The ISN/RPS 2016 classification predicts renal prognosis in patients with first-onset class III/IV lupus nephritis

Lupus nephritis (LN) is a life-threatening complication of systemic lupus erythematosus. The 2003 pathological classification of LN was revised in 2016; it quantitatively evaluates the interstitium in addition to the glomeruli. We performed a retrospective multi-centre cohort study and investigated the utility of the 2016 classification—including the activity index (AI), chronicity index (CI), and each pathological component to predict complete remission or renal function decline, defined as 1.5-fold increase in serum creatinine levels—and compare with that of the 2003 classification. Ninety-one consecutive adult patients with first-onset class III/IV LN who were newly prescribed any immunosuppressants were enrolled and followed up for a median of 51 months from January 2004. Cox regression analysis demonstrated the subclasses based on the 2003 classification, which mainly evaluate glomerular lesions, were not associated with clinical outcomes. After adjustments for estimated glomerular filtration rate and urinary protein levels, higher CI and higher interstitial fibrosis and lower hyaline deposit scores were associated with renal functional decline. Similarly, higher CI and interstitial inflammation scores were associated with failure to achieve complete remission. Therefore, the 2016 classification can predict the clinical outcomes more precisely than the 2003 classification.

www.nature.com/scientificreports/ or chronic lesions in the 2003 classification; subclass A was for purely active lesions, subclass A/C was for any combination of active and chronic lesion, subclass C was for purely chronic lesions 9 . AI includes pathological findings, such as endocapillary hypercellularity, neutrophils/karyorrhexis, fibrinoid necrosis, hyaline deposits, cellular and/or fibrocellular crescents, and interstitial inflammation. CI includes pathological findings, such as global/segmental sclerosis, fibrous crescents, interstitial fibrosis (IF), and tubular atrophy (TA). Of all these parameters, the scores of fibrinoid necrosis and cellular/fibrocellular crescents were set doubled weight. Notably, the 2016 classification incorporated the evaluation of tubulointerstitial lesions in the quantitative scoring system in the form of AI for interstitial inflammation and CI for IF/TA as opposed to the 2003 classification, which was merely based on the glomerular lesions. Several definitions of the pathological findings have also been revised.
To date, the clinical utility of the 2016 classification has not been fully investigated. The aim of the present study was to investigate the clinical usefulness of the 2016 classification with that of the 2003 classification by evaluating the achievement of complete remission (CR) and renal function decline in adult patients with first-onset class III/IV LN based on the Nagoya Kidney Disease Registry (N-KDR).

