New risk score for predicting progression of membranous nephropathy

Background Patients with Idiopathic membranous nephropathy (IMN) have various outcomes. The aim of this study is to construct a tool for clinicians to precisely predict outcome of IMN. Methods IMN patients diagnosed by renal biopsy from Shanghai Ruijin Hospital from 2009.01 to 2013.12 were enrolled in this study. Primary outcome was defined as a combination of renal function progression [defined as a reduction of estimated glomerular filtration rate (eGFR) equal to or over 30% comparing to baseline], ESRD or death. Risk models were established by Cox proportional hazard regression analysis and validated by bootstrap resampling analysis. ROC curve was applied to test the performance of risk score. Results Totally 439 patients were recruited in this study. The median follow-up time was 38.73 ± 19.35 months. The enrolled patients were 56 (15–83) years old with a male predominance (sex ratio: male vs female, 1:0.91). The median baseline serum albumin, eGFR-EPI and proteinuria were 23(8–43) g/l, 100.31(12.81–155.98) ml/min/1.73 m2 and 3.98(1.50–22.98) g/24 h, respectively. In total, there were 36 primary outcomes occurred. By Cox regression analysis, the best risk model included age [HR: 1.04(1.003–1.08), 95% CI from bootstrapping: 1.01–1.08), eGFR [HR: 0.97 (0.96–0.99), 95% CI from bootstrapping: 0.96–0.99) and proteinuria [HR: 1.09 (1.01–1.18), 95% CI from bootstrapping: 1.02–1.16). One unit increasing of the risk score based on the best model was associated with 2.57 (1.97–3.36) fold increased risk of combined outcome. The discrimination of this risk score was excellent in predicting combined outcome [C statistics: 0.83, 95% CI 0.76–0.90]. Conclusions Our study indicated that older IMN patients with lower eGFR and heavier proteinuria at the time of renal biopsy were at a higher risk for adverse outcomes. A risk score based on these three variables provides clinicians with an effective tool for risk stratification. Electronic supplementary material The online version of this article (10.1186/s12967-019-1792-8) contains supplementary material, which is available to authorized users.


Background
Idiopathic membranous nephropathy (IMN) is one of the most common types of adult-onset primary glomerulonephritis [1][2][3]. The incidence of IMN has increased dramatically recently at least in China [2,4,5] which maybe partly due to air pollution for example the increased level of PM2.5 in the air [5]. IMN is an immune complexmediated glomerular disease. The understanding of the pathophysiological mechanism underlying IMN has been greatly improved thanks to the discovery of anti-PLA2R and anti-THSD7A antibodies in IMN patients [6,7]. Interestingly, previous studies [8][9][10] based on western population have shown that the level of anti-PLA2R antibody in serum was helpful in the differential diagnosis and the prognosis prediction.
It was reported that approximately one-third of all IMN patients will develop end stage renal disease (ESRD). Both clinical variables including age, gender, serum creatinine, proteinuria and histological variables including tubulointerstitial fibrosis and focal segmental sclerosis(FSGS) at time of diagnosis were associated with renal function progression in IMN patients based on prior studies [11,12]. However, Trayanov et al. [13] failed to validate the correlation between FSGS and progressive renal disease. Zent et al. [14] found that elderly and young patients had similar rates of ESRD (12% vs 18%, P > 0.05) based on a cohort of 323 IMN patients. The discrepant findings suggested validating studies were necessary in independent cohorts with diverse populations since most of these studies were performed in Western countries. Finally, establishing a risk model to combine the independent predictors could potentially improve the accuracy of prediction since the effect of each single predictor is relatively small. In this study, we enrolled an extended Chinese IMN cohort to establish a risk score to precisely predict the outcome of these patients. This prediction tool will be helpful to clinicians for assessing the risk classification of IMN patients and to decide who need more aggressive treatments and more frequent follow-up.

Study population and study design
All the patients in this study were recruited at Shanghai Ruijin Hospital from 2009.01 to 2013.12. The inclusion criteria were as follows: (1) renal biopsy was required for the diagnosis of IMN; (2) age ≥ 15 years; (3) informed consent was obtained. The exclusion criteria were: (1) patients with secondary causes of membranous nephropathy, such as malignancy, autoimmune disease and hepatitis B. (2) Patients receiving immunosuppressive treatment before hospitalization in our nephrology service. (3) Patients with severe heart failure or hepatic failure.
The primary outcome was defined as a combination of renal function progression, ESRD or death. Renal function progression was defined as a reduction in eGFR greater than or equal to 30% compared with renal function at the time of renal biopsy [15]. ESRD was defined as the need for dialysis or kidney transplant.

Data collection
The patients' demographic characteristics, baseline and follow-up clinical data were collected. eGFR was calculated by eGFR-EPI formula [16]. Renal biopsies were evaluated and scored by 2 experienced nephropathologists. The IMN stages was assessed based on the criteria listed below. Pathological staging oF IMN [17].

Statistical analysis
Continuous variables that were normally distributed were expressed as the mean ± SD and compared with Student's t-test. Continuous variables that had a skewed distribution were presented as the median (Range) and compared with the Mann-Whitney U test. Categorical variables were compared with the Chi squared test. The proportional hazard assumption was checked by testing covariate-by-time interactions for each variable. The Cox proportional hazards models were built to test the associations between the variables and the outcomes. The best model was selected using a stepwise selection of variables using Akaike information selection criterion. Variables with P values less than 0.05 in the univariate analysis were included in the multivariate Cox proportional hazards models. A nonparametric bootstrapping resampling analysis with replacements was used to validate the risk factors obtained in multivariate Cox analysis. Only the independent predictors validated by bootstrapping resampling analysis were retained in the final model for the calculation of risk score. ROC curve were generated to compare the discrimination among the risk factors and risk scores. Two sided P value < 0.05 was considered statistically

Baseline demographic and clinical data
In total, 439 patients were recruited. The baseline characteristics of the enrolled patients are listed in Table 1 Fig. 1a).

