Peak oxygen uptake cut‐points to identify children at increased cardiometabolic risk – The PANIC Study

We aimed to develop cut‐points for directly measured peak oxygen uptake ( V˙O2peak ) to identify boys and girls at increased cardiometabolic risk using different scaling methods to control for body size and composition. Altogether 352 children (186 boys, 166 girls) aged 9‐11 years were included in the analyses. We measured V̇O2peak directly during a maximal cycle ergometer exercise test and lean body mass (LM) by bioelectrical impedance. We computed a sex‐ and age‐specific cardiometabolic risk score (CRS) by summing important cardiometabolic risk factors and defined increased cardiometabolic risk as >1 standard deviation above the mean of CRS. Receiver operating characteristics curves were used to detect V̇O2peak cut‐points for increased cardiometabolic risk. Boys with V̇O2peak <45.8 mL kg body mass (BM)−1 min−1 (95% confidence interval [CI] = 45.1 to 54.6, area under the curve [AUC] = 0.86, P < 0.001) and <63.2 mL kg LM−1 min−1 (95% CI =52.4 to 67.5, AUC = 0.65, P = 0.006) had an increased CRS. Girls with V̇O2peak <44.1 mL kg BM−1 min−1 (95% CI = 44.0 to 58.6, AUC = 0.67, P = 0.013) had an increased CRS. V̇O2peak scaled by BM−0.49 and LM−0.77 derived from log‐linear allometric modeling poorly predicted increased cardiometabolic risk in boys and girls. In conclusion, directly measured V˙O2peak <45.8 mL kg BM−1 min−1 among boys and <44.1 mL kg BM−1 min−1 among girls were cut‐points to identify those at increased cardiometabolic risk. Appropriately controlling for body size and composition reduced the ability of cardiorespiratory fitness to identify children at increased cardiometabolic risk.


| INTRODUCTION
An increased cardiometabolic risk in childhood has been associated with an elevated risk of metabolic syndrome and type 2 diabetes, increased arterial stiffness, endothelial dysfunction, and preclinical carotid atherosclerosis in adulthood. [1][2][3][4] Therefore, the early identification of children with increased cardiometabolic risk is important to prevent cardiometabolic diseases in adulthood. Decreased cardiorespiratory fitness (CRF), independent of the levels of physical activity, has been considered a strong determinant of increased cardiometabolic risk in children and adolescents. [5][6][7][8][9] However, only a few studies have investigated cut-points for CRF to identify children at increased cardiometabolic risk. [10][11][12][13][14][15][16][17] Previous studies have suggested that peak oxygen uptake (V · O 2peak ) lower than 44.0 mL kg body mass (BM) −1 min −1 in boys and lower than 39.5 mL kg BM −1 min −1 in girls are indicative of increased cardiometabolic risk. [10][11][12]17 Furthermore, the most recent FITNESSGRAM guidelines suggest that boys with V · O 2peak 37.3 to 41.2 mL kg BM −1 min −1 and girls with V · O 2peak 35.3 to 37.3 mL kg BM −1 min −1 depending on their age have an increased risk of metabolic syndrome. 13,18 However, these studies have measured workload or heart rate during a submaximal treadmill or cycle ergometer exercise test, a stage reached during a 20-m shuttle run test, or other types of exercise tests and converted these measures of performance into an estimate of V · O 2peak 10-13,15,17 instead of measuring V · O 2peak directly during a maximal exercise test continued until exhaustion. Estimated V · O 2peak obtained from these types of exercise tests is problematic in that it has, at best, 50% agreement with directly measured V · O 2peak . 19 Furthermore, V · O 2peak thresholds obtained from these studies are based on V · O 2peak divided by BM that is confounded by body fat content and may invalidate V · O 2peak as a measure of CRF in children with increased body mass and particularly adiposity. [10][11][12][13]17,[20][21][22] Allometric scaling of V · O 2peak by measures of body size and composition using log-linear regression can partly overcome the problems related to scaling of V · O 2peak by BM using the ratio standard method. 23 Nonetheless, allometrically scaled V · O 2peak for lean body mass (LM) is regarded superior to allometrically scaled V · O 2peak for BM in order to account for variance in body fat content in the expression of V · O 2peak among children and adolescents. [24][25][26][27][28][29] Thus, allometrically scaled V · O 2peak by LM has been recommended as the best approach in expressing V · O 2peak among children and adolescents. 29 However, there are few studies on the associations of CRF with cardiometabolic risk having appropriately controlled for body size and composition using the allometric methods or the ratio standard methods. 22 Using these approaches has attenuated the associations of CRF with cardiometabolic risk, 22 suggesting that CRF expressed in these manners has inferior predictive power compared with CRF scaled by BM.
The aim of this study was to provide cut-points for V · O 2peak measured directly during a maximal cycle ergometer exercise test among boys and girls to identify those who are at increased cardiometabolic risk. We used different methods for scaling V · O 2peak to control for body size and composition.