Results
Study participants. We screened 233 consecutive patients with LN in our real biopsy registry between January 2004 and December 2014. We enrolled patients who underwent the first renal biopsy, were ≥ 16 years of age, who met ≥ 4 American College of Rheumatology (ACR) criteria 26 of SLE, and were classified to have class III or IV LN. We excluded patients with missing medical or pathological records (n = 9), a history of renal function deterioration (n = 4), conservative treatment without immunosuppressive therapy (n = 1), immunosuppression before induction therapy for LN (n = 49), observational period less than a month (n = 1), and evaluable glomeruli less than six (n = 1). Finally, 91 patients were enrolled in this study. We assessed their pathological findings and renal function decline during the observational duration (Analysis 1). Of these, six patients were excluded because of missing adequate follow-up data, and 85 were assessed for CR (Analysis 2). The detailed flowchart is shown in Fig. 1.
Baseline characteristics. Baseline characteristics are summarized according to the eGFR 27 levels at baseline as lower eGFR group (eGFR < 60 ml/min/1.73 m 2 , n = 42 [46%]) and higher eGFR group (eGFR ≥ 60 ml/ min/1.73 m 2 , n = 49 [54%]) ( Table 1). Patients in the lower eGFR group were older, had heavier proteinuria, more severe haematuria and higher proportion of nephrotic syndrome than those in the higher eGFR group. Anti-dsDNA, serum C3 levels, and SLE disease activity index (SLEDAI) 28 (Fig. 2a), while there was no difference in the A and A/C subclasses (Fig. 2b). Both AI and CI were higher in the lower eGFR group (Fig. 2c,d) than those in the higher eGFR group. In the pathologi- Figure 1. Flow chart of patient selection. Ninety-one patients with first-onset lupus nephritis were enrolled in this study and assessed for pathological findings and renal function decline during the observational duration (Analysis 1). After excluding 6 patients with missing follow-up data, achievement of complete remission was evaluated for 85 patients (Analysis 2). LN lupus nephritis; ACR American College of Rheumatology.
Medications during the induction therapy, clinical outcomes, and adverse events. The overall median observation period was 51 (IQR: 23-77) months, and there was no statistically significant difference between the groups (p = 0.50) ( Table 3). Median interval from renal biopsy to start of the induction therapy was 1 (IQR: − 7-9) day, and there was no statistically significant difference between the two groups (p = 0.85). During induction therapy, prednisolone was prescribed for all patients. The proportion of patients in the lower eGFR group who received methylprednisolone pulse therapy was higher than that in the higher eGFR group (62% [26/42] vs. 51% [25/49], respectively). However, the proportion of patients who received any type of immunosuppressants was not statistically different between the two groups. Of all patients, five were lost to follow-up and four died during induction therapy. Of the remaining, 82 received maintenance therapy, and of these, 66 responded to induction therapy 29 . There was no statistically significant difference in the content of maintenance treatment between the groups. Overall, 54/85 patients achieved CR; the cumulative incidence of CR in the lower eGFR group was lower (38%, 15/39) than that in the higher eGFR group (55%, 39/46). Overall (n = 91), 16 patients developed 1.5-fold increase in sCr, eight patients had doubling of sCr, and two patients reached endstage renal disease (ESRD) during the entire observation period. Six patients died, and all of them were in the lower eGFR group. Regarding the adverse events after the initiation of induction therapy, the incidence of steroids-induced diabetes was significantly higher in the lower eGFR group (52%, 22/42) than that in the higher eGFR group (31%, 15/46) (see Supplementary Table S2 online).
Survival curves for renal function decline and CR. The cumulative incidence of renal event (1.5-fold increase in sCr)-free survival and CR are illustrated in Fig. 3. Time to CR was assessed within 5 years from the initiation of induction therapy because none of the patients achieved CR after 5 years. The baseline eGFR levels were not associated with renal function decline (p = 0.80) (Fig. 3a), but patients in the higher eGFR group were  www.nature.com/scientificreports/ more likely to achieve CR than were those in the lower eGFR group (p < 0.001) (Fig. 3b). Similarly, the presence of nephrotic syndrome was not associated with renal function decline (p = 0.84) (Fig. 3c), but patients without nephrotic syndrome were also more likely to achieve CR than those with nephrotic syndrome (p = 0.006) (Fig. 3d). www.nature.com/scientificreports/  Table 4).

