The Effect of Ignoring Statistical Interactions in Regression Analyses Conducted in Epidemiologic Studies: An Example with Survival Analysis Using Cox Proportional Hazards Regression Model
- Corresponding Author:
- Mohammad H. Rahbar PhD
University of Texas Health Science Center at Houston
Component of Center for Clinical and Translational Sciences
6410 Fannin Street, UT Professional Building Suite 1100.05
Houston, TX 77030, USA
E-mail: [email protected]
Received Date: November 01, 2014 Accepted Date: January 12, 2015 Published Date: January 15, 2015
Citation:Vatcheva KP, Lee M, McCormick JB, Rahbar MH (2016) The Effect of Ignoring Statistical Interactions in Regression Analyses Conducted in Epidemiologic Studies: An Example with Survival Analysis Using Cox Proportional Hazards Regression Model. Epidemiol 6: 216. doi:10.4172/2161-1165.1000216
Copyright: © 2016 Vatcheva KP, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Objective: To demonstrate the adverse impact of ignoring statistical interactions in regression models used in epidemiologic studies.
Study design and setting: Based on different scenarios that involved known values for coefficient of the interaction term in Cox regression models we generated 1000 samples of size 600 each. The simulated samples and a real life data set from the Cameron County Hispanic Cohort were used to evaluate the effect of ignoring statistical interactions in these models.
Results: Compared to correctly specified Cox regression models with interaction terms, misspecified models without interaction terms resulted in up to 8.95 fold bias in estimated regression coefficients. Whereas when data were generated from a perfect additive Cox proportional hazards regression model the inclusion of the interaction between the two covariates resulted in only 2% estimated bias in main effect regression coefficients estimates, but did not alter the main findings of no significant interactions.
Conclusions: When the effects are synergic, the failure to account for an interaction effect could lead to bias and misinterpretation of the results, and in some instances to incorrect policy decisions. Best practices in regression analysis must include identification of interactions, including for analysis of data from epidemiologic studies.