Validating the oxford classification of IgA nephropathy.

The evolution of patients with IgA nephropathy (IgAN) varies greatly, and predictive tools are needed to assess confidently the risk for progression. Although in many studies clinical risk factors such as hypertension and proteinuria have consistently been linked to outcome, the independent value of pathology remains debated. Pathologists have developed a number of classifications over the years, at times with seemingly conflicting results. In 2009, the International IgA Nephropathy Network Group, working in collaboration with the Renal Pathology Society, proposed a set of four predictive renal biopsy findings independent of clinical assessment, the MEST score (M, mesangial hypercellularity; E, endocapillary proliferation; S, segmental glomerulosclerosis/adhesion; T, tubular atrophy/interstitial fibrosis) (1,2). This study differed from previous classification by using a stepwise methodology initially considering all plausible variables, subsequently eliminating those with poor reproducibility, avoiding highly collinear biopsy findings, testing their univariate predictive value and, finally, adjusting for clinical risk factors of progression. It also simultaneously considered two important, albeit different, outcomes: The rate of renal function decline and the survival from a 50% reduction of renal function or renal failure. Despite this meticulous approach, certain limitations were unavoidable. Its retrospective design and uncontrolled treatment allocation could have influenced the results. Validation of the Oxford classification has been carried out in children as well as adults (3,4). Verification is necessary because premature implementation of biased conclusions can harm patients and mislead future research. It has been emphasized repeatedly that a validation study should consist of an adequate sample of different but related patients compared with the derivation population (5). “Relatedness” is defined by patients at risk for the same event. Whether a decision rule can be extrapolated to a population at different risk can never be confidently answered from the initial study, regardless of its quality. The original Oxford study excluded patients with an (1) initial estimated GFR (eGFR) 30 ml/min per 1.73 m, (2) proteinuria 0.5 g/d, and (3) follow-up period of less than 12 months. On the basis of these criteria, one can argue that three important subgroups of patients with IgAN were disregarded from the classification: Patients with advanced disease, patients with low-grade proteinuria unlikely to progress over a few years, and, finally, the rare but important patient presenting with a rapidly progressive form of IgAN. In this issue of CJASN, Alamartine et al. (6) carried a validation endeavor in 183 adult patients who had IgAN and were followed for an average of 77 months. Pathologic assessment was assessed and covariates gathered identically to the original study. The authors stated that very mild and very severe forms of the disease were not a priori excluded. Many wished the original study had included them, and addressing these groups is clinically relevant. Consequently, differences in the study populations have surfaced. Mesangial and endocapillary hypercellularity was present in 81% and 42% in the Oxford cohort, respectively, as opposed to 21% and 14% in this study. The validation cohort was 13 years older without children, and the initial proteinuria was 1.24 g/d as opposed to 1.7 g/d originally. Finally, 13 and 12 patients presented with stages 4 and 5 chronic kidney disease (CKD), respectively. Treatment allocation seemed similar. Clinicopathologic correlations mirrored those found in the Oxford study with the notable exception of the M lesion showing no relation with proteinuria. The E, S, and T lesions also predicted the survival from a combined event by univariate analysis, but none remained significant when adjusting for clinical variables. In the multivariate model, only baseline eGFR was a positive predictive factor for doubling of serum creatinine or ESRD. Interestingly, proteinuria and BP also failed to predict survival independently from a combined event when factoring in the initial eGFR. Both were strongly associated with outcome univariately, as expected. Table 3 offers a glimpse of the important collinearity between tubulointerstitial disease and eGFR when the hazard ratio for the T lesion jumps from 1.24 to 5.11 when eGFR is removed from the equation. The original study also showed this attenuation, to a lesser extent, perhaps having excluded patients with stages 4 and 5 CKD. Using the rate of renal function decline (slope) as an outcome allows the impact of the initial eGFR that weights strongly on renal survival analyses to be distinguished. However, estimating the slope can be difficult and biased by a single outlier. Consistency using both outcomes offers reassurance. *Department of Nephrology, Hôpital du Sacré-Coeur de Montréal, University of Montreal, Montreal, Québec, Canada; and Division of Nephrology and Hypertension, Mayo Clinic, Rochester, Minnesota