Att 11_published article describing SEARCH methods

Att 11_published article describing SEARCH methods.pdf

SEARCH for Diabetes in Youth Study

Att 11_published article describing SEARCH methods

OMB: 0920-0904

Document [pdf]
Download: pdf | pdf
Controlled Clinical Trials 25 (2004) 458 – 471
www.elsevier.com/locate/conclintrial

SEARCH for Diabetes in Youth: a multicenter study
of the prevalence, incidence and classification of diabetes
mellitus in youth
The SEARCH Study Group1
Received 7 November 2003; accepted 17 August 2004

Abstract
SEARCH for Diabetes in Youth is an observational, multicenter study focusing on physician-diagnosed
diabetes in individuals b20 years old. The study will estimate the population prevalence and incidence of diabetes
by type, age, gender, and ethnicity and develop practical approaches to diabetes classification in 5 million children
(~6% of the b20 U.S. population) with wide ethnic and socioeconomic representation from four geographically
Corresponding author. Ronny Bell. Wake Forest University School of Medicine, Department of Public Health Sciences,
Medical Center Blvd., Winston-Salem, NC 27157-1063, USA. Tel.: +1 336 716 9736; fax: +1 336 713 4300.
E-mail address: [email protected].
1
Writing group: Dr. David J. Pettitt (Chair), Sansum Medical Research Institute, Santa Barbara, CA; Dr. Ronny Bell,
Wake Forest University School of Medicine; Dr. Dana Dabelea, University of Colorado Health Sciences Center, Denver, CO;
Dr. Lawrence Dolan, Cincinnati Children’s Hospital, Cincinnati, OH; Dr. Giuseppina Imperatore, Centers for Disease Control
and Prevention, Atlanta, GA; Dr. Jean M. Lawrence, Kaiser Permanente Southern California, Pasadena, CA; Dr. Angela D.
Liese, University of South Carolina, Columbia, SC; Dr. Lenna L. Liu, Children’s Hospital and Regional Medical Center,
Seattle, WA; Ms. Beth Waitzfelder, Pacific Health Research Institute, Honolulu, HI. Search steering committee: Pacific Health
Research Institute, Honolulu: Dr. Beatriz L. Rodriguez, Dr. Teresa Hillier, Ms. Beth Waitzfelder; Children’s Hospital and
Regional Medical Center, Seattle: Dr. Catherine Pihoker, Dr. Irl Hirsch, Dr. Carla Greenbaum, Dr. Lenna Liu; Kaiser
Permanente Southern California, Pasadena: Dr. Diana B Petitti, Dr. Jean M Lawrence, Dr. Ann Kershnar, Dr. David J. Pettitt;
University of Colorado Health Sciences Center, Denver: Dr. Richard F. Hamman, Dr. Dana Dabelea, Dr. Georgeanna J.
Klingensmith, Dr. Marian Rewers, Dr. Jonathon Krakoff, Ms. Patricia V. Nash, Ms. Carissa M. Smith; Children’s Hospital
Medical Center, Cincinnati: Dr. Lawrence Dolan, Ms. Debra A. Standiford, Dr. Stephen R. Daniels; University of South
Carolina School of Public Health, Columbia: Dr. Elizabeth J. Mayer-Davis, Dr. Angela Liese, Dr. John Oeltmann; Northwest
Lipid Research Laboratories, Seattle: Dr. Santica Marcovina, Mr. Alan Aldrich; Wake Forest University School of Medicine:
Dr. Timothy Morgan, Dr. Lyn Hardy, Ms. Susan Vestal, Dr. Ronny Bell; National Institute of Diabetes and Digestive and
Kidney Diseases: Dr. Barbara Linder. Centers for Disease Control and Prevention: Dr. Giuseppina Imperatore, Dr. Michael
Engelgau, Dr. Henry Kahn, Dr. Venkat Narayan, Dr. Jinan Saaddine, Dr. Rodolfo Valdez, Dr. Desmond Williams.
0197-2456/$ - see front matter D 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.cct.2004.08.002

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

459

defined populations and two health plans. An estimated 6000 prevalent and 800 incident diabetes cases per year
will be identified with annual follow-up. Cases will be ascertained through clinical and nonclinical resources or
partnerships at each site. Data collection involves patient interviews, physical examinations, laboratory
measurements (diabetes autoantibodies, fasting/stimulating C-peptide, hemoglobin A1c, blood glucose, lipids,
urine albumin, creatinine), medical records reviews, and documentation of risk factors for complications and
processes of care.
D 2004 Elsevier Inc. All rights reserved.
Keywords: Type 1 diabetes; Type 2 diabetes; Children; Adolescents; Ethnic groups

