Development of a clinical prediction model of chronic kidney disease in primary Focal Segmental Glomerulosclerosis


 Background: Focal segmental glomerulosclerosis (FSGS) is one of the important causes of end stage kidney disease (ESKD). We evaluated the risk factors of progression of primary FSGS to chronic kidney disease (CKD) or ESKD with a predictive model including clinical and histological predictors. Methods: 201 patients with primary FSGS (59% male, mean age: 38±15 years), were studied. Time-dependent Cox model and C statistics were used for the predictive model. Interaction and correlation between independent variables were estimated. Results: During 55±27 months of follow-up, 82 patients (41%) developed CKD (46) or ESKD (36) patients. In adjusted model, 1 unit of higher serum creatinine (SCr) at baseline (HR:1.39, 95%CI: 1.15-1.70) and 1% increase in glomeruli with segmental glomerulosclerosis (SGS) (HR: 1.03, 95% CI: 1.02-1.04) or interstitial fibrosis/tubular atrophy (IF/TA) (HR: 1.03, 95% CI: 1.01-1.05) increased the risk of CKD/ESKD. In adjusted model, higher baseline proteinuria and collapsing variant were not associated with risk of CKD/ESKD. By adding SGS and IF/TA scores to baseline SCr in the model, discrimination by C statistics was 0.83 (95%CI: 0.77-0.90) for prediction of CKD/ESKD . Median renal survival was 3.1 years (95% CI: 2.2-4.1 years) in patients with highest risks score (baseline eGFR<25 ml/min/1.73 m 2 + IF/TA/SGS> 50%), and 8.1 years (95% CI: 7.7-8.6 years).in those with lowest score (baseline eGFR>75 ml/min/1.73 m 2 + IF/TA/SGS <5%). Conclusion: In primary FSGS, higher baseline SCr, increased SGS and IF/TA were the predictors for CKD/ESKD. Baseline proteinuria did not predict the risk of CKD/ESKD. Collapsing variant did not increase the risk of CKD/ESKD after adjustment for IF/TA score. These findings indicated the importance of baseline GFR and the degree of chronicity at biopsy as predictors of kidney outcome .


Background
Idiopathic or primary focal segmental glomerulosclerosis (FSGS) is a syndrome with proteinuria, mostly nephrotic range, focal and segmental glomerular sclerosis lesions and foot process effacement (1)(2)(3)(4). (FSGS) is known as a one of the leading glomerular causes of end stage kidney disease (ESKD) in most parts of the world (5)(6)(7). In a recent report by our group the prevalence of FSGS rose to second rank among primary glomerular diseases (8). However in reports from many countries FSGS is the most common glomerulopathy diagnosed by kidney biopsy (9, 10). In patients with primary FSGS 3 response to treatment or progression to chronic kidney disease (CKD) are very diverse and it has been reported that more than half of the patients progressed to end-stage kidney disease (ESKD) after 10 years of follow-up (7).
Columbia morphologic classification, which was defined in 2004, is the most popular method of classification of FSGS, and classifies FSGS to five pathologic variants including collapsing, cellular, tip lesion, perihilar and not otherwise specified (NOS) (1).It has been suggested that the histologic subtypes of FSGS correlate with remission and renal outcome (11,12).
Identifying the risk factors that predict the progression of primary FSGS to CKD/ESKD could enable appropriate patient care and improved individualized decision-making. The aim of this study was to

Pathological findings definitions:
The criteria for enrollment were at least 5 glomeruli in light microscopic field, at least 1 segmentally sclerotic glomerulus. Categorical data were presented as numbers and percentages, and continuous data as mean ± SD.
To compare the different clinical and laboratory characteristics of FSGS variants we used the Chi-2, one-way ANOVA and Post hoc Bonferoni tests, where appropriate. The correlation between SGS and IF/TA and laboratory findings were evaluated by Pearson test.
Continuous and categorical prognostic factors were analyzed by Cox regression model to estimate the risk of CKD/ESKD. Hazard Ratio (HR) with 95% confidence interval (CI) was estimated by univariate and multivariate analyses. Possible interactions between significant variables were tested. A significant interaction was found between SGS and IF/TA percentages (p=0.01) and also baseline SCr and IF/TA percentage (p=0.05). As a result we calculated the product of two independent variables in model by using centered interaction.
According to the sum of the baseline scores of eGFR, SGS and IF/TA, the patients were classified into three groups of low risk (total score 0-3), medium risk (total score: 4-6) and high risk (total score: 7-9). Finally Kaplan-Meier and the log-rank methods were used to estimate renal survival in each group (Supplementary Tab 1 and 2).
In prognostic model, discrimination was evaluated using the C-statistic, which represents the area under the receiver operating characteristic curve (ROC) curve. Accuracy of the test was described as area under the ROC curve (AUC). An area of 1.0 reflects perfect discrimination (sensitivity and specificity both 100%) and C-Statistic less than 0.5 is equivalent to random guessing.
In Multivariate analyses each main independent variable was adjusted for other variables. In addition, due to significant interaction between IF/TA percentages and SGS and IF/TA and SCr at baseline, we added the main effect of their products in the model, as explained in methodology.
Clinical and laboratory findings at baseline: In 201 patients, the mean age was 38±15 years and 6 119 were male (59%). Eighty patients (40%) were hypertensive at admission, however 70% of the patients with collapsing variant were hypertensive.

