Examination of Early Childhood Temperament of Shyness and Social Avoidance and Associations With Cardiometabolic Health in Young Adulthood

Key Points Question Is a childhood temperament characterized by shyness and avoidance of social interactions associated with cardiometabolic health? Findings In this cohort of 9491 children from the United Kingdom, children with an avoidant-shy temperament engaged in less physical activity in adolescence and attained lower social standing compared with children with other temperaments. These factors were associated with poor cardiometabolic health in adulthood. Meaning These findings suggest that early temperamental differences among children are associated with long-term health behaviors and psychosocial development, which carry significant implications for cardiometabolic risks.


eAppendix 3. Longitudinal Temperament Profiles Using Repeated Measures of Shyness and Sociability
First, we determined the internal consistency and temporal stability of shyness and sociability scales in this sample. The two scales showed high internal consistency (αshyness= .79 to .83; αsociability= .63 to .68), as well as high temporal stability across time, r'sshyness= .57 to .73; r'ssociability= .46 to .59 (Supplemental Table S2 below).
To examine whether different temperaments exist across childhood, joint trajectories of shyness and sociability from ages 3-6 were estimated using a non-parametric k-means longitudinal clustering method in the R package "kml3d" 4 . K-means is a simple form of unsupervised machine learning that uses an exploratory hill-climbing algorithm to detect clusters of homogenous subgroups within a larger heterogenous population; "kml3d" uses these principles to detect subgroups based on patterns of several repeatedly measured variables simultaneously. In this procedure, each observation is first arbitrarily assigned to a cluster based on the co-evolution of the variables, then optimal clustering is achieved by repeatedly calculating the mean of each cluster/trajectory and reassigning each observation to its nearest means until no further changes occur. The estimations were repeated 5000 times to obtain optimal solutions (1000 times for each of the 2-, 3-, 4-, 5-, 6-cluster solutions). This analysis included 9491 participants, who had temperament data in at least two of the three time points; the missing data point was imputed through linear interpolation.
Selection of the cluster solution was based on theory and fit/quality of partition statistics, including the Calinski-Harabasz, Davies-Bouldin index, BIC, and global average of posterior-probabilities (i.e., the probability that each participant belongs to their assigned trajectory). Supplemental Table S3 shows the fit/quality of partition statistics. The 4-cluster solution was selected, as the subgroups support prior theory about the four different types of temperament (see Figure 1 in main text): Avoidant-shy, Conflicted-shy, Extraverted, and Introverted). This solution also showed relatively good reliability and separation across indices.  Measurement and Processing. At ages 11, 13, and 15, children who attended the research clinics were asked to wear an Actigraph AM7164 2.2 accelerometer (Actigraph LLC, FL, USA) around the waist for a week, during waking hours, except when showering, bathing, or playing water sports. Accelerometry measures of physical activity has been validated in both children and adolescents 5 6 . The accelerometer detects acceleration and deceleration in a vertical plane as a combined function of movement frequency and intensity. Data were recorded as counts, averaged over one-minute periods. The data were downloaded with the Actigraph Reader Interface unite RIU-41A with RIU software (version 2.26B, Actigraph LLC). Data were considered valid and reliable, if the accelerometer had been worn for at least 10 hours a day for at least three days of the week, after deletion of missing data (i.e., periods with ≥10 minutes of successive zeros was regarded as non-wear time and were deleted) 6 .
MVPA is defined as any activity that is equivalent to the physiological stress of brisk walking for a person of average fitness 7 . Activities that pass the threshold of moderate intensity in vigor include any aerobic activity that causes increased heart rate and potential sweating (e.g., running, cycling, dancing). Other activities, including jogging, hill walking and racket sports, are also considered vigorous. In the Actigraph recordings, time spent in MVPA is identified as minutes with > 3600 accelerometer counts. The threshold of 3600 counts per minute was determined from a calibration study conducted in a subsample of 246 ALSPAC children 6 . This threshold is four times greater than the resting metabolic rate, approximates activity from brisk walking, and has predictive validity as it is associated with reduced fat mass in these children 7 . Weekly MVPA was calculated as the average minutes with >3600 counts per valid day; Logtransformation was used to reduce positive skewness in weekly MVPA values and used in further analyses. Weekly MVPA was moderately correlated across adolescence (r's=.39 to .46, p's <.001).
Confirmatory factor analysis of adolescent physical activity. To maximize the available data and provide a robust and valid measure of physical activity to use in analyses, we used a confirmatory factor analysis (CFA). Adolescent physical activity was measured as a latent construct using weekly MVPA measured at ages 11, 13, and 15 as indicators (Supplemental Figure S1). CFA of the measurement model revealed good fit and all loadings were > .40 (Supplemental Table S4 Note. N= 9457 in analyses, excluding those who reported type-1 diabetes and pregnant females. All analyses adjusted for sex, child ethnicity, childhood SES, maternal depressive symptoms, birthweight z-score, and BMI at age 24.