Carbon footprint of industry-sponsored late-stage clinical trials

Abstract

Objectives

To quantify the carbon footprint from a sample of pharma industry sponsored phase III trials. To develop an approach that can readily be applied to future trials by AstraZeneca and other trial sponsors.

Design

Life cycle assessment including all the sources of carbon emissions associated with a completed, an ongoing and a planned clinical trial. The methodology followed the guidance on appraising the sustainability of Care Pathways, developed by the UK National Health Service in collaboration with parties across the healthcare system.

Setting

Three multicentre late phase trials. One completed heart failure trial, one ongoing oncology trial and one asthma trial with the addition of devices to be representative of current practice.

Participants

The three trials had a total number of 7412 participants.

Main outcome measures

Total carbon emissions from each trial, the drivers of those emissions and the emissions per patient.

Results

The total carbon footprint for the cardiovascular trial was calculated as 2498 tonnes carbon dioxide equivalents (CO₂e), the first 3 years of the oncology trial resulted in 1632 tonnes CO₂e and the respiratory trial 1437 tonnes CO₂e.

Conclusions

We have shown that it is feasible to perform a retrospective life cycle assessment to appraise the carbon footprint of large clinical trials and confirmed that phase III trials result in significant emissions. Having identified all the drivers of emissions and their magnitude, we are well placed to develop a plan for achieving net-zero carbon clinical trials. Now it is possible to expand the use of life cycle assessment to planned studies so that scientific aims can be achieved with a minimum of carbon emissions. We encourage other trialists to apply the same methodology as a necessary first step in reducing the carbon footprint of clinical trials.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This study is based on three relatively large, late phase trials across three therapeutic areas. Early phase trials, however, remain unexplored.

• The study appraises all mapped trial activities using standard life cycle assessment methods.

• Due to a lack of published life cycle data for some trial materials, devices and procedures, it has been necessary to estimate their carbon impact.

• Greenhouse gas emissions reported as carbon footprint are accounted for.

• Other environmental impacts have not been appraised.

Introduction

Healthcare is responsible for 4%–5% of greenhouse gas (GHG) emissions globally. GHG are a group of gases that contribute to global warming and climate change. We have used internationally agreed factors based on their relative radiative forcing over a specified time period (100 years) to aggregate them into a single global warming contribution, reported as a mass of carbon dioxide equivalents (CO₂e) and referred to as the carbon footprint.^{1 2}

National healthcare systems are making ambitious decarbonisation commitments, such as the UK’s National Health Service (NHS) objective to reach net zero by 2045.³ In 2019, the NHS carbon footprint was 25 million tonnes CO₂e, of which 5 million tonnes came from the supply of pharmaceuticals and chemicals.⁴

Over 75% of clinical trials are sponsored by the pharmaceutical industry.⁵ Many pharmaceutical companies are now taking steps to reduce their GHG emissions, in line with the Paris Agreement, committing to limit global warming to <1.5°C.⁶ In order to be fully effective, these efforts must include plans to reduce emissions associated with clinical trials. This requires a reliable method for measurement of their footprint. Life cycle assessment (LCA) is a methodology for assessing environmental impacts associated with all the stages of the life cycle of a product, process or service. While work has been done on some academic clinical trials,^7–9 there are to date no reported LCA for pharmaceutical-sponsored trials. Subaiya et al speculated that industry led trials will have a much larger carbon footprint.⁹ The number of active clinical trials is on the rise globally, with over 5550 planned new trial starts in 2021 reported, a 19% increase on 2019.¹⁰

