Reliability of stromal markers multiplex immunofluorescent staining: pathologist assessment compared to quantitative image analysis.

Stroma may play an important role in breast carcinogenesis. There is no data on the expression of stromal markers αSMA, FAP, MMP14, TNC, and s100a6 in the breast tissue of cancer-free women. We compared the multiplex immunofluorescence (IF) expression assessment for these markers in normal terminal duct-lobular units (TDLUs) by an expert pathologist with the automated image analysis results and assessed the homogeneity of the markers across multiple cores pertaining to each woman. We included 73 cancer-free women with biopsy-confirmed benign breast disease in the Nurses' Health Study (NHS) and NHSII cohorts. IF was conducted with commercial antibodies (αSMA: 1:400 dilution; FAP: 1:50; MMP14: 1:150; TNC: 1:200; s100a6: 1:300). For each tissue microarray core, the percent positivity was assessed by the pathologist and inForm v2.6.0. Using the pathologist scores as the gold standard, correlations between pathologist and inForm scores were evaluated with Spearman correlation (for categorical positivity: 0, >0 - <1, 1 - 10, >10 - 50, and >50%) and sensitivity/specificity (for binary positivity defined with 1%, 10% and 25% cut-offs). Pathologist and inForm readings were available for 149 and 134 cores, respectively; 105 cores had both. The correlation in the expression across available cores for a woman (median =3, range 1-6) was strong for FAP, MMP14, and s100a6 (Intra-class correlation [ICC]=0.69, 0.72, 0.63, respectively), moderate for αSMA (ICC=0.35), and poor for TNC (ICC=0.21). Correlation between pathologist and inForm was strong for s100a6, FAP, and MMP14 (correlation coefficient =0.77, 0.70, and 0.78, respectively) and moderate for αSMA (0.37) and TNC (0.42). With 1% positivity cut-off, sensitivity was the lowest for TNC (0.30) and ranged between 0.84-0.93 for other markers. Specificity ranged between 0.43-0.98 across all markers with the lowest estimates for αSMA. Sensitivity declined for all markers while using 10% and 25% cut-offs, while specificity increased. Our findings show that computational assessments for αSMA, FAP, MMP14, TNC, and s100a6 exhibit variable correlations with manual assessment. These findings support the use of computational platforms for IF evaluation of most, but not all, stromal markers in large-scale epidemiologic studies and the importance of pilot studies for identification of appropriate cut-offs for defining staining positivity.