| Study design and study population
The Physical Activity and Nutrition in Children (PANIC) Study is an ongoing physical activity and dietary intervention study (ClinicalTrials.gov NCT01803776) in a population sample of primary school children living in the city of Kuopio, Finland. Altogether 736 children 6-9 years of age who had been registered for the first grade in one of the 16 public schools of the city of Kuopio were invited for baseline examinations in 2007-2009. Altogether 512 children (248 girls, 264 boys), who accounted for 70% of those invited, participated in the baseline examinations in 2007-2009. The participants did not differ in sex distribution, age, or body mass index standard deviation score (BMI-SDS) from all children who started the first grade in the city of Kuopio in 2007-2009 based on data from the standard school health examinations performed for all Finnish children before the first grade. 30 The present analyses are based on the 2-year follow-up data. We had complete data on variables needed in the analyses for 352 children (186 boys, 166 girls) 9-11 years of age. Of these children, 99.1% are Caucasians.
The PANIC Study protocol was approved by the Research Ethics Committee of the Hospital District of Northern Savo. A written informed consent was acquired from the parent or caregiver of each child and every child provided assent to participation.

| Assessment of cardiorespiratory fitness
We assessed CRF by a maximal exercise test using the Ergoselect 200 K ® electromagnetic cycle ergometer coupled with a pediatric saddle module (Ergoline, Bitz, Germany), as explained in detail earlier. 30 The children and their parents and caregivers were informed about the exercise test in the invitation letter. A research nurse and a research physician gave the children instructions on how to perform the exercise test. The children were familiarized with the exercise test protocol two years earlier during baseline examination. They were also allowed to practice cycling with the ergometer and using the pediatric mask 10 minutes before the exercise test. The exercise test protocol, supervised by the research physician and assisted by the research nurse, included a 2.5-minute anticipatory period with the child sitting on the ergometer; a 3-minute warm-up period with a workload of 5 watts; a 1-minute steady-state period with a workload of 20 watts; an exercise period with an increase in the workload of 1 watt per 6 seconds until exhaustion, and a 4-minute recovery period with a workload of 5 watts.
The children were asked to keep the cadence stable and within 70-80 revolutions per minute (rpm). The children were verbally encouraged to exercise until voluntary exhaustion. Heart rate was measured continuously throughout the exercise test using a 12-lead electrocardiogram (ECG) registered by the Cardiosoft ® V6.5 Diagnostic System (GE Healthcare Medical Systems, Freiburg, Germany). The exercise test was considered maximal if the peak heart rate was at least 185 beats per minute and the respiratory exchange ratio was at least 1.0. 31 However, the research physician also adjudged the exercise test maximal among 21 (6%) children with a peak heart rate of 179-184 beats per minute, because the cadence dropped below 65 rpm although the children still had the motivation to continue and the reason for terminating the test suggested a maximal effort had been provided. 30 The peak workload was defined as the workload at the end of the exercise test.
The respiratory gas analysis was performed during the exercise test from the beginning of the 2.5-minute anticipatory period before the exercise test to the end of the 4-minute recovery period after the exercise test using the Oxycon Pro ® respiratory gas analyzer (Jaeger, Hoechberg, Germany) and the Hans-Rudolph ® pediatric mask (Shawnee, Kansas, USA). V · O 2peak was measured using the breath-by-breath method and was averaged over consecutive 15-second periods. V · O 2peak and the respiratory exchange ratio were defined as the highest 15-seconds average values recorded during the last minute of the exercise test. 30

