Carbon cost of pragmatic randomised controlled trials: retrospective analysis of sample of trialsBMJ 2009; 339 doi: http://dx.doi.org/10.1136/bmj.b4187 (Published 30 October 2009) Cite this as: BMJ 2009;339:b4187
- Katy Lyle, research fellow1,
- Louise Dent, acting principal research fellow1,
- Sally Bailey, senior programme manager1,
- Lynn Kerridge, chief executive officer1,
- Ian Roberts, professor of epidemiology and public health2,
- Ruairidh Milne, director of strategy and development1
- 1NIHR Evaluation, Trials and Studies Coordinating Centre, University of Southampton, Southampton SO16 7NS
- 2London School of Hygiene and Tropical Medicine, London WC1E 7HT
- Correspondence to: K Lyle
- Accepted 1 September 2009
Objective To calculate the global warming potential, in carbon dioxide (CO2) equivalent emissions, from a sample of pragmatic randomised controlled trials.
Design Retrospective analysis.
Data source Internal data held by NIHR Evaluation, Trials and Studies Coordinating Centre.
Studies included All eligible pragmatic randomised controlled trials funded by the NIHR Health Technology Assessment programme during 2002 and 2003.
Main outcome measure CO2 equivalents for trial activities calculated with standard conversion factors.
Results 12 pragmatic randomised controlled trials involving more than 4800 participants and a wide range of technologies were included. The average CO2 emission generated by the trials was 78.4 (range 42.1-112.7) tonnes. This is equivalent to that produced in one year by approximately nine people in the United Kingdom. Commuting to work by the trial team generated the most emissions (average 21 (11.5-35.0) tonnes per trial), followed by study centres’ fuel use (18 (9.3-32.2) tonnes per trial), trial team related travel (15 (2.0-29.0) tonnes per trial), and participant related travel (13 (0-46.7) tonnes per trial).
Conclusions CO2 emissions from pragmatic randomised controlled trials are generated in areas where steps could be taken to reduce them. A large proportion of the CO2 emissions come from travel related to various aspects of a trial. The results of this research are likely to underestimate the total CO2 emissions associated with the trials studied, because of the sources of information available. Further research is needed to explore the additional CO2 emissions generated by clinical trials, over and above those generated by routine care. The results from this project will feed into NIHR guidelines that will advise researchers on how to reduce CO2 emissions.
Global warming is heavily influenced by human activity, is considered to be the biggest global health threat of the 21st century, and is already having impacts on health.1 2 The Climate Change Act,3 which outlines the government’s commitment to reduce the United Kingdom’s carbon output by 80% by 2050, became law in 2008. The National Health Service is committed to reducing its carbon output,4 5 to which clinical research contributes. This is manifest in the NHS carbon reduction strategy, which sets the NHS a target of reducing its 2007 carbon footprint by 15% by 2015.5
Total public funding for clinical research will be increased to £1.7 billion (€1.9bn; $2.8bn) by 2010.6 With more funding than ever before available for clinical research, considering its environmental impacts is increasingly important. As far as we are aware, only one previous study has attempted to quantify the environmental impact of a clinical trial. That study calculated the carbon dioxide (CO2) equivalent emissions of the CRASH trial.7 Total CO2 emissions for the CRASH trial were estimated at 629 tonnes, or 63 kg per participant. As the CRASH trial was international and recruited more than 10 000 participants, it is unlikely to be representative of pragmatic randomised controlled trials in the UK.
As a leading funder of research in the NHS, the Health Technology Assessment programme of the National Institute for Health Research (NIHR) is interested in the CO2 emissions from the research it funds. On behalf of the Health Technology Assessment programme, the NIHR Evaluation, Trials and Studies Coordinating Centre (NETSCC) completed a feasibility study to quantify the CO2 emissions of a randomised controlled trial in the Health Technology Assessment programme. The analysis was then extended to a sample of trials to determine which elements of the trials produce the most emissions. This paper presents the results of that research.
All randomised controlled trials that were funded by the Health Technology Assessment programme during 2002 and 2003 were eligible for inclusion in the study. We stipulated that the final report of the project must have been submitted by September 2008 to ensure that the trials were complete before data collection began.
We used information that is routinely held by NETSCC. The main sources of data were the project files and the published monograph, or the final report of the project if it had yet to be published. Project files are used for monitoring purposes and contain documents such as grant application forms, six monthly progress reports, extension requests, and minutes of steering group meetings.
