The carbon footprint of clinical trials
A clinical trial's carbon footprint comes from running its sites, the travel it causes and the supplies and samples it ships, and which of these dominates differs from one trial to the next. The few published audits put figures on it that depend on what each one counted and when, and each source of emissions has its own ways of being reduced.
How much is known about trial emissions
Clinical research is one of the pharmaceutical industry's key carbon-intensive activities. Thousands of clinical trials are conducted globally each year, and as of 13 January 2022 ClinicalTrials.gov listed 400,873 studies. Yet in 2022 the regulatory landscape for healthcare products did not take the environmental impact of clinical trials into account. Little is known about that impact.
A PubMed search on 30 December 2021 for "clinical trials AND carbon footprint", with no filters, returned 12 publications, of which four were judged eligible, and a manual search of the identified publications added one more. All five papers were by UK researchers: two commentaries, two pieces of original research on the CRASH trials, and a meta-analysis of 12 pragmatic randomised controlled trials. Between them they report carbon metrics for 14 trials, expressed in carbon dioxide equivalents (CO2e) and broken down into three broadly defined activities: operations, travel and supplies. The literature on the greenhouse gas emissions of clinical trials is sparse, and this dearth of publications indicates an underserved domain.
How the emissions were counted
Three of the papers reported data on specific trials and followed a similar method. Data were collected retrospectively, according to the greenhouse gas reporting protocol developed by the World Business Council for Sustainable Development, on every trial element that would generate emissions, and the emissions were expressed in CO2e with the protocol's calculation tools and generally accepted conversion factors. The sources were grouped roughly as operation of the coordination centre or study site (fuel for electricity, waste disposal, water), travel (trial staff commuting and trial-related journeys), and trial supplies (manufacture and distribution of drugs, documents and other equipment). All the calculations followed the same protocol, but scoping and conversion factors could still differ between them.
How large a trial's footprint is
The Sustainable Trials Study Group, convened by the London School of Hygiene and Tropical Medicine to find ways of reducing greenhouse gas emissions from clinical trials, published in 2007 what is probably the first report quantifying a trial's emissions. Its subject was the CRASH trial, a multicentre international study run from April 1999 to May 2004 on the effect of corticosteroids on death and disability in adults with head injury. On a one-year carbon audit the whole trial was estimated to have emitted 630 tonnes CO2e, and with 10,008 participants and 1,945 primary endpoint events that is 63 kg CO2e per participant or 324 kg per primary endpoint event. For the trial's 5.1 years the report gives 126 tonnes a year. It names trial documentation as a contributor to a study's footprint, which links medical writing directly to carbon emissions.
In March 2022 the first and only meta-analysis of trial emissions known to have been published was Lyle and colleagues' analysis of 2009. It covered 12 pragmatic randomised controlled trials funded by the UK National Institute for Health Research's Health Technology Assessment programme from 2002 to 2003, for which the CRASH trial was not eligible: more than 4,800 participants, UK centres only, a mean of 16 centres and 402 participants per trial, and a wide range of interventions including pharmaceuticals, devices and psychological therapies. Mean emissions were 306 kg CO2e per participant, with a range of 80.0 to 883.7, and 78 tonnes per trial, with a range of 42.1 to 112.7, for a total of 941.2 tonnes across the 12 trials. Per year, the trials emitted 222.3 tonnes in total and a mean of 18.1 tonnes each, with a range of 8.9 to 30.1, and the analysis also reported a mean of 0.1 kg per pound spent and 5.6 tonnes per full-time member of staff. Beyond site operation, travel and supplies, it counted the information technology equipment used in the trials. It did not count freight distribution of trial documentation, probably because the documents were electronic.
A 2011 paper by Subaiya and colleagues followed up the 2007 one by comparing the original CRASH trial, designated CRASH-1, with CRASH-2. CRASH-2 was of similar design and ran from May 2005 to February 2010, starting one year after CRASH-1 ended, but it made a greater effort to reduce its footprint, using several of the strategies outlined in the NIHR carbon reduction recommendations. It recruited roughly twice as many participants, 20,211, and emitted 73% less carbon per randomised patient, 25 kg against 92 kg CO2e, the main drivers being more efficient study design, recruitment and conduct and more compact trial supplies. CRASH-1, over 5.1 years, came to 181.3 tonnes a year and 924.6 tonnes for the whole trial, and CRASH-2, over 4.7 years, to 108.2 tonnes a year and 508.5 tonnes. These CRASH-1 figures differ from the 2007 paper's 63 kg per participant and 630 tonnes for the same trial. Because carbon calculation tools are regularly updated, the difference was most likely due to different metrics, such as updated tools and conversion factors.
