Table 1 Examples of types of data-sharing models

Usually to address one or more specific research questions

Negotiate provision of IPD from trialists

IPD provided directly to data requestor

Usually pseudonymised

Usually fully collaborative

Single-collaboration IPD reviews [40, 41]

Retrospective:

Assembly of multiple existing IPD databases within a topic area to address new questions. May include new trials as they become available

Negotiate provision of IPD from trialists

IPD provided directly to data requestor

Usually pseudonymised

Collaborative

ACCENT [14]; ARCAD [21]; EBCTCG [11]; ICECaP M0 [20]

Prospective:

Plan to assemble IPD from all trials within a topic area, whilst most of the trials are still ongoing or yet to report, to address planned questions or new questions arising

STOPCAP M1 [24]

Data available on demand with unrestricted access

Data available to download online (usually only registration is required)

IPD provided directly to data requestor; usually via direct download

Usually fully anonymised

Not usually collaborative

freeBIRD [34]; International Stroke Trial (IST) database [42]

Moderated access

Subject to the approval of a formal research proposal

IPD provided directly to data requestor

May be fully anonymised or pseudonymised

Potential to be collaborative

CSDR [37]; YODA [38]; VIVLI [39]; NCTN/NCORP [36]

Moderated and restricted (“locked-box”) access

Subject to the approval of a formal research proposal

IPD available only within secure environment

May be fully anonymised or pseudonymised

Not usually collaborative

CSDR [37]; YODA [38]; VIVLI [39]; NCTN/NCORP [36]

Analyses taken to the data

Subject to approval of a formal research proposal

IPD interrogated remotely; question(s) and/or analysis software provided by the data requestor

Not applicable— no direct access to data

Not usually collaborative

DataSHIELD [35]