Hemoglobin A1c and Type 2 Diabetes Incidence Among Adolescents With Overweight and Obesity

Key Points Question What is the risk of future type 2 diabetes (T2D) among adolescents with overweight and obesity as indicated by hemoglobin A1c (HbA1c) levels? Findings In this cohort study of 74 552 adolescents aged 10 to 17 years with overweight or obesity, T2D incidence increased from 1 to 69 individuals per 1000 person-years as baseline HbA1c increased from less than 5.5% to 6.3% to 6.4%, with the greatest increase beyond HbA1c 6.0%. In multivariable analyses, T2D risk was 9-fold, 23-fold, and 72-fold higher for baseline HbA1c levels of 5.9% to 6.0%, 6.1% to 6.2%, and 6.3% to 6.4%, respectively, compared with a baseline level below 5.5%. Meaning These findings suggest that T2D surveillance in adolescents should be tailored based on HbA1c level, among other risk factors.


Introduction
Over the past 2 decades, prediabetes and type 2 diabetes (T2D) have increased among adolescents, parallel to the increase in childhood obesity. 1 From 1999 to 2018, the prevalence of prediabetes in adolescents aged 12 to 19 years increased from 12% to 28%. 2 From 2002 to 2015, the incidence of T2D in adolescents aged 10 to 19 years increased by 5% per year to 14 per 100 000 person-years. 3 From 2001 to 2017, the prevalence of T2D increased by 95% to 67 per 100 000 individuals. 4 Compared with adolescents with type 1 diabetes (T1D), adolescents with T2D have substantially greater cardiovascular risk burden (hypertension, obesity, or dyslipidemia) and microvascular complications (kidney disease, retinopathy, or neuropathy) [5][6][7] and become young adults with much higher rates of cardiovascular disease and mortality. 8,9Thus, it is important to identify adolescents with the highest risk of developing T2D who may benefit from increased surveillance, targeted lifestyle intervention, and other treatment considerations before adulthood.
The American Diabetes Association (ADA) recommends diabetes screening for at-risk adolescents with overweight or obesity after pubertal onset or age 10 years, whichever occurs earlier, based on several criteria. 10,11These criteria include a body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) at or above the 85th percentile for age and sex and at least 1 additional risk factor, such as a maternal history of diabetes during the child's gestation, family history of T2D, race and ethnicity (American Indian or Alaska Native, Asian or Pacific Islander, Black, or Hispanic), or a sign or condition associated with insulin resistance (acanthosis nigricans, hypertension, dyslipidemia, polycystic ovarian syndrome, or small-for-gestational-age birth weight). 10,11These guidelines would currently target approximately one-quarter of US adolescents. 12abetes screening can be performed using fasting glucose, 2-hour glucose during an oral glucose tolerance test (OGTT), or hemoglobin A 1c (HbA 1c ). 10,11Of these screening measures, nonfasting HbA 1c assessment is more practical for adolescents because it is more convenient, time efficient, less variable, and more reproducible. 13In addition, use of HbA 1c results has been shown to increase diabetes screening among at-risk adolescents. 14continuous association between HbA 1c and risk of T2D has been established in adults.15,16 Based on these observations, adults with an HbA 1c level of 5.7% to 6.4% are considered to have increased risk of developing T2D and are classified as having prediabetes.11 The ADA recommends these same HbA 1c thresholds to classify prediabetes in adolescents, although few population studies have validated the exact HbA 1c cut point of 5.7% in adolescents as indicative of future T2D.10,11 Data regarding T2D risk at higher HbA 1c levels are also limited and may be important for guiding screening frequency among adolescents with overweight and obesity, especially in racially and ethnically diverse populations.This study aimed (1) to determine T2D incidence by baseline HbA 1c levels in a diverse population of adolescents with overweight and obesity and (2) to identify clinically relevant HbA 1c thresholds associated with increased risk of T2D so that surveillance of high-risk populations can be optimized.

Design, Setting, and Study Population
This retrospective observational cohort study was conducted at Kaiser Permanente Northern California (KPNC) using electronic health record data (January 1, 2010, to December 31, 2019).The large KPNC integrated health care delivery system provides care to 4.4 million members in northern California, and approximately one-fifth of these individuals are aged younger than 20 years. 17The KPNC Institutional Review Board approved this study, and a waiver of informed consent was

