Validation of the Oxford classification of IgA nephropathy
2011; Elsevier BV; Volume: 80; Issue: 3 Linguagem: Inglês
10.1038/ki.2011.126
ISSN1523-1755
AutoresAndrew M. Herzenberg, Agnes B. Fogo, Heather N. Reich, Stéphan Troyanov, Nüket Bavbek, Alfonso Eirin, Tracy E. Hunley, Michelle Hladunewich, Bruce A. Julian, Fernando C. Fervenza, Daniel Cattran,
Tópico(s)Systemic Lupus Erythematosus Research
ResumoThe Oxford classification of IgA nephropathy (IgAN) identified four pathological elements that were of prognostic value and additive to known clinical and laboratory variables in predicting patient outcome. These features are segmental glomerulosclerosis/adhesion, mesangial hypercellularity, endocapillary proliferation, and tubular atrophy/interstitial fibrosis. Here, we tested the Oxford results using an independent cohort of 187 adults and children with IgAN from 4 centers in North America by comparing the performance of the logistic regression model and the predictive value of each of the four lesions in both data sets. The cohorts had similar clinical and histological findings, presentations, and clinicopathological correlations. During follow-up, however, the North American cohort received more immunosuppressive and antihypertensive therapies. Identifying patients with a rapid decline in the rate of renal function using the logistic model from the original study in the validation data set was good (c-statistic 0.75), although less precise than in the original study (0.82). Individually, each pathological variable offered the same predictive value in both cohorts except mesangial hypercellularity, which was a weaker predictor. Thus, this North American cohort validated the Oxford IgAN classification and supports its utilization. Further studies are needed to determine the relationship to the impact of treatment and to define the value of the mesangial hypercellularity score. The Oxford classification of IgA nephropathy (IgAN) identified four pathological elements that were of prognostic value and additive to known clinical and laboratory variables in predicting patient outcome. These features are segmental glomerulosclerosis/adhesion, mesangial hypercellularity, endocapillary proliferation, and tubular atrophy/interstitial fibrosis. Here, we tested the Oxford results using an independent cohort of 187 adults and children with IgAN from 4 centers in North America by comparing the performance of the logistic regression model and the predictive value of each of the four lesions in both data sets. The cohorts had similar clinical and histological findings, presentations, and clinicopathological correlations. During follow-up, however, the North American cohort received more immunosuppressive and antihypertensive therapies. Identifying patients with a rapid decline in the rate of renal function using the logistic model from the original study in the validation data set was good (c-statistic 0.75), although less precise than in the original study (0.82). Individually, each pathological variable offered the same predictive value in both cohorts except mesangial hypercellularity, which was a weaker predictor. Thus, this North American cohort validated the Oxford IgAN classification and supports its utilization. Further studies are needed to determine the relationship to the impact of treatment and to define the value of the mesangial hypercellularity score. IgA nephropathy (IgAN) is the most common glomerular disease worldwide. The outcome of patients with IgAN varies greatly.1.Bartosik L.P. Lajoie G. Sugar L. et al.Predicting progression in IgA nephropathy.Am J Kidney Dis. 2001; 38: 728-735Abstract Full Text Full Text PDF PubMed Scopus (297) Google Scholar, 2.Woo K.T. Edmondson R.P. Wu A.Y. et al.The natural history of IgA nephritis in Singapore.Clin Nephrol. 1986; 25: 15-21PubMed Google Scholar, 3.Droz D. Kramar A. Nawar T. et al.Primary IgA nephropathy: prognostic factors.Contrib Nephrol. 1984; 40: 202-207Crossref PubMed Google Scholar, 4.Geddes C.C. Rauta V. Gronhagen-Riska C. et al.A tricontinental view of IgA nephropathy.Nephrol Dial Transplant. 2003; 18: 1541-1548Crossref PubMed Scopus (150) Google Scholar Clinical variables available to physicians treating patients with IgAN account for less than half of the variability in outcome.1.Bartosik L.P. Lajoie G. Sugar L. et al.Predicting progression in IgA nephropathy.Am J Kidney Dis. 2001; 38: 728-735Abstract Full Text Full Text PDF PubMed Scopus (297) Google Scholar Assessing prognosis is challenging and has contributed to the lack of consensus regarding the management of these patients. The additive value of pathology features to the known clinical parameters of proteinuria, serum creatinine, and blood pressure in predicting long-term outcome has long been debated.5.D'Amico G. Natural history of idiopathic IgA nephropathy: role of clinical and histological prognostic factors.Am J Kidney Dis. 2000; 36: 227-237Abstract Full Text Full Text PDF PubMed Scopus (432) Google Scholar, 6.Donadio J.V. Bergstralh E.J. Offord K.P. et al.Clinical and histopathologic associations with impaired renal function in IgA nephropathy. Mayo Nephrology Collaborative Group.Clin Nephrol. 