Predictors of Dyslipidemia Over
Time in Youth With Type 1
Diabetes: For the SEARCH for
Diabetes in Youth Study
Diabetes Care 2017;40:607–613 | DOI: 10.2337/dc16-2193
OBJECTIVE
Understanding the risk factors associated with progression and regression of
dysli pidem ia in youth with type 1 diabetes may guide treatments.
RESEARCH DESIGN AND METHODS
We studied 1,478 youth with type 1 diabe tes (age 10.8 6 3.9years,50%male,77%
non-Hispanic white, not on lipid-lowering medications) at baseline and at a
mean follow-up of 7.1 6 1.9 years in the S EARCH for Diabetes in Youth (SEARCH)
study. Progression to dyslipidemia was defined as no rmal lipid concen tration s at
baseline and abnormal at follow-up (non–HDL-cholesterol [C] >130 mg/dL or
HDL-C <35 mg/dL). Regression was defined as abnormal lipids at baseline and
normal at follow-up. Multivariable logistic regression was used to evaluate
factors associated with progression and regression compared with stable nor-
mal and stable abnormal, respectively. An area under the curve (AUC) variable
was used for the time-varying covariates A1C and waist-to-height ratio
(WHtR).
RESULTS
Non– HDL-C progressed, regressed, was stable normal, and stable abnormal in
19%, 5%, 69% , and 7% of youth with type 1 diabetes, respectively. Corresponding
percentages for HDL-C were 3%, 3%, 94%, and 1%, respe ctivel y. Factors associated
with non–HDL-C progression were higher A1C AUC and h igher WHtR A UC in males.
Non–HDL-C regression was associated with lower WHtR AUC, and HDL-C progres-
sion was associated with male sex and higher WHtR AUC. HDL-C regression was
not modeled due to small numbers.
CONCLUSIONS
A1C and WHtR are modifiable risk factors associated with change in dyslipidemia
over time in youth wit h type 1 diabetes.
Cardiovascular disease is the leading cause of death in adults with type 1 diabetes
(1,2). This process begins in youth (3,4), and dyslipidemia is a major contributing risk
factor (4).
Dyslipidemia has been well docume nted among youth with type 1 dia betes (5–9).
However, few longitudinal studies exist, and those that have been published are
limited by their retrospective nature, small sample size, inclusion of nonfasting lipid
measurements, an d relatively short duration of follow-up (10–14).
1
Department of Pediatrics, Cincinnati Children’s
Hospital and University of Cincinnati, Cincinnati,
OH
2
Department of Pediatrics, University of Colo-
rado School of Medicine, Aurora, CO
3
Department of Biostatistical Sciences, Wake
Forest School of Medicine, Winston-Salem, NC
4
Division of Diabetes Translation, Centers for
Disease Control and Prevention, Atlanta, GA
5
Depa rtment of Epidemiology and Prevention,
Wake Forest School of Medicine, Winston-Salem,
NC
6
Department of Epidemiology and Biostatistics,
University of South Carolina, Columbia, SC
7
Department of Research & Evaluation, Kaiser
Permanente Southern California, Pasadena, CA
8
Department of Pediatrics, University of
Washington, Seattle, WA
9
Northwest Lipid Metabolism and Diabetes Re-
search Laboratories, University of Washington,
Seattle, WA
10
Department of Epidemiology, Colorado School
of Public Health, University of Colorado Denver,
Aurora, CO
Corresponding author: Amy S. Shah, amy.shah@
cchmc.org.
Received 12 October 2016 and accepted 11
January 2017.
© 2017 by the American Diabetes Association.
Readers may use this article as long as the work
is properly cited, the use is educational and not
for profit, and the work is not altered. More infor-
mation is available at http://www.diabetesjournals
.org/content/license.
