Physiological within subject variability and test-retest reliability of deconvolution analysis of luteinizing hormone release

Deconvolution analysis is a powerful analytical tool, which can provide valuable information regarding hormone secretion. However, as with all analytical tools, it must be properly applied to maximize reliability. The investigator should use a standardized approach including the assessment of statistical confidence limits. Optimal experimental conditions must be utilized in collecting the data in order not only to maximize sensitivity and specificity of the deconvolution method, but also to ensure the reliability and validity of individual parameter estimates. Finally, biological variability of a system output (e.g., pulsatile LH secretion) over time must be recognized. Indeed, estimates of the intrinsic physiological reliability of a system will vary depending on the features being observed; for example, mean (24-hr) serum LH concentrations are quite stable across observation sessions, whereas apparent LH secretory pulse frequency and amplitude can vary physiologically over time in the same individual. The extent of such expected serial nonuniformity and the nature of the specific experimental questions posed will jointly control the overall statistical power of any particular study designed to prove or disprove significant intervention or treatment effects on the neuroendocrine axis of interest.