Diabetes is the third most common chronic disease of childhood [1]. However, major gaps exist in
knowledge of the types, frequency, pathophysiology, natural history, and processes of care. Until the past
decade, types of diabetes other than type 1 were rarely diagnosed in children and adolescents. Recently,
several reports describe type 2 diabetes as a pediatric disease [2,3]. However, outside of some AmericanIndian groups and limited data in African-American and Hispanic populations [4], there are virtually no
population-based studies of childhood type 2 diabetes, and the population prevalence of type 2 diabetes
is not known. Also, it is unclear whether the frequency of type 1 diabetes is increasing. Between 1989
and 1994, there was an increase in the average annual incidence of type 1 diabetes in Europe of 3.4%
overall and 6.3% in children aged 0 to 4 years [5]. However, this varied from country to country. In the
United States, the incidence of type 1 diabetes appears to be stable in Colorado and Chicago but has been
reported to be increasing in Allegheny County, Pennsylvania, and Hawaii [6,7]. The reasons for these
discrepancies are not clear.
To date, there are no gold-standard definitions of different types of diabetes presenting in youth.
Clinical phenotypes at onset frequently overlap. Obesity and diabetic ketoacidosis can be found in
both type 1 and type 2 diabetes [2,8,9], and age at diagnosis poorly differentiates between types.
Therefore, a diagnosis based on pathogenesis is a more effective tool for identifying diabetes type
[10,11]. Finally, there is a paucity of data concerning not only the frequency of complications but
also processes of care by diabetes type and by race/ethnicity and gender among youth with
diabetes.
To address these issues, SEARCH for Diabetes in Youth (SEARCH) was initiated in 2000. The
primary aims of SEARCH are (1) to estimate the population prevalence and incidence of type 1, type 2,
and other types (or hybrids, i.e., having evidence of autoimmunity as well as continuing insulin
secretion) of diabetes overall and by age, gender, and race/ethnicity; (2) to develop efficient and practical
approaches to the classification of diabetes type for prevalent and incident cases; and (3) to describe and
compare clinical presentation and course of type 1, type 2, and other types (or hybrids) of diabetes. The
secondary aims of the study are to describe by type (1) the distribution of risk factors for selected
microvascular and macrovascular disease complications; (2) the distribution of selected acute and
chronic complications; and (3) the health care utilization, processes of care, and quality of life. Much
work has been done previously in type 1 diabetes, hence SEARCH will be confirming and updating
findings in type 1 while breaking new ground for type 2 and other types of diabetes. Thus, SEARCH will
bridge existing gaps in the knowledge of diabetes types, prevalence, pathogenesis, frequency of the
complications, quality of care, and health utilization for diabetes in youth in the United States. These
data are essential for the development of public health strategies to prevent diabetes, effectively treat

460

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

diabetes and its complications, and limit the personal and financial burden on the patient and the
financial burden on health care resources.

1. Methods
SEARCH for Diabetes in Youth is an observational, multicenter, population-based study focusing on
cases of physician-diagnosed diabetes in individuals b20 years old. The study is attempting to identify
and enroll all eligible cases of diabetes that are (a) prevalent in the year 2001 and (b) newly diagnosed
(incident) on and after January 1, 2002 through 2004. A nationally standardized data collection effort
therefore builds on the local case ascertainment. SEARCH will then estimate the population prevalence
and incidence of diabetes by type, age, gender, and race/ethnicity. In addition, the project has a
prospective cohort component that involves annual follow-up.
1.1. Study populations and denominator definitions
Six clinical centers are participating in the SEARCH study (Table 1): Cincinnati Children’s Hospital
Medical Center, Cincinnati, OH; University of Colorado Health Sciences Center, Department of
Preventive Medicine and Biometrics, Denver, CO; Seattle Children’s Hospital and Regional Medical
Center, Division of Endocrinology, Seattle, WA; University of South Carolina, Department of
Epidemiology and Biostatistics, Columbia, SC; Kaiser Permanente Southern California, Pasadena,
CA; and Pacific Health Research Institute, Honolulu, HI. Four sites (Cincinnati, Colorado, Seattle, South
Carolina) are identifying cases of diabetes in geographically defined populations. Two sites (Hawaii and
Southern California) are identifying cases of diabetes in membership-based health plans. The ethnic
distribution of the Southern California site is very close to that of the population of all of Southern
California, and because Kaiser Permanente subsidizes the dues for families that cannot otherwise afford
health insurance and accepts children whose health insurance is paid by Medicaid, the full
socioeconomic range is represented. In Hawaii, approximately 95% of the population of the state is
enrolled in one of the health plans that is participating in this study. At some sites, surveillance
populations are larger for the incidence than for the prevalence component to assure an adequate number
of incident cases in the study.
SEARCH includes a wide range of racial and ethnic groups. Within the base population of
eligible SEARCH participants, the estimated racial and ethnic distribution is 65% non-Hispanic
White, 11% Hispanic, 13% African-American, 6% Asian, 2% Pacific Islander, and 3% Native
American.
To estimate the total number of persons in the denominator in 2001, the four geographically based
sites (Cincinnati, Colorado, Seattle, South Carolina) plan to use the nonmilitary, noninstitutionalized
2000 census denominators if these are available at the level of detail required. For years 2002 and
beyond, the geographically based centers are using projections of population changes based on the
2000 Census to estimate denominators for incidence. At the membership-based sites, address data
geocoded to the 2000 census data at the block level will be used to estimate the ethnicity distribution
of the member population for the 2001 prevalence year and for incidence years over a period of time
during which the ethnicity distribution remains stable as determined based on consultation with local
and census experts.