Identification of clinical predictors of CR.
Baseline renal function decline was associated with achieving CR, while nephrotic syndrome, anti-dsDNA antibody, and serum C3 levels were not (  In the present study, CI was associated with renal function decline and CR independently of eGFR and urinary protein levels mainly due to its high correlation with the scores of interstitial lesions. Both AI and CI were predictive of CR. Of the components of AI, interstitial inflammation was associated with CR, and of the components of CI, IF was independently associated with renal function decline. Therefore, it is crucial to assess interstitial lesions in order to predict renal prognosis in patients with LN. In contrast, AI was not associated with CR after adjusting for eGFR and urinary protein levels. Cellular crescents, which were highly correlated with AI, had moderate correlation with eGFR and urine protein levels. These correlations probably attenuate the association of AI and CR. Therefore, we demonstrated the utility of CI and importance of assessing interstitial regions in predicting renal prognosis, as previously reported [19][20][21][22][23] . In our study, however, we did not identify active glomerular lesions as potential risk factors of poor renal prognosis. Crescentic lesions and fibrinoid necrosis were not associated with renal function decline in our study, although previous reports showed them as indicators for poor renal prognosis 4,19,23,29,30 . Hyaline deposits were rather inversely correlated with renal function decline in the present study while Austin et al. 25 adopted it as an active indicator associated with prognosis. A recent research for clinical and histopathologic predictors of renal Table 4. Associated factors for renal function decline. eGFR estimated glomerular filtration rate, HR hazard ratio, CI confidence interval. *Model 1 is adjusted by activity index and chronicity index by the baseline eGFR levels (cut-off of 60 ml/min/1.73 m 2 ) and the presence of nephrotic syndrome. **Model 2 is adjusted by pathological components by eGFR and urinary protein levels. Pathological variables associated with 1.5-fold increase in serum creatinine were identified on forward-backward stepwise regression analysis. † Number of missing data: N = 27. ‡ Log-transformed. www.nature.com/scientificreports/ outcomes for LN demonstrated that wire loops, or hyaline deposits, were associated with eGFR recovery rather than decline 23 . This is consistent with our results. There are two possible reasons for these discrepancies. One is the improvement of treatment for LN over time. Our patients received immunosuppressant therapy depending on their disease activities, and as high as 80.5% of them responded to the treatments accordingly. Of active glomerular lesions, hyaline deposits, or subendothelial deposits, might represent an early pathological change of LN that was likely to heal easily by immunosuppressive treatment. Another reason is the differences in the background of patients. Most of the previous studies included first-onset LN patients as well as those who had already been treated for SLE. In contrast, we included only first-onset LN patients without previous immunosuppressive treatments. Because active glomerular lesions of LN were considered to be reversible, we believe that they did not reflect the long-term renal prognosis in our study. We suggest that the pathological classification system should be improved by investigating the effects of each pathological component through an evidence-based process such as the MEST score in the Oxford classification of IgA nephropathy 31 . Our results suggest that treatment resistance factors, such as interstitial lesions and treatment response factors, such as hyaline deposits should be considered separately. Further investigations are required to identify the pathological findings that are associated with the clinical outcomes and determine their weightages in the scoring system.

Univariable analysis
There were several limitations to this study. First, this was a retrospective observational study. However, to the best of our knowledge, this is the largest multi-centre cohort study of adult patients with first-onset class III/ IV LN. These results can be generalizable in various clinical settings. Second, there might have been substantial differences in the treatment strategies between the hospitals. There was no unified protocol for the treatment, Table 5. Associated factors for complete remission. eGFR estimated glomerular filtration rate, HR hazard ratio, CI confidence interval. *Model 1 is adjusted by activity index and chronicity index by the baseline eGFR levels (cut-off of 60 ml/min/1.73 m 2 ) and the presence of nephrotic syndrome. **Model 2 is adjusted by a pathological component by eGFR and urinary protein levels. Pathological variables associated with complete remission were identified on forward-backward stepwise regression analysis. † Number of missing data: N = 29. ‡ log-transformed. www.nature.com/scientificreports/ and it was decided at the discretion of the doctors. The potential differences in the treatment strategies over the course of the study period might have also affected the clinical course of LN. However, these results reflect the real-world data and have high generalizability.

Univariable analysis
In conclusion, we demonstrated that comprehensive and quantitative assessments of the renal biopsy specimen based on the 2016 classification can provide useful information to predict the renal prognosis in patients with first-onset class III/IV LN. Of the pathological findings, interstitial lesions were strong predictors of both short-and long-term renal prognoses. Further prospective validation studies are currently underway.