The small number of published clinical trial carbon footprint calculations vary in order of magnitude, from tens to hundreds of tonnes CO₂e, demonstrating the challenge in generalising between trials. The Sustainable Trials Study Group performed a carbon footprint audit of the academic-sponsored, global, multicentre trial CRASH-1 (ISRCTN74459797) which evaluated the effect of corticosteroid administration on patient outcomes after traumatic brain injury. They estimated total CO₂e emissions of the 5 years, on the pdf ten is too far from other numbers, here figure 8 is much bigger than other figures 10 008-participant trial at 630 tonnes.⁷ A subsequent audit by Subaiya et al found this to be even higher, at 924.6 tonnes CO₂e, equivalent to 681 round trip flights from London to New York for one passenger.⁹

In the same audit, Subaiya et al compared the footprint of CRASH-1 with a second trial of similar design, CRASH-2 (ISRCTN86750102). Using National Institute for Health Research carbon reduction guidelines for clinical research, and improving some of the GHG hotspots identified in CRASH-1 (eg, trial management and distribution of treatment packs), the sponsors were able to reduce carbon emissions per randomised patient from 92 kg CO₂e in CRASH-1 to 25 kg CO₂e in CRASH-2.⁹

Lyle et al retrospectively calculated the carbon footprint of 12 pragmatic randomised controlled trials involving more than 4800 participants.⁸ The average CO₂e emission was 78.4 tonnes (range 42.1–112.7) per trial. Average emissions of 306 kg CO₂e per participant reported by Lyle were 3 or 12 times greater than the per patient footprint calculated for the CRASH trials.⁹ Nevertheless Lyle et al believed their assessment underestimated the carbon footprint for trials due to missing information and approximations based on literature and recommended further work to address this. All analyses of CRASH-1 and 2 footprints were based on extrapolation from a 1-year audit period which may have introduced some inaccuracies.⁹

To the best of our knowledge, in the decade since Subaiya et al’s analysis, no further carbon footprint assessments for clinical trials have been published. In a recent Lancet Commentary, Adshead et al highlighted the need to address and manage the carbon footprint of trials and to embed this within the funding and publication processes.¹¹ The development of a carbon footprinting tool to aid the appraisal of trials was seen as an essential step.

AstraZeneca aims to achieve net zero for its directly controlled emissions in 2025 and for its supply chain by 2030.¹² As part of an extensive programme to refine measurement of AstraZeneca’s carbon footprint and implement carbon reductions, we report here, for the first time, results of a LCA of three representative pharmaceutical-sponsored clinical trials. We hope this work will provide useful input and guidance to other pharmaceutical companies and groups such as the Sustainable Markets Initiative¹³ who are currently developing a clinical trial footprinting tool.

The research described here aimed to quantify the carbon emissions from three representative phase III clinical trials in order to:

• Assess their magnitude.

• Identify the key drivers.

• Identify opportunities for carbon footprint reduction.

• Provide the basis for a method for others to follow.

Methods

We chose one trial from each of AstraZeneca’s main therapeutic areas: a heart failure trial, a lung cancer trial and an asthma trial. The trials were chosen to include all the main design elements from a typical phase III trial. The use of LCA techniques ensures that impacts upstream and downstream of healthcare interventions are considered: this minimises the risks of omitting significant emission sources or advocating actions that would result in greater emissions than those avoided.

In 2015, the NHS in collaboration with parties across the healthcare system developed guidance on appraising the sustainability of Care Pathways.¹⁴ The guidance is founded on the international Standards for LCA (ISO14040 and ISO14044),¹⁵ the Greenhouse Gas Protocol Product Life Cycle Accounting and Reporting Standard (Product Standard) and the Greenhouse Gas Accounting Sector Guidance for Pharmaceutical Products and Medical Devices (Sector Guidance).

The guidance is intended for anyone interested in understanding further the sustainability of health and care systems globally and those wanting to improve or understand the sustainability of changing models of care. Clinical trials are a key component of healthcare activity, focused on the delivery of care and a first step in the development of a new care pathway. The guidance sets out a consistent approach to life cycle-based appraisal of healthcare related activities.

The guidance specifies the method for assessing GHG emissions, water use and waste generation, using a holistic life cycle approach aligned with the above standards. The method requires consideration of all distinct activities or services required to achieve the objective of the trial, including patient, clinical and management activities.

