The TrialNet Natural History Study of the Development of Type 1 Diabetes: objectives, design, and initial results

Article (PDF Available)inPediatric Diabetes 10(2):97-104 · October 2008with140 Reads
DOI: 10.1111/j.1399-5448.2008.00464.x · Source: PubMed
Abstract
TrialNet's goal to test preventions for type 1 diabetes has created an opportunity to gain new insights into the natural history of pre-type 1 diabetes. The TrialNet Natural History Study (NHS) will assess the predictive value of existing and novel risk markers for type 1 diabetes and will find subjects for prevention trials. The NHS is a three-phase, prospective cohort study. In phase 1 (screening), pancreatic autoantibodies (glutamic acid decarboxylase, insulin, ICA-512, and islet cell antibodies) are measured. Phase 2 (baseline risk assessment) includes oral glucose tolerance tests (OGTTs) in antibody-positive subjects and estimation of 5-yr diabetes risks according to the OGTT and number of confirmed positive antibody tests. Phase 3 (follow-up risk assessments) requires OGTTs every 6 months. In phases 2 and 3, samples are collected for future tests of T-lymphocyte function, autoantibody isotypes, RNA gene expression, and proteomics. The primary outcome is diabetes onset. Of 12 636 relatives screened between March 2004 and December 2006, 605 (4.8%) were positive for at least one biochemical antibody. Of these, 322 were confirmed antibody positive and completed phase 2, of whom 296 subjects were given preliminary 5-yr diabetes risks of <25% (n = 132), > or =25% (n = 36), and > or =50% (n = 128) where the latter two categories represent different subjects based on number of confirmed positive antibodies (2, > or =25%; 3 or more, > or =50%) and/or an abnormal OGTT (> or =50%). The NHS is identifying potential prevention trial subjects and is assembling a large cohort that will provide new natural history information about pre-type 1 diabetes. Follow-up to diabetes will help establish the biological significance and clinical value of novel type 1 diabetes risk markers.

Figures

Figure
Figure
Figure

Full-text (PDF)

Available from: Clinton Thompson
Pediatric Diabetes 2009: 10: 97–104
doi: 10.1111/j.1399-5448.2008.00464.x
All rights reserved
#2008 The Authors
Journal compilation #2008 Blackwell Munksgaard
Pediatric Diabetes
Original Article
The TrialNet Natural History Study of the
Development of Type 1 Diabetes: objectives,
design, and initial results
Mahon JL, Sosenko JM, Rafkin-Mervis L, Krause-Steinrauf H, Lachin
JM, Thompson C, Bingley PJ, Bonifacio E, Palmer JP, Eisenbarth GS,
Wolfsdorf J, Skyler JS, for the TrialNet Natural History Committee and
Type 1 Diabetes TrialNet Study Group. The TrialNet Natural History
Study of the Development of Type 1 Diabetes: objectives, design, and
initial results.
Pediatric Diabetes 2009: 10: 97–104.
Objectives: TrialNet’s goal to test preventions for type 1 diabetes has
created an opportunity to gain new insights into the natural history of
pre-type 1 diabetes. The TrialNet Natural History Study (NHS) will
assess the predictive value of existing and novel risk markers for type 1
diabetes and will find subjects for prevention trials.
Research design and methods: The NHS is a three-phase, prospective
cohort study. In phase 1 (screening), pancreatic autoantibodies (glutamic
acid decarboxylase, insulin, ICA-512, and islet cell antibodies) are
measured. Phase 2 (baseline risk assessment) includes oral glucose
tolerance tests (OGTTs) in antibody-positive subjects and estimation of
5-yr diabetes risks according to the OGTT and number of confirmed
positive antibody tests. Phase 3 (follow-up risk assessments) requires
OGTTs every 6 months. In phases 2 and 3, samples are collected for
future tests of T-lymphocyte function, autoantibody isotypes, RNA gene
expression, and proteomics. The primary outcome is diabetes onset.
Results: Of 12 636 relatives screened between March 2004 and December
2006, 605 (4.8%) were positive for at least one biochemical antibody. Of
these, 322 were confirmed antibody positive and completed phase 2, of
whom 296 subjects were given preliminary 5-yr diabetes risks of ,25%
(n ¼132), 25% (n ¼36), and 50% (n ¼128) where the latter two
categories represent different subjects based on number of confirmed
positive antibodies (2, 25%; 3 or more, 50%) and/or an abnormal
OGTT (50%).
Conclusions: The NHS is identifying potential prevention trial subjects
and is assembling a large cohort that will provide new natural history
information about pre-type 1 diabetes. Follow-up to diabetes will help
establish the biological significance and clinical value of novel type 1
diabetes risk markers.
Jeffrey L Mahon
a
,
Jay M Sosenko
b
,
Lisa Rafkin-Mervis
b
,
Heidi Krause-Steinrauf
c
,
John M Lachin
c
,
Clinton Thompson
c
,
Polly J Bingley
d
,
Ezio Bonifacio
e
,
Jerry P Palmer
f
,
George S Eisenbarth
g
,
Joseph Wolfsdorf
h
and
Jay S Skyler
b
, for the TrialNet
Natural History Committee*
and Type 1 Diabetes TrialNet
Study Group
a
Division of Endocrinology and
Metabolism, University of Western
Ontario, London, ON, Canada;
b
Division
of Endocrinology, University of Miami,
Miami, FL, USA;
c
The Biostatistics Center,
George Washington University,
Washington, DC, USA;
d
Division of
Endocrinology and Metabolism,
University of Bristol, Bristol, UK;
e
Center
for Regenerative Therapies, Dresden
University of Technology, Dresden,
Germany;
f
Department of Medicine,
University of Washington, Seattle, WA,
USA;
g
Barbara Davis Center for
Childhood Diabetes, University of
Colorado, Denver, CO, USA; and
h
Division of Endocrinology, Children’s
Hospital Boston, Harvard Medical
School, Boston, MA, USA
Key words: pre-type 1 diabetes – natural
history
Corresponding author:
Jeffrey Mahon, MD, MSc
Room E5-16, Robarts Clinical Trials
P.O. Box 5015, 100 Perth Drive
London, ON
Canada N6A 5K8.
