Content uploaded by Maria Inês Schmidt
Author content
All content in this area was uploaded by Maria Inês Schmidt on Nov 23, 2015
Content may be subject to copyright.
Identifying Individuals at High Risk for
Diabetes
The Atherosclerosis Risk in Communities study
MARIA INˆES SCHMIDT,
MD, PHD
1,2
BRUCE B. DUNCAN,
MD, PHD
1,2
HEEJUNG BANG,
PHD
3
JAMES S. PANKOW,
PHD
4
CHRISTIE M. BALLANTYNE,
MD
5
SHERITA H. GOLDEN,
MD, MHS
6
AARON R. FOLSOM,
MD
4
LLOYD E. CHAMBLESS,
PHD
3
FOR THE ATHEROSCLEROSIS RISK IN
COMMUNITIES INVESTIGATORS
OBJECTIVE — To develop and evaluate clinical rules to predict risk for diabetes in middle-
aged adults.
RESEARCH DESIGN AND METHODS — The Atherosclerosis Risk in Communities is
a cohort study conducted from 1987–1989 to 1996–1998. We studied 7,915 participants
45–64 years of age, free of diabetes at baseline, and ascertained 1,292 incident cases of diabetes
by clinical diagnosis or oral glucose tolerance testing.
RESULTS — We derived risk functions to predict diabetes using logistic regression in a
random half of the sample. Rules based on these risk functions were evaluated in the other half.
A risk function based on waist, height, hypertension, blood pressure, family history of diabetes,
ethnicity, and age was performed similarly to one based on fasting glucose (area under the
receiver-operating characteristic curve [AUC] 0.71 and 0.74, respectively; P⫽0.2). Risk func-
tions composed of the clinical variables plus fasting glucose (AUC 0.78) and additionally in-
cluding triglycerides and HDL cholesterol (AUC 0.80) performed better (P⬍0.001). Evaluation
of scores based on the metabolic syndrome as defined by the National Cholesterol Education
Program or with slight variations showed AUCs of 0.75 and 0.78, respectively. Rules based on all
these approaches, while identifying 20–56% of the sample as screen positive, achieved sensi-
tivities of 40–87% and specificities of 50–86%.
CONCLUSIONS — Rules derived from clinical information, alone or combined with simple
laboratory measures, can characterize degrees of diabetes risk in middle-aged adults, permitting
preventive actions of appropriate intensity. Rules based on the metabolic syndrome are reason-
able alternatives to rules derived from risk functions.
Diabetes Care 28:2013–2018, 2005
P
revention of diabetes and its associ-
ated burden has become a major
health priority worldwide (1). Re-
cent clinical trials demonstrate that life-
style (2–4) and pharmaceutical (2,5,6)
interventions in individuals with im-
paired glucose tolerance (IGT) can pre-
vent the development of diabetes,
providing a rationale for the identification
of high-risk subjects so as to institute
early lifestyle interventions.
Because these trials focused primarily
on individuals with IGT, an oral glucose
tolerance test (OGTT) was required to
identify individuals meriting interven-
tion. The inconveniences and costs asso-
ciated with this test (7) have stimulated
the development of simple rules involving
readily available clinical information ca-
pable of predicting diabetes with equal or
better diagnostic properties than IGT.
Currently reported investigations are lim-
ited to Mexican Americans and non-
Hispanic whites (8), Japanese Americans
(9), and Finns (10).
The purpose of this study is to de-
velop and evaluate rules to predict high
risk of developing diabetes in middle-aged,
white, and African-American adults using
readily available clinical information.
RESEARCH DESIGN AND
METHODS — In 1987–1989, the
Atherosclerosis Risk in Communities
(ARIC) study recruited a population-
based cohort of 15,792 men and women
45–64 years of age from four U.S. com-
munities (11). Human subjects research
review committees approved the study,
and all participants gave written consent.
Follow-up visits, ⬃3 years apart, were
conducted in 1990–1992, 1993–1995,
and 1996–1998.
We excluded 2,018 individuals who
at baseline had diabetes on the basis of
clinical diagnosis, diabetes medication
use, or fasting glucose ⱖ7.0 mmol/l (an
OGTT was not performed at baseline), 95
from underrepresented minority groups,
210 not fasting for at least 8 h, 1,696 with
missing information on risk factors at
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
From the
1
Graduate Studies Program in Epidemiology, School of Medicine, Federal University of Rio Grande
do Sul, Porto Alegre, Brazil; the
2
Department of Epidemiology, School of Public Health at Chapel Hill,
University of North Carolina, Chapel Hill, North Carolina; the
3
Department of Biostatistics, School of Public
Health, University of North Carolina, Chapel Hill, North Carolina; the
4
Division of Epidemiology, School of
Public Health, University of Minnesota, Minneapolis, Minnesota; the
5
Department of Medicine, Baylor
College of Medicine, Houston, Texas; and the
6
Department of Medicine, Johns Hopkins University School of
Medicine, Baltimore, Maryland.
