Principal component analysis reveals gender-specific predictors of cardiometabolic risk in 6th graders

The prevalence of obesity has increased significantly and is a primary cause of cardiometabolic comorbidities[1]. Evidence from the National Health and Nutrition Examination Survey (NHANES)[2] has indicated that approximately two-thirds of adolescents present with one or more risk factors, and that nearly 10% of this population has clustering of three or more cardiometabolic abnormalities[3]. Moreover, obesity and insulin resistance (IR) during childhood are demonstrated to track into adulthood[4‐6], and appear to be antecedents to adult-onset type 2 diabetes[7] and cardiovascular complications[8]. Evidence further suggests that risk for cardiovascular disease (CVD) originates during childhood[9], and that several risk factors such as obesity, glucose intolerance, and hypertension are associated with premature mortality[10]. However, the clinical value of defining and identifying pediatric patients with the metabolic syndrome (MetS) has been heavily debated, due largely to the instability of a dichotomized diagnosis[11].

Findings also reveal that MetS in childhood is a robust predictor of adult cardiovascular disease[8], and thus early prevention efforts are crucial among pediatric populations. However, adequate guidelines are at present lacking to delineate the structure and norms for risk factor stratification in pediatrics[11]. Barriers to a universally accepted definition of MetS include the fact that clinical end-points emerge infrequently at an early age, and that there is substantial variability in the established normal values across different ages and races[11]. Until recently[12], virtually no research has delineated the sex-specific patterns of cardiometabolic risk among adolescents. We have previously shown that sexual-dimorphic differences indeed have a genetic underpinning[13] that likely also impacts cardiometabolic risk beginning in early phases of life. As a result, the use of adult criteria to classify children and adolescent patients with cardiometabolic abnormalities may significantly misrepresent underlying risk in boys and girls.

Although previous studies have incorporated continuous MetS scores to improve robustness of cardiometabolic risk assessment[14], the respective “weight” of each component on absolute cardiometabolic risk is not accounted for, and thus limits the value of creating a MetScore for risk determination[15]. A more robust method of principal components analysis (PCA) has been proposed[16] that accounts for the loading coefficient of each variable. Although the loading coefficients are limited to the specific data set from which they are derived, this strategy enables the aggregation of key MetS risk components that are known to cluster, and also provides a weighted, continuous MetS risk outcome that is demonstrated to remain stable into young adulthood[17]. PCA has been recently used in pediatric cohorts to aid in the interpretation of statistical MetS clustering[18, 19] and subsequent reduction into orthogonal factors. However, the technique has never been incorporated for the purpose of identifying independent, biological and behavioral cardiometabolic risk predictors specific to boys and girls. Indeed, using a sex-stratified analysis will provide a better indication of the unique pattern of cardiometabolic component clustering, which may further identify at-risk pediatrics as well as inform tailored approaches to prevent the establishment of cardiometabolic comorbidities in adulthood. Therefore, the purpose of this investigation was to assess the influence of biological, behavioral and parental factors on sex-specific pediatric cardiometabolic risk, using a PCA-derived continuous MetScore.

Descriptive characteristics are presented in Table1. Approximately 19% of children were overweight and 24% were obese. Female participants were younger, had smaller WCs, higher triglyceride levels, lower fasting glucose levels, lower CRF (estimated VO₂), and were less likely to be physically active (see Additional file2 and Additional file3: Table S2).

We used principal components analyses to conduct a comprehensive investigation of cardiometabolic risk in a large cohort of 6th grade boys and girls. Our results reveal that both BMI and CRF are the two most influential independent predictors of overall risk in boys and girls, even after controlling for PA, family history of CVD, and parental characteristics (e.g. BMI, smoking status, age, etc.). The finding that CRF is inversely associated with metabolic risk has been previously shown[24, 25]; however, given that PA serves as a moderating variable, the extent to which CRF may impact metabolic risk is often debated. Nevertheless, our data substantiate the importance of CRF as an independent determinant of metabolic risk in children. Moreover, we also identified novel sex-specific biological, behavioral, and parental predictors of childhood cardiometabolic risk. Specifically, among girls parental age and paternal family history of CVD were each independently associated with an increased Metscore. For boys, maternal smoking was the only additional independent predictor of an increased MetScore. Certainly, future research is needed to provide added insight into the mechanisms behind sex differences in cardiometabolic risks and divergent associations with environmental, familial, and behavioral influences.

