The contribution of physical fitness to individual and ethnic differences in risk markers for type 2 diabetes in children: The Child Heart and Health Study in England (CHASE)

Background The relationship between physical fitness and risk markers for type 2 diabetes (T2D) in children and the contribution to ethnic differences in these risk markers have been little studied. We examined associations between physical fitness and early risk markers for T2D and cardiovascular disease in 9‐ to 10‐year‐old UK children. Methods Cross‐sectional study of 1445 9‐ to 10‐year‐old UK children of South Asian, black African‐Caribbean and white European origin. A fasting blood sample was used for measurement of insulin, glucose (from which homeostasis model assessment [HOMA]‐insulin resistance [IR] was derived), glycated hemoglobin (HbA1c), urate, C‐reactive protein (CRP), and lipids. Measurements of blood pressure (BP) and fat mass index (FMI) were made; physical activity was measured by accelerometry. Estimated VO2 max was derived from a submaximal fitness step test. Associations were estimated using multilevel linear regression. Results Higher VO2 max was associated with lower FMI, insulin, HOMA‐IR, HbA1c, glucose, urate, CRP, triglycerides, LDL‐cholesterol, BP and higher HDL‐cholesterol. Associations were reduced by adjustment for FMI, but those for insulin, HOMA‐IR, glucose, urate, CRP, triglycerides and BP remained statistically significant. Higher levels of insulin and HOMA‐IR in South Asian children were partially explained by lower levels of VO2max compared to white Europeans, accounting for 11% of the difference. Conclusions Physical fitness is associated with risk markers for T2D and CVD in children, which persist after adjustment for adiposity. Higher levels of IR in South Asians are partially explained by lower physical fitness levels compared to white Europeans. Improving physical fitness may provide scope for reducing risks of T2D.


| INTRODUCTION
Higher levels of physical fitness in adults are associated with a lower risk of developing type 2 diabetes (T2D) 1,2 and lower rates of cardiovascular disease (CVD) mortality and all-cause mortality. 3,4 South Asian adults are at higher risk of developing T2D, stroke and coronary heart disease compared to white Europeans in the UK and other western countries. 5,6 Evidence suggests that ethnic differences in risk markers for T2D (particularly insulin resistance) are apparent in prepubertal children with higher levels of T2D risk markers in South Asian children compared to white Europeans. 7 Ethnic differences in physical fitness are also apparent in childhood; we have shown in a recent report lower levels of physical fitness in South Asians compared to white Europeans. 8 These differences in physical fitness could potentially help to explain ethnic differences in risk markers for T2D. We therefore examined associations between physical fitness and risk markers for T2D and CVD in a study of British school children and examined the contribution of differences in physical fitness to ethnic differences in risk markers for T2D between South Asians and white Europeans. It has been suggested that physical fitness and physical activity could have separate and independent effects on metabolic risk in childhood. 9 Given the complex relationship between physical activity and fitness, we also examine the independent associations between physical activity, physical fitness, and risk for T2D and CVD.  7,10,11 . In brief, the study was based on a random sample of 200 state primary schools, half were drawn from a sampling frame of schools with a high prevalence of UK South Asian children and half were drawn from a sampling frame with a high prevalence of UK black African-Caribbean children. This report is based on the final phase of the study carried out between January  5 ) was derived to be uncorrelated with height (r = −0.04) which has been shown previously to provide a more valid measure of body fatness in this multi-ethnic study population. 13 Systolic and diastolic blood pressure (BP) were measured twice in the right arm using an Omron HEM-907 (Omron Electronics, Milton Keynes, UK). Mean systolic and diastolic BP were adjusted for cuff size using a previously validated method. 14 Participants provided a blood sample after an overnight fast; children who reported having eaten breakfast were excluded from the analysis. Breakfast was provided after the sample was collected.
Serum was separated and frozen on dry ice immediately after collection for measurement of insulin. Urate was measured in serum using an enzymatic assay. 15 Samples were shipped to a central laboratory within 48 hours of collection. Details of insulin, glucose, glycated hemoglobin (HbA1c), C-reactive protein (CRP) and blood lipids assays have been previously reported. 7,10 Homeostasis model assessment (HOMA) equations 16 were used to derive insulin resistance from fasting insulin and glucose measurements.
An 8-minute submaximal step test was performed by participating children as previously described. 17,18 In brief, participants followed an audible prompt which instructed them to increase their step frequency progressively from 15 to 32.5 body lifts/min on a 150-mm high step while wearing a combined heart rate (HR) and movement sensor (Actiheart, CamNtech, Papworth, UK). The test was terminated if the participant was unable to sustain the prescribed step frequency. At the end of the test, 2 minutes of seated recovery was recorded. ECG and acceleration waveforms were recorded at 128 and 32 Hz sampling, respectively. VO 2 max was estimated using a similar method to that used in the Health Survey for England 2008. 19 In brief, predicted workload was regressed against instantaneous HR (expressed above resting HR) and 1-minute recovery HR was extracted using quadratic regression against recovery time. This was combined with resting HR and test duration to calculate the submaximal relationship between workload and HR, which was extrapolated to predicted maximal HR to predict maximal work capacity. This result was converted to VO 2 max by adding an estimate of resting metabolic rate and then dividing by the energetic value of oxygen. 19 Objective measures of physical activity were made using an accelerometer recorded at 5 second epochs (GT1M; ActiGraph LLC, Pensacola, Florida), worn on the participant's left hip during waking hours for 7 days; the device was removed for water-based activities.
Non-wear time was defined as a period of at least 20 consecutive minutes of zero counts, and was excluded from the analyses. Physical activity was summarized as total daily counts. Participants with at least 1 day of valid data (defined as at least 600 minutes of registered time) were included in the analysis.
Ethnic origin was defined using information from a parental questionnaire on the ethnicity of both parents where available (63%), or using the parentally defined ethnic origin of the child (36%). In a small number of children (1%) where this information was not available, ethnic origin was defined using information on parental and grandparental place of birth provided by the child, cross-checked with observer assessment of ethnic origin. Children were defined as white European, South Asian, black African-Caribbean and "other" ethnicity.