Tel: 519-663-3756;
fax: 519-663-3211;
e-mail: jl.mahon@lhsc.on.ca
97
Submitted 6 February 2008. Accepted
for publication 30 July 2008
*Other Natural History Committee
members are J Barker, D Becker, S
Bennett-Johnson, P Gottlieb, S Harris,
E Leschek, K McCants, G Navidi,
A Pugliese, M Rewers, K Riley, D Schatz,
M Siegelman, D Wilson and A Ziegler.
Recognition that type 1 diabetes is caused by immune-
mediated loss of pancreatic b-cells has prompted
a search for interventions that protect b-cells and
improve long-term clinical outcomes (1, 2). The
feasibility of studies testing this strategy has been
proven in persons at risk for type 1 diabetes (prevention
trials) (3–5) and in persons with type 1 diabetes and
residual b-cell function (intervention trials) (6, 7). For
these reasons, the National Institutes of Health, in
cooperation with the American Diabetes Association
(ADA) and Juvenile Diabetes Research Foundation,
has formed Type 1 Diabetes – TrialNet to perform
further prevention and intervention trials (8). In
conducting prevention trials, TrialNet can also obtain
new information about the natural history of pre-type 1
diabetes in relatives of persons with the disease. We
describe in this study the aims, design, and initial results
of the TrialNet Natural History Study of the Develop-
ment of Type 1 Diabetes (NHS).
TrialNet is based on the Diabetes Prevention Trial-
Type 1 (DPT-1), which screened more than 100 000
relatives for islet cell antibodies (ICA). ICA-positive
subjects underwent diabetes risk assessments based
on oral and intravenous glucose tolerance tests
(IVGTTs) and were then entered into prevention
trials (3, 5). In planning the NHS, TrialNet used data
from a DPT-1 ancillary study (9), which showed that
biochemical antibodies [glutamic acid decarboxylase
antibodies (GADA), insulin autoantibodies (mIAA),
and ICA-512 antibodies (ICA512A)/islet antigen-2
antibodies] can replace the more labor intensive and
observer-dependent ICA assay as the first screening test.
We designed the NHS to address three main aims.
First, we will validate the predictive accuracy of a new
diabetes risk algorithm that uses biochemical anti-
bodies rather than ICA as the first screening test.
Second, we will collect samples for future assays of
novel type 1 diabetes risk markers. The specimens
(termed Ômechanistic samples’) are being collected and
stored according to procedures established in collab-
oration with the Immune Tolerance Network (10).
Expected uses include frozen peripheral blood mono-
cytes for T-lymphocyte functional assays, serum for
antibody isotypes, RNA for gene expression studies, and
plasma for proteomic studies. Third, the NHS will
identify subjects for TrialNet prevention and interven-
tion trials.
Methods
The protocol is available online (11). We based our
definition of an NHS on standard sources (12–15) and
are using a prospective cohort design divided into three
phases: screening (phase 1), baseline risk assessment
(phase 2), and follow-up risk assessments (phase 3).
Sites in participating countries (USA, Canada, UK,
Germany, Italy, Australia, and New Zealand) have
local institutional review board approval and appro-
priate assurance with the Office for Human Research
Protections. All test procedures and assays, including
ICA (16) and biochemical antibodies (17) and the cut-
points to define positive antibody tests, use established
methods. In the 2005 Diabetes Antibody Standardiza-
tion Program (DASP), the TrialNet biochemical
antibody laboratory had respective sensitivities and
specificities of 76 and 99% for GADA, 64 and 100% for
ICA512A, and 58 and 99% for mIAA. In the 1998
Combinatorial Islet Antibody Workshop, the TrialNet
ICA screening laboratory showed a sensitivity of 81%
and specificity of 96%.
Figure 1 shows phase 1 antibody testing procedures.
Inclusion criteria include age 1–45 yr and a relative
with type 1 diabetes (first-degree relatives and second-
degree or third-degree relatives for subjects less than
20 yr old). Type 1 diabetes in probands requires
diabetes onset before age 40 yr and prescription of
insulin within a year of diagnosis. The requirement for
insulin therapy within 1 yr in the proband reflects a com-
promise between a shorter interval (e.g., 3–6 months)
that will be more specific for type 1 diabetes, but that is
also more restrictive because it will disqualify some
probands who have type 1 diabetes, and a longer interval
that increases the chance for enrolling subjects with
probands who clearly have type 2 diabetes. It is also
consistent with the time to start of insulin therapy in
probands previously used in the DPT-1 prevention
trials (3, 5).
Subjects with at least two positive tests for any one of
the four antibodies in the phase 1 samples are eligible
for phase 2 (Fig. 1). Subjects who are discordant for
a specific antibody on the first two samples (e.g.,
GADA positive on the first test but GADA negative on
the second test) are asked to provide a third sample to
resolve their baseline antibody status. Thus, if two out
of three tests are positive for GADA, the subject is
98 Pediatric Diabetes 2009: 10: 97–104
Mahon et al.
defined as Ôconfirmed GADA positive’. Other phase 2
tests include an oral glucose tolerance test (OGTT),
hemoglobin A
1
C (A1C) level, and human leukocyte
antigen (HLA) typing. Subjects with protective type 1
diabetes genotypes are not excluded. The OGTT
includes samples for glucose levels at 0, 30, 60, 90, and
120 min. Insulin and C-peptide levels are also being
measured at these times for, respectively, estimates of
insulin sensitivity by the homeostatic model assessment –
insulin resistance and endogenous insulin secretion.
At completion of phase 2, glucose levels during the
OGTT and the number of confirmed positive anti-
bodies are being used to assign subjects to preliminary
5-yr diabetes risk categories of 50, 25, or ,25%
(Fig. 3) (see also Protocol modifications below). We are
not masking subjects or investigators with respect to
projected diabetes risks.