Amy S. Shah,
1
David M. Maahs,
2
Jeanette M. Stafford,
3
Lawrence M. Dolan,
1
Wei Lang,
3
Giuseppina Imperatore,
4
Ronny A. Bell,
5
Angela D. Liese,
6
Kristi Reynolds,
7
Catherine Pihoker,
8
Santica Marcovina,
9
Ralph B. D’Agostino Jr.,
3
and
Dana Dabelea
10
Diabetes Care Volume 40, April 2017 607
CARDIOVASCULAR AND METABOLIC RISK
Downloaded from http://diabetesjournals.org/care/article-pdf/40/4/607/548316/dc162193.pdf by guest on 26 August 2022
Thus, using 7 years of follow-up data
in a large cohort of youth with type 1
diabetes, we examined 1)howfasting
lipid levels track over time and 2) factors
that are associated with progression or
regression of dyslipidemia over time.
Identifying risk factors that are a ssoci-
ated with progression and regression
of dyslipidemia in youth with type 1 di-
abetes may guide treatments.
RESEARCH DESIGN AND METHODS
Study Participants
Participan ts for this study were enrolled
in the SEARCH for Diabetes in Youth
(SEARCH) study, a multicenter study ex-
amining the prevalence, incidence, and
complications for youth with all forms
of diabetes. Extensive details of t he
SEARCH study have been published
and are summarized in a recent publica-
tion by Hamman et al. (15). Youth in-
cluded in this analysis were diagnosed
with incident type 1 diabetes starting
in 2002 (baseline study visit occurred
9.0 6 6.1 months after diabetes diagno-
sis) and subsequently participated in a
SEARCH study follow-up visit (all visits
completed by 2015). At baseline all par-
ticipants had type 1 diab etes, defi ned as
diabetes autoantibody po sitivity (G AD,
islet antigen 2 [IA2], or zinc transporter
8 [ZnT8]) or no diabetes autoantibodies
with high insulin sensitivity, as previously
described by Dabelea et al. (16).
There were 2,004 SEARCH participants
who had a baseline and follow-up visit. We
excluded participants if they did not have a
fasting lipid profile at the baseline (n = 179)
or follow-up (n = 205) visit, if they did not
report being on insulin at the follow-up visit
(n = 29), if they reported taking lipid-
lowering drugs at either visit (n = 63) to
evaluate change in lipids without the in-
fluence of medications, or if they were
younger than 10 years old at the fol-
low-up visit (n = 50). Therefore, this re-
port includes 1 ,478 youth with type 1
diabetes. Of those, 1,356 had diabetes
autoantibody positivity and 1 22 had a
high insulin sensitivity score alone (16).
The study was reviewed and approved by
each of the local institutional review boards,
and all participants and parents provided
written informed assent and/or consent.
Anthropometric and Metabolic
Measurements
Race/ethnicity was self-reported, and par-
ticipants were categorized as non-Hispanic
white (NHW), non-Hispanic black, His-
panic, or other racial/ethnic group
(Asian, Pacific Islander, American Indian,
or other). Participants completed stan-
dardized questionnaires for medical his-
tory and medications. BMI was calculated
as weight (kg)/height (m
2
), and age- and
sex-specific BMI z scores were deriv ed
(17). Waist circumference was measured
using the National Health and Nutrition
Examination Survey (NHANES) protocol
(18) and divided by height in centimeters
to calculate the waist-to-height ratio
(WHtR). Measurements of hemoglobin
A
1c
(A1C), total cholesterol (TC), triglyc-
erides (TGs), and HDL-cholesterol (C) were
performed as previously described (19).
LDL-C was calculated by the Friedewald
equation or measured by the beta
quantifi cation procedure if TGs were
$400 mg/dL.
Definitions of Abnormal Lipids
The major outcomes for this analysis
were changes in dyslipidemia status for
non–HDL-C (computed as TC 2 HDL-C)
and HDL-C over time. Non–HD L-C was
selected because it accounts for the
cholesterol carried by al l particles con-
taining apolipoprotein B and outper-
forms the individual lipid parameters
(TC, TGs, and LDL-C) in predicting sub-
clinical atherosclerosis a nd cardio-
vascular disease (20–22). Abnormal
non–HDL-C was defined as .130 mg/dL,
and abnormal HDL-C was defined
as ,35 mg/dL, thresholds based on
current recommendations in adults and
children with diabetes (23,24). For each
ofthesetwomeasures,wedefined pro-
gression of dyslipidemia as normal lipid
concentrat ions at baseline (non–HDL-C
#130 mg/dL or HDL-C $35 mg/dL) and
abnormal at final follow-up, and regression
was defined as abnormal at baseline and
normal at final follow-up. Stable normal was
defined as normal at baseline and follow-up
and stable abnormal as abnormal at both
baseline and follow-up.
