Abstract
Objective:
To examine the social and behavioural correlates of metabolic phenotypes during ‘at-risk’ and ‘case’ stages of the metabolic disease continuum.
Design:
Cross-sectional study of a random population sample.
Participants:
A total of 718 community-dwelling adults (57% female), aged 18–92 years from a regional South Australian city.
Measurements:
Total body fat and lean mass and abdominal fat mass were assessed by dual energy x-ray absorptiometry. Fasting venous blood was collected in the morning for assessment of glycated haemoglobin, plasma glucose, serum triglycerides, cholesterol lipoproteins and insulin. Seated blood pressure (BP) was measured. Physical activity and smoking, alcohol and diet (96-item food frequency), sleep duration and frequency of sleep disordered breathing (SDB) symptoms, and family history of cardiometabolic disease, education, lifetime occupation and household income were assessed by questionnaire. Current medications were determined by clinical inventory.
Results:
36.5% were pharmacologically managed for a metabolic risk factor or had known diabetes (‘cases’), otherwise were classified as the ‘at-risk’ population. In both ‘at-risk’ and ‘cases’, four major metabolic phenotypes were identified using principal components analysis that explained over 77% of the metabolic variance between people: fat mass/insulinemia (FMI); BP; lipidaemia/lean mass (LLM) and glycaemia (GLY). The BP phenotype was uncorrelated with other phenotypes in ‘cases’, whereas all phenotypes were inter-correlated in the ‘at-risk’. Over and above other socioeconomic and behavioural factors, medications were the dominant correlates of all phenotypes in ‘cases’ and SDB symptom frequency was most strongly associated with FMI, LLM and GLY phenotypes in the ‘at-risk’.
Conclusion:
Previous research has shown FMI, LLM and GLY phenotypes to be most strongly predictive of diabetes development. Reducing SDB symptom frequency and optimising the duration of sleep may be important concomitant interventions to standard diabetes risk reduction interventions. Prospective studies are required to examine this hypothesis.
Similar content being viewed by others
Introduction
A considerable proportion of obese individuals do not co-express metabolic risk (metabolically healthy obese) and a considerable proportion of normal weight individuals express multiple metabolic risk factors (metabolically unhealthy lean).1 In both obese and lean individuals, the main predictors of metabolic ill-health in the U.S. NHANES were older age and lower participation in leisure-time physical activity.1
The co-expression of metabolic traits vary both in nature (the traits that co-express) and in magnitude (the level to which they express) across individuals and populations,2, 3, 4, 5 and likely result from interactions between genetic and environmental factors, which have not been fully characterised.4, 6 Understanding the specificity of socio-economic and behavioural determinants of metabolic trait co-expression at different stages of the metabolic disease continuum may aid in tailoring diabetes prevention and management programs to individuals and communities with particular metabolic trait or risk profiles.
Current understandings of the social and behavioural determinants of metabolic status have come from studies, which use either individually measured risk factors or metabolic syndrome classifications as outcomes.7, 8, 9 The former approach may be limited by the inability to account for the co-expression of metabolic traits in individuals and the clinical dichotomy of the latter excludes examination of either the nature or the magnitude of trait co-expression.
Standardised definitions for metabolic phenotypes (such as metabolically healthy obese and metabolically unhealthy lean) do not exist, which limits the comparability of findings across studies, and if they did, they would have the same limitations as definitions for metabolic syndrome. Since 1994 a substantial literature has emerged from factor or principal components analysis (PCA)10 that has identified four canonical (accepted) latent (underlying/unmeasured) metabolic factors (phenotypes) in the pathophysiology of diabetes: body/fat mass/insulinemia (FMI) glycaemia (GLY), lipidaemia and blood pressure (BP).11, 12, 13 Despite these phenotypes being studied as determinants of diabetes,12 to our knowledge no studies have extensively examined the social and behavioural determinants of phenotype expression.
The objective of this study was to examine the specificity of social and behavioural correlates of latent metabolic phenotypes during ‘at-risk’ and ‘case’ stages of the metabolic disease continuum. The study aimed to confirm the presence of the canonical latent factors and to examine their social and behavioural correlates in ‘at-risk’ and ‘case’ sub-populations.
Participants and methods
Research context
The Whyalla Intergenerational Study of Health (WISH) is a population-based cohort recruited between February 2008 and July 2009, within the city of Whyalla, an industrial outer regional city that is located on the eastern Eyre Peninsula of South Australia.14 The city had a population of over 22 000 in 2006.
When compared with South Australia and Australia overall, Whyalla has a workforce enriched with blue-collar workers, and the demand for such occupations may explain the higher proportion of adults who had not completed year 12 of high school (or equivalent). The proportion of UK/Irish born and people of Australian Aboriginal or Torres Strait Islander origin appear higher than in South Australia and Australia overall.14
Previous investigations of the health of this population have revealed obesity,15 asthma,15, 16 chronic lung disease,15, 16 chronic liver disease and lung cancer16 to be higher in Whyalla than in South Australia overall, or demographically and geographically comparative communities. A previous ecological study suggested that behavioural risk factors were unable to explain the apparent poorer respiratory and hepatic health outcomes in this population.16
Sampling and Recruitment
Whyalla households (n=2500) were randomly selected from the residential housing database of the State Planning Authority. The strength of this sampling frame was the completeness, however, approach and recruitment was a complex multi-stage process because of limited contact information in the databases (residential address only). This novel sampling frame was used instead of telephone listings (Electronic White Pages) owing to a high proportion of households without landline telephone connections in the city. Invitations to participate were mailed to ‘The Householder’ co-ordinated with a community-wide media campaign informed by a Community Advisory Group. Householders were invited to register online or by telephone, providing their telephone number and basic demographic information. One hundred seventy-eight (7%) households responded to this approach. The second stage successfully matched 1183 household addresses in the sample to names and telephone numbers in the Electronic White Pages. Third, remaining unmatched households were approached by door knocking, a minimum of two attempts to contact were made and calling cards were left.