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

461

Table 1
Description of base population and summary of source of estimated cases for prevelance component of SEARCH study
populations under surveillance
Study Site
Cincinnati

Colorado

Hawaii

Seattle
South
Carolina
Southern
California

Prevelance Component

Incidence Component

Base Population

Size

Base Population

Size

Cincinnati and 8 surrounding
urban counties (Hamilton,
Butler, Warren, Clermont OH;
Boone, Kenton, Campbell KY;
Dearborn IN)
Denver and 4 surrounding
counties (Adams, Douglas,
Jefferson, Boulder), Six rural
counties in South-Central
Colorado (Conejos, Costilla,
Alamosa, Sauguache, Mineral,
Rio Grande), Mesa county in
Western Colorado, Native
American reservations in Arizona
and New Mexico
Members of Hawaii Medical Service
Association, Kaiser Foundation
Health Plan-Hawaii and the Hawaii
State Department of Health Services
Med-Quest in Oahu county
King, Pierce, Snohomish, Kitsap,
Thurston counties of Washington State
4 counties (Richard, Lexington,
Orangeburg, Calhoun) surrounding
Columbia, South Carolina
Members of the Kaiser Permanente
Medical Care Program in
Southern California
except San Diego

550,430

Cincinnati and 8
surrounding urban counties
(Hamilton, Butler, Warren,
Clermont OH; Boone, Kenton,
Campbell KY; Dearborn IN)
All 63 counties in Colorado,
Native American reservations
in Arizona and New Mexico

550,430

Members of Hawaii Medical Service
Association, Kaiser Foundation Health
Plan-Hawaii and the Hawaii State
Department of Health Services
Med-Quest in all counties in Hawaii
King, Pierce, Snohomish, Kitsap,
Thurston counties
All 46 counties in South Carolina

300,327

All sites

808,503

240,260

982,920
179,238

700,450

3,444,039

Members of the Kaiser Permanente
Medical Care Program Health Plan
Southern California except San Diego

1,420,839

982,920
1,118,022

700,450

5,072,988

For all centers, race/ethnicity data will be grouped in a uniform manner and collapsed into groups
(non-Hispanic White, Hispanic, African-American, Asian, Pacific Islander, Native American, Other and
Unknown) using rules and conventions developed by the Census [12]. Other potential racial groupings
of scientific interest will be considered.
1.2. Case definition and validation
SEARCH cases are considered valid (or true cases) if (a) there is a physician diagnosis of diabetes; or,
(b) the parent or the youth self-reports a physician diagnosis of diabetes. A physician-diagnosed case of
diabetes is established if any of the following criteria are met: medical record review indicates a
physician diagnosis of diabetes; the diagnosis of diabetes is directly verified by a physician; the diabetes

462

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

case is breferredQ to the study by a physician; diabetes is listed as the underlying or contributing cause of
death on a death certificate; or the case is included in a clinical database that has a requirement for
verification of diagnosis of diabetes by a physician. This study will exclude cases of gestational diabetes.
1.3. Eligibility and exclusion criteria
To be eligible to participate in SEARCH (Table 2), participants must be in the index year (prevalent
2001 or incident 2002, 2003, 2004): (a) b20 years old and (b) resident of the population defined in the
index year for geographically based centers or member of the participating health plan in the index year
for membership-based centers. Because the denominators are derived from Census data for the four
geographically based centers, individuals who are active duty military or institutionalized are excluded.
1.4. Case ascertainment
The approaches for identification of prevalent and incident cases vary by site based on availability of
existing diabetes databases and access to clinics, physicians, and computer-stored data resources.
Despite different approaches, each geographic or membership-based site is attempting to identify every
case of diabetes within its purview. Geographic-based sites (Cincinnati, Colorado, Seattle, South
Carolina) have established active surveillance systems de novo for SEARCH, based on networks of
pediatric and adult endocrinologists, existing pediatric diabetes databases, hospitals, health plan
databases, and other health care providers. Membership-based sites (Hawaii and Southern California) are
using existing diabetes databases in addition to their administrative databases as the source for case
identification. Death certificate searches are conducted to identify missed fatal cases (and to assess
mortality from diabetes). In the population under surveillance, all deaths, regardless of cause, for persons
with diabetes b20 years old are being identified and recorded.
1.5. Assessment of completeness
In SEARCH, evaluation of completeness of case ascertainment is crucial and is being conducted
using both statistical methods and additional, targeted data collection efforts. Capture–recapture [13–16],

Table 2
Eligibility criteria for SEARCH
Prevalence
Physician diagnosed cases of diabetes mellitus
Prevalent in 2001
Age less than 20 years on December 31, 2001;
Born between 1/1/82–12/31/2001
Resident of the population at any time in 2001 or
member of the participating health plan at any time in 2001
Not active–duty military
Not living in an institution (Census definition)
Not Gestational Diabetes

Incidence
Physician diagnosed cases of diabetes mellitus
First clinical diagnosis of diabetes in a non-pregnant
state January 1 through December 31 in the incidence year
Age less than 20 years at diagnosis
Resident of the population at diagnosis or member of the
participating health plan at diagnosis
Not active–duty military
Not living in an institution (Census definition)
Not Gestational Diabetes

