Cubic splines to model relationships between continuous variables and outcomes: a guide for clinicians

Area of Science:

• Biostatistics
• Hematology
• Transplant Medicine

Background:

• Bone marrow transplantation research often analyzes relationships between variables like measurable residual disease (MRD) and cumulative incidence of relapse (CIR).
• Traditional analysis methods dichotomize continuous biological variables (e.g., MRD status) using arbitrary cut-points, potentially losing valuable data.
• This approach can lead to missing important associations, analogous to overlooking details in a complex dataset.

Discussion:

• Cubic splines provide a method to create smooth curves from irregular data points, offering a more nuanced representation than simple linear fits.
• This technique is conceptually similar to drawing smooth shapes by connecting data points, as demonstrated by software tools.
• The paper introduces Gauthier and co-workers' application of cubic splines for extracting maximum information from data that resists simple categorization.

Key Insights:

• Cubic splines enable the analysis of continuous biological variables, such as MRD, without arbitrary data reduction.
• This method enhances the correlation analysis between pre-transplant MRD and post-transplant outcomes like CIR.
• Researchers can gain deeper insights into transplant data by avoiding the loss of information inherent in dichotomization.

Outlook:

• The application of cubic splines presents a new paradigm for data analysis in bone marrow transplantation and other fields.
• This technique encourages a re-evaluation of how continuous data is handled in statistical modeling for clinical research.
• Further exploration of cubic splines may lead to more accurate predictions and improved understanding of treatment efficacy.