Background Studies have shown that assessing renal histopathology for lupus nephritis is subject to interobserver variability.1,2 Local pathologists’ readings of lupus nephritis biopsies are commonly used to determine entry into multicenter interventional trials. It is not known to what extent those that are enrolled in such trials are correctly being identified.
Objectives First, to assess the interobserver agreement between local pathologists and an expert panel of 5 European nephropathologists. Secondly, to provide guidance for designing future interventional trials for lupus nephritis.
Methods A total of 91 locally classified Class III or IV biopsies from 22 centers from 12 countries were scored by a panel of 5 European nephropathologists, who were blinded to the local readings and the clinical data. The biopsies consisted of 70 first and 21 repeat biopsies from 70 patients with proliferative LN, included in BELONG, a randomized controlled trial assessing ocrelizumab vs. placebo. Each biopsy was scored independently by 2 of the panel members using a standardized histological form, and any discrepancies were resolved with periodic meetings of all 5 pathologists. Percent agreement on the overall ISN/RPS classification (class/subclass/chronicity) for local reading vs. expert panel was calculated. The kappa statistic was then used to determine the interobserver agreement on the sub-components of the classification.
Results There was agreement on only 22 of the 91 biopsies (24%) for the overall ISN/RPS classification. This low overall agreement was a result of only fair interobserver agreement on the class of LN (kappa 0.34, see Table 1), and poor agreement on segmental vs. global for class IV (kappa 0.18, see Table 2). Furthermore, the expert panel classified 5 biopsies as not being class III or IV and an additional 6 biopsies as having only chronic lesions. These 11 biopsies would have met the exclusion criteria in BELONG.
Conclusions This study shows that there is very low interobserver agreement of LN histopathology using the ISN/RPS classification between the local readings and the expert panel. Until a more reproducible LN classification scheme is available, clinical trials would benefit from having biopsies sent to a central panel for readings prior to determining inclusion.
 Schwartz MM. Am J Kidney Dis. 1993 April. PMID 8465815
 Furness PN. Am J Surg Pathol 2006 August. PMID 16861976
Disclosure of Interest K. Rao Consultant for: Genentech, F. Ferrario Consultant for: Roche Products, Ltd., T. Cook Consultant for: Roche Products, Ltd., I. Bajema Consultant for: Roche Products, Ltd., J. Bruijn Consultant for: Roche Products, Ltd., L.-H. Noel Consultant for: Roche Products, Ltd., R. Maciuca Employee of: Genentech, P. Brunetta Employee of: Genentech, S. Shahdad Employee of: Roche Products, Ltd.