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

463

a statistical approach that attempts to estimate completeness from incomplete samples, is used in the
geographically based sites with multiple independent sources of cases (Cincinnati, Colorado, Seattle,
South Carolina). These sources may include hospital discharge, laboratory, pharmacy, ambulatory
billing, and pediatric endocrinology case lists. This method will not be applied in the two membershipbased centers, because sources of case identification are highly dependent. In addition, intensive casefinding based on a mailed survey to a defined sample of providers in specialties likely to see youth with
diabetes, who are not included in primary case ascertainment, will be conducted in geographically based
centers. The cases identified will be compared with cases that have already been identified by SEARCH
as permitted by local Institutional Review Boards (IRBs).
1.6. Protection of human subjects
The SEARCH protocol has been reviewed and approved by the Institutional Review Boards (IRBs) of
each of the participating institutions. Written informed consent is obtained from participants age N18
years or their parents or legal guardians if b18 years. Assent is obtained in accordance with local IRB
requirements. Subjects may agree to participate in the study at a number of levels, which is reflected in
staged consent and assent forms. In order to further protect the privacy of participants, SEARCH has
obtained a certificate of confidentiality for each of the sites.
Because subjects will be asked to appear fasting and some laboratory tests that are not generally a part
of standard diabetes care will be performed, alert values have been defined for clinically relevant data
measurements and processes established to provide appropriate medical care. Adverse events will be
reported to Wake Forest University School of Medicine (coordinating center), each clinical center, and
the local IRBs. The study protocol will be modified as required to maintain participant safety.
1.7. Data collection and measures
Data collection is organized into a series of sections that consist of one or more data collection
instruments or measurements. All staff have been trained and certified on the standardized protocol and
manual of procedures.
All data collection forms are available in English and Spanish. Staff is either bilingual in English and
the subject’s preferred language or arrangements for an on-site translator are made.
1.7.1. Initial patient survey
The Initial Patient Survey (IPS) is a questionnaire used to collect information to assist with case
validation and to confirm eligibility. It also collects information about race/ethnicity, diabetes type,
and preliminary treatment information. The IPS may be completed either as a self-administered
mailed questionnaire (by the parent, or depending on age, by the participant), during a telephone
interview, or at the time of the in-person visit. It is expected that some youth with diabetes will be
willing to complete the IPS by mail or telephone even if they are not willing to participate in other
parts of the study.
1.7.2. In-person visit
The in-person visit consists of a physical examination (anthropometry, blood pressure, acanthosis),
laboratory work (fasting blood and urine collections), and the administration of several questionnaires to

464

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

collect information on medical history, family history, quality of life, depression, and health behaviors.
This visit is designed so it can be conducted in clinical research settings, health clinics, or the
participants’ homes. For incident case participants, the in-person visit is conducted as soon as possible
after the subject becomes clinically stable (approximately one month after initial diagnosis).
Questionnaires are interviewer-administered (health, family history, quality of life—child report,
health behaviors) or self-administered after staff instruction (quality of life—parent report, depression,
diet). The primary respondents for the health and family history questionnaires are the parent or legal
guardian. Participants N10 years old are asked to complete additional questionnaires, either intervieweradministered or self-administered after staff instruction, focusing on health behaviors (diet, smoking,
sleeping patterns) and depression. The parents or legal guardians of participants b18 years old are asked to
waive their right to review their children’s responses to these questionnaires prior to completion.
Blood and urine specimens are collected from all prevalent and incident cases, and the physical
examination is conducted on those aged 3 years or older. The exceptions are cases with other specific
types of diabetes, e.g., diabetes known to be secondary to another illness, such as cystic fibrosis or
medications, such as steroids, in whom data collection is limited to brief, annual mailed surveys.
Laboratory samples (with the exception of diabetes autoantibodies{DAA}) are obtained under
conditions of metabolic stability, defined as no episode of diabetic ketoacidosis during the previous
month. The DAA that will be obtained include glutamic acid decaboxylase, IA-2, and insulin
autoantibodies. Specimens are processed locally at the sites and shipped within 24 h to the central
laboratory (Northwest Lipid Laboratory, University of Washington, Seattle, WA) where they are
analyzed. Serum or plasma samples are analyzed for GAD65 antibodies in a radioligand-binding assay.
Details of the assay have been published previously [17,18]. In short, human recombinant GAD65 is
produced by in vitro transcription and translation system (Promega, Madison, WI) and labeled with 35Smethionine [19]. Reduction in the background radioactivity is achieved by filteration of the assay buffers
through a 0.22-Am filter (Millipore, Bedford, MA). The labeled GAD65 is incubated with serum in
duplicate determinations. Protein-A-Sepharose (Zymed, San Francisco, CA) is used to separate the GADtracer–GAD-antibody complexes from the free 35S-GAD65. Following repeated washings in Millipore
Multiscreen micron filteration plates, plates are dried, added with microscint fluid, and radioactivity in
wells is determined using a Top Count (96-well plate Beta counter, Packard). The laboratory uses a set of
positive and negative standard in each plate. Concentrations of GAD65 Ab are expressed as a GAD65AB
index to correct for the interassay variation according to the following formula: GAD65Ab index=[counts
per minute (cpm) of the unknown sample-average cpm of two negative standards]/(cpm of the positive
standard-average of two negative standards). The positive and negative controls are run in duplicate in
each assay. Standards and quality controls: the assay uses a positive control which is the WHO standard
for islet cell antibodies. A negative control sample used for the assay was prepared from a pool of normal
sera, and therefore the pool can be reproduced as necessary. A signal-to-noise ratio of 10 or above is a
requirement before an assay is considered acceptable.
The IA2 antibody assay is identical to the GAD65Ab assay, except using 35S-labelled IA2 as a tracer.
IA2Ab index=(cpm of the unknown sample-average cpm of two negative standards)/(cpm of the positive
standard-average of two negative standards). The positive and negative controls are run in duplicate in
each assay. Standards and quality controls: the assay uses a positive control which is the WHO standard
for islet cell antibodies. A negative control sample used for the assay was prepared from a pool of normal
sera, and therefore the pool can be reproduced as necessary. A signal-to-noise ratio of 10 or above is a
requirement before an assay is considered acceptable.