| Statistical methods
Statistical analyses were carried out using Stata/SE software (Stata/ SE 13 for Windows; StataCorp LP, College Station, Texas). All risk markers were inspected for normality; all variables except systolic and diastolic BP followed a log-normal distribution and were therefore log-transformed in analyses. A regression calibration method to allow for measurement error in physical activity counts 20 was used to allow for within-child variation by day of the week and variation in the number of days of recording (between 1 and 7), which provides an unbiased average for each child. The majority of participants (83%) wore the ActiGraph for at least 4 days. We examined the shape of associations between VO 2 max and risk markers by deciles of VO 2 max and did not find any evidence of departures from linearity. Hence, associations between estimated VO 2 max and risk markers were assessed using multilevel linear models adjusted for sex, age (in quartiles), ethnic group, month of measurement and height (fitted as fixed effects) and school fitted as a random effect to take into account clustering of children within schools. The effect of additional adjustment for FMI was also examined. We also examined whether associations between VO 2 max and risk markers were consistent in boys and girls, and ethnic groups. To elucidate whether ethnic differences in estimated risk markers were explained by differences in physical fitness, we examined the effect of adjustment for VO 2 max on ethnic differences in risk markers. In separate models we investigated whether associations between physical fitness and risk markers were consistent across tertiles of physical activity.

