Original Article
Obesity
EPIDEMIOLOGY/GENETICS
The Contribution of Childhood Cardiorespiratory Fitness and Adiposity to Inflammation in Young Adults Cong Sun1,2, Costan G. Magnussen3,4, Anne-Louise Ponsonby1,2, Michael D. Schmidt5, John B. Carlin2,6, Quan Huynh3, Alison J. Venn3, and Terence Dwyer1
Objective: Cardiorespiratory fitness and adiposity may influence cardiovascular risk through their effects on inflammation. The long-term effects of these modifiable factors on adult inflammation remain uncertain. The associations of childhood and adulthood cardiorespiratory fitness and adiposity with adult inflammation [C-reactive protein (CRP), fibrinogen] were examined. Methods: 1,976 children examined in 1985 and re-examined as young adults in 2004-2006 were included. Cardiorespiratory fitness and adiposity were assessed at both waves. CRP and fibrinogen were measured at follow-up. Results: Higher childhood fitness was associated with lower adult inflammation in both sexes. After adjusting for childhood adiposity, the association with CRP attenuated in males, but remained in females (average reduction of CRP 18.1% (95% CI 11.3-24.4%) per 1-SD increase in childhood fitness). Higher adult fitness, adjusting for childhood fitness (an increase in fitness from childhood to adulthood), was associated with lower adult CRP in females and lower fibrinogen in males. Higher childhood and adulthood adiposity (an increase in adiposity from childhood to adulthood) were associated with higher adult inflammation in both sexes. Conclusions: Prevention programs to increase fitness and reduce adiposity in childhood, and maintain a favorable fitness and weight into adulthood, may lead to reduction in adult systemic inflammation. Obesity (2014) 22, 2598–2605. doi:10.1002/oby.20871
Introduction Inflammation biomarkers such as C-reactive protein (CRP) and fibrinogen are emerging subclinical biomarkers with elevated levels predicting development of cardiovascular disease (CVD), independent of traditional risk factors (e.g., lipids and body mass index (BMI)) (1). Furthermore, these biomarkers may improve CVD risk prediction beyond traditional risk factors (1,2). Studies have shown that obesity, insulin resistance, and type 2 diabetes are closely associated with chronic inflammation, which is the mechanistic core of atherosclerotic CVD (3). Apart from obesity, inflammatory processes could be affected by cardiorespiratory fitness, another modifiable risk factor (4,5). Many, but not all (6,7), studies suggest that cardiorespiratory fitness may
have a protective effect on CVD risk, irrespective of adiposity levels (8,9). Most studies examining the relative contribution of fitness and fatness on inflammation in adult populations, however, have been cross sectional. Studies also imply that the link between obesity and chronic inflammation may already exist in children (10). However, studies in the pediatric population have shown an inconsistent association between cardiorespiratory fitness and inflammation independent of adiposity (11,12). In light of the escalating prevalence of childhood obesity, which often coexists with low cardiorespiratory fitness, the need for a better understanding of the long-term effects of childhood cardiorespiratory fitness and adiposity on adult health is increasingly urgent and will help public
1 Environmental and Genetic Epidemiology Research Group, Murdoch Children’s Research Institute (MCRI), Royal Children’s Hospital, Melbourne, Victoria, Australia. Correspondence: Cong Sun ([email protected] or [email protected]) 2 Department of Paediatrics, University of Melbourne, Melbourne, Victoria, Australia 3 Menzies Research Institute Tasmania, University of Tasmania, Hobart, Tasmania, Australia 4 Research Centre of Applied and Preventive Cardiovascular Medicine, University of Turku, Turku, Finland 5 Department of Kinesiology, University of Georgia, Athens, Georgia, USA 6 Clinical Epidemiology and Biostatistics Unit, MCRI, Royal Children’s Hospital, Melbourne, Victoria, Australia.
