ORIGINAL CONTRIBUTION

Incidence of Diabetes in Youth

in the United States
The Writing Group for the SEARCH
Context Data on the incidence of diabetes mellitus (DM) among US youth accord-
for Diabetes in Youth Study Group* ing to racial/ethnic background and DM type are limited.

E
STIMATES OF THE INCIDENCE OF Objective To estimate DM incidence in youth aged younger than 20 years accord-
type 1 diabetes mellitus (DM) ing to race/ethnicity and DM type.
derived from population- Design, Setting, and Participants A multiethnic, population-based study (The SEARCH
based registries show an in- for Diabetes in Youth Study) of 2435 youth with newly diagnosed, nonsecondary DM in
crease in incidence worldwide during 2002 and 2003, ascertained at 10 study locations in the United States, covering a popu-
the past 2 decades,1-7 albeit less so in lation of more than 10 million person-years.
the United States.8,9 It has been as- Main Outcome Measure Incidence rates by age group, sex, race/ethnicity, and DM
sumed that type 1 DM identified in type were calculated per 100 000 person-years at risk. Diabetes mellitus type (type 1/type
these registries is mostly autoimmune- 2) was based on health care professional assignment and, in a subset, further character-
mediated, although this has not been ized with glutamic acid decarboxylase (GAD65) autoantibody and fasting C peptide measures.
confirmed with measured diabetes au- Results The incidence of DM (per 100 000 person-years) was 24.3 (95% confi-
toantibodies. dence interval [CI], 23.3-25.3). Among children younger than 10 years, most had type
Type 2 DM has traditionally been 1 DM, regardless of race/ethnicity. The highest rates of type 1 DM were observed in
viewed as a disorder of adults, most non-Hispanic white youth (18.6, 28.1, and 32.9 for age groups 0-4, 5-9, and 10-14
commonly observed in those persons years, respectively). Even among older youth (ⱖ10 years), type 1 DM was frequent
among non-Hispanic white, Hispanic, and African American adolescents. Overall, type
who are middle-aged or elderly. In- 2 DM was still relatively infrequent, but the highest rates (17.0 to 49.4 per 100 000
deed, onset of DM after 30 or 40 years person-years) were documented among 15- to 19-year-old minority groups.
has frequently been used to distin-
Conclusions Our data document the incidence rates of type 1 DM among youth of
guish type 2 from type 1 DM. How-
all racial/ethnic groups, with the highest rates in non-Hispanic white youth. Overall,
ever, as the prevalence of obesity has type 2 DM is still relatively infrequent; however, the highest rates were observed among
increased in recent decades, several adolescent minority populations.
studies have reported an increasing pro- JAMA. 2007;297:2716-2724 www.jama.com
portion of youth with apparent type 2
DM, especially among racial/ethnic mi- groups and DM types. The SEARCH METHODS
nority populations.10,11 The number of for Diabetes in Youth Study has been Structure of the SEARCH
population-based studies is small,12 specifically designed to identify inci- Incidence Study
many were conducted among Ameri- dent cases of DM among individuals The SEARCH study is a multicenter ob-
can Indians,13-15 and most used a clini- younger than 20 years to estimate the servational study conducting popula-
cal DM definition.16,17 population incidence of type 1, type
There are currently limited com- 2, and other types of DM overall and *The Writing Group and a list of the SEARCH for Dia-
prehensive population-based esti- by age and race/ethnicity. We report betes in Youth Study Group Investigators appear at
mates of DM incidence among US herein incidence estimates of DM in
the end of this article.
Corresponding Author: Dana Dabelea, MD, PhD, Uni-
youth covering all major racial/ethnic youth for the 2002-2003 period by versity of Colorado Health Sciences Center, Depart-
ment of Preventive Medicine and Biometrics, 4200 E
age group, sex, DM type, and race/ Ninth Ave, Box C245, Denver, CO 80262 (dana
For editorial comment see p 2760.
ethnicity. [email protected]).

2716 JAMA, June 27, 2007—Vol 297, No. 24 (Reprinted) ©2007 American Medical Association. All rights reserved.
 INCIDENCE OF DIABETES IN US YOUTH

tion-based ascertainment of cases of vice units or by participation in the re- sex, and date of diagnosis were avail-
physician-diagnosed DM in youth aged search study. A race-bridging model22 able on all cases. The clinical DM type
younger than 20 years.18 New DM cases was used to classify persons with at least assigned by the health care profes-
occurring in 2002 and 2003 were iden- 2 self-reported races into larger catego- sional was obtained from medical rec-
tified (1) in geographically defined ries. Race/ethnicity-specific estimates ords or physician reports and catego-
populations in Ohio, Washington, were pooled across sites using 5 cat- rized as follows: (1) type 1 (combining
South Carolina, and Colorado; (2) egories: non-Hispanic white, His- type 1, type 1a, and type 1b), (2) type
among health plan enrollees in Ha- panic, African American, Asian/ 2, and (3) other types (including hy-
waii (Hawaii Medical Service Associa- Pacific Islander, and American Indian. brid type, maturity onset diabetes of the
tion, Med-Quest, Kaiser Permanente The numerator included all youth young, secondary DM, type unknown
Hawaii) and California (Kaiser Perma- with nongestational DM newly diag- by the reporting source, type desig-
nente Southern California, excluding nosed in 2002 and 2003 who were nated as other, and missing type). Race/
San Diego); and (3) with coordination younger than 20 years on December 31 ethnicity was based on self-report or
by the Colorado center, among health of the index year. The case ascertain- medical record−based data for 95.9% of
service beneficiary rolls in 3 American ment approach involved networks of pe- cases, and on US Census block-level
Indian reservation-based populations in diatric and adult endocrinologists, ex- geocoding for the 4.1% cases with miss-
Arizona and New Mexico, and among isting pediatric DM databases, hospitals, ing race/ethnicity. All validated case re-
participants in the National Institute of health plan databases, and other health ports together with the corresponding
Diabetes and Digestive and Kidney Dis- care organizations.18 Geographic-based core variables described above were reg-
eases Pima Indian study in Arizona.19 centers established active surveillance istered anonymously with the coordi-
systems de novo. The American Indian nating center at Wake Forest Univer-
Incidence Study Population reservation-based populations used ex- sity in North Carolina.
The denominator included noninstitu- isting DM databases as the source for case
tionalized, nonmilitary youth aged identification. In addition to reporting of Additional Data Collection
younger than 20 years in the index year. cases by pediatric endocrinologists, the In addition to the case validation pro-
Because the 2000 US Census projec- membership-based sites identified cases cess and collection of core variables,
tions for youth residing in the partici- using information from linkage of com- youth with nonsecondary DM identi-
pating areas were similar in 2002 and puter data on prescriptions, hospitaliza- fied by SEARCH were asked to partici-
2003 (−0.2% change overall), the same tion with DM as the discharge diagno- pate in a research visit that included
denominator was used for both years. sis, and laboratory measures of glycated study-specific questionnaires, a brief
The study covered 10 031 888 person- hemoglobin A1c. physical examination, and a blood
years at risk, which represents 6.2% of draw.23 The questionnaires collected
the US population younger than 20 Case Validation and Collection information on participant’s medical
years. Derivation of the appropriate de- of Core Variables history and comorbid conditions,
nominators was a multistep process tak- All case reports were validated on the health services utilization, insurance,
ing into account racial/ethnic catego- basis of physician reports or medical and satisfaction with medical care.
rization and the civilian nature of the record reviews, or self-report of a phy- Blood was drawn for measurement of
study population.20 For the geographi- sician diagnosis of DM in 60 cases. A glutamic acid decarboxylase antibod-
cally based sites, nonmilitary age-, sex-, physician-diagnosed case of DM was es- ies and fasting C peptide levels, which
and race/ethnicity-specific denomina- tablished if any of the following crite- were used to further characterize the
tors were determined based on projec- ria were met: (1) medical record re- clinically assigned DM type. Of the
tions from the 2000 US Census (http: view indicated a physician diagnosis of 2435 cases invited to the visit, 1134
//www.cdc.gov/nchs/about/major/dvs DM, (2) the diagnosis of DM was di- (46%) had measurements of glutamic
/popbridge/popbridge.htm). For rectly verified by a physician, (3) the acid decarboxylase antibodies and
California, race/ethnicity-, age-, and sex- physician referred a youth with DM to fasting C peptide levels. The visit
specific denominators were based on the study, or (4) the case was in- occurred after an overnight fast, under
block-level geocoding of the health plan cluded in a clinical database that had a conditions of metabolic stability,
membership. 21 For Hawaii, racial/ requirement for verification of diagno- defined as no episode of diabetic keto-
ethnic denominators were based on sis of DM by a physician. For all vali- acidosis during the previous month.
proportional distributions from the US dated cases, core demographic and di- All medicines, except long-acting
Census for the health plan catchment agnostic information, including date of insulin, were discontinued the night
area. For the American Indian reserva- birth, sex, race/ethnicity, date of diag- before the visit. Mean (SD) DM dura-
tion-based populations, denomina- nosis, and DM type, were obtained from tion at the visit was 11.74 (7.25)
tors were defined by the health service medical records, usually as part of the months and was similar for subgroups
user population for the eligible ser- case validation process. Date of birth, of race/ethnicity and DM type.
©2007 American Medical Association. All rights reserved. (Reprinted) JAMA, June 27, 2007—Vol 297, No. 24 2717
INCIDENCE OF DIABETES IN US YOUTH

