AI applications depend on the availability of strong electronic health record (EHR) systems, which require substantial investment and they may have only a limited impact if the EHRs are not integrated with the local language and scripts used across countries. With respect to the developing countries and other resource-poor nations, it may be difficult to gather such datasets due to lack of infrastructure required for collection and storage of large amounts of data [1].
Global pharmaceutical industry must abide by a number of different regulations across the world regions, covering every aspect of the industry from development to commercialisation so as to ensure the safety and efficacy of new medicines and medical devices. One of the important areas of AI use in the pharmaceutical industry is pharmacovigilance, where already approved drugs are monitored post-approval to collect patient-specific data on drug safety and/or efficacy in real-world settings. This data is then reported to the regulatory authorities. Traditionally, this task of collecting and collating patient experiences with the drugs has been performed manually and include inputs from the physicians, scientists, and other healthcare providers.
The data is then interpreted regarding the adverse reactions after careful review on the part of experts. AI provides the possibility of streamlining this real-world data to ensure accuracy in reporting, answering other safety-related issues concerning disease characterization and prevalence, to better understand the actual standard of care, and to validate the clinical outcomes of short-term surrogate markers.4 An AI pharmacovigilance system can receive data files in both structured and unstructured formats (emails, medical records from doctors’ offices and hospitals), and social media feeds, filter the data according to relevance, and even identify errors before reporting to the regulatory authorities.
The use of RWD collection devices (smart devices) and AI platforms enabling data exchange and processing can help improve the design of clinical trials in real time and offer better efficiencies. It is expected that the use of RWD will make clinical trials adaptable to react to drug-safety signals identified in specific population subsets. These measures can further allow for dynamic estimation of sample sizes in clinical trials and may facilitate the required changes in protocol to adapt to responses emerging from the clinical data in real time. AI in health care has the ability to automatically identify similarities using patients’ medical records, and therefore, can be used to support researchers in identifying an optimal patient cohort for a specific clinical trial [2]. Furthermore, use of genetic aspects of an individual patient while recruiting, AI systems can take into account specific populations for enrolment to smaller clinical trials. These practices can reduce the clinical trial costs and can enable clinical trials for a small targeted population set, for example in case of orphan indications and other rare diseases.
Consumer technology and the use of AI in data processing is enabling prospective studies, through the use of wearables. As an example, the ongoing prospective, single arm study to detect atrial fibrillation in more than 4,19,093 consenting Apple watch owners [3]. The study is conducted virtually, with screening, consent and data collection performed electronically using the smartphone app. Study visits are performed by telehealth study physicians via video chat through the app, and ambulatory ECG patches are mailed to the participants.
Safety monitoring is moving beyond traditional approaches to use sophisticated AI algorithms that identify safety signals arising from rare adverse events. Furthermore, these signals could be captured from a variety of sources like Websites and search engines. Other sources can include electronic medical records, and consumer-generated media, which can be identified in real-time to identify early signals regarding safety issues of pharmaceutical products. A prompt and timely response on the part of the pharmaceutical manufacturer to physician and patient concerns could prevent regulatory and public-relations backlashes.
The current focus on managing regulatory information is to change the structure from silo systems into a single integrated platform. This will enable simplification of the system while granting higher flexibility to be operable across functionalities. AI can also help in bringing a holistic view of regulatory information requirements right from the discovery to product launch. It is then expected to fulfil the compliance aspects and ensure optimal operational efficiency.
Standardization for identification of medicinal products (IDMP), clinical trial registration initiative (CTRI), electronic common technical document (eCTD), and fast health information resources (FHIR) are some modes for simplification and automation of repetitive, time-consuming, and quality intensive processes, can enable accurate capturing of events, triggering action, down to automating the entire processes via management of workflows, notifications and record updates can be streamlined using various automation technologies like cognitive and deep machine learning (NLP, ML, and & AI), hence transforming regulatory operational efficiency , quality and compliance.
Use of Advanced Analytics: embracing AI-based analytics has enabled the pharma industry to look for insights to build global regulatory strategies for submission approval, enable decisions on early drug launch timing, and facilitate well thought regulatory decisions.
Business Intelligence: for operational reporting needs, visibility across regulatory functions, interdependent processes, and increased interactions with internal and external stakeholders is of utmost importance. AI can enable business intelligence capabilities across regulatory functions, organisations to achieve better operational efficiency which relies upon accurate analytics and insights.
Collaboration: cloud-based collaborative platforms with automated workflows enable better collaboration with partners/ stakeholders like CROs, software vendors and affiliates as new working models using interfaces with regulatory internal and external stakeholders gain importance.
RWE has been shown to have an important role in understanding drug efficacy and the identification of adverse events based on differences of metabolism in various racial and genetic groups. The use of new technologies allows the sponsors to conduct decentralised and virtual clinical trials. As an example, Sanofi conducted a clinical trial which previously required participants regular visits to the trial site for the participants’ weight, blood pressure, and blood glucose, by giving them connected sensors and wireless technology to record and share this data from their homes5. AOBiome Therapeutics completed a 12-week clinical trial on a new acne drug, which was proven to be safe and effective in patients who participated in the trial while at home.
The sponsor AOBiome mailed either the drug or a placebo, along with a smartphone equipped with a pre-loaded application for the participants the trial to share regular photographs of their acne and communicate with study organisers throughout the clinical trial.6 Similarly, GlaxoSmithKline has conducted a feasibility study to use a smartphone and an application to record survey data from rheumatoid arthritis patients. Furthermore, the phone’s accelerometer and was a more accurate measure than the motion-evaluation exercises performed in-person in the presence of the physician [4].