The method for determining the GHG emissions of our clinical trials was in line with section 2.3.1 of the guidance¹⁴ and was applied here as follows:

Step 1: define the objectives of the appraisal.

Step 2: define the clinical trial to be appraised.

Step 3: define the unit of analysis (functional unit) for the clinical trial.

Step 4: create a detailed map of the activities required in a clinical trial.

Step 5: complete a materiality assessment.

Step 6: collect activity data.

Step 7: source GHG emission factors for the activity data.

Step 8: combine activity and emission factors to calculate the GHG emissions per unit of analysis.

Step 9: interpret and report the findings.

Step 1: define the objectives of the appraisal

The assessment sought to appraise the carbon footprint of three late-stage phase III clinical trials on a life cycle basis with the following objectives:

• To identify key drivers of the carbon footprint for each clinical trial.

• To support future trial design and decision-making to reduce the carbon footprint of clinical trials.

• To support discussions with stakeholders.

• To inform carbon metrics and tools for the monitoring of clinical trials.

The intention was not to directly compare trials. Instead, the intention was to understand, by appraising three different trials in three different therapy areas, the drivers of the trial carbon footprint on a life cycle basis by considering both direct contributions to climate change and indirect contributions up and downstream of clinical trial activities.

Step 2: selection of trials

The selection of trials was influenced by the ability to convene the trial management teams and gain their input in capturing trial activities. The following three trials representing typical phase III trials in the respective indications were selected for the carbon footprint assessment:

• DAPA-HF (NCT03036124): this trial evaluated the effect of dapagliflozin on the incidence of worsening heart failureor cardiovascular death in patients with chronic heart failure. The study randomised 4744 patients with reduced ejection fraction to receive either oral SGLT2 inhibitor dapagliflozin or placebo, in addition to recommended therapy. The trial involved 456 investigational sites across 20 countries and ran for 39 months.¹⁶

• ADRIATIC (NCT03703297): this ongoing double-blind, multicentre, placebo-controlled trial has randomised 668 patients with limited stage small-cell lung cancer in 261 investigational sites across 19 countries. Patients receive either durvalumab in combination with tremelimumab, durvalumab+placebo or placebo alone. The primary objectives are progression-free survival and overall survival for durvalumab±tremelimumab versus placebo, and the trial is expected to complete in October 2025.

• Representative asthma trial: randomised, placebo-controlled trial design with 2000 patients, across 300 sites and 29 countries. The design was based on the structure of a recent trial of tezepelumab, a monoclonal antibody blocking thymic stromal lymphopoietin, in patients with severe, uncontrolled asthma.¹⁷ Several parameters were modified to include design elements of a more representative asthma trial, including dry particle inhalation devices for investigational product delivery and inclusion of cough monitoring.

Step 3: unit of analysis

With the exception of the lung cancer trial, the unit of analysis was the complete trial: Starting from the draft protocol, ending at completion of the written trial report.

1. A 39-month trial of 4744 randomised adult patients from 456 investigational sites across 20 countries with an established diagnosis of heart failure and at high risk of cardiovascular death.

2. The first 36 months of a lung cancer trial of 668 randomised adult patients who have not progressed following definitive, platinum-based chemotherapy concurrent with radiotherapy from 261 investigational sites across 19 countries.

3. A 26 month asthma trial designed for 2000 randomised adult patients with moderate and severe asthma from 300 sites across 29 countries.

In addition to the above, an alternative unit of analysis is used to aid interpretation of results, that is, ‘per randomised patient’. However, the use of this unit of analysis requires reader caution as it does not infer a level of comparability between trials given the difference in patients and therapy areas.

Step 4: map of activities

Clinical study teams and clinical operations experts generated a high level map of common clinical trial activities, capturing the sum of activities across the three different trials. This initial map was reviewed and amended during multiple sessions with the clinical study team and the sustainability experts team. The detailed map provided the basis for the materiality assessment (see online supplemental figure 1).

