Progress of Zelen’s design and its application value in clinical research

Objective To systematically reviewed the progress of Zelen’s design and its modifications in clinical research and clarified its methodological elements, advantages, and limitations. Methods A systematic literature search was conducted for Zelen’s design from databases. The data were extracted. Results One hundred and twenty-four trials were included. The dominant disease in this design was mental disorders, followed by osteoarthrosis diseases, cancer, cardiovascular diseases, and others. Regarding types of consent, more than half of the trials used a double-consent (71, 57.26%)，and 42 used a single-consent. Eleven trials used a modified Zelen’s design nested within an observational study. This design used a two-stage informed consent. Stage 1, patients were invited to participate in a cohort study; Stage 2, patients randomized in the experimental group were informed of the allocation result and asked whether they would like to follow the treatment. Five trials used the McNulty-Zelen design, which could be applied in cluster randomized controlled trials and overcome the potential bias of the Hawthorne effect. Intention-to-treat analysis was the main population used in Zelen’s design. Conclusion Zelen’s design has a broad application in the foreground in clinical trials. It could also be used to adapt to research needs by combining with various observational studies. Zelen’s design offers unique advantages in reducing recruitment difficulty, improving patient compliance, and minimizing bias. Although the randomization of patients without their prior consent raises potential ethical concerns, these can be addressed through methods such as nested observational studies or supplementary informed consent. In real-world applications of Zelen’s design, it is necessary to design a reasonable informed consent strategy and data statistical analysis method according to the research context. Attention must be paid to the impact of sample size, group shifting and selection of dataset on the results, to improve the interpretability and accuracy of the results.