were calculated per 100 000 person-

Table 1. Number of Youth With Nonsecondary DM (2002-2003), Population Denominators,
and Incidence Rates by Age Group, Sex, and Race/Ethnicity years at risk. Ninety-five percent con-
Population Incidence Rate fidence intervals (CIs) were calcu-
No. of Youth Denominator, per 100 000 lated on the basis of inverting the score
Characteristics With DM* Person-Years Person-Years (95% CI) test for a binomial proportion.30 All sta-
Total population 2435 10 031 888 24.3 (23.3-25.3) tistical analyses were performed by
Age group, y
0-4 345 2 405 348 14.3 (12.9-15.9)
using SAS version 9.0 (SAS Institute Inc,
5-9 560 2 446 750 22.9 (21.1-24.9)
Cary, NC). P⬍.05 was considered sta-
10-14 903 2 661 278 33.9 (31.8-36.2) tistically significant.
15-19 627 2 518 512 24.9 (23.0-26.9)
RESULTS
Sex
Male 1193 5 123 956 23.3 (22.0-24.6) A total of 2561 newly diagnosed patients
Female 1242 4 907 932 25.3 (23.9-26.8) aged younger than 20 years in 2002
Race/ethnicity† (n=1325) and 2003 (n=1236) were as-
Non-Hispanic white 1545 5 928 400 26.1 (24.8-27.4)
certained, which included 1905 youth
African American 365 1 434 750 25.4 (23.0-28.2)
with type 1 DM, 530 youth with type 2
Hispanic 323 1 603 502 20.2 (18.1-22.5)
DM, and 126 youth with other DM types
Asian/Pacific Islander 131 780 110 16.7 (14.1-19.9)
(62 youth with secondary forms of DM
American Indian 71 285 126 25.0 (19.8-31.5)
and 64 youth with other or hybrid types
Abbreviations: CI, confidence interval; DM, diabetes mellitus.
*Includes 1905 youth with type 1 DM and 530 youth with type 2 DM. or unknown or missing information on
†Based on self-report or medical record−based data for 95.9% of cases, and on US Census block-level geocoding for
the 4.1% cases with missing race/ethnicity.
DM type). Cases were ascertained at sites
covering populations of varying sizes
(Ohio:355casesamong1.1millionpopu-
Blood specimens were processed lo- professional). For each center, cases were lation; South Carolina: 539 among 2.2
cally and shipped within 24 hours to the identified from multiple sources (from million;Washington:509among1.9mil-
central laboratory (University of Wash- 13 to 41). A source was defined as any lion; Colorado: 655 among 2.5 million;
ington, Seattle). Samples were analyzed location in which cases were reported. American Indian reservations: 58 among
for glutamic acid decarboxylase antibod- Matching across sources to identify po- 200 000; California: 342 among 1.5 mil-
ies in radioligand-binding assays.24,25 The tential duplicate records was per- lion; and Hawaii: 103 among 500 000).
levels were expressed as a glutamic acid formed on a regular basis at the center Completeness of ascertainment was es-
decarboxylase antibodies index: counts level using available personal health iden- timated to be 93% across all 4 geographi-
perminute(cpm)oftheunknownsample tifiers. Once matching across sources was cally based sites and varied little with site
minus average cpm of 2 negative stan- accomplished, the sources were further (from 87% to 99%). Ascertainment was
dards divided by cpm of the positive grouped into 2 modes of ascertainment loweramongthe15-to19-year-oldgroup
standard minus average of 2 negative (health care professionals and in- (87%) than among the 0- to 4-year-old
standards. Fasting C peptide levels were patient hospital system records). The per- (95%), 5- to 9-year-old (94%), and 10-
measured by radioimmunoassay.26 The centage completeness of ascertainment to 14-year-old (93%) age groups.
assay precision had a coefficient of varia- for each site was taken as the number of A total of 2435 youth with nonsec-
tion between 6.6% and 10.7%, and a sen- observed cases divided by the number es- ondary DM newly diagnosed in 2002-
sitivity limit of 0.15 ng/mL. timated from the capture-recapture 2003 were identified across all study lo-
The study was approved by the appro- method. Estimates of the ascertain- cations in a population of more than 10
priate institutional review boards and ment rates pooled across clinical sites million person-years (TABLE 1). Over-
recruitment of participants followed pro- were produced from a global log-linear all, the incidence rate (per 100 000 per-
cedures compliant with the Health Insur- model,28 which allowed for separate in- son-years) of DM was 24.3 (95% CI,
ance Portability and Accountability Act. trasite performance. The rates were es- 23.3-25.3). The incidence rate was
All parents or participants that came to timated using maximum likelihood and highest among 10- to 14-year-old youth
thein-personvisitsignedage-appropriate the SEs used the ␦ method.29 Approxi- (33.9; 95% CI, 31.8-36.2), and slightly
informed consent or assent forms. mately 20% of the study population was higher in females vs males (25.3; 95%
ascertained in membership-based sites CI, 23.9-26.8; vs 23.3; 95% CI, 22.0-
Statistical Analyses where it was impossible to assess com- 24.6). Overall, the highest incidence
Completeness of case ascertainment was pleteness of ascertainment using capture- rates of DM were observed among non-
assessed for 2002 for geographically recapture analyses due to the lack of in- Hispanic white (26.1; 95% CI, 24.8-
based centers using the capture- dependent sources of case ascertainment. 27.4), African American (25.4; 95% CI,
recapture method27 and a 2-mode ascer- Incidence rates by age group, sex, 23.0-28.2), and American Indian youth
tainment model (hospital vs health care race/ethnicity, and clinical DM type (25.0; 95% CI, 19.8-31.5), with lower
2718 JAMA, June 27, 2007—Vol 297, No. 24 (Reprinted) ©2007 American Medical Association. All rights reserved.
 INCIDENCE OF DIABETES IN US YOUTH