Online supplemental figure 1 describes the boundaries of the assessment and table 1 details the included activities and carbon footprint contributions. The trial system includes the extraction of resources and the production of materials, distribution, fuels and energy consumed by all activities as well as the disposal of waste from the trial activities.

Table 1.

Activities included within the assessment

Category	Activity	Description
Study team	Facilities	Direct and indirect energy use at AstraZeneca sites that can be attributed to the clinical trial. The manufacture and disposal of office equipment at end of life are excluded.
	Staff commuting	Staff travel to and from AstraZeneca facilities for employees involved in the trial.
Study drug impact	Investigational product (IP) and placebo	Manufacture, packaging and distribution of IP and placebo to trial sites, as well as administration at trial sites (eg, use of inhalers) where relevant and waste treatment.
Patient journey	Patient visits	Patient travel to and from trial sites for trial related appointments.
Samples life cycle	Pregnancy tests	Manufacture of locally sourced pregnancy tests used during the trial.
	Testing kits	Manufacture and distribution of testing kits to trial sites plus waste treatment of unused kits.
	Global biobank storage	Long-term biobank storage and archiving of relevant samples.
	Samples for analysis and storage	Transport of samples to designated central laboratory, sample analysis and sample storage and, for selected samples, onward transport to the biobank.
Trial devices and software	Electronic patient-reported outcomes (ePRO) Platform for Clinical Trials, home spirometer and e-diary, on-site spirometry device, portable recorder of exhaled air, portable cough monitor	Manufacture and distribution of device to trial sites, impact of use during trial duration and waste treatment at end of trial.
Procedures	Imaging (MRI, CT, PET, X-ray, bone scans) and ECGs	Energy consumption and consumables required per scan, located in each trial country. The manufacture and disposal of hospital equipment at end of life are excluded.
Other procured services	Services including trial documentation, data management and devices and software	Activities of third-party vendors providing services to the clinical trial.
Investigator and trial site management	Site monitor visits	Travel of site monitors to and from trial sites for visits, including hotel nights where required. Virtual visits are also included.
	Monitor training	Initial training during setup and continued training for the duration of the trial.
	Site audits	Travel of auditors to trial sites, including hotel nights where required.
	Trial site utilities	Direct and indirect energy use at trial sites that can be attributed to the clinical trial. Consumables used by investigators other than those included in test kits are excluded.
	Investigator and support staff commuting	Trial site investigator and support staff commuting impacts, which can be attributed to the trial.
	Patient communication materials	Information packs and thank you cards provided to patients enrolled in the trial.
Global trial management	Investigator meetings	International and short haul flights, road transport and hotel nights required for face-to-face global investigator meetings. Attendance at virtual meetings also included.
	Monitor meetings	International and short haul flights, road transport and hotel nights required for face-to-face global monitor meetings. Attendance at virtual meetings also included.
	Data monitoring committee (DMC) meetings	International and short haul flights, road transport and hotel nights required for face-to-face DMC meetings. Attendance at virtual meetings also included.
	Executive committee (EC) meetings	International and short haul flights, road transport and hotel nights required for face-to-face EC meetings. Attendance at virtual meetings also included.
	Independent Review Board/Independent Ethical Committee (IRB/IEC) meetings	International and short haul flights, road transport and hotel nights required for face-to-face IRB/IEC meetings. Attendance at virtual meetings also included.
	Global study team meetings	International and short haul flights, road transport and hotel nights required for face-to-face global study team meetings. Attendance at virtual meetings also included.

PET, Positron emission tomography.

The carbon footprint of construction of buildings/infrastructure, manufacture of office equipment used by the trial management team and the work of national health authorities are excluded from the scope of the assessment. This is a limitation, driven by practicalities of obtaining data and the lack of influence to be had. Based on personal communication, our opinion is that the carbon footprint contribution of the authorities, infrastructure and office equipment is unlikely to be significant, given the scale of these activities in the context of the trials and the long active lifetime of equipment/infrastructure.