Address correspondence and reprint requests to Maria Ineˆs Schmidt, School of Medicine, UFRGS R.
Ramiro Barcelos, 2600/414 Porto Alegre, RS 90035-003, Brazil. E-mail: mischmidt@orion.ufrgs.br.
Received for publication 28 January 2005 and accepted in revised form 8 May 2005.
Abbreviations: ARIC, Atherosclerosis Risk in Communities; AUC, area under the receiver-operating
characteristic curve; IGT, impaired glucose tolerance; NCEP, National Cholesterol Education Program;
OGTT, oral glucose tolerance test; ROC, receiver-operating characteristic.
A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion
factors for many substances.
© 2005 by the American Diabetes Association.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby
marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Metabolic Syndrome/Insulin Resistance Syndrome/Pre-Diabetes
ORIGINAL ARTICLE
DIABETES CARE,VOLUME 28, NUMBER 8, AUGUST 2005 2013
baseline, 3,834 who had no follow-up or
incomplete information at the end of the
study to ascertain diabetes, and 24 who
had temporally inconsistent reporting of a
diagnosis of diabetes across visits, thus
leaving 7,915 participants for the analyses.
We assessed diabetes and hyperten-
sion medication use, smoking, and paren-
tal history of diabetes (in either parent) by
interview and obtained physical measures
with participants fasting and with an
empty bladder. BMI was calculated as
weight/height
2
(kg/m
2
), and obesity was
defined as a BMI ⱖ30 kg/m
2
. Waist girth
was measured at the umbilical level.
Blood pressure was determined as the
mean of two standardized measurements.
All analytes were determined at cen-
tral laboratories according to standard
protocols: plasma glucose by a hexoki-
nase assay, insulin by radioimmunoassay
(
125
Insulin Kit; Cambridge Medical Diag-
nosis, Billerica, MA), and triglycerides
and HDL cholesterol by enzymatic meth-
ods (12).
We defined incident diabetes by an
OGTT (fasting glucose ⱖ7.0 mmol/l or a
2-h glucose ⱖ11.1 mmol/l) at the end of
the follow-up (1996–1998) or as a report
of clinical diagnosis or treatment for dia-
betes during the follow-up period
(13,14).
In consonance with the National
Cholesterol Education Program (NCEP)
Adult Treatment Panel III definition of the
metabolic syndrome (15), we defined
central obesity as a waist circumference
⬎88 cm (35 in) for women and ⬎102 cm
(40 in) for men; high triglycerides as
ⱖ150 mg/dl (1.70 mmol/l); low HDL
cholesterol as ⬍40 mg/dl (1.03 mmol/l)
for men and ⬍50 mg/dl (1.29 mmol/l) for
women; impaired fasting glucose as a fast-
ing value from 6.1 to 6.9 mmol/l as well
as, following recent American Diabetes
Association recommendations (16), from
5.6 to 6.9 mmol/l; and raised blood pres-
sure as ⱖ130/85 mmHg or use of medi-
cation for hypertension.
We produced risk functions for de-
tecting incident diabetes on a randomly
selected half of the sample (training sam-
ple) using logistic regression models. Risk
factors considered were sex, ethnicity, pa-
rental history of diabetes, use of medica-
tion for hypertension, height, age, various
measures of obesity (waist, weight, BMI,
waist-to-hip ratio, each investigated one
at a time), systolic blood pressure, fasting
glucose, HDL cholesterol, triglycerides,
and fasting insulin. Continuous variables
were examined with their squared terms.
Models were built by including, first, eas-
ily obtained clinical variables, leaving
those requiring laboratory determination
for a second phase. Starting with variables
that predicted incident diabetes in univar-
iate models, we constructed multivariable
models in a forward manner, eventually
including all variables whose addition
produced an increment of at least 0.005
in the area under the receiver-operating
characteristic (ROC) curve (AUC) (17).
Once best models were defined, we
evaluated their diagnostic properties on
the other random half of the sample (test-
ing sample). To do so, we first estimated
each subject’s probability of developing
diabetes based on the derived risk func-
tions. We next established rules to char-
acterize differing degrees of risk based on
cut points of these probabilities. These cut
points were defined by fixing proportions
(20, 30, 40, and 50%) of the testing sam-
ple that would be deemed screen positive.