Statistical Analysis
Data are presented as mean 6 SD or
median (interquartile range) for continu-
ous variables, or frequencies (and per-
centages) for categorical variables.
Demographics, anthropometrics, and car-
diovascular risk factors were compared
across the four groups (stable n ormal,
stable abnormal, progression, and regres-
sion) by one-way ANOVA for continuous
variables and x
2
tests for categorical
variables.
We u sed separate multivariable logis-
tic regression models to examine fa ctors
associated with non–HDL-C and HDL-C
progression compared with stable normal
and those associated with non–HDL-C re-
gression compared with stable abnormal.
HDL-C regression was not modeled be-
cause of small numbers in the regression
and stable abnormal groups. Model cova-
riates included age at basel ine visit (in
years), race/ethnicity (NHW vs. other),
sex (female vs. mal e), and duration of
type 1 diabetes at baseline (in years).
A derived area under the curve (AUC)
summary statistic (a continuous vari-
able)forWHtRandA1Cwasalsoin-
cluded in the models. AUC summarizes
the longitudinal measures collected
over time adjust ing for the interval be-
tween each measure. WHtR was chosen
over other measures of adipo sity (BMI
z score or waist circumference) because
the former has been shown to be more
strongly associated with adverse cardio-
vascular risk factors in children and
adults (25,26). We also evaluated inter-
action terms (race/ethnicity or sex by
WHtR) to determine whether the asso-
ciations between WHtR and lipid pro-
gression and regression were different
by race/ethn ic group or sex. All models
were also adjusted f or clin ic site, time
interval between the visits, and seas on
of the baseline visit. Variables with
P values of ,0.05 were considered sta-
tistically significant. Statistical analyses
were performed using SAS 9.4 s oftware
(SAS Institute, Inc., C ary, NC).
RESULTS
Characteristics of SEARCH participants
with type 1 diabetes included in this
analysis at baseline and follow-up are
presented in Table 1. At baseline, the
cohort was a mean age of 10.8 6 3.9
years, the average disease duration
was 0.75 6 0.5 years, and mean A1C
was 7.6 6 1.5%. NHW comprised 77%
of the cohort, and 50% were male.
Follow-up data were ob tained an av-
erage of 7. 1 6 1.9 years later, when par-
ticipants were an average age of 17.9 6
4.1 years and had an average disease
duration of 7.8 6 1.9 years. The mean
A1C at follow-up was 9.2 6 1.8%. Non–
HDL-C progressed in 19%, regressed in
5%, and r emained stable abnormal in 7%
608 Lipids Over Time in Youth With Diabetes Diabetes Care Volume 40, April 2017
Downloaded from http://diabetesjournals.org/care/article-pdf/40/4/607/548316/dc162193.pdf by guest on 26 August 2022
and stable normal in 69%. HDL-C pro-
gress ed in 3%, regressed in 3%, and re-
mained stable abnormal in 1% and
stable normal in 94%.
Par ticipants who had pro gression of
non–HDL-C levels compared with those
who remained stable normal (Table 2)
were older, more likely to be female,
had greater adiposity (measured by
BMI z score or WHtR), a longer duration
of type 1 diabetes, a higher A1C, and
higher diastolic blood pressure (all P ,
0.05). Participants who r emained stable
abnormal were more likely to be non–
Hispanic black, Hispanic, or other race/
ethnicity, female, have more adiposity ,
and have higher A 1C than youth who
were stable normal (all P , 0.05).