Once telephone numbers were obtained, Computer Assisted Telephone Interviewing technology was used to recruit the adult (⩾18 years) in each household who last had their birthday (‘index adult’), and to collect demographic, health and risk factor data and schedule clinical assessments. Interviews were conducted by trained interviewers under established protocols.17 Participants provided written informed consent at the time of clinical assessment. Protocols and procedures were approved by the Human Research Ethics Committee of The University of South Australia and the Aboriginal Health Research Ethics Committee of South Australia.
Definitions of ‘at-risk’ and ‘case’ stratification
‘At-risk’ was defined as the absence of both known diabetes and pharmacological management with diabetes or cardiovascular system agents (Anatomical Therapeutic Classification codes A10, C01-C03 and C07-C10). ‘Case’ was defined as diagnosed diabetes or management with the aforementioned therapeutic classes.
Major outcome measures
Body composition was assessed using DXA (Lunar Prodigy, GE Medical Systems, Madison, WI, USA) with the manufacturer's software (enCORE 2003 version 7.52). Scans were performed according to the manufacturer's protocols with participants’ supine. Total fat and lean masses (g) were determined automatically from the manufacturer's software. Total abdominal and intra-abdominal fat masses were determined by defining two regions of interest in the abdomen by drawing quadrilateral boxes with the base touching the top of the iliac crest, the upper margins touching the most inferior aspect of the ribs and the lateral borders extending to the edge of the abdominal soft tissue for total abdominal fat, and the inner border of the rib cage for intra-abdominal fat.
Fasting, morning venous blood samples were collected (∼40 ml) and assayed by a National Association of Testing Authorities accredited laboratory. Fasting plasma glucose concentration was measured using a Chemistry analyser system (Olympus AU5400, Olympus Optical C Ltd, Japan) with inter-assay CVs of 2.41% at 3.41 mmol l−1 and 2.21% at 19.72 mmol l−1. HbA1c was assayed by cation-exchange high pressure liquid chromatography using instrumentation, kits and reagents from Bio-Rad laboratories (Hercules, CA, USA) with inter-assay CVs of 2.4% at A1c of 5.7 and 2.2% at A1c of 10.0. Fasting serum triglycerides and cholesterol lipoproteins were determined enzymatically using a Roche Hitachi 911 chemistry analyser (Roche, Boehringer, Germany) with inter-assay CV's of: 2.63% at 2.44 mmol l−1 and 3.16% at 1.0 mmol l−1 of triglyceride; 2.05% at 5.83 mmol l−1 and 2.83% at 2.55 mmol l−1 of total cholesterol; 3.29% at 2.04 mmol l−1 and 3.2% at 0.9 mmol l−1 of high-density lipoprotein (HDL) cholesterol. Fasting serum insulin concentration was measured in serum collected, separated and frozen at −80 °C within 2 h of collection. Radioimmunoassay was performed on an Abbott Architect immunoassay analyser (Abbott Park, IL, USA) with inter-assay CVs of 7.81% at 8.44 mU l−1, 5.87% at 57.44 mU l−1 and 8.53% at 97.90 mU l−1. Systolic and diastolic BP were measured using an automatic sphygmomanometer (Welch Allyn Spot Vital Signs 420 Series, Skaneateles Falls, NY, USA). Two measurements were taken after the participant had been seated and rested for 5 and 10 min. A third measurement was taken if readings differed by more than 10%. The mean of the two readings or the median of three readings was calculated for both pressures. Height was determined to within 1 mm using a wall stadiometer (Surgical and Medical Products No. 26SM, Mentone Education Supplies, Melbourne, Australia), body mass was measured to the nearest 100 g with an electronic scale (Tanita BF-681, Tanita, Arlington Heights, IL, USA) and waist circumference was measured twice to the nearest 0.5 mm using a steel girth tape (Lufkin W606PM) and the mean value obtained. A third measurement was taken if the two varied by greater than 1%, and the median was used.18 Body mass index (BMI) was calculated as weight (kg)/height (m2).
For comparison with national prevalence data, abdominal obesity was classified as waist circumference ⩾102 cm for males and ⩾88 cm for females and general obesity was classified as BMI⩾30 kg m−2.19 Elevated systolic and diastolic BP were defined as ⩾140 and ⩾90 mm Hg, respectively.20 Elevated fasting triglyceride concentrations were defined as ⩾2.0 mmol l−1 and low HDL cholesterol was defined as ⩽1.0 mmol l−1.20 Impaired fasting glucose was defined as a fasting plasma glucose concentration >5.5 mmol l−1(ref.21) and diabetes was indicated by a fasting plasma glucose concentration >7 mmol l−1.20 Undiagnosed diabetes was defined as having a fasting plasma glucose concentration >7 mmol l−1 without having reported a previous diagnosis of diabetes.