| Assessment of cardiometabolic risk factors
Cardiometabolic risk factors were assessed in the morning for two children, first of them arriving at 08:00, and second at 09:15. The research nurse measured body height, body weight, and waist circumference using standard protocols. 32 Body mass index (BMI) was calculated by dividing body weight with body height squared and BMI-SDS using national references. 33 The prevalence of overweight and obesity was defined using age-and sex-specific cut-points. 33 Total body fat mass, body fat percentage (BF%), and LM were measured twice; the children having fasted for 12 hours, voided the bladder; and standing in light underwear, using the InBody ® 720 bioelectrical impedance device (Biospace, Seoul, South Korea). We have found a good agreement between BF% and LM measured with bioelectrical impedance and those derived from dual-energy x-ray absorptiometry. 34 The research nurse measured blood pressure manually from the right arm by a calibrated Heine 130 Gamma G7 ® aneroid sphygmomanometer (Heine Optotechnik, Herrsching, Germany). The measurement protocol included a 5-minute rest and thereafter three measurements in a sitting position at 2-minute intervals. The mean of all three values was used as the systolic and diastolic blood pressure. The research nurse took venous blood samples using a standard protocol after a 12-hour fast and the children having seated for 10 minutes. The assessment of serum insulin and plasma glucose, total cholesterol, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, and triglycerides, have been explained in detail earlier. 35

| Assessment of puberty
The research physician assessed pubertal status using a 5stage scale described by Tanner. 36,37 Boys were defined having entered clinical puberty if their testicular volume assessed by an orchidometer was ≥4 mL (Tanner stage ≥2). 37 Girls were defined having entered clinical puberty if their breast development had started (Tanner stage ≥2). 36

| Calculation of cardiometabolic risk score
We calculated a continuous cardiometabolic risk score using population-specific and age-and sex-standardized Z-scores for waist circumference, insulin, glucose, HDL cholesterol, triglycerides, and the average of systolic and diastolic blood pressure by a formula: waist circumference +insulin + glucose − HDL cholesterol +triglycerides + the average of systolic and diastolic blood pressure. 38 We defined elevated cardiometabolic risk as ≥1 standard deviation above the mean of the cardiometabolic risk score in the present study population. The rationale for using this approach is the existing evidence on the ability of a continuous cardiometabolic risk score in children to predict cardiometabolic diseases in adulthood. 39 The clustering of risk factors may also provide a more sensitive and clinically more relevant evaluation of increased cardiometabolic risk than using individual risk factors, 40 and girls, we added the interaction term to the model. The interactions of sex with BM (P = 0.362) and LM (P = 0.932) were not statistically significant. These allometric models were able to remove the associations of V · O 2peak with BM (r = −0.072, P = 0.328) and LM (r = 0.022, P = 0.939), indicating their validity in scaling CRF.

| Statistical methods
Student's t-test, Mann-Whitney U test, and Chi-square test were used to compare basic characteristics between boys and girls. The associations of measures of CRF with the cardiometabolic risk score were studied using linear regression analyses. The associations were adjusted for age and study group and additionally for puberty. Receiver operating characteristics (ROC) curves were used to investigate V · O 2peak cut-points associated with increased cardiometabolic risk. We decided not to provide cut-points for allometrically scaled V · O 2peak because they would depend on the scaling exponent b that is specific to our study population. Moreover, a recent review has highlighted that no general scaling exponent b is available. 29 The area under the curve (AUC) was used as a measure of the effectiveness of the predictor variable to identify correctly children having a cardiometabolic risk score ≥+1 SD (sensitivity) and to identify correctly children having a cardiometabolic risk score <1 SD (specificity). An AUC of 1.0 represents the ability to identify perfectly children having a cardiometabolic risk score ≥+1 SD from other children, whereas an AUC of 0.5 indicates no greater predictive ability than chance. We also compared the cardiometabolic risk score in the categories of V · O 2peak scaled by BM and LM among boys and girls combined using analysis of covariance with Sidak correction adjusted for age and the study group.
In these analyses, we used sex-specific categories of V