We constructed multivariable logistic
regression models to examine factors
associated with progression and regres-
sion of dyslipidemia compared with
stable normal and stable abnormal , re-
spectively, after adjusting for cov ariates
(Table 3). Factors associated with non–
HDL-C progression were h igher A1C AUC
and higher WHtR AUC. Non–HDL-C re-
gression was associated with lower
WHtR AUC. HDL-C progression was as-
sociated with male sex and higher WHtR
AUC. HDL-C regression was not mod-
eled because of small numbers in the
regression (3%) and stable abnormal
(1%) groups.
We evaluated the interactions be-
tween race or sex and WHtR for each
of the outcomes. We found a significant
sex-b y-WHtR interaction (P = 0.0071) for
non–HDL-C progression such th at the
association between WHtR and n on–
HDL-C progression was stronger for
males (2.63; 95% CI 1.83, 3.77) than
for females (1.3 8; 95% CI 1.0 2, 1.87).
CONCLUSIONS
We report the natural evolution of dys-
lipidemia over 7 y ears in a large cohort
of youth with t ype 1 diabetes. After
adjusting for covariates, we identified
two modifiable risk factors, WHtR and
A1C burden over time, that were indepen-
dent predictors of unfavorable changes
in lipids or of stable abnormal levels
over time.
The prevalence of dyslipidemia in
youth with type 1 diabetes has been
well documented in two large multicen-
ter cross-sectional studies, the SEARCH
for Diabetes in Youth study and the Ger-
man prospective documentation and
quality management system (DPV) study
(5–7), as well several smaller cross-sec-
tional studies (8,9,27). Although a few
longitudinal studies exist, these studies
are retrospective, have small sample
sizes, include nonfasting lipid measure-
ments, and are of relatively short follow-up
duration (10–14). In 2007, Maahs et al.
(11) retrospectively examined lipids over
time in 360 youth with type 1 diabetes
(age range 2–21 years) with a mean follow-
up of 2.9 years. Using the thresholds for
non–HD L-C a nd HDL-C of $130 m g/dL
and ,35 mg/dL as abnormal, they re-
ported 27.8 and 3.3%, respectively, of
youthwithtype1diabeteshadsus-
tained dyslipidemia over time. In addi-
tion, they found that high er A1C was
positively associated with non– HD L-C
levels and that a higher BMI z score
was inversely related to HDL-C levels
(11). Using similar criteria, Edge et al.
(10) reported the frequency of dyslipide-
mia in 229 youth with type 1 diabetes in
the U.K. as 4.3% for non–HDL-C and 0%
for HDL-C. Furthermore, they showed
that a higher non–HDL-C concentration
was a ssociated with higher A1C and
longer duration of type 1 diabetes but
lacked measures of adiposity to evaluate
associations with lipids over time. Marco-
vecchio et al. (12) did find that sustained
non–HDL-C abnormalities were related to
older age, longer duration of type 1 dia-
betes, and higher BMI and A1C levels.
However, with loss of greater than 75%
of their cohort at the end of 2.3 years, this
precluded definitive conclusions. In con-
trast, Reh et al. (13) reported longitudinal
lipid levels in a cohort of 46 adolescents
and young adults with type 1 diabetes in
the U.S. (age range 12–25 years) during
3 years of follow-up and found that 0%
of the cohort had sustained abnormal
HDL-C ( defined as ,40 mg/dL) over
time; non–HDL-C was not reported.
Here, we report prospective lipid data
in youth with type 1 diabetes over a mean
follow-up of ;7years,thelongestfollow-
up published in this population to date.
