Review Articles

How to implement the ‘one patient, one vote’ principle under the framework of estimand

Naitee Ting

Department of Biostatistics and Data Sciences, Boehringer Ingelheim Pharmaceuticals, Inc, Danbury, CT, USA

Pages | Received 28 Nov. 2022, Accepted 30 Dec. 2022, Published online: 15 Apr. 2023,
  • Abstract
  • Full Article
  • References
  • Citations

The scientific foundation of a modern clinical trial is randomization – each patient is randomized to a treatment group, and statistical comparisons are made between treatment groups. Because the study units are individual patients, this ‘one patient, one vote’ principle needs to be followed – both in study design and in data analysis. From the physicians' point of view, each patient is equally important, and they need to be treated equally in data analysis. It is critical that statistical analysis should respect design and study design is based on randomization. Hence from both statistical and medical points of view, data analysis needs to follow this ‘one patient, one vote’ principle. Under ICH E9 (R1), five strategies are recommended to establish ‘estimand’. This paper discusses how to implement these strategies using the ‘one patient, one vote’ principle.

  • EMA (2011). Guideline on missing data in confirmatory clinical trials. European Medicine Agency. Retrieved from 
  • ICH E9 (1998). Statistical principles for clinical trials. Retrieved from 
  • ICH E9 (R1) (2019). Addendum on estimands and sensitivity analysis in clinical trials to the guideline on statistical principles for clinical trials. Retrieved from
  • Mallinckrodt, C. H., Lane, P. W., Schnell, D., Peng, Y., & Mancuso, J. P. (2008). Recommendations for the primary analysis of continuous endpoints in longitudinal clinical trials. Drug Information Journal42(4), 303–319. 
  • Rubin, D. B. (2008). Multiple imputation after 18+ years (with discussion). Journal of the American Statistical Association91(434), 473–489. 
  • Shao, J., & Zhong, B. (2003). Last observation carry-forward and last observation analysis. Statistics in Medicine22(15), 2429–2441. 
  • Shao, J., & Zhong, B. (2006). On the treatment effect in clinical trials with dropout. Journal of Biopharmaceutical Statistics16(1), 25–33. 
  • Siddiqui, O., Hung, H. M. J., & O'Neill, R. (2009). MMRM vs LOCF: A comprehensive comparison based on simulation study and 25 NDA datasets. Journal of Biopharmaceutical Statistics19(2), 227–246. 
  • Ting, N., Chen, D., Ho, S., & Capppelleri, J. (2017). Phase II clinical development of new drugs. Springer. 
  • Ting, N., Huang, L., Deng, Q., & Capppelleri, J. (2021). Average response over time as estimand: An alternative implementation of the while on treatment strategy. Statistics in Biosciences13, 479–494. 

To cite this article: Naitee Ting (2023) How to implement the ‘one patient, one vote’ principle under the framework of estimand, Statistical Theory and Related Fields, 7:3, 202-212, DOI: 10.1080/24754269.2023.2164943

To link to this article: