Article type
Year
Abstract
Background
Rapid-learning health systems are underpinned by timely sharing of data and synthesis of existing evidence. Individual participant data (IPD) meta-analysis (MA) allows robust evaluation of treatment effects and offers numerous advantages both clinically and statistically over aggregate data MA. Acquiring IPD from triallists has become more time consuming since the introduction of the General Data Protection Regulation (GDPR) in 2018. IPD may be regarded as incompatible with the principles of rapid-learning health systems.
Objectives
To present an example of the timeline required for collection of IPD for a 2-year UK NIHR-funded evidence synthesis hosted by 2 academic institutions.
Methods
Systematic review and IPD MA of placebo-controlled randomised controlled trials of testosterone therapy in people with low testosterone levels. Authors of all eligible trials were contacted to request IPD. Those interested were sent a GDPR-compliant data sharing agreement to be signed by a representative of the respective institution. The initial length of the agreement (1939 words, 6 pages) was prohibitive so it was revised to a more manageable length (1423 words, 4 pages). Nevertheless, various international institutions requested further specific amendments to the agreement, each of which required perusal by the contract departments of the 2 host universities and added to the original timeline.
Results
Thirty-five eligible trials were identified in the current literature. At 12 months since the start of the IPD study (September 2018), signed data sharing agreements (and datasets) from 5 collaborators had been received. At 18 months, 6 further signed agreements had been received, providing data from around one-third of the total number of participants enrolled (lower than the 60% recommended for robust MA of IPD). Two of the 11 received datasets did not have a corresponding signed agreement (subsequently obtained); one was not anonymised and was deleted and re-requested. At 18 months, negotiations regarding 2 further data sharing agreements were still ongoing. Another data sharing agreement was refused due to concerns about its legality. A further triallist, who had initially agreed to provide IPD, did not return the data sharing agreement or respond to follow-up emails.
Conclusions
Rapid-learning health systems may benefit from the advantages conferred by the robustness of IPD analyses. However, the process of obtaining data from international investigators by consenting to a data sharing agreement can be prohibitive in terms of time required for data collection and synthesis. Early recruitment of potential sources of IPD is recommended, including both eligible investigators and their respective contract department, in order to identify and address potential issues in an effective and timely manner.
Patient or healthcare consumer involvement
Opportunities may be missed of full utilisation of patient data and the potential benefits to healthcare consumers and policy makers.
Rapid-learning health systems are underpinned by timely sharing of data and synthesis of existing evidence. Individual participant data (IPD) meta-analysis (MA) allows robust evaluation of treatment effects and offers numerous advantages both clinically and statistically over aggregate data MA. Acquiring IPD from triallists has become more time consuming since the introduction of the General Data Protection Regulation (GDPR) in 2018. IPD may be regarded as incompatible with the principles of rapid-learning health systems.
Objectives
To present an example of the timeline required for collection of IPD for a 2-year UK NIHR-funded evidence synthesis hosted by 2 academic institutions.
Methods
Systematic review and IPD MA of placebo-controlled randomised controlled trials of testosterone therapy in people with low testosterone levels. Authors of all eligible trials were contacted to request IPD. Those interested were sent a GDPR-compliant data sharing agreement to be signed by a representative of the respective institution. The initial length of the agreement (1939 words, 6 pages) was prohibitive so it was revised to a more manageable length (1423 words, 4 pages). Nevertheless, various international institutions requested further specific amendments to the agreement, each of which required perusal by the contract departments of the 2 host universities and added to the original timeline.
Results
Thirty-five eligible trials were identified in the current literature. At 12 months since the start of the IPD study (September 2018), signed data sharing agreements (and datasets) from 5 collaborators had been received. At 18 months, 6 further signed agreements had been received, providing data from around one-third of the total number of participants enrolled (lower than the 60% recommended for robust MA of IPD). Two of the 11 received datasets did not have a corresponding signed agreement (subsequently obtained); one was not anonymised and was deleted and re-requested. At 18 months, negotiations regarding 2 further data sharing agreements were still ongoing. Another data sharing agreement was refused due to concerns about its legality. A further triallist, who had initially agreed to provide IPD, did not return the data sharing agreement or respond to follow-up emails.
Conclusions
Rapid-learning health systems may benefit from the advantages conferred by the robustness of IPD analyses. However, the process of obtaining data from international investigators by consenting to a data sharing agreement can be prohibitive in terms of time required for data collection and synthesis. Early recruitment of potential sources of IPD is recommended, including both eligible investigators and their respective contract department, in order to identify and address potential issues in an effective and timely manner.
Patient or healthcare consumer involvement
Opportunities may be missed of full utilisation of patient data and the potential benefits to healthcare consumers and policy makers.