1994; 41: 65-71PubMed Google Scholar, 7.Johnston P.A. Brown J.S. Braumholtz D.A. et al.Clinico-pathological correlations and long-term follow-up of 253 United Kingdom patients with IgA nephropathy. A report from the MRC Glomerulonephritis Registry.Q J Med. 1992; 84: 619-627PubMed Google Scholar, 8.Lee S.M. Rao V.M. Franklin W.A. et al.IgA nephropathy: morphologic predictors of progressive renal disease.Hum Pathol. 1982; 13: 314-322Abstract Full Text PDF PubMed Scopus (260) Google Scholar, 9.Haas M. Histologic subclassification of IgA nephropathy: a clinicopathologic study of 244 cases.Am J Kidney Dis. 1997; 29: 829-842Abstract Full Text PDF PubMed Scopus (386) Google Scholar A recent international effort made by a group of expert nephrologists and nephropathologists using clinical, laboratory, and pathology features identified a set of distinct pathological variables that demonstrated an independent prognostic value in patients with IgAN, known as the Oxford classification.10.Cattran D.C. Coppo R. Cook H.T. et al.The Oxford classification of IgA nephropathy: rationale, clinicopathological correlations, and classification.Kidney Int. 2009; 76: 534-545Abstract Full Text Full Text PDF PubMed Scopus (812) Google Scholar, 11.Roberts I.S. Cook H.T. Troyanov S. et al.The Oxford classification of IgA nephropathy: pathology definitions, correlations, and reproducibility.Kidney Int. 2009; 76: 546-556Abstract Full Text Full Text PDF PubMed Scopus (682) Google Scholar In 265 cases from centers across continents, four pathology features were found to be of independent value in terms of predicting the outcome of renal function: mesangial hypercellularity score (M; M0≤0.5, M1>0.5), the presence of endocapillary proliferation (E; E0: absent, E1: present), segmental glomerulosclerosis/adhesion (S; S0: absent, S1: present), and severity of tubular atrophy/interstitial fibrosis (T; T0≤25%, T1: 26–50%, T2>50%). The MEST score remained statistically significant for prediction of the outcome even after taking into account the clinical indicators available both at the time of biopsy and during the observation period. These findings have the potential to greatly affect nephrologists' approach to IgAN by improving the capacity to identify, at the time of biopsy, high-risk-of-progression patients in contrast to other indicators that require an observation period before helping with prognosis.12.Reich H.N. Troyanov S. Scholey J.W. et al.Remission of proteinuria improves prognosis in IgA nephropathy.J Am Soc Nephrol. 2007; 18: 3177-3183Crossref PubMed Scopus (370) Google Scholar Definitive pathological features would significantly modify not only the approach by the physician to the individual patient but also the framework for designing and interpreting therapeutic trials. However, before widespread adoption of this new Oxford classification system, validation of its findings is essential.13.Yamamoto R. Imai E. A novel classification for IgA nephropathy.Kidney Int. 2009; 76: 477-480Abstract Full Text Full Text PDF PubMed Scopus (19) Google Scholar, 14.Bleeker S.E. Moll H.A. Steyerberg E.W. et al.External validation is necessary in prediction research: a clinical example.J Clin Epidemiol. 2003; 56: 826-832Abstract Full Text Full Text PDF PubMed Scopus (428) Google Scholar, 15.Toll D.B. Janssen K.J. Vergouwe Y. et al.Validation, updating and impact of clinical prediction rules: a review.J Clin Epidemiol. 2008; 61: 1085-1094Abstract Full Text Full Text PDF PubMed Scopus (341) Google Scholar Validation is often neglected in medical prognostic models and has been rarely performed in kidney disease despite its obvious importance.15.Toll D.B. Janssen K.J. Vergouwe Y. et al.Validation, updating and impact of clinical prediction rules: a review.J Clin Epidemiol. 