**Department of Pathology,
Johns Hopkins Medical Institutions,
Baltimore Maryland 21205, USA
***University Institute Of Pathology,
University of Aarhus,
****Division of Nephrology and Immunology Department of Medicine,
Faculty of Medicine, University of Alberta
Edmonton, Canada T6G 2A9
On the occasion of the Third Banff Conference on Allograft Pathology to be held July 20-24, 1995 in Banff, Canada, it is useful to reflect on the progress since the first such meeting four years ago and the work that still remains to be done. First, the successes: The original Banff Classification of Renal Allograft Pathology  is in wide use and there have been extensive studies showing its clinical validity and reproducibility [2-10]. Extrapolating from these studies, it is likely that the classification has resulted in improvement in patient care. The Banff classification has been endorsed by the FDA and other regulatory agencies and has enabled the use of objective histologic endpoints for international clinical trials of new antirejection agents and other scientific studies.
Concepts which grew out of the development of the Banff classification are now being applied to liver allograft pathology and pathology of the native kidney.
As a background for deliberations at the Banff Conference ten problems
There will always be some inaccuracy in rejection diagnosis based on the inevitable sampling error problems inherent in percutaneous biopsies .
Any attempt to make more stringent the criteria for rejection diagnosis [see below] will increase this problem of underdiagnosis in patients with suspected rejection.
This overdiagnosis of rejection is particularly a feature of late graft biopsies, in which additional criteria should probably be established to raise the threshold for rejection diagnosis.
The work of Rush et al. [8, 9] clearly indicates that the phenomenon of treatable subclinical rejection exists. The difficulty is in establishing a reasonable dividing line between a true false positive diagnosis and the identification of subclinical rejection in which treatment should be considered.
We have written previously about the need to use well-performed PAS stains  to identify the key lesions of the Banff classification. For laboratories with routine access to immunoperoxidase procedures staining for common leukocyte antigen (anti CD-45) may add precision to lymphocyte identification. However we have observed that the Dako reagent which gives the crispest specific staining in paraffin sections does not stain all lymphocytes, whereas the Becton Dickinson antibody does stain all lymphocytes but also gives nonspecific reactivity in tubular cells. So presently it appears that there is no reagent which is ideal for this purpose.
In this regard we have become a victim of our own success. It was never intended that treatment decisions would be based entirely on biopsy findings divorced from clinical data. This is why the far right column in Table 5 (p. 417) of the original Banff classification paper is labeled "Possible clinical approach" . Any attempt to regard this column as a strong treatment recommendation certainly should be opposed. As a consequence of the false negatives due to sampling error mentioned above there frequently are patients with biopsies in the "borderline changes" category who deserve antirejection treatment. The strength of clinical suspicion of rejection should be the guide for therapy in such instances.
Pharmaceutical companies running international clinical trials of new anti-rejection agents represent very important "consumers" of the pathology diagnoses rendered using the Banff schema. Central slide review for such trials has been conducted thus far without knowledge of clinical history.
If the classification is modified to allow for different interpretation
in renal failure and protocol-normal function biopsies, clinical trails
could accommodate to this is one of two ways:
As an initial approach the recommendations of Paul et al  and Mihatsch et al  for defining chronic rejection will be followed but with additional incorporation of peritubular capillary changes as described by Monga et al. . There is a strong need to decide whether any estimation of peritubular capillary changes can be made by light microscopy or whether the diagnostic changes can only be revealed using electron microscopy.
Consideration needs to be given to the diagnosis of late-appearing antibody-mediated rejection in which glomerulitis, peritubular capillary accumulation of polymorphs, and severe vasculitic and thrombotic vascular lesions are three of the defining features [17, 18] (See Abstract below). An important decision to be made at the upcoming Banff meeting is whether to incorporate this entity into the classification in 1995 or wait until 1997 when it is better understood.