Metabolic phenotype expression scores
Intra-abdominal, abdominal and total fat masses, total lean mass, fasting serum concentrations of triglycerides, HDL and insulin, fasting plasma concentration of glucose, HbA1c, seated systolic and diastolic BP were subject to PCA for dimension reduction to the latent factors. This analysis assigned a score to each participant representing how strongly they expressed each latent phenotype.
Independent variables
Similarly, the consumption frequencies for ninety-six food and beverage items collected using an updated version (version 3.1) of the Australian Cancer Council of Victoria's Dietary Questionnaire22 were entered into a PCA for dimension reduction to latent dietary patterns, which assigned a score to each participant representing how strongly their diet reflected each of the identified patterns. From the same questionnaire, alcohol consumption was computed in grams. Smoking status was defined as never-, past- and current-smoker. Total time spent in leisure-time physical activities (LTPA) was determined by questionnaire. Intensity weights (3.5 for walking, 5.0 for moderate and 7.5 for vigorous intensity exercise) were used to compute metabolic equivalent hours (MET-h), where <1600 MET-h/week was defined as sedentary or low; and >1600 MET-h/week as moderate to high physical activity.23 Average hours spent sleeping each day was self reported. Owing to evidence that both short and long sleep durations are associated with increased risk of diabetes,24 sleep duration was categorised as short (<7 h per day) or long (⩾9 h per day) sleep and examined relative to normal sleep duration (7–<9 h per day). A sleep disordered breathing (SDB) symptom score (range 0–12) was computed as the sum of symptom frequency from zero (never) to four (>4 times per week) for three symptoms: snorting or gasping; loud snoring; and breathing cessation, choking or struggling for breath during sleep. Family history (parents and grandparents) of diabetes, high BP, stroke and heart disease were obtained by self-report. Pictograms25 were used to estimate maternal and paternal somatotypes at age 40 years. Fifteen year retrospective somatotyping of parents, by offspring, using these pictograms, correlated highly with previously measured height and weight.26 Educational attainment was categorised as Bachelor degree or higher, high school or vocational certificate, or did not complete high school. Lifetime primary occupation was coded according the Australian and New Zealand Classification of Occupations and reduced to ‘Manager/professional’, ‘White collar worker’ and ‘Blue collar worker’. Gross annual household income was collected in approximate $20 000 increments and reduced to three categories of $40 000 increments for analysis.
Standard population data
Baseline data from the Australian Obesity Diabetes and Lifestyle (AusDiab) Study, a stratified sample of adults 25 years and over from 42 randomly selected Census Collector Districts in Australia in 1999–2000 (n=11 247, pseudo-response rate 55%),27 were used to estimate standardised prevalence ratios (SPR) for cardiometabolic risk factors in the study population.
Australian and South Australian behavioural risk factor data from the 2007–08 Australian National Health Survey28 were used to estimate SPRs for behavioural risk factors in the study population.
Statistical analysis
Data analysis was performed using STATA 10.1 for Windows (StataCorp., College Station, TX, USA). For population prevalence estimates, data were weighted to the age and sex distribution of the Estimated Residential Population of Whyalla in 2007(ref. 29) and the likelihood of being recruited into the study, as per the equation: weight=hhld_adult × (bnh/lnh) × (ln/bn); where hhld_adult is the number of people aged 18 and over (adults) residing in the household; bnh is the adult population size of Whyalla by sex and age-group; bn is the total adult population of Whyalla; lnh is the adult sample size of WISH by sex and age-group; and ln is the total adult sample size of the WISH cohort. For SPRs, the stratum-specific prevalence estimates from the national standard population data (AusDiab and Australian National Health Survey) were used to calculate the expected number of cases in the study population (WISH). Expected cases were summed across strata, and compared with the number of observed cases, the SPR was calculated as the ratio of observed to expected cases.
PCA with oblique rotation (oblimin) was used to define the principal latent metabolic phenotypes and the principal latent dietary patterns in the population. Components were maintained for further analysis based on Eigen values >1 (see Supplementary Information, Figure 1) and limited to four for interpretability. To minimise the likelihood of over-specifying the number of components, potentially outlying and influential observations were examined using scatter-plots and bi-plots of the first and last two components as suggested by Jolliffe.30 Components scores were standardised for further analysis.
Hierarchical, multivariate (multiple outcomes) regression was used to model the associations of age, sex, education, lifetime primary occupational class, household income, family history of cardiometabolic disease, dietary pattern scores, alcohol consumption, past and current smoking, moderate-high LTPA, sleep duration and SRB symptoms with latent phenotypes.
Results
Participants
The participant flow and response rates are summarised in Figure 1. Overall, 32.2% of eligible adults participated in the clinical study. Married people were over-represented by ∼20% and divorced/separated people and those who had never married were under-represented by ∼10%. The sample was 10% over-represented by adults who owned their residences or rented privately and 10% under-represented by those renting from the State Housing administration. There was an ∼5% over-representation of adults born in the UK and Ireland and a 1.5% under-representation of Aboriginal or Torres Strait Islanders (data not shown).