rates among Hispanic (20.2; 95% CI, Pacific Islander (11.8 and 22.7, respec- FIGURE 1 and FIGURE 2 show race/
18.1-22.5) and Asian/Pacific Islander tively), and Hispanic youth (8.9 and ethnicity-specific incidence estimates of
youth (16.7; 95% CI, 14.1-19.9). 17.0, respectively), and were lowest type 1 DM and type 2 DM, respectively,
TABLE 2 shows incidence estimates among non-Hispanic white youth (3.0 by5-yearage groupsand sex. Thefemale-
(per 100 000 person-years) of DM by and 5.6, respectively). to-male incidence rate ratio (RR) was cal-
5-year age groups, race/ethnicity, and
DM type (type 1 and type 2). For chil-
dren aged 0 to 4 years and 5 to 9 years, Table 2. Incidence Rates of Diabetes (2002-2003) per 100 000 Person-Years by Age Group,
Race/Ethnicity, and Diabetes Mellitus Type
most DM was type 1, regardless of race/
Type 1 Diabetes Type 2 Diabetes
ethnicity. The incidence of type 1 DM
was highest among non-Hispanic white Total No. of No. of Incidence Rate No. of Incidence Rate
children (18.6 for 0-4 years and 28.1 Age Group Youth Youth (95% CI) Youth (95% CI)
for 5-9 years), and lowest among Ameri- 0-4 y
Non-Hispanic white 696 338 259 18.6 (16.5-21.0) 0 0.0 (0.0-0.3)
can Indian (4.1 and 5.5, respectively) African American 171 753 33 9.7 (6.9-13.6) 0 0.0 (0.0-1.1)
and Asian/Pacific Islander children (6.1 Hispanic 209 846 38 9.1 (6.7-12.5) 0 0.0 (0.0-0.9)
and 8.0, respectively), with intermedi- Asian/Pacific Islander 92 650 11 6.1 (3.5-10.9) 0 0.0 (0.0-2.1)
ate rates among Hispanic (9.1 and 15.7, American Indian 32 087 3 4.1 (1.3-13.0) 0 0.0 (0.0-6.0)
respectively) and African American All 1 202 674 345 14.3 (12.9-15.9) 0 0.0 (0.0-0.2)
children (9.7 and 16.2, respectively). 5-9 y
No children aged 0 to 4 years and only Non-Hispanic white 714 238 401 28.1 (25.5-31.0) 4 0.3 (0.1-0.7)
19 children aged 5 to 9 years had type African American 173 942 56 16.2 (12.5-21.0) 6 1.7 (0.8-3.7)
2 DM. Hispanic 204 809 65 15.7 (12.3-20.1) 5 1.3 (0.6-2.9)
Similarly, for older youth (10-14 Asian/Pacific Islander 95 675 15 8.0 (4.9-13.1) 4 2.2 (0.9-5.6)
years and 15-19 years), the incidence American Indian 34 712 4 5.5 (2.1-14.5) 0 0.0 (0.0-5.6)
of type 1 DM (per 100 000 person- All 1 223 376 541 22.1 (20.3-24.1) 19 0.8 (0.5-1.2)
years) was highest among non- 10-14 y
Non-Hispanic white 787 605 518 32.9 (30.2-35.8) 47 3.0 (2.3-4.0)
Hispanic white children (32.9 and
African American 194 635 75 19.2 (15.4-24.1) 87 22.3 (18.1-27.5)
15.1, respectively), followed by Afri-
Hispanic 205 436 72 17.6 (14.0-22.2) 37 8.9 (6.4-12.3)
can American (19.2 and 11.1, respec-
Asian/Pacific Islander 103 233 17 8.3 (5.2-13.3) 24 11.8 (7.9-17.5)
tively) and Hispanic youth (17.6 and
American Indian 39 729 6 7.1 (3.2-15.8) 20 25.3 (16.4-39.0)
12.1, respectively), and lowest among
All 1 330 638 688 25.9 (24.0-27.9) 215 8.1 (7.1-9.2)
American Indian (7.1 and 4.8, respec- 15-19 y
tively) and Asian/Pacific Islander Non-Hispanic white 766 019 231 15.1 (13.2-17.1) 85 5.6 (4.5-6.9)
youth (8.3 and 6.8, respectively). The African American 177 045 39 11.1 (8.1-15.1) 69 19.4 (15.3-24.5)
rates of type 2 DM were highest Hispanic 181 660 44 12.1 (9.1-16.4) 62 17.0 (13.3-21.8)
among American Indian youth (25.3 Asian/Pacific Islander 98 497 13 6.8 (4.0-11.5) 45 22.7 (16.9-30.4)
for 10-14 years and 49.4 for 15-19 American Indian 36 035 3 4.8 (1.7-13.2) 36 49.4 (35.6-68.5)
years), followed by African American All 1 259 256 331 13.1 (11.8-14.6) 296 11.8 (10.5-13.2)
(22.3 and 19.4, respectively), Asian/ Abbreviation: CI, confidence interval.

Figure 1. Incidence of Type 1 Diabetes Mellitus by 5-Year Age Groups, Sex, and Race/Ethnicity, 2002-2003

Non-Hispanic Whites African Americans Hispanics Asian/Pacific Islanders American Indians

60
Males
50
100 000 Person-Years

Females
Incidence per

40

30

20

10

0-4 5-9 10-14 15-19 0-4 5-9 10-14 15-19 0-4 5-9 10-14 15-19 0-4 5-9 10-14 15-19 0-4 5-9 10-14 15-19
Age Group, y Age Group, y Age Group, y Age Group, y Age Group, y

Error bars represent 95% confidence intervals. In Figures 1 and 2, y-axes in blue indicate an incidence range of 0 to 60 per 100 000 person-years.

©2007 American Medical Association. All rights reserved. (Reprinted) JAMA, June 27, 2007—Vol 297, No. 24 2719
INCIDENCE OF DIABETES IN US YOUTH

culated with data pooled from all age TABLE 3 presents the proportional dis- agnosis of type 1 DM had positive glu-
groups and racial/ethnic groups, for each tribution of type 1 and type 2 DM for tamic acid decarboxylase antibodies.
DM type. For type 1 DM (Figure 1), the each racial/ethnic group. Among youth For youth with a clinical diagnosis of
rates were very similar in females and younger than 10 years at diagnosis, most type 2 DM, 1 of only 3 participants
males (RR, 1.028; 95% CI, 1.025-1.030), DM is type 1, regardless of race/ (33.3%) in the younger age group (⬍10
although due to the large sample size, ethnicity. Among youth aged 10 years or years) had positive glutamic acid de-
the difference reached statistical signifi- older at diagnosis, type 1 DM repre- carboxylase antibodies, although in the
cance. Overall, across all racial/ethnic sents the major type among non- older age group (ⱖ10 years), 32
groups and sex, the highest rates of type Hispanic white adolescents (85.1%). In (21.2%) had positive glutamic acid de-
1 DM were observed among 5- to 9- addition, a notable proportion of minor- carboxylase antibodies. Overall, mean
year-old and 10- to 14-year-old youth ity adolescents (53.9% of Hispanic, (SD) fasting C peptide level was sig-
(P⬍.001 for each age group vs 0-4 years 42.2% of African American, 30.3% of nificantly higher in youth with type 2
and 15-19 years), although this was Asian/Pacific Islander, and 13.8% of DM than in those with type 1 DM, re-
largely driven by the age pattern in non- American Indian) have type 1 DM. As ex- gardless of age group (0-9 years: 1.80
Hispanic white youth. The incidence rate pected, type 2 DM is most common (1.42) vs 0.43 (0.48) ng/mL and 10-19
of type 2 DM (Figure 2) was higher in among minorities aged 10 to 19 years, years: 3.52 (2.12) vs 0.78 (0.65) ng/
females than in males (RR, 1.63; 95% CI, especially American Indian (86.2%) and mL, respectively; P⬍.001 for each com-
1.58-1.67; P⬍.001). Across all racial/ Asian/Pacific Islander (69.7%), but also parison). In addition, participants with
ethnicgroupsandsex,theincidencerates among African American (57.8%) and a clinical diagnosis of type 1 DM had
of type 2 DM were higher among 15- to Hispanic (46.1%) youth. similar fasting C peptide levels, and
19-year old youth than among 10- to TABLE 4 shows characteristics of similar proportions of nondetectable
14-year-old youth (P⬍.001 for non- youth with DM, by DM type and age fasting C peptide levels (ⱕ0.2 ng/mL
Hispanic whites, Hispanics, Asian/Pacific group, among participants to the re- [ⱕ0.07 nmol/L]) and current insulin
Islanders, and American Indians). This search visit. For both younger (⬍10 use, regardless of glutamic acid decar-
pattern was not consistent among years) and older (ⱖ10 years) youth, 238 boxylase antibody status. Fasting C pep-
African American youth with type (56.4%) and 330 (65.6%) partici- tide level and current insulin use also
2 DM. pants, respectively, with a clinical di- did not substantially differ according to

