Académique Documents
Professionnel Documents
Culture Documents
www.jahonline.org
Original article
Article history: Received July 12, 2013; Accepted December 18, 2013
Keywords: Adolescent; Polycystic ovarian syndrome; Screening; Diabetes
A B S T R A C T
IMPLICATIONS AND
CONTRIBUTION
Purpose: Screening for impaired glucose tolerance (IGT) is recommended for adolescents with
polycystic ovary syndrome (PCOS) with oral glucose tolerance test (OGTT). Whether glycated
Our study demonstrates
hemoglobin (HbA1c) can be used for screening in this patient population is unknown. We sought that both glycated hemo-
to determine the utility of HbA1c and 2-hour OGTT for diagnosing dysglycemia in adolescents globin (HbA1c) and oral
with PCOS. glucose tolerance test
Methods: This was a retrospective cohort study of 68 adolescents with PCOS seen in the Boston (OGTT) are useful for iden-
Childrens Hospital Division of Adolescent Medicine between 2008 and 2011 and not known to tifying adolescents with
have diabetes. Prevalence of dysglycemia (impaired fasting glucose, IGT, increased risk for diabetes, polycystic ovary syndrome
or diabetes mellitus as diagnosed by fasting plasma glucose, 2-hour OGTT, and/or HbA1c) and (PCOS) with dysglycemia
sensitivity and specicity of HbA1c for diagnosing dysglycemia compared with OGTT were and may identify different
assessed. populations of girls at risk for
Results: Twenty-four participants had abnormal glucose testing, including one participant (1.5%) long-term cardiometabolic
who met criteria for diabetes mellitus and 23 participants (34%) who met criteria for impaired consequences. Future stud-
fasting glucose/IGT/prediabetes. More patients were identied as having dysglycemia by HbA1c ies are necessary to deter-
than OGTT. Compared with OGTT, HbA1c had a sensitivity of 60% and a specicity of 69% for mine which test best predicts
diagnosing dysglycemia. future dysglycemia or pro-
Conclusions: In adolescents with PCOS, HbA1c had moderate sensitivity and specicity for gression to diabetes in lon-
detecting dysglycemia compared with OGTT. Clinicians should be aware that both tests have gitudinal follow-up.
benets and limitations, and the optimal test for follow-up requires further study.
2014 Society for Adolescent Health and Medicine. All rights reserved.
PCOS affects 5%e10% of reproductive-age women and often tolerance (IGT) and progression to diabetes mellitus (DM) [3e5].
presents during adolescence [1]. Many patients with PCOS are Because IGT can present as early as adolescence in patients with
insulin resistant even if they are not overweight or obese [2], PCOS [6,7], health care providers are tasked with initiating the
placing them at increased risk for developing impaired glucose evaluation for IGT and DM in these young women. The Androgen
Excess Society recommends IGT screening for adolescent and
Conicts of Interest: The authors have no conicts of interest to disclose. adult women with PCOS who are obese, over 40 years of age, or
* Address correspondence to: Holly Catherine Gooding, M.D., Division of have a personal history of gestational diabetes or a family history
Adolescent and Young Adult Medicine, Boston Childrens Hospital, 300 Long- of type 2 diabetes [8] with a 2-hour OGTT, as fasting plasma
wood Avenue, Boston, MA 02115.
E-mail address: holly.gooding@childrens.harvard.edu (H.C. Gooding).
glucose (FPG) alone is less sensitive in this population [3]. The
1054-139X/$ e see front matter 2014 Society for Adolescent Health and Medicine. All rights reserved.
http://dx.doi.org/10.1016/j.jadohealth.2013.12.020
80 H.C. Gooding et al. / Journal of Adolescent Health 55 (2014) 79e84
American Diabetes Association (ADA) and American Academy of electronic patient registration system. Family history was
Pediatrics recommend screening obese youth with two or more recorded by the treating clinician.
risk factors for diabetes (one of which can be PCOS) with FPG [9].
