Designing a trial for lower emissions
What a clinical trial emits, and how much of it can be cut, turns on the choices its protocol makes about visits, samples, data collection and supply, and on the digital solutions that shorten a trial or reduce its number of participants. Which of those choices a trial can make depends on whether regulators accept a smaller trial, and on how a sponsor weighs each option's carbon against its other needs.
Design questions for travel, samples, data and supply
Travel can be reduced by replacing some on-site visits with remote visits, virtual monitoring and investigator meetings, telemedicine visits or home monitoring devices, perhaps with a reduced visit schedule or visit frequency. Embedding the trial in standard of care minimises additional assessments and travel to site. Where possible, electric vehicles can ship the product to sites or transport participants to them.
Each sample should have a strong rationale for its inclusion. Samples that need long-term storage, a cold chain or expedited shipment should be minimised, and the pros and cons of local against central laboratories weighed. Before a sample is collected, it should be checked whether the answer to its scientific question is already available and whether a sample already in the biobank can serve instead.
A medical writer can really make a difference in deciding which data a specific question needs. Each item of data and each analysis should be questioned: whether it needs to be collected or performed at all, and whether regulatory approval or reimbursement needs it. If not, it is waste and should not be done. Many samples are collected and not analysed, and many exploratory endpoints bring nothing in return. So the study objectives should be optimised, exploratory objectives minimised and data collection optimised to address the study objectives properly.
The distribution of the investigational product should be optimised, and the clinical supply of both the investigational medicinal product and any commercial product used in the trial planned carefully.
The NIHR carbon reduction guidelines
The UK National Institute for Health Research's carbon reduction guidelines are part of the National Health Service's commitment to meet the targets set by the Climate Change Act of 2008. Their rationale is that, as the leading funder of health research in the NHS, the NIHR must play a part in reducing carbon emissions from health research. They were published on 30 July 2019, and as of March 2022 an update was planned.
They are aimed at researchers conducting NIHR-funded research, but their principles apply to all research regardless of type, funding source or geography. They are not mandatory: they offer a framework for reducing the carbon footprint of clinical research without adversely affecting its quality, validity or reliability. UK researchers developed them from the data in two papers: the Sustainable Trials Study Group's 2007 paper quantifying the emissions of the CRASH trial, and Lyle and colleagues' 2009 analysis of the carbon cost of a sample of pragmatic randomised controlled trials. A review of the published carbon footprint studies counts a third among those whose results were used to develop them: Subaiya and colleagues' 2011 comparison of the carbon footprints of the CRASH-1 and CRASH-2 trials.
The recommendations fall under two headings: sensible study design, and reducing the environmental impact of the NHS through research. Their high-level headings are:
- setting the research question and making full use of existing evidence
- efficient study design
- study set-up and conduct
- avoiding unnecessary data collection
- sensible clinical trial monitoring
- good practice in reporting research
- reducing the environmental impact of the NHS through research
Digital solutions that reduce a trial's emissions
The health systems task force of the Sustainable Markets Initiative advocates digital solutions as a way to reduce clinical trial emissions, and it identified five ways in which they could do so.
The first is a shorter trial timeframe, which reduces the emissions from electricity use and office waste. Running more trials in fewer facilities, and faster, would reduce emissions. Digital biomarkers, together with data and artificial intelligence tools for trial planning, site selection, patient screening and recruitment, could enable a shorter trial. For example, one sponsor's advanced analytics streamline site selection by aggregating multiple datasets to give evidence-driven recommendations on site effectiveness, which reduces the need to visit and audit several sites and so the emissions from that travel. Digital recruitment and screening could both reduce patient travel and help select patients more likely to complete the trial, which reduces wasted resources that would otherwise add to emissions.
The second is optimising employee work through electronic data capture and artificial-intelligence-driven automation, which could streamline processes and reduce staff requirements and workload by eliminating repetitive tasks and reducing data entry errors.
The third is fewer trial participants, through patient-reported outcomes and novel endpoints and through digital solutions. A synthetic control arm, a computer-generated control arm based on real-world evidence or on historical data from similar trials, can reduce the participant cohort by around 20 to 50%. Another option is a digital twin, which takes the place of the placebo group and uses clinical records to simulate how patients would be likely to respond in the placebo arm. Providers of digital twins use real-world evidence on how a general population may respond to treatments or present with a disease. They quote an average 12 to 25% reduction in trial participants, a figure that depends on the trial, the patient population and other aspects.
The fourth is reduced patient travel: decentralised trials, remote data monitoring and collection, wearables and telemedicine can reduce the need for participants and staff to travel to sites, which could reduce emissions by 25 to 40%.
