International IgA Nephropathy Prediction tool: A step toward personalized medicine

By Dr Pallavi Prasad
Professor, Nephrology VMMC and Safdarjung Hospital
New Delhi, India

GlomCon Editors with significant contributions to the development of this article include: Sayali Thakare, Haresh Selvaskandan, Paolo Nikolai So, and Nasim Wiegley.

Disclaimer: No conflict of interest


The International IgAN prediction tool was developed in 2018 to help predict the individual risk of disease progression in patients with primary IgAN. The tool uses clinical and histopathological markers at the time of biopsy to predict the risk of 50% decline in eGFR or kidney failure in patients with IgAN. It has been modified for use in the pediatric population and risk prediction at 1 and 2 years post kidney biopsy. Validation of the tool for various ethnicities has been done, but further studies are needed before it can be considered truly generalizable. Future modifications may also include biomarkers, which have been shown to improve the discriminatory power of the tool in small scale studies.


Immunoglobulin A nephropathy (IgAN) is a common primary glomerular disease with a heterogeneous clinical presentation. The rate at which IgAN progresses to kidney failure is highly variable, with the risk of kidney failure at ten years varying between 5%60%. The lifetime risk of kidney failure, however, appears to be much higher. A recent UK registry study including over 2000 adults and 100 children with biopsy-proven IgAN who had proteinuria > 0.5 grams/day found that over 80% developed kidney failure within 30 years unless the rate of estimated glomerular filtration rate (eGFR) decline was maintained at ≤ 1ml/min/1.73m². Crucially, progression to kidney failure within 10 years occurred even with low-grade proteinuria (< 0.88g/g).

A number of widely validated clinicopathological factors are known to influence the rate at which IgAN progresses to kidney failure. These include race, blood pressure, proteinuria, eGFR at the time of diagnosis, and histological findings (defined by the MEST-C score as detailed in Table 1). On their own, however, these estimate the risk of kidney failure at a population-wide level and provide little insight into the risk of progression of an individual patient. Individual risk prediction not only supports patient counseling and risk stratification but, in the future, may also help formulate individualized management plans. The International IgA Nephropathy prediction tool was developed to directly address some of these issues.

Table 1. The Oxford MEST-C SCORE

Pathology Description Score
The mesangial hypercellularity score is the mean
score for all glomeruli.

Score for each glomerulus is calculated as below:
<4 Mesangial cells/mesangial area = 0
4–5 Mesangial cells/mesangial area = 1
6–7 Mesangial cells/mesangial area = 2
>8 Mesangial cells/mesangial area = 3
M0: ≤0.5
M1: >0.5
Segmental glomerulosclerosisAny amount of the tuft involved in sclerosis, but not involving the whole tuft or the presence of an adhesion; presence or absence of podocyte hypertrophy/tip lesions in biopsy specimens with S1S0: Absent
S1: Present
Endocapillary hypercellularity Hypercellularity due to increased number of cells within glomerular capillary lumina causing narrowing of the luminaE0: Absent
E1: Present
Tubular atrophy/
interstitial fibrosis
Percentage of cortical area involved by the tubular atrophy or interstitial fibrosis, whichever is greaterT0: 0–25%
T1: 26–50%
T2: >50%
Crescents Percentage of total glomeruli with crescentsC0: no crescents
C1: <25% crescents
C2: ≥25% crescents

Adapted from: Cattran et al and Trimarchi et al.

The Original International IgAN Prediction Tool ( IIgANPT)

The IIgANPT integrates validated clinicopathological prognostic factors with treatments received at diagnosis to produce an individualized risk of disease progression (Figure 1). The model was derived from a multiethnic cohort of patients with biopsy-proven primary IgAN (n=2781 for derivation and n=1146 for external validation). Data were derived from established IgAN-related population studies, including cohorts from Europe, China, Japan, North and South America. Using data from the derivation cohort, three models were developed to predict a primary composite outcome of a 50% decline in eGFR or kidney failure (defined as eGFR <15 ml/min/1.73m2 by CKD EPI, dialysis or transplantation): the clinical, limited and full model. The clinical model incorporated known clinical predictors of outcomes [eGFR, proteinuria and mean arterial pressure (MAP)], and the limited model incorporated the MEST score into the clinical model. The full model, incorporated additional predictors of progression, including age, sex, race, crescents, body mass index (BMI), renin-angiotensin (RAS) blocker use at biopsy, immunosuppression at biopsy, and interaction terms between age and eGFR and between proteinuria and each of MAP, sex, RAS inhibition use at biopsy, and MEST. Race in the full model was specified as Chinese, Japanese, or White (since most of the derivation cohort hailed from these ethnicities). A full model without race was also developed. The two full models were found to have better discrimination, calibration and risk reclassification for predicting the primary outcome as compared to the clinical and limited models.

