Abstract
Objectives
To synthesise evidence on carbon dioxide equivalent (CO
₂
-eq) emissions associated with conducting clinical research, appraise its quality, explore variation and identify emission hotspots.
Design
Systematic review and quantitative synthesis.
Data sources
MEDLINE, Embase, Web of Science, HealthcareLCA, medRxiv and bioRxiv were searched from inception to November 2025 without language restrictions, with backward and forward citation searching.
Eligibility criteria
Original studies estimating CO
₂
-eq emissions attributable to clinical research were eligible, regardless of study design or assessment method.
Data extraction and synthesis
Two reviewers independently selected studies, extracted data on carbon-footprint assessments and the clinical studies they covered, and appraised the methodological and reporting quality of the assessments using a customized checklist. Emissions per clinical study were summarized descriptively by study and assessment characteristics and grouped into 10 activity categories to identify emission hotspots. Exploratory multivariable linear regression examined factors associated with log-transformed total CO₂-eq emissions per clinical study, accounting for clustering of clinical studies within assessments.
Ten carbon footprint assessments covering 45 clinical studies were included; six assessments were rated as high quality and four as medium quality. Emissions ranged from 8.8 to 3107 tCO₂-eq per clinical study, with a median of 78.0 tCO₂-eq (IQR 54.9-171.0). Descriptively, emissions per clinical study increased with the number of countries and sites and were higher when estimated using life cycle assessment (LCA) than non-LCA methods. In exploratory regression analyses, the number of countries, number of sites and assessment method accounted for the largest proportions of variation after adjustment for other factors (partial R²: 12.8%, 10.7% and 9.8%, respectively), although none of the associations was statistically significant. Meetings and travel, followed by clinical study unit operations, were the most frequent emission hotspots, although activity-level reporting was incomplete in several studies.
Conclusions
Descriptively, carbon-footprint estimates increased with the geographical scale of clinical studies and varied across assessment methods, although no independent, statistically significant associations were established. Meetings and travel, as well as clinical study unit operations, were most frequently identified as emission hotspots, offering potential targets for mitigation, although hotspot identification was limited by incomplete activity-level reporting. As estimating emissions from clinical studies is increasingly encouraged, a standardized framework for prospective calculation could improve the reliability and comparability of estimates, facilitate hotspot identification, distinguish variation attributable to study characteristics from that arising from assessment methods, and support the systematic integration of carbon considerations and mitigation measures into study design and conduct.