We then evaluated the risk functions and
their derived rules in terms of AUC, frac-
tion of total incident cases identified (sen-
sitivity), specificity, and positive and
negative predictive values.
We also examined similar diagnostic
proprieties of rules based on the NCEP
Adult Treatment Panel III metabolic syn-
drome definition and variations of it in
the testing sample. We estimated 95% CIs
for the AUCs, sensitivity, specificity, and
predictive values using 500 bootstrap
samples (18). All analyses were per-
formed with SAS software (SAS Institute,
Cary, NC).
RESULTS — At baseline, 56% of the
7,915 individuals studied were women,
85% were white, 27% were hypertensive,
and 20% were current smokers. Median
and interquartile range for various char-
acteristics were as follows: age 54 years
(49–59); BMI 26.6 kg/m
2
(23.7–30.1);
waist, women, 93 cm (84–103); waist,
men, 97 cm (91–104); height 168 cm
(161–175); systolic blood pressure 120
mmHg (108–133); fasting glucose 5.44
mmol/l (5.11–5.83). Comparing these
characteristics with those of the 5,764
subjects excluded, the only characteris-
tics suggesting possible important sys-
tematic differences were a greater
percentage of African Americans, smok-
ers, and hypertensive subjects among
those excluded. Only 2% of the final
sample was taking cholesterol-lowering
medication.
We ascertained 1,292 cases of inci-
dent diabetes: 189 (cumulative incidence
of 23.6%) among African-American
women, 93 (22.5%) among African-
American men, 532 (14.6%) among
white women, and 478 (15.7%) among
white men. Of these cases, 387 (30%)
were ascertained by self-report of clinical
diagnosis or medication use for diabetes.
Independent of this ascertainment, the
OGTT identified 1,156 (89%) case sub-
jects, 317 (24%) by fasting glucose alone,
511 (40%) by 2-h glucose alone, and 328
(25%) by both criteria. Of total case sub-
jects, all except 15 who were ascertained
by self-report of diabetes at ARIC interim
visits were present at the last follow-up
visit.
We initially defined two models: one
including only clinically detectable ele-
ments not requiring laboratory evaluation
and the other including only fasting glu-
cose. Next, we defined two further mod-
els. The first combined elements of the
two initial models and the second addi-
tionally contained HDL cholesterol and
triglycerides. Neither BMI nor fasting in-
sulin was included in these models be-
cause the additional contribution to the
AUC, although statistically significant,
was minimal for each. Models generating
risk functions separately for African
Americans and whites had generally sim-
ilar coefficients and are not reported.
The diagnostic properties of the risk
functions were next evaluated in the test-
ing sample. Figure 1Ashows the percent
of incident diabetes case subjects in each
decile of estimated risk. Risk functions in-
cluding laboratory measurements pro-
vided important risk stratification: 52% of
case subjects were distributed in the two
highest deciles of risk, and 15% were dis-
tributed in the five lowest deciles. Figure
1Billustrates the fraction of individuals in
each decile of estimated risk who devel-
oped diabetes (positive predictive value)
for each model. Participants classified in
the 9th and 10th deciles of estimated risk
by the model including lipids had, in fact,
a 33 and 52% risk of developing diabetes,
respectively; among individuals in inter-
mediate-risk categories (deciles 6– 8),
risk ranged from 13 to 24%. The other
risk functions performed slightly worse,
more so for that composed only of clinical
variables.
Table 1 presents diagnostic proper-
Prediction of incident diabetes
2014 DIABETES CARE,VOLUME 28, NUMBER 8, AUGUST 2005
ties for each risk function, displaying
rules based on cut points, chosen to per-
mit the percentage of the population
identified as at risk to vary from 20 to
50%.
Although differences are small, rules
based on risk functions including labora-
tory measurements performed generally
better, as reflected in the estimated AUCs
for these models (Table 1). Predictive
ability of the clinical variable only and
fasting glucose only models was not sig-
nificantly different (AUC 0.71 vs. 0.74,
P⫽0.2). Compared with the clinical vari-
able only model, the model combining
clinical elements with fasting glucose was
more predictive (AUC 0.78, P⬍0.001)
and that including lipids was the best pre-
dictor (AUC 0.80, P⬍0.001).
Table 2 shows the properties of rules
based on the presence of elements of the
metabolic syndrome. Rules attributing
equal weights for each element of the met-
abolic syndrome produced somewhat less
desirable diagnostic properties than rules
based on the risk function including lip-
ids (Table 1). For instance, the presence
of the metabolic syndrome (three or more
abnormalities) labeled 23% as positive
and identified 50% of future cases of dia-
betes (sensitivity), whereas a rule derived
from the risk function including lipids,
labeling as high risk a slightly lower sam-
ple fraction (20%), correctly identified
slightly more (52%) future cases. The rule
with NCEP cut points also showed less
overall predictive capacity than the risk
function including lipids (AUC 0.75 vs.