2008; 61: 1085-1094Abstract Full Text Full Text PDF PubMed Scopus (341) Google Scholar Our objective was to assemble an independent cohort of IgAN patients to test the validity of the Oxford pathology classification. A total of 187 patients were recruited from 4 different centers in North America (NA). In all, 44 patients (24%) were younger than 18 years at the time of biopsy, a proportion similar to that in the Oxford data set (Table 1). The median year of biopsy was 2002 for the NA validation cohort versus 1997 in the Oxford derivation data set (P<0.001).Table 1Age distribution and geographical origin of the NA validation cohortTotalAdultsChildren at biopsyn18714344Vanderbilt University551936Mayo Clinic38326University of Alabama at Birmingham22211University of Toronto72711Abbreviation: NA, North America. Open table in a new tab Abbreviation: NA, North America. The clinical characteristics of the validation cohort are presented in Table 2. Compared with the Oxford derivation cohort, there were no significant differences in age, estimated glomerular filtration rate (eGFR), proteinuria, and mean arterial pressure (MAP) at the time of biopsy. The NA validation cohort contained a greater proportion of female subjects (42 compared with 28%, P=0.001).Table 2NA validation cohort: clinical characteristics at the time of biopsy and follow-upAt the time of biopsyFollow-upAge (years)34 (18–45)Duration of follow-up (months)53 (36–77)Female42%MAP (mm Hg)92±11Pediatric at biopsy ( 0.1). At the time of biopsy, 58% of patients in the NA validation data set were on an antihypertensive medication versus 31% in the Oxford derivation data set (P=0.006). Moreover, at entry, a higher percentage of patients was receiving renin angiotensin system blockade (RASB) in the NA cohort (31 versus 20%, P=0.007). The NA validation cohort also received more therapy during the follow-up period with regard to immunosuppression (41% NA versus 29% Oxford, P=0.02), RASB (87 versus 74%, P=0.001), and number of anti-hypertensive medications (median of 1.1 versus 0.9, P 0.1). Kidney survival from a combined event was not statistically different. The agreements between the two pathologists (ABF and AMH) are shown in Table 3. The intraclass correlation coefficients (ICCs) were higher for each of the pathology variables assessed in comparison with the initial study, with the exception of the mesangial score.Table 3Intraclass correlation coefficients comparing the derivation Oxford and the validation North American cohortsOxford cohortNA cohortMesangial hypercellularity score0.640.37Percentage of total glomeruli showing segmental sclerosis0.460.94Percentage of total glomeruli showing adhesions0.200.76Percentage of total glomeruli showing endocapillary hypercellularity0.570.87Percentage of cortex showing tubular atrophy0.790.98Percentage of cortex showing interstitial fibrosis0.780.98Abbreviation: NA, North America. Open table in a new tab Abbreviation: NA, North America. The prevalence of lesions in the validation compared with the derivation data set is shown in Figure 1. Overall, the distributions of the pathology findings were similar in both cohorts with the exception of the mesangial score, with a median score of 1.07 (0.73–1.59) in the NA validation data set compared with 0.89 (0.60–1.30) in the Oxford derivation data set (P 0.1 in the validation cohort compared with the Oxford derivation hazard ratio of 0.07, 95% CI: 0.01–0.52, P=0.009). Interactions between the E lesion and immunosuppression from both cohorts are illustrated in Figure 6. The relationship between E lesions and the rate of renal-function decline is evident in those who did not receive immunosuppression and is lost in those who did. When both derivation and validation data sets with E lesions are combined, patients who never received immunosuppression during follow-up had a rate of renal-function decline of -5.0±9.6 ml/min per 1.73 m2 compared with -2.3±5.4 ml/min per 1.73 m2 in those who were treated (P=0.02). This difference remained significant after adjustment for eGFR, follow-up time-average proteinuria, and MAP. The discriminative strength of the logistic model was high in both cohorts with an area under the receiver-operating characteristic curve (or c-statistic) of 0.75 (95% CI: 0.68–0.82) in the NA validation cohort (0.70 using clinical variables only; 0.63 using pathology variables only). The c-statistic was originally 0.82 (95% CI: 0.77–0.87) in the Oxford data set (0.78 using clinical variables only; 0.70 using pathology variables only) (Figure 7). A recent international effort identified a set of distinct pathological variables with prognostic value, independent of all clinical and laboratory parameters in patients with IgAN.10.Cattran D.C. Coppo R. Cook H.T. et al.The Oxford classification of IgA nephropathy: rationale, clinicopathological correlations, and classification.Kidney Int. 2009; 76: 534-545Abstract Full Text Full Text PDF PubMed Scopus (812) Google Scholar, 11.Roberts I.S. Cook H.T. Troyanov S. et al.The Oxford classification of IgA nephropathy: pathology definitions, correlations, and reproducibility.Kidney Int. 2009; 76: 546-556Abstract Full Text Full Text PDF PubMed Scopus (682) Google Scholar Although these findings are currently insufficient to recommend specific treatment recommendations, they offer the possibility of a targeted approach to management based on an improved and semi-quantified risk assessment. Given