We show that 19% of the cohort pro-
gressed to abnormal non–HDL-C concen-
trations during this time. Also concerning
is that 7% of you th had sustained abnor-
mal non–HDL-C over time, but only 5%
had regressed. This stable abnormal fre-
quency is somewhat lower than previ-
ously reported by Maahs et al. (11),
where 27.8% of their adolescent cohort
with type 1 diabetes had sustained eleva-
tion in non–HDL-C. The lower frequency
of dyslipidemia reported here may be ex-
plained by our exclusion of t hose o n
lipid-lowering medication. Differences
Table 1—Study cohort at baseline and follow-up
Baseline Follow -up
n Mean 6 SD or n (%) n Mean 6 SD or n (%)
Age (years) 1,478 10.8 6 3.9 1,478 17.9 6 4.1
Race/ethnicity 1,477
Non-Hi spanic
White 1,141 (77.3) –
Black 140 (9.5) –
Hispanic 170 (1 1.5) –
Other 26 (1.8)
Male sex 1,478 743 (5 0.3) –
BMI z score 1,457 0.48 6 1.04 1,473 0.59 6 0.96
WHtR 1,358 0.48 6 0.06 1,472 0.51 6 0.08
Type 1 diabetes duration (years) 1,478 0.7 6 0.5 1,478 7.8 6 1.9
A1C (%) 1,472 7.6 6 1.5 1,474 9.2 6 1.8
A1C (mmol/mol) 1,472 59.8 6 16.1 1,474 76.6 6 19.9
TC (mg/dL) 1,478 159 6 27 1, 478 169 6 34
LDL-C (mg/dL) 1,478 91 6 22 1,478 96 6 28
HDL-C (mg/dL) 1,478 56 6 13 1,478 55 6 13
Non–HDL-C (mg/dL) 1,478 103 6 25 1, 478 114 6 35
TGs (mg/dL), median (Q1, Q3) 1,478 55 (42, 71) 1,478 75 (56, 105)
Systolic blood pressure (mmHg) 1,438 99 6 12 1,475 106 6 11
Diastolic blood pressu re (mmHg) 1,436 63 6 10 1,475 69 6 9
Mean interval between visits 7.1 6 1.9 years. Q, quartile.
care.d iabetesjournals.org Shah and Associates 609
Downloaded from http://diabetesjournals.org/care/article-pdf/40/4/607/548316/dc162193.pdf by guest on 26 August 2022
Table 2—Characteristics of participants at follow-up visit: comparisons using non–HDL-C*
Stable normal
n = 1,020
Stable abnormal
n = 105
Progression
n = 285
Regression
n =68 P valu e* P valu e* P valu e* P value*
Mean 6 SD
or n (%)
Mean 6 SD
or n (%)
Mean 6 SD
or n (%)
Mean 6 SD
or n (%)
Overall among
four groups
Progression vs.
stable normal
Stable abnormal
vs. stable normal
Regression vs.
stable normal
Age (years) 17.68 6 4.14 17.88 6 4.15 18.53 6 3.73 18.49 6 4.58 0.0111 0.0019 0.6198 0.1121
Race/ethnicity 0.0225 0.0530 0.0224 0.5495
Non-Hispanic
White 806 (79.1) 71 (67.6) 206 (72.3) 58 (85.3)
Black 88 (8.6) 18 (17.1) 30 (10.5) 4 (5.9)
Hispanic 110 (10.8) 14 (13.3) 40 (14.0) 6 (8.8)
Other 15 (1.5) 2 (1.9) 9 (3.2) 0 (0)
Male sex 535 (52.5) 42 (40.0) 130 (45.6) 36 (52.9) 0.0288 0.0412 0.0151 0.9375
BMI z score 0.52 6 0.93 1.01 6 0.99 0.77 6 0.92 0.32 6 1.11 ,0.0001 ,0.0001 ,0.0001 0.0924
WHtR 0.49 6 0.07 0.55 6 0.10 0.53 6 0.08 0.49 6 0.08 ,0.0001
,0.0001 ,0.0001 0.7894
Type 1 diabetes duration (years) 7.73 6 1.85 7.97 6 1.91 8.07 6 1.93 8.14 6 1.97 0.0184 0.0069 0.2082 0.0792
A1C (%) 8.90 6 1.66 9.80 6 1.96 9.92 6 2.02 8.78 6 1.97 ,0.0001 ,0.0001 ,0.0001 0.5943
TC (mg/ dL) 153.54 6 21.23 211.56 6 29.96 209.54 6 30.67 166.13 6 18.94 ,0.0001 ,0.0001 ,0.0001 ,0.0001
LDL-C (mg/dL) 82.57 6 17.08 136.57 6 26.32 128.26 6 22.63 95.74 6 12.09 ,0.0001 ,0.0001 ,0.0001 ,0.0001
HDL-C (mg/dL) 56.32 6 13.56 49.67 6 13.64 52.36 6 12.46 53.72 6 12.68 ,0.0001 ,0.0001 ,0.0001 0.1198
Non–HDL-C (mg/dL) 97.22 6 18.68 161.90 6 29.54 157.18 6 29.57 112.41 6 12.15 ,0.0001 ,0.0001 ,0.0001 ,0.0001
TGs† (mg /dL), median (Q1, Q3) 65.0 (51, 86) 107.0 (76, 157) 115.0 (89, 167) 74.5 (57.5, 94.5) ,0.0001
,0.0001 ,0.0001 0.0259
Blood pressure
Systolic (mmHg) 106.02 6 10.77 106.84 6 9.84 106.32 6 11.01 105.95 6 12.39 0.8801 ^ ^ ^
