Banff Initiative for Quality Assurance in Transplantation (BIFQUIT): reproducibility of polyomavirus immunohistochemistry in kidney allografts

Abstract

Immunohistochemistry (IHC) is the gold standard for diagnosing (positive vs. negative) polyomavirus BK (BKV) nephropathy and has the potential for disease staging based on staining intensity and quantification of infected cells. This multicenter trial evaluated the reproducibility of BKV IHC among 81 pathologists at 60 institutions. Participants stained tissue microarray slides and scored them for staining intensity and percentage of positive nuclei. Staining protocol details and evaluation scores were collected online. Slides were returned for centralized panel re-evaluation and kappa statistics were calculated. Individual assessment of staining intensity and percentage was more reproducible than combined scoring. Inter-institutional reproducibility was moderate for staining intensity (κ = 0.49) and percentage (κ = 0.42), fair for combined (κ = 0.25) and best for simple positive/negative scoring (κ = 0.78). Inter-observer reproducibility was substantial for intensity (κ = 0.74), percentage (κ = 0.66), positive/negative (κ = 0.78) and moderate for combined scoring (κ = 0.43). Inter-laboratory reproducibility was fair for intensity (κ = 0.37), percentage (κ = 0.40) and combined (κ = 0.24), but substantial for positive/negative scoring (κ = 0.67). BKV RNA copies/cell correlated with staining intensity (r = 0.56) and percentage (r = 0.62). These results indicate that BKV IHC is reproducible between observers but scoring should be simplified to a single-feature schema. Standardization of tissue processing and staining protocols would further improve inter-laboratory reproducibility.

Keywords: Clinical research/practice; infection and infectious agents; kidney (allograft) function/dysfunction; kidney disease: infectious; pathology/histopathology; viral: BK/JC/polyoma.

Figure 1

Strong staining of <10% of tubular nuclei (overall BKV score of C1) in a tissue core on the reference slide.

Figure 2

Bland-Altman plots for intensity of staining (A) and percentage of positive tubular nuclei (B). If a plot shows widely scattered data points, above and below zero, this suggests that there is no consistent bias of one observer (panel) versus the other (participants). These plots show that the panel assigned generally lower scores than the local participants. This is indicated by the fact that most points, each of which represents one of the 25 cores in the TMA, are below the zero line. This discrepancy was relatively greater for staining percentage (bias index = −0.21) than for staining intensity (bias index = −0.04). For staining percentage, the plot also shows that there is distinctly more discrepancy between the panel and participants for intermediate cases compared with negative and strongly positive cases. The frequency bars along the top and right-hand side of the graphs demonstrate the relative distribution of data points along both axes.

Figure 3

Scatter plots showing inter-observer versus inter-laboratory reproducibility for each participant (as represented by weighted kappa values) for BKV IHC staining intensity score (A) and staining percentage score (B). Distribution of the majority of the data points in the top-left quadrants indicates much better inter-observer reproducibility than inter-laboratory reproducibility. Kappa >0.4: at least moderate reproducibility; kappa >0.6: at least substantial reproducibility.