Figure 2. Incidence of Type 2 Diabetes Mellitus by 5-Year Age Groups, Sex, and Race/Ethnicity, 2002-2003

Non-Hispanic Whites African Americans Hispanics Asian/Pacific Islanders American Indians

90
Males
80
100 000 Person-Years

Females
70
Incidence per

60
50
40
30
20
10
0

0-4 5-9 10-14 15-19 0-4 5-9 10-14 15-19 0-4 5-9 10-14 15-19 0-4 5-9 10-14 15-19 0-4 5-9 10-14 15-19
Age Group, y Age Group, y Age Group, y Age Group, y Age Group, y

Error bars represent 95% confidence intervals. In Figures 1 and 2, y-axes in blue indicate an incidence range of 0 to 60 per 100 000 person-years.

Table 3. Proportional Distribution of Type 1 and Type 2 DM (2002-2003) by Age Group at Diagnosis and Race/Ethnicity
No. (%) of Youth

0-9 y 10-19 y

Non- Asian/ Non- Asian/

Hispanic African Pacific American Hispanic African Pacific American
DM White American Hispanic Islander Indian White American Hispanic Islander Indian
Type 1 660 (99.4) 89 (93.7) 102 (95.3) 26 (86.7) 7 (100) 749 (85.1) 114 (42.2) 116 (53.9) 30 (30.3) 9 (13.8)
Type 2 4 (0.6) 6 (6.3) 5 (4.7) 4 (13.3) 0 132 (14.9) 156 (57.8) 99 (46.1) 69 (69.7) 56 (86.2)
Total 664 (100) 95 (100) 107 (100) 30 (100) 7 (100) 881 (100) 270 (100) 215 (100) 99 (100) 65 (100)
Abbreviation: DM, diabetes mellitus.

2720 JAMA, June 27, 2007—Vol 297, No. 24 (Reprinted) ©2007 American Medical Association. All rights reserved.
 INCIDENCE OF DIABETES IN US YOUTH

glutamic acid decarboxylase antibody

Table 4. Biochemical Characteristics of Youth With DM (2002-2003), by Clinical DM Type,
status among youth with type 2 DM. Among Participants to the Research Visit
Type 1 DM Type 2 DM
COMMENT
With a population of more than 10 mil- GAD65 GAD65 GAD65 GAD65
⫹ − ⫹ −
lionperson-yearsforwhichDMincidence Age 0-9 y
isestimated,theSEARCHstudyrepresents No./Total No. (%) 238/422 (56.4) 184/422 (43.6) 1/3 (33.3) 2/3 (66.7)
the largest standardized registry of child- Fasting C peptide, 0.46 (0.54) 0.40 (0.38) 3.4 1.0 (0.42)
hood DM in the United States. In this mean (SD), ng/mL
study, which encompasses a large mul- Fasting C peptide 135 (56.7) 106 (57.6) 0 0
tiethnic population, the vast majority of ⱕ0.2 ng/mL, No. (%)
all new cases of DM in children younger Taking insulin, No. (%) 237 (99.6) 184 (100) 0 1 (50)

than 10 years had type 1 DM, regardless Age 10-19 y

ofrace/ethnicity.Evenamongolderyouth No./Total No. (%) 330/503 (65.6) 173/503 (34.4) 32/151 (21.2) 119/151 (78.8)
(ⱖ10years),type1DMisproportionately Fasting C peptide, 0.75 (0.60) 0.82 (0.74) 2.83 (1.8) 3.71 (2.2)
mean (SD), ng/mL
the most common form of DM for youth
Fasting C peptide 80 (24.2) 34 (19.7) 1 (3.1) 0
of non-Hispanic white and Hispanic ori- ⱕ0.2 ng/mL, No. (%)
gin. Although type 2 DM is still relatively Taking insulin, No. (%) 324 (98.2) 173 (100) 14 (43.8) 37 (31.1)
infrequent overall, it becomes more com- Abbreviations: DM, diabetes mellitus; GAD65, glutamic acid decarboxylase antibody.
SI conversion: To convert fasting C peptide to nmol/L, multiply by 0.333.
monafter10yearsofage,withhigherrates
among US minority populations than
among non-Hispanic white populations. studies difficult. However, taken to- lescents presenting to a Florida clinic
The SEARCH study estimates of type gether, these data suggest that the inci- with DM between 1994 and 1998,37 the
1 DM incidence are higher than the in- dence of type 1 DM may be increasing proportion with type 2 DM increased
cidence of insulin-dependent DM re- in the United States, consistent with from 9.4% of new cases to 20% of new
ported for the period 1990-1994 by the worldwide trends.7,34 In agreement with cases during the 5-year period. Con-
Diamond Study2 among children aged previous data,35,36 the SEARCH study sistent with previous articles,13,38 the
0-4, 5-9, and 10-14 years. Our rates of shows that incidence rates of type 1 DM SEARCH study demonstrates that type
type 1 DM are also 15% to 40% higher peak at ages 5 to 9 years and 10 to 14 2 DM contributes considerably to the
than insulin-dependent DM rates from years, and the risk is similar for males and overall DM incidence among minority
Allegheny County, across all ages for females. We estimate that the annual youth aged 10 years or older, and rates
non-Hispanic white and among 0- to number of newly diagnosed youth with are approximately 60% higher in fe-
9-year-old African American children, type 1 DM in the United States is ap- males than in males. Well-designed
but are lower than insulin-dependent proximately 15 000. studies from Europe39-41 indicate that
DM rates among 10- to 19-year-old Afri- There are limited population-based type 2 DM remains a rarity in these
can American youth.31 Similarly, the data on the incidence of type 2 DM in populations, accounting for only 1% to
SEARCH study rates of type 1 DM are youth, which makes comparison with 2% of all DM cases. In contrast, al-
30% to 80% higher than recently re- other studies difficult. Using data from though the SEARCH study data sup-
ported insulin-dependent DM rates from the medical records of 735 African port the notion that type 2 DM in youth
Philadelphia, across all ages for non- American and Latino children with type is predominantly occurring in high-
Hispanic white and approximately 20% 2 DM in Chicago,12 the incidence was risk ethnic groups, type 2 DM ac-
higher among 0- to 9-year-old African higher in African American vs Latino counts for 14.9% of all DM cases among
American children, but are lower than children (5.0 per 100 000 person- non-Hispanic white adolescents aged 10
insulin-dependent DM rates in 10- to 14- years in African American girls and 2.7 years or older. Although differences in
year-old African American youth.32 per 100 000 person-years in African obesity rates between US and Euro-
Among the Hispanic population, the American boys vs 2.4 per 100 000 per- pean youth are likely contributors, the
SEARCH study rates of type 1 DM are son-years in Latino girls and 1.8 per full explanation for these discrepan-
similar to insulin-dependent DM rates re- 100 000 person-years in Latino boys). cies remains uncertain. The SEARCH
ported in the 1990s for Puerto Rican Among 1027 consecutive patients with study estimates that the annual num-
Americans from Philadelphia,32,33 but are DM attending a diabetes clinic in Cin- ber of newly diagnosed youth with type
approximately 40% higher than those re- cinnati,17 a 10-fold increase in type 2 2 DM in the United States is approxi-
ported in the 1980s for Hispanics from DM incidence rates (from 0.7 per mately 3700.
Colorado.9 Different ascertainment meth- 100 000 person-years in 1982 to 7.2 per Similar to other population-based
ods and case definitions were used in 100 000 person-years in 1994) was ob- studies,2,13,42 these analyses focused on
these studies, making comparisons across served. Among 569 children and ado- type 1 and type 2 DM and used DM
©2007 American Medical Association. All rights reserved. (Reprinted) JAMA, June 27, 2007—Vol 297, No. 24 2721
INCIDENCE OF DIABETES IN US YOUTH

