|
Activity Number:
|
177
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2003 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #300579 |
|
Title:
|
Software for Variable Optimal Matching in Case-Control Studies
|
|
Author(s):
|
Erik J. Bergstralh*+ and Steven J. Jacobsen and Jon Kosanke
|
|
Companies:
|
Mayo Clinic and Mayo Clinic and Mayo Clinic
|
|
Address:
|
Biostatistics/Harwick 7, Rochester, MN, 55905-0002,
|
|
Keywords:
|
software ; matching ; optimal ; case ; control
|
|
Abstract:
|
In observational studies one often wishes to compare a group of index subjects (cases) to a group of controls regarding some endpoint Y, while controlling for potential confounders (X's). In classic case-control studies, where index subjects have the disease or event and controls do not, the "endpoint" might be previous exposure to some potential etiologic agent or biomarker. Control of confounding is often done using multivariate modeling or by matching each of the n cases to one or more of M potential controls on the basis of case-control distance (typically defined as a function of the difference in the X's). Recently it has been shown that substantial gains in bias reduction can be attained using optimal matching and allowing the number of controls per case to vary, as opposed to being fixed (Ming/Rosenbaum 2000). We previously developed an SAS macro (%match) to implement fixed optimal matching (Bergstralh/Kosanke/Jacobsen 1996). We now present new software to implement variable optimal matching (%vmatch) and calculate the n x M case-control distance matrix (%dist). An example is presented for a nested case-control study within a prostate cancer cohort.
|
