Miguel HernanMay 18, 2024

With #TargetTrial emulation becoming increasingly popular, it's important to understand what it can and cannot do.

In this podcast with JAMA Statistical Editor Roger Lewis, I discuss our commentary "Target Trial Emulation: A Framework for Causal Inference From Observational Data"
https://jamanetwork.com/journals/jama/article-abstract/2799678

We talk about how explicit emulation of a target trial can 1) improve causal inference from observational databases and 2) extend the inferences from randomized trials.
https://edhub.ama-assn.org/jn-learning/audio-player/18877235

Target Trial Emulation for Causal Inference From Observational Data

This Guide to Statistics and Methods describes the use of target trial emulation to design an observational study so it preserves the advantages of a randomized clinical trial, points out the limitations of the method, and provides an example of its use.

Show threadMiguel HernanJul 7, 2023

2/
We described this bias with simulations + real data.

We show that a vaccine booster will be associated with higher reinfection even if the booster has no harmful effect.
https://www.bmj.com/content/381/bmj-2022-074404.full

The good news: Preventing this self-inflicted #selectionbias is easy: Specify the #TargetTrial that the observational analysis tries to emulate.

Here the (unethical) target trial requires forced infection, so observational analyses will be biased unless they adjust for susceptibility to infection.

The imprinting effect of covid-19 vaccines: an expected selection bias in observational studies

Recent observational studies have found a higher risk of reinfection with the omicron variant of SARS-CoV-2 in people who received a third covid-19 booster dose. This finding has been interpreted as evidence of immune imprinting of covid-19 vaccines. This article proposes an alternative explanation: that the increased risk of reinfection in individuals vaccinated with a booster compared with no booster is the result of selection bias and is expected to arise even in the absence of immune imprinting. To clarify this alternative explanation, this article describes how previous observational analyses were an attempt to estimate the direct effect of vaccine boosters on SARS-CoV-2 reinfections—an effect that cannot be correctly estimated with observational data. Causal diagrams (directed acyclic graphs), data simulations, and analysis of real world data are used to illustrate the mechanism and magnitude of this bias, which is the result of conditioning on a collider. Antigenic variation in the SARS-CoV-2 omicron variant and subvariants are substantial compared with previous variants and covid-19 vaccines used until September 2022. The concerns are that past exposure to previous variants—through infection or vaccination—could alter the immunological response to an omicron related infection in such a way that the immune response to successive omicron infections would be impaired.123 This so-called “immune imprinting hypothesis” has been used to suggest that a vaccine booster in individuals who later are infected with omicron increases the risk of a second omicron infection.45 If this effect of immune imprinting truly exists, recommendations for additional vaccine doses may need to be re-evaluated. ### Summary points

The BMJ
Miguel HernanApr 20, 2023

For multidrug-resistant #tuberculosis, the optimal duration of bedaquiline treatment is unknown.

Our #TargetTrial emulation: Bedaquiline for more than 6 months doesn't increase the probability of success.

Cloning+censoring+#IPweighting

https://www.atsjournals.org/doi/10.1164/rccm.202211-2125OC

Luisa N. BorrellMar 22, 2023

#TargetTrial and observational data for cancer research. #epitwitter #epiverse
---
RT @_MiguelHernan
The European Organisation for Research and Treatment of Cancer (@EORTC) endorses #TargetTrial emulation with #RWD.

An organization with unmatched expertise in the design of #randomizedtrials embraces their explicit emulation when using observational data.
https://www.ejcancer.com/article/S0959-8049(23)00149-1/fulltext
https://twitter.com/_MiguelHernan/status/1638259606072750094

Defining the role of real-world data in cancer clinical research: the position of the European Organisation for Research and Treatment of Cancer

The emergence of the precision medicine paradigm in oncology has led to increasing interest in the integration of real-world data (RWD) into cancer clinical research. As sources of real-world evidence (RWE), such data could potentially help address the uncertainties that surround the adoption of novel anticancer therapies into the clinic following their investigation in clinical trials. At present, RWE-generating studies which investigate antitumor interventions seem to primarily focus on collecting and analyzing observational RWD, typically forgoing the use of randomization despite its methodological benefits.

European Journal of Cancer
Miguel HernanFeb 2, 2023

Observational studies are often the only option to
estimate effects of interventions during #pregnancy.

Thinking of emulating a #TargetTrial during pregnancy using healthcare databases?

Here we review challenges and propose solutions to common biases 👇
https://pubmed.ncbi.nlm.nih.gov/36722806/

Emulating a Target Trial of Interventions Initiated During Pregnancy with Healthcare Databases: The Example of COVID-19 Vaccination - PubMed

Some biases may be unavoidable in observational emulations, but others are avoidable. For instance, immortal time bias can be avoided by aligning the start of follow-up with the gestational age at the time of the intervention, as we would do in the target trial. Explicitly emulating target trials at …

PubMed
C. Harris FloudasDec 28, 2022

#oncology

RT @MiguelHernan: 1/

Problem: The best therapies for advanced pancreatic #cancer (Folfirinox, GN) haven't been compared in a head-to-head randomized trial due to cost and commercial barriers.

We emulated a #TargetTrial using observational data: Folfirinox is a bit better

https://www.sciencedirect.com/science/article/pii/S1047279722003088?dgcid=coauthor

Show threadMiguel HernanDec 28, 2022

Because randomized trials are great only when they exist.

Without trials, high-quality observational data + explicit #TargetTrial emulation is our best chance to help

people make the treatment decisions they must make

until we convince companies/funders to pay for the trial.

Miguel HernanDec 28, 2022

1/

Problem: The best therapies for advanced pancreatic #cancer (Folfirinox, GN) haven't been compared in a head-to-head randomized trial due to cost and commercial barriers.

We emulated a #TargetTrial using observational data: Folfirinox is a bit better

https://www.sciencedirect.com/science/article/pii/S1047279722003088?dgcid=coauthor

Jack Iwashyna 🫁Dec 13, 2022

I just spent the last hour with a brilliant young anesthesiologist imagining potential good questions in management of #IntraoperativeHypotension in #CardiacSurgery

I used @SPT_MD's example of specification and emulation of a #TargetTrial to help really clarify the questions of interest, and let me tell you, @SPT_MD's way of thinking is great

Miguel HernanDec 13, 2022

1/
Our summary of #TargetTrial emulation is out.

We illustrate it with another observational success:

A survival benefit of #tocilizumab in critically ill #COVID19 patients was quantified using observational data before a randomized trial confirmed it.
👇

https://jamanetwork.com/journals/jama/fullarticle/2799678

Target Trial Emulation for Causal Inference From Observational Data

This Guide to Statistics and Methods describes the use of target trial emulation to design an observational study so it preserves the advantages of a randomized clinical trial, points out the limitations of the method, and provides an example of its use.