Subjects completing phase 2 are eligible for phase 3
and are asked to return every 6 months for an OGTT,
A1C level, and antibody tests. We are also enrolling
former DPT-1 subjects who have not developed
diabetes into phase 3 to extend observations on the
long-term risk for diabetes in that cohort. At all phases,
residual blood samples including DNA are being stored
at a TrialNet or NIH repository laboratory. The mech-
anistic samples are being collected in phases 2 and 3.
Protocol modifications
Two protocol changes (effective from February 2007)
have been made since the NHS was implemented in
February 2004. First, the antibody criteria determining
phase 2 eligibility were modified. Whereas the previous
definition required that at least one specific antibody be
positive on two separate tests, the new definition (see
above) was made to match antibody criteria that are
determining subject eligibility for the TrialNet ran-
domized trial of oral insulin to prevent type 1 diabetes
that is now underway (www.diabetestrialnet.org, ac-
cessed 12 November 2007). We also projected that the
risk for diabetes in persons with at least two positive
tests for any antibodies warranted closer follow-up in
phase 3. In the Results below, the original antibody
criteria defining eligibility for phase 2 are reported.
Second, the original risk algorithm assigned subjects
to 5-yr diabetes estimates of ,25, 25–50, or .50%
based on results of the OGTT, number of confirmed
positive antibodies, and, in selected cases, an impaired
first-phase insulin response (FPIR) during the IVGTT.
The FPIR was used in subjects with a normal OGTT
and exactly two confirmed positive antibodies to
discriminate between 5-yr diabetes risks of 25–50 or
.50%. However, we found that few (10%) subjects
needed the FPIR for risk stratification and that the
First sample
GADA, ICA512A, mIAA (ICA tested if 1 other antibody is positive)
<18 yr old
Second sample
GADA, ICA512A, mIAA, ICA
Rescreen yearly until
18 yr old
Not eligible
Eligible for phase 2 (baseline risk assessment)
Annual rescreening
for 2 yr
1 antibody positive No antibodies positive
No antibodies positive
in 2nd sample
2 antibodies positive
18 yr old
1 antibody positive
in 2nd sample
Fig. 1. Phase 1 antibody screening procedure. Subjects with at least two positive antibodies on the first sample can proceed directly to phase 2
or can provide a second sample in phase 1 before entering phase 2. ICA positivity is defined by values 10 Juvenile Diabetes Foundation units
(16). Definitions for GADA, ICA512A, and mIAA positivity are based on values exceeding the 99% in normal controls (17). GADA, glutamic
acid decarboxylase antibodies; ICA, islet cell antibodies; ICA512A, antibodies to ICA-512; mIAA, insulin autoantibodies.
Pediatric Diabetes 2009: 10: 97–104 99
TrialNet pre-type 1 diabetes study
projected power to assess the added predictive value
of the FPIR over other risk markers would remain
low even upon screening thousands of subjects. Other
DPT-1 analyses also suggested that the FPIR added
little additional predictive accuracy over glucose
and C-peptide levels obtained during the OGTT in
antibody-positive relatives (18). Given this, and the
burden to subjects in performing the IVGTT, we
omitted the test from the NHS. As a result, using the
FPIR to refine risk beyond a base estimate of 25% in
subjects with exactly two confirmed positive antibodies
was no longer possible. We therefore altered the 5-yr
risk categories to 50, 25, or ,25% where the risk
categories of 50 and 25% identify different subjects
based on the number of confirmed positive antibodies
(three or more antibodies for subjects in the 50% group
and two antibodies for subjects in the 25% group)
(Fig. 3). In the Results below, we have reclassified
subjects according to the modified risk algorithm.
Statistical analysis and sample size estimation
The primary outcome is diabetes mellitus by ADA
criteria (19). The prevalence of each risk category will
be described as a function of subject characteristics.
Cumulative diabetes incidences and hazard rates will be
assessed within risk categories. Covariate effects on
diabetes risks (e.g., age, gender, and the time to starting
insulin in the proband) will be analyzed by proportional
hazards regression models (20). Formal analyses will be
conducted after 250 subjects in the 50% risk category
have developed diabetes. This will provide a 95%
confidence band of (55 and 64%) for a cumulative
incidence of 60% and will also provide 85% power to
detect a 30% difference in the hazard rate between two
equal-sized groups for a binary covariate. We will not
necessarily exclude prevention trial subjects from
natural history analyses, in part because subjects
allocated to untreated control arms in prevention trials
may still contribute unbiased natural history informa-
tion. To test this assumption, we will compare diabetes
risk between subjects who are eligible and randomized
to control groups with the risk in subjects who are
eligible but who do not enter a prevention trial.
Results
We report in this study the results for 12 636 TrialNet
subjects who entered phase 1 between March 2004 and
January 2006 and for 322 subjects who entered phase 2
by 31 December 2006 (Fig. 2). By December 2006, 219
TrialNet subjects and 54 former DPT-1 subjects had
entered phase 3, of whom 7 and 2 subjects, respectively,
developed diabetes. The time for follow-up, and there-
fore number of new diabetes cases, is insufficient to draw
meaningful conclusions about risk prediction. Diabetes
outcomes will be reported in future publications.
Table 1 shows baseline characteristics of phase 1 and
2 subjects. More than 90% had at least one first-degree
relative with T1D, and 55% were less than 21 yr old.
Most subjects (95%) were ineligible for phase 2 because
they were negative for the biochemical antibodies on
the first phase 1 sample (Fig. 2). Among 605 subjects
with at least one positive biochemical antibody, 72%
met eligibility criteria for phase 2, of whom 74% entered
phase 2 (Fig. 2).
Table 2 shows NHS antibody test results in the first
phase 1 sample and antibody results in a subset of more
than 17 000 subjects screened in DPT-1 (9). Approx-
imately 5% (605/12 636) of NHS subjects had at least
one positive biochemical antibody on their first sample.
