Quetelet Seminar by dr. Saskia Le Cessie

For whom
Students , Alumni , Employees
When
07-10-2021 from 16:00 to 17:00
Where
campus Sterre - S9 - classroom 3.1 or zoom
Language
Dutch
Organizer
Jan De Neve
Contact
Jan.DeNeve@UGent.be
Add to my calendar

Meta-analysis of aggregate continuous outcomes with baseline and follow up measurements

Meta-analysis of aggregate continuous outcomes with baseline and follow up measurements

by dr. Saskia le Cessie (Dept of Clinical Epidemiology, Leiden University Medical Center and Dept of Biomedical Data Sciences, Leiden University Medical Center).

Abstract

In many clinical trials with a continuous outcome, measurements are performed at two time points: at baseline, before treatment assignment, and at the end of follow up. The recommended approach to analyze this type of data is to use analysis of covariance (ANCOVA): a regression analysis with the follow up measurement as dependent variable and baseline measurement and treatment as independent variables. This approach is in general more efficient than using final measurements or change scores, and it may adjust for baseline differences between the treatment groups.  However in practice, many studies do not follow this approach and ANCOVA estimates are rarely reported.

This yields problems when performing a meta-analysis using this types of studies. Often only aggregated data (AD) are available, such as means and standard deviations of baseline and follow up measurements, and mean change with standard error.  We will discuss different ways of performing a meta-analysis using these data: using follow-up measurements, using change scores, reconstructing the ANCOVA estimates from the summary data, and the Trowman meta regression approach(1). We will compare these methods to a new approach, where pseudo individual participant data are being generated based on the AD(2). The advantage of the latter approach is that the data can be analysed as if they are real individual participant data, using the framework of linear mixed models. This allows for many different modeling options of increased complexity.

References:

(1)   Trowman et al., Journal of Clinical Epidemiology 2007; 60(12): 1229–1233.

(2)   Papadimitropoulou et al. Research Synthesis Methods 2020; 11(6), 780-794.

Venue

  • on-campus:

campus Sterre - S9 - classroom 3.1 or zoom 

  • zoom link:

https://ugent-be.zoom.us/j/99158326522?pwd=bFRTMHBTdTdzNDdLelhpV09Geng5QT09

Meeting ID: 991 5832 6522
Passcode: Ltdty35a