Reproducibility of the NEPTUNE descriptor-based scoring system on whole-slide images and histologic and ultrastructural digital images.

The multicenter Nephrotic Syndrome Study Network (NEPTUNE) digital pathology scoring system employs a novel and comprehensive methodology to document pathologic features from whole-slide images, immunofluorescence and ultrastructural digital images. To estimate inter- and intra-reader concordance of this descriptor-based approach, data from 12 pathologists (eight NEPTUNE and four non-NEPTUNE) with experience from training to 30 years were collected. A descriptor reference manual was generated and a webinar-based protocol for consensus/cross-training implemented. Intra-reader concordance for 51 glomerular descriptors was evaluated on jpeg images by seven NEPTUNE pathologists scoring 131 glomeruli three times (Tests I, II, and III), each test following a consensus webinar review. Inter-reader concordance of glomerular descriptors was evaluated in 315 glomeruli by all pathologists; interstitial fibrosis and tubular atrophy (244 cases, whole-slide images) and four ultrastructural podocyte descriptors (178 cases, jpeg images) were evaluated once by six and five pathologists, respectively. Cohen's kappa for inter-reader concordance for 48/51 glomerular descriptors with sufficient observations was moderate (0.40<kappa≤0.60) for 17 and good (0.60<kappa≤0.80) for 8, for 52% with moderate or better kappas. Clustering of glomerular descriptors based on similar pathologic features improved concordance. Concordance was independent of years of experience, and increased with webinar cross-training. Excellent concordance was achieved for interstitial fibrosis and tubular atrophy. Moderate-to-excellent concordance was achieved for all ultrastructural podocyte descriptors, with good-to-excellent concordance for descriptors commonly used in clinical practice, foot process effacement, and microvillous transformation. NEPTUNE digital pathology scoring system enables novel morphologic profiling of renal structures. For all histologic and ultrastructural descriptors tested with sufficient observations, moderate-to-excellent concordance was seen for 31/54 (57%). Descriptors not sufficiently represented will require further testing. This study proffers the NEPTUNE digital pathology scoring system as a model for standardization of renal biopsy interpretation extendable outside the NEPTUNE consortium, enabling international collaborations.