Separating Albuminuric Groups in Diabetic Nephropathy

• Activity Number: 10
• Type: Invited
• Date/Time: Sunday, August 3, 2003 : 2:00 PM to 3:50 PM
• Sponsor: ASA, San Francisco Chapter
• Author(s): Richard A. Olshen*+ and Derek B. Boothroyd and Kevin V. Lemley and Robert G. Nelson and Bryan Myers
• Companies: Stanford University and Stanford University and Stanford University and NIDDK and Stanford University
• Keywords: glomerular filtration ; Hotelling's T2 ; smoothing splines ; random coefficient regression splines ; sample reuse methods

Abstract

The data focus upon glomerular filtration (GFR) of Arizona Pima Indians with type 2 diabetes. Decline of GFR to sufficiently low levels entails dialysis or transplantation. Many Pimas with type 2 diabetes excrete albumin through urine. The excretion begins at times that vary by individual, from about seven years past the onset of diabetes to more than 20 years. Those who excrete large amounts of albumin are macroalbuminurics (macros); those who excrete less are microalbuminurics (micros). Many of those screened as micros progress, becoming macros. We followed GFR for at least four years for three groups: new macros (22), progressors (13), and micros (13). We wished to know GFR at a regular grid of points; however, measurements were made irregularly. One goal was to define a regression B-spline for each subject that begins six years after onset of diabetes. Coefficients are random by subject. Groups were compared by imputed values on the regular grid via suitable permutation distributions of Hotelling's T2. Various tools were adapted to our purpose: smoothing splines, estimation of the covariance structure of the coefficients, and estimation of subject-specific spline coefficents.