Climate footprint of industry-sponsored in-human clinical trials: life cycle assessments of clinical trials spanning multiple phases and disease areas.

STRENGTHS AND LIMITATIONS OF THIS STUDY

This is the first study comparing greenhouse gas (GHG) emissions from industry-sponsored clinical trials spanning all four phases of clinical development and across different disease areas.

Study limitations are associated largely with the available data, and the data gaps, such as lack of visibility to the associated GHG emissions of some drug products, were filled by proxy values or assumptions.

The involved trials were selected to show diversity and may not be representative of all trials of their type.

Due to resource limitations, the sample size remains limited.

Despite the limitations, overall, the assessment is a reasonable estimate of the GHG emissions and key drivers of GHG emissions for this subset of clinical trials.

Introduction

Historically, life cycle assessments (LCAs) of pharmaceuticals have been process or product driven, either focusing on part of the pharmaceutical supply chain 1 3 or assessing the footprint of a product. 4 However, LCA boundaries have gradually expanded, and a new focus has emerged. This new area recognises that pharmaceuticals as a product are just one element of a larger care pathway featuring healthcare provider visits, hospitalisation and/or outpatient care. 5 7 A limitation of the recent care pathway approach is that it has focused on typical care pathways in a commercial setting after a drug has received regulatory approval and has often neglected the environmental impact of clinical research required to bring those drugs to market. This research seeks to close that gap by shedding light on the greenhouse gas (GHG) emissions of clinical trials and the key underlying drivers of those emissions in clinical trial designs.

A phase 1 trial represents the first testing of a new drug in humans primarily to assess safety. A drug is given to a small group of healthy volunteers who are then closely monitored. If the safety profile of the drug is favourable, then it may advance to phase 2, where a small group of volunteers with the targeted disease are given the drug to further assess safety, understand how the drug is metabolised, and begin to assess efficacy. If the safety and efficacy profile continue to prove favourable, then a drug may advance to phase 3 where a larger population of volunteers with the targeted disease are exposed to the drug for continued evaluation of safety and efficacy. It is frequently the phase 3 trials that provide the safety and efficacy data required to obtain regulatory approval for the commercial marketing of a new drug; however, overly compelling phase 1 or phase 2 trials can also meet this goal through the breakthrough therapy designation. Phase 4 trials, or postmarketing trials, occur after a drug has received regulatory approval and seek to gather additional data regarding the long-term safety profile of the drug. The common purpose of all phases of clinical trials is to assess safety, but these trials come at a cost as they are carbon intensive. We have little understanding of their magnitude nor the underlying processes that are key drivers behind a trial’s overall GHG emissions.

LCA is a standardised method 8 9 for assessing the potential environmental impacts of product systems, taking into consideration all processes related to the product or service life cycle and all relevant environmental impacts. 10 Many companies, business associations and policy-makers use this method as a decision-support tool, providing a quantitative evaluation of environmental sustainability. While LCA methodology allows for the assessment of a broad range of environmental impacts such as land use, water acidification and toxicity, this paper focuses only on measuring the GHG emissions of the clinical trial system. The goals of this project were threefold: to benchmark the GHG emissions across an array of clinical trials spanning multiple trial phases and disease areas, identify primary drivers affecting GHG emissions and inform clinical trial sponsors and those involved in designing clinical trials of potential opportunities to mitigate the impacts of these hotspots.

This new research is intended to complement an earlier publication 11 by expanding our knowledge of clinical trial LCA emissions to include larger clinical trials involving multiple clinical trial sites and countries, and the additional resources and oversight they require.

Methods

This study uses LCAs to measure the GHG emissions of a set of seven clinical trials sponsored by Johnson & Johnson Innovative Medicine that span all four phases of clinical development as well as multiple disease areas. The LCAs were inclusive of clinical site utilities and other gaps observed in earlier research.

All in-scope clinical trials were conducted between 2018 and 2023. The operational details of the trials are outlined in table 1 . The trials were intentionally selected to span different disease areas, trial phases, as well as an attempt to ensure a global representation of clinical site locations (eg, Asia Pacific, Central Asia, Europe, Latin America and North America). All trials were recently completed trials, so that we had a full accounting of all consumables and could ensure access to data such as site personnel travel between their homes and the site which wouldn’t be captured in standard trial documentation and risked loss with time.