As in the DPT-1, GADA was the most frequent positive
antibody (3.6%) on the first test. Most (57%) antibody-
positive subjects had a single positive antibody; this
group comprised 2.8% of all screened subjects.
Confirmation rates for specific antibodies on sub-
sequent testing in the NHS ranged between 74% (ICA)
and 88% (GADA) (Table 2). As expected, a strong
association was seen between the number of positive
antibodies on the first test and confirmation of
12636 relatives screened for biochemical antibodies
605 1 positive biochemical antibody and eligible to provide 2nd
sample
434 antibody eligible for phase 2
322 entered phase 2
8 OGTT inadequate
18 diabetes-range OGTT
132 estimated diabetes risk/5 yr <25%
36 estimated diabetes risk/5 yr 25%
128 estimated diabetes risk/5 yr 50%
171 did not continue
138 failed to confirm as antibody positive
21 withdrew consent or lost to follow-up
12 developed diabetes before entry to phase 2
112 not enrolled into phase 2 as of December 2006:
12 withdrew consent
100 phase 2 visit pending
12031 not eligible to continue for one or more of
15 inadequate blood collection
33 diabetes at time of screening
239 non-qualifying proband
33 outside age range
11 withdrew consent
12016 biochemical antibody negative
Fig. 2. Enrollment of subjects from March 2004 to December 2006.
Criteria for antibody eligibility for phase 2 are based on the
definition used before February 2007 (see Methods). ÔLost to follow-
up’ refers to subjects who failed to return for further testing despite
repeated attempts to contact them or their caregiver. OGTT, oral
glucose tolerance test.
100 Pediatric Diabetes 2009: 10: 97–104
Mahon et al.
antibody positivity on repeat testing. Thus, 98% of
subjects with at least two positive antibodies on the first
test were positive for at least one antibody on sub-
sequent testing, whereas 74% of subjects with only one
positive antibody on the first test were confirmed
positive on subsequent testing (Table 2). In subjects
with a single positive biochemical antibody test on the
first sample, failure to confirm antibody positivity for
the specific biochemical antibody on subsequent testing
was associated with lower antibody levels on the first
test. Thus, the median GADA value among 44 subjects
positive on the first test but who did not confirm
GADA positive on a subsequent test was 0.061 vs. a
median GADA value of 0.300 in 318 subjects who sub-
sequently confirmed positive for GADA (p ,0.0001,
Wilcoxon two-sample test). Corresponding values for
ICA512A were 0.067 in 21 subjects who did not confirm
positive for ICA512A vs. 0.672 in 124 subjects who did
(p ,0.0001); for mIAA, the values were 0.018 in 40
subjects not confirming positive vs. 0.041 in 117
subjects who did (p ¼0.0002).
Of the 322 subjects enrolled in phase 2, 314 completed
the OGTT and 8 subjects’ OGTT were inadequate (e.g.,
samples hemolyzed, missing samples at 0 or 120 min).
In the 314 subjects with an adequate OGTT, the
projected 5-yr diabetes risks were ,25% (n ¼132),
25% (n ¼36), and 50% (n ¼128) based on the
modified risk algorithm (Fig. 3). Eighteen subjects had
an abnormal OGTT consistent with diabetes. The
OGTT results in the remaining subjects were normal
(n ¼217) or showed impaired glucose tolerance (IGT)
(n ¼46), impaired fasting glucose (IFG) (n ¼22), or
IGT and IFG (n ¼11). Subjects with an abnormal
OGTT including diabetes had a higher body mass index
(BMI) compared with subjects with normal OGTTs
(respective mean BMIs: 23.9 vs. 21.8 kg/m
2
,p ¼0.023,
Student’s t-test).
Discussion
Natural history studies in antibody-positive relatives of
persons with type 1 diabetes are neither new nor rare,
beginning with the Bart’s–Windsor study in 1978 that
showed ICA predicted diabetes (21). Since then, more
than 250 natural history studies have been reported
based on a literature scan for a systematic review of
natural history studies that we are currently undertak-
ing. These studies have defined and validated markers
Table 1. Phase 1 and phase 2 subjects’ characteristics
Phase 1
(n ¼12 636)
Phase 2
(n ¼322)
Mean age (SD) (yr) 21.8 (14.8) 19.4 (14.3)
Age range (yr) (%)
1–5 14 15
6–10 19 26
11–15 15 16
16–20 7 6
21–25 2 2
26–45 43 35
Female (%) 59 58
Race (%)
White 88 90
African-American 3 2
Asian 1 2
Others 9 6
Ethnicity (%)
Hispanic/Latino 13 9
Others 87 91
Family history of type 1 diabetes* (%)
First-degree relative 91 94
Second-degree relative 17 18
Third-degree relative 9 8
One affected relative 77 73
Two or more affected
relatives
20 27
HLA DQB1*0602 (%) Not done 5
*Subjects could have two or more relatives, of different
degrees, with type 1 diabetes.
Table 2. Antibody test results
NHS
(n ¼12 636)
DPT-1* (9)
(n ¼17 207)
Number of subjects with a positive Ab test on
first sample, n (%)
One Ab 347 (2.8) 1009 (5.9)
Two Abs 125 (1.0) 193 (1.1)
Three Abs 91 (0.7) 147 (0.8)
Four Abs 42 (0.3) 54 (0.3)
Number of subjects positive by specific Abs on
first sample, n (%)
GADA 452 (3.6) 688 (4.0)
ICA512A 205 (1.6) 276 (1.6)
mIAA 203 (1.6) 437 (2.5)
ICA* 178 651 (3.8)
Number of subjects positive
for one or more biochemical
Ab on first sample, n (%)
605 (4.8) 1076 (6.2)
Confirmation rates by Ab type, n (%)
GADA 318/362 (88)
ICA512A 124/145 (86)
mIAA 117/157 (75)
ICA 102/139 (74)
Confirmation rates by number of positive Abs on first
sample, n (%)
One Ab positive 211/284 (74)
Two Abs positive 91/94 (97)
Three Abs positive 67/67 (100)
Four Abs positive 33/33 (100)
Two or more Abs positive 191/194 (98)
Ab, antibody; DPT-1, Diabetes Prevention Trial-Type 1;
GADA, glutamic acid decarboxylase antibodies; ICA, islet
cell antibodies; ICA512A, antibodies to ICA-512; mIAA,
insulin autoantibodies.
