- Email: [email protected]
Validation of the Oxford classification of IgA nephropathy: valid or invalid?
letter to the editor
subjects, GLP-1 has no effect on GFR, suggesting a mechanism that may be dependent on the baseline filtration rate.4 To conclude, incretin-based therapies may reduce renal risk by targeting glomerular hypertension/hyperfiltration in patients with diabetes and increased baseline GFR. Longterm trials are, however, needed to further explore whether these and other potential ‘off-target’ effects of incretin-based therapies result in meaningful renoprotection.
progression. Indeed, incretin-based drugs ameliorated AKI via anti-apoptotic, anti-inflammatory, and anti-oxidative effects. Therefore, we consider that incretin-based therapy might have the potential to ameliorate DKD via additional mechanisms beyond glomerular hemodynamics. Further studies are needed to understand the underlying mechanisms of incretin-based therapy against DKD in humans. 1.
1. 2.
3.
4.
Tanaka T, Higashijima Y, Wada T et al. The potential for renoprotection with incretin-based drugs. Kidney Int 2014; 86: 701–711. Muskiet MH, Smits MM, Morsink LM et al. The gut-renal axis: do incretinbased agents confer renoprotection in diabetes? Nat Rev Nephrol 2014; 10: 88–103. Gutzwiller JP, Tschopp S, Bock A et al. Glucagon-like peptide 1 induces natriuresis in healthy subjects and in insulin-resistant obese men. J Clin Endocrinol Metab 2004; 89: 3055–3061. Skov J, Dejgaard A, Fr kiær J et al. Glucagon-like peptide-1 (GLP-1): effect on kidney hemodynamics and renin-angiotensin-aldosterone system in healthy men. J Clin Endocrinol Metab 2013; 98: E664–E671.
Lennart Tonneijck1, Mark M. Smits1, Danie¨l H van Raalte1 and Marcel H.A. Muskiet1 1 Diabetes Center, Department of Internal Medicine, VU University Medical Center, Amsterdam, The Netherlands Correspondence: Lennart Tonneijck, Diabetes Center, Department of Internal Medicine, VU University Medical Center, De Boelelaan 1117, 1081 HV Amsterdam, The Netherlands. E-mail: [email protected]
Kidney International (2015) 87, 660–661; doi:10.1038/ki.2014.398
The Authors Reply: We thank Tonneijck et al.1 for their interest in our recent paper2 and agree to their view that renoprotection by incretin-based therapy may be associated with a reduction in glomerular hypertension/hyperfiltration. Glomerular hyperfiltration is an early manifestation in diabetic kidney disease (DKD). Given that the clinical improvement in albuminuria appears, in selected patients, several weeks after the first administration of incretin-based drugs, it is expected that the plastic, not structural, changes, such as hemodynamic factors, might be involved. Indeed, a double-blind, placebo-controlled, crossover study demonstrated that glucagon-like peptide-1 (GLP-1) reduced glomerular hyperfiltration in obese subjects.3 Alterations in glomerular hemodynamics may be mediated by multiple factors. Incretin-based therapy facilitates natriuresis in the salt-sensitive population, which increases the amount of luminal fluid that reaches macula densa, thus leading to afferent arteriolar constriction. Incretins may also mediate vasodilation and improve endothelial dysfunction via its direct action on vascular endothelium. Importantly, a recent in situ hybridization study demonstrated predominant expression of GLP-1 receptor in vascular endothelium in the mouse kidney.4 Incretin-based therapy also exhibits reno-protective effects against rodent acute kidney injury (AKI) models, such as cisplatin-induced nephrotoxicity and ischemia–reperfusion injury.5–6 In these models, glomerular hypertension/ hyperfiltration is not a major determinant of disease Kidney International (2015) 87, 660–664
2. 3.
4.
5.
6.