Funding agencies: C.S. and C.G.M were supported by the Australian National Health and Medical Research Council (NHMRC) Early Career Public Health Fellowships (grant 1013535 and 1037559). The Childhood Determinants of Adult Health Study was supported by the NHMRC project grant (211316), the Australian National Heart Foundation, the Tasmanian Community Fund, Veolia Environmental Services, the Sanitarium Health Food Company, ASICS Oceania, and Target Australia. No funding organization had any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation of the manuscript. Disclosure: The authors declared no conflict of interest. Author contributions: T.D. and A.J.V. obtained study funding and oversaw data collection, C.S. analyzed the data and wrote the manuscript, J.B.C. advised on aspects of the analysis, all authors contributed to interpretation of the data, discussion, and critical revision of the manuscript and had final approval of the submitted and published versions. Received: 1 April 2014; Accepted: 23 July 2014; Published online 19 November 2014. doi:10.1002/oby.20871
2598
Obesity | VOLUME 22 | NUMBER 12 | DECEMBER 2014
www.obesityjournal.org
Original Article
Obesity
EPIDEMIOLOGY/GENETICS
tive turbidimetric immunoassay kit (Olympus System CRP Latex reagent, Olympus Life and Material Science Europa GmbH, Ireland) by MedVet (Institute of Medical and Veterinary Science, Adelaide, South Australia). Plasma concentration of fibrinogen was determined by the Clauss clotting method using the STA automated coagulation analyzer (STA-Fibrinogen reagent, Diagnostica Stago, Manufactured in Paris, France distributed from Parsippany, NJ). Inter-assay coefficient of variation ranged from 3.6% to 23.9% for CRP, and 2.4% to 6.4% for fibrinogen. Duplicate aliquots from a single blood draw were collected from the first participant scheduled for each clinic day (N 5 108) to examine measurement errors associated with collection, processing, and analysis of blood samples. The coefficients of variation were 17.6% for CRP, and 5.8% for fibrinogen. All duplicate assays were conducted by technicians blinded to the first-run results.
Figure 1 Flow chart of study population in the Childhood Determinants of Adult Health (CDAH) Study.
health practitioners develop well-timed optimal interventions for primary prevention. We are not aware of any prospective cohort study that has examined the effects of childhood cardiorespiratory fitness and adiposity on adult inflammation in a general population. To address this gap, we examined these associations with two adult acute-phase inflammatory biomarkers, CRP and fibrinogen, in the Childhood Determinants of Adult Health (CDAH) Study (13). We further investigated whether adult cardiorespiratory fitness (and adiposity) or change in cardiorespiratory fitness (and adiposity) from childhood to adulthood is associated with adult inflammation.
Methods Study population The CDAH Study is a prospective cohort study that from May 2004 to May 2006 followed a total of 5,170 young adults who participated as children (baseline sample 8,498, 7-15 years) in the Australian Schools Health and Fitness Survey in 1985. The study design and sample recruitment have been described in detail previously (13). Of those enrolled in the CDAH follow-up study, 2,410 attended study clinics held throughout Australia and provided physical measurements and blood samples and an additional 170 had blood collected remotely. A flow chart that shows participants at both waves of the study and those included and excluded in our analysis is illustrated in the Figure 1. We have excluded participants who reported infection in the 2 weeks before data collection and women who reported being pregnant. At baseline, consent from both parent and child was required; at followup, all participants gave written informed consent. The baseline study was approved by the State Directors General of Education and the follow-up survey was approved by the Southern Tasmania Health and Medical Human Research Ethics Committee.
