Oral Presentation Australian Diabetes Society and the Australian Diabetes Educators Association Annual Scientific Meeting 2017

Validation of models for predicting rapidly declining renal function in type 2 diabetes: the Fremantle Diabetes Study Phase II (#153)

Kirsten E Peters 1 2 , Wendy A Davis 1 , Jason Ito 2 , Kaye Winfield 2 , Thomas Stoll 2 , Scott D Bringans 2 , Richard J Lipscombe 2 , Timothy ME Davis 1
  1. Medical School, University of Western Australia, Perth, Western Australia, Australia
  2. Proteomics International, Perth, Western Australia, Australia

There is a need for earlier detection of individuals at risk of chronic kidney disease (CKD) to optimise timely intervention and monitoring of disease progression. In a previous study we identified circulating protein biomarkers (APOA4, CD5L, C1QB, and IBP3) that improved clinical models for predicting rapid decline in estimated glomerular filtration rate (eGFR). The aim of the present study was to validate these models in an independent cohort of patients with type 2 diabetes.

 

A mass spectrometry platform was used to measure baseline biomarkers in 792 participants from the longitudinal observational Fremantle Diabetes Study Phase II. Rapid eGFR decline was defined as i) incident CKD, ii) eGFR decline ≥30% over four years, iii) annual eGFR decline ≥5 mL/min/1.73m2, and iv)  eGFR declining trajectories. Prediction models were developed using multiple logistic regression in a training cohort (n=345) before validation in an independent cohort (n=447). Model performance was assessed in the validation cohort by comparing model predictions to actual outcomes using indices of discrimination (ROC-AUC) and calibration (Hosmer-Lemeshow goodness-of-fit test).

 

The development and validation cohorts had similar baseline eGFR (80.6±18.8 vs 82.7±16.9 ml/min/1.73m2, respectively), but differed in age and diabetes duration (P<0.05). During 4.2±0.3 years of follow-up, 5-10% of participants experienced rapid eGFR decline. Applied to the validation cohort, the best performing model was for incident CKD (AUC=0.88 (95%CI 0.84-0.93)); calibration chi-square=5.7, P=0.77). At the optimal score cut-off, this model provided 86% sensitivity, 78% specificity, 30% positive predictive value and 98% negative predictive value to predict four year risk of developing CKD.

 

The present study assessed and validated the prognostic utility of novel plasma biomarkers for prediction of rapidly declining kidney function in type 2 diabetes. These prediction models may be useful for risk stratification in future clinical trials and incorporation into clinical decision making.