| Characteristics of children
Boys had more LM, less fat mass, a lower BF%, a higher waist circumference, lower insulin, higher glucose, higher HDL cholesterol, and higher V · O 2peak scaled by BM and LM compared with girls (Table 1).

| Associations of peak oxygen uptake with cardiometabolic risk
Absolute V · O 2peak expressed in mL/min was directly associated with cardiometabolic risk in boys and girls adjusted for age and the study group (

| Cardiometabolic risk among children in categories of peak oxygen uptake
Cardiometabolic risk decreased in a dose-dependent manner with increasing categories of V · O 2peak expressed in mL kg BM −1 min −1 among children (P < 0.001 for linear trend) (Figure 1). Moreover, children in the highest category of V · O 2peak expressed in mL kg LM −1 min −1 also had a lower cardiometabolic risk than children in other categories of V · O 2peak (Figure 1). Our results are in agreement with previous findings that CRF measured in exercise test laboratories or using field tests and scaled by BM using the ratio standard method had a strong inverse association with cardiometabolic risk in children. [10][11][12] However, the inverse relationship between CRF scaled by BM and cardiometabolic risk is partly confounded by adiposity, because CRF divided by BM is a measure of both CRF and body fat content. We observed that using V and the view that V · O 2peak scaled allometrically or by fat-free mass would be a better measure to estimate the magnitude of the association between CRF and cardiometabolic risk. 22 Notwithstanding, the measures of CRF scaled by LM may also be influenced by adiposity because individuals with higher fat mass also have higher LM. 44 In the present crosssectional study, adjusting for puberty had no effect on the relationships between V · O 2peak and cardiometabolic risk in boys or girls. Therefore, longitudinal studies are needed to clarify the role of CRF in cardiometabolic health during growth and maturation. It is also important to note that CRF is strongly influenced by genetic factors 45 and some genetic variants have been reported to modify the relationship between CRF and cardiometabolic risk factors, such as adiposity, insulin resistance, and elevated blood pressure. 45 In our study, genetics may play a role in the inverse association between CRF controlled for body fat and cardiometabolic risk. 46 The cut-point for V · O 2peak of 45.8 mL kg BM −1 min −1 to identify boys aged 9-11 years at increased cardiometabolic risk in our study corresponds well with the previously reported cut-point for V · O 2peak of 43.6 mL kg BM −1 min −1 among boys aged 8-11 years. 10 However, the cut-point for V · O 2peak of 44.1 mL kg BM −1 min −1 to identify girls aged 9-11 years at increased cardiometabolic risk in our study was notably higher than the cut-point for V · O 2peak of 37.0 mL kg BM −1 min −1 observed among girls aged 9-10 years in a previous study. 15 The reason for a higher threshold for V · O 2peak among girls in our study than in the earlier study may be that Finnish girls aged 9-11 years are more fit than girls of the same age in other pediatric populations. However, Welk and co-workers 13 found that the cut-point for estimated V · O 2peak to identify children aged 10-11 years at increased cardiometabolic risk was 40.2 mL kg BM −1 min −1 in both sexes. The diversity in cutpoints may also be due to the different age ranges of children because most of the earlier studies have reported pooled data of various age groups. 10,11,14 There is some evidence that the cut-point for V · O 2peak decreases with increasing age in girls,

