Abstract Context The scarcity of robust long-term safety data has contributed to increasing restrictions on the use of gonadotropin-releasing hormone analog (GnRHa) treatment in transgender and gender diverse adolescents (TGDAs) in several countries. Methodological limitations of earlier studies have led to controversy and debate, underscoring the urgent need for high-quality evidence in this field Objective To present a study design using target trial emulation (TTE) to evaluate long-term safety of GnRHa treatment in TGDAs using observational data from national healthcare registries, and highlight methodological challenges. Methods We applied a stepwise TTE approach, to emulate a randomized trial using real-world data. The main methodological challenges were identified as the selection of the control group and the definition of time zero. To address these, we engaged independent experts in causal inference to review the concept design and provide feedback on potential solutions. Results and Discussion Expert feedback was systematically summarized and discussed. The resulting study design will select late-presenting TGDA individuals as the control group and defines time zero as the initiation of GnRHa treatment for the intervention group, with the equivalent age assigned as time zero in controls. Sensitivity analyses using alternative control groups and time zero definitions will be included to assess robustness. This approach allows for adjustment of confounders and aims to emulate the conditions of a randomized trial as closely as possible within the constraints of observational data. Conclusion This discussion paper serves as a methodological example of target trial emulation with observational data to study long-term safety of GnRHa in TGDA. By transparently addressing methodological challenges and incorporating expert input, this approach offers a feasible and ethically sound alternative to RCTs, and contributes to the evidence base needed for informed clinical and policy decisions in transgender healthcare.
Schuurman et al. (Tue,) studied this question.