0.80, P⬍0.001).
Lowering the NCEP cut point for fast-
ing glucose (ⱖ5.6 mmol/l) in the defini-
tion of the metabolic syndrome did not
improve overall predictability (AUC ⫽
0.75), but produced rules labeling the
greater fraction of the sample as at high
risk, and as such, had higher sensitivity
and lower specificity.
Also seen in Table 2 are the diagnostic
characteristics of rules based on an alter-
native metabolic syndrome approach, de-
rived from rounding of the coefficients
of an all-categorical variable model. We
assigned 1 point for the presence of each
element of the metabolic syndrome (ex-
cept impaired fasting glucose) present, 1
additional point for obesity (BMI ⱖ30 kg/
m
2
), and 2 points for a fasting glucose
ⱖ5.6 mmol/l (or 5 points when ⱖ6.1
mmol/l). Rules based on this approach
performed slightly better. For example, a
score ⱖ5, labeling 22% of participants as
at high risk, identified 54% of future cases
of diabetes; and a score ⱖ3, labeling 46%
of participants as high risk, identified
81% of future cases of diabetes. The AUC
for this approach (0.78) was greater than
that obtained for the original NCEP rule
(0.75, P⬍0.001).
When the best risk function, that in-
cluding lipids, was evaluated in sex and
ethnicity strata, AUCs were 0.79 (95% CI
0.76– 0.82) for men, 0.81 (0.78 – 0.83)
for women, 0.80 (0.78– 0.82) for whites,
and 0.76 (0.71–0.80) for African Ameri-
cans. Additional analyses showed that de-
veloping a risk function containing lipids
on a training sample using only African-
American participants did not improve its
performance in this ethnic group in the
testing sample (data not shown).
Diagnostic properties were better
when the analysis was based only on cases
ascertained by clinical diagnosis or treat-
ment. For example, the AUC for the risk
function including lipids increased from
0.80 to 0.87; that based only on clinical
variables increased from 0.71 to 0.78. Fi-
nally, we tested in whites a clinical score
developed in the San Antonio Heart Study
(8) composed of age, sex, family history of
diabetes, BMI, HDL cholesterol, and hy-
pertension and found an AUC of 0.80.
CONCLUSIONS — Clinical trials
demonstrate that high-risk individuals,
defined as having IGT, can reduce their
risk of diabetes by more than half when
offered a well-structured intensive life-
style modification program (2,3). Diag-
nosing IGT requires an OGTT, a test of
Figure 1—A: Percentage of incident cases of diabetes detected during the 9 years of follow-up in
each decile of estimated risk. Risks were estimated from models (see equations in the footnote of
Table 1) based only on readily available clinical information (solid bars), only on fasting glucose
(dark hatched bars), on clinical information plus fasting glucose (light hatched bars), and on these
elements plus HDL cholesterol and triglycerides (open bars). B: Percentage who actually devel-
oped diabetes during the 9 years of follow-up (positive predictive value), according to decile of
estimated risk. Risks were estimated from models (see equations in the footnote of Table 1) based
only on readily available clinical information (solid bars), only on fasting glucose (dark hatched
bars), on clinical information plus fasting glucose (light hatched bars), and on these elements plus
HDL cholesterol and triglycerides (open bars).
Schmidt and Associates
DIABETES CARE,VOLUME 28, NUMBER 8, AUGUST 2005 2015
high specificity (92%) but low sensitivity
(52%) in the prediction of diabetes (8).
Our results indicate that a rule defin-
ing high risk (9-year probability of devel-
oping diabetes ⱖ26%) based on the risk
function composed of multiple variables
including lipids had similar diagnostic
properties (sensitivity 52% and specificity
86%, respectively), labeling 20% of the
sample as high risk. These properties are
generally consistent with those previously
reported for similarly constructed rules
(8,9). Slight differences between results of
the three studies are probably accounted
for by differences in population charac-
teristics such as age and ethnicity, dura-
tion of follow-up, and diabetes definition.
Of note is that rules derived from
other risk functions (Table 1) and from
various clinical scores based on the met-
abolic syndrome definitions (Table 2)
had similar, though somewhat poorer,
diagnostic properties at cut points
labeling a similar fraction (⬃20%) as
positive.
