Reference curves based on non-parametric quantile regression.

Area of Science:

• Biostatistics
• Medical Statistics
• Quantitative Biology

Background:

• Establishing age-dependent reference curves is crucial in medicine but current statistical methods are often complex or inefficient.
• Classical and semi-parametric methods exist but have limitations in applicability and systematic construction.

Purpose of the Study:

• To propose a new, systematic, and efficient non-parametric methodology for estimating reference intervals.
• To develop methods applicable to continuous quantitative biological and clinical variables.
• To demonstrate the methodology using skin biophysical properties in healthy women.

Main Methods:

• Non-parametric estimation of conditional quantiles for constructing reference intervals.
• Application to continuous quantitative data, exemplified by skin biophysical properties.
• Comparison with established parametric (Royston's polynomial) and semi-parametric (LMS) methods.

Main Results:

• The proposed non-parametric method provides a robust approach for estimating reference intervals.
• The methodology is demonstrated to be applicable to biological variables, such as skin properties.
• Results are comparable to existing methods, offering a potentially more versatile alternative.

Conclusions:

• A novel non-parametric methodology for constructing reference intervals has been developed.
• This approach offers an efficient and systematic alternative for creating age-dependent reference curves in medicine.
• The method is broadly applicable to continuous biological measurements.