Computational protocol: Predictors and risks of body fat profiles in young New Zealand European, Māori and Pacific women: study protocol for the women’s EXPLORE study

Similar protocols

Protocol publication

[…] Study participants are adult NZ women from three ethnic groups (NZ European, Māori, and Pacific Island). A total sample size of 225 women per ethnic group, consisting of 75 per profile group, will provide 80% power at significance levels of p < 0.05 to detect a medium effect size f of 0.25 (G*Power 3.1.2) for comparing the “hidden fat” profile with the other two body composition profiles (“normal fat” and “apparent fat”) regarding metabolic disease risk markers, dietary and physical activity patterns, and miRNA expression levels.The medium effect size is relevant to all variables, and encompasses a variety of scenarios, as we wish to be able to explore how metabolic profile changes with body composition. For example, if the three groups have equally spaced means (μ - Δ, μ, μ + Δ), the difference in means will be detected with 80% power when Δ = 0.31 x σ, where σ is the within group standard deviation. For cholesterol, where preliminary data suggests σ = 0.98 mmol/L, we have 80% power to detect the difference when Δ = 0.30 mmol/L. Alternately, if two groups have the same mean μ and the third has mean μ + Δ, 80% power is achieved when Δ > 0.53 x σ, or 0.52 mmol/L in the case of cholesterol. A Δ = 0.30 – 0.52 mmol/L is estimated to be associated with a 9 – 15% lower relative risk of coronary heart disease (CHD)-related mortality (Gould et al. ). For HDL-C where preliminary data suggests σ = 0.38 mmol/L we have 80% power to detect a difference when Δ = 0.12 – 0.20 mmol/L. Every 0.1 mmol/L increase in HDL-C has been suggested to reduce CHD risk by between 8 – 15% (Gordon et al. ; Turner et al. ). For TG where preliminary data suggests σ = 0.45 mmol/L we have 80% power to detect a difference when Δ = 0.14 – 0.24 mmol/L. Studies in women showed a 1 mmol/L increase in TG was associated with 37% increase in risk of CVD (after adjustment for HDL-C and other risk factors)(Austin et al. ); Δ of 0.14 – 0.24 is thus estimated to be associated with 5.2 – 9% difference in CVD risk.The power for simple (one variable) logistic regression for the risk of having a “hidden fat” profile among people of normal BMI is equivalent to the power of the independent sample t-test (Vaeth and Skovlund ) for comparing the predictor variable mean between the hidden fat and normal fat profiles. With a sample size of 75 per group and equal variance within groups, 80% power is achieved for differences of 0.46 σ. For instance, preliminary data on total energy expenditure assessed using the Recent Physical Activity Questionnaire (RPAQ) estimates the standard deviation at 5.36 METs-h/day; if the true difference means between the body composition groups is 2.47 METs-h/day, the logistic regression coefficient will be significantly non-zero 80% of the time. Clinically, the difference of 2.47 METs-h/day equates to ~684 kJ/day (164 kcal/day)(Besson et al. ), which is considered within the target range of recommended energy expended each day in physical activity and/or exercise (Pescatello and American College of Sports Medicine ). We will consider logistic regression predictors from the dietary and activity pattern data, and separately for the miRNA measurements.Based on our pilot study that showed a prevalence of 21% of NZ European women having a “hidden fat” profile, (Kruger et al. ) a sample of ~1140 women will need to be screened (380 per ethnicity) to find ~75 women per profile group; or to explore new profiles. The study design and study procedures are illustrated in Figure .Figure 1Inclusion criteria for women are:age (16 to 45 years),being post-menarcheal or pre-menopausal (as defined by a continuous regular menstrual cycle for the past one complete year),ethnicity (being of NZ European, Māori, or Pacific ethnicity as defined by self-identification and having at least one parent from the same ethnicity).age (16 to 45 years),being post-menarcheal or pre-menopausal (as defined by a continuous regular menstrual cycle for the past one complete year),ethnicity (being of NZ European, Māori, or Pacific ethnicity as defined by self-identification and having at least one parent from the same ethnicity).Exclusion criteria for women are:pregnancy and lactation,presence of any diagnosed chronic illness particularly affecting metabolic health (e.g. T2DM),presence of dairy allergy as the taste solution is dairy based.pregnancy and lactation,presence of any diagnosed chronic illness particularly affecting metabolic health (e.g. T2DM),presence of dairy allergy as the taste solution is dairy based. [...] Statistical analysis will be performed using IBM SPSS statistics (IBM Corporation, New York, USA). Descriptive statistics will be used to describe the baseline population using mean (standard deviation), median (25, 75 percentile) or frequencies summary statistics. Normality of distribution will be evaluated using the Kolmogorov-Smirnov test and examining normality plots. Non-normally distributed variables will be transformed into approximately normal distributions by logarithmic transformations and again tested for normality. Primary statistical analyses will involve ANOVA tests with post-hoc analysis and Bonferoni adjustments comparing body composition profile groups regarding metabolic disease risk markers and dietary and physical activity patterns and miRNA levels; multiple logistic regression analysis to determine odds ratios of having a “hidden body fat” profile based on dietary and physical activity patterns; and, separately, miRNA expression levels. Principal components analysis will be performed on the miRNA data using the PCP directive in GenStat to identify linear combinations of the miRNAs that account for most of the variation between individuals (Zacharewicz et al. ); and Pearson correlations to determine correlation coefficients between dietary and physical activity patterns and miRNA expression profiles. A p-value of <0.05 will be considered significant. […]

Pipeline specifications

Software tools G*Power, SPSS
Applications Miscellaneous, Metabolic profiles analysis
Organisms Homo sapiens
Chemicals Glucose