type−assignments made by health care tempt to assess how much undiag- 100 000 person-years) by age and DM
professionals. This raises the issue of nosed DM exists among youth and did type based only on the 4 geographic
potential variation in health care pro- not screen for undiagnosed DM. We sites were similar to those computed
fessionals’ diagnostic norms across may, therefore, have underestimated the with data pooled across all centers (0-9
study locations. However, across all the true risk of type 2 DM in youth; how- years: for type 1 DM, 19.9 vs 18.3 per
SEARCH study sites, more than 60% of ever, limited screening studies sug- 100 000 person-years and for type 2
cases were reported by DM specialists gest that undiagnosed type 2 DM is rela- DM, 0.3 vs 0.4 per 100 000 person-
(pediatric or adult endocrinologists), for tively rare in youth.46 years; and 10-19 years: for type 1 DM,
which such variation is unlikely to be We used the capture-recapture 21.1 vs 19.7 per 100 000 person-years
substantial. In addition, DM type was method to estimate completeness of as- and for type 2 DM, 8.8 vs 9.9 per
further characterized in a subset of certainment; however, in the context of 100 000 person-years). This suggests
youth participating in the research visit, the current US health care system and that completeness of ascertainment was
using measurements of glutamic acid privacy regulations, several limita- equally high in membership-based and
decarboxylase antibodies and fasting C tions of the method were encoun- geographic-based sites.
peptide levels. With a clinical diagno- tered.27 Incomplete matching across In conclusion, our data document the
sis of type 1 DM, glutamic acid decar- sources due to restrictions on access to incidence rates of type 1 DM among
boxylase antibody positivity was ob- names in some sites and design of the youth of all racial/ethnic groups. The
served in more than 50% of individuals, case ascertainment system for effi- incidence of type 1 DM among non-
regardless of age. The glutamic acid ciency (thus avoiding sources of likely Hispanic white youth now exceeds 20
decarboxylase−negative participants duplicate cases) have been shown to per 100 000 person-years compared
with type 1 DM include patients who lead to an underestimate of complete- with 16.5 per 100 000 person-years in
may have lost glutamic acid decarbox- ness as assessed by the capture- Allegheny County in the early 1990s.31
ylase positivity, who may be positive for recapture method.27 We therefore be- Type 2 DM was found among adoles-
other autoantibodies, such as insuli- lieve that our estimates represent the cents of all racial/ethnic groups. Al-
noma-associated antibody or insulin au- lower bound on the completeness of as- though the evidence of the presence of
toantibody, who may have a form of un- certainment in the SEARCH study. The type 2 DM in youth is still developing,
diagnosed monogenic DM, or other analysis indicates a lower complete- it is consistent with the increasing
causes of insulin deficiency, as sug- ness among 15- to 19-year-old youth prevalence of type 2 DM in adults, and
gested by the American Diabetes Asso- (87%), a group with higher incidence the increasing prevalence of obesity in
ciation.43 With a clinical diagnosis of of type 2 DM. Had all these youth been both adults and children. Overall, type
type 2 DM, and similar to other smaller identified, 44 more cases with type 2 2 DM is still relatively infrequent in US
US studies,44,45 21.2% of the SEARCH DM would have been added, for an youth; however, the highest rates are
study participants aged 10 years or older overall unadjusted rate of type 2 DM observed among 15- to 19-year-old ado-
had positive glutamic acid decarbox- among 15- to 19-year-old adolescents lescent minorities, especially Ameri-
ylase antibodies. The majority of par- of 13.5 per 100 000 person-years. can Indian youth (49.4 per 100 000 per-
ticipants with type 2 DM and positive Although some of the SEARCH study son-years).
glutamic acid decarboxylase antibod- centers are membership-based, their The SEARCH study provides
ies were overweight (more than 75% base populations are very representa- unique population-based data on the
with a body mass index, calculated as tive of the geographic areas in which incidence of DM among youth of vari-
weight in kilograms divided by height they are located. The Kaiser Perma- ous racial/ethnic backgrounds, accord-
in meters squared, higher than the 95th nente Southern California member- ing to DM type. Continuing this sur-
percentile), of minority racial/ethnic ship is representative of the popula- veillance effort will document
background (68% minorities), and tion living in the greater Los Angeles temporal trends in the incidence of
more than half had glutamic acid de- metropolitan area with regard to de- DM among various racial/ethnic
carboxylase antibody titers less than 2 mographics, ethnicity, and socioeco- groups and accurately assess the
times the cutpoint used to define posi- nomic status. The Hawaii Medical Ser- future health care burden of DM and
tivity. This suggests that most of these vice Association, MedQuest, and Kaiser its complications in the US pediatric
participants have type 2 DM. Never- Permanente Hawaii cover more than and young adult population.
theless, the role of DM-related autoan- 90% of the Hawaiian population. In ad- SEARCH for Diabetes in Youth Study Writing Group:
tibody positivity in the etiology and dition, the capture-recapture method Dana Dabelea, MD, PhD (University of Colorado Health
natural evolution of DM among minor- could not be used in the membership Sciences Center, Denver); Ronny A. Bell, PhD, MS
(Wake Forest University School of Medicine, Winston-
ity youth with a clinical phenotype of sites because they had essentially 1 com- Salem, NC); Ralph B. D’Agostino,Jr, PhD (Wake For-
type 2 DM requires further exploration. bined reporting source for cases rather est University School of Medicine, Winston-Salem, NC);
Giuseppina Imperatore, MD, PhD (Centers for Dis-
This study has several potential limi- than the required 2 or more sources. ease Control and Prevention, Atlanta, Ga); Judith M.
tations. The SEARCH study did not at- However, the risk estimates (per Johansen, RN, BSN (Cincinnati Children’s Hospital

2722 JAMA, June 27, 2007—Vol 297, No. 24 (Reprinted) ©2007 American Medical Association. All rights reserved.
 INCIDENCE OF DIABETES IN US YOUTH

