Predicting renal survival in primary focal glomerulosclerosis from the time of presentation

Background. To predict the risk of developing chronic renal failure in patients with primary focal glomerulosclerosis (FGS) using predictors available at the time of presentation, a retrospective analysis was performed on 111 patients who were diagnosed at Christchurch Hospital from 1965 to 1998. Methods. The predictors of outcome included age, gender, systolic and diastolic blood pressure, serum albumin, plasma creatinine, presence of hematuria, and amount of proteinuria tall at the time of presentation). An injury score (combination of percentage of sclerosed glomeruli and proportion of tubulointerstitial fibrosis) was derived from a review of the initial kidney biopsy. Log-logistic accelerated failure time parametric models were used. Results. The median renal survival was 16.4 years (Kaplan-Meier estimate). The best single variable model was that using the proportion of tubulointerstitial fibrosis (global chi-square 55.99, P < 0.0001). However, inclusion of plasma creatinine significantly improved the fit of the model (global chi-square 65.04, P < 0.0001). This joint model was superior to the single-variable model. Both of the models were validated using jackknifing. Conclusion. For a patient with primary FGS, these models can be used to predict the risk of developing chronic renal failure at any time and the median renal survival, given the proportion of tubulointerstitial fibrosis and plasma creatinine at the time of presentation.