Advanced Methods for Controlling for Selection Bias

Propensity Score Analysis

Propensity score analysis is appropriate when there are a number of variables that may influence the choice of a par- ticular treatment being studied. In performing a propensity score analysis, you must first estimate the probability of receiving a particular treatment in a logistic regression model. This is the propensity score. Then patients who received treatment are matched to those who did not based on pro- pensity scores and then compared. This methodology can only reduce inherent bias in observable characteristics, as they need to be entered into the propensity score model.

Instrumental Variable (IV) Analysis

Instrumental variable (IV) analysis is a sophisticated method to help control for selection bias in observational studies. It is appropriate when potential confounding variables are either unknown or difficult to measure..The most critical element of an IV analysis is the instrumental variable itself. An IV is a measurable event or characteristic that gets a patient into a treatment group, but is not associated with the outcome. Theoretically, the IV is comparable to randomization in assigning individuals to control and treatment groups. The difficulty in using IV analysis is in identifying an instrumental variable. Many candidate variables are associated not only with the choice of treatment but with the outcome as well.

Summary

Understanding basic biostatistical methods is essential for both the research and clinical practice of a surgeon. Basic understanding of the methods discussed in this chapter will

Chapter 6

Animal Models for Surgical

Research

Andrea A. Hayes-Jordan

Keywords  Transgenic • Animal model • Xenograft models

• Hyperplastic islets • Angiogenic islets

As a surgeon, you will be the best equipped to conceptualize and construct a unique animal model that closely mimics the human clinical presentation. When beginning your basic sci- ence research career, establishing your own novel animal model is one of the very first things you must do. After you decide what field of interest and what specific clinical prob- lem or disease you wish to study,you must validate and estab- lish an animal model. This model will define your research and associate it with your body of work such that even when a surgery technician or resident in your lab or anyone is pre- senting data from your lab, because your animal model is recognized as your “signature,” the audience will know you are the principal investigator. The animal model you choose should afford ready visualization of the effects, be easily reproducible, and chosen specifically to highlight your dis- ease of interest. In this chapter, we will explore several

A.A. Hayes-Jordan

Department of Surgical Oncology, UT MD Anderson Cancer Center,

Houston,TX, USA

DOI 10.1007/978-0-85729-313-8_6,

© Springer-Verlag London Limited 2012