# Estimating Tumor Proportion in Smear Slides for Reliable Molecular Analysis

## TL;DR

This study assesses how consistently pathologists estimate tumor cell percentages in smear slides, finding variability that highlights the need for better standards.

## Contribution

The study evaluates inter-observer reliability in tumor cell estimation using digital images and predefined categories.

## Key findings

- The molecular cytopathologist showed the highest consistency with the gold standard (Kappa = 0.69).
- The lowest agreement occurred in the 11–20% tumor proportion category, with frequent overestimation.
- Substantial agreement was observed in the 71–100% category, aligning in over 95% of cases.

## Abstract

Objective:
The use of molecular pathology is critical in diagnostics and theranostics. Today, cytological smears are utilized for molecular testing more often than ever. Accurate tumor cell percentage estimation is essential for reliable molecular testing, but its consistency remains uncertain. This study evaluates the reliability of tumor cell percentage estimations among an expert cytopathologist, a molecular cytopathologist, and a molecular pathologist.

Material and Methods:
Digital images from May-Grünwald-Giemsa (MGG)-stained smear slides of ten EBUS-guided mediastinal lymph node samples were selected. Five regions per slide were evaluated (50 areas from 10 patients). Three pathologists independently estimated tumor cell percentages using predefined categories (0–10%, 11–20%, 21–50%, etc.). Cells were also counted manually as the gold standard.

Results:
The molecular cytopathologist (Observer 1 -) showed the highest consistency (Kappa = 0.69), followed by the expert cytopathologist (Observer 3 -, Kappa = 0.64), both demonstrating substantial agreement with the gold standard. The molecular pathologist (Observer 2 -) displayed moderate consistency (Kappa = 0.52). Agreement was most significant in the 71–100% category, aligning in over 95% of cases. The lowest value occurred in the 11–20% category. In this category, tumor proportions were frequently overestimated compared to the gold standard.

Conclusion:
Variability in tumor percentage estimations shows the need for standardized protocols and training. Substantial agreement was reached in specific categories. However, discrepancies in borderline cases highlight the importance of accurate assessments. More research is needed to improve estimation methods.