The LCA boundary conditions excluded overhead functions such as IT services, legal, healthcare compliance and other roles critical to overall clinical research conduct and broader sponsor operations. While these functions are important, they did not directly report the time and resources they expend in support of any one clinical study making it impossible to amortise a portion of their emissions to any one trial. The exclusion of these functions aligns with the boundary conditions of more recently published research. 11 12

Included trials

We have previously investigated the GHG emissions of a phase 1 trial, TMC114FD1HTX1002. 11 We include the results of that analysis here to compare to the results with those of phase 2, 3, and 4 trials. This new research is intended to complement the earlier publication by expanding our knowledge of clinical trial LCA emissions to include larger clinical trials involving multiple clinical trial sites and countries, and the additional resources and oversight they require.

We selected trials that we perceived to represent the spectrum of potential GHG emissions, including two phase 2 trials, three phase 3 and two phase 4 trials. Further diversity was introduced by including trials from different disease areas.

VAC18193RSV3006 was a randomised, double-blind, placebo-controlled, phase 3 trial to investigate the safety and immunogenicity of the Ad26.RSV.preF based vaccine in adults 18–59 years of age who are healthy or at risk for severe respiratory syncytial virus (RSV) disease, compared with adults 65 years and above. The trial involved the administration of a single dose of either active or placebo vaccine followed by laboratory assessments at seven or more time points per patient shipped frozen to a central lab.

54767414MMY3012 was a randomised phase 3 trial intended to demonstrate that a subcutaneous administration of investigational drug JNJ-54767414 was non-inferior to intravenous administration terms of the overall response rate and maximum trough concentration. It was a global trial involving clinical sites across North American, European, Asia Pacific and Latin American regions. The average patient visited the site more than 26 times. The trial involved skeletal scans (CT scans, MRI, X-ray per local standard of care) and bone marrow aspirates to monitor disease progression and treatment response. Additionally, it involved frequent laboratory assessments with the shipment of frozen samples to central labs.

R092670PSY3016 was a double-blind, randomised, active-controlled, parallel group, phase 3 trial to assess the long-term safety and tolerability of paliperidone administration via a dose once every 6-month formulation. It was a global trial involving clinical sites in North America, European, Asia Pacific and Latin America regions. Of note to this trial is that it involved a relatively insignificant amount of central laboratory testing, as trial endpoint measures were based on psychiatric assessments and labs that could be performed locally.

77242113PSO2001 was a dose-ranging trial to evaluate the long-term clinical response of investigational drug JNJ-77242113 in participants with moderate-to-severe plaque psoriasis. It involved a broad mix of sites in the Asia Pacific, European and North American regions.

42756493BLC2002 was a phase 2 study to evaluate safety, efficacy, pharmacokinetics and pharmacodynamics of various regimens of investigational drug JNJ-42756493 in subjects with metastatic or locally advanced urothelial cancer. It was a global study with sites in Asia Pacific, Central Asia, European, Latin American and North American regions. Of note is that this trial originally started in January 2018 with trial operations outsourced to a contract research organisation (CRO). The sponsor company decided to take back management of trial operations and bring them in-house. The sponsor assumed full management of trial operations by April 2019, but this left a 14-month gap from January 2018 until April 2019 when it was under the control of the CRO. Because data around staff locations and commuting were not available during the period the CRO was in control, sponsor location and commuting data were substituted to represent how trial conduct would have appeared with the sponsor managing the trial from the beginning.

28431754DIA4032 was an open-label, single-arm, phase 4 trial to evaluate the safety and efficacy of canagliflozin and metformin hydrochloride immediate-release fixed-dose combination as an adjunct to diet and exercise to improve glycaemic control in Indian adult patients with type 2 diabetes mellitus when treatment with both canagliflozin and metformin is appropriate. It was a single-country trial involving sites in India. All laboratory samples were processed locally.

Patient and public involvement

While participants and the public were involved in the 77242113PSO2001 (ClinicalTrials.gov: NCT05364554 ), 42756493BLC2002 (ClinicalTrials.gov: NCT03473743 ), 54767414MMY3012 (ClinicalTrials.gov: NCT03277105 ), VAC18193RSV3006 (ClinicalTrials.gov: NCT05070546 ), R092670PSY3016 (ClinicalTrials.gov: NCT04072575 ) and 28431754DIA4032 (ClinicalTrials.gov: NCT04288778 ) clinical trials, the LCAs were performed as an independent retrospective analysis after the clinical trial was completed. None of the participating trial subjects were involved specifically in the LCA analysis nor was any personal identifying information from the trial subjects collected or shared.