Using the methods of the CRASH audit,7 the NHS footprinting report,5 and the greenhouse gas reporting protocol developed by the World Business Council,8 we compiled a list of elements of trials that we believed would generate CO2 emissions and a list of the activities within each element that would generate the emissions. We searched the data sources for information on these activities and any other activities not included in the list that we believed had a CO2 value. The box shows the elements and activities of trials for which we collected data.
Trial elements and activities for which data were collected
Participant related travel
Trial team commuting
Staff commuting to work
Trial team related travel
Travel for steering group meetings
Travel for data monitoring committee meetings
Trial coordinator’s travel
Travel for training
Travel to conferences
Travel for Health Technology Assessment monitoring visits
Other trial meetings/travel
Information technology equipment
Energy use in manufacture
Energy use in manufacture
Distribution and deliveries
We used standard conversion factors to convert the data into CO2 values. The information contained in the data sources was not always sufficient to allow us to do the calculations. In these instances, we made informed assumptions to supplement the information and allow it to be converted into CO2. The conversion factors and assumptions used are detailed below.
National travel—We used a conversion factor of 0.25 to calculate the CO2 emissions per kilometre travelled. This conversion factor comes from the NHS footprinting report and accounts for the direct and indirect emissions from all modes of transport.5 We used Google maps UK to calculate all distances. We assumed that staff travelled to meetings from their workplace. In the absence of detailed information about travel, we calculated CO2 from the amount of money spent on journeys. We used a conversion factor of 0.43 kg of CO2 per pound spent, which we derived from the NHS footprinting report.5
International travel—We calculated the CO2 value of any flights that were made during the trial by using a CO2 calculator available at the climate care website (www.climatecare.org/roughguides/). The tool calculates the CO2 emissions from a return flight between any two cities. We assumed that travellers departed from the nearest airport to their place of work and flew directly to the airport of the city to which they were travelling.
Primary care—We assumed the distance travelled for primary care visits to be 2.4 km. This is the average distance that patients live from their general practice, according to data from an Ipsos-MORI survey for the Department of Health (DH, personal communication, 2008).
Secondary care—We assumed the distance travelled to secondary care visits to be 17.4 km. We took this from a study using hospital episode statistics data, which found that 17.4 km was the median distance that patients travel for elective hospital visits.9
Staff commuting—We assumed the distance travelled by staff commuting to work on the trial to be 14 km, which was found to be the average distance commuted to work in the national travel survey 2006.10 We assumed that trial employees would take 40 days’ leave pro rata, including 30 days’ annual leave and 10 days’ sick leave.11
Energy consumption at study centres
We calculated energy consumption at the study centres by using conversion factors available from the Department for Food Environment and Rural Affairs.12 The amount of energy used differs according to the type of accommodation. Table 1⇓ shows the conversion factors we used for different types of accommodation, which represent the amount of energy used per square metre of office space in a year. We assumed that each employee would be allocated 10 m2 of office space, which was recommended by the Carbon Trust (Carbon Trust, personal communication, 2008).
Manufacture and delivery of trial technologies
We calculated the energy used in the manufacture and delivery of the trial technologies from the amount of funding allocated for the purchase of the technologies. This method was used in the NHS report.5 We assumed that 95% of the money allocated to trial technologies in the full proposal was used to purchase the technology and that the remaining 5% was spent on delivering the technology. We did not find any evidence on which to base these figures. Table 2⇓ shows the conversion factors used to calculate the emissions for different types of technology.
Information technology equipment
We calculated the CO2 emissions relating to the production of information technology equipment for the trials from the amount of money spent on the purchase of such equipment during the trial. We used a conversion factor of 0.21, which came from the NHS footprinting report.5
In our analysis, we calculated the total CO2 emissions of the trials, including those that would have occurred as part of routine practice as well as those solely attributable to the trial, as we could not separate these emissions in our data. One could argue that staff commuting to work should not be included in the analysis, as the researchers would still need to commute to work even if they were working in another field. To account for this, we did a sensitivity analysis to calculate the CO2 emissions of the trials excluding those generated by staff’s commuting.
Twelve randomised controlled trials met the inclusion criteria (fig 1⇓), involving more than 4800 participants and a wide range of technologies, including medical devices, pharmaceuticals, and psychological therapies. The average number of participants was 402, the average cost per participant was approximately £2500, and the average number of study centres was 16.