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

465

The assay for insulin autoantibody (IAA) utilizes a competition of insulin antibodies in a serum or
plasma sample for unlabeled and labeled insulin for quantitative determination of the antibody levels
[20]. 125 I-Insulin (Amersham) of 20,000 cpm is incubated with 5 Al of serum with and without cold
insulin diluted in buffer A (20 mM Tris, 150 mM NaCl, 1% BSA, 0.15% Tween-20, and 0.1% sodium
azide). Following a 3-day incubation at 4 8C, 50 Al of 50% protein-A/8% protein-G-Sepharose
(Pharmacia) mixture is added to the incubation in a Multiscreen-NOB 96-well filteration plate (Millipore
plate), which has been precoated with buffer A overnight at room temperature. The plate is shaken for 45
min at 4 8C followed by extensive washing in buffer B (buffer A containing 0.1% BSA) using a vacuumoperated plate washer. After washing, 40 Al of scintillation fluid (Microscint-20, Packard) is added to
each well, and radioactivity is determined using a Top Count (96-well plate Beta counter, Packard). The
results are calculated based on the difference in counts per minute (Dcpm) between the well without cold
insulin and the well with insulin and expressed as an index (IAA index).
1.7.3. C-peptide testing
C-peptide is measured fasting and following a mixed meal challenge with BoostR (Mead Johnson and
Company, Evansville, IN) with blood samples drawn at 30, 60, and 90 min after the liquid meal. The
subgroups of participants who undergo this test and the frequency of this test are described under link to
Classification of Diabetes Type. There are three objectives to stimulated C-peptide testing: (1) to develop
better ways to distinguish type 1 and type 2 diabetes; (2) to better define diabetes type in SEARCH
participants; and (3) to understand the evolution of insulin secretion in children with diabetes. C-peptide
in serum or plasma is measured in the central laboratory by an in-house-developed radioimmunoassay
[21]. Plasma or serum samples are incubated with limiting concentrations of an in-house-produced
guinea pig antibody specific to human C-peptide and 125I-labeled purified C-peptide. The C-peptide in
patient samples and the 125I-labeled C-peptide compete for the antibody binding sites. The decrease in
the amount of 125I-labeled C-peptide is proportional to the concentration of C-peptide in the samples.
The antigen–antibody complex is precipitated by the addition of a goat antiguinea pig IgG and
polyethylene glycol. Supernatants are decanted and the specific radioactivity in the immunocomplexes
counted using an auto gamma counter (Packard Instruments). The concentration of C-peptide in the
samples is determined using a standard curve generated with known concentrations of purified human Cpeptide. The assay is linear up to a 5000 pg/ml concentration. To monitor the long-term consistency of
the C-peptide results and to avoid assay drift, the laboratory has prepared in 1995 two lyophilized plasma
pools representing high and low C-peptide reference ranges. Six replicates of these two samples are
included in an assay at the end of each month, and obtained values are plotted against the expected
values. Over 300 vials of each pool are stored at 80 8C and should last for an extended period of years.
Assay precision is excellent, with a CV of 6.6 for the high quality control and 10.7 for the low quality
control. Sensitivity limit for this assay is 0.15 ng/ml. The values are reported in nanograms per milliliter.
These can be converted to nanomole per liter by multiplying with a factor of 0.333.
DNA and biologic samples are being stored for future studies designed to meet the primary and
secondary aims of SEARCH only for participants for whom written informed consent and assent has
been obtained according to local IRB guidelines.
1.7.4. Medical record review
For incident cases, medical records are reviewed to collect information on clinical presentation, initial
clinical course, and utilization of health care services. Specific information regarding tests for DAA, C-