Baseline Variables
Height and weight measurements from ambulatory visits were used to calculate BMI, with BMI percentiles for age and sex determined using the US Centers for Disease Control and Prevention growth chart reference data. 18The BMI categories for age and sex were defined as overweight (BMI 85th to <95th percentile) and obesity (BMI Ն95th percentile), subcategorized as moderate obesity (BMI 100% to <120% of 95th percentile) and severe obesity (BMI Ն120% of 95th percentile).Race and ethnicity was determined from patient-or family-reported data in health records and administrative databases and was classified as Asian or Pacific Islander, Black, Hispanic, non-Hispanic White (hereinafter White), or other race or ethnicity (including American Indian or Alaska Native, multiple races or ethnicities, or unknown race or ethnicity).These data were collected because T2D risk varies by race and ethnicity.Because neighborhood socioeconomic factors can influence T2D risk, 19 we obtained Neighborhood Deprivation Index (NDI) scores, which ranged from −2.1 (lower estimated deprivation) to 4.4 (greater estimated deprivation) in our cohort.The NDI measure is derived using area of residence and US Census tract-level data relating to several socioeconomic factors (neighborhood wealth, income, education, occupation, and housing conditions). 20seline HbA 1c levels were reported from a single KPNC regional laboratory using the following analyzers: Modular P Tina-quant (2010), Integra 800 Tina-quant Gen. 2 (2011-2017), and Cobas c513 Tina-quant Gen. 3 (2017-2018; all from Roche Diagnostics), with all standardized to the National Glycohemoglobin Standardization Program (NGSP).During the period of use, the mean HbA 1c biases for each analyzer were −0.02, 0.07, and 0.01, respectively, compared with NGSP reference values between 5.4% and 6.4%. 21Using baseline HbA 1c data from this cohort, the mean (SD) HbA 1c levels for each analyzer were 5.35% (0.28%), 5.52% (0.26%), and 5.30% (0.26%), respectively.Other glycemic measures included fasting glucose, random glucose, and 2-hour glucose level during an OGTT.Diabetes range values were defined with HbA 1c (Ն6.5%;48 mmol/mol), fasting glucose (Ն126 mg/dL; 7.0 mmol/L), random glucose (Ն200 mg/dL; 11.1 mmol/L), or 2-hour glucose during an OGTT (Ն200 mg/dL; 11.1 mmol/L) using ADA-recommended glycemic thresholds. 11

Outcome Ascertainment
Individuals with possible incident diabetes were initially identified from clinical diagnoses of diabetes by health care providers or any glycemic measure in the diabetes range occurring after baseline HbA 1c measurement.Medical record review was then conducted by a pediatric endocrinologist (F.M.H.) to identify individuals with incident diabetes, diagnosis date, and diabetes type.A random sample of 10% of individuals was also reviewed by a second pediatric endocrinologist (L.C.G.) to confirm diagnostic concordance.Individuals with incident diabetes were defined based on at least 1 glycemic measure in the diabetes range, using the aforementioned ADA glycemic thresholds (but not the ADA criteria of Ն2 glycemic measures in the diabetes range).To reduce false-positivity rates, fasting glucose levels of 126 to 199 mg/dL (7.0-11.0mmol/L) were excluded if the following occurred: (1) the individual was documented to be not fasting or the test was performed after 12 PM, (2) repeat fasting glucose was less than 100 mg/dL (5.6 mmol/L), (3) 2-hour glucose during an OGTT was less than 140 mg/dL (7.8 mmol/L), or (4) HbA 1c was less than 5.9% within the following 2 weeks and diabetes was not subsequently confirmed.Random glucose at or above 200 mg/dL (11.1 mmol/L) was also excluded if measured in an inpatient, emergency department, or ambulatory procedure setting and diabetes was not subsequently confirmed, unless the encounter was related to diabetes.Individuals were classified as having T1D if there was at least 1 positive diabetes autoantibody (eg, glutamic acid decarboxylase 65 [GAD65], insulinoma-associated protein 2, islet cell, or insulin autoantibodies) or based on physician-assigned diagnosis if diabetes autoantibodies were not measured (only in 7% of individuals with T1D).All other individuals were classified as having T2D except for those with maturity-onset diabetes of the young (MODY) or secondary diabetes.
Secondary diabetes included steroid-induced hyperglycemia and cases related to chronic pancreatitis, asparaginase exposure, or hyperglycemia associated with major surgery.Individuals who developed gestational diabetes that resolved after pregnancy were not classified as having diabetes.

Statistical Analysis
Baseline differences between subgroups were compared using analysis of variance for continuous variables and the χ 2 test for categorical variables.Individuals were followed through 2019, with follow-up censored at membership disenrollment (gap >6 consecutive months), death, or development of diabetes.The Kaplan-Meier method was used to determine 5-year cumulative incidence of T2D by HbA 1c level.In addition, the T2D incidence rate was calculated per 1000 personyears with 95% CIs.Cox proportional hazard regression analyses were performed to examine the association of baseline HbA 1c and risk of T2D, accounting for sex, age, BMI category, race and ethnicity, and NDI quartile, reporting adjusted hazard ratios (HRs) and 95% CIs.Stratified analyses were also conducted by sex, BMI category, and race and ethnicity.In sensitivity analyses, we limited the multivariable analyses to the subset with at least 1 follow-up glycemic measure and additionally conducted analyses censoring at the last glycemic measure.All analyses were conducted using SAS, version 9.4 (SAS Institute Inc).A 2-sided P < .05 was used as the threshold for statistical significance.
Data abstraction and analyses were conducted from January 1, 2020, to November 16, 2023.
The overall incidence rate of T2D during follow-up was 2.1 (95% CI, 1.9-2.3)per 1000 personyears.As baseline HbA 1c increased (from <5.5% to 5.9%-6.0%,6.1%-6.2%,and 6.3%-6.4%),T2D incidence increased exponentially from 0.8 (95% CI, 0.6-0.9) to 8.1 ( ), were independently associated with T2D risk.Findings were similar in sensitivity analyses that excluded adolescents with no follow-up glycemic measure and additionally censored follow-up time at the last glycemic measure (Table 3).In analyses stratified by race and ethnicity (Table 4), T2D risk increased with increasing HbA 1c levels for all groups, but the magnitude of increase at higher HbA 1c levels tended to be less for Black and Hispanic adolescents compared with Asian or Pacific Islander and White adolescents.
The overall incidence of T2D among adolescents with overweight and obesity was low (2.1 per 1000 person-years) and was generally low even among adolescents with baseline HbA 1c in the lower prediabetes range (HbA 1c , 5.7%-6.0%), in which the T2D incidence was 3.8 per 1000 person-years.
However, larger incremental differences in risk were apparent for baseline HbA 1c above 6.0% and support findings from a previous study that identified 368 predominantly Hispanic adolescents with overweight or obesity and baseline HbA 1c of 5.7% to 6.4%, in which the incidence of subsequent diabetes (HbA 1c Ն6.5%) increased from 25 to 50 per 1000 person-years for baseline HbA 1c of 5.7% to 5.9% and 6.0% to 6.4%, respectively. 226][27][28][29][30] Although Black and Hispanic adolescents in