We calculated the total CO2 emissions generated by the 12 trials as 941.2 tonnes, which is equivalent to that produced in one year by about 103 people in the UK.13 Table 3⇓ shows the CO2 emissions for each trial, which ranged from 42.1 tonnes to 112.7 tonnes.
Our analysis shows that the average amount of CO2 produced by the trials was 78.4 tonnes, equivalent to that produced in one year by about nine people in the UK.13 The average amount of CO2 produced in one year of a trial was 18.1 (range 8.9-30.1) tonnes. On average, the trials produced 306.2 (80.0-883.7) kg of CO2 per participant, 0.1 (0.07-0.15) kg of CO2 per pound spent, and 5.6 (3.8-9.5) tonnes of CO2 per whole time staff member.
Across all of the trials, commuting to work by trial teams accounted for the most CO2 emissions (26%), followed by use of fuel at study centres (23%) and trial team related travel (19%). Participants’ travel accounted for 16% of emissions, trial technologies accounted for 14% of emissions, and information technology equipment accounted for 2% of emissions. The amount of CO2 that was generated from different trial elements varied between the randomised controlled trials (fig 2⇓).
The CO2 emissions generated by participant related travel ranged from 0 to 46.7 tonnes (average 12.6 tonnes per trial). This variation can in part be explained by the difference in the number of visits that were required in the different trials, which ranged from none to 36. Additionally, participants’ visits took place in a variety of locations in secondary care, primary care, and the community, which corresponds to the use of different distances in our calculations.
The amount of CO2 produced by trial team related travel ranged from 2.0 to 29.0 tonnes (average 14.7 tonnes per trial), owing to the different activities in the trials and the organisation of the trials. For example, the number of steering group meetings needed for the trials ranged from three to 10, and the number of people attending ranged from four to 13. The number of site visits the trial coordinators made during the trials also varied greatly. The average number of conferences that the trial teams attended ranged from one to 21 (total 78, of which 16 were international).
The CO2 generated from use of fuel at the study centres ranged from 9.3 to 32.2 tonnes (average 18.1 tonnes per trial), and that generated from the trial team commuting to work ranged from 11.5 to 35.0 tonnes (average 20.9 tonnes per trial). These calculations were based on the number of staff working on the trial, the length of time they were employed on the trial, and the type of accommodation they occupied.
The CO2 emissions related to the trial technology ranged from 0 to 35.9 tonnes (average 10.7 tonnes per trial), and emissions related to information technology equipment ranged from 0.4 to 2.9 tonnes (average 1.5 tonnes per trial). The CO2 emissions from these elements are directly proportional to the cost of the technologies.
A sensitivity analysis excluding the emissions generated by staff commuting to work found that the average amount of CO2 produced by the trials was 57.6 (27.9-80.3) tonnes, which is equivalent to that produced in one year by about six people in the UK.13 On average, the trials produced 13.3 (5.9-20.7) tonnes of CO2 a year, 218.0 (53.2-561.9) kg of CO2 per participant, 0.08 (0.06-0.12) kg of CO2 per pound spent, and 4.2 (2.4-8.1) tonnes of CO2 per whole time staff member. Table 3⇑ shows the individual results of the sensitivity analysis for each trial.
We found that most of the CO2 emissions produced by pragmatic randomised controlled trials come from trial related travel (including commuting by the trial team and participant and trial team related travel) and use of fuel at the study centres. The trials varied in the proportion of CO2 generated by different elements of the trial.
Strengths and weaknesses
This is the first study we are aware of that has attempted to calculate the CO2 emissions from a sample of randomised controlled trials involving a range of technologies and groups of participants. The trials took place in various settings and were based in multiple centres throughout the UK. Perhaps a more accurate way to calculate the CO2 emissions of a trial would be to prospectively collect information on activities with a CO2 value for the duration of the trial. Although this might provide a more accurate estimate, it would be very burdensome on time and resources. We therefore believe that, although our analysis may not represent the exact CO2 emissions produced by these trials, it does provide a good guide as to the level of CO2 emitted and the trial related activities that generate the most emissions.
The study has some limitations. We did the study retrospectively using data that was not designed for this purpose, which had several implications. Use of project files and reports meant that we attributed CO2 values to the trial related activities reported, rather than all of the activities executed. Therefore, the results of this research are likely to be an underestimate of the total CO2 emissions associated with the trials studied. The information we needed to calculate the CO2 value was not always available, which meant that data had to be supplemented with assumptions. The amount of information contained within the files varied between projects, resulting in a variable number of assumptions being used for each trial.
