At mandatory school faculty meetings, we presented information about the study. Faculty members then invited study team members to their classrooms. Administrators permitted access only to elective courses containing health objectives (eg, child development, health occupations, Junior Reserve Officer Training Corps), which included less than half of the student body. In addition to giving presentations, the study team promoted the study through health fairs, information tables in high-traffic areas, flyers, and public address announcements.

Study personnel maintained a presence on campus during the data collection phase. Although school officials were supportive, they prohibited students from leaving class to participate. The school nurse and study coordinator instead scheduled appointments. Participants completed the study before school or during free time. The study coordinator and principal investigator collected data, including a fasting blood specimen, and height, weight, and waist circumference (WC) measurements. Participants rescheduled appointments if they arrived nonfasting. After we obtained fasting measures, participants received juice and crackers. The school nurse coordinated counseling and follow-up for abnormal anthropometric values.

The sample included students aged 15 to 18 years. Students who self-reported pregnancy or physical conditions influencing anthropometric measures, such as amputation, were excluded. Participants reported to the nurse's office for data collection. Data were collected prospectively on 325 participants (16% participation rate) Monday through Friday between February 2004 and March 2005, except in summer months when school was not in session.

Students self-reported age, birth date, grade, and ethnicity. We categorized age as early adolescence (aged <15 y), middle adolescence (aged 15-17 y), and late adolescence (>17 y).

We used standard equipment and methods to measure weight and height. Participants removed their shoes and stood motionless on a portable electronic scale with feet slightly apart and weight equally distributed. We recorded height by using a portable stadiometer with participants standing tall, holding their breath, and looking straight ahead. We calculated body mass index (BMI) as weight in kilograms divided by height in meters squared (kg/m²). We classified participants with a BMI in the 85th percentile to less than 95th percentile by age and sex as overweight and those with a BMI in the 95th percentile or higher as obese, on the basis of Centers for Disease Control and Prevention growth charts (20).

We determined central adiposity by measuring WC as participants remained in a standing position, breathing normally. We used Fernandez and colleagues' (21) estimates of WC percentiles by age, sex, and height for Mexican American adolescents, which were derived from the National Health and Nutrition Examination Survey. We considered WC in the 75th percentile or higher for age, sex, and height as high (21).

We measured blood pressure according to the Fourth Report on the Diagnosis, Evaluation, and Treatment of High Blood Pressure in Children and Adolescents protocol (22). For analysis, we used the mean of 2 systolic values taken a few minutes apart by the same examiner. Using standards from the National High Blood Pressure Education Program that account for sex, age, and height, we considered systolic blood pressure in the 90th percentile or higher as high (22).

After participants fasted for 10 hours, we collected blood to measure fasting blood glucose, insulin, and lipids. We used standard equipment and methods to determine blood glucose and insulin levels. We estimated IR with the formula for homeostasis model assessment for insulin resistance (HOMA-IR) (23,24). We also used standard equipment to measure plasma lipids, including triglycerides, total cholesterol, high-density lipoprotein (HDL) cholesterol, and low-density lipoprotein (LDL) cholesterol.

We used the 2010 American Diabetes Association Standards of Medical Care to define fasting blood glucose values of at least 7.0 mmol/L as elevated, and values between 5.6 and 6.9 mmol/L as impaired fasting glucose (25). We defined IR as HOMA-IR of at least 3.16 (23). We considered triglyceride values of at least 11.1 mmol/L (logarithmically transformed for analysis) as high (26,27). We used cutpoints for HDL, LDL, and total cholesterol in adolescents recommended by the National Cholesterol Education Program Adult Treatment Panel for analyses (26,27).

No observations were eliminated because of missing values; however, where values for individual variables were missing, we specified denominators. For univariate analyses, separate logistic regression models used IR as the dependent variable; BMI, WC, age, sex, HDL cholesterol, LDL cholesterol, triglycerides, and total cholesterol were the independent variables. Multivariate logistic regression with IR as the dependent variable controlled for potential confounding and effect modification. We used SAS version 9.1.3 (SAS Institute, Inc, Cary, North Carolina) to perform the analyses.

Of 2,064 enrolled students, 337 adolescents consented to participate; 325 completed 1 data collection examination appointment. We used standard protocols with reminder telephone calls and attempts to reschedule missed appointments. Missing values generally resulted from inability to obtain sufficient blood volume specimens or from incomplete examinations. Twelve participants failed to arrive for their appointments. After 5 attempts to reschedule, we withdrew those participants from the study.

Of the 325 adolescents, 65% (n = 211) were girls, 92% (n = 298) reported being of Mexican American descent, and 66% (n = 213) reported that all 4 grandparents were born in Mexico. Most participants (average age, 16 y) attended 9th (43%) or 10th (25%) grade compared with the district's enrollment for 9th and 10th grade of 37% and 28%, respectively.

Half of participants were categorized as overweight or obese. The largest proportion, 27%, were obese (boys 33%, girls 24%). More than one-third of participants exhibited WC measures indicative of central adiposity.

Abnormal metabolic values occurred with regularity in these adolescents who might otherwise be considered healthy (Table 1). Obese participants were nearly 10 times more likely to exhibit IR than participants who were not obese (Table 2). Overweight participants were 2.7 times more likely to exhibit IR than those who were not overweight. Although significant, total cholesterol's association with IR was one of the weakest relationships noted at the univariate stage. We found that central adiposity consistently demonstrated close association with IR. Of participants with high WC, 53% exhibited IR, compared with 10% of participants with normal WC. Therefore, WC was strongly associated with IR and contributed to identifying adolescents in this sample who were most at risk for type 2 diabetes. We found no significant association between IR and sex, age, or fasting blood glucose.

On the basis of these findings, we tested IR as the outcome variable in a multivariate model with WC and HDL cholesterol (Table 2). Age, sex, and BMI remained in the model because of their associations with IR in other populations, even though BMI correlated with WC (r²= 0.60; P < .001). The multivariate model excluded total cholesterol, triglycerides, and LDL cholesterol because of weaker association in univariate analysis and because these lipid values covary inversely with HDL cholesterol. Because of the high number of participants with elevated HDL cholesterol (66%), which is considered modifiable with the use of nonpharmacologic physical activity interventions, this variable remained in the model. The association with IR remained nonsignificant for age and sex in the multivariate model. Participants with high WC measures and elevated HDL values were more likely to exhibit IR than those with normal values.