the potential relevance of this proposed classification and its distinctive nature compared with standard classifications, we believed that validation was essential before recommending its implementation on a global scale. Three of the four pathological variables identified in the Oxford derivation cohort predicted a rapid rate of renal-function decline in the NA validation data set, supporting the original findings. Additional support of its validity is the closeness of fit across the quintiles of risk when the observed-to-predicted rate of renal-function decline is compared in both cohorts (Figure 4). The exception was the mesangial hypercellularity score (M-score), with a predictive value that underperformed in the NA data set. Although the added value of pathology to longitudinal clinical assessment may seem moderate (Figure 7), biopsy findings have the advantage of immediate long-term predictive value. The M-score deserves additional comment. Not only was its long-term predictive value reduced but also the cross-sectional correlation with proteinuria at the time of biopsy was absent. Possible explanations, other than a spurious relationship identified in the derivation study, are multiple. The lower ICC observed for this variable in this study may have resulted in error in measurement, leading to a non-significant finding. Alternative explanations include the different therapeutic strategies in the two cohorts, such as the greater anti-hypertensive treatment and RASB in the NA cohort, both during follow-up and before renal biopsy. Certainly others studies have shown that the effect of the renin-angiotensin system is associated with mesangial cell inhibition and prevention of renal-function decline in both human and experimental glomerulonephritis.16.Minutolo R. Balletta M.M. Catapano F. et al.Mesangial hypercellularity predicts antiproteinuric response to dual blockade of RAS in primary glomerulonephritis.Kidney Int. 2006; 70: 1170-1176Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar, 17.Miyake-Ogawa C. Miyazaki M. Abe K. et al.Tissue-specific expression of renin-angiotensin system components in IgA nephropathy.Am J Nephrol. 2005; 25: 1-12Crossref PubMed Scopus (37) Google Scholar As in the derivation data set, we found an interaction between the E-score and the use of immunosuppressive therapy and the rate of renal-function decline in the validation cohort. In the original derivation study, because of the small number of subjects, we could not determine the independent predictive value of the E-score, given this important potential confounder of treatment effect. However, by combining the two cohorts, we were able to provide further evidence that supports the independent predictive value of endocapillary proliferation (Figure 6). Our findings strongly corroborate the value of the inclusion of the E-score in the Oxford classification and address previous questions in this regard.13.Yamamoto R. Imai E. A novel classification for IgA nephropathy.Kidney Int. 2009; 76: 477-480Abstract Full Text Full Text PDF PubMed Scopus (19) Google Scholar These results also illustrate the potential role of renal biopsy findings in the selection of IgA patients requiring aggressive therapy. Although a recent study did not show a predictive value of endocapillary lesions, the prevalence of such findings was not reported, and potential interaction with immunosuppression was not assessed.18.Walsh M. Sar A. Lee D. et al.Histopathologic features aid in predicting risk for progression of IgA nephropathy.Clin J Am Soc Nephrol. 2010; 5: 425-430Crossref PubMed Scopus (72) Google Scholar Prediction models estimate an outcome using a mathematical equation combining multiple variables. Three steps guide their clinical utility: (1) development of a prediction equation, (2) validation of the accuracy and generalizability of findings, and (3) studying the clinical impact on physician behavior and patient outcome after implementation in daily practice. Although articles proposing predictive models based on risk factors are numerous, very few have been validated or have addressed how their implementation has affected patient care.19.Reilly B.M. Evans A.T. Translating clinical research into clinical practice: impact of using prediction rules to make decisions.Ann Intern Med. 2006; 144: 201-209Crossref PubMed Scopus (531) Google Scholar This failure to validate has potential serious consequences, because premature implementation of an imprecise model or inappropriate generalization of conclusions can cause harm to patients, and can misdirect treatment algorithms and clinical trials. True validation of any type is complex and has its own set of rules often not understood or followed. There are also various methodologies to test validity.20.Altman D.G. Royston