# Comprehensive Agreement Analysis of Colorimetric and Turbidimetric Total Protein Assays in Cerebrospinal Fluid

## TL;DR

This study compares two methods for measuring total protein in cerebrospinal fluid and finds they agree well for most cases, though some differences exist.

## Contribution

The study provides a detailed agreement analysis between colorimetric and turbidimetric methods for CSF total protein measurement.

## Key findings

- The two methods showed substantial agreement overall with a concordance correlation coefficient of 0.9881.
- The pathological subgroup had better agreement than the normal subgroup.
- Cohen’s Kappa indicated substantial clinical agreement between the methods.

## Abstract

Background/Objectives: Accurate measurement of total protein (TP) in cerebrospinal fluid (CSF) is crucial for diagnosing various neurological conditions. This study aims to evaluate the concordance between a routine colorimetric method and a recently introduced turbidimetric method for measuring CSF TP. Methods: We measured 161 CSF samples using both methods, analyzing the whole population and two subgroups: normal (≤500 mg/L) and pathological (>500 mg/L). Agreement was assessed using Lin’s Concordance Correlation Coefficient (CCC), Bland–Altman, and Deming regression, while clinical concordance was determined with Cohen’s Kappa. Results: The concentrations obtained from the two methods did not differ significantly and were well-correlated across the population and subgroups. The CCC for the entire dataset was 0.9881 (substantial agreement), while the Bland–Altman analysis showed a mean bias of 4.467 mg/L. For the “normal” subgroup (n = 97), the CCC was 0.8722 (poor agreement), with a mean bias of 7.668 mg/L. In the “pathological” subgroup (n = 64), the CCC was 0.9858 (substantial agreement) with a mean bias of −3.838 mg/L. Demin regression did not show statistically significant proportional or constant bias in the whole population. However, a stratified analysis revealed a significant negative constant bias in the “normal” subgroup in absence of significant bias in the “pathological” subgroup. Cohen’s kappa was 0.804, indicating substantial agreement. Conclusions: Both methods showed substantial agreement for quantifying CSF TP and clinical classification, supporting their potential interchangeability for diagnostic purposes. Nonetheless, laboratories should note the presence of bias, particularly for samples near the clinical cut-off value.