Going Beyond Generative AI And Templates To Create Informed Consent Documents

By Kamila Novak, KAN Consulting

A voluntarily signed informed consent document (ICD) became the prerequisite for study participation in 1948 after the Nuremberg trials and is regarded as one of the pillars of research ethics acknowledging the autonomy of individuals.

We all know that ICDs should be easy to understand and written in plain language with terms explained. Nevertheless, today’s ICDs are excessively lengthy, complex, and difficult to read. Despite the notion that participants understand what they sign, literature brings evidence to the contrary.^1-3

The concepts best understood by participants include voluntary participation, blinding (except for investigators’ blinding), and freedom to withdraw at any time. Per the literature, placebo concepts, randomization, safety issues, risks, and side effects are among the least understood.

During audits, I noticed another area with poor understanding — the purpose of various examinations, especially blood draws, except for blood count and routine blood chemistry. Since safety and efficacy are the main objectives in Phase 2-3 clinical trials, poor understanding increases the risks of missing data and compromised trial results. We often observe many skipped visits, blood draw refusals, and low AE/SAE reporting by participants.

Most ICDs are still on paper, with no information visually attractive, such as:

• Pictures
• Diagrams or flowcharts
• Tables
• Bullet points
• Bold or color fonts

Finding important pieces of information is challenging. Participants’ immediate understanding may be checked using a quiz; however, continuing understanding is rarely reinforced in subsequent re-education.

Even some electronic ICDs follow the typical paper ICD format. They look the same; the only difference is that they are accessible electronically and enable electronic signatures. Honestly, I prefer paper format in this case since I can flip back easily and put pages side by side if I want to compare something.

The poor ICD design by sponsors and omitted re-education of participants by site personnel are among the factors contributing to missing data and risks to participants’ safety. A year ago, I published an article about informed consent⁴ where I provided specific examples of problematic ICDs as well as a few thoughts about possible ways for improvements.

In this article, I focus on ICD designs, our typical templates, and the recent FDA Guidance for Industry Key Information and Facilitating Understanding in Informed Consent, released as a draft for public consultation in March 2024.⁵

Let us think about the templates. We love templates because we believe they help standardization and save time, thus improving efficiency. We tend to forget the dark side of templates — they promulgate imitation and limit innovation.

Who Develops Today’s ICDs And How?

Clinical operations typically customize standard ICD templates that have been recycled from one study to another. Under pressure to produce the master ICD as soon as possible, ClinOps prioritizes speed and thus minimizes room for critical thinking, common sense, and a thorough review.

Templates are rarely reviewed carefully. Usually, they are patched with additional text due to new or revised regulations, e.g., related to privacy and confidentiality. Except for regulatory affairs and legal departments, nobody else is consulted and nobody else attempts to associate the ICD’s user-friendliness with issues from previous studies.

Customized master ICDs undergo quality control by ClinOps. However, the reviewers focus on copyediting (spelling, grammar, and punctuation) and the inclusion of all study-specific information, but readability and effective delivery of information are rarely part of the review process.

The results are petrified, subpar ICDs propagating from study to study.

How Should Future ICDs Be Developed?

The FDA has issued guidance on facilitating understanding in informed consent, which has the potential to encourage innovation and trigger changes needed in the design of ICDs. The key points are logical and straightforward:

1. Start with key information and make it stand out.
2. Provide supplemental information if needed.
3. Facilitate understanding using visual prompts (e.g., bubbles), logical flow of information, and plain language. Consider using a glossary of terms.

The key information includes:

• Voluntary participation and right to discontinue participation
• Purpose of the research, expected duration, and procedures to be followed
• Reasonably foreseeable risks and discomforts
• Reasonably expected benefits
• Appropriate alternative procedures
• Compensation and medical treatments for research-related injuries
• Costs related to participation

At the end, the guidance provides an example from a hypothetical clinical trial.

It sounds easy. Nevertheless, we will encounter challenges, such as including information related to privacy and confidentiality, sample and/or data retention for future research, explaining all alternative therapies, etc. These all significantly contribute to the current ICD length. We usually include large chunks of complex text, but we don’t have to. We can use focused appendices and use the key-principle approach described above to convey important information.

About eight years ago, I attempted to write an ICD implementing some of these recommendations. I expected the client to reject the draft and I was surprised to see an easy acceptance. Perhaps it was our notion that these changes would not be accepted, either by clients or by regulators, that has been holding us back. Now, with regulatory guidance backing up innovative approaches, we will get more courageous.

Who Should Develop ICDs And Why?

Our team should be colored by the diverse representatives of different functional teams and stakeholders.