Exposure variables. Although maximal ventilatory oxygen uptake (VO2 max) is a universally accepted measure of cardiorespiratory fitness, it is unsuitable for field work in large population-based studies. Accordingly, a submaximal fitness test, time (minutes) to complete a 1.6-km run, was used as the cardiorespiratory fitness measure in childhood. 1.6-km run time is a widely used measure that is strongly correlated with VO2 max (20.85 to 20.73; negative correlations indicate the longer the run time the lower the fitness) (14). In both childhood and adulthood, cardiorespiratory fitness was also assessed using physical working capacity at a heart rate of 170 beats/min (PWC170) on a bicycle ergometer (Monark Exercise AB, Vansbro, Sweden) pedalled at a rate of 60 revolutions per minute (15). However, PWC170 was measured only in a subsample of children aged 9, 12, and 15 years, whereas 1.6-km run time was completed by all children and thus was used as the primary measure of childhood cardiorespiratory fitness in this analysis. The correlation between unstandardized 1.6 km rum time and PWC170 was 20.36 for males (P < 0.001) and 20.26 for females (P < 0.001) (standardized r 5 20.13, P 5 0.03 for males; r 5 20.17, P 5 0.004 for females). 1.6-km run time was transformed into an age- and sexspecific z score based on the whole baseline sample and then reverse scored as the childhood fitness variable. In adulthood, triceps, biceps, subscapular, and suprailiac skinfold thickness were measured. Skinfolds at four sites were only assessed for children aged 9, 12, and 15 years (16). We then calculated percent body fat (17) from body density as estimated by the log of the sum of four skinfold thickness measures using age-specific regression equations (18,19). Height, weight, and waist circumference were measured in both waves. BMI was calculated as kg/m2. Ageand sex-specific z scores for childhood BMI and waist circumference were generated using smoothed reference values based on data available from the 1990 British Growth Reference [20,21). Anthropometric measures including BMI, waist circumference, and percent body fat were surrogate measures of adiposity. All three measures showed essentially similar effects, for brevity we display the results in full for BMI only.
Covariates.
Study measures Outcome measure. At follow-up, serum CRP was determined using an automated analyzer (Olympus AU5400) and a highly sensi-
www.obesityjournal.org
At baseline, 24-h food consumption was recorded for participants aged 10-15 years. At follow-up: Information about food habits and food intake frequency over the past 12 months was collected. The frequency of nine alcoholic beverages consumption and their average alcohol concentration was used to estimate weekly
Obesity | VOLUME 22 | NUMBER 12 | DECEMBER 2014
2599
Obesity
Fitness, Fatness, and Inflammation Sun et al.
TABLE 1 Characteristics of study participants in childhood
and young adulthood by sex Males (N 5 1,016)
Females (N 5 960)
Childhood (1985) Age, y 11.2 6 2.5 10.9 6 2.6 1.6-km run time z score 20.3 (20.7, 0.3) 20.2 (20.8, 0.4) Body mass index z score 0.3 6 0.9 0.1 6 0.9 Waist circumference z score 0.6 6 0.9 0.5 6 1.1 Cigarette smoker ever, % 18.3 (138) 17.5 (112) Socioeconomic status Highest quartile, % 30.0 (221) 28.4 (179) 3rd quartile, % 25.8 (190) 27.3 (172) 2nd quartile, % 33.7 (248) 38.4 (242) Lowest quartile, % 10.6 (78) 6.0 (38) Young adulthood (2004–2006) Age, y 31.6 6 2.6 31.3 6 2.6 Body mass index, kg/m2 26.5 6 4.1 24.9 6 5.2 Waist circumference, cm 89.5 6 10.3 77.9 6 11.2 Hormonal contraceptive use, % 40.3 (377) Cigarette smoking status Never, % 64.9 (629) 64.6 (591) Former, % 16.2 (157) 20.4 (187) Current, % 19.0 (184) 15.0 (137) Alcohol consumption (g/week) 59.3 30.0 (23.0, 111.0) (12.5, 69.3) Education Higher degree, % 37.8 (358) 48.1 (450) Undergraduate diploma, % 36.0 (341) 25.2 (236) School only, % 26.3 (249) 26.7 (250) Fasting insulin, mU/l 6.2 (4.3, 9.2) 5.8 (4.4, 8.3) Fasting glucose, mmol/l 5.2 (4.9, 5.4) 4.8 (4.6, 5.1) HOMA insulin sensitivity index 1.4 (1.0, 2.1) 1.3 (0.9, 1.8) C-reactive protein, mg/l 1.1 (0.5, 2.4) 1.6 (0.6, 3.9)a Fibrinogen, g/l 2.9 6 0.6 3.2 6 0.7a Data show crude mean 6 standard deviation, median (interquartile range), or proportions (n). a 0.9 (0.4, 2.3) and 3.2 6 0.7 among females with no hormonal contraceptive use for C-reactive protein and fibrinogen, respectively.