Histological findings:
The average number of glomeruli per one biopsy specimen was 19.8 ± 9.8. Risk category and renal survival: As mentioned before, the prognostic risk score was calculated according to the sum of the baseline scores of eGFR, SGS, IF/TA, and divided into three categories of low risk (≤ 3), medium risk (4-6 points), and high risk (7-9 points). 92(46%) were classified as low risk, 83 (41%) as medium risk, and 26 (13%) as high risk. Renal survival by risk category is shown in Figure 1.The median renal survival was calculated at 6.7 years for all patients (95% CI, 6.1 to7.2 years), 8.1 years for low risk patients (95% CI, 7.7 to 8.6), 5.8 years for medium-risk patients (95% CI, 5.1 to 6.5 years), and 3.3 years for high-risk patients (95% CI, 2.3 to 4.2 years; P< 0.001).

Five years prediction of CKD/ESKD:
The five years prediction of CKD/ESKD is given in figure 2. Receiver-operating characteristic curves (ROC-curve) with C-Statistic were used to estimate discrimination of the risk predictors. Baseline proteinuria as a predictor had C-Statistic less than 0.5 (0.40, 95% CI: 0.35-0.50) (figure 2). Discussion 8 FSGS is caused by podocyte loss and is mostly a progressive glomerular disease (13,14). Renal outcome is very diverse and progression to advanced kidney failure is common (15,16).
In the present study we assessed the predictors of CKD/ESKD in patients with primary FSGS. Fortyfive percent of our patients did not respond to treatment and 41% developed CKD or ESKD. The link between remission and renal survival was very high and 90% of patients without remission developed CKD/ESKD. In a study by Chun et al. the 10 years' kidney survival rates of patients with FSGS was 92% in patients with remission compared to 33% in those without remission (2).
Despite including 6 patients younger than 18 years, the demographics of our patients are almost similar to other studies of adult FSGS, with a mean age of 38±15 years in our study and an age at diagnose of 35-50 years old in other studies (11,15).The prevalence of idiopathic FSGS was somewhat more in men in our study similar to others (14,17).
Most of the patients in this study had mild renal impairment at baseline (MDRD eGFR: 65±38 ml/min/1.73m 2 ). Renal function at admission is very diverse in adult primary FSGS but it is mostly reported as mild to moderate renal dysfunction at admission (18,19). Our finding like others showed higher renal function at admission was positively related to better renal outcome (2,20).
In our study the most prevalent variant of FSGS was NOS (68%), followed by tip lesion (22%), perihilar (6%), collapsing (3%), and cellular types (1%). Demographic features are different between various histologic subtypes of FSGS. Collapsing lesion has been reported with a higher prevalence in Afro-American patients and both tip lesion and collapsing variants have been more common in teenagers and adults compared to children (11,12,21).
Most studies have reported NOS variant as the most prevalent type of FSGS which is similar to our study (12,21). However it is suggested that longer duration of FSGS may be correlated with NOS variant, probably due to evolution of segmental sclerosing lesions from tip lesion over time, as a part of chronicity process (22).
Regarding renal outcome, collapsing lesion has been associated with worst renal outcome and more common in non-responders (23). In our study all of the seven patients with collapsing lesion 9 developed CKD/ESRD, including the two patients who ended in dialysis.
Tip lesion variant has been associated with a significantly better kidney survival (24,25). Among 44 patients with tip lesion, only two did not attain complete or partial remission and developed CKD. In the current study 58% of patients with NOS did not response to treatment and more than half (52%) developed CKD/ESRD, which shows a worse prognosis in our NOS patients compared to other studies (11,21,26). This may have been the center effect and due to referral of more complicated patients to our kidney hospital, the effect of later referral compared to the mentioned studies or a basically different epidemiology of the disease in our country, which needs further investigation. Of the eleven patients with perihilar lesion, two progressed to CKD (18%) and were defined as non-responder. In the study by Thomas  We showed that in the collapsing variant the risk of CKD/ESKD is significantly high in univariate analyses, however after adding IF/TA score in the adjusted multivariate model, this risk was no longer significant. So the high frequency of bad outcome in collapsing FSGS may be due to higher chronicity at diagnosis and this should be noticed and examined in other series.
There are some limitations in using Columbia classification as a predictor of clinical outcome. Different location of segmental lesions makes overlapping and heterogeneity between variants, which can lead to misdiagnosis by kidney biopsy (27)(28)(29). The accuracy of diagnosis highly depends on the number of glomeruli and sample size in biopsy (30,31).
We showed that high IF/TA score at diagnosis strongly predicted the risk of CKD/ESKD in primary FSGS. The term interstitial fibrosis/tubular atrophy was first introduced in kidney allograft biopsy (32, 33).However today it is used as a predictor for progression of kidney failure, regardless of the type of the glomerular disease and interstitial fibrosis and tubular atrophy are well identified hallmarks for development of CKD (34)(35)(36) The presence of global and segmental sclerosis in kidney biopsy indicates worse prognosis in kidney survival (4,37). Global sclerosis is highly age-related and segmental sclerosis is the most important pathological finding in the typical FSGS (38). In this study we used the percentage of segmental sclerosis in the biopsy specimen as a predictor. Our finding showed that the higher number of glomeruli with segmental sclerosis on biopsy highly predicts the risk of CKD/ESKD.
In this study baseline proteinuria was not associated with increase in the risk of CDK/ESKD. This finding is contrary to the previous finding of proteinuria as an important risk factor for development of renal failure (39,40).On the other hand in a very recent study on 466 patients with primary FSGS and proteinuria, followed for 1,4 and 8 months, by using a novel definition of remission, reduction of proteinuria was associated with better long-term renal outcomes (41). In a sub-analysis in our patients

Conclusion
Our study showed that regardless of the histologic variant, kidney function at admission and percentages of glomeruli with segmental sclerosis and of interstitial fibrosis and tubular atrophy at kidney biopsy have high sensitivity and specificity to predict the kidney outcome. This is a retrospective observational study and these findings need to be examined by further prospective studies, which should consider the effect of therapeutic regimens.