Here's an interesting article from JAMA Psychiatry:
Estimating Treatment Effects From Observational Data—Calling It a Target Trial Does Not Make It One Krista F. Huybrechts, MS, PhD1,2; Sonia Hernández-Díaz, MD, DrPH2
Find it here. It might come with an access version here.
- See also NEJM, Hubbard 2024, on "target trial emulation." - here.
- And, Hernan, the target trial framework, causal inference from observational data, Ann Intern Med 2025 here. (My blog re Hernan here.)
- There's also Wilson, Real World Data - Trials to Practice, Lancet 2024 here.
- And Wang, Emulating RCTs with Databases, JAMA 2023, here. (Op Ed, Sheldrick, here.)
- Wang has already been cited 241 times.
Find an AI summary of HUYBRECHTS below.
Find an integrated AI discussion of all five or six papers as a sidebar - here.
###
"Calling It a Target Trial Doesn’t Make It One" — Huybrechts & Hernández-Díaz (JAMA Psychiatry, 2025)
This editorial underscores that simply labeling an observational study as a "target trial emulation" does not ensure its rigor. [Ha!]
The authors explain how true target trial emulation requires careful alignment of key elements: eligibility, treatment assignment, and the start of follow-up ("time zero")—just as in a randomized controlled trial (RCT).
They warn of immortal time bias, a design flaw common in real-world data studies, where periods in which outcomes cannot occur are improperly handled, leading to inflated treatment benefits. For instance, including prevalent users of a drug (rather than new users) can bias results, as only survivors remain in the sample. [patients who tolerate the drug, for example]
Target trial emulation is a valuable framework, but only when fully applied: researchers must explicitly define the hypothetical trial’s protocol and transparently map each component using real-world data.
Otherwise, the label becomes a veneer over flawed methods. The takeaway: design trumps terminology, and emulation demands rigorous planning—not just a rhetorical nod to RCTs.