The tool was externally validated in a cohort of 1146 patients and found to have good discrimination ( C statistic 0.82; 95% CI, 0.81-0.83 with race/ethnicity; 0.81; 95% CI, 0.80-0.82 without race/ethnicity) and model fit R2D (both 35.3%) with excellent calibration as compared to the derivation cohort in predicting the composite outcome. The model is recommended for risk prediction at five years post-biopsy but can be used to predict risk up to seven years; five and seven-year timepoints were the 50th and 75th percentiles of follow-up duration in the derivation cohort.

The IIgANPT is yet to be validated as a tool for guiding treatment decisions, although it has been shown to predict the risk of disease progression in an individual patient more accurately than proteinuria alone. Nevertheless, The Kidney Disease Improving Global Outcomes (KDIGO) Management of Glomerular Diseases 2021 guidelines acknowledge its value in supporting patient counseling and facilitating shared decision-making and encourage the use of the IIgANPT for these purposes as a practice point (Practice Point 2.2.1)

Figure 1. IgAN risk prediction tools and its modifications

Crescents and the IIgANPT

It is now clear that the presence of crescents in IgAN is a predictor of poor outcomes in adults and in children. Crescents were adopted in the Oxford classification of IgA nephropathy in 2016 after the working group found that the percentage of glomeruli containing cellular or fibrocellular crescents was an independent risk factor for predicting a composite endpoint of 50% decline in eGFR or kidney failure (derived from a multi-ethnic cohort, n = 3096). Crescents correlated with the use of immunosuppression after biopsy; however, the risk of the composite outcome remained statistically significant in those with >25% glomerular crescents, regardless of immunosuppression.

The absence of crescents as a variable in the IIgANPT stems from incomplete data being available at the time the model was derived and validated. Crescents were initially incorporated into the model as either being present or absent, in contrast to the three levels assigned in the MEST-C score (C0 = 0% glomeruli with crescents, C1 <25% crescents and C2 ≥25% crescents). These granular data were not available for all cohorts, and when incorporated on the basis of presence alone, crescents strongly correlated with race ( Japanese > Chinese > White), which was an independent risk factor for risk prediction in the model. In the model without race, crescents did not satisfy the criterion for selection in the prediction model. When the variable for the use of immunosuppression was combined with the presence of crescents, it was found to satisfy the criterion for selection in the prediction model, but the use of immunosuppression post-biopsy on its own did not satisfy the primary criterion for inclusion as a predictor variable. For these reasons, crescents were excluded from the current IIgANPT.

Validation of the Tool Across Different Time Points

For decades, the only treatment options available for IgAN were RAS inhibition or, controversially, immunosuppression. These measures were often initiated after diagnosis and are expected to modify the percentage risk of progression to kidney failure. The original tool consistently underestimated the risk of disease progression when used at one or two years post-biopsy, perhaps due to the evolution of histology over time or due to a shift of time lag bias of eGFR by as many years. To account for changes over time, the IIgANPT was modified and externally validated for use one and two years after the biopsy. The updated tool incorporates the same clinicopathological parameters as used in the original tool but is recorded at the landmark time post-biopsy to predict the risk of the primary outcome ( e.g., proteinuria at one year for risk prediction at one year post-biopsy).

Notably, Japanese race and E1 on tissue histology had a higher hazard ratio for the composite endpoint, while eGFR, T1 and T2 lesions on biopsy had a lower hazard ratio compared to the original tool. The use of immunosuppression post-biopsy was found to be more frequent in patients with E1 lesions and Japanese race.