Methods
Patient selection and study design. This study was a retrospective, multi-center cohort study. Primary LN was diagnosed in 233 consecutive patients from N-KDR between January 2004 and December 2014. Inclusion criteria were as follows: (1) diagnosed at first-biopsy, (2) aged over 16 years, (3) fulfilled 4 and more ACR criteria 26 , and (4) diagnosed with class III/IV LN. Exclusion criteria were as follows: (1) no medical or pathological records, (2) history of renal function decline, (3) no induction therapy, (4) previous immunosuppression, (5) less than 1-month observation period, and (6) total evaluable number of glomeruli less than 6. A history of renal function deterioration was defined as follows: (1) renal atrophy at diagnosis or (2) continuous decline in estimated glomerular filtration rate (eGFR) < 60 ml/min/1.73 m 2 within 3 months prior to diagnosis. Induction therapy was defined as the 6-month immunosuppressive medications for remission induction for LN. Previous immunosuppression was defined as history of other immunosuppressive therapies before ≥ 2 weeks of initiation of induction therapy for LN. Overall, 91 patients with first-onset class III/IV LN and new prescriptions of any immunosuppression were observed between January 2004 and July 2016; the observations were performed until ESRD or death, whichever was early, or the last available data of urinary proteins or sCr. All of them were followed up at the following 16 27 , anti-dsDNA antibody level, serum C3 level, 24-h urinary protein excretion (g/day) or urinary protein-to-creatinine ratio (g/gCr), haematuria, and SLEDAI score 28 . The severity of haematuria expressed as −/+/++/+++. Nephrotic syndrome was defined as urinary protein ≥ 3.5 g/day or urinary protein-to-Cr ratio ≥ 3.5, and serum albumin < 3.0 mg/dl. Medications during induction and maintenance therapy. All drugs used during induction and maintenance therapy were investigated. Induction therapy was defined as the immunosuppressive therapy for the first 6-month of treatment for LN. Maintenance therapy was defined as the immunosuppressive therapies administered after the 6-month induction therapy. The drugs included prednisolone, methyl prednisolone pulse, calcineurin inhibitors (cyclosporine or tacrolimus), cyclophosphamide, azathioprine, mizoribine, mycophenolate mofetil and rituximab.
Adverse events. Adverse events after the initiation of induction therapy included cardiovascular disease, cerebrovascular disease, femoral head osteonecrosis, steroids-induced diabetes, gastric ulcers, first infectious disease that required hospitalization, herpes zoster or cytomegalovirus infections that required medications, and cancer. Steroids-induced diabetes was defined as initiating new antidiabetic medications after the initiation of induction therapy.
Definition of clinical outcomes. The primary outcome was renal function decline, which was defined as 1.5-fold increase in sCr or 50% increase in sCr from the baseline level. The secondary outcome was the achievement of CR, which was defined as achievement of both proteinuria < 0.5 g/gCr or g/24 h and recovery of normal renal function 32 . Normal renal function was defined as (1) returning to the sCr levels before the onset of LN or (2) sCr < 1.0 mg/dl (if male) and < 0.7 mg/dl (if female) if the past sCr level was unknown. Treatment response to induction therapy was assessed at 6 months after the initiation of induction therapy, which was defined as both ≥ 50% decrease in proteinuria from the baseline to at least sub-nephrotic levels and stabilization (± 25%) or improvements in sCr (but not completely reverting to normal) 33  www.nature.com/scientificreports/ level from the baseline value. ESRD was defined as the disease stage that required initiation of dialysis or renal transplantation.

Statistical analysis.
Continuous variables with asymmetric distribution are presented as median [IQR].
Categorical variables are expressed as percentages. Spearman's correlation coefficients were used to examine the relationships between the continuous variables. The cumulative probability of attaining the outcomes was calculated using the Kaplan-Meier method, and log-rank test was employed for hypothesis testing. The timeto-clinical outcomes were calculated between the date of the initiation of induction therapy and the date of the clinical outcomes. Loss to follow-up, ESRD, and all-cause death were censored. In order to use the 2016 classification for quantitative prognostic evaluation, we performed exploratory investigation of their mutual correlation and relevance to the renal prognosis using Rs. The proportional hazards assumption for covariates was tested using scaled Schoenfeld residuals. Both baseline and pathological data were examined using univariable and multivariable Cox's proportional hazards models in order to identify independent predictors associated with the clinical outcomes. Covariates included both the clinical and pathological findings, and we selected pathological components using a stepwise method to avoid multicollinearity of these findings. All statistical models were performed using complete case analysis. The level of statistical significance was set at p value < 0.05. All statistical analyses were performed using Stata SE v14.0 (STATA Corp, 4905 Lakeway Drive College Station, Texas 77845-4512, USA, www.stata .com).

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.