Adshead and colleagues' 2021 commentary extrapolated the earlier estimates to the roughly 350,000 trials registered in ClinicalTrials.gov as of 16 June 2021 and arrived at an estimated cumulative 27.5 million tonnes of emissions attributable to clinical trials globally. In 2024 a speaker from one trial sponsor put clinical trial emissions at up to 100 million tonnes of CO2 a year, equivalent to Belgium's yearly emissions.
AstraZeneca's life-cycle assessments of three large phase 3 trials, which led to a publication in 2022, were by the company's account probably the first life-cycle assessments conducted on industry-sponsored trials. The three trials were in different therapeutic areas: DAPA-HF, a cardiovascular outcome trial in heart failure that randomised 4,744 patients in 20 countries; NAVIGATOR, an asthma trial of 2,000 patients across 300 sites in 29 countries; and an oncology study that had enrolled 668 patients with small-cell lung cancer across 261 sites in 19 countries by 2024.
Judged by its length and number of participants, the cardiovascular outcome trial was expected to have the highest emissions, and it did, at 2,498 tonnes of CO2. The oncology study had the highest emissions per patient, 2,454 kg against 526 kg in the cardiovascular outcome trial, because different trials had different drivers of emissions.
Where the emissions come from
In AstraZeneca's cardiovascular outcome trial, 24% of emissions came from investigator and trial site management, mostly travel to sites for audits, investigator site meetings and visits by clinical research associates. Site emissions, the sample life cycle and global trial management also contributed. In the asthma trial, 30% came from the trial devices used with asthma patients, and investigator and trial site management came second. In the oncology study, 55% came from site emissions, because the sites were in countries such as the United States and China, whose power sources are more carbon-heavy.
No single driver of emissions was found across the three trials, and emissions need to be assessed trial by trial. The size of a trial, its number of participants, is a key driver. The assessments grouped the drivers by activity:
- trial visits, including site monitoring and investigator meetings, driven by travel and transport by plane and car
- study drug supply, driven by air freight and distribution
- site operations, driven by the sites' power sources
- the sample life cycle: air freight of frozen samples to central laboratories, used kits, and storage of samples for future analysis
Across the five papers' 14 trials, the main activities driving emissions were study-site operation, trial-related travel and trial supplies. The key recommendations included more efficient study design and conduct and less trial-related travel. Each paper's breakdown and recommendations:
- The 2007 CRASH audit. Coordination-centre operations 39%, the largest share, distribution of drugs and documents to sites 28%, trial-related travel 23%, trial team commuting 5% and deliveries related to producing the trial drugs 5%; recommendations to reduce bureaucracy at regulatory agencies and ethics committees, simplify study designs, choose better research questions, reduce travel, avoid unnecessary data collection, save electricity through renewable energy, and use systematic reviews to answer research questions before proposing new trials, of which simplifying study design and processes and minimising travel are the key ones.
- The 2009 meta-analysis. Trial team commuting 26% and study-centre operations 23%, the two largest shares, staff trial-related travel 19%, participants' travel 16%, manufacture and distribution of supplies 14% and information technology 2%, the lowest; recommendations to minimise trial-related travel, reduce the number of face-to-face study visits, and develop tools and methods that let the carbon cost of a trial be considered at the planning stage, such as using the NIHR carbon reduction guidelines.
- The CRASH-1 and CRASH-2 comparison. One trial's emissions led with distribution of trial drugs (48%), coordination-centre operation (30%), trial-related travel (21%) and commuting (1%), and the other's with coordination-centre operation (37%), drug distribution (32%) and travel (29%); recommendations to improve trial efficiency in recruitment, data entry, validation and monitoring, reduce travel through web-based training and teleconferences, and improve logistics through more compact materials and lighter packaging.
- Pencheon's 2011 commentary. Embed sustainability as a core part of research governance, take a more holistic view of the process of conducting research, incorporate environmental as well as financial cost into commissioning, and make valid comparisons with consistent metrics.
- The 2021 commentary. Confirm through systematic reviews that a trial is necessary, make carbon footprint measures part of study design, provide funding incentives for carbon reduction, use the NIHR guidelines, involve regulatory bodies, ethics committees and journals in policing the footprint, and develop a tool that measures a trial's footprint reliably and shows which elements are carbon-heavy.
What reduces a trial's emissions
Pencheon's 2011 commentary drew heavily on the CRASH comparison and took its recommendations beyond clinical trials to the whole life cycle of health interventions, proposing to calculate potential health gain per tonne of carbon expended. It calls for the environmental impact of trials to be made more transparent and for the need for a trial to be justified by a thorough environmental cost-benefit assessment based on a systematic review of the literature and of trial registries. One of the papers proposes that regulatory agencies, ethics committees and biomedical journals police trials' carbon emissions, and its authors concede that this adds bureaucratic burden. The need for reliable clinical data has to be weighed against the urgency of the climate crisis.