LCA methodology

The LCA was conducted under a dynamic and iterative approach in accordance with ISO 14040/44 standards and peer reviewed by an independent third party. The system boundary of the LCA is summarised in table 2 and visually represented in online supplemental figure A .

Data sources

According to the definitions of the Life Cycle Initiative, 13 the foreground system consists of processes/activities which are under the control of the decision-maker for which an LCA is carried out. The background system consists of processes on which no or, at best indirect, influence can be exercised by the decision-maker. For this LCA, the foreground system is the activities/processes presented in online supplemental figure A . The background system is all supporting activities/processes, including electricity production, energy/heat production, transport operations and incineration processes.

For foreground data, certain key parameters (eg, distances, number of visits of participants) are provided as primary data by the case study. Foreground data such as distance and mode of transport for participants, site staff and sponsor staff were based on administered surveys and typical regional behaviour data. Participant responses were anonymised by site staff with only distance and mode of transport provided for analysis.

As no LCAs had previously been performed for any of the drug products included in the clinical trials, proxy data needed to be used. For non-biological drug products, data from the Association of British Pharmaceutical Industry tools 14 were used as a proxy for the climate change impacts of the active pharmaceutical ingredient (API) (1500 kg carbon dioxide equivalents (CO 2 e) per kg). Proxy data from the Ecoinvent v3.8 emissions-factor database were then scaled so that results for all the environmental impact categories could be generated. For biological drug products, data were sourced from Budzinski et al 15 to develop a proxy climate change impact for the API of 22.7 t CO 2 e per kg.

Other sources of foreground data were based on estimations or calculations, such as utilities consumption at clinical trial sites. While many sites expressed an interest in sustainability, most lacked the bandwidth to support our data collection efforts. For utilities consumption at the trial site and at other sites (eg, clinical management), a time-related allocation was considered where specific activity consumption data were not available. For instance, considering the number of full-time equivalent (FTE) employees working at the specific sites (for the clinical trial) in relation to the total number of FTE at the specific sites.

Background data characterising activities in the background system were drawn from the Ecoinvent Database V.3.8 16 and used as alternative data to represent foreground processes where more reliable primary data (or good estimations) are not available.

All data used in our analysis are publicly available through DRYAD. 17

LCA Method

The LCA model is created using the SimaPro V.9.5 Software system for life cycle engineering, developed by PRé Consultant BV. 18 The environmental impact of the study, expressed by the global warming potential, was calculated according to the Intergovernmental Panel on Climate Change 100 years method based in kilograms of CO 2 e. 19

Results

The TMC114FD1HTX1002 phase 1 trial was the smallest trial both in terms of number of patients (39) and sites (1) and had the smallest emissions at 17 648 kgCO 2 e. The 54767414MMY3012 phase 3 trial was not the largest trial in terms of number of participating patients (517) but had the largest number of participating sites (129) and had the largest emissions at 3,107,436 kg CO 2 e. Across all seven trials analysed, the mean emissions per patient were 3260 kg CO 2 e. When the overall trial footprints are broken down by phase, the phase 2 mean per patient was 5722 kg CO 2 e and the phase 3 mean per patient emissions were 2499 kg CO 2 e. The five largest contributors of GHG emissions were drug product (50% mean), patient travel (10% mean), travel for on-site monitoring visits (10% mean), collection and processing of laboratory samples (9% mean) and sponsor staff commuting (6% mean). Patient travel was the only consistent GHG hotspot across all seven trials, as other hotspots appeared intermittently in some trials but not others based on variations in trial design. Across the multisite phase 2, 3 and 4 trials we analysed, a combination of the observed five largest contributors to GHG emissions was responsible for no less than 79% of GHG emissions for any one trial. Onlinesupplemental tables 3 4 provide a breakdown of the major sources of GHG emissions across the trials from both total emissions and per-patient emissions perspectives, while online supplemental figure B provides a graphical representation of the primary contributors to the overall GHG emissions for the trial.