|
whereas it remains relatively stable in boys. 31 One explanation for this sex difference may be that body fat content increases more in girls than in boys during maturation that introduces more confounding by adiposity in the measurement of CRF in girls. 47 Another reason for the array of cutpoints for V · O 2peak scaled by BM may be that the prevalence of overweight has varied among the study populations. 13 Moreover, different assessments of CRF and cardiometabolic risk factors and different definitions of increased risk may explain the incongruence in CRF cut-points among these studies. [10][11][12][13]17 In consonance with earlier studies, 10,12,15 the prevalence of increased cardiometabolic risk defined by ≥+1 SD of the cardiometabolic risk score in our study was 25% among boys and 34% in girls. We found moderate sensitivity and specificity of the measures of CRF in predicting increased cardiometabolic risk which corresponds to those of previous studies. [10][11][12]15 Sensitivity for V · O 2peak scaled by BM was 75% in boys and 69% in girls. This sex difference may be due to a stronger inverse association between CRF and cardiometabolic risk in boys than in girls. 13 Nonetheless, the false-positive rate would be too high for screening children with increased cardiometabolic risk using V · O 2peak scaled by BM. There is a trade-off between false-positive and false-negative rates. Classification accuracy may lead to the problem of fictitious interpretation when applying cut-points with a high false-positive rate than those with a high false-negative rate. Improving CRF may decrease cardiometabolic risk; however, a large number of false-positive cases would result in failure to correctly identify children at increased risk, in contrast to recommending increased physical activity to improve CRF in false-negative cases. 10 It is important that children are not subject of the social stigma associated with being erroneously classified as being at increased cardiometabolic risk. In addition, it is better to err on the side of caution so that children who truly are at increased risk are not deprived of health care. 10 Loftin and co-workers suggested that V · O 2peak should be allometrically scaled for LM due to the involvement of skeletal muscle in locomotion. 29  scaled by BM in these children aged 9-11 years; however, we cannot extrapolate our findings to other age groups. Our study participants were Caucasian children, so the cut-points may not be generalized to children of different ethnic groups. A limitation of the study is its cross-sectional design that does not allow us to arrive at a conclusion regarding the causality of the association between CRF and cardiometabolic risk. Therefore, longitudinal studies are warranted in order to investigate whether a decrease in adiposity-independent measures of CRF is associated with an increase in cardiometabolic risk over time among children and adolescents. In addition, it would be important to provide evidence for the effects of growth and maturation on the cut-points for CRF using different methods to scale CRF for body size and composition.
In conclusion, we found that directly measured V · O 2peak less than 45.8 mL kg BM −1 min −1 in boys 9-11 years of age and less than 44.1 mL kg BM −1 min −1 in girls 9-11 years F I G U R E 1 Differences in cardiometabolic risk score according to sex-specific V · O 2peak distribution using analysis of covariance (ANCOVA) with Sidak correction, adjusted for age, and study group | 23 AGBAJE Et Al.
of age was associated with increased cardiometabolic risk with moderate sensitivity and specificity. The association of CRF scaled by BM with cardiometabolic risk was markedly weaker than that of CRF scaled by LM because scaling by LM reduced the dependence of the measure of CRF on adiposity. Appropriately controlling for body size and composition reduced the ability of CRF to identify boys and girls at increased cardiometabolic risk.

| PERSPECTIVES
Cardiometabolic risk tracks from childhood into adulthood and the early identification of individuals at increased risk is essential in developing public health actions targeted at preventing cardiometabolic diseases. Our results showed that CRF scaled by BM, which is partly confounded by adiposity, had a strong inverse association with cardiometabolic risk among children.
Appropriately controlling for body size and composition markedly attenuated the predictive ability of CRF. The strong inverse association between CRF scaled by BM and cardiometabolic risk suggests that CRF scaled by BM can be used in screening children at increased cardiometabolic risk. However, children may be erroneously classified as being at increased risk, which may subject them to social stigma. Hence, there should be cautious interpretation and utilization of CRF thresholds so that children who truly are at increased cardiometabolic risk are not deprived of appropriate intervention. A markedly weakened relationship between CRF and increased cardiometabolic risk when adiposity was appropriately controlled for raises the question of whether there is an etiological link between CRF and cardiometabolic health in children. Hence, longitudinal research is needed to establish whether decreased CRF, using appropriate scaling methods to control for body size and composition, increases cardiometabolic risk among children.