More data is needed on the rapidity of onset of hyaline arteriolar change  and the relative significance of "specific" and "nonspecific" arteriolar thickening as described by Mihatsch . Since the studies of both Mihatsch et al  and Marcussen et al.  show poor reproducibility in assessment of this lesion an alternative approach to quantitation should be sought. Location of the lesions within the arteriolar tree may be helpful in differential diagnosis [19, 20].
Experience with central slide review in international clinical trials clearly shows the great value of a baseline implantation biopsy [21-23].
Many more such biopsies should be performed to improve the accuracy of assessment of vascular lesions in later biopsies, but moderate degrees of scarring and vascular disease should not be interpreted as contraindications to use of the kidney. In the studies of Curschellas et al  62% of implantation biopsies showed significant morphologic changes, most commonly fibrous intimal thickening or hyaline arteriolar thickening. Despite these changes 82% of transplant recipients had a good initial and long term function. The authors suggest that since age-related vascular disease does not seem to compromise renal function after transplantation, a more liberal choice of donors should be considered.
After three months post-transplant we are considering differentiating biopsies obtained at time of suspected rejection from those obtained in grafts with stable function, and applying different histologic thresholds for diagnosis of rejection in these two settings. In normally functioning grafts with tubulitis but no intimal arteritis the proposal is that two further criteria must be met before a diagnosis of rejection can be made:
Incorporating the Colvin criteria would mean that to diagnose rejection in the presence of tubulitis without arteritis, two of the following three lesions would have to be present: edema, tubular injury (in association with the inflammatory infiltrate), or activated lymphocytes (judged on size and appearance on PAS and H&E).
Excluding tubulitis in atrophic tubules as defined by Mihatsch  would mean not counting lymphocytic invasion in tubules with (p. 27) thickened tubular basement membranes and reduced diameter. In the absence of thickened basement membranes atrophy would be regarded as present when tubular diameter was reduced to less than 50% of normal values, as determined by comparison with adjacent tubules." Tubulitis in atrophic tubules would also not be counted as evidence of acute rejection in biopsies obtained after one year post-transplant. However, for biopsies obtained in the setting of suspected rejection in the early post-transplant period the basic construct of the Banff Schema would remain the same, thus maintaining the integrity of ongoing clinical trials.
Alloantibody frequently appears during the immune response to alloantigens in renal transplant recipients. We studied whether the presence of antibody against donor class I antigens correlated with clinical and pathologic features of acute rejection episodes. We identified patients who had (i) clinical evidence of acute rejection; (ii) a renal biopsy showing pathologic features of acute rejection, defined by the Banff criteria; and (iii) pre- and post-transplant sera screened against donor T cells. We divided these patients into those with or without donor specific alloantibody reactive with donor T cells, presumed to be anti class I. Of 44 patients meeting these criteria, 20 were antibody negative (Ab-R) and 24 were antibody positive (Ab+R). In patients with multiple rejections only the first episode meeting these criteria was tested. The Ab+R patients had a higher incidence of severe vasculitis (V3) (p<.0009) and glomerulitis (p<.02). Ab-R patients more often had tubulitis than Ab+R patients: Moderate (T2) and severe(T3) tubulitis was present in 19/20 (95%) of Ab-R patients, versus 12/24 (50%) of Ab+R patients (p<.003). No renal biopsies in the Ab-R group had infarction, compared with 9/24 in the Ab+R group (p<.003). Polymorphs in peritubular capillaries were more frequent in Ab+R patients (p<.003). Fibrin thrombi in glomeruli and/or vessels, fibrinoid necrosis, and dilatation of peritubular capillaries were also more frequent in the AB+R group. Graft loss was increased in the Ab+R patients, particularly graft loss before 3 months (12/24 compared with 3/20, p<.025).
These results indicate that, during biopsy proven acute rejection episodes, the presence of anti-class I antibody in the early post-transplant period correlates with and probably causes severe vascular lesions, glomerulitis, infarction, and early graft loss. Thus the findings of fibrin thrombi in small blood vessels, fibrinoid necrosis in vessel walls, glomerulitis, and polymorphs in peritubular capillaries in renal transplant biopsies should suggest that the rejection is associated with anti class I antibody.