Project

Back to overview

Model-Based Recursive Partitioning for Stratified Medicine

English title Model-Based Recursive Partitioning for Stratified Medicine
Applicant Hothorn Torsten
Number 163456
Funding scheme Project funding (Div. I-III)
Research institution Institut für Epidemiologie, Biostatistik und Prävention Universität Zürich
Institution of higher education University of Zurich - ZH
Main discipline Mathematics
Start/End 01.01.2016 - 31.12.2018
Approved amount 185'453.00
Show all

Keywords (6)

Clinical trials; Random forests; Semi-parametric regression; Personalised medicine; Subgroup analyses; Regression trees

Lay Summary (German)

Lead
Im Rahmen klinischer Studien wird der Behandlungserfolg einer bestimmtenTherapie üblicherweise für relativ große Gruppen von Patienten in uniformer Art und Weise evaluiert. Innerhalb dieser Gruppen kann es jedoch sehr wohl Patienten geben, welche sehr gut oder sehr schlecht auf diese Therapieansprechen. Ziel des Projektes ist die Entwicklung von datengetriebenenVerfahren zur Identifikation solcher Patientensubgruppen.
Lay summary
(2) Inhalt und Ziel des Forschungsprojektes
Im Rahmen des Projektes ``Modellbasiertes Rekursives Partitionieren in der  
Stratifizierten Medizin'' werden statistische Verfahren zur Vorhersage eines
individuellen Behandlungseffektes zum Zeitpunkt des Therapiebeginns
entwickelt, implementiert und evaluiert.  Basierend auf zum Zeitpunkt des
Beginns der Therapie vorliegenden klinischen Parametern wird versucht, auf
den zu erwartenden Behandlungseffekt einer oder mehrerer potentiell zur
Verfügung stehender Therapien zu schliessen.  Das Hauptaugenmerk liegt dabei
auf Verfahren des rekursiven Partitionierens, also der datengesteuerten
Segmentierung von Modellen, welche einen Behandlungseffekt beschreiben.
Solche Verfahren erlauben damit eine datengetriebene Einteilung von   
Patienten in Subgruppen mit einem hohen bzw.  niedrigen zu erwartenden
Behandlungseffekt.  In einem zweiten Schritt soll dieses Verfahren erweitert
werden, um durch eine Aggregierung mehrerer solcher Stratifikationsmodelle
eine glatte Schätzung des individuellen zu erwartenden Behandlungseffektes
zu ermöglichen.

(3) Wissenschaftlicher und gesellschaftlicher Kontext des
Forschungsprojektes
Verfahren zur Subgruppenidentifikation ermöglichen die Planung von Studien  
mit dem Ziel, Therapien besser auf die einzelnen Patienten anzupassen und   
damit die Behandlung von Patienten mit voraussichtlich wenig wirksamen
Verfahren zu vermeiden.

Direct link to Lay Summary Last update: 28.09.2015

Responsible applicant and co-applicants

Employees

Publications

Publication
Estimating Patient-specific Treatment Advantages in the ``Treatment for Adolescents with Depression Study''
Foster Simon, Mohler-Kuo Meichun, Tay Lynette, Hothorn Torsten, Seibold Heidi (2019), Estimating Patient-specific Treatment Advantages in the ``Treatment for Adolescents with Depression Study'', in Journal of Psychiatric Research, 112, 61-70.
model4you: An R Package for Personalised Treatment Effect Estimation
Heidi Seibold, Achim Zeileis, Torsten Hothorn (2019), model4you: An R Package for Personalised Treatment Effect Estimation, in Journal of Open Research Software, 17.
Detecting treatment-subgroup interactions in clustered data with generalized linear mixed-effects model trees.
Fokkema M, Smits N, Zeileis A, Hothorn T, Kelderman H (2018), Detecting treatment-subgroup interactions in clustered data with generalized linear mixed-effects model trees., in Behavior research methods, 50(6), 2016-2034.
Subgroup identification in dose-finding trials via model-based recursive partitioningSubgroup identification in dose-finding trials
Thomas Marius, Bornkamp Björn, Seibold Heidi (2018), Subgroup identification in dose-finding trials via model-based recursive partitioningSubgroup identification in dose-finding trials, in Statistics in Medicine, 37(10), 1608-1624.
Individual treatment effect prediction for amyotrophic lateral sclerosis patients
Seibold Heidi, Zeileis Achim, Hothorn Torsten (2018), Individual treatment effect prediction for amyotrophic lateral sclerosis patients, in Statistical Methods in Medical Research, 27(10), 3104-3104.
Model-Based Recursive Partitioning for Subgroup Analyses.
Seibold Heidi, Zeileis Achim, Hothorn Torsten (2016), Model-Based Recursive Partitioning for Subgroup Analyses., in The international journal of biostatistics, (1), 45-63.
Generalised Linear Model Trees with Global Additive Effects
Seibold Heidi, Hothorn Torsten, Zeileis Achim (2016), Generalised Linear Model Trees with Global Additive Effects, in ArXiv e-prints.

Collaboration

Group / person Country
Types of collaboration
Axel Benner, Division of Biostatistics, DKFZ Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Achim Zeileis, Universität Innsbruck Austria (Europe)
- Publication
- Exchange of personnel
Björn Bornkamp, Novartis Statistical Methodology Group Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
CEN ISBS Vienna 2017 Talk given at a conference Individual treatment effect prediction for ALS patients 28.08.2017 Vienna, Austria Seibold Heidi;
Statistical Computing 2017 Talk given at a conference Model-Based Recursive Partitioning for Stratified and Personalised Treatment Effect Estimation 23.07.2017 Guenzburg, Germany Seibold Heidi;
Heidelberger Kolloquium Medizinische Biometrie, Informatik und Epidemiologie Individual talk Stratified and personalised medicine using model-based recursive partitioning 07.11.2016 Heidelberg, Germany Seibold Heidi;
useR! 2016 Talk given at a conference Predicting individual treatment effects with the partykit package 27.06.2016 Stanford, United States of America Seibold Heidi;


Knowledge transfer events

Active participation

Title Type of contribution Date Place Persons involved
Biogen Symposium on Statistical Methods in MS Talk 08.11.2017 Boston, United States of America Seibold Heidi;
Novartis Science VC Talk 20.09.2017 Basel, Switzerland Seibold Heidi;


Awards

Title Year
Arthur-Linder Prize (https://www.ibs-roes.org/home-en/arthur-linder-prize/) 2017

Use-inspired outputs

Associated projects

Number Title Start Funding scheme
184603 A Lego System for Transformation Inference 01.01.2020 Project funding (Div. I-III)

Abstract

This research proposal aims at developing methods for inference onpatient-specific treatment effects in the context of stratified medicine.The term ``stratified medicine'' refers to methods ``using a biomarker tomatch a patient to a cohort that has exhibited a differential response to atreatment'' according to the US Food and Drug Administration report ``Pavingthe Way for Personalized Medicine'' published in 2013. Establishedstatistical procedures based on stratified or interaction models areroutinely used when both biomarkers and cohorts have already been identifiedby previous research. This proposal targets statistical procedures tailoredfor the identification of such cohorts from a larger number of potentiallyuseful biomarkers. From a conceptual point of view, we distinguish betweenthree separate steps of such an analysis: (1) the selection of one (or afew) biomarker(s) which is predictive for a differential treatment effect, (2) the identification of patient subgroups (or cohorts) from these selectedbiomarkers, and (3) the estimation of a subgroup-specific treatment effect.Established stratified or interaction models for (3) are commonly refered toas ``primary subgroup analyses'' in stratified, individualised, orpersonalised medicine. With this research plan, we propose statisticalmethods for biomarker selection and subgroup identification and improvedestimation procedures for differential treatment effects in steps (1-3).
-