Although the same cohorts as used in the original tool were employed for derivation and validation of the updated tool, sufficient follow-up data was present in only one of the two Japanese cohorts. Further validation studies among Japanese and other ethnicities are needed to consider this model to be truly generalizable. The tool is most accurate when used one year post-biopsy. The median follow-up after the second year landmark time was 3.6 years, and hence risk calibration when the post-biopsy model was used at the second year (landmark time), especially in high-risk patients (risk of primary outcome >30%), may not be as accurate as when used at one year post-biopsy.

Validation of the tool for pediatric IgAN

In 2021, the IIgANPT was updated for use in children (age <18 years) with data collected from a multiethnic cohort recruited from China, Japan, Russia, Europe and North America. eGFR trajectory in children and adolescents with IgAN varies from that in adults – an initial rise in GFR is observed until about 18 years of age, followed by a linear fall (corresponding to the trajectory seen in adults). Due to this unique feature in children, the original prediction tool overestimated the risk in children.

Due to the unique trajectory of eGFR in the pediatric cohort, the composite primary endpoint was achieved in only a small percentage of those included (52/1060= 4.9%), and hence, a secondary outcome of 30% decline in eGFR or kidney failure was used for the prediction model in children. The survival curves also showed that the risk of a 30% decline in eGFR in children was analogous to a 50% decline in eGFR in adults. The pediatric tool is therefore used at the time of kidney biopsy for prediction of risk of 30% decline in eGFR or kidney failure over a maximum of 80 months post-biopsy, with similar factors as used in the adult tool, except for a few modifications made for use in the pediatric population. These changes include a change in the primary composite endpoint(as already mentioned earlier), using a full age spectrum formula for eGFR, body surface area normalized proteinuria quantification, and MAP standardized for age (summarized in Figure 1). Other iterations of the IIgANPT have not been validated in pediatric cohorts to date.

The characteristics of the tools are summarized in Table 2.

Table 2. IIgAN prediction tool- derivation studies and characteristics

Studies on tool development
Name of study with yearPopulationTool developedPrimary outcomeRemarks
Barbour et al 2019Derivation n=2781
Validation n=1146

IIgAN PT50% decline in eGFR or kidney failureStrengths:
- Use of large multiethnic cohorts
- Endorsed by KDIGO for risk prediction in IgAN

- Percentage crescents not incorporated in model due to data unavailability
- Effect of post-biopsy treatment on risk prediction not accounted for
Barbour et al 2020Derivation n=1060

IIgAN PT- pediatric30% decline in eGFR or kidney failureStrengths:
- First tool developed for risk prediction in pediatrics

- No external validation
- 30% decline in eGFR is not considered a hard kidney endpoint
- Percentage crescents not incorporated in model due to data unavailability

- In pediatric IgAN, an initial rise in eGFR followed by subsequent decline was observed
Barbour et al 2022Derivation n= 2507
Validation n= 722

IIgAN PT post biopsy50% decline in eGFR or kidney failureStrengths:
- Incorporates use of immunosuppression and RASi after biopsy in risk prediction

- Data regarding dose & duration of immunosuppression/RASi not available
- Percentage crescents not incorporated in model due to data unavailability
- Risk prediction at 2 year after landmark time not as accurate especially in high risk group

- Change in coefficients of eGFR, T1, T2, E1, and Japanese race as compared to original tool helped in better risk prediction at landmark time post biopsy
Incorporation of biomarkers with tools
Name of studyBiomarker/s evaluatedTool developedRemarks
Pawluczyk I et al 2021miR-150-5p
miR -155-5p
miR -146b-5p
IIgAN PT with miR- Expression of each miR significantly correlated with the predicted 5-year risk
- Improved discrimination with increased C-statistic when expression of each miR was added individually to the IIgANPT model with or without race
- Small sample size
Pawluczyk I et al 2021miR-204 expressionIIgAN PT with miR- Expression of miR-204 was significantly lower in IgAN patients at high risk of progression
- Improved discrimination with increased C-statistic when expression of miR 204 was added to the IIgANPT model without race
- Improved reclassification as assessed by significant event and non-event net reclassification improvement
- Small sample size

Abbreviations: eGFR: estimated glomerular filtration rate, IIgAN PT: International IgAN prediction tool, KDIGO: Kidney Disease Improving Global Outcomes, miR: micro-RNA.