Table 1 describes the metabolic, behavioural and socio-economic characteristics of the ‘At-risk’ and ‘Case’ sub-populations by sex. The prevalence of both abdominal obesity (48.4% (95% CI 44.1−52.7%)) and low HDL cholesterol (17.7% (95% CI 14.2−21.8%)) was higher than national estimates (SPR for abdominal obesity 1.6 (95% CI 1.4−1.8), SPR for low HDL cholesterol 1.3 (95% CI 1.1−1.6)). All other metabolic risk markers were equivalent to the national prevalence. Hypertension prevalence was 31.3% (95% CI 27.7−35.1%); 8.7% (95% CI 6.7−11.4%) were untreated and 55.8% (95% CI 47.6−63.7%) were adequately controlled below 140/90 mmHg. The prevalence of hypertriglyceridaemia was 22.1% (95% CI 18.8−25.9%). After excluding diabetes that was only associated with pregnancy (13 cases), the prevalence of self-reported doctor-diagnosed diabetes was 7.5% (95% CI 5.8−9.8%). Four percent (4.3%) of the population had biochemical evidence of diabetes (fasting plasma glucose ⩾7 mmol l−1) at clinical assessment. Based on self-reported diagnoses and biochemical evidence, the estimated prevalence of likely undiagnosed diabetes was 1.5% resulting in an estimated population prevalence for total diabetes of 9.1% (95% CI 7.1−11.6%). This equated to one undiagnosed diabetic for every 5 diagnosed cases in Whyalla. The estimated population prevalence of impaired fasting glucose was 6.4% (95% CI 4.5−9.2%). The prevalence of moderate-high LTPA (35.0% (95% CI 30.8−39.2%)) was significantly higher than the national prevalence (SPR 1.3 (95% CI 1.1−1.5)). Current smoking (23.0% (95% CI 19.6−26.7%)), past smoking (28.6% (95% CI 25.1−32.4%)) and never smoking (48.4% (95% CI 44.1−52.8%)) were similar to national and state estimates.
Latent metabolic phenotypes and dietary factors
Four latent metabolic phenotypes were confirmed in both ‘at-risk’ and ‘case’ groups: FMI; BP; lipidaemia/lean mass (LLM) and GLY (Table 2 ). Moderate correlations existed between all latent phenotypes except for BP in ‘cases’, which was uncorrelated with the other phenotypes. Four latent dietary patterns were identified that are described as: Mediterranean; Fruit; Anglo; and Junk (See Supplementary data file).
Socio-economic, familial and behavioural correlates
The covariate-adjusted models of phenotype expression in ‘at-risk’ and ‘case’ sub-populations are shown in Tables 3 and 4, respectively (see Supplementary data file for the sequential models). In the ‘at-risk’ strata, high school and/or vocational education and a unit increase in SDB symptom score were associated with a 0.51 and 0.11 s.d. increase in FMI expression, respectively. Blue-collar occupation and a unit increase in the maternal endomorphic-ectomorphic scale were associated with a 0.38 s.d. and 0.13 s.d. decrease. In ‘cases’, a 1-year age increase and long sleep was associated with a 0.03 s.d. and 0.56 s.d. decrease, and taking diuretics with a 0.62 s.d. increase in FMI expression.
In the ‘at-risk’, female gender and current smoking were associated with a 0.84 s.d. and 0.35 s.d. decrease in BP expression, respectively. A 1-year age increase was associated with a 0.02 s.d. increase. In ‘cases’, a one unit increase in ‘fruit’ diet pattern score was associated with a 0.08 s.d. increase, and taking diuretics with a 0.60 s.d. decrease in BP expression.
In the ‘at-risk’, female gender, household income up to $40 000 p.a. and a 100 g per month increase in alcohol consumption (10 Australian standard drinks) were associated with a 1.17 s.d., 0.28 s.d. and 0.01 s.d. decrease in LLM expression, respectively. Current smoking and a unit increase in SDB symptom score were associated with a 0.31 s.d. and 0.05 s.d. increases, respectively. In ‘cases’, female gender and a 1-year age increase were associated with a 1.34 s.d. and 0.02 s.d. decrease in LLM expression, respectively. Lipid-lowering medications were not associated with LLM expression.
In the ‘at-risk’, female gender and blue-collar occupation were associated with a 0.51 s.d. and 0.28 s.d. decrease in GLY expression, respectively. A 1-year age increase, a 1 unit increase in SDB symptom score and short sleep duration were associated with a 0.03 s.d., 0.05 s.d. and 0.38 s.d. increase, respectively. In ‘cases’, being female was associated with a 0.40 s.d. decrease and family history of diabetes, moderate-high LTPA, short sleep and taking medication for diabetes were associated with a 0.37 s.d., 0.41 s.d., 0.34 s.d. and 1.06 s.d. increase in GLY expression, respectively.
Discussion
Metabolic health profile of the population
Almost one in two adults in the study population had abdominal obesity, which, when standardised to 10-year old Australian national estimates, represented 1.6 cases to every case nationally. Despite the excess of abdominal obesity, which may be expected given the age of the reference data, the prevalence of diabetes, hypertriglyceridaemia and hypertension were equivalent to estimates from the same national data. This suggests that metabolic risk in Whyalla was either well controlled or had not emerged from the observed abdominal obesity. Whyalla had higher prevalence of controlled and lower prevalence of untreated hypertension suggesting that the incongruence between abdominal obesity and hypertension may be due to improvements in detection and management over the past decade, as has been demonstrated internationally.31 Improved pharmacological management may also explain the incongruent hyperlipidemia prevalence, however, it is unlikely to explain the incongruent impaired fasting glucose and diabetes prevalence as the use of insulin sensitising or beta-cell preserving agents for diabetes prevention is not yet widespread despite emerging positive evidence.32, 33, 34 The prevalence of moderate-high LTPA was higher than the contemporary national prevalence, which may confer metabolic risk reduction independent of fat loss.35, 36 Understanding the determinants of this (and other) particular community metabolic profiles has implications for the design of diabetes prevention programs.
