Zoccali Carmine, Tripepi Giovanni
Renal Research Institute, New York, NY, USA.
Institute of Molecular Biology and Genetics (Biogem), Ariano Irpino, Italy.
J Nephrol. 2025 Jan;38(1):11-23. doi: 10.1007/s40620-024-02158-5. Epub 2024 Nov 27.
Trial emulation, also known as target trial emulation, has significantly advanced epidemiology and causal inference by providing a robust framework for deriving causal relationships from observational data. This approach aims to reduce biases and confounding factors inherent in observational studies, thereby improving the validity of causal inferences. By designing observational studies to mimic randomized controlled trials (RCTs) as closely as possible, researchers can better control for confounding and bias. Key components of trial emulation include defining a clear time-zero, simulating random assignment using techniques like propensity score matching and inverse probability treatment weighting, assessing group comparability by standardized mean differences and establishing a clear comparison strategy. The increasing availability of large-scale real-world databases, such as research cohorts, patient registries, and hospital records, has driven the popularity of target trial emulation. These data sources offer information on patient outcomes, treatment patterns, and disease progression in real-world settings. By applying the principles of target trial emulation to these rich data sources, researchers can design studies that provide robust causal inferences about the effects of interventions, informing clinical guidelines and regulatory decisions. Despite its advantages, trial emulation faces challenges like data quality, unmeasured confounding, and implementation complexity. Future directions include integrating trial emulation with machine learning techniques and developing methods to address unmeasured confounding. Overall, trial emulation represents a significant advancement in epidemiology, offering a valuable tool for deriving accurate and reliable causal inferences from observational data, ultimately improving public health outcomes.
试验模拟，也称为目标试验模拟，通过提供一个强大的框架来从观察性数据中推导因果关系，极大地推动了流行病学和因果推断的发展。这种方法旨在减少观察性研究中固有的偏差和混杂因素，从而提高因果推断的有效性。通过设计观察性研究以尽可能紧密地模仿随机对照试验（RCT），研究人员可以更好地控制混杂和偏差。试验模拟的关键组成部分包括定义明确的时间零点，使用倾向得分匹配和逆概率处理加权等技术模拟随机分配，通过标准化均值差异评估组间可比性以及建立明确的比较策略。大规模真实世界数据库（如研究队列、患者登记册和医院记录）的日益普及推动了目标试验模拟的流行。这些数据源提供了真实世界环境中患者结局、治疗模式和疾病进展的信息。通过将目标试验模拟的原则应用于这些丰富的数据源，研究人员可以设计出能够对干预效果提供有力因果推断的研究，为临床指南和监管决策提供依据。尽管试验模拟具有优势，但它也面临数据质量、未测量的混杂因素和实施复杂性等挑战。未来的方向包括将试验模拟与机器学习技术相结合，以及开发解决未测量混杂因素的方法。总体而言，试验模拟代表了流行病学的一项重大进展，为从观察性数据中得出准确可靠的因果推断提供了一个有价值的工具，最终改善公共卫生结果。