As of 2023, the UK National Institute for Health and Care Research's carbon reduction guidelines for clinical trials recommend virtual meetings.
One contract research organisation, which is given the investigational product, has little effect on how the product is made but can affect how it is shipped. Depending on how involved it is in the study design, it can suggest changes to the sponsor once it knows which activities are carbon-intensive: where a protocol requires five MRIs, a very carbon-intensive activity, it can ask whether five are needed or show how much carbon reducing them to three or four would save. Much of a trial lies outside its control, and the areas it judged it could affect, at least at first, were patient travel, monitoring visits and site utilities, where in its view finding suppliers that use electricity rather than natural gas is beneficial. Its overall approach has been to minimise the carbon cost of shipping and of patient travel, use renewable electricity and increase recycling.
Patient travel
A lot of carbon is emitted in getting the patient to the trial site. Decentralised designs, in which the patient need not go to the doctor's office or can have certain tests run at a more local clinic rather than travel to the principal investigator's site, can make a big difference both to the patient's emissions and to their retention in the trial. Every clinical trial enrols an excess of participants because some drop out, and making participation easier makes a big difference, whether it is done for carbon reduction or for another reason. Decentralised trials can also allow sample sizes to be reduced.
Monitoring visits
Monitoring visits by clinical research associates make up a relatively big part of one contract research organisation's carbon footprint, and their travel is one of its more carbon-intensive activities. Doing the visits remotely is not always appropriate, but sometimes it can be, and that can make a difference. Typically 91% of the carbon of its clinical research associates' trips comes from air travel. The share varies by region and country, and it is lower in Europe, where train travel, which is significantly more carbon-efficient than air travel, is a more common substitute and air journeys tend to be a little shorter. The same organisation's sustainability lead holds that clinical trials have a lower carbon footprint in Europe than in America, partly because clinical research associates travel by train more often in Europe than in the United States. Communicating this to clinical research associates can also make a difference.
Kits and shipping
The kits sent to clinical investigators can make a very big difference to a trial's emissions. Enrolling patients as quickly as possible, to finish the trial and get the drug to market sooner, produces a lot of excess kits if speed is the only metric. Repurposing excess kits and better systems for predicting where kits will be needed are ways to reduce that wastage, and multi-visit kits reduce shipping.
Most shipments have to go frozen on dry ice, and shippers made of recyclable materials have a significantly reduced carbon footprint compared with expanded polystyrene, which is generally not recyclable. Shipping usually has to go by air, because kits have to arrive quickly and samples usually have to reach the laboratory within 24 or 48 hours. That leaves two levers: shortening the shipping distance by placing kitting facilities in different parts of the world, and using sustainable aviation fuel for air shipments, which adds a cost for the contract research organisation and for the customers who choose it but gives them that option.
Sample storage
A researcher in academic clinical research reports considerable efforts there to optimise the storage temperature of biobank samples, and holds that storing ultra-frozen samples at −70 °C rather than −80 °C can save up to 30% of energy use, with the planning phase key to this in clinical trials. One contract research organisation has some flexibility here, with some freezers at −70 °C and some at −80 °C, and it prefers the −70 °C freezers where they can be used. Freezers use a fair amount of electricity, and their refrigerants can have a negative carbon impact. In the organisation's relatively large biorepositories, limited freezer space had already led to a fairly strong process for keeping the freezers full and disposing of samples promptly once a study is done, and that process brings a sustainability benefit.
In the contract research organisation's view, people in research and development laboratories and academic groups do not want to throw samples away, and one organisation that works with academic groups makes an inventory of what the freezers hold and how many freezers are needed one of its main sustainability measures. Care over the type of freezer a sample goes in and how long it is kept can save a fair amount of energy, carbon and cost.
Limits of the evidence and open questions
Several caveats may limit how far the five papers' findings generalise. Only one database, PubMed, was searched. All five papers came from the UK, and the meta-analysis covered only UK pragmatic trials funded by one NIHR programme. The papers found up to the end of 2021 gave no data from industry-sponsored explanatory trials. Every study was performed before 2020, and trial conduct changed drastically during the pandemic, with restricted travel, remote monitoring and virtual meetings.
Gaps that warrant more research include harmonised and validated metrics for quantifying a trial's footprint; carbon metrics and reduction strategies built into trial planning and design; involvement of funders, regulatory agencies, ethics committees, biomedical journals and other governance bodies in disclosing and managing a trial's carbon profile; data from other countries, especially the United States, China and the European Union; and data on other research types and study designs.