This study examined the dominant metabolic phenotypes, their coexpression patterns and their associations with socio-economic characteristics, health-related behaviours and family cardiometabolic health history in ‘at-risk’ and ‘case’ sub-populations. The dominance of the widely accepted underlying metabolic phenotypes, similar to those previously described in diabetics and non-diabetics12 were confirmed in this population. GLY and FMI phenotypes have been shown to predict diabetes equivalently to NCEP defined metabolic syndrome; LLM is slightly inferior and BP has the lowest predictive value.12 Consistent with this finding, the BP phenotype in ‘cases’ was uncorrelated with the other phenotypes, and in counter distinction with the GLY phenotype, explained less of the metabolic variance in the ‘case’ versus ‘at-risk’ sub-population.
Correlates of metabolic phenotypes in ‘cases’: implications for management
These differences likely reflect the impact of pharmacological regulation of metabolic traits in ‘cases’ as opposed to physiological regulation in the ‘at-risk’. The absence of associations between phenotype expression and medication may indicate successful treatment to target, that is, those taking medication have, on average, the same phenotype expression as those not taking medication. This was the case for lipid modifying agents, but not diuretics, which were strongly associated with lower BP and higher FMI expression, suggesting that they lower BP below the untreated average and increase fat mass and fasting insulin above the untreated average. Diuretics such as amiloride can raise fasting insulin37 and may predict type 2 diabetes in people with impaired glucose tolerance.38 In particular, thiazide diuretics increase hepatic fat content and c-reactive protein, both of which are associated with reduced insulin sensitivity.39 Thus, while diuretics may lower BP, they can have adverse metabolic effects, in particular reducing insulin sensitivity. In contradistinction, diabetic medication was associated with higher GLY expression (and therefore, failure to achieve normoglycaemia). Lowering glucose concentrations to normal in diabetics, however, may not be the treatment goal as the benefits versus risks of aggressive management are still unclear.40, 41 Similarly, moderate-high LTPA was not associated with GLY expression in the ‘at-risk’, but was associated with higher expression in ‘cases’, suggesting that it fails to downregulating expression to normoglycaemia in ‘cases’.
In both ‘at-risk’ and ‘cases’, moderate-high LTPA was inversely associated with FMI expression, consistent with studies, which have shown improvements in fat mass and insulin regulation across the metabolic disease continuum42, 43 through increasing cardiorespiratory fitness.44 As moderate-high LTPA was associated with the lower FMI expression only, it does not explain the incongruence between abdominal obesity and metabolic risk in this population. Instead, the higher prevalence of self-reported moderate-high LTPA may reflect greater perceived exertion due to physical activities in the less metabolically fit.45 This highlights the importance of physical activity for secondary prevention during progression along the metabolic disease continuum.
Interestingly, long sleep was associated with a more favourable FMI expression in ‘cases’, which is consistent with the mounting evidence that sleep curtailment promotes weight gain and the onset of obesity.46 Current clinical trials in obese individuals aim to investigate the effect of extending sleep duration on obesity and cardiometabolic risk factor trajectories.47 Consistent with studies showing and increased likelihood of having diabetes in short and long sleepers,48, 49 short sleep in this study was associated with less favourable GLY expression in both ‘cases’ and ‘at-risk’ subpopulations. This suggests a negative impact on glucose regulation throughout the metabolic disease continuum or alternatively, shortened sleep as a result of even sub-clinical derangements in glycaemic regulation. The association between sleep duration, obesity and glycaemic dysregulation is potentially mediated by an increase in appetite.50 Intervening on diet and exercise in overweight people with Impaired Glucose Tolerance improve body weight and insulin sensitivity to similar degrees in both short and long sleepers, but appeared to produce greater reductions in the incidence of type 2 diabetes in long sleepers.51
Correlates of metabolic phenotypes in the ‘at-risk’: implications for prevention
Of all the socio-economic, family history and behavioural variables examined, the frequency of SDB symptoms was the only variable associated with multiple phenotypes, after adjustment for sleep duration and other covariates. More frequent symptoms were associated with greater expression of FMI, LLM and GLY phenotypes, suggesting that these modifiable symptoms may be partly responsible for phenotype co-expression, leading to greater risk of diabetes. Consistent with this interpretation, the BP phenotype, which holds the lowest predictive value for future diabetes,12 was not associated with SDB in ‘at-risk’ adults. Associations of clinically defined SDB with glucose intolerance and insulin resistance have been previously reported.52 Although there are fewer than 20 prospective studies in this field, current evidence suggests that SDB may be a prevalent cause of diabetes53, 54 and poor glycaemic control in diabetics, which can be ameliorated by continuous positive airway pressure therapy.55 Both SDB symptoms and risk of diabetes are modifiable by weight loss in very fat people.56 The causal role of SDB in dyslipidaemia is less clear.57
In addition, mid-level education, family history of diabetes and maternal adiposity were major correlates of FMI expression in the ‘at-risk’. Higher adherence to the ‘fruit’ and ‘Anglo’ diet patterns were associated with lower GLY expression, however, this was not independent of family history, other behaviours or SDB. These findings are contrary to those of other studies linking Mediterranean diet to reduced risk of metabolic syndrome.58 This may be owing to differences in what this dietary pattern represents locally versus in the international literature on ‘Mediterranean diet’. Higher alcohol consumption and not being a current smoker were associated with lower LLM expression, consistent with the beneficial effect of alcohol consumption on increasing HDL cholesterol59 and negative effects on muscle mass and function.60
The strengths of this study include the metabolic characterisation of participants, particularly the use of DXA to quantify fat and lean masses. Limitations include the characterisation of behaviour by self-report. This study provides insight into the specificity of proximal social and behavioural factors as correlates of metabolic phenotypes underlying diabetes development. The findings have implications for tailoring diabetes prevention programs in communities and specific interventions in individuals with different phenotype expression. Reducing the frequency of SDB symptoms may be effective in reducing the expression of the three latent phenotypes that most strongly predict diabetes development. Moreover, management of optimal sleep duration may have added benefits for the glycaemic regulation across the full continuum of metabolic health. These hypotheses require prospective examination due to the possibility of reverse causality despite the stratification used in this study.