Tonneijck L, Smits M, van Raalte D et al. Incretion-based drugs and renoprotection—is hyperfiltration key. Kidney Int 2015; 87: 660–661. Tanaka T, Higashijima Y, Wada T et al. The potential for renoprotection with incretin-based drugs. Kidney Int 2014; 86: 701–711. Gutzwiller JP, Tschopp S, Bock A et al. Glucagon-like peptide 1 induces natriuresis in healthy subjects and in insulin-resistant obese men. J Clin Endocrinol Metab 2004; 89: 3055–3061. Fujita H, Morii T, Fujishima H et al. The protective roles of GLP-1R signaling in diabetic nephropathy: possible mechanism and therapeutic potential. Kidney Int 2014; 85: 579–589. Glorie LL, Verhulst A, Matheeussen V et al. DPP4 inhibition improves functional outcome after renal ischemia-reperfusion injury. Am J Physiol Renal Physiol 2012; 303: F681–F688. Katagiri D, Hamasaki Y, Doi K et al. Protection of glucagon-like peptide-1 in cisplatin-induced renal injury elucidates gut-kidney connection. J Am Soc Nephrol 2013; 24: 2034–2043.
Yoshiki Higashijima1, Tetsuhiro Tanaka1 and Masaomi Nangaku1 1 Division of Nephrology and Endocrinology, The University of Tokyo Graduate School of Medicine, Tokyo, Japan Correspondence: Masaomi Nangaku, Division of Nephrology and Endocrinology, The University of Tokyo Graduate School of Medicine, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113 8655, Japan. E-mail: [email protected]
Kidney International (2015) 87, 661; doi:10.1038/ki.2014.410
Validation of the Oxford classification of IgA nephropathy: valid or invalid? To the Editor: Coppo et al.1 assess the validity of the Oxford classification of IgA nephropathy by evaluating what they call the predictive value of renal pathology variables on the rate of renal function decline and renal survival. What they appear to be doing, however, is reporting the ability of the said variables to ‘explain’ the observed outcomes in the samples generating the model. When the authors state that they validated the multivariate models from the original Oxford classification study, does that mean that the identified models were used on patients not generating the models? The variability in a predictive model is likely to be greater than in an explanatory model and the form different,2 was that the case? Trying to understand what has then in fact been ‘explained’ is not easy. In Table 4 is presented the results of ‘prediction’ of rapid progression versus observed rapid progression. For those not on immunosuppression therapy and using both clinical and pathological variables, a positive predictive value (PPV) of 0.77 and a negative predictive value (NPV) of 0.65 can be calculated compared with 0.63 for both PPV and NPV for those on immunosuppression. This is 661
letter to the editor
not impressive, especially as it uses the same data twice: in generating and assessing the model. The authors conclude that immunosuppression abolishes the ‘predictive’ effect of addition of pathological variables to clinical variables. This is interesting to know but less so than whether the treatment in fact helped the patients. Was that possible to tell? The rate of renal function decline was obtained by fitting straight line, which could lead to problems if in fact treatment was effective. ‘Obvious outliers were censored’—were they only few and evenly distributed? When the authors state that an early renal biopsy may be an advantage as these patients with well-preserved glomerular filtration rate on reninangiotensin system blockade and with little proteinuria have a destiny, which is not determined by advanced renal damage as it is absent, the overall perspective of the study seems questionable. How could in the real world an early biopsy by itself result in a better prognosis? The predictive value of the mesangial hypercellularity (M), endocapillary proliferation (E), segmental glomerulosclerosis (S), and tubular atrophy/interstitial fibrosis (T) (MEST) criteria relies on accurate histological classification, and pathology is by far no exact science. Thus, as noted by one of the authors in a recent paper concerning the Oxford Classification of IgAN in a European cohort study: ‘yagreement between local and review pathologists in the scoring of lesions was good for tubular atrophy/interstitial fibrosis and segmental sclerosis, moderate for mesangial hypercellularity and poor for endocapillary hypercellularity.’3 In other words, in the absence of a pathology review board, assessment of histological data is suboptimal; yet, this is what guides the routine diagnostic practice. If the predictive value (if even it were investigated) of all our measurements and biopsies is as weak as it appears from this great endeavor, then we certainly appear to look in the wrong places, both for explanation and for prediction.4 1.