Discussion

The goal of this research is to provide a better understanding of the scale of GHG emissions from industry-sponsored clinical trials and the key drivers of those emissions. In our analysis, we observed two significant sources of GHG emissions which until recently were largely absent from previous research 20 . These were patient travel, which contributed to 10% of emissions on average in our sample with a maximum observed contribution of 29%, and clinical site operations, which contributed to 5% of emissions on average in our sample with a maximum observed contribution of 25%. Until recently, previous research has fallen into two categories: those that underestimated the GHG emissions of patient travel or those that neglected or underestimated the impact of GHG emissions from clinical trial sites.

In the former category, Lyle et al 21 assumed that participant travel to and from the site was like that of a typical general practice in the UK, with travel of 2.4 km for primary care and 17.4 km for secondary care visits. This contradicts other research 22 23 where average patient travel for clinical site visits averaged between 80 km and 83 km, a near fivefold increase in distance.

The second category of publications 24 25 excludes GHG emissions at the clinical sites under the rationale that hospitals or clinics would exist and continue to operate without the clinical trial in question, and any incremental increase in emissions from the clinical trial would be negligible. During the conduct of our research, we observed that many hospitals have dedicated clinical trial units that have separate operations from the main hospital and are dedicated to clinical trial conduct. The GHG emissions of such clinical trial units should therefore be considered very distinct and separate from the normal operations of the hospitals, as they would not exist in the absence of clinical trials.

The most comprehensive analysis found to date has been that of MacKillop et al 12 and the previous publication of the TMC114FD1HTX1002 trial. 11 MacKillop et al examined the GHG emissions of three phase 3 clinical trials sponsored by Astra Zeneca. Both analyses thoroughly assessed patient travel and clinical site utilities and sought to create a comprehensive approach that would be replicated in future clinical trial LCAs. Our research aligns with the scope and depth of these analyses.

Across the phase 2, 3, and 4 studies in our samples, we saw a common set of five activities that contributed to no less than 79% of trial emissions. Patient travel was the only consistent GHG hotspot across all seven trials, as other hotspots appeared intermittently in some trials but not others based on variations in trial design. For example, laboratory samples were a negligible source of GHG emissions in the R092670PSY3016 phase 3 schizophrenia trial. The trial was focused on new therapeutic interventions for schizophrenia, a disease that lacks any blood-based biomarkers, so the use of laboratory samples was limited to pregnancy tests and other simple laboratory measures involved assessing patient eligibility to participate in the trial and none during the actual patient treatment phase.

Comparing the LCA results of the phase 2, 3 and 4 studies in our sample to the previously published LCA results for the TMC114FD1HTX1002 phase 1 HIV trial, we see some more significant differences in the key sources of emissions. These are explained by two important differences in the design and execution of the TMC114FD1HTX1002 trial compared with the others. First, the TMC114FD1HTX1002 trial involved only a single clinical trial site, and the trial timelines for recruitment and treatment of the trial participants were well within the shelf-life of the drug products used. This provided the trial sponsor with near certainty of the amount of drug product required for the trial, and they were able to supply the drug product needs of the trial with negligible waste. In our sample of phase 2, 3, and 4 trials, all trials involved multiple clinical trial sites; and for many of the trials, the overall timelines for trial execution extended well beyond the shelf-life of the involved drug products. When clinical trials involve multiple clinical trial sites, it poses a forecasting challenge in predicting how many patients will be enrolled into the trial at each clinical trial site and the timing of that enrolment. This forces trial sponsors to maintain high levels of safety-stock inventory to counter variability, and that poses a high risk of obsolescence when actual enrolment comes in below forecast. As a result of these challenges, it has been estimated that industry-wide obsolescence is as high as 50% 26 and in some of the trials in our sample, we saw obsolescence rates in this range. The second significant difference between the TMC114FD1HTX1002 trial and the phase 2, 3 and 4 trials in our sample was that the TMC114FD1HTX1002 trial involved a dedicated phase 1 clinical trial site, with oversight provided by monitors already located on-site. This avoided the need for travel associated with on-site monitoring visits. These design differences between the TMC114FD1HTX1002 trial and the other trials in our sample allowed it to avoid the strong influence of drug product (mean 50% of GHG emissions) and on-site monitoring visit travel (mean 10% of GHG emissions) seen in the phase 2, 3 and 4 trials.