466

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

peptide, diabetes-related genes, diabetes complications and comorbidities, and prescription medications
are recorded. Information is collected from all provider visits (inpatient and outpatient) that occur within
2 months prior to and up to 6 months following the diabetes diagnosis. For prevalent cases, an
abbreviated medical record review is conducted if information is needed to establish diabetes type in
untypeable cases (see Classification of Diabetes Type).
1.7.5. Annual follow-up
All incident cases, prevalent hybrid cases (defined below in Classification of Diabetes Type), and
those with known genetic beta cell defects are followed with annual in-person visits. These include the
physical examination and laboratory tests and questionnaire information about all factors that could
change over time, except DAA and stimulated C-peptide testing in type 1A cases. All prevalent cases
and all cases initially classified as other specific diabetes types are asked to complete an annual survey
by mail. This survey gathers information on health care utilization and updates contact information to
facilitate ancillary studies.
1.8. Classification of diabetes type
As recommended by the American Diabetes Association (ADA) Expert Committee on the Diagnosis
and Classification of Diabetes Mellitus [22], SEARCH has developed a systematic approach to triage for
classification of diabetes based on pathogenesis. The guiding principles for this triage system, described
below and in Fig. 1, follow a hierarchical procedure. Data collected using SEARCH laboratory values
will supercede those from other laboratories that will in turn supercede clinical data alone.
Initially, data measured in the SEARCH laboratory (DAA {antibodies to glutamic acid decarboxylase,
IA-2, and insulin autoantibodies} and fasting plasma C-peptide concentration) define diabetes type, and
an assignment to type 1A, type 1, type 2, or hybrid diabetes is made for incident and prevalent cases. All
other cases in which this determination cannot be made are considered untypeable and investigated
further as outlined below.
Stimulated C-peptide tests are done on prevalent cases N8 years that are initially untypeable and on
cases with mixed features of autoimmunity and insulin resistance. The initial study plan was to use the
same criteria for stimulated C-peptide testing in incident cases. During the second year of data collection,
it was decided to invite all incident cases N8 years to undergo stimulated C-peptide testing, because there
was an unexpectedly large number of incident cases that could not be categorized based solely on
autoantibody data and fasting C-peptide.
Cases age N8 years old that are initially untypeable will undergo a stimulated C-peptide test using cut
points that will assure making misdiagnosis highly unlikely [23]. Diabetes type will then be assigned by
comparing the biochemical and clinical phenotypes of the untypeables to the biochemical and clinical
phenotypes of two reference groups: (1) all laboratory-defined incident type 2 cases from SEARCH N8
years old, and (2) a sample of cases that are drawn from all incident SEARCH type 1A cases, which will
be matched to untypeable cases by age and date of ascertainment of specimens. Subjects who decline
measurement of DAA and C-peptide as part of the study but who have had DAA and/or C-peptide
measured in the past in a non-SEARCH laboratory are classified as shown in Fig. 1.
Participants for whom only clinical data are available are classified based on the clinical phenotype
defined in the reference group described previously. In those who still cannot be typed by this method,
the following clinical definitions will be used: type 1 diabetes—diagnosis of diabetes made when the

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

467

Fig. 1. SEARCH initial diabetes classification approach.

subject was b10 years old with a weight less than the 25th percentile for chronologic age or body mass
index less than the 50th percentile for chronologic age at diagnosis; type 2 diabetes—duration of
diabetes greater than 1 year and no insulin therapy for 1 month without an episode of diabetic
ketoacidosis, or duration of diabetes greater than 6 months and never treated with insulin.
Subjects previously identified as having a genetic defect in beta cell functioning will undergo yearly
DAA measurement and a stimulated C-peptide test. For subjects classified with other specific types of
diabetes [22], except those with a previously diagnosed genetic defect in beta cell functioning, the type
of diabetes is recorded and no further testing for typology performed.
As new information, such as new DAA or new markers of beta cell destruction, becomes available over
the course of the study, additional testing will be performed, and the approach to classification of diabetes
type will be modified to reflect the most accurate and current methods of classifying the types of diabetes.
1.9. Statistical considerations
Based on presumed negligible changes in population size and characteristics between 2000 and 2001,
prevalence estimates are calculated per 1000 persons less than 20 years of age in 2001 by dividing the

468

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

total number of validated, eligible prevalent cases in 2001 by the total number of persons aged less than
20 years who were residents or members of the denominator population in 2000. Incidence estimates are
calculated per 100,000 persons less than 20 years of age per year by dividing the total number of
validated, eligible cases of diabetes with onset in the incidence year by the estimated total of persons
resident or member of a given populations in the incidence year. Prevalence in 2001 and annual
incidence are also calculated by type, age, race/ethnicity (non-Hispanic White, Hispanic, AfricanAmerican, Asian, Pacific Islander, Native American) and gender. National estimates of prevalence and
incidence are derived by applying the age, gender, and race/ethnicity specific estimates of prevalence
and incidence derived from SEARCH to counts of the U.S. national population in 2000 (for prevalence)
and projections of the U.S. national population for subsequent years.
The number of expected prevalent and incident cases cannot be estimated with precision using
published data, especially considering the ethnic diversity of the base population. It is anticipated that
there will be 6200 to 6800 prevalent cases and at least 800 incident cases annually.
Furthermore, SEARCH aims to develop efficient and practical approaches to classification of diabetes
type for prevalent and incident cases (aim 2). An assessment of the accuracy of different approaches is
based on misclassification ratios. The number of misclassifications is used to estimate the proportion of
false positives, false negatives, sensitivity, and specificity. In evaluation of potential measures that are
continuous or ordinal, receiver operator curves (ROC) is used to evaluate and test the usefulness of the
diagnostic measure.
The third major aim of the study is to describe and compare clinical presentation and course of type 1,
type 2, and other types or hybrids of diabetes. Patient characteristics and clinical presentation will be
compared between types of diabetes. The statistical significance of these comparisons will be tested
using the chi-square tests for categorical measures, Wilcoxon rank-sum tests for ordinal measures, and
analysis of variance for continuous measures. Analyses are performed separately for prevalent and
incident cases. Analysis of covariance procedures will be used to compare complications and risk factors
for complications (e.g., hypertension, microalbuminuria, hyperglycemia) between types of diabetes
adjusting for possible confounders (e.g., age, sex, race/ethnicity).
The clinical course of incident diabetes cases (acute complications, quality of life, insulin production,
glycemia, lipidemia, development of microalbuminuria, presence of autoantibodies) is described using
longitudinal data collected once per year. The statistical significance of differences in clinical course by
diabetes type will be tested using repeated measures analysis of covariance and mixed models [24].
Maximum likelihood will be used to fit these models, because this increases precision and minimizes
bias associated with varying lengths of follow-up among participants.