Step 5: materiality assessment

The materiality assessment was performed on the DAPA-HF trial. It was used to inform data collection efforts and to ensure that all activities that have a significant impact and that could be influenced by AstraZeneca were included in the appraisal.

The materiality assessment was conducted by the trial team estimating the quantity of each activity identified in the trial map and using available emission factors to calculate the likely carbon footprint contribution of the different activities identified (see online supplemental table 1). This process helped identify key data gaps and established data collection priorities for the main study.

The key data gaps identified were associated with third-party vendors who provided a large variety of services. In the absence of supplier activity emission factors, their carbon footprint contribution in the materiality assessment was estimated based on the expenditure for vendor services, using GHG emission factors for industry sectors per USD of output.¹⁸ Using this methodology, the vendors contributed 53% of the estimated total trial carbon footprint.

In response to the materiality assessement no trial activities were excluded from the main assessment of the three trials, but importantly we collected activity data from the third party vendors for all three trials.

Step 6: collection of activity data for the LCA

The extent of each activity in terms of location, quantity of supplies, number of staff and patients, attendance and location of meetings was established through interviews with global trial management teams and reference to trial management records. It was necessary to estimate patient time and patient travel, investigator person hours and commuting, hotel nights, travel and meals for meetings as this was a retrospective assessment. No data were collected from trial sites. Table 2 in online supplemental material details activity data and table 3 in online supplemental material details the assumptions made.

The assessment involved working with third-party vendors for the trial to determine the carbon footprint contribution associated with all the activities underlying the services supplied. Where third-party vendors were unable to supply their carbon footprint contribution, additional data were requested for their activities allowing the authors to calculate the carbon footprint contribution based on activity data.

Emission factors for services provided to the lung cancer and asthma trials were generated using the third-party vendor data collected for the heart failure trial. Once all activity data had been quantified or estimated for each clinical trial (see online supplemental table 2), GHG emission factors were sourced or estimated for each activity (see online supplemental tables 3–5).

Patient and public involvement

Patients or the public were not involved in the design, or conduct, or reporting, or dissemination of our research.

Results

The total carbon footprint for the DAPA-HF trial was calculated as 2498 tonnes CO₂e (table 2), equivalent to the emissions associated with driving an average petrol car 10 million km.¹⁹ Site monitor visits, study drug supply, study team facilities and the sample life cycle are the main drivers of impact. Travel and transport make the largest contributions to these impact drivers, for example, the site monitor visits impact is dominated by car travel of monitors to sites. The study drug and samples contributions are dominated by the air freight distribution of study drugs to trial sites, and the frozen shipment of collected samples from the trial sites to the central laboratories.

Table 2.

Results of LCA for DAPA-HF heart failure trial

Category	Activity	Carbon (tCO2e)	Contribution%
Study team	Facilities	361.4	14.5
	Staff travel	70.1	2.8
Study drug impact	IP and placebo	391.5	15.7
Patient journey	Patient visits	143.4	5.7
	Patient communication materials	1.1	0.04
Sample life cycle	Global biobank storage	3.3	0.1
	Samples for analysis and storage	265.6	10.6
	Pregnancy tests	0.2	0.01
	Testing kits	91.3	3.7
Other procured services (third-party vendors)	Documentation and communication	64.3	2.6
	Data management	33.9	1.4
	Device, software and online tools	2.7	0.1
	Other services	13.0	0.5
Trial devices and software	ePRO	70.9	2.8
Investigator and trial site management	Site monitor visits	519.4	20.8
	Trial site utilities	37.6	1.5
	Investigator and support staff commuting	24.4	1.0
Global trial management	Investigator and monitor meetings	182.8	7.3
	Data monitoring committee meetings	27.3	1.1
	Executive committee meetings (seven international f2f meetings)	57.6	2.3
	IRB/IEC meetings	121.8	4.9
	Global study team meetings	14.5	0.6
Total		2498	100
	Per randomised patient (kg of CO2e)	527

CO_2e, carbon dioxide equivalent; ePRO, electronic patient-reported outcome; IP, investigational product; IRB/IEC, Independent Review Board/Independent Ethical Committee; LCA, life cycle assessment.