consumption of number of standard drinks (10 g of alcohol) (22). Scores from three questions (What type of milk do you use? What spread do you usually use on bread? How often is the meat you eat trimmed of fat?) were combined to form a surrogate fat intake variable. Fasting glucose was measured by the Olympus AU5400 automated analyzer. Fasting insulin was measured by immunoassays with inter-assay standardization (16). We estimated insulin sensitivity using the homeostatic model assessment (HOMA) index (fasting glucose 3 fasting insulin/22.5) (23). We asked female participants whether they were currently using hormonal contraceptives. At both waves, information on cigarette smoking was derived from a self-administered questionnaire. Childhood and adulthood arealevel socioeconomic status (SES) of each participant was derived
2600
Obesity | VOLUME 22 | NUMBER 12 | DECEMBER 2014
from their residential postcode or census collection district using the Australian Bureau of Statistics Index of Relative Socioeconomic Disadvantage based on either the 1986 (24) or 2006 population census (25). Percent body fat was used to calculate lean body mass by subtracting fat mass from total body mass.
Statistical analysis Categorical variables were described using percentages and continuous variables using means and standard deviations or medians and inter-quartile ranges if skewed in distribution. We conducted analyses stratified by males and females due to the effect of hormonal contraceptive use on CRP and fibrinogen among females (26). We assessed the evidence for sex differences in the associations between fitness, adiposity, and inflammation by including fitness/adiposity and sex cross-product (interaction) terms in the multivariable models. CRP was natural log transformed due to its skewed distribution. Using linear regression models, we examined associations of childhood fitness and adiposity with adult CRP and fibrinogen. For CRP, logscale regression coefficients were interpreted as percent change in outcome per 1-SD change in exposure variable (% change 5 1 2 exp(b) for negative b or exp(b) 2 1 for positive b). We inspected residuals graphically after fitting linear regression models to check for nonlinear associations between continuous exposures and outcomes (confirming that linear models were adequate: results not shown). A range of potential confounders were considered in the model building process. We used relatedness of potential confounders (age, childhood and adulthood SES, smoking, fat intake and alcohol consumption, education, and hormonal contraceptive use for females) to exposures and outcomes with a 10% change-in-estimate criterion for inclusion in models. We adjusted for lean body mass in linear regression models when PWC170 was used as the fitness variable given that the absolute workload achieved is a function of muscle mass (27,28). For each of the two exposures, childhood fitness and adiposity, we initially adjusted for identified confounders in Model 1. In Model 2, we included both exposure variables to investigate whether the effect of each was independent of the other, adjusted for confounders. We also adjusted for adulthood insulin sensitivity in the models to test its potential mediation effect. We tested for potential interaction effects in the relationships by including fitness tertile and adiposity (BMI) cross-product term in Model 2 in both sexes. We noted that our sample included both prepubertal and pubertal children. Although we did not measure pubertal status, we tested for potential interaction effects in the relationships by including categorical age variable (7-9 years vs. 10-15 years groups for girls; 7-11 years vs. 12-15 years groups for boys; roughly corresponding to pubertal status) and exposure cross-product term in the models. In sensitivity analyses, we repeated all analyses among female participants with no hormonal contraceptive use. Similar analyses were performed among nonsmokers to assess a potential residual confounding effect of adult smoking on the associations. We also conducted sensitivity analyses using population-weighted models to assess whether the results were influenced by differences in demographic and adiposity characteristics between the 1,976 participants analyzed and those excluded. Weights were used to adjust the analyzed sample to be representative of the reference group (all
www.obesityjournal.org
Original Article
Obesity
EPIDEMIOLOGY/GENETICS