Given that 70% of adolescents with PCOS are obese, the majority Measures
will qualify for screening [10]. Despite these recommendations,
pediatric adolescent and gynecology providers report evalu- Body mass index (BMI) was calculated using the formula:
ating only 25%e60% of adolescents diagnosed with PCOS for weight (kg)/height2 (m). Sexual maturity rating of pubertal
IGT/DM [11]. development for both breast and pubic hair was recorded by the
In 2010, the ADA endorsed the use of HbA1c for the diagnosis treating clinician. Blood pressure was obtained by GE Dinamap
of diabetes in adults [12]. HbA1c is a reection of the average DPC100X-US, measured in the right arm resting at heart level
blood glucose levels over the life of a red blood cell (approxi- with both feet on the ground per standard clinic policy.
mately 90 days) and is closely associated with microvascular
disease in diabetes [13]. HbA1c is less subject to intraindividual Laboratory assays
variation and does not require that the patient be fasting. This
latter advantage may be particularly relevant for adolescent An OGTT was performed in the morning on a random day of
patients, for whom there may be barriers to access to care and the menstrual cycle in 68 fasting individuals who were instruc-
who may be less likely to return for follow-up [14]. However, the ted to consume a normal amount of carbohydrates for the 3 days
Androgen Excess Society concluded in their 2010 position paper preceding the test in either the BCH clinical laboratory or another
that there were insufcient data on the use of HbA1c testing in certied laboratory (n 4). Blood glucose was measured at
PCOS to recommend its use at that time [8]. baseline (FPG) and at 120 minutes (BG 120) after a 75 g oral
Several recent studies have compared screening for dysglyce- glucose load. ADA criteria [12] were used to classify participants
mia by HbA1c with either FPG or OGTT in adult [15e21] and as having normal fasting glucose or glucose tolerance (FPG <
pediatric [22,23] groups, including adult women with PCOS 100 mg/dl or BG 120 < 140 mg/dl), impaired fasting glucose (IFG)
[24,25]. Most of these studies have found only fair agreement or IGT (FPG, 100e125 mg/dl or BG, 120 140e199 mg/dl), or dia-
among HbA1c, FPG, and OGTT. The role of HbA1c as a screening betes (FPG 126 mg/dl or BG 120 200 mg/dl) based on their
tool in adolescents with PCOS has not been claried and may differ OGTT results. For HbA1c results (turbidimetric inhibition
from its use in adults due to the lower prevalence of dysglycemia immunoassay, BCH clinical laboratory), we classied subjects as
in adolescents. We thus sought to examine the prevalence of normal if HbA1c < 5.7%, increased risk for diabetes if HbA1c
dysglycemia as diagnosed by HbA1c and OGTT in adolescents with 5.7%e6.4%, or diabetic if HbA1c 6.5% based on ADA criteria.
PCOS to help clinicians identify strategies for clinical practice. Additional laboratory values (total cholesterol, low density
lipoprotein cholesterol, high density lipoprotein cholesterol, tri-
Methods glycerides, alanine aminotransferase, total and free testosterone
levels, dehydroepiandrosterone sulfate, and sex hormone bind-
Participants ing globulin) were obtained using standard assays in either the
BCH clinical laboratory or another certied laboratory. Elevation
We used International Classication of Diseases (ICD-9) billing in testosterone levels was dened as values greater than the
codes to identify patients seen in our Adolescent and Young Adult normative range in that laboratory for a given patients sexual
Medicine Clinic from 2008 to 2011 who were diagnosed with PCOS maturity rating.
by their treating clinician and underwent both a 2-hour OGTT and
HbA1c testing within 90 days (N 68, 45 of whom had both tests Statistical analysis
on the same day), as well as N 125 individuals with PCOS who
did not have both OGTT and HbA1c testing for a comparison Descriptive statistics are presented as mean (SD) or frequency
sample. If multiple OGTT or HbA1c results were recorded within (%). Median (interquartile range) are presented in lieu of mean
the study period, those closest to the initial evaluation were used (SD) for continuous factors with non-normal distributions upon
for analysis. Data on demographic characteristics, social and family visual inspection. Demographic and clinical factors at initial visit
history, physical examination ndings, laboratory studies, pelvic were compared by HbA1c categories. All tests were performed at
ultrasound results, and treatment were collected through a stan- an alpha level of .05 using SAS software, version 9.3 (SAS Institute
dardized chart abstraction tool and entered into the RedCap Inc., Cary, NC). Independent two-tailed t tests were used to assess
database [26]. All participants met 1990 National Institutes of differences between participants classied as normal and
Health (NIH) Consensus Criteria for PCOS and thus had irregular abnormal by HbA1c on continuous factors. Wilcoxon rank-sum
menses and either clinical or biochemical hyperandrogenism [27]. nonparametric test was used to assess differences between
Participants were excluded if they had a prior diagnosis of type I or HbA1c categories on continuous factors with a non-normal dis-
type II diabetes or IGT or if they were taking medications known to tribution where indicated in results and tables. Chi-square or
affect the results of an OGTT or HbA1c (such as metformin). This Fishers exact test (where appropriate) was used to assess dif-
study was approved by the Boston Childrens Hospital (BCH) ferences between the HbA1c categories on categorical factors.
Committee on Clinical Investigations. Diagnostic characteristics of FPG and HbA1c were compared
using OGTT in terms of the predictive ability of these tests to
Demographic information classify participants as having normal or abnormal glucose
ndings. Measures of the diagnostic ability calculated were
Age (in months/years) was dened by subtracting the date of sensitivity, specicity, positive and negative likelihood ratios, and
chart extraction from the participants date of birth. Race/ the kappa statistic as a measure of overall agreement between the
ethnicity was obtained from self-reported data from the test and the gold standard. Empirically derived cut points that
H.C. Gooding et al. / Journal of Adolescent Health 55 (2014) 79e84 81
maximize sensitivity and specicity were also generated from the Table 1
data. Receiver operating characteristic (ROC) curves were con- Characteristics of study population at rst test by HbA1c categories (N 68)
structed using the binary OGTT result classication of normal Participant characteristic Normal Abnormal p Valueb
and abnormal as the gold standard to determine the predictive (<5.7%) (5.7%)
(n 44) (n 24)
ability of FPG and HbA1c at the range of all possible cut points
for these tests. To determine whether the test had good pre- n (%)a or n (%)a or
mean (SD) mean (SD)
dictive ability to discriminate glucose ndings, the areas under
the ROC curve (AUC) were compared with the null value of .5 Age (years) 16.6 (2.4) 17.1 (2.8) .43
(discrimination due to chance) using a ManneWhitney test to Race .03
White 30 (79%) 8 (21%)
construct 95% condence intervals (CIs) for the AUC. A test of the
Black or African-American 3 (38%) 5 (63%)
contrast between the ROC curves for HbA1c to FPG in reference Hispanic or Latino 3 (43%) 4 (57%)
to OGTT was also computed to discern whether the tests differed Other/Multiracial 2 (50%) 2 (50%)
from each other in terms of discriminatory ability. Family history of diabetes 13 (54%) 11 (46%) .18
Family history of PCOS 3 (75%) 1 (25%) .99
BMI z-score 1.8 (1.1) 1.9 (.7) .84
Results BMI categoryc .99
Normal 5 (71%) 2 (29%)
Demographic and clinical characteristics of the 68 patients Overweight 3 (60%) 2 (40%)
who completed both OGTT and HbA1c testing within 90 days, Obese 30 (63%) 18 (38%)
