Aims: The histopathological diagnosis of infection in periprosthetic tissue from loose total joint endoprosthesis has been the subject of controversy. The aim was to define a histological criterion that would best differentiate between aseptic and septic endoprosthesis loosening.
Methods and results: Neutrophilic granulocytes (NG) were enumerated histopathologically in 147 periprosthetic membranes obtained from aseptic and septic revision surgery, using periodic acid-Schiff (PAS) stains and CD15 immunohistochemistry. Cell numbers were correlated with the results of microbiological culture and the clinical diagnoses. Using receiver-operating characteristics, an optimized threshold was found at 23 NG in 10 high-power fields (HPF). Using this threshold, histopathological examination had a sensitivity of 73% and specificity of 95% when compared with microbiological diagnosis (area under the curve 0.881), and a sensitivity of 77% and specificity of 97% when compared with clinical diagnosis (area under the curve 0.891).
Conclusions: We therefore recommend a counting algorithm with a threshold of > or =23 NG in 10 HPF (visual field diameter 0.625 mm) for the histopathological diagnosis of septic endoprosthesis loosening. If the enumeration of NG is difficult in conventional haematoxylin and eosin-stained slides, CD15 immunohistochemistry should be performed, whereas the PAS stain has not proven to be helpful.