Limitations and Strengths
This study had several limitations.First, we cannot exclude the possibility of selection bias regarding who had baseline HbA 1c testing.Hemoglobin A 1c can also be affected by health conditions (eg, hemoglobinopathies, anemia, and glucose-6-phosphate dehydrogenase deficiency), which were not examined in this study. 11,13Second, we required only 1 glycemic measure in the diabetic range to define incident diabetes, rather than 2 glycemic measures, similar to other large epidemiologic studies. 12,36  Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); HbA 1c , hemoglobin A 1c ; HR, hazard ratio; NDI, Neighborhood Deprivation Index.
a Includes American Indian or Alaska Native race, multiple races or ethnicities, and unknown race or ethnicity.
c Scores ranged from −2.1 to 4.4; higher scores represent greater estimated neighborhood deprivation.Scores were not available for 0.3% of children.

JAMA Network Open | Pediatrics
HbA 1c and Type 2 Diabetes Incidence Among Adolescents With Overweight and Obesity obtained [exemption category 4, with criteria §46.104(d)(4)(iii) met].The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.The study cohort included KPNC members aged 10 to 17 years who had at least 1 HbA 1c measurement during 2010 to 2018 and a BMI at or above the 85th percentile for age and sex within the 12 months before or at the time of baseline HbA 1c measurement.Individuals were excluded if they had possible or confirmed preexisting diabetes, as evidenced by (1) a diagnosis of diabetes mellitus (International Classification of Diseases, Ninth Revision, Clinical Modification code 250.x or International Classification of Diseases, Tenth Revision, Clinical Modification codes E08.x to E13.x), (2) glycemic measurement in the diabetes range, or (3) receipt of glucose-lowering medication (eg, metformin, sulfonylurea, insulin, or other diabetes pharmacotherapy) before or at the time of baseline HbA 1c measurement based on pharmacy records.Individuals who were pregnant at the time of HbA 1c measurement were also excluded.A flowchart depicting identification and cohort assembly is shown in the eFigure in Supplement 1.

Figure .
Figure.Cumulative Incidence of Type 2 Diabetes Over 5 Years of Follow-Up by Baseline Hemoglobin A 1c (HbA 1c ) Level

Table 2 .
35cidence Rate per 1000 Person-Years (95% CI) of Type 2 Diabetes by Baseline HbA 1c Level, Stratified by BMI Category, Race and Ethnicity, and Sex a be considered for follow-up diabetes screening less frequently than once per year, as currently recommended by the ADA for individuals with prediabetes.35 c Includes American Indian or Alaska Native race, multiple races or ethnicities, and unknown race or ethnicity.dScores ranged from −2.1 to 4.4; higher scores represent greater estimated neighborhood deprivation.Scores were not available for 0.3% of children.could

Table 4 .
Third, 44.3% of individuals did not have a follow-up glycemic measure performed (required for T2D diagnosis), but sensitivity analyses restricted to those with at least 1 follow-up glycemic measure showed similar findings.Nonetheless, we cannot exclude potential follow-up bias and T2D underdiagnosis.Fourth, although many individuals with T2D did not have diabetes autoantibodies measured, only a very small subset (4.2%) persistently required insulin therapy, where we cannot entirely exclude the possibility of T1D.Finally, other factors that can influence risk of T2D, such as change in weight or BMI, cardiometabolic conditions, and pubertal timing (incompletely captured during routine care), were not examined.A major strength of our study is the inclusion of a large and diverse cohort receiving care in the same integrated health care setting where laboratory and pharmacy data, as well as clinical diagnoses, could be tracked in electronic health records.Our findings are notable in that 15.2% of individuals who developed T2D required insulin beyond 6 months of diagnosis.Previous studies have Multivariable Association of HbA 1c Levels and Risk of Type 2 Diabetes by Race and Ethnicity