Systolic BP (mm Hg) 116.8 (14.8) 116.0 (14.4) .83
stratied by HbA1c category, are listed in Table 1. The median Diastolic BP (mm Hg) 65.8 (10.5) 66.6 (10.5) .79
time between OGTT and HbA1c testing was 0 days (interquartile Total cholesterol (mg/dl) 150.5 (24.8) 166.6 (23.0) .10
range 23.5). Findings did not differ when analysis was restricted LDL cholesterol (mg/dl) 79.4 (25.9) 104.8 (24.9) .04
to the 45 individuals who had OGTT and HbA1c testing on the HDL cholesterol (mg/dl)* 44.1 (18.3) 41.55 (13.05) .75
Triglycerides (mg/dl)* 93.5 (90) 105 (64.5) .75
same day. Compared with 125 patients with PCOS seen during
ALT (IU/L)* 21 (14) 19 (8.5) .34
the same period who did not have both OGTT and HbA1c testing, Total testosterone (ng/dl) 49.7 (22.4) 44.8 (21.9) .43
patients who had both tests performed had higher BMI Z-scores Free testosterone (pg/dl)* 7.85 (6.7) 7.05 (6.55) .70
(1.83 vs. 1.27, p < .01). However, patients with both tests per- DHEAS (mg/dl)* 204.1 (141.5) 222.3 (340.8) .38
formed were no more likely to have a family history of DM and SHBG (nmol/L)* 16.0 (21.0) 18.0 (10.0) .78
PCOS diagnosisd
did not differ from other PCOS patients with respect to race/ Clinical hyperandrogenism 39 (64%) 22 (36%) .99
ethnicity, age, or free testosterone (pg/dl). (n 61)
Subjects with an HbA1c 5.7% were more likely to be black Biochemical 37 (67%) 18 (33%) .19
(p .03) and to have higher low-density lipoprotein cholesterol hyperandrogenism (n 55)
(p .04) compared with individuals with a normal HbA1c. There Median (IQR) reported for continuous factors noted above by * due to non-
were no statistically signicant differences by HbA1c category in normal distribution.
other cardiometabolic risk factors, including blood pressure and ALT alanine aminotransferase; BMI body mass index; BP blood pressure;
DHEAS dehydroepiandrosterone sulfate; HbA1c hemoglobin A1c; HDL
BMI, nor in androgen levels or criteria met for PCOS diagnosis.
high density lipoprotein; IQR interquartile range; LDL low density lipopro-
Subjects with an HbA1c 5.7% had higher mean FPG levels and tein; PCOS polycystic ovarian syndrome; SD standard deviation; SHBG sex
were more likely to have an abnormal 2-hour glucose (Table 2). hormone binding globulin.
a
There were no statistically signicant differences by OGTT category Percentages are of those with complete data on a given variable. Totals for
in demographic variables, androgen concentrations, homeostasis each characteristic do not necessarily equal 68 due to missing data.
b
Independent two-tailed t tests were used to assess differences between
model assessment of insulin resistance, or cardiometabolic risk participants classied as normal and abnormal by HbA1c on continuous
factors (data not shown). factors. Wilcoxon rank-sum test was used to assess differences between the
One subject was classied as having diabetes; this subject met HbA1c categories on continuous factors with non-normal distribution (indicated
criteria for diabetes by both HbA1c and OGTT. Three patients were by *). Chi-square or Fishers exact test (where appropriate) was used to assess
differences between the HbA1c categories on categorical factors.
classied as having IFG on the basis of an abnormal FPG, while nine c
BMI categories were determined by pediatric percentiles for subjects under
subjects were classied as having IGT based upon BG 120. Twenty- 18 years of age and by adult cutoffs for subjects 18 years of age or older.
three patients were classied at increased risk for diabetes by d
All subjects met 1990 NIH criteria for PCOS and thus had oligomenorrhea and
HbA1c. Due to the low number of patients with DM in our sample, either clinical or biochemical hyperandrogenism. Forty-nine (75%) of subjects
DM and IFG/IGT/increased risk for diabetes were considered had both clinical and biochemical hyperandrogenism.
Table 2
OGTT and FPG results by HbA1c categories (N 68)
FPG fasting plasma glucose; HbA1c hemoglobin A1c; OGTT oral glucose tolerance testing; SD standard deviation.
a
Independent two-tailed t tests were used to assess differences between participants classied as normal and abnormal by HbA1c on continuous factors. Chi-
square or Fishers exact test (where appropriate) was used to assess differences between the HbA1c categories on categorical factors.
