Non-linear regression modelling for medical professionals; making curved paths straightforward

Area of Science:

• Biostatistics
• Medical Research Methodology

Background:

• Regression models are crucial for understanding patient factors, diagnoses, and outcomes in evidence-based medicine.
• Standard regression models often assume linear relationships and group continuous variables, which has limitations.
• Flexible non-linear regression techniques are underutilized in medical research due to complexity and statistical jargon.

Purpose of the Study:

• To introduce non-linear regression, specifically restricted cubic splines (RCS), to medical researchers.
• To demonstrate how RCS can capture non-linear relationships between predictors and outcomes within familiar regression frameworks.
• To provide practical tools and understanding for applying non-linear modeling in medical research.

Main Methods:

• Discussion of pitfalls associated with traditional linear regression and variable grouping.
• Introduction to restricted cubic splines (RCS) for modeling non-linear relationships.
• Implementation of RCS in a case study with accompanying R scripts and a new R package ('rmsMD').

Main Results:

• RCS allows for flexible fitting of curved relationships within standard regression models.
• The 'rmsMD' R package simplifies the application of RCS for medical researchers.
• The study provides a case study and example scripts to facilitate adoption.

Conclusions:

• Non-linear modeling using RCS can be intuitively understood and applied by medical researchers.
• Accessible tools like 'rmsMD' can overcome barriers to using advanced regression techniques.
• This approach enhances the ability to model complex relationships, advancing evidence-based medicine.