Interobserver agreement for post mortem renal histopathology and diagnosis of acute tubular necrosis in critically ill patients.

Abstract

The renal histopathology of critically ill patients dying with acute kidney injury (AKI) in intensive care units of high income countries remains uncertain. Retrospective observational assessment of interobserver agreement in the reporting of renal post mortem histopathology, and the ability of pathologists blinded to the clinical context to independently identify the presence of pre-mortem AKI from digital images of histological sections from 34 critically ill patients dying in teaching hospitals in Australia and Canada. We identified a heterogeneous cohort with a median age of 65 years (interquartile range [IQR], 56.5-77), APACHE II score of 27 (IQR, 19-33), and sepsis as the most common admission diagnosis (12/34; 35%). The most common proximate causes of death were cardiovascular (19/34; 56%) and respiratory (7/34; 21%) failure. AKI was common, with 23 patients (68%) developing RIFLE-F AKI, and 21 patients (62%) receiving renal replacement therapy. Structured reporting for tubular inflammation showed excellent agreement (kappa = 1), but no other subdomain demonstrated better than moderate agreement (kappa < 0.6). Only fair agreement (55.9% of cases; kappa = 0.23) was demonstrated on the diagnosis of moderate to severe acute tubular necrosis (ATN). Pathologist A predicted RIFLE-I or worse AKI with the diagnosis of ATN, with an overall accuracy of 61.8%; pathologist B predicted AKI with an accuracy of 35.3%. Post mortem assessment of the renal histopathology in critically ill patients is neither robust nor reproducible; independent pathologists agree poorly on the diagnosis of ATN, and their structural assessment appears dissociated from ante-mortem renal function.