We excluded two eligible trials from the study, as the information available was insufficient to allow us to do the calculations. We know that these studies involved a lot of travel, and they may have had a larger CO2 value than the other studies that were included.
Comparisons with other studies
The only previous work in this area was a carbon audit of the CRASH trial,7 which was estimated to have produced 629 tonnes of CO2. Our estimate of the average CO2 emission per trial was lower at 78 tonnes. However, the CRASH trial included 10 000 participants, whereas the average number of participants per trial in our sample was 402. The average CO2 emission per participant in our study was 306 kg, whereas the corresponding figure in the CRASH trial was 63 kg, suggesting that per participant the CRASH trial was much less carbon intensive. The CRASH trial was a large simple trial of an acute care intervention. It involved no extra tests, and data collection was minimal and based entirely on the patients’ clinical notes. Statistical data monitoring reduced the need for multiple site visits, which would have involved considerable amounts of extra travel. The CRASH trial did not involve any extra travel by participants, whereas most of the trials in our study involved several visits. These factors may account for the difference in CO2 emissions per participant between our study and the CRASH trial. Nevertheless, the proportions of CO2 emissions from different trial elements were similar between the two studies. Two of the most CO2 intensive elements of the CRASH study were use of fuel at the coordinating centre and trial related travel, which is consistent with our findings.
Meaning of study
The government is committed to drastically reducing CO2 emissions,3 and all sectors must contribute to this reduction. Clinical research contributes to the NHS’s carbon footprint, and the research community should be aware of the CO2 emissions generated by their trials. This research shows that most CO2 emissions from randomised controlled trials are generated in areas in which steps could be taken to reduce them. When aiming to minimise CO2 emissions, trialists and funders should focus particular attention on the emissions generated by participants’ travel and travel by the trial team, which can be reduced with simple changes. The number of visits by participants that are required for a trial should be kept to a minimum. Careful consideration of each visit and its purpose may show that the trial could be completed with fewer visits or that several visits could be combined. Alternative methods that do not require face to face encounters, such as telephone interviews, may be available for some assessments of participants. Some of the trials in our study have shown that simple steps can reduce the amount of travelling staff do during the trial and therefore lower CO2 emissions. For example, some trials held meetings by teleconference to reduce travel; in one study, meetings were held in a location that was within walking distance for people attending. Technologies such as teleconferencing or videoconferencing could be used more and have the potential to reduce the carbon output from trials. This research will feed into NIHR guidelines that will expand on these recommendations and advise researchers on how to reduce CO2 emissions from their trials and other research.
Unanswered questions and further research
In our analysis, we were unable to separate excess CO2 emissions from those that occur during routine care. Further work is needed to explore the additional CO2 emissions generated by randomised controlled trials over and above those generated by routine care. Our analysis included a range of randomised controlled trials, but we did not explore the effect of different research designs on CO2 output, so further research is needed to explore the CO2 emissions from other study designs. This study has developed methods to calculate the CO2 emissions associated with trials. These methods could be expanded on and developed into a tool for use by researchers and funders to allow the carbon cost of a trial to be considered at the planning stage.
What is already known on this topic
One previous study examined the CO2 emissions of a large, international randomised controlled trial
This was not representative of pragmatic trials carried out in the NHS
What this study adds
CO2 emissions in randomised controlled trials were mainly generated from areas such as trial related travel, for which steps can be taken to reduce them
The results from this research will feed into NIHR guidelines, which will advise researchers on how to reduce CO2 emissions from their trials and other research
Cite this as: BMJ 2009;339:b4187
We thank the following people for their hard work in collecting data for the project: Syd Anstee, Lisa Douet, Marina Malaffo, Claire Rosten, and Tannaze Tinnati. We also thank David Pencheon for his advice and support.
Contributors: KL and LD were involved in all stages of the study. SB assisted with data collection and was a member of the steering group. LK and IR contributed to the design of the project through their involvement in the steering group. RM was a member of the steering group and was involved in revising the article critically and approving it for publication. All authors critically reviewed and agreed the final draft. KL and LD are the guarantors.
Funding statement: This study was carried out by NIHR Evaluations Trials and Studies Coordinating Centre (NETSCC), which manages the Health Technology Assessment programme. KL, LD, SB, LK, and RM are employed by NETSCC. No additional funding was sought for this research
Competing interests: None declared.
Ethical approval: Not needed.
Data sharing: No additional data available.
This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode.