Défis statistiques en recherche en santé respiratoire


Andrea Benedetti

CUSM - Institut thoracique de Montréal


Domaine : santé des populations

Programme chercheurs-boursiers - Junior 1

Concours 2014-2015

My research over the next four years will focus on statistical challenges with applications in respiratory health research.  The Collaborative Group for Meta-Analysis of Individual Patient Data in Multi-Drug Resistant Tuberculosis (IPD-MDRTB) has assembled individual patient data on treatment outcomes in multiple drug resistant (MDR) tuberculosis (TB) patients from 32 observational studies comprising over 9000 individual patients.  This database offers a unique opportunity to address methodologic issues in IPD-MA and to investigate additional questions related to MDR-TB.
The overall goals of the primary project described in this proposal are to (1) advance statistical methodology for the analysis of individual patient data meta-analysis and using the insights developed, (2) investigate the effectiveness of surgery as treatment for MDR-TB.
IPD-MA are the gold standard of meta-analytic methods, and increasingly performed.  However, several questions regarding the analytic approach remain unanswered which I intend to address. The guidelines that I develop will allow data analysts to make the best decisions when analyzing this type of data, and enhance the robustness of their conclusions.  The methods developed here will have a wide applicability in many different domains.

These methodologic developments will also let us make the best use of the IPD-MDRTB.  The use of surgery as adjunctive treatment remains controversial, given the substantial risks of the procedures, varied results, and likely confounding by indication.
Over the next 4 years, my research plan will also include:
1. Statistical challenges in molecular epidemiology with applications to TB
2. Smoothing in the context of mixed models with applications to respiratory health

All three projects will result in guidelines that will enable researchers to use novel and sophisticated statistical methods to their greatest advantage, while maintaining statistical integrity and validity of the analyses.