P. What do we mean by validating a prognostic model?.Stat Med. 2000; 19: 453-473Crossref PubMed Scopus (1056) Google Scholar Internal validation can be performed by splitting a study cohort into two groups and using one for validation. More sophisticated methods repeat this procedure (bootstrap) many times on random data subsets to identify a more robust model.21.Schumacher M. Hollander N. Sauerbrei W. Resampling and cross-validation techniques: a tool to reduce bias caused by model building?.Stat Med. 1997; 16: 2813-2827Crossref PubMed Scopus (115) Google Scholar External validation is a superior method but does require the use of a new database, ideally from different centers and scientists.15.Toll D.B. Janssen K.J. Vergouwe Y. et al.Validation, updating and impact of clinical prediction rules: a review.J Clin Epidemiol. 2008; 61: 1085-1094Abstract Full Text Full Text PDF PubMed Scopus (341) Google Scholar A validation study classically requires an adequate sample of a different but related patient population as that found in the original derivation cohort. We followed our original Oxford inclusion and exclusion criteria to maintain this element of validation in our study. We recognize that, as discussed in the original paper, the elimination of the subsets of the IgAN patient population displaying extremes of the histological and/or clinical features, that is, those with very rapid progression or with no significant proteinuria may mean that our results cannot be applied to these subsets. It may also not apply to other ethnicities, such as African Americans, Hispanics, and Asians. There are two additional elements that can be examined to strengthen the validation process; these include studying the predictive value of a model across time and across geographic boundaries. There was a 5-year difference between the time of biopsy in the Oxford derivation cohort (1997) versus the North American validation cohort (2002), thus fulfilling this criterion. It is likely that this resulted in the previously discussed differences in terms of blood pressure medications, RASB usage, and immunosuppression; however, despite these changes, the predictive capacity of the Oxford classification remained largely intact. We also fulfill the different origins of the data elements as our NA cohort is geographically different from the derivation data set being restricted to a North American base versus a more varied origin in the Oxford derivation cohort. In general, any classification needs to be reproducible. This capacity for replication of the interpretation of the pathology findings to span a wide range of nephropathologists, from academic specialists to those in the community, has been a notable limitation in many previous studies that have attempted to quantify the added value of pathology in renal diseases.22.Weening J.J. D'Agati V.D. Schwartz M.M. et al.The classification of glomerulonephritis in systemic lupus erythematosus revisited.J Am Soc Nephrol. 2004; 15: 241-250Crossref PubMed Scopus (1359) Google Scholar, 23.Solez K. Colvin R.B. Racusen L.C. et al.Banff 07 classification of renal allograft pathology: updates and future directions.Am J Transplant. 2008; 8: 753-760Crossref PubMed Scopus (1553) Google Scholar, 24.Furness P.N. Taub N. Interobserver reproducibility and application of the ISN/RPS classification of lupus nephritis-a UK-wide study.Am J Surg Pathol. 2006; 30: 1030-1035Crossref PubMed Scopus (78) Google Scholar, 25.Furness P.N. Taub N. International variation in the interpretation of renal transplant biopsies: report of the CERTPAP Project.Kidney Int. 2001; 60: 1998-2012Abstract Full Text Full Text PDF PubMed Scopus (194) Google Scholar In all, 10 of the 24 initial variables from the original study were judged to be insufficiently reproducible for further analysis; after assessing co-linearity, only 6 were sufficiently independent from one another to warrant predictive testing.11.Roberts I.S. Cook H.T. Troyanov S. et al.The Oxford classification of IgA nephropathy: pathology definitions, correlations, and reproducibility.Kidney Int. 2009; 76: 546-556Abstract Full Text Full Text PDF PubMed Scopus (682) Google Scholar In our NA cohort, the ICCs were very good and higher than those in the Oxford derivation data set. This finding supports its reproducibility. The agreement based on the cutoffs proposed by the Oxford classification was also high. Part of this is because of the simplicity of the chosen variables. Two, namely the presence of endocapillary proliferation or segmental glomerulosclerosis, are based on a simple 'present' or 'absent' assessment. Thus, a single glomerulus with either of these lesions correlated with a worse prognosis during follow-up. Tubulointerstitial fibrosis, assessed with a semi-quantitative score with the Oxford