Causal inference methods for evaluating comparative effectiveness: target trial emulation

Causal inference is grounded in contrasts between potential outcomes under alternative interventions. Randomized trials are the reference standard for estimating average causal effects because randomization renders treatment assignment independent of potential outcomes, eradicating confounding by design. However, many clinically important questions in rheumatology cannot feasibly be addressed through randomized experiments due to practical, ethical, or temporal constraints. In such settings, observational data can inform decisions. This paper argues that principles derived from randomized trials should continue to anchor causal reasoning and proposes the target trial framework as a structured approach to strengthen causal inference from observational data. By explicitly specifying the protocol of the hypothetical randomized trial that would answer the question and emulating it using real-world (observational) data, investigators can clarify eligibility criteria, treatment strategies, time zero, outcomes, causal contrasts, and identifying assumptions. By describing how each part of the target trial is emulated in observational data, they increase transparency, clarify the causal estimand, and make assumptions, limitations, and sources of bias explicit. This design-based perspective helps prevent common biases, including immortal time bias, selection of prevalent users, and inappropriate conditioning on post-treatment variables, and aligns reporting with clearly defined causal contrasts and effect measures. Ultimately, credibility of causal inference from observational data depends on whether the assignment mechanism is plausibly reconstructed, design choices are made without access to outcome data, and analyses target explicitly defined causal contrasts under stated assumptions, all of which are explicitly addressed within the target trial emulation framework.