TABLE 2 Associations of childhood cardiorespiratory fitness and adiposity with adult C-reactive protein (CRP), from multivariable linear regression analysis of childhood exposures on lnCRP
Childhood exposures Cardiorespiratory fitness Body mass index z score Cardiorespiratory fitness Body mass index z score Cardiorespiratory fitness Body mass index z score Cardiorespiratory fitness Body mass index z score Cardiorespiratory fitness Body mass index z score
Crude model b (95% CI)
P-value
Model 1 adjusted b (95% CI)
P-value
Males (N 5 1,016) 0.005 20.11 (20.19, 20.03) 0.005
Obesity
EPIDEMIOLOGY/GENETICS
The Contribution of Childhood Cardiorespiratory Fitness and Adiposity to Inflammation in Young Adults Cong Sun1,2, Costan G. Magnussen3,4, Anne-Louise Ponsonby1,2, Michael D. Schmidt5, John B. Carlin2,6, Quan Huynh3, Alison J. Venn3, and Terence Dwyer1
Objective: Cardiorespiratory fitness and adiposity may influence cardiovascular risk through their effects on inflammation. The long-term effects of these modifiable factors on adult inflammation remain uncertain. The associations of childhood and adulthood cardiorespiratory fitness and adiposity with adult inflammation [C-reactive protein (CRP), fibrinogen] were examined. Methods: 1,976 children examined in 1985 and re-examined as young adults in 2004-2006 were included. Cardiorespiratory fitness and adiposity were assessed at both waves. CRP and fibrinogen were measured at follow-up. Results: Higher childhood fitness was associated with lower adult inflammation in both sexes. After adjusting for childhood adiposity, the association with CRP attenuated in males, but remained in females (average reduction of CRP 18.1% (95% CI 11.3-24.4%) per 1-SD increase in childhood fitness). Higher adult fitness, adjusting for childhood fitness (an increase in fitness from childhood to adulthood), was associated with lower adult CRP in females and lower fibrinogen in males. Higher childhood and adulthood adiposity (an increase in adiposity from childhood to adulthood) were associated with higher adult inflammation in both sexes. Conclusions: Prevention programs to increase fitness and reduce adiposity in childhood, and maintain a favorable fitness and weight into adulthood, may lead to reduction in adult systemic inflammation. Obesity (2014) 22, 2598–2605. doi:10.1002/oby.20871
Introduction Inflammation biomarkers such as C-reactive protein (CRP) and fibrinogen are emerging subclinical biomarkers with elevated levels predicting development of cardiovascular disease (CVD), independent of traditional risk factors (e.g., lipids and body mass index (BMI)) (1). Furthermore, these biomarkers may improve CVD risk prediction beyond traditional risk factors (1,2). Studies have shown that obesity, insulin resistance, and type 2 diabetes are closely associated with chronic inflammation, which is the mechanistic core of atherosclerotic CVD (3). Apart from obesity, inflammatory processes could be affected by cardiorespiratory fitness, another modifiable risk factor (4,5). Many, but not all (6,7), studies suggest that cardiorespiratory fitness may
have a protective effect on CVD risk, irrespective of adiposity levels (8,9). Most studies examining the relative contribution of fitness and fatness on inflammation in adult populations, however, have been cross sectional. Studies also imply that the link between obesity and chronic inflammation may already exist in children (10). However, studies in the pediatric population have shown an inconsistent association between cardiorespiratory fitness and inflammation independent of adiposity (11,12). In light of the escalating prevalence of childhood obesity, which often coexists with low cardiorespiratory fitness, the need for a better understanding of the long-term effects of childhood cardiorespiratory fitness and adiposity on adult health is increasingly urgent and will help public
1 Environmental and Genetic Epidemiology Research Group, Murdoch Children’s Research Institute (MCRI), Royal Children’s Hospital, Melbourne, Victoria, Australia. Correspondence: Cong Sun ([email protected] or [email protected]) 2 Department of Paediatrics, University of Melbourne, Melbourne, Victoria, Australia 3 Menzies Research Institute Tasmania, University of Tasmania, Hobart, Tasmania, Australia 4 Research Centre of Applied and Preventive Cardiovascular Medicine, University of Turku, Turku, Finland 5 Department of Kinesiology, University of Georgia, Athens, Georgia, USA 6 Clinical Epidemiology and Biostatistics Unit, MCRI, Royal Children’s Hospital, Melbourne, Victoria, Australia.