*ICA is only tested in TrialNet subjects with at least one
positive biochemical antibody but was tested in all DPT-1
subjects irrespective of biochemical antibody results.
Based on up to three samples being tested for anti-
bodies in phase 1 and phase 2 more than 1 yr, where
a confirmed positive antibody required that a specific
antibody be positive on at least two samples. Subjects
could be confirmed positive for more than one antibody.
Pediatric Diabetes 2009: 10: 97–104 101
TrialNet pre-type 1 diabetes study
of future type 1 diabetes to the point where definitive
prevention trials have become possible (3–5).
Despite this, important natural history questions
remain unanswered. For example, the capacity of T-
lymphocyte functional assays to predict future type 1
diabetes, and thus the exact role T-lymphocytes play in
human b-cell loss, is unclear beyond general agreement
that they are centrally important (22). Similarly, the
impact of insulin sensitivity on progression to type 1
diabetes is uncertain (23–26). Answers to these and
other questions will help TrialNet develop type 1
diabetes preventions in two ways. First, better under-
standing of the natural history of pre-type 1 diabetes
can direct decisions about treatments worth testing.
For example, testing interventions that improve insulin
sensitivity, of which there are now several in routine use
in clinical practice in patients with type 2 diabetes, will
become attractive if insulin resistance is confirmed to
increase the risk for future type 1 diabetes over other
readily available risk markers. Second, the efficiency of
clinical trials can be improved through more accurate
diabetes prediction in subjects allocated to control
groups within trials. Additional justification for the
NHS arises from earlier detection of type 1 diabetes
through antibody screening and, with that, reduction in
risk for diabetic ketoacidosis at first presentation (27).
The feasibility of prevention trials depends heavily
upon antibody screening rates and the frequency of
antibody-positive relatives. For example, for each
subject randomized in the DPT-1 trials, some 145
persons were tested for ICA (3, 5). The NHS’s practical
aim of finding subjects for prevention trials therefore
makes the rates of antibody screening and antibody
positivity important determinants of TrialNet’s success.
Till December 2006, the NHS screened 23 951 relatives
for antibodies over 30 months (750 subjects/month). As
a benchmark, the DPT-1 screened 990 subjects/month
over 8 1/3 yr (5). However, expansion of TrialNet to
countries outside USA and the availability in 2007 of the
oral insulin prevention trial as an option for some
antibody-positive NHS subjects have led to an increase
in screening such that the 2007 rate (1340/month)
exceeds that of the DPT-1.
Comparison of antibody positivity rates on the first
screening test between the DPT-1 and the NHS must
consider the major difference in screening procedures:
the DPT-1 screened with ICA (9), whereas the NHS
screens with biochemical antibodies. The difference
between the two studies in rates of single antibody-
positive subjects on the first test is one example of this,
where at least some of the difference in rates (NHS ¼
2.8% vs. DPT-1 ¼5.9%) can be accounted for by
exclusively ICA-positive subjects. This group, which
cannot be ascertained by the NHS screening procedure,
occurred in 1.9% (n ¼327) of screened DPT-1 subjects
(9). However, there is more congruence in rates of
antibody positivity on the first screening test between
the two studies in respect to specific biochemical
antibodies and the number of subjects who are positive
for two or more antibodies (Table 2). The failure to
confirm antibody-positive tests in some 25% of single
biochemical antibody subjects in the NHS was clearly
associated with marginally positive test results on the
first sample. Raising the cut-point to define antibody
positivity would reduce the number of false positives
but comes at the cost of missing individuals who would
confirm positive on two tests. This finding affirms the
value of repeating antibody testing before judging
individuals’ risks or enrolling them in a prevention trial.
The decision to screen persons more than age 25 yr in
the NHS follows from TrialNet’s plan to enter
antibody-positive older relatives into prevention trials.
Abnormal OGTT
(IFG and/or IGT) Normal OGTT
2 Abs confirmed
positive
1 Ab confirmed
positive
50% 5-yr
diabetes risk
25% 5-yr
diabetes risk
<25% 5-yr
diabetes risk
3 Abs confirmed
positive
Fig. 3. Five-year diabetes risk estimates. Confirmed antibody positivity requires that the specific antibody (GADA, ICA512A, mIAA, or
ICA) be positive on at least two separate tests in phase 1. The risk categories of 25 and 50% apply to mutually exclusive subject groups
according to the number of confirmed positive antibodies and presence or absence of an abnormal glucose tolerance test. GADA, glutamic
acid decarboxylase antibodies; ICA, islet cell antibodies; ICA512A, antibodies to ICA-512; IFG, impaired fasting glucose; IGT, impaired
glucose tolerance; mIAA, insulin autoantibodies; OGTT, oral glucose tolerance test.
102 Pediatric Diabetes 2009: 10: 97–104
Mahon et al.
This decision was made despite the fact that screening
older relatives is less cost-effective than screening
younger relatives, where cost-effectiveness refers in this
context to the costs incurred to identify and enroll
enough subjects to detect a treatment effect of interest.
The cost-effectiveness of a trial that includes older
subjects will not be as good because the rates of posi-
tive antibodies, including multiple antibodies, and risk
for future diabetes are lower in older compared with
younger subjects. For example, in the European
Nicotinamide Diabetes Intervention Trial (ENDIT),
more than 80% of subjects progressing to diabetes within
5 yr were less than 25 yr old at screening (28). Thus, at
least three times as many older subjects relative to
younger subjects needed to be screened to find one
person who progressed to diabetes.