classification, was also robustly scored in this study. Different morphometric approaches and various staining parameters have been advocated,26.Hunter M.G. Hurwitz S. Bellamy C.O. et al.Quantitative morphometry of lupus nephritis: the significance of collagen, tubular space, and inflammatory infiltrate.Kidney Int. 2005; 67: 94-102Abstract Full Text Full Text PDF PubMed Scopus (31) Google Scholar but the semi-quantitative 'eyeballing' assessment on a light microscopy slide by an experienced pathologist is a straightforward, reliable, and robust morphological parameter.27.Mengel M. Chapman J.R. Cosio F.G. et al.Protocol biopsies in renal transplantation: insights into patient management and pathogenesis.Am J Transplant. 2007; 7: 512-517Crossref PubMed Scopus (83) Google Scholar The ICC for mesangial proliferation was lower than that in the derivation data set. This could have occurred because the scoring process is more time-consuming and complex compared with the other features as it involves counting mesangial cell nuclei in each glomerulus.11.Roberts I.S. Cook H.T. Troyanov S. et al.The Oxford classification of IgA nephropathy: pathology definitions, correlations, and reproducibility.Kidney Int. 2009; 76: 546-556Abstract Full Text Full Text PDF PubMed Scopus (682) Google Scholar In large part, this study validates the Oxford classification in IgAN and provides strong support for a more widespread adoption of this model in routine pathology practice. In addition, the validation process indicates that utilization of this new classification will improve the ability to identify early those patients who are most likely to progress to renal failure and may respond to immunosuppressive treatment. The classification should significantly modify not only our approach to the individual patient but also our framework for designing future clinical trials in IgAN. This is a retrospective study of 187 patients with biopsy-proven IgAN followed in 1 of 4 North American centers: the University of Toronto Glomerulonephritis Registry, the University of Alabama at Birmingham, the Vanderbilt University Medical Center in Tennessee, and the Mayo Clinic in Minnesota. The inclusion criteria as required in validation studies were identical to those of the initial study. Patients of any age with native-kidney biopsy-proven IgAN were included, provided their initial eGFR was ≥30 ml/min per 1.73 m2, proteinuria was ≥0.5 g/24 h (≥0.5 g/24 h per 1.73 m2 in children), follow-up was at least 12 months, and an adequate renal biopsy was available. Patients with isolated hematuria, Henoch-Schoenlein purpura, or co-existing conditions such as diabetes mellitus were excluded. Demographics collected included gender, ethnicity, and age at the time of biopsy. Children were defined as patients aged <18 years at biopsy. Clinical parameters collected initially (within 3 months of the date of biopsy) and during follow-up included systolic and diastolic blood pressures, weight, height, serum creatinine, and 24-h urine protein or urine protein-to-creatinine ratio. To provide consistency between measurements in adults and children, proteinuria was expressed in g/24 h per 1.73 m2 in children and g/24 h in adults. Drug exposure was also collected, including immunosuppressive agents, fish oil, and the number and type of antihypertensive medications, including angiotensin-converting enzyme inhibitor and angiotensin receptor blocker. Data verification occurred by communication between the two of the writing committee authors (ST and HNR) and the contributing centers. Biopsy adequacy was defined as a minimum of eight glomeruli available for examination by light microscopy. IgAN was confirmed as predominant or codominant immunoglobulin in the mesangial deposits. Two experienced renal pathologists, both members of the original Oxford study, independently scored every feature according to the full Oxford score sheet.11.Roberts I.S. Cook H.T. Troyanov S. et al.The Oxford classification of IgA nephropathy: pathology definitions, correlations, and reproducibility.Kidney Int. 2009; 76: 546-556Abstract Full Text Full Text PDF PubMed Scopus (682) Google Scholar The first pathologist scoring a given case chose the most representative section within biopsy slides. This section was marked and independently scored by the second pathologist. We derived an MEST score based on these results. Discordant scores were observed in a minority of cases and they were resolved by a head-to-head meeting between the pathologists (ABF and AMH). The eGFR was calculated using the four-variable Modification of Diet in Renal Disease Study formula in adults and the Schwartz formula in children (using the constant 0.55). End-stage renal disease was defined as eGFR <15 ml/min per 1.73 m2. A combined event was defined by