2. 3. 4.
Coppo R, Troyanov S, Bellur S et al. Validation of the Oxford classification of IgA nephropathy in cohorts with different presentations and treatments. Kidney Int 2014; 86: 828–836. Shmueli G. To explain or to predict? Stat Sci 2010; 25: 289–310. Roberts IS. Pathology of IgA nephropathy. Nat Rev Nephrol 2014; 10: 445–454. Kiryluk K, Novak J. The genetics and immunobiology of IgA nephropathy. J Clin Invest 2014; 124: 2325–2332.
Christian D. Peters1 and Troels Ring1 1 Department of Nephrology, Aalborg University Hospital, Aalborg, Denmark Correspondence: Christian D. Peters, Department of Nephrology, Aalborg University Hospital, M lleparkvej 4, 9000 Aalborg, Denmark. E-mail: [email protected]
Kidney International (2015) 87, 661–662; doi:10.1038/ki.2014.401
proposal and address the uncertainties surrounding the MEST score: mainly, its applicability to individuals with lower-grade proteinuria and/or advanced chronic kidney disease cohorts excluded in the original classification. In this regard, we agree that the study is not a true validation as it applies to different presentations, as stated in the title of our article. We also recognize how imprecise VALIGA (and Oxford) is. In our opinion, few models in glomerular disease are explanatory: they all have low R2 values––i.e., they can rarely pinpoint with confidence long-term events. The causes are multiple, but mainly that measurement errors in long retrospective surveys are always additive.3 The more important question is whether biases are spread equally. We wonder how random imprecisions, or the exclusion of rare (o1%) outliers in glomerular filtration rate within a patient, can lead to false-positive results. They are more likely to favor rejecting the null hypothesis. The limitations of using the rate of renal function decline are an excellent example: simplifying episodes of active disease with periods of stability into a straight line may well contribute to disappointing predictive values and low net reclassification index. However, if the authors apply this reasoning to survival analysis, they would realize that the elegant Kaplan–Meier curves disguise even greater uncertainty. What would a patient prefer to know, that he has a 50% risk of ESRD at 10 years or an average 4 ml/min/1.73 m2 yearly drop in renal function? We think both are important. The imperfections of the slope are no excuse to discard it. In addition, we did not create and validate the same model using the same patients (we used the Oxford model). We do agree that the differences in biopsy interpretation between central and local pathologists are immensely important and will be the subject of a future manuscript; however, the data presented in the paper were centrally reviewed and confirmed, thus giving the biopsy findings a firm baseline. Despite decades of research in IgA nephropathy, VALIGA does confirm our ongoing inability to confidently predict outcome; however, remember that recent guidelines4 take no stance on the therapeutic influence of pathology, the definition and management of corticosteroid resistance, retreatment, and the value of maintenance regimens because of the absence of published data. How could prospective trials address these essential questions be intelligently conceived without observational studies such as ours to guide them? The authors are correct in identifying many issues in the paper but not in their conclusion regarding the lack of relevance of the information provided. We encourage them to stay tuned as the interpretation of VALIGA is ongoing. 1.
The Authors Reply: We thank the authors Peters and Ring1 for their thoughtful letter. Since the initial article of the Oxford classification, many validation studies have been published.2 It thus became clear to the VALIGA working group that it would have to deviate from their original grant 662
2.
3. 4.
Peters C, Ring T. Validation of the Oxford classification of IgA nephropathy: valid or invalid? Kidney Int 2015; 87: 661–662. Lv J, Shi S, Xu D, Zhang H, Troyanov S, Cattran DC, Wang H. Evaluation of the Oxford Classification of IgA nephropathy: a systematic review and meta-analysis. Am J Kidney Dis 2013; 62: 891–899. Katz MH. Multivariable analysis: a primer for readers of medical research. Ann Intern Med 2003; 138: 644–650. KDIGO Clinical Practice Guideline for Glomerulonephritis. Chapter 10: Immunoglobulin IgA nephropathy. Kidney Int Suppl 2012; 2: 209–217. Kidney International (2015) 87, 660–664
subjects, GLP-1 has no effect on GFR, suggesting a mechanism that may be dependent on the baseline filtration rate.4 To conclude, incretin-based therapies may reduce renal risk by targeting glomerular hypertension/hyperfiltration in patients with diabetes and increased baseline GFR. Longterm trials are, however, needed to further explore whether these and other potential ‘off-target’ effects of incretin-based therapies result in meaningful renoprotection.