However, the cost-effectiveness of a trial in those terms
does not address the larger question of cost-effectiveness
of screening and preventing type 1 diabetes in older
persons. For example, a safe and effective prevention for
type 1 diabetes in subjects between age 25 and 45 yr
could yield sufficient cost savings that justify, if not
completely outweigh, the added costs needed to identify
and treat at-risk older persons. Higher costs to screen
older subjects could also be offset by a prevention that is
more effective in older subjects. This is plausible because
the autoimmune process appears to be less aggressive in
older persons. Although modeling studies can estimate
the cost-effectiveness of different screening strategies
according to age, they depend on important assumptions
about treatment effects. Such assumptions are best
confirmed, or refuted, in randomized trials that test
preventive treatments in subjects across a wide age range.
An important methodological issue in the NHS
extends from our plan to include data from subjects
who enter prevention trials in natural history analyses.
The advantage in combining data across prevention
trials and the NHS cohort is greater power to detect
relationships between risk markers and subsequent
diabetes, including more power to detect independent
relationships between diabetes and multiple risk
markers that may be confounded. This validity of this
approach depends on whether there are differences in
the natural history of subjects who enter a prevention
trial compared with those who do not, where an
important clinical measure of the natural history is
diabetes outcome. A recent Cochrane Systematic
Review has addressed this question more generally,
namely, ÔDo outcomes in subjects who participate in
randomized controlled trials differ compared with
similar subjects who do not participate?’. There were
no differences in outcomes between the two types of
subjects across a variety of different problems and
diseases (29).
However, if real differences in diabetes risk exist
between subjects who enter prevention trials vs. those
who do not, then natural history analyses based on
combined data will yield biased risk estimates. Such
differences can be ascribed to Ôtrial effects’ and include
placebo responses, unmeasured baseline differences in
subjects who choose not to enter a prevention trial
despite fulfilling all other entry criteria, and differ-
ences in procedures to ascertain diabetes between the
NHS and the prevention trials. We have minimized the
latter by matching our approach to diabetes ascer-
tainment in the NHS to the approach that is being used
in current and future prevention trials including the
oral insulin prevention trial. We cannot rule out
placebo effects or unmeasured differences in baseline
characteristics beyond testing for and excluding
heterogeneity in diabetes risks between subjects who
are eligible for but who do not enter a prevention trial
vs. eligible subjects randomized to a trial’s control
group. We also do not exclude a similar analytic
strategy, as has been used in recent natural history
publications from the DPT-1 and ENDIT Study
Groups (28, 30), that include data from subjects
randomized to active treatment arms in prevention
trials but in which it is found that the test therapies do
not affect diabetes risk.
Other large natural history studies, and prevention
trials that will yield important natural history informa-
tion, are now underway including The Environmental
Determinants of Diabetes in the Young (TEDDY) (31),
The Trial to Reduce IDDM in the Genetically at Risk
(TRIGR) (32), and the Finnish IDDM Prediction and
Prevention Project (DIPP) (33). It is useful to place the
TrialNet NHS in the context of those other studies. Each
has a focus on genetically high-risk subjects starting with
a positive family history of type 1 diabetes, although
TEDDY and DIPP are also enrolling subjects without
a family history but with high-risk HLA genotypes.
Important differences include ascertainment of the
cohorts at birth in TEDDY, TRIGR, and DIPP vs.
enrolling subjects after age 1 yr in the NHS; the use of
antibodies as a baseline determinant of eligibility in
the NHS vs. development of antibodies as important
secondary outcomes in TEDDY, TRIGR, and DIPP;
and a focus in TEDDY, DIPP, and TRIGR not shared
by the NHS on environmental factors that may cause b-
cell autoimmunity. These studies have much in common
including overlap among investigators across the study
groups. In aggregate, these and other studies will
generate large, prospectively followed cohorts that will
comprehensively assess the natural history of pre-type 1
diabetes in high-risk persons from birth to middle age.
In conclusion, the TrialNet NHS is identifying sub-
jects for participation in type 1 diabetes prevention
trials and is assembling a large cohort of at-risk persons
that will yield new natural history information about
pre-type 1 diabetes. Follow-up to diabetes onset is
underway and will establish the biological significance
and clinical value of both established and novel type 1
diabetes risk markers.
Pediatric Diabetes 2009: 10: 97–104 103
TrialNet pre-type 1 diabetes study
Acknowledgements
TrialNet is funded by the National Institutes of Health through
the National Institute of Diabetes and Digestive and Kidney
Diseases, National Institute of Allergy and Infectious Diseases,
National Institute of Child Health and Human Development,
and General Clinical Research Centers Program; the Juvenile
Diabetes Research Foundation International; and the ADA.
References
1. ATKINSON MA, EISENBARTH GS. Type 1 diabetes: new
perspectives on disease pathogenesis and treatment.
Lancet 2001: 353: 221–229.
2. GREEENBAUM CJ, HARRISON LC. Guidelines for inter-
vention trials in subjects with newly diagnosed type 1
diabetes. Diabetes 2003: 52: 1059–1065.
3. THE DIABETES PREVENTION TRIAL –TYPE 1STUDY GROUP.
Effects of insulin in relatives of patients with type 1
diabetes mellitus. N Engl J Med 2002: 346: 1685–1691.
4. EUROPEAN NICOTINAMIDE DIABETES INTERVENTION TRIAL
(ENDIT) GROUP. ENDIT: a randomized controlled trial
of intervention before onset of type 1diabetes. Lancet
2004: 363: 925–931.
5. THE DIABETES PREVENTION TRIAL –TYPE 1STUDY GROUP.
Effects of oral insulin in relatives of patients with type 1
diabetes. Diabetes Care 2005: 28: 1068–1076.
6. THE CANADIAN-EUROPEAN RANDOMIZED CONTROL TRIAL
GROUP. Cyclosporine-induced remission of IDDM after
early intervention. Diabetes 1988: 37: 1574–1587.
7. HEROLD KC, HAGOPIAN W, AUGER JA et al. Anti-CD3
monoclonal antibody in new-onset type 1 diabetes
mellitus. N Engl J Med 2002: 346: 1692–1698.
8. SKYLER JS. Prediction, and prevention of type 1
diabetes: progress, problems, and prospects. Clin
Pharmacol Ther 2007: 81: 768–771.