Diastolic (mmHg) 68.31 6 8.55 69.93 6 8.53 70.28 6 9.29 67.49 6 9.50 0.0022 0.0008 0.0707 0.4512
Q, quartile. *Comparisons among groups evaluated using one-way AN OVA (continuous) or x
2
tests (categorical); ^pairwise tests are not reported where the overall test across four groups is not statistically
signifi cant (P . 0.05); †tested using log (TGs).
610 Lipids Over Time in Youth With Diabetes Diabetes Care Volume 40, April 2017
Downloaded from http://diabetesjournals.org/care/article-pdf/40/4/607/548316/dc162193.pdf by guest on 26 August 2022
may also be explained by lower baseline
BMI and A1C in our cohort (11). Progres-
sion to abnormal HDL-C was 3% and sus-
tained abnormal HDL -C was 1% in this
study, consistent with previous reports
(10,11,13).
We used the WHtR AUC to explore
the association between burden of adi-
posity over time and dyslipidemia,
which has not been ass essed in longitu-
dinal studies of youth with type 1 diabe-
tes to date. We show that although
a higher WHtR AUC is independently
associated with non–HDL-C and HDL-C
progression, a lower WHtR AUC ratio is as-
sociated with higher odds of non–HDL-C
regression. Furthermore, we show that
the association b etween WHtR AUC
and non –HDL-C progression is stronger
for males compared with females. These
data suggest that similar to youth with-
out diabetes (28), adiposity is an im-
portant in dependent risk factor for
dyslipidemia among youth with type 1
diabetes. Future work is needed to
determine whether reductions in ab-
dominal adiposity improve lipid levels
over time in youth with type 1 diabe-
tes and whether these effects are more
pronounced in males.
We show that glycemic control over
time is another important modifiable
risk factor that is associated with higher
odds of non–HDL-C progression. These
findin gs are consistent with prior cross-
sectional and longitudinal studies in youth
with type 1 diabetes (5,7,8,11) as well as
data from adults who participated in the
Diabetes Control and Complication s Trial
(DCCT) (29). Although worse glycemic con-
trol over time appears to adversely affect
lipid levels, lowering of A1C through
intensive insulin therapy has been shown
to negatively affect weight (30), although
not in all studies (31). These results point
to a delicate b alance between achiev-
ing glycemic control and maintaining
body weight that affects lipids and re-
mains to be elucidated.
One potential mechanism linking adi-
posity, glycemic control, and dysli pide-
mia in type 1 diabetes may be insulin
resistance. Although insulin resistance
among those with type 1 diabet es ap-
pears counterintu itive, because they
arebydefinition insulin deficient, prior
work has shown that yo uth with type 1
diabetes exhibit insulin resistance
(32,33). The etiology of insulin resis-
tanceintype1diabetesisnotclear,
but adiposity, physical inactivity, and/
or chronic exogeno us insulin use may
all play a role. Therefor e, determining
the optimal level of insulin needed to
achieve glycemic control while avoiding
weight gain appear s critic al to decreas-
ing the progression of dyslipidemia in
youth with type 1 diabetes. Unfortu-
nately, we were not able to assess or
estimate insulin resistance or sensitivity
in this study. Prio r SEARCH studies hav e
used an equation that incorporates A1C,
waist circumference, and TGs (32) to es-
timate insulin sensitivity, but the cur-
rent study used insulin s ensitivity to
define th e cohort , included A1C and
WHtR AUC in the models, and TGs are
included in the outcome non–HDL-C.
