Censored data considerations and analytical approaches for salivary bioscience data.

Area of Science:

• Salivary bioscience
• Biostatistics
• Biobehavioral research

Background:

• Left censoring occurs when salivary analyte levels are below assay detection limits.
• Conventional statistical methods for censored data can introduce systematic bias.
• Specialized censored data methods are underutilized in salivary research.

Purpose of the Study:

• To compare conventional and specialized statistical approaches for handling left-censored salivary data.
• To assess the impact of censoring levels (9% and 15%) on statistical estimates.
• To evaluate the efficiency and appropriateness of different methods for salivary C-reactive protein data.

Main Methods:

• Comparison of descriptive statistics, correlation coefficients, and regression estimates.
• Utilized salivary C-reactive protein data with two levels of left censoring.
• Assessed method sensitivity to sample size variations.

Main Results:

• Statistical estimates were broadly similar between conventional and specialized approaches.
• Specialized censored data methods demonstrated greater efficiency and appropriateness.
• Implementation of specialized methods required fewer data manipulations.

Conclusions:

• Specialized censored data statistical approaches are recommended for salivary bioscience.
• These methods offer a more efficient and appropriate way to reduce bias in salivary data analysis.
• Preliminary recommendations are provided for investigators using left-censored salivary biomeasure data.