Our findings suggest a critical need for tailoring gender-specific approaches to combat the increased prevalence of pediatric obesity, in order to ameliorate the impending comorbidity burden into adulthood. Our data are aligned with a recent investigation which demonstrated the independent contribution of CRF to MetS diagnosis among European children[26]. According to those results, individuals categorized with MetS were reported to have approximately 20–25% lower CRF than children without. This is also consistent with results from the European Youth Heart Study[27] which revealed that even after adjusting for habitual PA, CRF was a robust independent predictor of cardiovascular disease risk. In a large U.S. cohort of children we now confirm the heightened importance of addressing CRF independent of PA and BMI. This is aligned with recent findings by Jago et al.[28], which revealed that in 6th graders both fatness and fitness were independently associated with cardiometabolic risk.

Sex-differences in obesity prevalence were also observed, such that a higher proportion of boys were categorized as obese compared to girls (26 vs. 22%). This finding has been observed in prior population surveys (i.e. NHANES data[29]), and suggests that males and females indeed follow a divergent sex-specific trajectory in the development of risk for CVD[30] and related comorbidities (e.g. non-alcoholic fatty liver disease[31]). Although at present there is a lack of cut-points for WC in pediatrics (thus forcing the use of sex- and age-adjusted BMI cut points), abdominal obesity has been indicated as a primary driver of cardiometabolic risk[32]. In the current cluster analysis, WC was associated with all cardiometabolic risk factors and carried the strongest loading coefficient within the continuous MetScore outcome. Previous work has demonstrated that WC in children is the best predictor of adulthood MetS[33], and thus our findings bolster the importance of monitoring abdominal adiposity early and throughout pediatric clinical care.

Cardiometabolic risk factors are known to aggregate in the family, which may be attributable to both hereditary and environmental influences[34, 35]. In addition to a highly correlated BMI among parents and children, mothers who were physically active were more likely to have active children, as compared to those with inactive mothers (Figure1). This was not the case for father’s activity level. Nevertheless, this is consistent with previous data indicating “familial clustering” in PA patterns[36], and suggests that optimizing children’s health is best achieved through comprehensive behavioral programs that involve the family-unit, and especially parental role modeling (particularly in mothers) through early adolescence.

Although maternal and paternal smoking status were significantly correlated with each other in our study, only maternal smoking was associated with MetScore in boys, and demonstrated a strong linear trend among girls (p = 0.06) (Figure2). The adverse impact of parental smoking on children’s cardiovascular risk has been well established[37], particularly that of maternal smoking during pregnancy[38]. Nevertheless, our data extend these previous findings, by revealing the negative influence of current maternal smoking status on children’s cardiometabolic health. It is possible that current maternal smoking status may be reflective of an increased level of secondhand exposure among children. Environmental tobacco smoke exposure has been demonstrated to be independently associated with the MetS in adolescents[39], and it could be speculated that second hand exposure is substantially higher around mothers given that they are frequently the primary caregiver. On the contrary, it may also suggest a longstanding history of smoking, including smoking during pregnancy. Future research is certainly warranted to unravel the mechanistic influences of maternal smoking and second hand/environmental exposure on children’s cardiometabolic health risk.

Despite the clear connection between environmental influences and children’s cardiometabolic health, we cannot ignore the impact of non-modifiable heritable attributes when considering MetS risks[40]. In the current analyses parental family history of early CVD was independently associated with multiple individual MetS risk factors among girls and boys. Family history of early CVD has long been recognized as a risk factor in offspring[41], although evidence comparing the differential influence of maternal versus paternal family history on children’s risk for disease has been controversial. The “fetal origin hypothesis of adult chronic disease”[42] implies that stronger maternal-offspring associations for CVD risks are expected than for paternal-offspring associations. Conversely, our results suggest that paternal family history confers greater risk to girl’s cardiometabolic health than maternal family history, as this was the only non-modifiable familial factor associated with MetScore. This observation is consistent with a Framingham Heart Study[43], which revealed that paternal premature CVD was more robustly associated with coronary artery and abdominal aortic calcification among third generation young adults, as compared to maternal premature CVD.

The current findings using a robust analytic approach revealed numerous modifiable biological and behavioral factors that shared strong associations with boys’ and girls’ cardiometabolic health risk. Accordingly, both BMI and CRF represent ideal targets for prevention/treatment therapies in boys and girls, to reduce the cardiometabolic health burden. Moreover, the novel sex-specific risk factors identified such as parental age, paternal family history of CVD, and maternal smoking, also influence metabolic risk in girls and boys, respectively. Additional exploration of these factors and other fitness attributes (e.g. muscular strength[44]) may lead to a better understanding of family-based and gender-specific cardiometabolic risk in childhood, and sustainable approaches for chronic disease prevention.