9. KRISCHER JP, CUTHBERTSON DD, YUL et al. Screening
strategies for the identification of multiple antibody-
positive relatives of individuals with type 1 diabetes. J
Clin Endocrinol Metab 2003: 88: 103–108.
10. ROTROSEN D, MATTHEWS JB, BLUESTONE JA. The im-
mune tolerance network: a new paradigm for developing
tolerance-inducing therapies. J Allergy Clin Immunol
2002: 110: 17–23.
11. TYPE 1DIABETES-TRIALNET STUDY GROUP.NaturalHis-
tory Study of the development of type 1 diabetes (Protocol
TN-01). (available from http://www.diabetestrialnet.org/
doc/nhs_pro.pdf). Accessed 17 April 2007.
12. LAST JM ed. A Dictionary of Epidemiology. New York:
Oxford University Press, 2001: 121.
13. SACKETT DL, HAYNES RB, GUYATT GH, TUGWELL P.
Clinical Epidemiology: A Basic Science for Clinical Medi-
cine. Boston, MA: Little, Brown, and Company, 1991.
14. FLETCHER RH, FLETCHER SW, WAGNER EH. Clinical
Epidemiology: The Essentials. Baltimore, MD: Williams
and Wilkins, 1988.
15. ROTHMAN KJ, GREENLAND S. Modern Epidemiology.
Philadelph ia, PA: Lippincott, Wil liams, and Wilkins, 1 998.
16. SCHATZ D, KRISCHER J, HORNE G et al. Islet cell
antibodies predict insulin-dependent diabetes in United
States school age children as powerfully as unaffected
relative. J Clin Invest 1994: 93: 2403–2407.
17. WANG J, MIAO D, BABU S et al. Prevalence of
autoantibody-negative diabetes is not rare at all ages
and increases with older age and obesity. J Clin
Endocrinol Metab 2007: 92: 88–92.
18. SOSENKO JM, KRISCHER JP, PALMER JP et al., FOR THE
DIABETES PREVENTION TRIAL –TYPE 1 (DPT-1) STUDY
GROUP. A risk score for type 1 diabetes derived from
autoantibody-positive participants in the DPT-1. Dia-
betes Care 2008: 31: 528–533.
19. EXPERT COMMITTEE ON THE DIAGNOSIS AND CLASSIFICATION
OF DIABETES MELLITUS. Report of the expert committee
on the diagnosis and classification of diabetes mellitus.
Diabetes Care 1997: 20: 1183–1197.
20. LACHIN JM. Biostatistical Methods: The Assessment of
Relative Risks. New York: John Wiley and Sons, 2000.
21. GORSUCH AN, SPENCER KM, LISTER J et al. Evidence for
a long prediabetic period in type I (insulin-dependent)
diabetes mellitus. Lancet 1981: 2: 1363–1365.
22. ROEP BO. The role of T-cells in the pathogenesis of type
1 diabetes: from cause to cure. Diabetologia 2003: 46:
305–321.
23. FOURLANOS S, NARENDRAN P, BYRNES GB, COLMAN PG,
HARRISON LC. Insulin resistance is a risk factor for
progression to type 1 diabetes. Diabetelogia 2004: 47:
1661–1667.
24. MRENA S, VIRTANEN SM, LAIPPALA P, HANNILA ML,
AKERBLOM HK, KNIP M. Models for predicting type 1
diabetes in siblings of affected children. Diabetes Care
2006: 29: 662–667.
25. XUP, CUTHBERTSON D, GREENBAUM C, PALMER JP,
KRISCHER JP, FOR THE DIABETES PREVENTION TRIAL-TYPE 1
STUDY GROUP. Role of insulin resistance in predicting
progression to type 1 diabetes. Diabetes Care 2007: 30:
2314–2320.
26. BINGLEY PJ, MAHON JL, GALE E, FOR THE ENDIT
STUDY GROUP. Insulin resistance and progression to type
1 diabetes in the European Nicotinamide Diabetes
Intervention Trial (ENDIT). Diabetes Care 2008: 31:
146–150.
27. BARKER JM, GOEHRIG SH, BARRIGA Ketal.FOR THE
DAISY STUDY. Clinical characteristics of children diag-
nosed with type 1 diabetes through intensive screening
and follow-up. Diabetes Care 2004: 27: 1399–1404.
28. BINGLEY PJ, GALE EA; EUROPEAN NICOTINAMIDE DIABE-
TES INTERVENTION TRIAL (ENDIT) GROUP. Progression to
type 1 diabetes in islet cell antibody-positive relatives in
the European Nicotinamide Diabetes Intervention Trial:
the role of additional immune, genetic and metabolic
markers of risk. Diabetologia 2006: 49: 881–890.
29. VIST GE, HAGEN KB, DEVEREAUX PJ, BRYANT D,
KRISTOFFERSEN DT, OXMAN AD. Outcomes of patients
who participate in randomized controlled trials com-
pared to similar patients receiving similar interventions
who do not participate. Cochrane Database Syst Rev
2007: 2: MR000009.
30. SOSENKO JM, PALMER JP, GREENBAUM CJ et al. Patterns
of metabolic progression to type 1 diabetes in the
Diabetes Prevention Trial-Type 1. Diabetes Care 2006:
29: 643–649.
31. HAGOPIAN WA, LERNMARK A, REWERS MJ et al.
TEDDY – The Environmental Determinants of Diabe-
tes in the Young: an observational clinical trial. Ann N
Y Acad Sci 2006: 1079: 320–326.
32. THE TRIGR STUDY GROUP. Study design of the Trial to
Reduce IDDM in the Genetically at Risk (TRIGR).
Pediatr Diabetes 2007: 8: 117–137.
33. KIMPIMA
¨KI T, KUPILA A, HA
¨MA
¨LA
¨INEN AM et al. The
first signs of beta-cell autoimmunity appear in infancy
in genetically susceptible children from the general
population: the Finnish Type 1 Diabetes Prediction and
Prevention Study. J Clin Endocrinol Metab 2001: 86:
4782–4788.
