|
Activity Number:
|
439
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 12, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #300833 |
|
Title:
|
The Impact of Twin Births on Inference in Randomized Clinical Trials of Pre-term Infants
|
|
Author(s):
|
Michele L. Shaffer*+ and Kristi L. Watterberg
|
|
Companies:
|
Pennsylvania State University and University of New Mexico
|
|
Address:
|
College of Medicine, 600 Centerview Drive, Hershey, PA, 17033,
|
|
Keywords:
|
correlated data ; mixed-effects model ; generalized estimating equations
|
|
Abstract:
|
The multiple-gestation rate in pre-term births is considerably higher than in full-term births so that twin births can account for an appreciable percentage (10-20%) of a study sample for a trial targeted at pre-term infants. Failure to account for the correlation within complete twin pairs may impact the precision of estimated treatment effects or lead to incorrect decisions regarding the significance of treatment. The current clinical literature rarely mentions if twins were taken into consideration for trial design or analysis. Well-known methods exist for analyzing correlated data, but the operating characteristics of these methods are not fully understood for such a mix of correlated and independent data. A simulation study is conducted to compare mixed-effects models and generalized estimating equations to standard methods, which assume all subjects are independent, for continuous outcomes in two-armed clinical trials. Data from an NICHD Neonatal Research Network trial are used as motivation.
|