We found that male sex was associated
with higher odds of HDL-C progression.
Longitudinal data in healthy children, in-
cluding work from the Bogalusa Heart
Study, have shown that HDL-C levels,
particularly for NHW males, decline at
age14yearsandcontinuetodrop
until age 26 years, unlike NHW females,
who have little decrease in HDL-C (34).
Therefore, it is unclear whether the
higher odds of HDL-C progression ob-
served in this cohort of predominantly
NHW males is a result of type 1 diabetes
or normal t racking of lipids through
adolescence.
Strengths of this study include a large
cohort of youth with type 1 diabetes, stan-
dardized lipid measurements, follow-
up data over 7 y ears, and the ability
to evaluate the associations between
burden of risk factors and l ipids over
time. Limitations of the study include a
lack of more frequent lipid assessments
during the 7 years of follow-up, relatively
small numbers of participants in each
category that limited our ability to ex-
plore H DL-C regression, and lack of
some variables, including thyroid status,
family histor y of hyper/dyslipidemia,
and pubertal status, each of which is
knowntoinfluence lipids. In addition,
physical activity, diet history, and smok-
ing status were not obtained on all par-
tici pants at the baseline visit and thus
could not be i ncluded as covariates to
evaluate change in lipids overtime, al-
though it is possible physical activity
and diet may be reflected by changes in
adiposity. Future studies should include
these variables.
In conclusion, we demonstrate ap-
proximately one-quarter of youth with
type 1 diabetes has progression of dysli-
pidemia or abnormal lipids that persists
over time. Risk factors that influence pro-
gression include both increased abdomi-
nal adiposity and worse glycemic control
over time. Until the complex interpla y
Table 3—Multivariable logistic regression models for dyslipidemia progression and regression
Non–HDL-C progression compared
with stable normal
n = 1,288 (281 events)
Non–HDL-C regression compared
with stable abnormal
n = 170 (67 events)
HDL progression compared with
stable normal
n = 1,405 (38 e vents)
Variable OR (95% CI) P value OR (95% CI) P value OR (95% CI) P value
Age at initial visit: 1 year increase 1.03 (0.99, 1. 07) 0.1815 1.07 (0.98, 1.18) 0.1205 1.02 (0.94, 1.11) 0.6354
Race/ethnicity: other vs. NHW 1.22 (0.85, 1.75) 0.2726 0.48 (0.19, 1.23) 0.1274 1.10 (0.48, 2.53) 0.8172
Sex: female vs. male 1.03 (0.77, 1.38) 0.8199 0.71 (0.32, 1.54) 0.3815 0.44 (0.22, 0.91) 0.0257
Type 1 diabetes duration at initial visit:
1 year increase 0.98 (0.74, 1.31) 0.8985 1.77 (0.93, 3.33) 0.0800 1. 48 (0.79, 2.76) 0.2183
A1C (AUC): 1% unit increase 1.39 (1.25, 1.55) <0.00 01 0.84 (0.65, 1.08) 0.1734 1.07 (0.83, 1.37) 0.5894
WHtR (AUC): 0.1 unit increase 1.81 (1.42, 2.29) <0. 0001 0.49 (0.29, 0.84) 0.0089 1.64 (1.03, 2.59) 0.0353
Variables included in the models: age and type 1 di abetes duration at initial visit, race/ethnicity, sex, A1C AUC, and WHtR AUC. Each model also
adjusted for clinical site, the time int erval between the baseline and follow-up visit, and season at the baseline visit. Sta tistically si gnificant covariates
appear in boldface type. OR, odds ratio.
care.d iabetesjournals.org Shah and Associates 611
Downloaded from http://diabetesjournals.org/care/article-pdf/40/4/607/548316/dc162193.pdf by guest on 26 August 2022