Key Words: morphologic criteria, prognosis, glomerulonephritis
Since the earliest application of the technique, renal biopsies have been performed to guide therapy and to render prognosis. Progressive correlation of morphologic features with clinical presentation and outcome led to a categorical system of diagnostic classification [1-3]. Criteria, including immuno-fluorescence and electron microscopic features, have become well established, and within several categories, "staging" based upon observations on sequential biopsies is also included. Particularly with respect to glomerulonephritides, experimental studies and clinical series have associated particular diagnostic categories with pathogenic mechanisms. All of these efforts are steeped in the traditional perspective that posits a relation between structure and function of biological systems. Nonetheless, the predictive power of the renal biopsy falls short of expectation. Witness, for example, the attempts to improve resolution of prognosis in lupus nephritis. Solez et al have recently noted limitations intrinsic to the current approach to renal biopsy, and have argued eloquently on the need for improved standardization of morphologic features and for a finer discrimination that might be afforded by supplementing morphologic evaluation with nucleic acid or protein phenotypic markers .
Certainly, sharper distinction among entities that are in the same category can develop from more quantitative and/or molecular approaches, if in fact uniform standards can be agreed upon. However, a major yet unspoken limitation to the employment of morphologic features for the establishment of prognosis and/or pathogenesis relates to the lack of a monatomic relationship between morphologic classification and either clinical course or pathogenesis. Membranous nephropathy is perhaps the lesion with the most consistent and standardized criteria for diagnosis among centers, and many agree that the pathogenetic mechanism that underlies membranous nephropathy is "in situ" formation of immune deposits. Yet, the diagnosis of membranous nephropathy carries virtually no prognostic power, in that roughly one-third of patients with this diagnosis progress to end stage, another one third maintain good function but with heavy proteinuria, and another one third spontaneously remit.
The problem, briefly stated, is not necessarily related to limitations posed by a morphologic approach, although this is certainly an important potential factor. Rather, the problem may lie with uniform extraction of appropriate morphologic features. As a codified set of features that are highly standardized and reproducible with defined criteria [4, 5], the Banff schema offers an opportunity to classify native renal biopsies as well as allograft renal biopsies. We examined the potential utility of the Banff schema for guiding therapy and rendering prognosis of patients undergoing biopsies of native kidney.
The age and sex of the patient were noted. Serum creatinine and cholesterol values and urinary protein excretion were recorded with reference to the time of biopsy, and the slope of reciprocal creatinine against time was calculated. Based on their functional status five years after biopsy, patients were classified in terms of renal functional impairment as 0 (less than 150% of the serum creatinine predicted from lean body weight and surface area), 1 (150-300% of predicted), 2 (300-500% of predicted), 3 (more than 500% predicted but not on renal replacement therapy) or 4 (on maintenance dialysis or transplanted). The interval from biopsy to end-stage was also noted for patients assigned to class 4.
The "traditional" categorical diagnosis [1-3] for each biopsy was noted, and all biopsies were re-reviewed and scored exactly according to the Banff schema for each of the acute and chronic features , with the exception that chronic glomerulopathy was based on capillary wall thickening and mesangial matrix expansion only, without the requirement for capillary wall splitting or cellular infiltration. The number of patent, partially sclerotic and totally sclerotic glomeruli were each recorded, and the percents of each category and the sum of the two categories of sclerosis were calculated.
A correlation matrix [Table 1] indicated that tubular atrophy and involution (ct) and chronic vascular change (cv) were highly correlated with each other, and well correlated with interstitial lymphohistiocytic infiltration (i) and arteriolar hyaline deposition (ah); in all cases, the correlation coefficients exceeded 0.58 (all p<0.001). Chronic glomerulopathy was equally well correlated with glomerulitis (r=0.77, p<0.001). Interstitial fibrosis correlated less strongly (0.4 < r < 0.5) but still significantly (all p<0.01) with tubular atrophy (ct), arteriolar fibroelastosis (cv) and tubulitis, and tubulitis (t) also correlated with interstitial lymphohistiocytic infiltration (i) to a similar degree. The percentage of totally sclerotic glomeruli and the sum of totally and partially sclerotic glomeruli correlated significantly with i, ah, ci, ct and cv.
When multiple linear regression was applied to control for the correlations among the variables, interstitial infiltration (i) and arteriolar hyalinosis (ah) correlated with deteriorated renal function at five years as independent variables (partial F's > 6.2); these variables, and no others in the matrix, also proved significantly correlated with function in a stepwise regression model. Arteriolar hyalinosis also correlated directly and independently with erosion of glomerular filtration, estimated by a negative correlation with the slope of reciprocal serum creatinine plotted against the time after renal biopsy (F = 4.2); glomerulitis correlated positively with the slope of reciprocal creatinine versus time after biopsy, also as an independent variable (F=7.7). None of the other variables in the matrix proved to be independently correlated with erosion of glomerular filtration rate. Interestingly, by analysis of variance, there was no significant difference among the various traditional diagnostic groups in terms of renal functional status five years after biopsy, and serum cholesterol, initial serum creatinine and urinary protein had no significant correlation.
The observations presented here offer insights with ramifications for the approach to nosology, the appreciation of mechanisms operative in the progression of primary glomerular disease, and the identification of suitable therapeutic goals. As suggested elsewhere by other investigators [6-9], the functional outcome of glomerular disease is poorly predicted by the diagnostic category. However, individual morphologic features, formally a part of the Banff schema and typically an inform al part of the more traditional interpretation of biopsies, prove portentous. In this regard, the presence and severity of interstitial infiltrates of mononuclear cells and arteriolar hyalinosis represent potential prognosticators that are readily available and easily standardized from current practice. It seems that at least these elements of the Banff schema have a much better predictive value for outcome of native renal biopsies, at least those derived from primary glomerular diseases, than traditional diagnostic criteria.
While this concept is certainly not original with this contribution [1-5], application of the Banff schema facilitates uniform quantification. It is worthwhile to note that even if the interstitial infiltrates and arteriolar hyalinosis are not intrinsic parts of the lesion, they may be important factors in the attendant progressive loss of renal function. Therefore, anti-inflammatory therapy that reduces interstitial inflammation, and rigorous control of blood pressure that may attenuate arterial hyalinosis, may be powerful therapeutic modalities, even if they do not ameliorate the underlying intraglomerular process. At the same time, i and ah scores may be useful in gauging therapeutic response.
The data compiled in our center are too scanty to permit conclusions as such. Conclusive ratification of morphologic parameters such as interstitial nephritis must await both increased numbers of cases and stratification of patients within diagnostic categories, so that patients that are similar in other respects but who differ in their Banff "i" score can be compared for a variety of primary disease categories. Among our series of biopsies, the worst functional outcome was observed in patients with diffuse proliferative lupus nephritis and focal and segmental glomerulosclerosis, which are also lesions notorious for a conspicuous tubulointerstitial nephritis component. Conceivably, the correlation of the "i" score with severely deteriorated function after five years' follow-up simply reflects the greater interstitial nephritis component observed in patients with these categorical diagnoses.
Notwithstanding these limitations, the Banff schema can be applied readily and with some apparent utility, to the classification of native renal biopsies. If additional refinements based on phenotypic or molecular markers appear to bear value, they can be added to a native biopsy classification, just as the same or similar markers might, in the future, be applied to allograft biopsies . Given the difficulties in gathering large numbers of patients in a single center, multi-center longitudinal studies, prospective or retrospective, seem mandated. In this regard, the reproducible and defined criteria intrinsic to the Banff schema can facilitate inter-institutional studies.
Table 1: Correlation Among Pathologic Parameters Scored