2. Discussion
SEARCH is uniquely poised to generate vital information required to develop clinical interventions
and public health policies designed to reduce the incidence and improve the outcomes of diabetes in
youth. The six study centers provide a study population that is larger and more diverse in terms of race/
ethnicity, geography, and age than any previous study of diabetes in children. Importantly, this study
systematically employs a uniform methodology of diabetes classification based on pathogenic criteria,
reflecting the recommendations of the American Diabetes Association (ADA) Expert Committee on the
Diagnosis and Classification of Diabetes Mellitus [22]. The use of this systematic method of

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

469

classification with this large and diverse study population will enable this study to achieve its primary
objective, to provide accurate estimations of the incidence and prevalence of diabetes by type, age,
gender and race/ethnicity, and enhances the generalizability of this approach in other settings. The
study is powered to identify differences in the presentation and course of each type of diabetes by age,
sex, and race/ethnicity if they are present. In addition, the study is designed to facilitate the longitudinal
study of this pediatric population, which, if further funding is obtained, has the potential to illuminate
factors associated with the onset and progression of diabetes-related complications. Finally, DNA and
biological samples obtained by this study will provide an invaluable resource for genetic and
biochemical studies.
SEARCH does however have limitations. The study will determine the prevalence and incidence of
diagnosed diabetes only. No attempt will be made to determine how much undiagnosed diabetes exists in
youth or whether undiagnosed cases vary by age or ethnicity. While it is recognized that type 2 diabetes
may be present for years prior to diagnosis, SEARCH will not screen for undiagnosed cases. Growing
awareness of the presence of type 2 diabetes among youth may influence screening or diagnostic
approaches by health care providers over time. In the absence of an extension beyond the year 2005, the
study will have a limited ability to address temporal trends in the incidence of diabetes given that it will
only encompass 3 years.
SEARCH is also faced with a number of practical challenges. The completeness of case ascertainment
depends on the cooperation of multiple health care providers and organizations. To assure complete
cases ascertainment, all potential sources of care for children with diabetes have been enumerated in the
geographic sites, and the study monitors the ability of sites to obtain data from these sites. Linkage of
computer-stored records from a large number of sources (pharmacy, hospitalizations, outpatient records,
laboratory) is used in an attempt to assure complete case ascertainment in the membership-based sites.
Capture–recapture methods are being explored as a way to assess completeness of case ascertainment in
settings where all potential cases that derive from each data source can be determined to be validated,
unique cases of diabetes or ruled out as validated unique cases and where reasonable independence of
information from the various sources can be assured. Concern about the completeness of case
ascertainment is unlikely to be completely laid to rest.
The recruitment of this study population also presents some unique obstacles. The study population is
by definition a vulnerable population due to the age composition. It is anticipated that there will be
missing data resulting from patients who refuse enrollment. This is particularly problematic in
calculating prevalence and incidence, and distribution of diabetes type, if refusal is more common
among specific groups of youth (e.g., ethnic minorities versus whites, older versus younger, girls versus
boys). Although the study is likely to yield substantial benefits in terms of the knowledge it will
generate, the benefit to individual participants is somewhat limited, and participation entails a significant
amount of time and minimally invasive procedures.
Issues of data confidentiality are addressed in the study design by reducing the data required to
document a case to a minimal set of variables without individual identifiers. The study design also
incorporates the use of multiple data sources as needed and as possible at each of the sites, including
information from parents and patients, health care providers, medical records, and administrative health
care data, in an effort to reduce the amount of missing data.
The recent marked increases in type 2 diabetes in youth are occurring in parallel with increases in
adults and appear to be directly related to nationwide increases in obesity in all ages and ethnic groups.
Reasons for the regional increases in type 1 diabetes have not been clearly identified. The SEARCH

470

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

study will increase our understanding of the public health burden of diabetes for youth, who are at high
risk for long-term complications from this disease.

Acknowledgements
SEARCH for Diabetes in Youth is funded by the Centers for Disease Control and Prevention (PA
number 00097) and supported by the National Institute of Diabetes and Digestive and Kidney
Diseases.