104 Pediatric Diabetes 2009: 10: 97–104
Mahon et al.
    • "The study was approved by the San Raffaele Hospital Ethics Committee (protocol DRI- 003). First-degree relatives (FDR) of patients with T1D were enrolled in the T1D TrialNet Pathway to Prevention Trial (TN01 trial, former TrialNet Natural History Study) (18, 19). The overall objective of this study is to perform baseline and repeated assessments over time of the immunologic and metabolic status of individuals at risk for T1D. "
    [Show abstract] [Hide abstract] ABSTRACT: The tolerogenic anti-CD3ε monoclonal Abs (anti-CD3) are promising compounds for the treatment of type 1 diabetes. Anti-CD3 administration induces transient T cell depletion both in preclinical and in clinical studies. Notably, the said depletion mainly affects CD4(+) but not CD8(+) T cells. Moreover, type 1 diabetes reversal in preclinical models is accompanied by the selective expansion of CD4(+)Foxp3(+) T regulatory (Treg) cells, which are fundamental for the long-term maintenance of anti-CD3-mediated tolerance. The mechanisms that lead to this immune-shaping by affecting mainly CD4(+) T effector cells while sparing CD4(+)Foxp3(+) Treg cells have still to be fully elucidated. This study shows that CD3 expression levels differ from one T cell subset to another. CD4(+)Foxp3(-) T cells contain higher amounts of CD3 molecules than do CD4(+)Foxp3(+) and CD8(+) T cells in both mice and humans. The said differences correlate with the anti-CD3-mediated immune resetting that occurs in vivo after anti-CD3 administration in diabetic NOD mice. Additionally, transcriptome analysis demonstrates that CD4(+)Foxp3(+) Treg cells are significantly less responsive than are CD4(+)Foxp3(-) T cells to anti-CD3 treatment at a molecular level. Thus, heterogeneity in CD3 expression seems to confer to the various T cell subsets differing susceptibility to the in vivo tolerogenic anti-CD3-mediated modulation. These data shed new light on the molecular mechanism that underlies anti-CD3-mediated immune resetting and thus may open new opportunities to improve this promising treatment. Copyright © 2015 by The American Association of Immunologists, Inc.
    Full-text · Article · Feb 2015
    • "At present, carrying a " protective " DQB*0602 allele is considered as a criterion of exclusion for enrolling first-degree relatives of diabetic patients in clinical trials, such as the Diabetes Prevention Trial 1 (DPT-1), which is being carried out in the U.S. This trial has been designed to prevent the progression to the clinical onset of T1DM in individuals considered at high risk to develop the disease (DPT-1 Study Group, 1995; Mahon et al., 2009). However, this does not mean that carrying the DQB1*0602 allele confers protection from developing the disease (Greenbaum et al., 2000b; Pugliese et al., 1999). "
    [Show abstract] [Hide abstract] ABSTRACT: Type 1 diabetes mellitus (T1DM) is an autoimmune disease arising through a complex interaction of both genetic and immunologic factors. Similar to the majority of autoimmune diseases, T1DM usually has a relapsing remitting disease course with autoantibody and T cellular responses to islet autoantigens, which precede the clinical onset of the disease process. The immunological diagnosis of autoimmune diseases relies primarily on the detection of autoantibodies in the serum of T1DM patients. Although their pathogenic significance remains uncertain, they have the practical advantage of serving as surrogate biomarkers for predicting the clinical onset of T1DM. Type 1 diabetes is a polygenic disease with a small number of genes having large effects, (i.e. HLA) and a large number of genes having small effects. Risk of T1DM progression is conferred by specific HLA DR/DQ alleles [e.g., DRB1*03-DQB1*0201 (DR3) or DRB1*04-DQB1*0302 (DR4)]. In addition, HLA alleles such as DQB1*0602 are associated with dominant protection from T1DM in multiple populations.
    Full-text · Article · Jan 2015
    • "By analyzing a large cohort of pediatric patients affected by T1D, we have demonstrated that the circulating levels of TRAIL are significantly decreased in T1D with respect to control subjects. On the other hand, in autoantibodypositive individuals (AutoAb POS /T1D NEG ), who have a higher risk of developing T1D than do autoantibody-negative individuals[26], circulating TRAIL levels were not significantly different from healthy control subjects. Diabetic ketoacidosis at the diagnosis of T1D is a lifethreatening situation that represents the main cause of morbidity and mortality in pediatric patients with T1D[3]. "
    [Show abstract] [Hide abstract] ABSTRACT: Experimental evidence in animal models suggests that TNF-related apoptosis-inducing ligand (TRAIL), a member of the TNF superfamily, might play an important role in type 1 diabetes (T1D). We have performed a retrospective study by analyzing the sera of a cohort of pediatric subjects (age ≤18 years; n = 507) consisting of (1) patients diagnosed with T1D (n = 387), (2) healthy individuals (n = 98, considered as controls), and (3) healthy autoantibody-positive subjects (n = 22). Patients with T1D exhibited significantly decreased levels of circulating TRAIL with respect to the control healthy subjects, as well as to the healthy autoantibody-positive subjects. Within the T1D group, no differences in the levels of circulating TRAIL were observed between patients with or without other concomitant autoimmune pathologies. Of note, the levels of TRAIL were significantly lower in the T1D patients analyzed at onset, although reduction in TRAIL levels persisted also in patients analyzed after disease onset (>1 year from diagnosis). In particular, T1D patients who exhibited ketoacidosis at onset showed significantly lower levels of circulating TRAIL with respect to patients without ketoacidosis at onset. Moreover, the levels of TRAIL at diagnosis correlated inversely with the insulin requirement up to 21 months of follow-up. This is the first study demonstrating that the levels of circulating TRAIL are significantly decreased in T1D, with the lowest levels of TRAIL being observed in patients with ketoacidosis at the onset and with the highest insulin requirement.
    Article · Aug 2013
Show more