BMC and BMD
were examined as indicators of bone health. Total body, lumbar spine, and proximal femur scans were performed using a Hologic QDR4500 bone densitometer (Hologic, Inc., Bedford, MA) and analyzed using software release 12.4. As BMC has been advocated as the appropriate measure to use during growth, total body BMC was the primary outcome owing to interest in factors that affect peak bone mass (22
). Lumbar spine, total hip, and femoral neck BMD were obtained because they are common sites of osteoporotic fracture.
were measured with the self-report Children’s Depression Inventory (CDI) (23
). The 27-item CDI demonstrated high reliability (α = 0.89). T-scores were used in the analyses (mean = 50; SD = 10), with scores ≥ 65 considered clinically significant.
symptoms were measured by the State Trait Anxiety Inventory for Children (STAIC). Two versions were used: the STAIC (24
) for ages <12 years and the STAI (25
) for ages ≥ 12 years. Reliability was high in this sample (α =.85–.89). Trait anxiety (T-score) was used in the analyses because it is stable and more likely to be related to bone health.
Smoking behavior was determined by questionnaire asking whether participants had ever smoked tobacco in their life. Categories included never (n = 104), one puff to two cigarettes (n = 54), three to 99 cigarettes (n = 53), or more than 100 cigarettes (n = 51). These categories are referred to as “graded smoking.”
was determined by the Diagnostic Interview Schedule for Children (DISC) (26
) using the adolescent report. The DISC is a standardized computer interview focusing on diagnostic symptoms of psychopathology using DSM
). Participants were categorized as no drinks (n
= 135), one to five drinks (n
= 59), and six drinks or more (n
= 67) within the past year (28
). These categories are referred to as “graded alcohol.”
Other participant characteristics were measured and examined as potential covariates. Race
was determined by parental report. Socioeconomic status (SES)
was estimated by parent report of occupation and education, (29
) with possible scores ranging from 8 (lower) to 66 (higher). Height was obtained by wall-mounted stadiometer (Holtain Ltd., Crosswell, United Kingdom), and weight was measured by digital scale (Scaletronix, Carol Stream, IL). Measures were obtained in triplicate by trained nursing personnel, and the mean was used. Pubertal maturation
was determined by physical examination by trained clinicians using visualization and breast palpation and inspection of pubic hair (30
) and was categorized by Tanner criteria. A nonrandom sample (n
= 27) examined by two raters showed 100% agreement on both stages. Gynecological age
was obtained by clinician interview asking each girl whether she ever had a menstrual period and, if yes, how old (year and month) she was at that time. Methodology to enhance accuracy was utilized (31
). The date of menarche was subtracted from the date of the visit to determine gynecological age. Hormone contraceptive history
was obtained by clinician interview including oral contraceptive pills (OCPs), depot medroxyprogesterone acetate (DMPA), transdermal patch, and intravaginal ring. Duration of use was categorized as 1 = 1–3 months, 2 = 4–6 months, 3 = 7–9 months, 4 = 10–12 months, 5 = >1 year but < 2 years, 6 = ≥ 2 years but < 3 years, and 7 ≥ 3 years. Due to differential effects of DMPA versus other contraceptive methods containing estradiol and a progestin on bone health, DMPA duration was examined separately. The duration of use for estradiol-containing methods (e.g., OCPs, transdermal patch, intravaginal ring) was summed to represent a combined duration of exposure. Summed scores were capped at 7 representing more than 3 years of combined use. Calcium intake
was measured by food frequency questionnaire reflecting intake of 19 calcium-rich foods (32
). Calcium supplement use was collected also. Physical activity
was estimated using the Physical Activity Questionnaire for Older Children (PAQ-C) (33
). Participants recalled their performance of moderate to vigorous physical activities within the last 7 days, and mean scores were calculated (1 [low] to 5 [high]). Serum 25-hydroxy-vitamin D [25(OH)D] concentration
was analyzed using radioimmunoassay. Inter- and intra-assay coefficients of variation ranged from 3.5%–4.4% and 11.1%–16-9%, respectively; sensitivity was 1.5 ng/mL.
Participant characteristics were compared among groups using ANOVA and Chi-square tests. Primary aims were examined using series of multiple regression analysis. Specifically, the effects of anxiety, depression, smoking, and alcohol intake on BMC and BMD were examined with separate regression models for each dependent variable: TB BMC and BMD of the lumbar spine, total hip, and femoral neck. Independent variables were smoking and alcohol use status and depressive and anxiety symptoms. Interactions between substance use and symptoms (e.g., smoking status by depressive symptoms) also were tested; all interaction terms were created from centered variables. As age is an important confounder of substance use and bone health, all analyses adjusted for chronological age. Additionally potential covariates (defined in the measures section) were tested. After accounting for age, only significant covariates (p < .05, 2-sided) were retained in subsequent models (age, race, height, weight, Tanner breast stage).