Even before publication of the Banff schema  (fig. 1), a study was underway to assess intra- and interobserver variability in application of the schema by pathologists. That study has identified problem areas in semiquantitative biopsy scoring of acute changes according to Banff criteria (fig. 2), areas which must be addressed in optimizing the schema  (fig. 3]).
Glomerulitis is the accumulation of monocytes and lymphocytes in glomerular capillaries with swelling of endothelial cells (fig. 4]). To improve agreement in grading of glomerulitis, silver stain of PAS stain, which outline the glomerular capillary loops, may help in localizing inflammatory cells in the glomerulus. An additional strategy which could substantially improve agreement is immunostaining for Common Leukocyte Antigen (CLA) or other lymphocyte/monocyte markers. Intraglomerular mononuclear cells could then be reliably quantitated, and potentially expressed as number of cells per glomerulus, with scoring based on agreed-upon thresholds for mild, moderate, and severe, The significance of glomerulitis in the rejection process, and it's implication for long-term prognosis, remain controversial; by carefully defining and quantitating glomerulitis, investigators should be able to better address the importance of this lesion.
Interstitial inflammation is graded by the percent of renal parenchyma I the biopsy with an inflammatory infiltrate (non-polymorphonuclear). Agreement among observers in the grading of interstitial mononuclear cell inflammation was reasonable, with good agreement with morphometric assessment of volume fraction of inflammatory infiltrate. The schema grades mononuclear infiltrates by definition. However, in substantial numbers of cases, eosinophils are a significant component of the inflammatory infiltrate in the allograft. Since numerous eosinophils may portend a poor prognosis () and/or may ultimately herald the need for altered therapeutic strategies, they are included in grading of the interstitial infiltrate. Plasma cells in the infiltrate, particularly if numerous and if there is no tubulitis (fig. 5), must be evaluated carefully to rule out possible PTLD; PTLD may occur within weeks of transplantation. If the cells are not atypical, and if accompanied by tubulitis, the plasma cells are likely a component of the rejection process. If eosinophils and/or plasma cells are present is significant numbers, an asterisk could be added to the quantitative grade, to flag these interesting cases. The presence of polymorphonuclear leukocytes (PMN) in the allograft infiltrate may be part of a rejection process, and in particular may signal an antibody-mediated rejection (fig. 6). However, PMN are so non-specific, that they are not graded as part of the rejection process.
Tubulitis defined by lymphocytic infiltration across the tubular basement membrane into the tubular epithelium (fig. 7), is a cardinal feature in grading acute rejection by the Banff schema. Agreement among pathologists in grading this lesion was adequate, though there is clearly room for improvement in evaluation of this critical component of the grading schema. Tubulitis is most easily quantitated using PAS or silver stains, enabling quantitation of inflammatory cells which have breached the basement membrane; these stains also help in identifying fragments of basement membrane in cases with "vanished" tubules, tubules which have been overrun and largely destroyed by inflammatory cells (figs. 8 & 9). Failure to recognize residual tubular structures in these lesions may lead to underscoring of severity. Tubulitis in atrophic tubules (fig. 10), may not correlate with allograft dysfunction. While perhaps not irrelevant to the evolution of chronic allograft changes, this finding apparently has no acute implications for the graft, and "tubulitis" in moderately to severely atrophic tubules, as defined for example by tubular diameter, should probably not be included in the grading of severity of tubulitis. There is some concern that apoptotic cells in the epithelium may mimic infiltrating cells, though this problem is lessened by adherence to strict morphologic criteria. Clearly, use of immunostaining for CLA or other markers, especially with PAS counter-stain to define boundaries of the tubules, would enhance recognition and quantitation of this lesion as well.
Vasculitis is graded in most cases by severity and extent of intimal arteritis (figs. 11 & 12). Problems in agreement ma arise when focal vasculitis lesions are not seen by one observer, of when other changes such as interstitial hemorrhage (fig. 13), or necrosis of vessel walls (fig. 14) are interpreted differentially. Necrosis of vessel walls without inflammation may be seen in antibody-mediated rejection reactions, but may also be due to drug reaction, e.g. the HUS-like lesion produced by cyclosporine; severe hypertension; or recurrent HUS. Interstitial hemorrhage present as an isolated finding may be due to rejection-related vascular injury not sampled by the biopsy However, other causes, including parenchymal infarction not due to rejection, may cause interstitial hemorrhage, and care must also be taken not to misinterpret hemorrhage and parenchymal injury at a previous biopsy site. Staining of additional sections may resolve this differential diagnosis. The presence of inflammation in arterioles (figs. 15 & 16), not classically involved in rejection vasculitis, also needs further study.
Summary - In general, the Banff Working Formulation provides objective criteria for diagnosis and grading of acute rejection, and should be a reliable tool in the hands of most pathologists. However, as with any such schema, discrepancies may arise, and improvement and refinement are both possible and desirable. Agreement should be enhanced by more careful definitions of morphologic criteria. In addition, immunostaining, even nonspecific staining, for lymphocyte/monocyte, would be a useful adjunct for identifying and quantitating glomerular, interstitial, and tubular lesions. More specific immunostains, e.g. for effector cell populations, cytokine, and/or adhesion molecules, may also eventually be validate and incorporated into this rejection assessment schema.