References
Wildman RP, Muntner P, Reynolds K, McGinn AP, Rajpathak S, Wylie-Rosett J et al. The obese without cardiometabolic risk factor clustering and the normal weight with cardiometabolic risk factor clustering: prevalence and correlates of 2 phenotypes among the US population (NHANES 1999–2004). Arch Intern Med 2008; 168: 1617–1624.
Forouhi NG, Sattar N . CVD risk factors and ethnicity--a homogeneous relationship? Atheroscler Suppl 2006; 7: 11–19.
O’Dea K, Rowley KG . Macrovascular disease risk factors and insulin resistance in Aboriginal and Torres Strait Islander people. J Diabetes Complications 2002; 16: 9–16.
Narayan KM, Aviles-Santa L, Oza-Frank R, Pandey M, Curb JD, McNeely M et al. Report of a National Heart, Lung, And Blood Institute Workshop: heterogeneity in cardiometabolic risk in Asian Americans In the US Opportunities for research. J Am Coll Cardiol 2010; 55: 966–973.
Rostambeigi N, Shaw JE, Atkins RC, Ghanbarian A, Cameron AJ, Forbes A et al. Waist circumference has heterogeneous impact on development of diabetes in different populations: longitudinal comparative study between Australia and Iran. Diabetes Res Clin Pract 2010; 88: 117–124.
Guilloume M, Bjorntorp P . Obesity in children. Environmental and genetic aspects. Horm Metab Res 1996; 28: 573–581.
Hata Y, Nakajima K . Life-style and serum lipids and lipoproteins. J Atheroscler Thromb 2000; 7: 177–197.
Gall SL, Jamrozik K, Blizzard L, Dwyer T, Venn A . Healthy lifestyles and cardiovascular risk profiles in young Australian adults: the Childhood Determinants of Adult Health Study. Eur J Cardiovasc Prev Rehabil 2009; 16: 684–689.
Shi L, Morrison JA, Wiecha J, Horton M, Hayman LL . Healthy lifestyle factors associated with reduced cardiometabolic risk. Br J Nutr 2011; 105: 747–754.
Meigs JB . Invited commentary: insulin resistance syndrome? Syndrome X? Multiple metabolic syndrome? A syndrome at all? Factor analysis reveals patterns in the fabric of correlated metabolic risk factors. Am J Epidemiol 2000; 152: 908–911; discussion 912.
Meigs JB, D’Agostino Sr RB, Wilson PW, Cupples LA, Nathan DM, Singer DE . Risk variable clustering in the insulin resistance syndrome. The Framingham Offspring Study. Diabetes 1997; 46: 1594–1600.
Hanson RL, Imperatore G, Bennett PH, Knowler WC . Components of the ‘metabolic syndrome’ and incidence of type 2 diabetes. Diabetes 2002; 51: 3120–3127.
Edwards KL, Burchfiel CM, Sharp DS, Curb JD, Rodriguez BL, Fujimoto WY et al. Factors of the insulin resistance syndrome in nondiabetic and diabetic elderly Japanese-American men. Am J Epidemiol 1998; 147: 441–447.
Haren MT, Stokes A, Taylor J, McDermott RA . Community participation in the Whyalla Intergenerational Study of Health (WISH): looking back on the possible influences of knowledge, trust and power. Public Health Bulletin SA 2011; 8: 41–48.
Dal Grande E, Dempsey P, Johnson G, Taylor A . Whyalla Health Survey. South Australian Department of Human Service: Adelaide, 2000.
The Whyalla Health Impact Study. South Australian Department of Health: Adelaide, 2006.
Grant JF, Chittleborough CR, Taylor AW, Dal Grande E, Wilson DH, Phillips PJ et al. The North West Adelaide Health Study: detailed methods and baseline segmentation of a cohort for selected chronic diseases. Epidemiol Perspect Innov 2006; 3: 4.
International Society for Advancement of Kinanthropometry. International Conference : Melbourne A, Marfell-Jones M, Olds T, Commonwealth International Sport Conference : Melbourne A. Kinanthropometry X: proceedings of the 10th International Society for Advancement of Kinanthropometry Conference, held in conjunction with the 13th Commonwealth International Sport Conference / edited by Mike Marfell-Jones and Tim Olds. Routledge: London; New York, 2008.
Obesity: preventing and managing the global epidemic: report of a WHO consultation on obesity, Geneva, 3–5 June, 1997, World Health Organisation: Geneva, 1998.
Alberti KG, Zimmet PZ . Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation. Diabet Med 1998; 15: 539–553.
Alberti KG, Zimmet P, Shaw J . The metabolic syndrome--a new worldwide definition. Lancet 2005; 366: 1059–1062.