References
[1] Rewers M, LaPorte RE, King H, Tuomilehto J. Trends in the prevalence and incidence of diabetes: insulin-dependent
diabetes mellitus in childhood. World Health Stat Q 1988;41:179 – 89.
[2] Fagot-Campagna A, Pettitt DJ, Engelgau MM, Burrows NR, Geiss LS, Valdez R, et al. Type 2 diabetes among North
American children and adolescents: an epidemiologic review and a public health perspective. J Pediatr 2000;136:664 – 72.
[3] Rosenbloom AL, Joe JR, Young RS, Winter WE. Emerging epidemic of type 2 diabetes in youth. Diabetes Care
1999;22:345 – 54.
[4] Dabelea D, Pettitt DJ, Jones KL, Arslanian S. Type 2 diabetes mellitus in minority children and adolescents: an emerging
problem. In: Arslanian S, editor. Endocrinology and metabolism clinics of North America, vol. 28. 1999. p. 709 – 29.
[5] EURODIAB ACE Study Group. Variation and trends in incidence of childhood diabetes in Europe. Lancet 2000;355:
873 – 6.
[6] Onkamo P, Vaananen S, Karvonen M, Tuomilehto J. Worldwide increase in incidence of Type I diabetes—the analysis of
the data on published incidence trends. Diabetologia 1999;42:1395 – 403.
[7] Lipton R, Keenan H, Onyemere KU, Freels S. Incidence and onset features of diabetes in African-American and Latino
children in Chicago, 1985–1994. Diabetes Metab Res Rev 2002;18:135 – 42.
[8] Ferrara A, Imperatore G, Quesenberry CP, Karter AJ, Fagot-Campagna A, Ergas IJ, et al. Characteristics and quality of
health care of type 1 and type 2 diabetic youths: the Northern California Kaiser Permanente Diabetes Registry. Diabetes
2002;51(Suppl. 2):A26.
[9] Barrett TG, Ehtisham S, Smith A, Hattersley AT. UK Diabetes Survey shows type 2 diabetes present in 0.4% of newly
diagnosed children, associated with overweight, female, and ethnic minority status. Diabetes 2002;51(Suppl. 2):A25.
[10] American Diabetes Association: report of the expert committee on the diagnosis and classification of diabetes mellitus.
Diabetes Care 1997;20:1183 – 97.
[11] Alberti KG, Zimmet PZ. Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1:
diagnosis and classification of diabetes mellitus provisional report of a WHO consultation. Diabet Med 1998;15:539 – 53.
[12] US Census Bureau. United States Census 2000, available at: http://www.census.gov/).
[13] Fienberg SE, Johnson MS, Junker BW. Classical multilevel and Bayesian approaches to population size estimation using
multiple lists. J R Stat Soc, Ser A 1999;162:383 – 406.
[14] Verlato G, Muggeo M. Capture–recapture method in the epidemiology of type 2 diabetes. Diabetes Care 2000;23:759 – 64.
[15] Capture–recapture and multiple-record systems estimation. I: history and theoretical development. International Working
Group for Disease Monitoring and Forecasting. Am J Epidemiol 1995;142:1047 – 58.
[16] Hook EB, Regal RR. The value of capture–recapture methods even for apparent exhaustive surveys. The need for
adjustment for source of ascertainment intersection in attempted complete prevalence studies. Am J Epidemiol
1992;135:1060 – 7.
[17] Falorni A, Ortqvist E, Persson B, Lernmark A. Radioimmunoassays for glutamic acid decarboxylase (GAD65) and
GAD65 autoantibodies using 35S or 3H recombinant human ligands. J Immunol Methods 1995;186:89 – 99.
[18] Marcovina SM, Landin-Olsson M, Essen-Moller M, Palmer JP, Lernmark A. Evaluation of a novel radioimmunoassay
using 125I labeled human recombinant gad65 for the determination of glutamic acid decarboxylase (GAD65)
autoantibodies. Int J Clin Lab Res 2000;30:21 – 6.

SEARCH Study Group / Controlled Clinical Trials 25 (2004) 458–471

471

[19] Gruben CF, Daniels T, Toivola B. A novel radioligand binding assay to determine diagnostic accuracy of isoform specific
glutamic acid decarboxylase antibodies in childhood IDDM. Diabetologia 1994;37:344 – 50.
[20] Yu L, Robles DT, Abiru N, Kaur P, Rewers M, Kelemen K, et al. Early expression of antiinsulin autoantibodies of
humans and the NOD mouse: evidence for early determination of subsequent diabetes. Proc Natl Acad Sci
2000;97:1701 – 6.
[21] Heding LG. Radioimmunological determination of human c-peptide in serum. Diabetologia 1975;11:541 – 8.
[22] American Diabetes Association. Report of the Expert Committee on the diagnosis and classification of diabetes Mellitus.
Diabetes Care 2000;23:S4 – 19.
[23] Allen AF, Klingensmith GJ, Jensen P, Simoes E, Hayward A, Chase HP. Effect of Bacillus Calmette–Guerin vaccination
on new-onset type 1 diabetes. A randomized clinical study. Diabetes Care 1999;22(10):1703 – 7.
[24] Laird NM, Ware JH. Random-effects models for longitudinal data. Biometrics 1982;38:963 – 74.


File Typeapplication/pdf
File Modified2004-09-25
File Created2004-09-25

© 2024 OMB.report | Privacy Policy