progression. Indeed, incretin-based drugs ameliorated AKI via anti-apoptotic, anti-inflammatory, and anti-oxidative effects. Therefore, we consider that incretin-based therapy might have the potential to ameliorate DKD via additional mechanisms beyond glomerular hemodynamics. Further studies are needed to understand the underlying mechanisms of incretin-based therapy against DKD in humans. 1.
1. 2.
3.
4.
Tanaka T, Higashijima Y, Wada T et al. The potential for renoprotection with incretin-based drugs. Kidney Int 2014; 86: 701–711. Muskiet MH, Smits MM, Morsink LM et al. The gut-renal axis: do incretinbased agents confer renoprotection in diabetes? Nat Rev Nephrol 2014; 10: 88–103. Gutzwiller JP, Tschopp S, Bock A et al. Glucagon-like peptide 1 induces natriuresis in healthy subjects and in insulin-resistant obese men. J Clin Endocrinol Metab 2004; 89: 3055–3061. Skov J, Dejgaard A, Fr kiær J et al. Glucagon-like peptide-1 (GLP-1): effect on kidney hemodynamics and renin-angiotensin-aldosterone system in healthy men. J Clin Endocrinol Metab 2013; 98: E664–E671.
Lennart Tonneijck1, Mark M. Smits1, Danie¨l H van Raalte1 and Marcel H.A. Muskiet1 1 Diabetes Center, Department of Internal Medicine, VU University Medical Center, Amsterdam, The Netherlands Correspondence: Lennart Tonneijck, Diabetes Center, Department of Internal Medicine, VU University Medical Center, De Boelelaan 1117, 1081 HV Amsterdam, The Netherlands. E-mail: [email protected]
Kidney International (2015) 87, 660–661; doi:10.1038/ki.2014.398
The Authors Reply: We thank Tonneijck et al.1 for their interest in our recent paper2 and agree to their view that renoprotection by incretin-based therapy may be associated with a reduction in glomerular hypertension/hyperfiltration. Glomerular hyperfiltration is an early manifestation in diabetic kidney disease (DKD). Given that the clinical improvement in albuminuria appears, in selected patients, several weeks after the first administration of incretin-based drugs, it is expected that the plastic, not structural, changes, such as hemodynamic factors, might be involved. Indeed, a double-blind, placebo-controlled, crossover study demonstrated that glucagon-like peptide-1 (GLP-1) reduced glomerular hyperfiltration in obese subjects.3 Alterations in glomerular hemodynamics may be mediated by multiple factors. Incretin-based therapy facilitates natriuresis in the salt-sensitive population, which increases the amount of luminal fluid that reaches macula densa, thus leading to afferent arteriolar constriction. Incretins may also mediate vasodilation and improve endothelial dysfunction via its direct action on vascular endothelium. Importantly, a recent in situ hybridization study demonstrated predominant expression of GLP-1 receptor in vascular endothelium in the mouse kidney.4 Incretin-based therapy also exhibits reno-protective effects against rodent acute kidney injury (AKI) models, such as cisplatin-induced nephrotoxicity and ischemia–reperfusion injury.5–6 In these models, glomerular hypertension/ hyperfiltration is not a major determinant of disease Kidney International (2015) 87, 660–664
2. 3.
4.
5.
6.