Although the best diagnostic proper-
ties found here were those derived from a
risk function including lipid variables
(AUC 0.80), the gain is small compared
with those derived from a risk function
without lipids (AUC 0.78) and, depend-
ing on the setting, may not justify the in-
creased resources needed for the lipid
measures. In settings in which HDL cho-
lesterol and triglyceride measurements
are readily available, rules based on the
metabolic syndrome definition (Table 2)
are valid, although slightly less predictive,
alternatives to rules based on risk func-
tions including these variables. Similar
findings have been described recently
(19). Yet, for those who prefer to classify
Table 1—Diagnostic characteristics in the testing sample of rules predicting risk of incident diabetes in the ARIC study
%⫹Sensitivity Specificity
Positive
predictive
value
Negative
predictive
value
Models and rules
Clinical information only AUC ⫽0.71 (0.69–0.73)
Pr(DM)
ⱖ0.23 20 40 (37–44) 84 (83–85) 32 (29–36) 88 (87–89)
ⱖ0.19 30 54 (51–58) 75 (74–75) 29 (27–32) 90 (88–91)
ⱖ0.16 40 67 (64–71) 65 (64–66) 27 (25–30) 91 (90–92)
ⱖ0.14 50 77 (74–80) 55 (54–56) 25 (23–27) 93 (91–94)
Fasting glucose only (mmol/l) AUC ⫽0.74 (0.71–0.76)
Pr(DM)
ⱖ0.24 (ⱖ5.88) 20 50 (46–53) 84 (83–86) 38 (35–42) 90 (88–91)
ⱖ0.19 (ⱖ5.72) 30 60 (57–64) 76 (74–76) 32 (29–35) 91 (90–92)
ⱖ0.15 (ⱖ5.55) 40 70 (66–73) 63 (62–66) 27 (25–29) 92 (90–93)
ⱖ0.13 (ⱖ5.44) 50 77 (74–80) 54 (52–55) 24 (22–26) 92 (91–94)
Clinical ⫹glucose AUC ⫽0.78 (0.76–0.80)
Pr(DM)
ⱖ0.26 20 51 (47–54) 86 (85–87) 41 (37–45) 90 (89–91)
ⱖ0.18 30 65 (61–68) 77 (76–77) 35 (31–37) 92 (91–93)
ⱖ0.14 40 75 (72–78) 67 (66–67) 30 (28–32) 93 (92–94)
ⱖ0.11 50 83 (80–85) 56 (56–57) 27 (24–28) 94 (93–95)
Clinical ⫹glucose ⫹lipids AUC ⫽0.80 (0.78–0.82)
Pr(DM)
ⱖ0.26 20 52 (49–56) 86 (85–87) 42 (39–46) 90 (89–91)
ⱖ0.18 30 67 (64–70) 77 (76–78) 36 (33–39) 92 (91–93)
ⱖ0.14 40 77 (73–80) 67 (66–68) 31 (29–33) 94 (93–95)
ⱖ0.10 50 85 (81–88) 57 (56–57) 27 (25–29) 95 (94–96)
Data in parentheses are 95% CIs. %⫹, percentage of sample identified as screen positive by the detection rule; Pr(DM), probability of developing diabetes, derived
from the prediction model, used as the high-risk cut point. The following are parameter estimates for the models estimating the probability of developing diabetes
over the 9-year follow-up period:
Pr(DM) ⫽1/(1 ⫹e
⫺x
), where x⫽
Clinical variables only model:⫺7.3359 ⫹0.0271 ⫻age (years) ⫹0.2295 ⫻black ⫹0.5463 ⫻parental history of diabetes ⫹0.0161 ⫻systolic blood pressure
(mmHg) ⫹0.0412 ⫻waist (cm) ⫺0.0115 ⫻height (cm).
Fasting glucose only:⫺11.7303 ⫹1.7996 ⫻fasting glucose (mmol/l).
Note: When using traditional units, the coefficient for fasting glucose (mg/dl) is 0.0999.
Clinical variables plus fasting glucose:⫺12.2555 ⫹0.0168 ⫻age (years) ⫹0.2631 ⫻black ⫹0.5088 ⫻parental history of diabetes ⫹1.6445 ⫻fasting glucose
(mmol/l) ⫹0.0120 ⫻systolic blood pressure (mmHg) ⫹0.0328 ⫻waist (cm) ⫺0.0261 ⫻height (cm).
Note: When using traditional units, the coefficient for fasting glucose (mg/dl) is 0.0913.
Clinical variables plus fasting glucose and lipids:⫺9.9808 ⫹0.0173 ⫻age (years) ⫹0.4433 ⫻black ⫹0.4981 ⫻parental history of diabetes ⫹1.5849 ⫻fasting
glucose (mmol/l) ⫹0.0111 ⫻systolic blood pressure (mmHg) ⫹0.0273 ⫻waist (cm) ⫺0.0326 ⫻height (cm) ⫺0.4718 ⫻HDL cholesterol (mmol/l) ⫹0.2420 ⫻
triglycerides (mmol/l).
