The objective of this analysis was to test the validity of the estimates of energy expenditure and sedentary lifestyle obtained through a self-administered questionnaire of physical activity for Spanish-speaking people adapted from US questionnaires (Nurses' Health Study and Health Professionals' Follow-up Study) using a triaxial accelerometer (RT3 Triaxial Research Tracker) as the reference.

Design and setting: Validation study, calculating the non-parametric correlation coefficients between the level of physical activity and sedentary lifestyle collected by the self-administered questionnaire and the triaxial accelerometer measurements. Percentage of misclassification and kappa coefficients were also calculated.

The study population consisted of a sample of 40 obese women who were participants of the SUN (Seguimiento Universidad de Navarra) project (a prospective cohort study among Spanish university alumni). They were selected because of their peculiar metabolic characteristics, in the search for a sub-optimal scenario for validity.

Physical activity during leisure time (estimated as MET-h week-1) derived from the self-administered questionnaire moderately correlated with kcal day-1 assessed through the accelerometer (Spearman's ρ = 0.507, 95% confidence interval (CI) 0.232, 0.707). The Spearman correlation between the ratio of sedentary lifestyle to physical activity obtained through the questionnaire and the direct estimation (RT3) was -0.578 (95% CI -0.754, -0.325). The kappa index was 0.25 (P = 0.002) when assessing the cross-classification into quintiles and 0.41 for the dichotomous estimation of a sedentary lifestyle. Only 2.5% of participants were misclassified by the questionnaire more than two quintiles apart from the estimates of the RT3.

The moderate values obtained for correlation in a sub-optimal scenario for validity and the low percentage of extreme misclassification suggest the validity of the questionnaire to assess physical activity in Spanish-speaking women aged 20–50 years.

To produce prediction equations for basal metabolic rate (BMR) derived from weight and height covering the age range from birth to old age.

Cross-sectional data on BMR, sex, age, weight, height, ethnicity and measurement technique from the Oxford Brookes BMR database.

Worldwide.

Data for 13 910 men, women and children from 174 papers published between 1914 and 2001.

Absolute and proportional regression models were developed for each sex, showing a steep rise in predicted BMR with age until 15 years, more pronounced in males than females, then a gradual fall through adulthood. Predicted BMR increased by 6% and 1.4%, respectively, per standard deviation increase in weight and height. Predicted BMR in Caucasians was 4% higher than in non-Caucasians, though the effect size was sensitive to the inclusion or exclusion of data from certain influential publications. The effect of measurement technique on BMR, closed-circuit versus open-circuit, was small, near 1%.

It is possible to develop prediction equations that avoid splitting the data into arbitrary age groups. Heterogeneity between publications is greater than might be expected by chance, probably due to undocumented differences in technique.

To cross-validate three predictive set of equations for basal metabolic rate (BMR) developed by Schofield (Schofield database), Henry (Oxford database) and Cole (Oxford database) using mean values for age, weight, height and BMR of published studies.

Literature review of studies published from 1985 to March 2002.

All studies selected used appropriate methods and followed conditions that met the criteria established for basal metabolism, were performed in healthy adults, and were not part of the Schofield or Oxford database.

A total of 261 groups of men and women from 175 studies were selected and categorised in three age groups (18.5–29.9, 30.0–59.9, ≥60 years old) and three body mass index (BMI) groups (normal weight, overweight and obese).

Linear regression and concordance correlation analysis showed that the three sets of equations had the same association and agreement with measured BMR, across gender, age, and BMI groups. The agreement of all equations was moderate for men and poor for women. The lowest mean squared prediction errors (MSPRs) were given by Henry equations in men and Cole equations in women. Henry and Cole equations gave lower values than Schofield equations, except for men over 60 years of age. Henry equations were the most accurate in men. None of the three equations performed consistently better in women.

These results support the use of Henry equations in men with a wide range of age and BMI. None of the proposed predictive equations seem to be appropriate to estimate BMR in women.