Reliability of two indices of the biologic variability in glycosylation among children and adolescen

1
Reliability of two indices of the biologic
variability in glycosylation among children and
adolescents with T1DM
Darrell M Wilson1, Rosanna Fiallo-Scharer2,
Dongyuan Xing3, Timothy Wysocki4, Jennifer
Block1, Stu Weinzimer5, Craig Kollman3, Roy
Beck3, Katrina Ruedy3, William Tamborlane5, and
the Diabetes Research in Children Network
(DirecNet) Study Group. 1Stanford, CA 2Denver,
CO 3Tampa, FL 4Jacksonville, FL 5New Haven, CT
2
ABSTRACT
Individuals appear to glycosylate proteins at
different rates. Among subjects with diabetes,
higher glycosylation rates may predict increased
risk for long-term complications independently of
glycemic control. A DirecNet study of 200
children (mean age 12.5 yrs, 54 male) allowed us
to test the reliability of two indices of
biologic variability of glycosylation. A1c was
measured (DCA2000) after 3 days of CGMS use at 0,
3, and 6 m. Home glucose meter measurements
(HGM) were downloaded for the 12 wks preceding
A1c values at 3 and 6 m. We calculated the
glucose area under the curve (AUC) using the
trapezoidal rule over 3 days from CGMS data
(AUC-CGMS) and 12 wks from HGM data (AUC-meter).
Analysis was restricted to the 123 subjects who
averaged 4 daily HGMs for each of the 6 m. The
hemoglobin glycation index (HGI observed A1c
predicted A1c) was calculated as the residual
from a repeated measures regression model of A1c
versus each of the glucose AUC controlling for
age, insulin modality and clinical site. A second
measure, the glycosylation index (GI) was
calculated as A1c divided by the preceding AUC.
An individuals HGI values at the 3 time points
were well correlated (r 0.56 to 0.62). The GI
values were well correlated only when the
AUC-meter, and not when the AUC-CGMS were used.
These finding are consistent with HGI being a
reasonable reliable measure of individual
variability in glycosylation. While current A1c
measurements are quite robust, each glucose AUC
method has problems. The CGMS provides only 3
days of glucose data that is less accurate than
meters. The AUC-meter can be biased if
pre-prandial measurements predominate. Long-term
glucose sensors are needed to calculate the
unbiased estimate of the glucose AUC over weeks
required to fully assess the biologic variability
in glycosylation.
3
INTRODUCTION

• Individuals (both with and without diabetes) with
  the same average glucose have different A1c
  levels.
• This implies that subjects may glycosylate
  proteins at inherently different rates which are
  stable over time.
• An individuals rate of glycosylation may be
  associated with likelihood of developing the
  microvascular complications associated with
  diabetes independent of the overall level of
  glycemic control.

4
PURPOSE

• We tested the reliability of two different
  measures of the rate of glycosylation in 200
  children and adolescents with Type 1 diabetes
  over 6 months.

5
METHODS

• We used data from DirecNets recently published
  study of the GlucoWatch G2 Biographer (Cygnus
  Inc, Redwood City, CA)
• This home environment, six month, randomized,
  clinical trial involved 200 children with Type 1
  diabetes at 5 clinic sites
• Because the study did not impact hemoglobin A1c
  (A1c), the two treatment groups were combined for
  these analyses

6
Participants

• Eligibility
• Age 7 to 17 years
• Type 1 diabetes with use of insulin for at least
  1 yr
• HbA1c between 7.0 and 11.0 inclusive
• Stable insulin regimen (either a pump or at least
  2 injections per day) for the two months prior
• Exclusion Criteria
• Prior use of a GlucoWatch
• Current or past use of corticosteroids in the
  last 6 months
• Cystic fibrosis or another chronic illness

7
Glycemic Control

• Evaluation of glycemic control, done at 0, 3, and
  6 months, included
• A1c measured by DCA 2000
• 48 to 72 hr CGMS (Medtronic MiniMed, Northridge,
  CA) profile
• 8-point self-monitored blood glucose (SMBG)
  (before and 2 hours after each meal, bedtime and
  3 AM) for at least two days using OneTouch
  UltraSmart (Lifescan, Inc, Milpitas, CA)
• All subjects were given a computer for download
  of essentially all of their SMBG data (available
  prior to only the 3 and 6 month visits)

8
Glucose AUCs

• Glucose area under the curves (AUC) were
  calculated using
• 48 to 72 hr CGMS data immediately prior to the
  A1c at 0, 3, and 6 month visits
• 8-point SMBG data immediately prior to the A1c at
  0, 3, and 6 month visits using the trapezoidal
  method
• Essentially all of their SMBG data for 3 months
  prior to the 3 and 6 month visits using the
  trapezoidal method