Validation of risk prediction tools across various cohorts & ethnicities

The IIgANPT has been externally validated in different ethnicities (Table 3) with varying outcomes. External validation in a Norwegian, a Chinese and another Chinese-Argentenian cohort have been published, and the models were found to have good discrimination and model fit. However, the risk probability over 3 years was overestimated in the full model (with race) in the Chinese-Argentenian cohort. Among patients from India, the tool had reasonable discrimination but underestimated the risk of progression across all risk groups. A validation study in a Korean cohort was also found to underestimate risk, more so in the model without race. The underestimation of risk in Indian and Korean cohorts further highlights the variability in risk progression in various ethnicities and possibly also the use of different management practices in different areas of the world. The tool may need further modification to be able to predict outcomes in ethnicities not adequately represented in the derivation or validation cohorts. In an increasingly multicultural world, ethnicity is often difficult to define, and categorization of risk progression on the basis of ethnicity alone should probably be replaced by categorization based on genetic makeup. A genetic risk score has recently been proposed for identifying patients prone to rapid progression of kidney disease. Also, environmental influences may also play a role in disease progression and need to be studied in more detail in the future.

There is a single validation study of the pediatric tool in Chinese population (Table 3) , and the post-biopsy tool has not yet been externally validated in other studies.

Table 3. Validation studies of IIgAN prediction tool (adult and pediatric)

Name of studyPopulation (n)Validation results
DiscriminationCalibrationModel FitReclassification
Adult tool validation
Bon G et al 2023French
AUC of 0.833 and 0.817 without and with raceCalibration analysis was good for both models up to 15 years after diagnosis.R²D 29% and 28% without and with race respectivelyNA
Haaskjold YL et al 2023Norwegian
C-statistic: 0.80 (model type not specified)Overall acceptable but underestimates risk at 5 years and overestimates risk at 20 years for patients with the highest observed riskNANA
Shaffi SK et al 2023Hispanics, American Indians
C-statistic of 0.79 (without race)Model overestimated risk in those with estimated risks between 20 and 50% and underestimated risk in those with estimated risks above 50%NANA
Bagchi S et al 2022Indian
C-statistic 0.845 for model with and without raceUnderestimated risk of progression across all risk groups. Both models demonstrated poor calibration for predicting risk at 2.8 and 5 years.R²D 44.7 % and 49.9 % without and with race respectivelyLimited improvement in risk reclassification at 5 and 2.8 years when comparing model with and without race.
Joo YS et al 2022 (validation and development of Korean coefficient)Korean
C-statistic of 0.81 in model with and without race respectivelyModels underestimated the risk of the primary outcome, with lesser underestimation for the model with race.R²D 44.8% and 45.2% without and with race respectivelyModel with race had better performance in reclassification compared to the model without race (NRI 0.13). Updated model with Korean coefficient showed good agreement between predicted risk and observed outcome.
Papasotiriou M et al 2022Greek
C-statistic 0.70 and 0.71 without and with race respectivelyCalibration not acceptable for both the models, with risk probability overestimationR²D of 35% and 39% for the full models without and with race respectivelyNA
Hwang D et al 2021 (validation and model development for Korean population))Korean
C-statistic 0.69 without race, 0.67 with race
Internally derived clinical and full models (for Korean population) had AUC values of 0.78 and 0.84
Not well calibrated, significant overestimation especially in patients with a higher riskNANRI and IDI analyses showed that discrimination and reclassification performance of the international model was inferior to the internally derived models.
Saulīte AJ et al 2021Latvian
C-statistic of 0.7NANANA
Zhang J et al 2020Chinese
C-statistic >0.85 for model with and without raceModel without race underestimated risk
Full model with race well calibrated for predicting 5-year risk
NAFull model with race had significant improvement in reclassification, as assessed by the net reclassification improvement- NRI (0.49; 95% confidence interval, 0.41 to 0.59) and integrated discrimination improvement IDI (0.06; 95% confidence interval, 0.04 to 0.08).
Zhang Y et al 2020Chinese (1169)
Argentinian (106)
C-statistic >0.81 model with and without raceCalibration acceptable for the full model without race. Risk probability over 3 years overestimated in model with raceR-squared (R²D) values 37.6% and 42.2% for model without and with race respectivelyNA
Pediatric tool validation
Yu X et al 2023Chinese
C-statistic 0.64 and 0.685 for model without and with race respectivelyNeither model showed sufficient calibrationR²D values 56.2% and 66.5% for model without and with race respectivelyNA