Medical Center, Cincinnati, Ohio); Barbara Linder, MD, tion Program). Hawaii: Beatriz L. Rodriguez, MD, PhD, in the United States? Diabetes Care. 1993;16:1606-
PhD (National Institute of Diabetes and Digestive and Beth Waitzfelder, PhD, Wilfred Fujimoto, MD, J. David 1611.
Kidney Diseases, Bethesda, Md); Lenna L. Liu, MD, Curb, MD, Fiona Kennedy, RN, Greg Uramoto, MD, 7. EURODIAB ACE Study Group. Variation and trends
MPH (University of Washington Child Health Insti- Sorrell Waxman, MD, Teresa Hillier, MD, Richard in incidence of childhood diabetes in Europe. Lancet.
tute, Seattle); Beth Loots, PhD (Children’s Hospital and Chung, MD (The Pacific Health Research Institute). 2000;355:873-876.
Regional Medical Center, Seattle, Wash); Santica Ohio: Lawrence M. Dolan, MD, Debra Standiford, 8. LaPorte RE, Fishbein HA, Drash AL, Kuller LH. The
Marcovina, PhD (University of Washington, Seattle); MSN, CNP, Stephen R. Daniels, MD, PhD, Judith M. Pittsburgh Insulin-Dependent Diabetes Mellitus (IDDM)
Elizabeth J. Mayer-Davis, MSPH, PhD (University of Johansen, RN, BSN (Cincinnati Children’s Hospital Registry: the incidence of insulin-dependent diabe-
South Carolina, Columbia); David J. Pettitt, MD (San- Medical Center). South Carolina: Elizabeth J. Mayer- tes mellitus in Allegheny County, Pennsylvania
sum Diabetes Research Institute, Santa Barbara, Calif ); Davis, PhD, Angela D. Liese, PhD, MPH, Robert (1965-1976). Diabetes. 1981;30:279-284.
and Beth Waitzfelder, PhD (Pacific Health Research McKeown, PhD, John Oeltmann, PhD, Joan Thomas 9. Kostraba JN, Gay EC, Cai Y, et al. Incidence of in-
Institute, Honolulu, HI). MS, RD, Deborah Truell, RN, CDE, Gladys Gaillard- sulin-dependent diabetes mellitus in Colorado.
Author Contributions: Dr Dabelea had full access to McBride, RN, CFNP, Deborah Lawler, MT (ASCP); Epidemiology. 1992;3:232-238.
all of the data in the study and takes responsibility for Lynne Hartel, MA, Robert R. Moran, PhD, April Irby, 10. Dabelea D, Pettitt DJ, Jones KL, Arslanian SA. Type
the integrity of the data and the accuracy of the data BS, James Amrhein, MD, Yaw Appiagyei-Dankah, MD, 2 diabetes mellitus in minority children and adoles-
analysis. Pam Clark, MD, Howard Heinze, MD, Lyndon Key, cents: an emerging problem. Endocrinol Metab Clin
Study concept and design: Dabelea, D’Agostino, MD, Andy Muir, MD, Mark Parker, MD, I David North Am. 1999;28:709-729.
Imperatore, Linder, Liu, Marcovina, Mayer-Davis, Schwartz, MD, Steve Willi, MD. Washington: Cather- 11. Kitagawa T, Mano T, Fujita H. The epidemiol-
Pettitt. ine Pihoker, MD, Lisa Gilliam, MD, PhD, Irl Hirsch, MD, ogy of childhood diabetes mellitus in Tokyo metro-
Acquisition of data: Dabelea, Johansen, Liu, Lenna L. Liu, MD, MPH, Carolyn Paris, MD, MPH (Uni- politan area. Tohoku J Exp Med. 1983;141:171-179.
Mayer-Davis, Waitzfelder. versity of Washington); Beth Loots, MPH, MSW, Jenny 12. Lipton R, Keenan H, Onyemere KU, Freels S. In-
Analysis and interpretation of data: Dabelea, Bell, Tseng, PhD, and Shirley Vacanti, RN, BSN (Seattle Chil- cidence and onset features of diabetes in African-
D’Agostino, Imperatore, Linder, Liu, Loots, Marcovina, dren’s Hospital and Regional Medical Center); and American and Latino children in Chicago, 1985-1994.
Mayer-Davis, Pettitt. Carla Greenbaum, MD (Benaroya Research Insti- Diabetes Metab Res Rev. 2002;18:135-142.
Drafting of the manuscript: Dabelea, D’Agostino, tute); Centers for Disease Control and Prevention: Gi- 13. Dabelea D, Hanson RL, Bennett PH, Roumain J,
Imperatore, Liu, Mayer-Davis, Pettitt. useppina Imperatore, MD, PhD, Desmond E. Wil- Knowler WC, Pettitt DJ. Increasing prevalence of type
Critical revision of the manuscript for important in- liams, MD, PhD, Michael M. Engelgau, MD, Henry II diabetes in American Indian children. Diabetologia.
tellectual content: Bell, D’Agostino, Imperatore, Kahn, MD, and K. M. Venkat Narayan, MD, MPH; 1998;41:904-910.
Johansen, Linder, Liu, Loots, Marcovina, Mayer-Davis, National Institute of Diabetes and Digestive and Kid- 14. Dean HJ, Young TK, Flett B, Wood-Steiman P.
Pettitt, Waitzfelder. ney Diseases, National Institutes of Health: Barbara Screening for type-2 diabetes in aboriginal children in
Statistical analysis: Dabelea, D’Agostino, Imperatore. Linder, MD, PhD; Central Laboratory (University of northern Canada. Lancet. 1998;352:1523-1524.
Washington): Santica Marcovina, PhD, and Greg 15. Harris SB, Gittelsohn J, Hanley A, et al. The preva-
Obtained funding: Dabelea, Bell, Linder, Mayer-Davis,
Strylewicz, MS; Coordinating Center (Wake Forest Uni- lence of NIDDM and associated risk factors in native
Waitzfelder.
versity School of Medicine): Ronny Bell, PhD, MS, Canadians. Diabetes Care. 1997;20:185-187.
Administrative, technical, or material support: Bell,
Ralph D’Agostino, Jr, PhD, Timothy Morgan, PhD, Su- 16. Pihoker C, Scott CR, Lensing SY, Cradock MM,
Johansen, Marcovina, Mayer-Davis.
san Vestal, BS, and Bharathi Zvara, BS. Smith J. Non-insulin dependent diabetes mellitus in
Study supervision: Dabelea, Mayer-Davis. Disclaimer: The contents of this article are solely the
Financial Disclosures: None reported. African-American youths of Arkansas. Clin Pediatr
responsibility of the authors and do not necessarily rep-
Funding/Support: The SEARCH for Diabetes in Youth (Phila). 1998;37:97-102.
resent the official views of the funding agencies.
Study is funded by the Centers for Disease Control and 17. Pinhas-Hamiel O, Dolan LM, Daniels SR, Stan-
Acknowledgment: The SEARCH for Diabetes in Youth
Prevention through a cooperative agreement (PA No. diford D, Khoury PR, Zeitler P. Increased incidence of
Study is indebted to the many youth and their fami-
00097 and DP-05-069) and supported by the Na- non-insulin-dependent diabetes mellitus among
lies, and their health care professionals, whose par-
tional Institute of Diabetes and Digestive and Kidney adolescents. J Pediatr. 1996;128:608-615.
ticipation made this study possible. We thank the Gen-
Diseases. The site contract numbers are California (U01 18. The SEARCH Study Group. SEARCH for Diabe-
eral Clinical Research Centers at the following
DP000246), Colorado (U01 DP000247), Hawaii (U01 tes in Youth: a multicenter study of the prevalence,
institutions for their involvement in the SEARCH for
DP000245), Ohio (U01 DP000248), South Carolina incidence and classification of diabetes mellitus in youth.
Diabetes in Youth Study: Medical University of South
(U01 DP000254), Washington (U01 DP000244), and Control Clin Trials. 2004;25:458-471.
Carolina (grant M01 RR01070); Cincinnati Chil-
the Coordinating Center (U01 DP000250). The co- dren’s Hospital (grant M01 RR08084); Children’s Hos- 19. Knowler WC, Bennett PH, Hamman RF, Miller
operative agreement between study centers and fund- pital and Regional Medical Center and the University M. Diabetes incidence and prevalence in Pima Indi-
ing agencies supported the work of study investiga- of Washington School of Medicine (grants ans: a 19-fold greater incidence than in Rochester,
tors, collaborators, and staff, including the writing M01RR00037 and M01RR001271); and the Colo- Minnesota. Am J Epidemiol. 1978;108:497-505.
group members and the individuals named in the Ac- rado Pediatric General Clinical Research Center (grant 20. Liese AD, D’Agostino RB Jr, Hamman RF, et al;
knowledgment section. M01 RR00069). SEARCH for Diabetes in Youth Study Group. The bur-
Role of the Sponsors: Through the cooperative agree- den of diabetes among U.S. youth: prevalence esti-
ment mechanisms, investigators from the 2 funding mates from the SEARCH for Diabetes in Youth Study.
agencies participated in the design of the study, in the Pediatrics. 2006;118:1510-1518.
REFERENCES
analysis and interpretation of the data, and in the 21. Chen W, Petitti DB, Enger S. Limitations and po-
preparation, review, and approval of the study manu- 1. Green A, Gale EAM, Patterson CC. Incidence of tential uses of census-based data on ethnicity in a di-
script. They did not participate in the data collection childhood-onset insulin-dependent diabetes melli- verse community. Ann Epidemiol. 2004;14:339-345.
and management component of the study. tus: the EURODIAB ACE study. Lancet. 1992;339:905- 22. Ingram DD, Parker JD, Schenker N, et al. United
The SEARCH for Diabetes in Youth Study Writing 909. States Census 2000 population with bridged race cat-
Group acknowledges the contributions of the fol- 2. Karvonen M, Viik-Kajander M, Moltchanova E, Lib- egories: data evaluation and methods research. Vital
lowing individuals to the work of the SEARCH for Dia- man I, LaPorte R, Tuomilehto J. Incidence of child- Health Stat 2. 2003;(135):1-55. DHHS Publication No
betes in Youth Study: California: Jean M. Lawrence, hood type 1 diabetes worldwide: Diabetes Mondiale (PHS) 2003-1335.
ScD, MPH, MSSA, Diana B. Petitti, MD, MPH, Ann (DiaMond) Project Group. Diabetes Care. 23. Petitti DB, Imperatore G, Palla SL, et al. Serum lip-
K. Kershnar, MD, and Kimberly J. Holmquist, BS (Kai- 2000;23:1516-1526. ids and glucose control: the SEARCH for Diabetes in
ser Permanente Southern California); and David J. Pet- 3. Rewers M, LaPorte R, Walczak M, Dmochowski Youth Study. Arch Pediatr Adolesc Med. 2007;161:
titt, MD (The Sansum Diabetes Research Institute). K, Bogaczynska E. Apparent epidemic of insulin- 159-165.
Colorado: Dana Dabelea, MD, PhD, Richard F. Ham- dependent diabetes mellitus in midwestern Poland. 24. Marcovina SM, Landin-Olsson M, Essen-Moller
man, MD, DrPH, Emelin Martinez, MSN, FNP, and Lisa Diabetes. 1987;36:106-113. A, Palmer JP, Lernmark A. Evaluation of a novel ra-
Testaverde, BS (Department of Preventive Medicine 4. Diabetes Epidemiology Research International dioimmunoassay using 125I-labelled human recom-
and Biometrics, University of Colorado at Denver and (DERI) Group. Secular trends in incidence of chilhood binant GAD65 for the determination of glutamic acid
Health Sciences Center); Georgeanna J. Klingen- IDDM in 10 countries. Diabetes. 1990;39:858- decarboxylase (GAD65) autoantibodies. Int J Clin Lab
smith, MD, and Marian J. Rewers, MD, PhD (The Bar- 864. Res. 2000;30:21-26.
bara Davis Center for Childhood Diabetes); Clifford 5. Tuomilehto J, Virtala E, Karvonen M, et al. In- 25. Falorni A, Ortqvist E, Persson B, Lernmark A. Ra-
A. Block, MD (The Pediatric Endocrine Associates); crease in incidence of insulin-dependent diabetes melli- dioimmunoassays for glutamic acid decarboxylase
Jonathan Krakoff, MD, and Peter H. Bennett, MB, FRCP tus among children in Finland. Int J Epidemiol. 1995; (GAD65) and GAD65 autoantibodies using 35S or 3H
(The NIDDK Pima Indian Study); and Joquetta A. De- 24:984-992. recombinant human ligands. J Immunol Methods.
Groat, BA (The Navajo Area Indian Health Preven- 6. Dokheel TM. An epidemic of childhood diabetes 1995;186:89-99.