Identification of risk factors and establishing risk scores
The proportional hazard assumption was checked by testing covariate-by-time interactions for each variable (Additional file 1: Table S2) showing all these variables agreed with proportional hazard assumption. In the univariate analysis, age, serum albumin, proteinuria, eGFR and severe interstitial fibrosis were associated with primary outcomes. In multivariate analysis, the best model included age  Table 2).

Assessment of the risk score performance
The ROC curve was generated to compare the prognostic values of the identified risk factors with the risk scores. The risk score (C statistics: 0.83, 95% CI 0.76-0.90) was better than each of the 3 risk factors alone including age(C statistics: 0.79, 95% CI 0.72-0.86), eGFR-EPI (C statistics: 0.81, 95% CI 0.73-0.89) and proteinuria (C statistics: 0.69, 95% CI 0.61-0.77), which suggested that the risk score was had better discrimination in predicting adverse outcomes in IMN patients (Fig. 2a). Besides, we compared the predicted risk versus observed rate of primary outcome. The discriminative slope was calculated as the difference between the mean predicted probability between IMN patients with or without primary outcome (1.80) indicating good performance of the risk score (Fig. 2b). Next we divided the patients into 2 groups based on median of risk scores at − 0.29. A Kaplan-Meier curve revealed that the patients in the high risk subgroup had a 5.97-fold increased risk for developing primary outcomes (HR: 5.97, 95% CI 2.32-15.35) compared with patients in the low risk subgroup (Fig. 2c).

Associations between serum PLA2R antibody, renal PLA2R staining and primary outcome
There was no significant difference in serum PLA2R antibody titer (30.40 RU/L vs 12.17 RU/L, P = 0.09) and renal PLA2R positive staining (100% vs 84.27%, P = 1.00) between patients with or without primary outcome occurrences. Then, the patients were divided evenly into two groups (high vs low serum PLA2R antibody groups) based on median serum PLA2R antibody titers. A Kaplan-Meier curve revealed a similar tendency toward primary outcomes in the two groups (P = 0. 19). No significant difference was observed in the occurrence of primary outcomes in patients with positive or negative renal PLA2R staining.

Discussion
The disease course of IMN is quite variable. An effective tool for clinicians to decide which patients need more aggressive therapy and more frequent follow-up would be helpful in clinical practice. However, the risk factors reported by previous studies [11,12]  The risk score showed a good discriminating based on ROC curve. One unit increasing of the risk score was associated with 2.57 (1.97-3.36) fold increasing risk of primary outcome. To our knowledge, this is the first risk prediction tool based on baseline parameters that has been proposed for risk stratification in IMN patients and is helpful to improve clinical practices. In our study, we found that age, eGFR and proteinuria were 3 independent risk factors for unfavorable outcome in IMN patients. The associations between age and ESRD in IMN patients was controversial based on previous studies. Shiiki et al. [3] enrolled 949 Japanese IMN patients and found that male gender, older age (≥ 60 years), higher serum creatinine concentration (≥ 1.5 mg/dl) and tubulointerstitial changes were associated with ESRD. However, no correlation was found between age and renal progression in Zent's study [14]. He found although the mortality rate was higher, the ESRD rate was similar in elderly IMN patients than in young patients. It's probably due to the competing risk of death and renal progression that masked the effect of age on renal outcome. Our study confirmed the correlation between age and a combined outcome consisting of renal function progression, ESRD and death. The associations between creatinine, proteinuria and inferior renal outcome in IMN patients were reported by several studies [8,18]. In our study, we validated these findings in Asian population by using eGFR-EPI instead of creatinine which is a better way in evaluating baseline renal function.
In our study, we did not validate the correlations between baseline serum anti-PLA2R antibody levels, renal PLA2R antigen and adverse outcomes in IMN patients. Kanigerchela et al. [19] found that high level of PLA2R antibody was correlated to disease activity and a higher risk for renal function deterioration based a study of 90 IMN patients. More recently, studies [9,10] suggested that dynamic measurement of serum anti-PLA2R antibodies was helpful to predict treatment response and relapses. Considering the limited number of patients with serum anti-PLA2R antibody and renal PLA2R antigen measurements in our study, a more extended study is needed to draw further conclusions.
Cattran et al. has proposed a predictive model of renal prognosis in IMN patients based on dynamic changes of proteinuria and creatinine [20][21][22]. This model need follow up patients for a period of time before the risk stratification. In our study, we constructed a risk predicting tool based on baseline parameters, which can be directly applied into newly diagnosed IMN patients. The risk score showed a good discrimination based on ROC curve and discriminative slope between IMN patients with or without primary outcome. Thus, our risk score will be useful for clinicians to perform risk stratification for IMN patients.
Our study has several limitations. An external cohort of more ethnically diversified patients is needed for further validation. In addition, our study was based on a retrospective review of IMN patients, which needed to be further evaluated in a prospective cohort.

Conclusion
Our study indicated that older IMN patients with lower eGFR and heavier proteinuria at the time of renal biopsy were at a higher risk for having adverse outcomes. A new risk score based on these three variables provides clinicians an effective tool for risk stratification.