Tonneijck L, Smits M, van Raalte D et al. Incretion-based drugs and renoprotection—is hyperfiltration key. Kidney Int 2015; 87: 660–661. Tanaka T, Higashijima Y, Wada T et al. The potential for renoprotection with incretin-based drugs. Kidney Int 2014; 86: 701–711. Gutzwiller JP, Tschopp S, Bock A et al. Glucagon-like peptide 1 induces natriuresis in healthy subjects and in insulin-resistant obese men. J Clin Endocrinol Metab 2004; 89: 3055–3061. Fujita H, Morii T, Fujishima H et al. The protective roles of GLP-1R signaling in diabetic nephropathy: possible mechanism and therapeutic potential. Kidney Int 2014; 85: 579–589. Glorie LL, Verhulst A, Matheeussen V et al. DPP4 inhibition improves functional outcome after renal ischemia-reperfusion injury. Am J Physiol Renal Physiol 2012; 303: F681–F688. Katagiri D, Hamasaki Y, Doi K et al. Protection of glucagon-like peptide-1 in cisplatin-induced renal injury elucidates gut-kidney connection. J Am Soc Nephrol 2013; 24: 2034–2043.
Yoshiki Higashijima1, Tetsuhiro Tanaka1 and Masaomi Nangaku1 1 Division of Nephrology and Endocrinology, The University of Tokyo Graduate School of Medicine, Tokyo, Japan Correspondence: Masaomi Nangaku, Division of Nephrology and Endocrinology, The University of Tokyo Graduate School of Medicine, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113 8655, Japan. E-mail: [email protected]
Kidney International (2015) 87, 661; doi:10.1038/ki.2014.410
Validation of the Oxford classification of IgA nephropathy: valid or invalid? To the Editor: Coppo et al.1 assess the validity of the Oxford classification of IgA nephropathy by evaluating what they call the predictive value of renal pathology variables on the rate of renal function decline and renal survival. What they appear to be doing, however, is reporting the ability of the said variables to ‘explain’ the observed outcomes in the samples generating the model. When the authors state that they validated the multivariate models from the original Oxford classification study, does that mean that the identified models were used on patients not generating the models? The variability in a predictive model is likely to be greater than in an explanatory model and the form different,2 was that the case? Trying to understand what has then in fact been ‘explained’ is not easy. In Table 4 is presented the results of ‘prediction’ of rapid progression versus observed rapid progression. For those not on immunosuppression therapy and using both clinical and pathological variables, a positive predictive value (PPV) of 0.77 and a negative predictive value (NPV) of 0.65 can be calculated compared with 0.63 for both PPV and NPV for those on immunosuppression. This is 661
letter to the editor
not impressive, especially as it uses the same data twice: in generating and assessing the model. The authors conclude that immunosuppression abolishes the ‘predictive’ effect of addition of pathological variables to clinical variables. This is interesting to know but less so than whether the treatment in fact helped the patients. Was that possible to tell? The rate of renal function decline was obtained by fitting straight line, which could lead to problems if in fact treatment was effective. ‘Obvious outliers were censored’—were they only few and evenly distributed? When the authors state that an early renal biopsy may be an advantage as these patients with well-preserved glomerular filtration rate on reninangiotensin system blockade and with little proteinuria have a destiny, which is not determined by advanced renal damage as it is absent, the overall perspective of the study seems questionable. How could in the real world an early biopsy by itself result in a better prognosis? The predictive value of the mesangial hypercellularity (M), endocapillary proliferation (E), segmental glomerulosclerosis (S), and tubular atrophy/interstitial fibrosis (T) (MEST) criteria relies on accurate histological classification, and pathology is by far no exact science. Thus, as noted by one of the authors in a recent paper concerning the Oxford Classification of IgAN in a European cohort study: ‘yagreement between local and review pathologists in the scoring of lesions was good for tubular atrophy/interstitial fibrosis and segmental sclerosis, moderate for mesangial hypercellularity and poor for endocapillary hypercellularity.’3 In other words, in the absence of a pathology review board, assessment of histological data is suboptimal; yet, this is what guides the routine diagnostic practice. If the predictive value (if even it were investigated) of all our measurements and biopsies is as weak as it appears from this great endeavor, then we certainly appear to look in the wrong places, both for explanation and for prediction.4 1.
2. 3. 4.