©2007 American Medical Association. All rights reserved. (Reprinted) JAMA, June 27, 2007—Vol 297, No. 24 2723
INCIDENCE OF DIABETES IN US YOUTH

26. Heding LG. Radioimmunological determination of 33. Lipman TH, Chang Y, Murphy KM. The epide- of diabetes in 0-30 year olds: a hospital based study
human C-peptide in serum. Diabetologia. 1975;11: miology of type 1 diabetes in children in Philadelphia in Leeds, UK. Arch Dis Child. 2003;88:676-679.
541-548. 1990-1994. Diabetes Care. 2002;25:1969-1975. 41. Ortega-Rodriguez E, Levy-Marchal C, Tubiana
27. Verlato G, Muggeo M. Capture-recapture method 34. Onkamo P, Vaananen S, Karvonen M, Tu- N, Czernichow P, Polak M. Emergence of type 2 dia-
in the epidemiology of type 2 diabetes: a contribu- omilehto J. Worldwide increase in incidence of type I betes in an hospital based cohort of children with dia-
tion from the Verona Diabetes Study. Diabetes Care. diabetes: the analysis of the data on published inci- betes mellitus. Diabetes Metab. 2001;27:574-578.
2000;23:759-764. dence trends. Diabetologia. 1999;42:1395-1403. 42. Green A, Patterson CC. Trends in the incidence
28. Bishop YMM. FSHPW: Discrete Multivariate 35. Rewers M, Norris J, Dabelea D. Epidemiology of of childhood-onset diabetes in Europe 1989-1998.
Analysis. Cambridge, Mass: MIT Press; 1975. type 1 diabetes mellitus. Adv Exp Med Biol. 2004;552: Diabetologia. 2001;44:B3-B8.
29. Espeland MA. A general class of models for dis- 219-246. 43. American Diabetes Association. Report of the Ex-
crete multivariate data. Commun Statist Simula. 1986; 36. Gale EAM, Gillespie KM. Diabetes and gender. pert Committee on the Diagnosis and Classification
15:405-424. Diabetologia. 2001;44:3-15. of Diabetes Mellitus. Diabetes Care. 1997;20:
30. Agresti A, Coull BA. Approximate is better than 37. Macaluso CJ, Bauer UE, Deeb LC, et al. Type 2 1183-1197.
“exact’’ for interval estimation of binomial proportions. diabetes mellitus among Florida children and adoles- 44. Brooks-Worrell BM, Greenbaum CJ, Palmer JP,
Am Stat. 1998;52:119-126. cents, 1994 through 1998. Public Health Rep. 2002; Pihoker C. Autoimmunity to islet proteins in children
31. Libman IM, LaPorte RE, Becker D, Dorman JS, 117:373-379. diagnosed with new-onset diabetes. J Clin Endocri-
Drash AL, Kuller L. Was there an epidemic of diabe- 38. Scott CR, Smith JM, Cradock MM, Pihoker C. Char- nol Metab. 2004;89:2222-2227.
tes in nonwhite adolescents in Allegheny County, acteristics of youth-onset noninsulin-dependent dia- 45. Hathout EH, Thomas W, El Shahawy M, Nahab
Pennsylvania? Diabetes Care. 1998;21:1278- betes mellitus and insulin-dependent diabetes melli- F, Mace JW. Diabetic autoimmune markers in chil-
1281. tus at diagnosis. Pediatrics. 1997;100:84-91. dren and adolescents with type 2 diabetes. Pediatrics.
32. Lipman TH, Jawad AF, Murphy KM, et al. Inci- 39. Schober E, Frisch H. Incidence of childhood dia- 2001;107:E102.
dence of type 1 diabetes in Philadelphia is higher in betes mellitus in Austria 1979-1984. Acta Paediatr 46. Dolan LM, Bean J, D’Alessio D, et al. Frequency
black than white children from 1995 to 1999: epi- Scand. 1988;77:299-302. of abnormal carbohydrate metabolism and diabetes
demic or misclassification? Diabetes Care. 2006; 40. Feltbower RG, McKinney PA, Campbell FM, Ste- in a population-based screening of adolescents.
29:2391-2395. phenson CR, Bodansky HJ. Type 2 and other forms J Pediatr. 2005;146:751-758.