dysglycemia. The majority were identied as at risk for dia- results on OGTTs performed on two separate days [30]. Thus while
betes by an abnormal HbA1c, but four individuals with dysgly- the sensitivity and specicity of HbA1c for diagnosing IGT and DM
cemia were identied only by 2-hour glucose levels and one was are indeed suboptimal in our study and other studies of adoles-
identied with FPG only. Compared with OGTT, HbA1c had a cents compared with OGTT, HbA1c testing may actually identify
sensitivity of 60.0% and a specicity of 69.0% for diagnosing some individuals not identied on the OGTT or FPG due to the
dysglycemia. intraperson variability of day-to-day glucose processing [31]. This
Similar to our study, two recent studies have looked at HbA1c may be particularly true for adolescents who often consume high
as a screening tool for diabetes in adult women with PCOS [24,25] glucose loads from sugar-sweetened beverages and other foods
and found only fair agreement between HbA1c and OGTT results, high in carbohydrate content [32]. Alternatively, elevated HbA1c
reporting a kappa statistic of .29 [24], a sensitivity of 35%, and a results may represent false-positive results. Long-term follow-up
specicity of 99% for the diagnosis of diabetes [25]. Interestingly, studies examining whether HbA1c or OGTT measured in adoles-
Magnussen et al. found HbA1c had a stronger correlation with cence best predicts the development of microvascular and mac-
other important cardiovascular risk factors (such as BMI and lipid rovascular complications of diabetes in adulthood are needed to
proles) than the OGTT [25]. We did not nd signicant differ- resolve this distinction.
ences in cardiometabolic risk factors by HbA1c or OGTT category Study limitations should be acknowledged. Our study was
in our adolescent sample, which may reect the relatively normal based upon a retrospective design and included patients with
cardiometabolic parameters in this younger age group. Impor- PCOS evaluated at one academic medical center, and thus results
tantly, neither test was more likely to identify dysglycemia in the may not be generalizable to other adolescent populations. Test
small number of nonobese adolescents with PCOS in our sample. ordering and completion were at the discretion of the provider
A recent editorial cautioned against the premature adoption of and patient, and not surprisingly, both tests were obtained more
HbA1c as a screening strategy in youth, noting that a criteria of often in adolescents who were obese, consistent with screening
6.5% for diagnosis of diabetes may be too high in children and guidelines. We report results based on one test date and as
adolescents [28]. Others have noted that at thresholds of 5.7% and intraindividual variation with glucose testing occurs, the ADA
6.5%, HbA1c testing may not be as cost effective as OGTT as a recommends two tests to conrm a diagnosis of diabetes [12].
screening test for dysglycemia [29]. However, it is important to Although OGTT and HbA1c testing were done on the same day in
recognize that the reproducibility of the OGTT in overweight only 66% of cases, the two tests are often done at different times
children is itself poor, with only 30% of youth having concordant in actual clinical practice, and our results did not change when
we restricted to the individuals who had both tests on the same
day. Although data are not available regarding the timing of
testing within the menstrual cycle, the majority of patients with
PCOS are anovulatory and thus expected to be in the follicular
phase. Finally, some participants may have already engaged in
lifestyle modications at the instruction of their treating clini-
cian before testing. Despite these limitations, this study provides
important data on both the high prevalence of abnormal glucose
testing in a population of predominantly overweight and obese
adolescents with PCOS and on the utility of HbA1c compared
with OGTT for diagnosing early problems with glucose homeo-
stasis in these young women.
Given the high prevalence of dysglycemia and the lack of a
single measure identied in our study and other studies [24,25],
a clear recommendation for a screening strategy in adolescent
Figure 1. Diagnosis of dysglycemia by FPG, 2-hour OGTT, or HbA1c in 29 of 68 women with PCOS cannot be made. We propose two possible
patients with PCOS. strategies, both of which require additional study. The rst,
H.C. Gooding et al. / Journal of Adolescent Health 55 (2014) 79e84 83
Table 3
Sensitivity and specicity of HbA1c and FPG for diagnosing dysglycemia in adolescents with PCOS compared with 2 hour OGTT (N 68)
HbA1c
Consensus cut point (5.7%) 60.0 69.0 1.93 .58 .18 .645
Empirically determined cut point (>5.6%) 60.0 69.0 1.93 .58 .18 .645
Fasting glucose
Consensus cut point (>99 mg/dl) 10.0 97.0 3.33 .93 .09 .533
Empirically determined cut point (>87 mg/dl) 80.0 62.0 2.11 .32 .23 .710
FPG fasting plasma glucose; HbA1c hemoglobin A1c; LR positive likelihood ratio; LR negative likelihood ratio; OGTT oral glucose tolerance test; PCOS
polycystic ovary syndrome.
