gms | German Medical Science

MAINZ//2011: 56. GMDS-Jahrestagung und 6. DGEpi-Jahrestagung

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V.
Deutsche Gesellschaft für Epidemiologie e. V.

26. - 29.09.2011 in Mainz

Simulating outcomes of untreated patients – can we reduce the placebo group?

Meeting Abstract

  • Manfred Berres - RheinAhrCampus Remagen, FH Koblenz, Remagen
  • René Spiegel - Universitätsspital, Basel
  • Andreas U. Monsch - Universitätsspital, Basel
  • André R. Miserez - Universitätsspital, Basel

Mainz//2011. 56. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 6. Jahrestagung der Deutschen Gesellschaft für Epidemiologie (DGEpi). Mainz, 26.-29.09.2011. Düsseldorf: German Medical Science GMS Publishing House; 2011. Doc11gmds076

DOI: 10.3205/11gmds076, URN: urn:nbn:de:0183-11gmds0767

Published: September 20, 2011

© 2011 Berres et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.



Introduction: We consider the scenario of a new drug with a potential to stop the progression of a disease with no effective competitor being available. The routinely used design for phase III clinical studies is a randomized placebo-controlled trial. This means that many patients receive placebo for the duration of the study, often 18 months or longer. As an example, we consider patients with minimal cognitive impairement (MCI). Ethical problems arise, when such patients are given placebo while they progress to Alzheimer’s disease. Instead of referring to historical controls, we propose a novel study design, termed Placebo Group Simulation Approach (PGSA), using statistical models to forecast outcomes of pre-symptomatic AD patients from their own baseline data. Forecasted outcomes are compared with outcomes observed on candidate drugs, thus replacing a concomitant placebo group.

Methods: Using the MCI sample of the ADNI database, we developed regression models and mixed effects models to predict the ADAS-cog score and a neuropsychological test score at 24 months and the trajectory up to 36 months after recruitment. Predictors of the model are demographic, clinical and neuropsychological baseline variables.

Results: Although these models are not able to predict single outcomes, they appear to be sufficiently precise to estimate sample means for sample sizes commonly used in clinical trials. New outcomes have been simulated from the predictors. Their distribution is very similar to the observed outcomes. Hence, these models are able to simulate outcomes under placebo conditions.

Conclusion: Given a sample of patients who are actually treated in a trial, the observed outcomes under treatment could be compared to the simulated outcomes under placebo. Analytical calculations and simulations show that such trials would be very powerful. Models derived from the ADNI data are currently applied to other data sets.


Spiegel R, Berres M, Miserez AR, Monsch AU. For debate: substituting placebo controls in long-term Alzheimer’s prevention trials. Alzheimer’s Research & Therapy. 2011;3:9.