The total carbon footprint for the first 3 years of the ADRIATIC lung cancer trial are calculated to be 1638 tonnes CO₂e (table 3). The key driver of impact is the study team facilities accounting for 838 tonnes of CO₂e (51.1%) and global trial management meetings accounting for 234 tonnes of CO₂e (14.3%). Although the lung cancer and heart failure trials involved the same number of person hours the emissions contribution is double in the lung cancer trial; this difference is due to a larger percentage of staff working at locations where the required utilities are powered from a more carbon-intense grid, such as in the USA or China. The impacts from global trial management meetings are driven by both the transport of attendees, such as international long and short haul flights for monitors and investigators to attend the meetings, and also the provision of meals for attendees of the institutional review board/institutional ethics committee meetings (modelling assumed 33 attendees per meeting with each receiving lunch with an estimated carbon footprint of 5.9 kg CO₂e per meal).²⁰

Table 3.

Results of LCA for ADRIATIC lung cancer trial

Category	Activity	Carbon (tCO2e)	Contribution
Study team	Facilities	838	51.14%
	Staff commuting	70	4.29%
Study drug impact	IP and placebo	33	2.03%
Patient journey	Patient visits	11	0.67%
	Patient communication materials	0.4	0.02%
Sample life cycle	Global biobank storage	1.1	0.06%
	Samples for analysis and storage	44	2.68%
	Pregnancy tests	0.1	0.00%
	Testing kits	14	0.84%
Other procured services (third-party vendors)	Documentation and communication	14	0.87%
	Data management	18	1.09%
	Device, software and online tools	90	5.47%
	Other services	5.5	0.33%
Trial devices and software	ePRO devices	95	5.78%
Procedures	MRI scans	51	3.10%
	CT scans	9.4	0.57%
	Other diagnostics (X-ray, ECG, bone and PET scans)	0.9	0.05%
Investigator and trial site management	Site monitor visits (and audits)	68	4.17%
	Trial site utilities	26	1.61%
	Investigator and support staff commuting	15	0.90%
Global trial management	Global investigator meetings	77	4.73%
	Global monitor meetings	18	1.11%
	IRB/IEC meetings (486 local f2f IEC meetings, 306 virtual)	139	8.47%
	Data monitoring committee meetings and global study team meetings	ND	ND
Total		1638	100%
	Per randomised patient (kg of CO2e)	2452

CO_2e, carbon dioxide equivalent; ePRO, electronic patient-reported outcome; IP, investigational product; IRB/IEC, independent review board/independent ethical committee; LCA, life cycle assessment; PET, Positron emission tomography.

The total carbon footprint for the designed asthma trial is calculated to be 1437 tonnes CO₂e (table 4). The key drivers of the impact for this trial are the trial devices (443 tonnes CO₂e, 30.8%), samples (179 tonnes CO₂e, 12.5%) and the external trial management conducted by CROs (166 tonnes CO₂e, 11.6%). The device emissions come largely from their distribution (by air freight) from the site of manufacture to all global trial sites. However, for devices where there is a shorter lifetime, such as the home spirometry device, the impacts associated with their manufacture are fully allocated to their use in the trial and not spread out over a longer lifetime or multiple trial usage. The manufacturing footprint, therefore, makes a significant contribution to the impact of these devices for this trial. Again, the impact associated with samples is driven by the frozen shipment of collected samples from the trial sites to central laboratory. The outsourced trial management, conducted by CROs, had a significant contribution as this impact is analogous to the contributions of study team facilities in the previous two trials.

Table 4.

Results of LCA for representative asthma trial

Category	Activity	Carbon (tCO2e)	Contribution
Study team	Facilities	19	1.30%
	Staff commuting	10	0.69%
Study drug impact	IP and placebo (including inhaler device)	24	1.66%
Patient journey	Patient visits	128	8.91%
	Patient communication materials	0.4	0.03%
Sample life cycle	Global biobank storage	2	0.15%
	Samples for analysis and storage	179	12.50%
	Pregnancy tests	0.4	0.03%
	Testing kits	21	1.48%
Other procured services* (third-party vendors)	Documentation and communication	27	1.88%
	Data management	16	1.09%
	Device, software and online tools	10	0.66%
	External trial management, CRO (estimated)	166	11.60%
	Other services	6	0.44%
Trial devices and software	Home spirometry device with eDiary (assumed 1-year lifetime)	260	18.10%
	On-site spirometry device (assumed 7-year lifetime)	76	5.30%
	FeNO measurement device (assumed 1-year lifetime)	53	3.70%
	Cough monitor device (subgroup of 50 patients)	0.31	0.02%
	ePRO devices	54	3.75%
Investigator and trial site management	Site monitor visits	165	11.50%
	Trial site utilities	114	7.91%
	Investigator and support staff commuting	87	6.06%
Global trial management	Investigator and monitor meetings	0.69	0.05%
	DMC meetings	0.14	0.01%
	IRB/IEC meetings (30 local f2f and 10 virtual, 57 hybrid)	18	1.26%
	Global study team meetings	ND	ND
Total		1437	100%
	Per randomised patient (kg of CO2e)	718

CO₂e, carbon dioxide equivalent; CRO, Clinical Research Organisation; DMC, data monitoring committee; ePRO, electronic patient-reported outcome; IP, investigational product; IRB/IEC, Independent Review Board/Independent Ethical Committee; LCA, life cycle assessment; ND, no data.

Figure 1 shows the carbon footprint drivers for each trial. Note that the results reported are on a per patient basis and the drivers of carbon footprint vary across the trials. There are several common drivers across the three trials, in particular the study team (for the asthma trial included in procured services), the sample life cycle (laboratory kits and distribution, sample shipment, analysis and storage) and the in-person meetings associated with investigator and trial site management (site monitoring visits). Providing devices to investigators and patients will have a significant influence on CO₂e footprint. When considering opportunities to reduce emissions, it is worth noting that a significant percentage of sample and study drug impact comes from supply that is never used.

Figure 1.

Drivers of CO₂e per patient for each trial.

A significant finding in our analysis was a large disparity between financial proxies for carbon as used in the materiality assessment compared with collecting and using activity-based data and corresponding carbon emission factors. In the heart failuretrial, a 10-fold overestimation was observed using the financial surrogate markers for the external vendor contribution to the carbon footprint. This observation confirms the need for increasing primary data coverage when calculating and reporting supply chain (scope 3) emissions.

Discussion

Our initial materiality assessment of the phase III heart failure trial suggested it was associated with significant carbon emissions. We refined the initial materiality assessment and then confirmed a similar scale of emissions in two further typical phase III trials. There were some drivers of emissions that were consistent across the three trials. These hotspots included the travel associated with in-person meetings, sample life cycle including supply, particularly shipments between continents and unused materials such as sampling kits and later in the process, shipments and storage of samples for future analysis. In our experience, clinical trials ship and store many frozen samples for future exploratory research but historically few are used. Based on our findings, we suggest those hotspots as the first focus for those wishing to reduce the emissions from a large trial. Other drivers varied in magnitude between the trials, with some of those drivers related to the disease area. However, all activities within a trial were shown to have the potential to drive the overall footprint. This diversity suggests the need to assess emissions on a trial-by-trial basis as no single driver, obvious surrogate or key performance indicator, is apparent as the sole basis for estimating the carbon footprint of a trial. At present, AstraZeneca has employed two main metrics; first, the total carbon emissions from the trial which will allow us to understand the total contribution to climate change that trials make, and second emissions per patient to enable comparison of trial designs. A next step for us is expanding existing IT tools to replicate the assessments reported here to allow design teams to include carbon emissions as one factor in the choice of trial design.

