Prediction of kidney-related outcomes in patients with type 2 diabetes

Meg J. Jardine, Jun Hata, Mark Woodward, Vlado Perkovic, Toshiharu Ninomiya, Hisatomi Arima, Sophia Zoungas, Alan Cass, Anushka Patel, Michel Marre, Giuseppe Mancia, Carl E. Mogensen, Neil Poulter, John Chalmers

Research output: Contribution to journalArticle

Abstract

Background: Tools are needed to predict which individuals with diabetes will develop kidney disease and its complications. Study Design: An observational analysis of a randomized controlled trial. Setting & Participants: The ADVANCE (Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation) Study followed up 11,140 participants with type 2 diabetes for 5 years. Predictor: Readily available baseline demographic and clinical variables. Outcomes: (1) Major kidney-related events (doubling of serum creatinine to <2.26 mg/dL [<200 μmol/L], renal replacement therapy, or renal death) in all participants, and (2) new-onset albuminuria in participants with baseline normoalbuminuria. Measurements: Cox proportional hazard regression models predicting the outcomes were used to generate risk scores. Discrimination of the risk prediction models was compared with that of models based on estimated glomerular filtration rate (eGFR) alone, urinary albumin-creatinine ratio (ACR) alone, and their combination. Results: Risk scores for major kidney-related events and new-onset albuminuria were derived from 7- and 8-variable models, respectively. Baseline eGFR and ACR were dominant although models based on the 2 factors, alone or combined, had less discrimination (P < 0.05) than the risk prediction models containing additional variables (risk prediction model C statistics of 0.847 [95% CI, 0.815-0.880] for major kidney-related events, and 0.647 [95% CI, 0.637-0.658] for new-onset albuminuria). Novel risk factors for new-onset albuminuria included Asian ethnicity and greater waist circumference, and for major kidney-related events, less education. The risk prediction models had acceptable calibration for both outcomes (modified Hosmer-Lemeshow test, P = 0.9 and P = 0.06, respectively). Limitations: The follow-up period was limited to 5 years. Results are applicable to people with type 2 diabetes at risk of vascular disease. Conclusions: Risk scores have been developed for early and late events in diabetic nephropathy. Although eGFR and urinary ACR are important components of the prediction models, the extra variables considered add significantly to discrimination and, in the case of new-onset albuminuria, are required to achieve satisfactory calibration.

Original languageEnglish (US)
Pages (from-to)770-778
Number of pages9
JournalAmerican Journal of Kidney Diseases
Volume60
Issue number5
DOIs
StatePublished - Nov 1 2012

    Fingerprint

Keywords

  • Diabetic nephropathy
  • chronic kidney disease
  • kidney failure
  • prediction model
  • risk score

ASJC Scopus subject areas

  • Nephrology

Cite this

Jardine, M. J., Hata, J., Woodward, M., Perkovic, V., Ninomiya, T., Arima, H., Zoungas, S., Cass, A., Patel, A., Marre, M., Mancia, G., Mogensen, C. E., Poulter, N., & Chalmers, J. (2012). Prediction of kidney-related outcomes in patients with type 2 diabetes. American Journal of Kidney Diseases, 60(5), 770-778. https://doi.org/10.1053/j.ajkd.2012.04.025