Linear regression estimation of minimal detectable concentration. Thyrotropin as an example

Area of Science:

• Clinical chemistry
• Analytical chemistry
• Immunoassay development

Background:

• Minimal detectable concentration (MDC) is a critical performance metric for clinical immunoassays.
• Accuracy, defined as a consistent linear relationship between observed and actual analyte concentration, is vital for meaningful MDC values.

Purpose of the Study:

• To develop and validate a novel linearity regression protocol for accurately determining the minimal detectable concentration (MDC) in immunoassays.
• To assess the performance of the new protocol by evaluating its accuracy and accounting for between-run variability.

Main Methods:

• Developed a linearity regression protocol using serial twofold dilutions and polynomial regression (linear, second-, and third-order).
• Employed t-tests and F-tests to identify the linear dynamic range and statistically significant beta coefficients.
• Iteratively refined the linear model by removing and reintroducing data points based on analytical significance and predicted vs. observed values.

Main Results:

• Applied the protocol to two automated thyrotropin immunoassay systems.
• One system demonstrated linearity and accuracy down to 0.02 mU/L (77% of the time) and 0.01 mU/L (68% of the time).
• The second system showed infrequent linearity, with usefulness down to 0.02 mU/L only about 20% of the time.

Conclusions:

• The proposed accuracy-based linearity regression protocol provides a robust method for determining MDC.
• This approach offers a significant advantage over traditional empiric methods relying solely on interassay variability.