While this is the first study detailing the emissions of several different trial types, these trials were selected to show diversity and may not be representative of all trials of their type. Due to resource limitations, the number of trials analysed does not represent a statistically relevant sample size, and the LCAs themselves were limited by our data sources. Despite these limitations, for multisite studies, we believe that the five activities of drug product manufacture, packaging and distribution; patient travel; on-site monitoring visit travel; the collection, transport and processing of laboratory samples; and sponsor staff commuting between their homes and the office; reflect the primary sources of GHG emissions. Addressing these hotspot activities in both the design of new clinical trials and through targeted action to reduce their emissions presents the strongest levers for reducing the GHG emissions of clinical trials.

Conclusion

Based on our LCAs of seven clinical trials spanning all four phases of development and multiple disease areas, there are five activities that drive no less than 79% of the average clinical trial’s GHG footprint. These are drug product manufacture, packaging and distribution; patient travel; on-site monitoring visit travel; the collection, transport and processing of laboratory samples; and sponsor staff commuting between their homes and the office. Understanding the activities that drive GHG emissions in clinical trials can both guide trial designers in avoiding or minimising reliance on these activities when designing new trials and guide trial sponsors in taking targeted actions to reduce GHG emissions from these activities where their use cannot be avoided.

supplementary material

Funding: This study was sponsored and funded by Johnson & Johnson Innovative Medicine, LLC.

Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online ( https://doi.org/10.1136/bmjopen-2024-085364 ).

Provenance and peer review: Not commissioned; externally peer reviewed.

Patient consent for publication: Not applicable.

Ethics approval: The underlying clinical trials were performed in accordance with the Declaration of Helsinki and Guidelines for Good Pharmacoepidemiology Practice. All participating investigators were required to obtain full governing board approval for conducting research involving humans. Sponsor approval and continuing review were obtained through the appropriate Institutional Review Board/Ethics Committee and Health Authority channels. For academic investigative sites that did not receive authorisation to use the central IRB, full board approval was obtained from their respective governing IRBs, and documentation of approval was submitted to Johnson & Johnson Innovative Medicine, LLC before the site’s participation and initiation of any trial procedures. All registry participants provided written informed consent and authorisation before participating.

Data availability free text: [Dataset] LaRoche, Jason. 'LCA Reports for Janssen-sponsored phase-2, 3, 4 clinical studies'. Dryad, 2025. DOI: 10.5061/dryad.n5tb2rc45.

Patient and public involvement: While participants and the public were involved in the 77242113PSO2001 (ClinicalTrials.gov: NCT05364554 ), 42756493BLC2002 (ClinicalTrials.gov: NCT03473743 ), 54767414MMY3012 (ClinicalTrials.gov: NCT03277105 ), VAC18193RSV3006 (ClinicalTrials.gov: NCT05070546 ), R092670PSY3016 (ClinicalTrials.gov: NCT04072575 ), 28431754DIA4032 (ClinicalTrials.gov: NCT04288778 ) clinical trials, the life cycle assessment was performed as an independent retrospective analysis after the clinical trial was completed. None of the participating trial subjects were involved specifically in the LCA analysis nor was any personal identifying information from the trial subjects collected or shared.

References

1 De Soete W Dewulf J Cappuyns P et al Exergetic sustainability assessment of batch versus continuous wet granulation based pharmaceutical tablet manufacturing: a cohesive analysis at three different levels Green Chem 2013 15 3039 10.1039/c3gc41185k

2 De Soete W Boone L Willemse F et al Environmental resource footprinting of drug manufacturing: Effects of scale-up and tablet dosage Resour Conserv Recycl 2014 91 82 8 10.1016/j.resconrec.2014.08.002

3 Renteria Gamiz AG De Soete W Heirman B et al Environmental sustainability assessment of the manufacturing process of a biological active pharmaceutical ingredient J Chem Tech Biotech 2019 94 1937 44 10.1002/jctb.5975

4 Van der Vorst G Aelterman W De Witte B et al Reduced resource consumption through three generations of Galantamine·HBr synthesis Green Chem 2013 15 744 10.1039/c3gc36854h