Note: When using traditional units, the coefficient is 0.0880 for fasting glucose (mg/dl), 0.0122 for HDL cholesterol (mg/dl), and 0.00271 for triglycerides (mg/dl).
Black ⫽1 if African American, 0 if white, and parental history of diabetes ⫽1 if at least one parent has diabetes or 0 if not.
Prediction of incident diabetes
2016 DIABETES CARE,VOLUME 28, NUMBER 8, AUGUST 2005
risk based on the metabolic syndrome
definition rather than by entering num-
bers into a clinical calculator or webpage,
the losses are small: lower sensitivity (2%)
and specificity (4%) and a slightly greater
percent of sample deemed positive (3%).
The properties we found for the
NCEP definition (sensitivity of 50% and
specificity of 82%) are similar to those
found for IGT (52 and 92%, respectively)
and for the same NCEP rule (53 and 85%,
respectively) in a cohort of white and
Mexican-American men and women (20).
The sensitivity of the NCEP definition
found in Finnish men was lower (41%)
and specificity was higher (90%), perhaps
because the NCEP waist cut point was too
high in that setting (21). Lowering the cut
point of impaired fasting glucose to 5.6
mmol/l, as recently recommended by the
American Diabetes Association (16), pro-
duced an equally predictive, in terms of
AUC, but more sensitive NCEP-based
rule. Slight manipulations of the NCEP
definition improved its predictive power
and might serve as alternatives to those
clinicians who prefer not to use risk
functions.
A risk function built strictly on clini-
cal variables, although having less overall
predictability (AUC 0.71), was able to de-
rive rules with sensitivities ranging from
40 to 77% and corresponding specificities
ranging from 84 to 55%. A reported score
composed of a slightly different set of
clinical variables also had good diagnostic
properties (10), although, in that study,
case definition was more stringent. Thus,
rules based only on clinical information
may be of value, for example, as a first step
in serial diagnostic strategies for primary
prevention in community settings.
Whether rules with ⬃50% sensitiv-
ity, such as IGT and those mentioned
above, detect an adequate number of fu-
ture cases of diabetes for prevention is de-
batable. The main point against rules with
greater sensitivity is the consequent in-
crease in resources necessary for interven-
tions. However, to optimize resource use,
one could categorize more than just high-
and low-risk groups and implement
graded intensities of interventions, ac-
cording to the degree of risk. Our data
suggest that cut points for such categori-
zation of risk in middle-aged U.S. popu-
lations might be between deciles 5 and 6,
and 8 and 9. As illustrated in Fig. 1, par-
ticipants with estimated risk in the 9th
and 10th deciles had, in fact, a risk of
developing diabetes over 9 years of ⬃30
and 50%, respectively. In contrast, the
risk of developing diabetes among indi-
viduals in the first five deciles ranged from
⬃1% to ⬃9%.
The large community-based sample
of white and African-American men and
women followed over the current epi-
demic phase of diabetes in the U.S. and
the use of split samples to generate and
validate rules presented strengthen the
validity and generalizability of our find-
ings. The nearly equivalent predictability
of similar equations reported in other eth-
nic groups (8,9) suggests that these rules
may also be applicable to other U.S. eth-
nic groups. In fact, the equation de-
veloped in non-Hispanic whites and
Mexican Americans of the San Antonio
Heart Study (8), when applied to our
ARIC sample, produced an AUC (0.80)
equal to that found when including a sim-
ilar set of variables.
Yet, some limitations need to be con-
sidered. Losses to follow-up were not
small. Although those lost presented a
risk profile generally similar to those
studied, their exclusion could possibly
bias the diagnostic properties described.