a 50% reduction in eGFR or end-stage renal disease. MAP was defined as diastolic pressure plus a third of the pulse pressure. For each patient, average MAP and proteinuria were determined for each year of observation. Time-average MAP and proteinuria represent the average of these annual values. Immunosuppressive treatment is reported as intent to treat, regardless of the type or duration of therapy. RASB indicates any exposure to angiotensin-converting enzyme inhibitor, angiotensin receptor blocker, or both. Normally distributed variables were expressed as mean±s.d. and compared using Student's t-test or one-way ANOVA (analysis of variance). Non-parametric variables were expressed as median with interquartile range and compared using either Mann–Whitney or Kruskal–Wallis test. Categorical variables were expressed in percentages and compared using the Pearson χ2 test. Reproducibility was assessed for each variable of the extended pathology data set using the ICC, a measure of reproducibility applicable to multiple raters. By convention, an ICC of 0.40–0.59 is moderate inter-rater reliability; 0.60–0.79, substantial; and ≥0.80, outstanding.28.Armstrong G.D. The intraclass correlation as a measure of interrater reliability of subjective judgments.Nurs Res. 1981; 30 (320A): 314-315PubMed Google Scholar The rate of renal-function decline (slope) was used as the main outcome. It was determined by fitting a straight line through available data for eGFR using the principle of least squares. This line was plotted and visually examined in each patient. Obvious outliers were censored. The original study used three multivariate methodologies to assess the independent value of pathology findings: linear regression (slope), logistic regression (slope categorized by slow or rapid rate of renal-function decline based on the median value), and Cox regression (survival from either end-stage renal disease or a 50% decrease in eGFR). However, the slope distribution in this cohort was flatter and more skewed than the normal 'bell-shaped' curve; hence, the assumption of normality could not be met and therefore linear regression analysis was not applied. We had also assessed kidney survival from a 50% decrease in eGFR or ERSD but could not adjust for multiple variables in the NA cohort because the number of these events was too low in comparison with the Oxford cohort. Therefore, we could only use the logistic model to test whether the added value of pathology features remained independent of the clinical assessment. Validation of the Oxford clinicopathological classification using this new cohort was tested in a series of predefined successive steps: (1)Using the logistic regression equation from the Oxford data set, we calculated for each patient from both cohorts the predicted probability of a rapid rate of renal-disease progression.(2)We then categorized these predicted probabilities into five groups with increasing risk (quintiles) for each data set.(3)We compared the observed rate of renal-function decline among quintiles. A consistent relationship between predicted risk and observed rate of renal-function decline in both cohorts validates the Oxford logistic model.(4)To further address differences between the two data sets and to determine whether any of the specific pathological findings differed in the predictive value compared with the original, we performed a new logistic regression on the validation data set using the same variables as in the initial Oxford model and compared graphically the adjusted odds ratio of each pathological variable with those reported in the original classification paper.(5)Finally, we calculated the area under the receiver-operating characteristic curves (c-statistic) from the Oxford and North American logistic regression models. Survival from a 50% reduction in renal function or end-stage renal disease was performed using the log-rank test. All P-values were two-tailed, and values <0.05 were considered statistically significant. CIs included 95% of predicted values. Analyses were carried out using SPSS software (version 16, SPSS, Chicago, IL). HNR is the recipient of a KRESCENT clinician scientist fellowship from the Kidney Foundation of Canada and the Canadian Institute of Health Research. ST's research efforts are supported by the Fonds de la Recherche en Santé du Québec. This study was supported in part by a grant from the Canadian Institutes of Health Research, Net Grant on Genes, Gender and Glomerulonephritis (no. 452773L), the Fonds de la Recherche en Santé du Québec (no. 14395), and the NIH grants DK082753, DK078244, and DK080301. We thank the following coordinators who were of substantial help in collecting the data: from Toronto, N. Ryan and P. Ling; and from Birmingham, C.V. Barker and C. Snyder.
Referência(s)