Funding agencies: C.S. and C.G.M were supported by the Australian National Health and Medical Research Council (NHMRC) Early Career Public Health Fellowships (grant 1013535 and 1037559). The Childhood Determinants of Adult Health Study was supported by the NHMRC project grant (211316), the Australian National Heart Foundation, the Tasmanian Community Fund, Veolia Environmental Services, the Sanitarium Health Food Company, ASICS Oceania, and Target Australia. No funding organization had any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation of the manuscript. Disclosure: The authors declared no conflict of interest. Author contributions: T.D. and A.J.V. obtained study funding and oversaw data collection, C.S. analyzed the data and wrote the manuscript, J.B.C. advised on aspects of the analysis, all authors contributed to interpretation of the data, discussion, and critical revision of the manuscript and had final approval of the submitted and published versions. Received: 1 April 2014; Accepted: 23 July 2014; Published online 19 November 2014. doi:10.1002/oby.20871
2598
Obesity | VOLUME 22 | NUMBER 12 | DECEMBER 2014
www.obesityjournal.org
Original Article
Obesity
EPIDEMIOLOGY/GENETICS
tive turbidimetric immunoassay kit (Olympus System CRP Latex reagent, Olympus Life and Material Science Europa GmbH, Ireland) by MedVet (Institute of Medical and Veterinary Science, Adelaide, South Australia). Plasma concentration of fibrinogen was determined by the Clauss clotting method using the STA automated coagulation analyzer (STA-Fibrinogen reagent, Diagnostica Stago, Manufactured in Paris, France distributed from Parsippany, NJ). Inter-assay coefficient of variation ranged from 3.6% to 23.9% for CRP, and 2.4% to 6.4% for fibrinogen. Duplicate aliquots from a single blood draw were collected from the first participant scheduled for each clinic day (N 5 108) to examine measurement errors associated with collection, processing, and analysis of blood samples. The coefficients of variation were 17.6% for CRP, and 5.8% for fibrinogen. All duplicate assays were conducted by technicians blinded to the first-run results.
Figure 1 Flow chart of study population in the Childhood Determinants of Adult Health (CDAH) Study.
health practitioners develop well-timed optimal interventions for primary prevention. We are not aware of any prospective cohort study that has examined the effects of childhood cardiorespiratory fitness and adiposity on adult inflammation in a general population. To address this gap, we examined these associations with two adult acute-phase inflammatory biomarkers, CRP and fibrinogen, in the Childhood Determinants of Adult Health (CDAH) Study (13). We further investigated whether adult cardiorespiratory fitness (and adiposity) or change in cardiorespiratory fitness (and adiposity) from childhood to adulthood is associated with adult inflammation.
Methods Study population The CDAH Study is a prospective cohort study that from May 2004 to May 2006 followed a total of 5,170 young adults who participated as children (baseline sample 8,498, 7-15 years) in the Australian Schools Health and Fitness Survey in 1985. The study design and sample recruitment have been described in detail previously (13). Of those enrolled in the CDAH follow-up study, 2,410 attended study clinics held throughout Australia and provided physical measurements and blood samples and an additional 170 had blood collected remotely. A flow chart that shows participants at both waves of the study and those included and excluded in our analysis is illustrated in the Figure 1. We have excluded participants who reported infection in the 2 weeks before data collection and women who reported being pregnant. At baseline, consent from both parent and child was required; at followup, all participants gave written informed consent. The baseline study was approved by the State Directors General of Education and the follow-up survey was approved by the Southern Tasmania Health and Medical Human Research Ethics Committee.