5 De Soete W Jiménez-González C Dahlin P et al Challenges and recommendations for environmental sustainability assessments of pharmaceutical products in the healthcare sector Green Chem 2017 19 3493 509 10.1039/C7GC00833C

6 Debaveye S De Soete W De Meester S et al Human health benefits and burdens of a pharmaceutical treatment: Discussion of a conceptual integrated approach Environ Res 2016 144 19 31 10.1016/j.envres.2015.10.027 26544901

7 Penny T Collins M Whiting A et al Care pathways: Guidance on appraising sustainability Coalition for Sustainable Pharmaceuticals and Medical Devices (CSPM) 2015

8 International Standards Organization (ISO) ISO 14040: environmental management – life cycle assessment – requirements and guidelines International Organization for Standardization: Geneva 2006 Available https://www.iso.org/standard/37456.html

9 ISO 14044: environmental management – life cycle assessment – requirements and guidelines International Standards Organization (ISO) Geneva International Organization for Standardization 2006 Available https://www.iso.org/standard/38498.html

10 Guinée J Handbook on life cycle assessment operational guide to the ISO standards. I: LCA in perspective. IIa: guide. IIb: operational annex. III: scientific background Dordrecht Kluwer Academic Publishers 2002

11 LaRoche JK Alvarenga R Collins M et al Climate footprint of industry-sponsored clinical research: an analysis of a phase-1 randomised clinical study and discussion of opportunities to reduce its impact BMJ Open 2024 14 e077129 10.1136/bmjopen-2023-077129 PMC10806794 38216192

12 Mackillop N Shah J Collins M et al Carbon footprint of industry-sponsored late-stage clinical trials BMJ Open 2023 13 e072491 10.1136/bmjopen-2023-072491 PMC10445340 37604634

13 LCI 2021 Available https://www.lifecycleinitiative.org/resources/life-cycle-terminology-2/

14 Association of British Pharmaceutical Industry (ABPI) Carbon trust – blister pack carbon evaluation tool, version 1.2

15 Budzinski K Constable D D’Aquila D et al Streamlined life cycle assessment of single use technologies in biopharmaceutical manufacture N Biotechnol 2022 68 28 36 10.1016/j.nbt.2022.01.002 35007778

16 Ecoinvent database v3.8 n.d. Available https://ecoinvent.org/the-ecoinvent-database/data-releases/ecoinvent-3-8

17 LaRoche J LCA reports for Johnson & Johnson innovative medicine sponsored phase-2, 3, 4 clinical trials Dryad 2025 10.5061/dryad.n5tb2rc45

18 SimaPro v9.5 software, PRé Consultant BV n.d. Available https://pre-sustainability.com/

19 Intergovernmental Panel on Climate Change (IPCC) Sixth assessment report – AR6 Available https://www.ipcc.ch/assessment-report/ar6

20 Billiones R Carbon footprint of clinical trials: a high-level literature review Med Writ 2022

21 Lyle K Dent L Bailey S et al Carbon cost of pragmatic randomised controlled trials: retrospective analysis of sample of trials BMJ 2009 339 b4187 10.1136/bmj.b4187 19880529 PMC2771077

22 Borno HT Zhang L Siegel A et al At What Cost to Clinical Trial Enrollment? A Retrospective Study of Patient Travel Burden in Cancer Clinical Trials Oncologist 2018 23 1242 9 10.1634/theoncologist.2017-0628 29700209 PMC6263122

23 LabCorp A clinical development paradigm to improve patient engagement 2020 Available https://www.labcorp.com/clinical-development-paradigm-improve-patient-engagement

24 Burnett J Clarke M Darbyshire J Towards sustainable clinical trials BMJ 2007 334 671 3 10.1136/bmj.39140.623137.BE 17395948 PMC1839193

25 Subaiya S Hogg E Roberts I Reducing the environmental impact of trials: a comparison of the carbon footprint of the CRASH-1 and CRASH-2 clinical trials Trials 2011 12 31 10.1186/1745-6215-12-31 21291517 PMC3045899

26 McKinsey & Company Clinical supply chains: how to boost excellence and innovation 2021 Available https://www.mckinsey.com/industries/life-sciences/our-insights/clinical-supply-chains-how-to-boost-excellence-and-innovation

Data availability statement

Data are available in a public, open access repository.