1. Project manager: The project manager leads the team, controls timelines, and keeps the ICD development on track.
2. Patient representative and/or investigator: We must understand our patient population. The patients are the consumers of the ICD; we should develop it to meet their need for information and bridge any possible limitations to their understanding. It may be challenging to include a patient representative; however, the investigator can be a suitable substitute. Investigators know their patients well and will be consenting them. Per the U.S. federal regulations, the investigator is responsible for writing study protocols and informed consent, not the CRO or the sponsor, hence, their seat at the table is justified. Yet, it is very rarely done. They are seen as mere reviewers, sometimes not even that.
3. Clinical operations and sponsor’s medical expert: These people know the study and its context, can explain reasons for performing specific procedures, and have access to needed information. They perform quality control of the document before it is internally approved for submission to IRBs and regulatory authorities.
4. Regulatory affairs: These specialists advise on predicate rules for regulatory filing and local regulations.
5. Data protection officer: This expert advises on privacy and confidentiality requirements and which rights are suspended in research.
6. Medical writer: They convert complex medical and legal information into plain language that can be understood by potential study participants. These SMEs know readability rules and should be aware of glossaries of terms released by the Council for International Organizations of Medical Sciences (CIOMS) and the FDA in collaboration with the NIH.^{6, 7} They write the document.
7. Graphic designer: Hand in hand with medical writers, they help visualize key messages to make them stand out and be easy to understand and remember.
8. Quality assurance: They may look like a superfluous member of the team; however, they have the big-picture experience from audits and inspections to advise on issues related to consent, safety, data quality, and overall study integrity.

If an electronic ICD, or eConsent, is used, we should also include an IT representative, cybersecurity expert, application developer, quiz maker, and data manager.

This team must have adequate time to develop a concise, meaningful, and effective ICD that is fit for purpose.

Can Generative AI Write The ICD?

AI, and especially GenAI, have been met with excitement by technology enthusiasts. Since GenAI can produce materials, e.g., articles that have not existed, it is presented as a universal solution to all our pains. However, we should keep in mind that the “new” article written by the GenAI tool in a couple of seconds is a compilation of existing information available in the public domain. The article can be good, or it can be a hallucination. The data set used for the AI tool training is the crucial factor. The training data provenance and reliability determine the quality of the tool’s product. Although a new template produced by the GenAI may look better than our typical one, the algorithm does not know our diverse patient population with their specific needs. And while GenAI may be a handy assistant, it should not be used to produce yet another template to recycle.

Conclusion

Developing a fit-for-purpose ICD is an effort of a multidisciplinary team focusing on the study-specific patient population in their social and socioeconomic environment, their needs, and possible limitations. If one team member is absent, their expertise and unique contribution are missing.

References:

1. Pietrzykowski, T., et al. The reality of informed consent: empirical studies on patient comprehension—systematic review. Trials 22, 57; 2021 https://doi.org/10.1186/s13063-020-04969-w.
2. 2. Tam, N. T., et al., J. (2015). Participants' understanding of informed consent in clinical trials over three decades: systematic review and meta-analysis. Bulletin of the World Health Organization, 93(3), 186–98H;2015 https://doi.org/10.2471/BLT.14.141390.
3. 3. Bertoli A. M., et al. Lack of correlation between satisfaction and knowledge in clinical trials participants: A pilot study, Contemporary Clinical Trials, Volume 28, Issue 6, 2007, Pages 730-736, ISSN 1551-7144, https://doi.org/10.1016/j.cct.2007.04.005.
4. Kamila Novak. Study Participant Consenting — Can We Do A Better Job?, Clinical Leader, Guest Column, July 24, 2023, https://www.clinicalleader.com/doc/study-participant-consenting-can-we-do-a-better-job-0001
5. Key Information and Facilitating Understanding in Informed Consent, Guidance for Sponsors, Investigators, and Institutional Review Boards, FDA, March 2024, Docket Number: FDA-2022-D-2997
6. Glossary of ICH terms and definitions, CIOMS, 2024, https://doi.org/10.56759/eftb6868  
7. FDA-NIH Terminology For Clinical Research, Glossary of Terms and Definitions, FDA-NIH Clinical Research Working Group, 2024, https://osp.od.nih.gov/wp-content/uploads/2024/05/Glossary-Terms-for-RFI_Final_Draft.pdf

About The Author:

Kamila Novak, MSc, has been involved in clinical research since 1995, having worked in various positions in pharma and CROs. Since 2010, she has run her consulting company, focusing mostly on GXP auditing. She has first-hand experience with countries in Europe, the Middle East, Africa, and North America. Kamila Novak chairs the DIA Clinical Research Community and the SQA Beyond Compliance Specialty Section, leads the DIA Working Group on System Validation, serves as a mentor at the SQA and the DIA MW Community. In addition, Kamila is a member of the CDISC, the European Medical Writers’ Association, the Florence Healthcare Site Enablement League, the Continuing Professional Development UK, the Association for GXP Excellence, and the Rare Disease Foundation. She publishes articles and speaks at webinars and conferences. She received the SQA Distinguished Speaker Award in 2023 and 2024 and the DIA Global Inspire Award for Community Engagement in 2024. She and her company actively support capacity-building programs in Africa.