Coppo R, Troyanov S, Bellur S et al. Validation of the Oxford classification of IgA nephropathy in cohorts with different presentations and treatments. Kidney Int 2014; 86: 828–836. Shmueli G. To explain or to predict? Stat Sci 2010; 25: 289–310. Roberts IS. Pathology of IgA nephropathy. Nat Rev Nephrol 2014; 10: 445–454. Kiryluk K, Novak J. The genetics and immunobiology of IgA nephropathy. J Clin Invest 2014; 124: 2325–2332.
Christian D. Peters1 and Troels Ring1 1 Department of Nephrology, Aalborg University Hospital, Aalborg, Denmark Correspondence: Christian D. Peters, Department of Nephrology, Aalborg University Hospital, M lleparkvej 4, 9000 Aalborg, Denmark. E-mail: [email protected]
Kidney International (2015) 87, 661–662; doi:10.1038/ki.2014.401
proposal and address the uncertainties surrounding the MEST score: mainly, its applicability to individuals with lower-grade proteinuria and/or advanced chronic kidney disease cohorts excluded in the original classification. In this regard, we agree that the study is not a true validation as it applies to different presentations, as stated in the title of our article. We also recognize how imprecise VALIGA (and Oxford) is. In our opinion, few models in glomerular disease are explanatory: they all have low R2 values––i.e., they can rarely pinpoint with confidence long-term events. The causes are multiple, but mainly that measurement errors in long retrospective surveys are always additive.3 The more important question is whether biases are spread equally. We wonder how random imprecisions, or the exclusion of rare (o1%) outliers in glomerular filtration rate within a patient, can lead to false-positive results. They are more likely to favor rejecting the null hypothesis. The limitations of using the rate of renal function decline are an excellent example: simplifying episodes of active disease with periods of stability into a straight line may well contribute to disappointing predictive values and low net reclassification index. However, if the authors apply this reasoning to survival analysis, they would realize that the elegant Kaplan–Meier curves disguise even greater uncertainty. What would a patient prefer to know, that he has a 50% risk of ESRD at 10 years or an average 4 ml/min/1.73 m2 yearly drop in renal function? We think both are important. The imperfections of the slope are no excuse to discard it. In addition, we did not create and validate the same model using the same patients (we used the Oxford model). We do agree that the differences in biopsy interpretation between central and local pathologists are immensely important and will be the subject of a future manuscript; however, the data presented in the paper were centrally reviewed and confirmed, thus giving the biopsy findings a firm baseline. Despite decades of research in IgA nephropathy, VALIGA does confirm our ongoing inability to confidently predict outcome; however, remember that recent guidelines4 take no stance on the therapeutic influence of pathology, the definition and management of corticosteroid resistance, retreatment, and the value of maintenance regimens because of the absence of published data. How could prospective trials address these essential questions be intelligently conceived without observational studies such as ours to guide them? The authors are correct in identifying many issues in the paper but not in their conclusion regarding the lack of relevance of the information provided. We encourage them to stay tuned as the interpretation of VALIGA is ongoing. 1.
The Authors Reply: We thank the authors Peters and Ring1 for their thoughtful letter. Since the initial article of the Oxford classification, many validation studies have been published.2 It thus became clear to the VALIGA working group that it would have to deviate from their original grant 662
2.
3. 4.
Peters C, Ring T. Validation of the Oxford classification of IgA nephropathy: valid or invalid? Kidney Int 2015; 87: 661–662. Lv J, Shi S, Xu D, Zhang H, Troyanov S, Cattran DC, Wang H. Evaluation of the Oxford Classification of IgA nephropathy: a systematic review and meta-analysis. Am J Kidney Dis 2013; 62: 891–899. Katz MH. Multivariable analysis: a primer for readers of medical research. Ann Intern Med 2003; 138: 644–650. KDIGO Clinical Practice Guideline for Glomerulonephritis. Chapter 10: Immunoglobulin IgA nephropathy. Kidney Int Suppl 2012; 2: 209–217. Kidney International (2015) 87, 660–664