Exposure variables. Although maximal ventilatory oxygen uptake (VO2 max) is a universally accepted measure of cardiorespiratory fitness, it is unsuitable for field work in large population-based studies. Accordingly, a submaximal fitness test, time (minutes) to complete a 1.6-km run, was used as the cardiorespiratory fitness measure in childhood. 1.6-km run time is a widely used measure that is strongly correlated with VO2 max (20.85 to 20.73; negative correlations indicate the longer the run time the lower the fitness) (14). In both childhood and adulthood, cardiorespiratory fitness was also assessed using physical working capacity at a heart rate of 170 beats/min (PWC170) on a bicycle ergometer (Monark Exercise AB, Vansbro, Sweden) pedalled at a rate of 60 revolutions per minute (15). However, PWC170 was measured only in a subsample of children aged 9, 12, and 15 years, whereas 1.6-km run time was completed by all children and thus was used as the primary measure of childhood cardiorespiratory fitness in this analysis. The correlation between unstandardized 1.6 km rum time and PWC170 was 20.36 for males (P < 0.001) and 20.26 for females (P < 0.001) (standardized r 5 20.13, P 5 0.03 for males; r 5 20.17, P 5 0.004 for females). 1.6-km run time was transformed into an age- and sexspecific z score based on the whole baseline sample and then reverse scored as the childhood fitness variable. In adulthood, triceps, biceps, subscapular, and suprailiac skinfold thickness were measured. Skinfolds at four sites were only assessed for children aged 9, 12, and 15 years (16). We then calculated percent body fat (17) from body density as estimated by the log of the sum of four skinfold thickness measures using age-specific regression equations (18,19). Height, weight, and waist circumference were measured in both waves. BMI was calculated as kg/m2. Ageand sex-specific z scores for childhood BMI and waist circumference were generated using smoothed reference values based on data available from the 1990 British Growth Reference [20,21). Anthropometric measures including BMI, waist circumference, and percent body fat were surrogate measures of adiposity. All three measures showed essentially similar effects, for brevity we display the results in full for BMI only.
Covariates.
Study measures Outcome measure. At follow-up, serum CRP was determined using an automated analyzer (Olympus AU5400) and a highly sensi-
www.obesityjournal.org
At baseline, 24-h food consumption was recorded for participants aged 10-15 years. At follow-up: Information about food habits and food intake frequency over the past 12 months was collected. The frequency of nine alcoholic beverages consumption and their average alcohol concentration was used to estimate weekly
Obesity | VOLUME 22 | NUMBER 12 | DECEMBER 2014
2599
Obesity
Fitness, Fatness, and Inflammation Sun et al.
TABLE 1 Characteristics of study participants in childhood
and young adulthood by sex Males (N 5 1,016)
Females (N 5 960)
Childhood (1985) Age, y 11.2 6 2.5 10.9 6 2.6 1.6-km run time z score 20.3 (20.7, 0.3) 20.2 (20.8, 0.4) Body mass index z score 0.3 6 0.9 0.1 6 0.9 Waist circumference z score 0.6 6 0.9 0.5 6 1.1 Cigarette smoker ever, % 18.3 (138) 17.5 (112) Socioeconomic status Highest quartile, % 30.0 (221) 28.4 (179) 3rd quartile, % 25.8 (190) 27.3 (172) 2nd quartile, % 33.7 (248) 38.4 (242) Lowest quartile, % 10.6 (78) 6.0 (38) Young adulthood (2004–2006) Age, y 31.6 6 2.6 31.3 6 2.6 Body mass index, kg/m2 26.5 6 4.1 24.9 6 5.2 Waist circumference, cm 89.5 6 10.3 77.9 6 11.2 Hormonal contraceptive use, % 40.3 (377) Cigarette smoking status Never, % 64.9 (629) 64.6 (591) Former, % 16.2 (157) 20.4 (187) Current, % 19.0 (184) 15.0 (137) Alcohol consumption (g/week) 59.3 30.0 (23.0, 111.0) (12.5, 69.3) Education Higher degree, % 37.8 (358) 48.1 (450) Undergraduate diploma, % 36.0 (341) 25.2 (236) School only, % 26.3 (249) 26.7 (250) Fasting insulin, mU/l 6.2 (4.3, 9.2) 5.8 (4.4, 8.3) Fasting glucose, mmol/l 5.2 (4.9, 5.4) 4.8 (4.6, 5.1) HOMA insulin sensitivity index 1.4 (1.0, 2.1) 1.3 (0.9, 1.8) C-reactive protein, mg/l 1.1 (0.5, 2.4) 1.6 (0.6, 3.9)a Fibrinogen, g/l 2.9 6 0.6 3.2 6 0.7a Data show crude mean 6 standard deviation, median (interquartile range), or proportions (n). a 0.9 (0.4, 2.3) and 3.2 6 0.7 among females with no hormonal contraceptive use for C-reactive protein and fibrinogen, respectively.