While medicine is a science, in many particulars it can-

not be exact, so baffling are the varying results of vary-
ing conditions of human life.
—Charles H. Mayo (1865-1939)

2724 JAMA, June 27, 2007—Vol 297, No. 24 (Reprinted) ©2007 American Medical Association. All rights reserved.
 LETTERS

3. Carrasco JL, Sandner C. Clinical effects of pharmacological variations in selec- ies of individual antidepressants are needed. While the reason
tive serotonin reuptake inhibitors: an overview. Int J Clin Pract. 2005;59(12):1428-
1434. that the efficacy of fluoxetine as an antidepressant may be
4. Whittington CJ, Kendall T, Fonagy P, Cottrell D, Cotgrove A, Boddington E. superior to that of the other SSRIs is unclear, the extant data
Selective serotonin reuptake inhibitors in childhood depression: systematic review
of published versus unpublished data. Lancet. 2004;363(9418):1341-1345.
support its use as the first-line treatment for major depres-
sion in children and adolescents.
In Reply: Dr Waslick is concerned that our analysis might
Jeffrey A. Bridge, PhD
inflate the risk-benefit ratio for antidepressants by concen-
Columbus Children’s Research Institute
trating only on the risk of treatment-emergent suicidal ide- Columbus, Ohio
ation and attempts. Our intent was not to mislead, but rather Boris Birmaher, MD
to explicitly compare the benefit of antidepressants to the David A. Brent, MD
risk of treatment-emergent suicidal ideation/suicide at- [email protected]
tempt, because this is the adverse effect that is the most fright- Western Psychiatric Institute and Clinic
ening, has engendered the most negative publicity, has re- Pittsburgh, Pennsylvania
sulted in a black box warning from the US Food and Drug Financial Disclosures: Dr Bridge reported having received honoraria for an in-
Administration, and has been associated with a decrease in vited paper from Current Opinion in Psychiatry/Lippincott Williams & Wilkins and
that from 2001-2004 he participated as a coinvestigator of an open-label trial of
use of antidepressants in children and adolescents.1,2 We ex- citalopram for treatment of pediatric recurrent abdominal pain. The study was funded
plicitly acknowledged this limitation in the Comment sec- by an investigator-initiated grant from Forest Labs ( John V. Campo, MD, princi-
pal investigator); Dr Bridge reported having received no financial support of any
tion. Although we could have been clearer in defining what kind from Forest or from Dr Campo for his participation. Salary support to Dr Bridge
was meant by a risk-benefit ratio, we assume that most read- was provided by National Institute of Mental Health grants MH55123 and sub-
ers regard treatment-emergent suicidal ideation and behav- sequently MH66371, Advanced Center for Interventions and Services Research
for Early-Onset Mood and Anxiety Disorder (Dr Brent, principal investigator). Dr
ior to be in a different category of concern than discontinu- Birmaher reported having received royalties for publication of New Hope for Chil-
ation of treatment because of adverse somatic symptoms. dren and Teens with Bipolar Disorder from Random House Inc; and having re-
ceived remuneration from the University of Cincinnati for participation in the writ-
We do agree that a complete analysis of other adverse ef- ing of algorithms for the treatment of children with bipolar disorder (Kowatch RA,
fects associated with short- and long-term antidepressant Fristad M, Birmaher B, et al. Treatment guidelines for children and adolescents with
bipolar disorder. J Am Acad Child Adolesc Psychiatry. 2005;44(3):213-235), spon-
treatment is warranted. Since individual trials were all un- sored by the Child and Adolescent Bipolar Foundation and supported by unre-
derpowered to compare rates of less common adverse events, stricted educational grants from Abbott Laboratories, AstraZeneca Pharmaceuti-
cals, Eli Lilly and Co, Forest Pharmaceuticals, Janssen Pharmaceuticals, Novartis,
the pooling of individual patient data from available ran- and Pfizer. No other financial disclosures were reported.
domized controlled trials (“mega-analysis”) may be an ef-
1. Nemeroff CB, Kalali A, Keller MB, et al. Impact of publicity concerning pedi-
fective strategy for identifying clinically important, but rare, atric suicidality data on physician practice patterns in the United States. Arch Gen
safety outcomes.3 Psychiatry. 2007;64(4):466-472.
2. Libby AM, Brent DA, Morrato EH, Orton HD, Allen R, Valuck RJ. Decline in
Dr Edwards and colleagues raise the important question treatment of pediatric depression after FDA advisory of risk of suicidality with SSRIs.
of whether fluoxetine is more efficacious for major depres- Am J Psychiatry. 2007;164(6):884-891.
sion than either paroxetine or citalopram/escitalopram. 3. Thase ME, Greenhouse JB, Frank E, et al. Treatment of major depression with
psychotherapy or psychotherapy-pharmacotherapy combinations. Arch Gen
Fluoxetine is the only agent that has been shown to have Psychiatry. 1997;54(11):1009-1015.
efficacy for the treatment of depression in children younger
than 12 years, which may explain the overall difference in
efficacy compared with other agents. Several possible ex-
planatory factors may be confounded—the longer half-life CORRECTIONS
of fluoxetine; investigation in relatively more academic medi- Incorrect Comparison Group: In the Original Contribution entitled “Effects of a
cal centers compared with studies investigating other agents; Low–Glycemic Load vs Low-Fat Diet in Obese Young Adults: A Randomized Trial”
and average number of sites in the studies, which in turn published in the May 16, 2007, issue of JAMA (2007;297[19]:2092-2102), an in-
correct comparison group was provided. On page 2096, in Figure 1, the first line
may affect study quality. While we agree that an analysis of in the box on the right under “73 Randomized” should be “37 Randomized to
individual antidepressants as a potential moderator of out- Receive a Low-Fat Diet.”
come is important, such analyses at this time would not be Incorrect Author Degree: In the Original Contribution entitled “Incidence of Dia-
meaningful because of the limited number of trials con- betes in Youth in the United States” published in the June 27, 2007, issue of JAMA
(2007;297[24]:2716-2724), there was an incorrect author degree. On page 2723,
ducted for several antidepressants. Consequently, we con- in the SEARCH for Diabetes in Youth Study Writing Group, “Beth Loots, PhD”
cluded that, with the exception of paroxetine, further stud- should have read “Beth Loots, MPH, MSW.”

©2007 American Medical Association. All rights reserved. (Reprinted) JAMA, August 8, 2007—Vol 298, No. 6 627