Hodge A, Patterson AJ, Brown WJ, Ireland P, Giles G . The Anti Cancer Council of Victoria FFQ: relative validity of nutrient intakes compared with weighed food records in young to middle-aged women in a study of iron supplementation. Aust N Z J Public Health 2000; 24: 576–583.
Craig CL, Marshall AL, Sjostrom M, Bauman AE, Booth ML, Ainsworth BE et al. International physical activity questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003; 35: 1381–1395.
Yaggi HK, Araujo AB, McKinlay JB . Sleep duration as a risk factor for the development of type 2 diabetes. Diabetes Care 2006; 29: 657–661.
Stunkard A, Sorensen T, Schulsinger F . Use of the Danish Adoption Register for the study of obesity and thinness. In: Kety S, Roland L, Sidman R, Matthysse S (eds). The genetics of neurological and psychiatric disorders. Raven Press: New York, 1983.
Sorensen TI, Stunkard AJ, Teasdale TW, Higgins MW . The accuracy of reports of weight: children's recall of their parents’ weights 15 years earlier. Int J Obes 1983; 7: 115–122.
Cameron AJ, Welborn TA, Zimmet PZ, Dunstan DW, Owen N, Salmon J et al. Overweight and obesity in Australia: the 1999–2000 Australian Diabetes, Obesity and Lifestyle Study (AusDiab). Med J Aust 2003; 178: 427–432.
Australian Bureau of Statistics. National Health Survey: Summary of Results (2007–2008) reissue. Commonwealth of Australia, 2009.
Australian Bureau of Statistics. Population by age and sex South Australia (2007). Commonwealth of Australia: Canberra, 2008.
Jolliffe IT . Principal Component Analysis 2nd edn. Springer-Verlag: New York, 2002.
Andersen UO, Jensen GB . Trends and determinant factors in hypertension control in a population study with 25 years of follow-up. J Hypertens 2010; 28: 1091–1096.
Wan Q, Wang F, Wang F, Guan Q, Liu Y, Wang C et al. Regression to normoglycaemia by fenofibrate in pre-diabetic subjects complicated with hypertriglyceridaemia: a prospective randomized controlled trial. Diabet Med 2010; 27: 1312–1317.
Rhee MK, Herrick K, Ziemer DC, Vaccarino V, Weintraub WS, Narayan KM et al. Many Americans have pre-diabetes and should be considered for metformin therapy. Diabetes Care 2010; 33: 49–54.
DeFronzo RA, Abdul-Ghani M . Type 2 diabetes can be prevented with early pharmacological intervention. Diabetes Care 2011; 34 (Suppl 2): S202–S209.
Hamer M, O’Donovan G . Cardiorespiratory fitness and metabolic risk factors in obesity. Curr Opin Lipidol 2010; 21: 1–7.
Maraki MI, Aggelopoulou N, Christodoulou N, Anastasiou CA, Toutouza M, Panagiotakos DB et al. Lifestyle intervention leading to moderate weight loss normalizes postprandial triacylglycerolemia despite persisting obesity. Obesity (Silver Spring) 2011; 19: 968–976.
Eriksson JW, Fowelin J, Urbanavicius V . Insulin concentrations and insulin sensitivity after short-term amiloride in healthy subjects. Eur J Clin Pharmacol 1994; 46: 469–472.
Rajala U, Qiao Q, Laakso M, Keinanen-Kiukaanniemi S . Antihypertensive drugs as predictors of type 2 diabetes among subjects with impaired glucose tolerance. Diabetes Res Clin Pract 2000; 50: 231–239.
Duarte JD, Cooper-DeHoff RM . Mechanisms for blood pressure lowering and metabolic effects of thiazide and thiazide-like diuretics. Expert Rev Cardiovasc Ther 2010; 8: 793–802.
Johnson JA, Bowker SL . Intensive glycaemic control and cancer risk in type 2 diabetes: a meta-analysis of major trials. Diabetologia 2011; 54: 25–31.
Brown A, Reynolds LR, Bruemmer D . Intensive glycemic control and cardiovascular disease: an update. Nat Rev Cardiol 2010; 7: 369–375.
Jorge ML, de Oliveira VN, Resende NM, Paraiso LF, Calixto A, Diniz AL et al. The effects of aerobic, resistance, and combined exercise on metabolic control, inflammatory markers, adipocytokines, and muscle insulin signaling in patients with type 2 diabetes mellitus. Metabolism 2011; 60: 1244–1252.
Ryan AS . Exercise in aging: its important role in mortality, obesity and insulin resistance. Aging health 2010; 6: 551–563.
Totsikas C, Rohm J, Kantartzis K, Thamer C, Rittig K, Machann J et al. Cardiorespiratory fitness determines the reduction in blood pressure and insulin resistance during lifestyle intervention. J Hypertens 2011; 29: 1220–1227.
Aadahl M, Kjaer M, Jorgensen T . Perceived exertion of physical activity: negative association with self-rated fitness. Scand J Public Health 2007; 35: 403–409.
Itani O, Kaneita Y, Murata A, Yokoyama E, Ohida T . Association of onset of obesity with sleep duration and shift work among Japanese adults. Sleep Med 2011; 12: 341–345.
Cizza G, Marincola P, Mattingly M, Williams L, Mitler M, Skarulis M et al. Treatment of obesity with extension of sleep duration: a randomized, prospective, controlled trial. Clin Trials 2010; 7: 274–285.