[18] Lipska KJ, De Rekeneire N, Van Ness PH, et al. Identifying dysglycemic providing translational research informatics support. J Biomed Inform 2009
states in older adults: Implications of the emerging use of hemoglobin A1c. Apr;42:377e81.
J Clin Endocrinol Metab 2010 Dec;95:5289e95. [27] Zawadski J, Dunaif A. Diagnostic criteria for polycystic ovary syndrome:
[19] Olson DE, Rhee MK, Herrick K, et al. Screening for diabetes and pre- Towards a rational approach. In: Dunaif A, Givens JR, Haseltine FP, et al., eds.
diabetes with proposed A1C-based diagnostic criteria. Diabetes Care Polycystic Ovary Syndrome. Boston, MA: Blackwell Scientic Publications;
2010 Oct;33:2184e9. 1992:377e84.
[20] James C, Bullard KM, Rolka DB, et al. Implications of alternative denitions of [28] Kester LM, Hey H, Hannon TS. Using hemoglobin A1c for prediabetes and
prediabetes for prevalence in U.S. adults. Diabetes Care 2011 Feb;34:387e91. diabetes diagnosis in adolescents: Can adult recommendations be upheld
[21] Mann DM, Carson AP, Shimbo D, et al. Impact of A1C screening criterion on for pediatric use? J Adolesc Health 2012 Apr;50:321e3.
the diagnosis of pre-diabetes among U.S. adults. Diabetes Care 2011 Oct; [29] Wu EL, Kazzi NG, Lee JM. Cost-effectiveness of screening strategies for
33:2190e5. identifying pediatric diabetes mellitus and dysglycemia. JAMA Pediatr
[22] Nowicka P, Santoro N, Liu H, et al. Utility of hemoglobin A(1c) for diagnosing 2013;167:32e9.
prediabetes and diabetes in obese children and adolescents. Diabetes Care [30] Libman IM, Barinas-Mitchell E, Bartucci A, et al. Reproducibility of the oral
2011 Jun;34:1306e11. glucose tolerance test in overweight children. J Clin Endocrinol Metab 2008
[23] Lee JM, Wu EL, Tarini B, et al. Diagnosis of diabetes using hemoglobin A1c: Nov;93:4231e7.
Should recommendations in adults be extrapolated to adolescents? [31] Inzucchi SE. Clinical practice. Diagnosis of diabetes. N Engl J Med 2012 Aug
J Pediatr 2011 Jun;158:947e52. e941-943. 9;367:542e50.
[24] Hurd WW, Abdel-Rahman MY, Ismail SA, et al. Comparison of diabetes [32] Forbes LE, Storey KE, Fraser SN, et al. Dietary patterns associated with
mellitus and insulin resistance screening methods for women with poly- glycemic index and glycemic load among Alberta adolescents. Appl Physiol
cystic ovary syndrome. Fertil Steril 2011 Oct;96:1043e7. Nutr Metab 2009 Aug;34:648e58.
[25] Velling Magnussen L, Mumm H, Andersen M, et al. Hemoglobin A1c as a [33] Herman WH, Cohen RM. Racial and ethnic differences in the relationship
tool for the diagnosis of type 2 diabetes in 208 premenopausal women between HbA1c and blood glucose: Implications for the diagnosis of dia-
with polycystic ovary syndrome. Fertil Steril 2011 Nov;96:1275e80. betes. J Clin Endocrinol Metab 2012 Apr;97:1067e72.
[26] Harris PA, Taylor R, Thielke R, et al. Research electronic data capture [34] Lee JM, Eason A, Nelson C, et al. Screening practices for identifying type 2
(REDCap)ea metadata-driven methodology and workow process for diabetes in adolescents. J Adolesc Health 2013 Aug 19.