Because an OGTT was not done at base-
line, some cases detected, especially early
on, could be prevalent ones. Yet this, in
fact, may increase the clinical relevance of
our predictive equations, because unde-
tected cases of diabetes not meeting the
diagnostic criteria of fasting hyperglyce-
mia are common in clinical practice. Ad-
ditionally, our reported diagnostic
properties would have been higher if only
clinically diagnosed cases had been in-
cluded. Finally, the sensitivities and spec-
ificities presented here for middle-aged
adults, whites, and African Americans
may not be applicable to younger or older
groups or to those in other settings, re-
Table 2—Diagnostic characteristics in the testing sample of metabolic syndrome–based rules in predicting high risk of diabetes in the ARIC
study
%⫹Sensitivity Specificity
Positive
predictive
value
Negative
predictive
value
Metabolic syndrome rules
NCEP (IFG ⱖ6.1 mmol)* AUC ⫽0.75 (0.73–0.77)
ⱖ3 23 50 (47–55) 82 (81–84) 36 (33–39) 90 (89–91)
ⱖ2 47 80 (77–83) 59 (58–61) 27 (25–29) 94 (93–95)
NCEP (IFG ⱖ5.6 mmol) AUC ⫽0.75 (0.73–0.77)
ⱖ3 32 64 (59–66) 74 (74–77) 32 (30–36) 91 (90–92)
ⱖ2 56 87 (83–89) 50 (50–53) 25 (24–27) 95 (94–96)
NCEP (augmented)† AUC ⫽0.78 (0.76–0.80)
ⱖ6 15 42 (38–46) 90 (89–92) 46 (42–50) 89 (88–90)
ⱖ5 22 54 (50–59) 84 (83–85) 40 (37–43) 91 (89–92)
ⱖ4 32 68 (65–72) 75 (74–77) 35 (32–37) 93 (91–93)
ⱖ3 46 81 (78–84) 61 (60–63) 29 (27–31) 94 (93–95)
Data in parentheses are 95% CIs. %⫹, percentage of sample identified as screen positive by the detection rule. *NCEP metabolic syndrome rules: 1 point each for
high waist circumference (women ⬎88 cm or 35 in, men ⬎102 cm or 40 in), raised blood pressure (⬎130/85 mmHg or using antihypertensive medication), low
HDL cholesterol (⬍40 mg/dl for men and ⬍50 mg/dl for women), high triglycerides (⬎150 mg/dl), and hyperglycemia (fasting glucose ⱖ6.1 mmol/l or ⱖ5.6
mmol/l). †Augmented metabolic syndrome score: 1 point for each element of the metabolic syndrome present (as above), except for fasting glucose (2 points when
fasting glucose ⱖ5.6 mmol/l, or 5 points when fasting glucose ⱖ6.1 mmol/l); additionally, 1 point for obesity (BMI ⱖ30 kg/m
2
).
Schmidt and Associates
DIABETES CARE,VOLUME 28, NUMBER 8, AUGUST 2005 2017
sulting in a different underlying risk of
developing diabetes.
In conclusion, rules derived from
readily available clinical information,
alone or combined with simple laboratory
measures, can characterize groups of
middle-aged adults as having various de-
grees of diabetes risk. This categorization
permits grading the intensity of preven-
tive actions according to the degree of risk
of each patient. Though further validation
of these rules in other samples is impor-
tant, they have immediate application. In
addition to their use in clinical encoun-
ters, they can be applied by managed care
organizations to existing databases to
identify high-risk individuals. Algorithms
based on these rules can also facilitate en-
rollment in clinical trials testing new strat-
egies to prevent diabetes.
Acknowledgments— Support for this study
was provided by National Heart, Lung, and
Blood Institute Contracts N01-HC-55015,
N01-HC-55016, N01-HC-55018, N01-HC-
55019, N01-HC-55020, N01-HC-55021, and
N01-HC-55022 and National Institute of Dia-
betes and Digestive and Kidney Diseases Grant
5R01-DK56918-03. M.I.S. and B.B.D. re-
ceived support from a Centers of Excellence
Grant of CNPq (the Brazilian National Council
for Scientific and Technological Develop-
ment).
The authors thank the staff and partici-
pants in the ARIC study for their important
contributions.
References
1. Venkat NK, Gregg EW, Fagot-Campagna
A, Engelgau MM, Vinicor F: Diabetes: a
common, growing, serious, costly, and
potentially preventable public health
problem. Diabetes Res Clin Pract 50
(Suppl. 2):S77–S84, 2000
2. Knowler WC, Barrett-Connor E, Fowler
SE, Hamman RF, Lachin JM, Walker EA,
Nathan DM: Reduction in the incidence of
type 2 diabetes with lifestyle intervention
or metformin. N Engl J Med 346:393–403,
2002
3. Tuomilehto J, Lindstrom J, Eriksson JG,
Valle TT, Hamalainen H, Ilanne-Parikka
P, Keinanen-Kiukaanniemi S, Laakso M,
Louheranta A, Rastas M, Salminen V,
Uusitupa M: Prevention of type 2 diabetes
mellitus by changes in lifestyle among
subjects with impaired glucose tolerance.