Abbreviations: AUC: Area under Curve, IDI: Integrated Discrimination Improvement, NA: not available, NRI: Net Reclassification Improvement.

Potential uses of the tool

The IIgANPT can be used for risk stratification and prognostication of patients. It can not be used in its current form for treatment decisions.

Currently, most trial recruitment is based on eGFR, proteinuria and histopathological parameters as discrete and independent criteria to attempt to capture those with potentially reversible disease at greatest risk of progression. A complex interplay of various factors, some as yet unidentified, is involved in disease progression, as was seen in the derivation of the IIgAN PT. Although it is not approved in its current form for decision-making regarding immunosuppression, the tool may perform better than proteinuria for treatment allocation and would allow integration of risk factors to better capture those at the highest risk of progression for clinical trials in the future. However, underestimation or overestimation of risk in populations that were not well represented in the original cohorts may lead to recruitment bias. Incorporation of the tool in the trial recruitment process must be done carefully and with a conscious effort to prevent inequities in allocation or trial recruitment based on ethnicity.

A caveat to the use of this tool for therapeutic decision-making is the high risk of progression conveyed by both markers of fibrosis and activity. Although both of these may point towards a higher risk of progression, the treatment decision may be completely divergent in patients with fibrosis versus those with active lesions. It can also be modified to be used as a surrogate marker of response to therapy with a decrease in predicted risk post-treatment, signifying an improvement.

Future directions

Research in the field of IgAN is evolving rapidly, with new drugs being tested in both the disease modifying and enhanced supportive care domains. Drug efficacy trials in IgAN should ideally be able to demonstrate a change in the rate of loss of kidney function in patients treated with a drug. This often led to prolonged trials since GFR decline in IgAN may take years to develop. A Kidney Health Initiative project workgroup was established to look for surrogate endpoints for trials of IgAN. The workgroup found an association between treatment effects on percent reduction of proteinuria and treatment effects on a composite of time to doubling of serum creatinine, kidney failure, or death. Proteinuria reduction was hence established as a reasonably likely surrogate endpoint for a treatment’s effect on progression to kidney failure in IgAN. This facilitated the use of this endpoint for accelerated drug approvals by US FDA with the requirement of post marketing confirmatory endpoints to establish drug efficacy in IgAN.

With the acceptance of proteinuria as a surrogate endpoint in trials of IgAN, many new drugs may soon become a part of standard of care in IgAN. Recently, sparsentan (dual endothelin and angiotensin receptor blocker) and oral budesonide (targeted release formulation) have received accelerated approval for use in IgAN whereas iptacopan (small molecule factor B inhibitor) has shown results which may lead to its accelerated approval. Considering each could decrease rate of disease progression in IgAN, and maybe used in combination therapy in the future, the tool will need modifications incorporating their use in the therapeutic armamentarium of IgAN.

Combining the tool with biomarkers, including microRNA expression, has been shown to improve the prediction performance of the tool, although this needs to be validated in larger cohorts. Although the current tool will probably always be useful considering its practical applicability in all, settings, a modified tool combined with biomarkers may be useful in trial recruitments and research settings to identify subsets of patients who benefit most from a particular therapy.


The IgAN PT combines histological and clinical parameters in adults and children to estimate the risk of progression of disease in IgAN. It is a step towards personalized medicine in IgAN, since the estimation of risk can form the basis of important therapeutic decisions for individual patients and serve as a useful parameter for designing trials with the barrage of novel drugs purported to be useful in this disease. The tool needs validation across various ethnicities and possibly a modification of version to incorporate treatment-responsive versus treatment-resistant lesions before we use it to guide therapeutic decision-making.

Share This Story

Leave A Comment