consumption of number of standard drinks (10 g of alcohol) (22). Scores from three questions (What type of milk do you use? What spread do you usually use on bread? How often is the meat you eat trimmed of fat?) were combined to form a surrogate fat intake variable. Fasting glucose was measured by the Olympus AU5400 automated analyzer. Fasting insulin was measured by immunoassays with inter-assay standardization (16). We estimated insulin sensitivity using the homeostatic model assessment (HOMA) index (fasting glucose 3 fasting insulin/22.5) (23). We asked female participants whether they were currently using hormonal contraceptives. At both waves, information on cigarette smoking was derived from a self-administered questionnaire. Childhood and adulthood arealevel socioeconomic status (SES) of each participant was derived
2600
Obesity | VOLUME 22 | NUMBER 12 | DECEMBER 2014
from their residential postcode or census collection district using the Australian Bureau of Statistics Index of Relative Socioeconomic Disadvantage based on either the 1986 (24) or 2006 population census (25). Percent body fat was used to calculate lean body mass by subtracting fat mass from total body mass.
Statistical analysis Categorical variables were described using percentages and continuous variables using means and standard deviations or medians and inter-quartile ranges if skewed in distribution. We conducted analyses stratified by males and females due to the effect of hormonal contraceptive use on CRP and fibrinogen among females (26). We assessed the evidence for sex differences in the associations between fitness, adiposity, and inflammation by including fitness/adiposity and sex cross-product (interaction) terms in the multivariable models. CRP was natural log transformed due to its skewed distribution. Using linear regression models, we examined associations of childhood fitness and adiposity with adult CRP and fibrinogen. For CRP, logscale regression coefficients were interpreted as percent change in outcome per 1-SD change in exposure variable (% change 5 1 2 exp(b) for negative b or exp(b) 2 1 for positive b). We inspected residuals graphically after fitting linear regression models to check for nonlinear associations between continuous exposures and outcomes (confirming that linear models were adequate: results not shown). A range of potential confounders were considered in the model building process. We used relatedness of potential confounders (age, childhood and adulthood SES, smoking, fat intake and alcohol consumption, education, and hormonal contraceptive use for females) to exposures and outcomes with a 10% change-in-estimate criterion for inclusion in models. We adjusted for lean body mass in linear regression models when PWC170 was used as the fitness variable given that the absolute workload achieved is a function of muscle mass (27,28). For each of the two exposures, childhood fitness and adiposity, we initially adjusted for identified confounders in Model 1. In Model 2, we included both exposure variables to investigate whether the effect of each was independent of the other, adjusted for confounders. We also adjusted for adulthood insulin sensitivity in the models to test its potential mediation effect. We tested for potential interaction effects in the relationships by including fitness tertile and adiposity (BMI) cross-product term in Model 2 in both sexes. We noted that our sample included both prepubertal and pubertal children. Although we did not measure pubertal status, we tested for potential interaction effects in the relationships by including categorical age variable (7-9 years vs. 10-15 years groups for girls; 7-11 years vs. 12-15 years groups for boys; roughly corresponding to pubertal status) and exposure cross-product term in the models. In sensitivity analyses, we repeated all analyses among female participants with no hormonal contraceptive use. Similar analyses were performed among nonsmokers to assess a potential residual confounding effect of adult smoking on the associations. We also conducted sensitivity analyses using population-weighted models to assess whether the results were influenced by differences in demographic and adiposity characteristics between the 1,976 participants analyzed and those excluded. Weights were used to adjust the analyzed sample to be representative of the reference group (all
www.obesityjournal.org
Original Article
Obesity
EPIDEMIOLOGY/GENETICS
TABLE 2 Associations of childhood cardiorespiratory fitness and adiposity with adult C-reactive protein (CRP), from multivariable linear regression analysis of childhood exposures on lnCRP
Childhood exposures Cardiorespiratory fitness Body mass index z score Cardiorespiratory fitness Body mass index z score Cardiorespiratory fitness Body mass index z score Cardiorespiratory fitness Body mass index z score Cardiorespiratory fitness Body mass index z score
Crude model b (95% CI)
P-value
Model 1 adjusted b (95% CI)
P-value
Males (N 5 1,016) 0.005 20.11 (20.19, 20.03) 0.005