9
Glucose AUC Example 8 point testing versus CGMS
Area under the curve (AUC) for 8-point testing
(blue triangles) depicted by the yellow region.
CGMS Tracing
10
Glucose AUC Example 8 point testing with CGMS
AUC for CGMS on the same day depicted by the
green region. (8-point testing data shown as blue
triangles)
11
Rates Of GlycosylationCalculations

• Hemoglobin glycation index (HGI)
• Calculated as the observed A1c predicted A1c
  (the residual from a repeated measures regression
  model of A1c versus each of the glucose AUC
  controlling for age, insulin modality and
  clinical site).
• Glycosylation index (GI)
• Calculated as A1c divided by each of the
  preceding AUC.

12
RESULTS200 Subjects at Baseline
13
Glucose AUC Completion

• 123 (62) of subjects averaged at least 4
  glucoses per day on the SMBG downloads
• Results were very similar for the whole group
  when compared with these 123
• 113 (57) of subjects had at least 75 of the
  glucoses requested for the 3 days of 8 point
  testing at baseline, 64 (32) at 3 months and 59
  (30) at 6 months
• 192 (96) of subjects had at least 48 hr of CGMS
  data at baseline, 158 (79) at 3 months and 165
  (83) at 6 months

14
Summary of AUC Glucose Values
Data presented as mean SD
15
Summary of GI Values
Data presented as mean SD
16
Summary of HGI Values
Data presented as mean SD
17
GI and HGI Across Time Using Glucose AUC by CGMS
(r values)
The correlations across the 3 visits are much
higher for the HGI (purple highlight) than for
the GI (orange highlight), although even the
highest r value is only 0.65. The correlation
between GH and HGI is only fair at the same visit
(green highlight), and much weaker at across the
different visits. BOLD indicates p lt 0.05.
18
Baseline A1c vs CGMS AUC
19
HGICGMS Baseline vs 6 Months
20
GI and HGI Across Time Using Glucose AUC by 8 Pt
(r values)
The correlations across the 3 visits are much
higher for the HGI (purple highlight) than for
the GI (orange highlight), although even the
highest r value is only 0.64. The correlation
between GH and HGI is only fair at the same visit
(green highlight), and much weaker at across the
different visits. BOLD indicates p lt 0.05.
21
GI and HGI Across Time Using Glucose AUC by SMBG
(r values)
The correlations across the 2 visits are very
similar for the HGI (purple highlight) and the GI
(orange highlight). The correlation between GH
and HGI is fair at the same visit (green
highlight), and weaker at across the different
visits. BOLD indicates p lt 0.05.
22
CONCLUSIONS

• Across the 6 month duration of the study,
  individual subjects had reasonably stable
  glycosylation indices
• This supports the hypothesis that individuals
  have inherently different and consistent rates of
  protein glycosylation
• Regardless of glucose detection method used
  (CGMS, 8 point testing, routine SBGM) to
  estimate the glucose AUC, the HGI was more
  repeatable than the GI
• Unfortunately, the HGI is computationally much
  more difficult than the GI

23
DISCUSSION

• Each glucose detection method used to estimate
  the glucose AUC has difficulties
• CGMS provided 3 days of moderately accurate data
• 8 point testing, while technologically accurate,
  captures only a small portion of the day
• Routine SBGM over 12 weeks tends to be biased
  toward pre-prandial glucose measurements
• Better estimates of an individual glycosylation
  rate will be possible with more accurate longer
  term glucose sensors

24

• Stanford University
• Bruce Buckingham
• Darrell Wilson
• Jennifer Block
• Paula Clinton
• Yale University
• William Tamborlane
• Stuart Weinzimer
• Elizabeth Doyle
• Kristen Sikes
• Amy Steffen
• Jaeb Center for Health Research
• Roy Beck
• Katrina Ruedy
• Craig Kollman
• Dongyuan Xing

• Barbara Davis Center
• H. Peter Chase
• Rosanna Fiallo-Scharer
• Jennifer Fisher
• Barbara Tallant
• University of Iowa
• Eva Tsalikian
• Michael Tansey
• Linda Larson
• Julie Coffey
• Nemours Childrens Clinic
• Tim Wysocki
• Nelly Mauras
• Larry Fox
• Keisha Bird
• Kelly Lofton

Supported by NIH Grants HD041919, HD041915,
HD041890, HD041918, HD041908, HD041906, RR00069,
RR00059, RR06022, RR00070-41, and the Nemours
Research Programs.