Table 1.

Summary of key components of the target trials that, in principle, can be emulated in observational data using methods that adjust for measured confounders compared to classical instrumental variable methods

Target trial feature	Target trials that can be emulated using methods that adjust for measured confounders	Target trials that can be emulated using classical instrumental variable methods
Eligibility criteria	No specific restrictions	When estimating a “local” effect, being a “complier” is an additional eligibility criterion (note: this further criterion would be impossible to evoke in a real trial)
Treatment strategies	No specific restrictions (can emulate trials of point interventions or sustained treatment strategies)	Can only emulate trials of point interventions (including initiating a sustained treatment strategy) Ignoring some of the indicated treatment strategies existing in the data requires stronger assumptions
Assignment procedures	Can only emulate random assignment without blinding	Can only emulate random assignment without blinding
Follow-up period	No specific restrictions	No specific restrictions
Outcome	No specific restrictions (frequently, blind ascertainment cannot be emulated)	No specific restrictions (frequently, blind ascertainment cannot be emulated)
Causal contrast of interest	No specific restrictions (can be used to estimate the effect of initiating or sustaining a sustained treatment strategy, or the effect of a point intervention)	Can be used to estimate the effect of a point intervention (including the effect of initiating a sustained treatment strategy), but generally not to estimate the per-protocol effect of sustaining a treatment strategy