Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Models for earlier prognosis of renal decline in polycystic kidney disease (PKD) patients

Shi, Tiange (2018) Models for earlier prognosis of renal decline in polycystic kidney disease (PKD) patients. Master's Thesis, University of Pittsburgh. (Unpublished)

Submitted Version

Download (822kB) | Preview


Polycystic kidney disease (PKD) is a genetic condition that leads to increased formation and growth of kidney cysts, and thus may lead to rapid onset of end-stage renal disease (ESRD). In 2017, more than 15% adults in the US were estimated to have inherited PKD. About half of PKD patients require dialysis or renal replacement therapy by 60 years of age. Prior to these terminal outcomes, chronic kidney disease (CKD), which is defined as a progressive loss of kidney function, represents the primary outcome of interest for PKD patients. Treatments of CKD are currently being developed, bur need to be administered earlier in the process. Therefore, earlier prognosis of PKD patients at risk for renal decline provides an opportunity to prevent or delay the progression of ESRD and decrease morbidity and mortality.
In this study, CKD stage 3B with glomerular filtration rate (GFR) less than 45 ml/min was considered the endpoint of highest clinical interest since stage 3B is early enough to identify patients before rapid decline, but late enough to represent a clinically meaningful outcome. We evaluated earlier prognostic ability of factors available at birth for CKD stage 3B among currently healthy PKD patients; use of only factors available at birth is a novel approach and could lead to early identification of PKD patients who subsequently experience clinical outcomes (e.g. later stages of CKD or ESRD).
Training data were collected from the Consortium for Radiologic Imaging Studies of Chronic Kidney Disease (CRISP). Multivariable logistic regression was initially employed to predict renal decline. A pruned classification tree model showed similar prognostic ability as logistic regression based on overlapping 10-fold cross validation AUC confidence intervals. Random forests, however, showed significant improvement in prognostic ability.
This study also validated results using a completed clinical trial of similar PKD patients (the HALT Progression of Polycystic Kidney Disease Study). Both CRISP cross validation and HALT validation results agreed on the best model (random forests) for prognostic ability.
In terms of public health significance, random forests could help estimate the probability of PKD patients reaching renal failure at given age, and thus inform prevention efforts.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Shi, Tiangetis42@pitt.edutis42
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairLandsittel,
Committee MemberBandos,
Committee MemberYouk,
Date: 28 June 2018
Date Type: Publication
Defense Date: 13 April 2018
Approval Date: 28 June 2018
Submission Date: 5 April 2018
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 80
Institution: University of Pittsburgh
Schools and Programs: School of Public Health > Biostatistics
Degree: MS - Master of Science
Thesis Type: Master's Thesis
Refereed: Yes
Uncontrolled Keywords: logistic regression; classification tree; random forests; ploycystic kidney disease; prognosis; chronic kidney disease; PKD; CKD
Date Deposited: 28 Jun 2018 20:13
Last Modified: 28 Jun 2018 20:13


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item