Conference abstract

Validation of two non-invasive risk models for predicting prevalent undiagnosed chronic kidney disease in diabetic patients receiving care at the Douala General Hospital

Pan African Medical Journal - Conference Proceedings. 2017:2(26).05 Sep 2017.
doi: 10.11604/pamj-cp.2017.2.26.58
Archived on: 05 Sep 2017
Contact the corresponding author
Keywords: Chronic kidney disease, risk model, validation, calibration, discrimination
Oral presentation

Untitled Document

Validation of two non-invasive risk models for predicting prevalent undiagnosed chronic kidney disease in diabetic patients receiving care at the Douala General Hospital

Nina Lubeka1, André Pascal Kengne2, François Folefack Kaze3, Siméon Pierre Choukem4,&

1Department of Internal Medicine, Faculty of Health Sciences, University of Buea, Buea, Cameroon, 2Department of Medicine, Groote Schuur Hospital, University of Cape Town, Cape Town, South Africa, 3Department of Internal Medicine and Specialties, Faculty of Medicine and Biomedical Sciences, University of Yaounde I, Yaounde, Cameroon, 4Department of Internal Medicine, Faculty of Health Sciences, University of Buea, Buea, Cameroon

&Corresponding author
Siméon Pierre Choukem, Department of Internal Medicine, Faculty of Health Sciences, University of Buea, Buea, Cameroon

Abstract

Introduction: undiagnosed CKD is common in diabetes mellitus. The main attraction of non-invasive CKD risk models in diabetic patients will help direct confirmatory screening only to those who are more likely to be diagnosed with CKD. We opted to validate the Korean model and the Thai non-invasive CKD risk prediction models by assessing the discrimination, calibration performance and effects of simple recalibration through intercept adjustment on those performance measures.

Methods: we recruited diabetic patients receiving care at the Douala General Hospital, using records in medical files. Risk models were identified through a systematic review. The MDRD and CKD-EPI equations were used to estimate the glomerular filtration rate (eGFR). The outcome of CKD was defined as CKD (eGFR < 60ml/min/1.73 m2) and any nephropathy (eGFR < 60ml/min/1.73 m2 and/or proteinuria). Discrimination was assessed and compared using c-statistics and non-parametric methods. Calibration performance was assessed via calibration curves before and after recalibration through intercept adjustment.

Results: we included 733 participants in our study, with 421 (57.4%) being women. The mean age was 57.0 overall and 56.5 and 57.7 for men and women respectively (p = 0.118). The MDRD equation diagnosed 223participants with CKD and 377participants with ‘any nephropathy’ whereas the CKD-EPI equation diagnosed lesser participants with any nephropathy. The original Korean model had the highest c-statistics of 0.696 for the outcome of eGFR < 60ml/min/1.73 m2 (CKD-EPI equation). Discrimination was better in subgroups of men, older and overweight participants. Intercept adjustment markedly improved calibration with Expected/Observed event rate being 0.85 (95% CI) for the Korean model and 0.97 (95%CI, any nephropathy) for the Thai model.

Conclusion: both models have modest discrimination and good calibration with modest adjustment in predicting undiagnosed CKD in diabetic patients receiving care at the Douala General Hospital. Only the Korean model appears to have acceptable enough performance for possible consideration for use in clinical practice in this setting.