Dichotomizing continuous predictors in multiple regression: a bad idea

2005/10/11 by Patrick Royston, Douglas G. Altman, Willi Sauerbrei · 15 citations
Mathematics · Medicine · #Liver Disease Diagnosis and Treatment #Liver Diseases and Immunity #Statistical Methods and Inference

paper · doi:10.1002/sim.2331

openalex publication_date 2005/10/11 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/31

Abstract

In medical research, continuous variables are often converted into categorical variables by grouping values into two or more categories. We consider in detail issues pertaining to creating just two groups, a common approach in clinical research. We argue that the simplicity achieved is gained at a cost; dichotomization may create rather than avoid problems, notably a considerable loss of power and residual confounding. In addition, the use of a data-derived 'optimal' cutpoint leads to serious bias. We illustrate the impact of dichotomization of continuous predictor variables using as a detailed case study a randomized trial in primary biliary cirrhosis. Dichotomization of continuous data is unnecessary for statistical analysis and in particular should not be applied to explanatory variables in regression models.

Cited by

• A flexible class of latent variable models for the analysis of antibody response data
• Trait coevolution and causal inference using generalized dynamic phylogenetic models
• A tale of two scripts: Applying the principle of least complexity to simplified and traditional Chinese
• A Parsimonious Explanation of the Resilient, Undercontrolled, and Overcontrolled Personality Types
• The Pick-the-Winner-Picker Heuristic: Preference for Categorically Correct Forecasts
• A Two-Step Test to Identify Zero-Inflated Biomarkers in Early-Phase Clinical Trials
• Significance of Th1 and Th2 Cell Densities and Th1/Th2 Cytokine Profiles in Colorectal Cancer
• Cubic splines to model relationships between continuous variables and outcomes: a guide for clinicians
• Learning About the Binding Nature of the Law: Police Violence, Criminal Offending and Adolescent Legal Socialization
• Nonparametric Scanning For Nonrandom Missing Data With Continuous Instrumental Variables
• External validation of the S-GRAS score for predicting recurrence in patients with adrenocortical carcinoma: implications for adjuvant mitotane therapy
• Is (poly-) substance use associated with impaired inhibitory control? A mega-analysis controlling for confounders
• Evaluating the Compulsive Buying Scale in a Canadian non-clinical sample: A psychometric analysis using differential item functioning
• Comparison between continuous and discrete doses using Escalation With Overdose Control
• Measuring frailty in the elderly: an indicator based on a super-classifier

Related

• Proposal of a general framework to categorize continuous predictor variables
• Categorisation of continuous covariates for stratified randomisation: How should we adjust?
• Explanation of multicollinearity using the decomposition theorem of ordinary linear regression models
• Distinguishing Between Latent Classes and Continuous Factors: Resolution by Maximum Likelihood?
• Controlling for multiple covariates