| RESULTS
The participation rate for children taking part in this phase of the study was 63%; of those who took part 1979 (89%) provided a fasting blood sample and had complete data. Estimated VO 2 max values were available for a total of 1445 participants (50% female with a mean age of 10.0 years [95% reference range 9.2, 10.7 years]) who had provided a fasting blood sample and had measurements of FMI.
Of these children, 1083 provided a valid estimate of physical activity.
The reasons for missing physical activity data were refusal to wear the device, failure to return the device or providing less than 1 day with 600 minutes of wear time of the device. Adjusted mean levels of VO 2 max and risk markers for T2D and CVD are shown in Table 1.
Boys had higher estimated levels of VO 2 max than girls; South Asians had lower levels and black African-Caribbeans had higher levels of estimated VO 2 max compared to white Europeans.
Associations between physical fitness (estimated VO 2 max ) and risk markers for T2D and CVD are shown in Table 2. Estimated VO 2 max was inversely associated with FMI, fasting insulin, HOMA-insulin resistance (IR), HbA1c, fasting glucose, urate, CRP, triglyceride, LDLcholesterol, and systolic and diastolic BP, and positively associated with HDL-cholesterol. Associations between estimated VO 2 max and insulin, HOMA-IR, glucose, urate, CRP and triglyceride were attenuated by adjustment for FMI between 32% and 63%. Further adjustment for fat free mass index (FFMI) did not further attenuate these associations (data available from authors). Associations with HbA1c, HDL-and LDL-cholesterol were completely explained by adjustment for FMI and those with systolic and diastolic BP were attenuated by 31% and 16%, respectively. After adjustment for FMI, a 1 IQR increase in estimated VO 2 max was associated with lower fasting insulin, HOMA-IR, fasting glucose, urate, CRP, and triglyceride of 7.7%, 7.7%, 0.7%, 2.2%, 20.8%, and 5.3%, respectively. Similarly, a 1 IQR increase in estimated VO 2 max was associated with decreases in systolic and diastolic BP of 1.7 and 2.7 mm Hg, respectively. These associations were broadly similar in boys and girls (Table S1) and in South Asians, black African-Caribbeans, white Europeans and "other" ethnic groups (Table S2) with the exception of LDL-cholesterol which was inversely associated with VO 2 max in all ethnic groups except black African-Caribbeans.
In a subset of 1083 children with objective measures of both physical fitness and physical activity, associations between risk markers for T2D/CVD and physical activity and VO 2 max are shown in Table S3. Associations were stronger between risk markers for T2D and estimated VO 2 max compared to those for overall physical activity. Following adjustment for FMI, however, associations between fasting insulin, HOMA-IR and physical activity were stronger than those for estimated VO 2 max . To explore whether the associations between physical fitness and risk markers were modified by physical activity level, associations between VO 2 max and risk markers were examined by tertiles of physical activity in Table S4. The inverse associations between estimated VO 2 max and fasting insulin, HOMA-IR and triglyceride were strongest among children in the lowest tertile of physical activity and weakest among children in the highest tertile; a graphical representation for fasting insulin is shown in Figure S1.
Ethnic differences in risk markers for T2D between South Asians and white Europeans and the effect of adjustment for physical fitness are shown in Table 3. South Asians had higher levels of fasting insulin, HOMA-IR, HbA1c, fasting glucose and triglyceride, and lower levels of HDL-cholesterol. Adjustment for VO 2 max reduced these ethnic differences in insulin, HOMA-IR, glucose and triglyceride by approximately 11%; differences in HbA1c and HDL-cholesterol were reduced by 4% and 13%, respectively. The borderline ethnic difference in CRP was reduced by 50% by adjustment for estimated VO 2 max . In the subset of 1083 participants with measurements of physical activity, Table S5 shows whether adjustment for physical activity counts in addition to VO 2 max would further explain the ethnic differences in risk markers for T2D. For fasting insulin, HOMA-IR, triglycerides and HDL-cholesterol adjustment for physical activity as well as VO 2 max , further reduced the difference between South Asians and white Europeans by 5% to 7%.

| Main findings
This study provides strong evidence that low levels of physical fitness are associated with risk markers for T2D and CVD in this multi-ethnic population of school children. These associations were attenuated but remained statistically significant after adjustment for measures of body fat, suggesting that associations were at least partly reduced by the effect of adjustment for adiposity, though additional adjustment for fat free mass had no material effect. Moreover, ethnic differences in risk markers for T2D, particularly increased levels of insulin resistance in South Asians compared to white European children, were partially explained by physical fitness (approximately 11%). Stratified analysis suggested that the association between physical fitness and risk markers differed by level of physical activity, with stronger associations being observed at lower levels of physical activity.

