Risk Prediction for patients with Chronic Kidney Disease.
Background: Chronic kidney disease is a major public health problem with increasing
incidence and prevalence worldwide. Kidney disease severity can be classified by
estimated glomerular filtration rate (eGFR) and albuminuria, but more accurate
information regarding risk for progression to kidney failure is required for clinical
decisions about testing, treatment and referral. ... read moreObjectives: To perform a systematic
review of the risk prediction literature for important clinical outcomes in CKD; To
develop and validate a new risk prediction instrument for progression of CKD to kidney
failure; and To evaluate a novel method for risk prediction for modeling CKD progression
Methods: We systematically searched MEDLINE for articles that included patients with
CKD, and predicted kidney failure, all cause and cardiovascular mortality. We then
examined the relevant articles for quality of reporting and evidence of clinical
utility. Subsequently, we developed and validated a laboratory based risk prediction
model for progression of CKD to kidney failure. Finally, we modified our published lab
based prediction model to include time dependent covariates and internally validated a
dynamic model for progression of CKD. Results: Our systematic review identified 10
studies describing 14 models in patients with CKD. Six studies (8 models) predicted
kidney failure, four studies (4 models) predicted all cause mortality, and two studies
(2 models) predicted cardiovascular events. Study quality was heterogeneous, with higher
quality models available for prediction of kidney failure than for other outcomes. Our
predictive model included age, sex, eGFR, albuminuria, serum calcium, phosphate,
bicarbonate and albumin and accurately predicted CKD progression. (C-statistic 0.92 in
development and 0.84 in validation). Finally, our dynamic model demonstrated an
incremental improvement in model performance (C statistic 0.91 vs 0.90, IDI 1.4 %, NRI
18.4 %). Discussion: In summary, we performed a systematic review of the relevant
literature and identified a need for more accurate risk prediction in CKD. We then
developed and validated laboratory based predictive models that accurately predict
kidney failure, and are easily translated to the bedside. Our future efforts are focused
on more widespread external validation and demonstration of the clinical utility of
these models using decision analyses and cluster randomized
Thesis (Ph.D.)--Tufts University, 2013.
Submitted to the Dept. of Clinical & Translational Science.
Advisor: David Kent.
Committee: David Kent, Andrew Levey, Lesley Inker, David Naimark, and John Griffith.
Keyword: Epidemiology.read less