Previously published emission footprints include two global multicentre trials CRASH-1 and CRASH-2,⁹ and 12 smaller pragmatic randomised controlled trials,⁸ all in an academic setting. Our study of recent or current phase III trials in the pharma setting includes the measurement of a much more complete set of activities and therefore higher calculated emissions per patient.

The cardiovascular and oncology studies required very similar internal staffing, but the calculated emissions were almost double for the US-centric team versus the Sweden-centric team due to the differences in the country specific impact of energy use. By developing a strategy to decarbonise our facilities, with an aim to be net zero, this difference can be addressed. This illustrates that footprint calculations are to some extent organisation specific and also the importance of carefully allocating each activity to the correct country. However, it raises a moral dilemma that by excluding certain areas of the world, carbon dioxide emissions can be reduced. Therefore, we must carefully plan, engage with partners and provide a focus to decarbonisation efforts in order to maintain geographical inclusion and diversity while minimising carbon emissions.

We find that understanding the underlying drivers for carbon enables us to put strategies in place to reduce our trial emissions both by looking at drivers relevant to all trials and also by specific factors in each disease area. Carbon reductions are enabled by optimising trial design and delivery. Carbon reduction is easiest where the savings have cobenefits, such as reduced financial cost or reduced complexity for investigator sites and trial participants. Robust measurement of carbon emissions would allow for their impact to join the traditional quantifiable factors that drive investment into a trial design and its delivery. Ultimately efficient pricing of carbon could allow it to be included in financial models throughout drug development. Meanwhile emissions will have to be traded off against all the other purposes and factors that influence the decisions on clinical trials.

With at least 2100 phase III trials ongoing globally,²¹ extrapolating our analysis gives a contribution of 3–5.2 million tonnes of CO₂e. Phase I–II studies will add more emissions.

We have shown that a life cycle approach can be applied to estimate the total carbon footprint of a clinical trial and that a typical phase III trial results in significant carbon emissions. We have identified hotspots such as travel and sample shipments that are likely to be common drivers of emissions in pharma industry trials while other drivers of emissions may be more related to disease area. In the selected three trials, we found financial data to be a poor proxy for activity data, that is, a large overestimation using financial proxies. Such poor estimates could result in misdirected reduction strategies and investments. Our findings demonstrate the benefit of bottom-up life cycle measurement approaches as opposed to estimating impact purely based on spend data. We expect other trial sponsors will make the same finding. Importantly, all the sources of emissions can be reduced by different choices during trial design and operational delivery. As a result of this study, we have demonstrated a process by which we can make the carbon cost as visible as financial cost to everyone designing and making investment decisions for phase III studies and programmes. Having identified all the drivers of emissions and their magnitude we are well placed to develop a plan for achieving net-zero carbon clinical trials. We believe we have shown that it is feasible for industry or academic sponsors to calculate the footprint of a trial. Like Adshead et al,¹¹ we encourage others to follow a similar approach.

Data availability statement

All data relevant to the study are included in the article or uploaded as online supplemental information. Data obtained from a third party may not be freely publicly available. Financial and third-party data related to some of the findings described in this manuscript may be considered commercial in confidence and not publicly available. Data underlying the findings described in this manuscript may be obtained in accordance with AstraZeneca’s data sharing policy described at https://astrazenecagrouptrials.pharmacm.com/ST/Submission/Disclosure.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

This study does not involve human participants. The trials appraised for their carbon footprint in this study were approved by the ethics committee at each site. All participants provided written informed consent.