| Relation to previous studies
The finding in the present study that objective measures of physical fitness are inversely associated with risk markers for T2D and CVD in childhood is consistent with a small number of studies, which have used similar objective methodologies. 21 In particular, findings from the European Youth Heart Study (EYHS) showed that maximal ergometer assessments of cardiorespiratory fitness were inversely related to metabolic risk markers at both 9 and 15 years of age. 21,22 These associations have since been demonstrated in prospective studies, providing further evidence of a causal relationship. 23 We have previously reported that both physical activity and adiposity are associated with risk markers for T2D and CVD in this study population, 20,24 and that these associations are partially, but not wholly, mutually independent.
The present study confirms the association between physical fitness and cardiometabolic risk markers, but is novel in suggesting that the association observed might be altered at different levels of physical activity, with stronger associations between physical fitness and metabolic risk at lower levels of physical activity. This is consistent with a report based on 589 Danish children in the EYHS but this was not confirmed in the full EYHS study population, which suggested little or no modifying influence of physical activity. 21 Intuitively it would seem plausible that physical fitness-metabolic risk associations might be modified by levels of physical activity, given the established link between activity and fitness, where previous physical activity interventions have been shown to improve physical fitness. [25][26][27] However, this apparent interaction could also reflect differences in the precision of measurement of physical activity and physical fitness, and the possibility of non-linear associations between them. A sensitivity analysis examining these associations in children with at least 6 days of physical activity data (54% of the sample) showed that stronger associations between risk markers and physical fitness persisted at lower levels of physical activity (data available from authors).
This study suggests that adiposity is an important potential mediator of the associations between physical fitness and risk markers for Previous studies examining the consistency of early physical fitness and T2D/CVD risk marker associations have suggested that these are similar in boys and girls, and show little regional variations among populations of European ancestry. 31 The present study confirms that associations were similar by sex, and extends knowledge by showing consistent associations in children of more diverse ethnic

| Strengths and limitations
There are a number of strengths and weaknesses worthy of consideration. A strength of the study is the objective measure of physical fitness using a submaximal fitness test to estimate VO 2 max, 8 which has previously been used in a nationally representative large-scale study including this age group. 19 The use of a submaximal fitness test means that VO 2 max must be extrapolated, however, this test was chosen for its suitability in large-scale studies 19 and to encourage participation among a target population with low levels of physical activity. 11 This was accompanied by objective assessment of physical activity, using waist worn accelerometry over a 7-day period, which provides a reliable assessment of habitual levels of physical activity. 20 While both these measures provide accurate assessment, their comparative accuracy should be considered. The former provides a single measure of a long-term attribute (indicative of fitness levels over a month or longer), whereas accelerometry provides a measure of a short-term behavior, which may not fully capture habitual levels of physical activity behavior. Hence, while the fitness-metabolic or cardiovascular risk marker associations might be considered robust, the predictive value of physical activity may have been underestimated.
Specifically fitness is partly a function of long-term sustained physical activity, so it may be representing effects of past activity and/or more intense activity.
Another key strength is the measurement of important early risk markers for T2D (including HOMA insulin resistance, HbA1c, triglyceride, and urate) and CVD (including systolic and diastolic BP and LDL-cholesterol). In addition, adiposity was measured using bioelectrical impedance with ethnic-and gender-specific equations for the prediction of fat mass in this age group; this method has been shown to provide valid assessments of adiposity in a multi-ethnic population. 12 While the response rate for the present study was modest, the sample contained balanced representation of South Asians, black African-Caribbeans and white Europeans, and of boys and girls. Furthermore, risk markers were similar in those who completed the fitness test and those who did not (data not presented).
Ethnic differences in metabolic risk markers in the CHASE study have previously been reported in the whole study population 32 ; patterns of ethnic differences were broadly similar in this subset of data.

| Implications
The results presented in the present study show that ethnic differences in physical fitness account for up to 11% of the difference in insulin resistance and glycaemia markers between South Asians (who had higher levels of these markers) and white Europeans. This has potential implications for the early prevention of T2D in South Asians. Clinical trials are needed to determine whether efforts to improve physical fitness (potentially through increases in vigorous intensity physical activity) could help to improve metabolic health in children, perhaps especially among South Asian children/adolescents in whom risks of insulin resistance and T2D are particularly high. 5,6