Buxton OM, Marcelli E . Short and long sleep are positively associated with obesity, diabetes, hypertension, and cardiovascular disease among adults in the United States. Soc Sci Med 2010; 71: 1027–1036.
Chao CY, Wu JS, Yang YC, Shih CC, Wang RH, Lu FH et al. Sleep duration is a potential risk factor for newly diagnosed type 2 diabetes mellitus. Metabolism 2011; 60: 799–804.
Spiegel K, Tasali E, Penev P, Van Cauter E . Brief communication: Sleep curtailment in healthy young men is associated with decreased leptin levels, elevated ghrelin levels, and increased hunger and appetite. Ann Intern Med 2004; 141: 846–850.
Tuomilehto H, Peltonen M, Partinen M, Lavigne G, Eriksson JG, Herder C et al. Sleep duration, lifestyle intervention, and incidence of type 2 diabetes in impaired glucose tolerance: The Finnish Diabetes Prevention Study. Diabetes Care 2009; 32: 1965–1971.
Punjabi NM, Shahar E, Redline S, Gottlieb DJ, Givelber R, Resnick HE . Sleep-disordered breathing, glucose intolerance, and insulin resistance: the Sleep Heart Health Study. Am J Epidemiol 2004; 160: 521–530.
Muraki I, Tanigawa T, Yamagishi K, Sakurai S, Ohira T, Imano H et al. Nocturnal intermittent hypoxia and the development of type 2 diabetes: the Circulatory Risk in Communities Study (CIRCS). Diabetologia 2010; 53: 481–488.
Al-Delaimy WK, Manson JE, Willett WC, Stampfer MJ, Hu FB . Snoring as a risk factor for type II diabetes mellitus: a prospective study. Am J Epidemiol 2002; 155: 387–393.
Babu AR, Herdegen J, Fogelfeld L, Shott S, Mazzone T . Type 2 diabetes, glycemic control, and continuous positive airway pressure in obstructive sleep apnea. Arch Intern Med 2005; 165: 447–452.
Grunstein RR, Stenlof K, Hedner JA, Peltonen M, Karason K, Sjostrom L . Two year reduction in sleep apnea symptoms and associated diabetes incidence after weight loss in severe obesity. Sleep 2007; 30: 703–710.
Drager LF, Jun J, Polotsky VY . Obstructive sleep apnea and dyslipidemia: implications for atherosclerosis. Curr Opin Endocrinol Diabetes Obes 2010; 17: 161–165.
Kastorini CM, Milionis HJ, Esposito K, Giugliano D, Goudevenos JA, Panagiotakos DB . The effect of Mediterranean diet on metabolic syndrome and its components: a meta-analysis of 50 studies and 534 906 individuals. J Am Coll Cardiol 2011; 57: 1299–1313.
Husemoen LL, Jorgensen T, Borch-Johnsen K, Hansen T, Pedersen O, Linneberg A . The association of alcohol and alcohol metabolizing gene variants with diabetes and coronary heart disease risk factors in a white population. PLoS One 2010; 5: e11735.
Urbano-Marquez A, Fernandez-Sola J . Effects of alcohol on skeletal and cardiac muscle. Muscle Nerve 2004; 30: 689–707.
Acknowledgements
We acknowledge the continued support, guidance and perspective of the WISH Community Advisory Group and all the participants who have given their time to the study. The clinic staff Kate Warren, Tracey Patterson, Deb Misan, Leanne Gibbons, Chris Creber, Corey Beinke-Heath and Helen Mills and administrative staff Stephanie Coombes and Hayley-Maree Ewbank. We acknowledge also the participant recruitment efforts of Helen Pickhaver, Olivia Grove-Jones, Sarah Murray, Mick Jong, Delys and Wayne Champion. We gratefully acknowledge the contribution of the North West Adelaide Health Study team in their intellectual input into the processes, protocols and measurement methods included in this study and for the training of staff. We are particularly grateful to Janet Grant and Alicia Montgomerie. The WISH investigators thank Professor Graham Giles of the Cancer Epidemiology Centre of the Cancer Council Victoria, for permission to use the Dietary Questionnaire for Epidemiological Studies (Versions 3 and 3.1): The Cancer Council Victoria, 2007. We acknowledge the Premiers Science Research Fund as the public source of funds for this study through the South Australian Population Health Intergenerational Research (SAPHIRe) collaborative. We acknowledge our collaboration partners, the North West Adelaide Health Study, Florey Adelaide Male Ageing Study and the Australian Longitudinal Study of Ageing. MTH is supported by a Post-doctoral Training Fellowship (Public Health) from the National Health and Medical Research Council (NHMRC) of Australia (no. 511345). RAM is supported by a NHMRC Practitioner Fellowship.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Competing interests
The authors declare no conflict of interest.
Additional information
Supplementary Information accompanies the paper on the Nutrition and Diabetes website
Supplementary information
Rights and permissions
This work is licensed under the Creative Commons Attribution-NonCommercial-No Derivative Works 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-nd/3.0/
About this article
Cite this article
Haren, M., Misan, G., Grant, J. et al. Proximal correlates of metabolic phenotypes during ‘at-risk’ and ‘case’ stages of the metabolic disease continuum. Nutr & Diabetes 2, e24 (2012). https://doi.org/10.1038/nutd.2011.20
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/nutd.2011.20
Keywords
This article is cited by
-
Abdominal adiposity and obstructive airway disease: testing insulin resistance and sleep disordered breathing mechanisms
BMC Pulmonary Medicine (2012)