N Engl J Med 344:1343–1350, 2001
4. Pan XR, Li GW, Hu YH, Wang JX, Yang
WY, An ZX, Hu ZX, Lin J, Xiao JZ, Cao
HB, Liu PA, Jiang XG, Jiang YY, Wang JP,
Zheng H, Zhang H, Bennett PH, Howard
BV: Effects of diet and exercise in prevent-
ing NIDDM in people with impaired glu-
cose tolerance: the Da Qing IGT and
Diabetes Study. Diabetes Care 20:537–544,
1997
5. Chiasson JL, Josse RG, Gomis R, Hanefeld
M, Karasik A, Laakso M: Acarbose for pre-
vention of type 2 diabetes mellitus: the
STOP-NIDDM randomised trial. Lancet
359:2072–2077, 2002
6. Buchanan TA, Xiang AH, Peters RK, Kjos
SL, Marroquin A, Goico J, Ochoa C, Tan
S, Berkowitz K, Hodis HN, Azen SP: Pres-
ervation of pancreatic beta-cell function
and prevention of type 2 diabetes by phar-
macological treatment of insulin resis-
tance in high-risk Hispanic women.
Diabetes 51:2796–2803, 2002
7. Stern MP, Williams K, Haffner SM: Do we
need the oral glucose tolerance test to
identify future cases of type 2 diabetes?
Diabetes Care 26:940–941, 2003
8. Stern MP, Williams K, Haffner SM: Iden-
tification of persons at high risk for type 2
diabetes mellitus: do we need the oral glu-
cose tolerance test? Ann Intern Med 136:
575–581, 2002
9. McNeely MJ, Boyko EJ, Leonetti DL, Kahn
SE, Fujimoto WY: Comparison of a clini-
cal model, the oral glucose tolerance test,
and fasting glucose for prediction of type
2 diabetes risk in Japanese Americans. Di-
abetes Care 26:758–763, 2003
10. Lindstrom J, Tuomilehto J: The diabetes
risk score: a practical tool to predict type 2
diabetes risk. Diabetes Care 26:725–731,
2003
11. ARIC Investigators: The Atherosclerosis
Risk in Communities (ARIC) Study. Am J
Epidemiol 129:687–702, 1989
12. Friedewald WT, Levy RI, Fredrickson DS:
Estimation of the concentration of low-
density lipoprotein cholesterol in plasma,
without use of the preparative ultracentri-
fuge. Clin Chem 18:499–502, 1972
13. Expert Committee on the Diagnosis and
Classification of Diabetes Mellitus: Report
of the Expert Committee on the Diagnosis
and Classification of Diabetes Mellitus.
Diabetes Care 20:1183–1197, 1997
14. Alberti KG, Zimmet PZ: Definition, diagno-
sis and classification of diabetes mellitus
and its complications. Part 1: diagnosis and
classification of diabetes mellitus provi-
sional report of a WHO consultation. Diabet
Med 15:539–553, 1998
15. Expert Panel on Detection, Evaluation,
and Treatment of High Blood Cholesterol
in Adults: Executive Summary of the
Third Report of the National Cholesterol
Education Program (NCEP) Expert Panel
on Detection, Evaluation, and Treatment
of High Blood Cholesterol in Adults
(Adult Treatment Panel III). JAMA 285:
2486–2497, 2001
16. Genuth S, Alberti KG, Bennett P, Buse J,
Defronzo R, Kahn R, Kitzmiller J, Knowler
WC, Lebovitz H, Lernmark A, Nathan D,
Palmer J, Rizza R, Saudek C, Shaw J,
Steffes M, Stern M, Tuomilehto J, Zimmet
P: Follow-up report on the diagnosis of
diabetes mellitus. Diabetes Care 26:3160 –
3167, 2003
17. Campbell G: Advances in statistical meth-
odology for the evaluation of diagnostic
and laboratory tests. Stat Med 13:499–
508, 1994
18. Efron B, Tibshirani RJ: Introduction to the
Bootstrap. New York, Chapman and Hall,
1993
19. Stern MP, Williams K, Gonzalez-Villal-
pando C, Hunt KJ, Haffner SM: Does the
metabolic syndrome improve identifica-
tion of individuals at risk of type 2 dia-
betes and/or cardiovascular disease?
Diabetes Care 27:2676–2681, 2004
20. Lorenzo C, Okoloise M, Williams K, Stern
MP, Haffner SM: The metabolic syndrome
as predictor of type 2 diabetes: the San
Antonio heart study. Diabetes Care
26:3153–3159, 2003
21. Laaksonen DE, Lakka HM, Niskanen LK,
Kaplan GA, Salonen JT, Lakka TA: Meta-
bolic syndrome and development of dia-
betes mellitus: application and validation
of recently suggested definitions of the met-
abolic syndrome in a prospective cohort
study. Am J Epidemiol 156:1070–1077,
2002
Prediction of incident diabetes
2018 DIABETES CARE,VOLUME 28, NUMBER 8, AUGUST 2005