While RWE promises to unlock significant value across the drug development life cycle, realising its potential relies on the quality of the data under consideration. RWD is valuable only if it is of a sufficient quality and is collected in a standardised manner allowing it to be portable and reusable in different contexts. Generating large datasets alone does not carry as much value as achieving meaningful insights from combination of different datasets collected over a period of time (longitudinal datasets). On the other hand, commercial interests can be an important hurdle in data sharing between pharmaceutical industry and the regulators, restrict access to important information.
To expand the use of data beyond clinical trials, pharmaceutical companies are also creating proprietary data networks to gather, analyse, share, and respond to real-world outcomes and claims data. It is important to realise that partnerships with insurers, healthcare providers, and other institutions are critical for the success of these efforts.
The United States (US) The 21st Century Cures Act passed by the United States Congress mandates the US Food and Drug Administration (FDA) to develop guidance to evaluate the use of real-world evidence (RWE) to support the regulatory process. The FDA has invested in the evaluation of electronic health records through the Sentinel Initiative, which is a legally mandated electronic-surveillance system that collates and performs analysis of the health-care data from multiple sources. The US Congress has further mandated that the FDA increase focus on RWE for regulatory decision making. In 2018, the FDA published the framework for its Real-World Evidence Program, providing guidance on the incorporation of RWD into clinical trials to generate meaningful RWE. Specifically, the Sentinel System is intended to help define, test, and shape regulatory-grade data, methods, and analytical standards for the drug development process. For example, while the Sentinel System could help link disease registries that have longitudinal data about patient populations with other RWD sources, the registries could become more effective at engaging patients for clinical trials [6]. FDA also plans to work with the pharmaceutical industry by establishing partnerships to use RWD in regulatory decision making, including using synthetic control arms for clinical trials.7
The European Union (EU) The European Medicines Agency (EMA) defines RWD as defined as “routinely collected data relating to a patient's health status or the delivery of health care from a variety of sources other than traditional clinical trials” and has been keen on using RWD regulatory decision making. The ‘Adaptive Pathways’ initiative aims to use current European regulatory framework to speed up patient access to medicines by encouraging the use of principles such as iterative development/approvals, RWD as a supplement to the clinical trial data, and the use of RWE by the health- technology-assessment bodies. The EMA is routinely uses three real world databases for in-house studies and has commissioned 15 external studies. It recognises that RWD is underutilised to assess the public health impact of risk minimisation measures, health technology assessment, and to support the pricing and reimbursement decisions. The EMA recently accepted an RWE- based control arm during their analysis of Alecensa effectiveness compared with the standard of care [7].
Japan The Japanese Pharmaceuticals and Medical Devices Agency (PMDA) has updated its regulations and published guidelines encouraging the use of RWD to satisfy post-safety requirements. The PMDA has shown a willingness to expand its application to regulatory assessments other than safety [8] by initiating the MID-NET® project for the real-time assessment of drug safety, which was launched in April 2018 and currently enrols more than 4 million patients. The PMDA has coordinated the project cooperation with partner hospitals to establish a high quality database, (remote participation), development of novel endpoints, and easier incorporation of patient reported outcomes [9].
While the use of AI has generated lot of possibilities to utilise data, it is important to realise that RWE is not generated with a particular study question in mind but generally for clinical care and billing purposes. Appropriate use of RWE has to be supported by well-designed guidelines and regulations so as to ensure the quality of data collected that is based on accurate, unbiased findings.
The industry is enthusiastic to use observational RWD as a substitute for well-conducted clinical trials, but the data generated should be used with caution as the RWD is unable to compare outcomes of nonrandomised groups. Establishing the IT capability remains the single most challenging aspect in order to generate the capacity to receive, manage and analyse datasets to discover insights on the safety, efficacy and use of medicines and explore the validity of claims made by the industry.
When used appropriately, RWE has the potential to supplement traditional clinical research to aid therapeutic development, clinical decision making and efficiency gains in healthcare, while improving the access to underserved populations. On the other hand, use of compromised data in an incorrect fashion can result in spurious approvals, and loss of valuable resources and time while harming the patients and decreasing overall efficiencies of the healthcare system.
1. Wahl, B., et al., Artificial intelligence (AI) and global health: how can AI contribute to health in resource-poor settings? BMJ global health, 2018. 3(4): p. e000798-e000798. 2. Sharafoddini, A., J.A. Dubin, and J. Lee, Patient Similarity in Prediction Models Based on Health Data: A Scoping Review. JMIR Med Inform, 2017. 5(1): p. e7. 3. Turakhia, M.P., et al., Rationale and design of a large-scale, app-based study to identify cardiac arrhythmias using a smartwatch: The Apple Heart Study. Am Heart J, 2019. 207: p. 66-75. 4. Crouthamel, M., et al., Using a ResearchKit Smartphone App to Collect Rheumatoid Arthritis Symptoms From Real-World Participants: Feasibility Study. JMIR Mhealth Uhealth, 2018. 6(9): p. e177. 5. Iwai, Y., et al., Cancer immunotherapies targeting the PD-1 signaling pathway. Journal of Biomedical Science, 2017. 24(1): p. 26. 6. New, J., The Promise of Data-Driven Drug Development C.f.D. Innovation, Editor. 2019. 7. Davies, J., et al., Comparative effectiveness from a single-arm trial and real-world data: alectinib versus ceritinib. J Comp Eff Res, 2018. 7(9): p. 855-865. 8. Andre, E., et al., Trial designs using real-world data: The changing landscape of the regulatory approval process. Pharmacoepidemiology and Drug Safety, 2019. 9. EMA, EMA Regulatory Science to 2025 - Strategic Reflection. 2018.
Poonam Balani has a PhD in biomedical sciences with 14 years of experience in medical and academic writing.