The fifth is better recruitment and retention, a pain point for most clinical trials. Decentralised trials enabled by digital solutions could reduce travel costs and the commitment required of participants and let a wider audience take part. Higher retention would reduce waste in the manufacture, supply and transport of drugs, and recruiting from broader areas would make the cohort more representative of the broader population. So digital solutions would benefit the quality of trials as well as reduce their emissions.
Regulatory acceptance of a smaller trial
One practitioner has seen study designers concerned about whether regulators will accept synthetic control arms and digital twins, a relatively novel approach. In 2024 one sponsor saw no willingness among regulators to compromise on trial size, as distinct from quality, on sustainability grounds. It hoped this would change, so that a sponsor could tell a regulator that the study as the regulator wants it would emit a given amount more carbon dioxide, and ask whether that is really needed. Until then, in its view, the best course is a conversation with the health authorities, using scientific justification for why synthetic control arms or digital twins may be appropriate. In its view they are probably appropriate where the patient population is small, which may make oncology and rare diseases the more appropriate areas, and those are the areas where it has seen them used more.
Reducing emissions is very hard in a trial the size of the cardiovascular prevention outcome trial that sponsor was running in 2024, which expected 8,000 patients. It was fairly confident that its cardiovascular and metabolism trials did not use digital twins, because of the nature of the diseases and of the trials, while it thought its oncology trials might use synthetic control arms. In its view, what the task force advocates was at the time very much aspirational, and difficult to put into practice because the framework to implement such digital solutions was not yet there.
Guidance was available at local level, such as FDA guidance on control arms and real-world evidence, the EMA reflection paper on artificial intelligence and the procedure for qualifying digital biomarkers, but its requirements differed by region. There was some guidance on synthetic control arms and none on digital twins. The task force encourages wider acceptance of digital solutions in clinical trials. It sees a need for more international alignment, so that global trials can apply digital solutions; more detailed guidance on digital twins and synthetic control arms; new endpoints, such as patient-reported outcomes and surrogate endpoints, that would help to optimise study size and reduce the number of participants; and considerations for embedding the carbon footprint in drug development.
Publishing validation studies of external control arms meets, in one view, a chicken-and-egg problem. In that view, regulators will be more open to the technique once more validation studies are published, including in harder therapeutic areas such as cardiology, while journal editors, especially at clinical journals, may not yet know the technique well, may be somewhat sceptical of it and look to the regulators to say it is acceptable. It may take more time before the technique becomes mainstream. Another view holds that regulators are the main hurdle, and that an editor will generally publish a good, scientifically valid paper, particularly one with something new in it. Clinical trials have been conducted with a control group for decades, so changing that concept is quite difficult, and the holder of that view knows of some published work on digital twins using real-world evidence, but not across a wide range of therapeutic areas. A study could be run both the traditional way and with a digital twin, but running it two ways carries a large cost, and results that differ then create a conundrum when put in front of a regulator.
The sponsor that saw no willingness to compromise on trial size answers the chicken-and-egg problem by continuing to talk with the health authorities, explaining the sponsor's views and what it is doing and taking them along, because they are learning as the sponsors are. It engages with policy makers on sustainability in drug development, attends business pipeline meetings with health authorities, and meets them to present on sustainability and what it is doing. It presented its life-cycle assessment to the EMA to seek feedback, and hopes that the agency will take more of a position as more companies do so and more publications become available.
Weighing design options with a carbon calculator
The same sponsor incorporates carbon dioxide reduction into the design of new clinical trials, involving the contract research organisations it works with, and has used a carbon emission calculator during study design since 2022. It links carbon dioxide considerations to clinical trial innovation, such as digital decentralised trials and new endpoints.
One contract research organisation's approach to a trial's emissions has four parts: model where the emissions come from, count the carbon as transparently as possible, work out how to decarbonise, and give sponsors options for further reduction. Many things in a trial are not necessarily under its control, but depending on its involvement in the study design, it can suggest changes to the sponsor once it knows which activities are carbon-intensive. For example, where a protocol requires five magnetic resonance imaging scans, a very carbon-intensive activity, the question can be asked whether five are needed, or the sponsor can be shown how much carbon reducing them to three or four would save in that trial, to help it meet its goals.
The same organisation was working with the Pistoia Alliance and the Sustainable Healthcare Coalition, part of the Sustainable Markets Initiative, on a very detailed calculator of a clinical trial's carbon emissions, and was trying to take it a step further into a simplified model with far fewer inputs for the study design phase. The simplified model looks at about 25 possible changes at most to reduce a trial's carbon. It is meant to let a sponsor see how much carbon a change to the design would save and balance that against speed, regulatory submission requirements, cost and the other things that matter for the study's outcomes.